Citation
Heterogeneous distribution of microvascular blood flow contributes to impaired skeletal muscle oxygenation in diabetes

Material Information

Title:
Heterogeneous distribution of microvascular blood flow contributes to impaired skeletal muscle oxygenation in diabetes
Creator:
McClatchey, Penn Mason ( author )
Language:
English
Physical Description:
1 electronic file (170 pages) : ;

Subjects

Subjects / Keywords:
Non-insulin-dependent diabetes ( lcsh )
Microcirculation ( lcsh )
Perfusion (Physiology) ( lcsh )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Review:
People with type 2 diabetes mellitus (T2DM) suffer excess morbidity and mortality. The strongest clinical predictor of morbidity and mortality in the general population is reduced aerobic exercise capacity. T2DM causes impaired exercise capacity, and traditional explanations for this impairment involve reduced blood flow and/or reduced mitochondrial capacity. However, recent studies indicate that exercise capacity may be impaired independently from either tissue demand or bulk blood flow. The most likely explanation for this disconnect is heterogeneous distribution of microvascular blood flow. Under conditions of heterogeneous blood flow, some capillaries are over-perfused and thus saturate their capacity for oxygen delivery, while others are under-perfused and thus cannot support local tissue demand. Local measures reveal excess skeletal muscle deoxygenation during exercise in T2DM despite normal limb blood flow, and skeletal muscle deoxygenation is more heterogeneous in T2DM than in overweight controls. Sensitivity analyses building from established principles in mass transport reveal that heterogeneous blood flow alone is sufficient to cause both impaired skeletal muscle oxygenation and insulin resistance in T2DM. A more detailed version of this model was applied to oxygen transport in the obese Zucker rat (OZR, a common animal model of T2DM), and accurately predicted the degree of perfusion heterogeneity observed in the OZR. A novel software technique for quantifying capillary blood flow and its distribution reveals that high fat feeding (an experimental model of insulin resistance) causes heterogeneous capillary blood flow in mice. Finally, a combined analysis drawing from both first principles in microfluidics and empirical measurements of blood viscosity reveals that diabetes-induced degradation of the endothelial glycocalyx (a gel-like layer of macromolecules lining the interior surface of blood vessels) can account for heterogeneous distribution of microvascular blood flow. Collectively, these findings help to explain impaired oxygen extraction despite reduced blood flow in T2DM, and also offer a potential explanation as to why exercise capacity would predict morbidity and mortality: the proposed mechanism could plausibly apply to all tissues, not just to skeletal muscle.
Thesis:
Thesis (Ph.D..)--University of Colorado Denver.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: Adobe Reader.
Statement of Responsibility:
by Penn Mason McClatchey, Jr.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
on10000 ( NOTIS )
1000053517 ( OCLC )
on1000053517

Downloads

This item has the following downloads:


Full Text
HETEROGENEOUS DISTRIBUTION OF MICRO VASCULAR BLOOD FLOW
CONTRIBUTES TO IMPAIRED SKELETAL MUSCLE OXYGENATION IN DIABETES
by
PENN MASON MCCLATCHEY, JR.
M.S., University of Colorado Denver, 2015 B.S., Georgia Institute of Technology, 2013
A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirements for the degree of Doctor of Philosophy Bioengineering Program
2017


ii
This thesis for the Doctor of Philosophy degree by Penn Mason McClatchey, Jr. has been approved for the Bioengineering Program by
Kendall S. Hunter, Chair Jane E. B. Reusch, Advisor Richard K. P. Benninger Timothy A. Bauer Jefferson C. Frisbee Mary C. M. Weiser-Evans Eric P. Schmidt
Date: May 13, 2017


Ill
McClatchey Jr., Penn Mason (Ph.D. Bioengineering)
Heterogeneous Distribution of Microvascular Blood Flow Contributes to Impaired Skeletal
Muscle Oxygenation in Diabetes
Thesis directed by Professor Jane E.B. Reusch
ABSTRACT
People with type 2 diabetes mellitus (T2DM) suffer excess morbidity and mortality. The strongest clinical predictor of morbidity and mortality in the general population is reduced aerobic exercise capacity. T2DM causes impaired exercise capacity, and traditional explanations for this impairment involve reduced blood flow and/or reduced mitochondrial capacity. However, recent studies indicate that exercise capacity may be impaired independently from either tissue demand or bulk blood flow. The most likely explanation for this disconnect is heterogeneous distribution of microvascular blood flow. Under conditions of heterogeneous blood flow, some capillaries are over-perfused and thus saturate their capacity for oxygen delivery, while others are under-perfused and thus cannot support local tissue demand. Local measures reveal excess skeletal muscle deoxygenation during exercise in T2DM despite normal limb blood flow, and skeletal muscle deoxygenation is more heterogeneous in T2DM than in overweight controls. Sensitivity analyses building from established principles in mass transport reveal that heterogeneous blood flow alone is sufficient to cause both impaired skeletal muscle oxygenation and insulin resistance in T2DM. A more detailed version of this model was applied to oxygen transport in the obese Zucker rat (OZR, a common animal model of T2DM), and accurately predicted the degree of perfusion heterogeneity observed in the OZR. A novel software technique for quantifying capillary blood flow and its distribution reveals that high fat feeding (an experimental model


IV
of insulin resistance) causes heterogeneous capillary blood flow in mice. Finally, a combined analysis drawing from both first principles in microfluidics and empirical measurements of blood viscosity reveals that diabetes-induced degradation of the endothelial glycocalyx (a gel-like layer of macromolecules lining the interior surface of blood vessels) can account for heterogeneous distribution of microvascular blood flow. Collectively, these findings help to explain impaired oxygen extraction despite reduced blood flow in T2DM, and also offer a potential explanation as to why exercise capacity would predict morbidity and mortality: the proposed mechanism could plausibly apply to all tissues, not just to skeletal muscle.
The form and content of this thesis are approved. I recommend its publication.
Approved: Jane E. B. Reusch


V
ACKNOWLEDGEMENTS
I would like to acknowledge each of my thesis committee members for their contributions to this project and for the mentorship they provided me during my pursuit of this degree: Jane Reusch, Richard Benninger, Kendall Hunter, Timothy Bauer, Jefferson Frisbee, Mary Weiser-Evans, and Eric Schmidt were each invaluable mentors in their own ways. I would also like to thank my collaborators and co-authors outside of my thesis committee who helped to produce the research included in this dissertation, including Judith Regensteiner, Irene Schauer, Amy Huebschmann, Fan Wu, Mark Olfert, Christopher Ellis, Daniel Goldman, Sara Hull, Ian Williams, David Wasserman, and Michal Schafer. In addition, I would like to thank the many people not mentioned above who provided feedback and guidance in the course of this project, including Pete Watson, Robert Roach, Andrew Subudhi, Rebecca Scalzo, Leslie Knaub, and Amy Keller. Finally, I would like to thank the friends and family to helped in me in innumerable ways- particularly my mother, Anne McClatchey (who homeschooled me and taught me to recognize the inherent mathematical relationships in natural systems) and my wife, Clio McClatchey (who proofread many of my papers and helped me translate my ideas into lay-English). This dissertation bears my name, but it required intensive efforts from dozens of others to come to fruition.


VI
TABLE OF CONTENTS
CHAPTER
I. EXERCISE, BLOOD FLOW, AND THE SKELETAL MUSCLE MICROCIRCULATION IN
DIABETES MELLITUS.................................................................1
1. Preface.................................................................1
2. Abstract................................................................1
3. Introduction............................................................2
4. Cardiac Output..........................................................3
5. Skeletal Muscle Blood Flow..............................................4
6. Microvascular Perfusion Heterogeneity...................................5
7. Considering Causality...................................................8
II. DISSOCIATION OF LOCAL AND GLOBAL SKELETAL MUSCLE OXYGEN TRANSPORT
METRICS IN TYPE 2 DIABETES.......................................................10
1. Preface................................................................10
2. Abstract...............................................................10
3. Introduction...........................................................11
4. Methods................................................................14
5. Results................................................................18
6. Discussion.............................................................23
III. A CONCEPTUAL FRAMEWORK FOR PREDICTING AND ADDRESSING THE
CONSEQUENCES OF DISEASE-RELATED MICROVASUCLAR DYSFUNCTION........................28
1. Preface................................................................28
2. Abstract...............................................................28
3. Introduction...........................................................29
4. Methods................................................................32
5. Results................................................................35


vii
6. Discussion...............................................................48
IV. IMPAIRED TISSUE OXYGENATION IN METABOLIC SYNDROME REQUIRES
INCREASED MICRO VASCULAR PERFUSION HETEROGENEITY...................................59
1. Preface..................................................................59
2. Abstract.................................................................59
3. Introduction.............................................................60
4. Materials and methods....................................................64
5. Results..................................................................74
6. Discussion...............................................................79
V. FULLY AUTOMATED SOFTWARE FOR COMPREHENSIVE QUANTITATION OF
CAPILLARY STRUCTURE AND PERFUSION..................................................84
1. Preface..................................................................84
2. Abstract.................................................................84
3. Introduction.............................................................86
4. Methods..................................................................89
5. Results..................................................................95
6. Discussion...............................................................98
VI. THE ENDOTHELIAL GLYCOCALYX PROMOTES HOMOGENOUS BLOOD FLOW
DISTRIBUTION WITHIN THE MICRO VASCULATURE.........................................104
1. Preface.................................................................104
2. Abstract................................................................104
3. Introduction............................................................105
4. Methods.................................................................109
5. Results.................................................................114
6. Discussion..............................................................119
VII. CONCLUSIONS AND FUTURE DIRECTIONS
126


Vlll
1. Abstract.....................................................................126
2. A perfusion-centric paradigm for inflammation................................127
3. Clinical and laboratory translation of microvascular perfusion findings......131
4. Potential microvascular perfusion therapies..................................137
5. Conclusions..................................................................139
REFERENCES............................................................................140


IX
LIST OF TABLES
TABLE
1. Summary of model parameters used in the present study-----------------73
2. Summary of microvascular perfusion measures in the primary literature-87


X
LIST OF FIGURES
FIGURE
1. Illustration of NIRS measures used for analysis----------------------------16
2. Impaired muscle oxygenation despite normal blood flow and hemoglobin reserve in
T2DM-----------------------------------------------------------------------19
3. Hemoglobin reserve and muscle deoxygenation do not correlate to V02peak in T2DM -----------------------------------------------------------------------21
4. Mutual correlations between blood flow, hemoglobin reserve, and V02peak are
abolished in T2DM---------------------------------------------------------------22
5. Solutes vary in their degree fractional equilibration with the interstitium during
capillary transit---------------------------------------------------------------34
6. Simulated distribution of blood flow in an idealized microvascular network-------35
7. Effects of blood flow and diffusion capacity on microvascular solute flux--------36
8. Relationship between baseline fractional equilibration and sensitivity to blood flow- 38
9. Relationship between baseline fractional equilibration and sensitivity to diffusion
capacity--------------------------------------------------------------------------40
10. Relationship between baseline fractional equilibration and sensitivity to perfused
capillary density-----------------------------------------------------------------42
11. Relationship between baseline fractional equilibration and sensitivity to arteriolar
perfusion heterogeneity-----------------------------------------------------------45
12. Complex phenotypes influence solute flux through a variety of mechanisms--------47
13. Flowchart of recommended steps to test for microvascular contributions to solute flux
defects---------------------------------------------------------------------------53
14. Previous studies establish impaired oxygen diffusion and microvascular perfusion
heterogeneity in the OZR----------------------------------------------------------63
15. Visualizations of simulations performed------------------------------------------66
16. Heatmaps visualizing look-up tables (LUTs) of venous oxygenation as a function of capillary hematocrit and relative blood flow-------------------------------------------70


XI
17.
18.
19.
20. 21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
Heatmaps visualizing look-up tables (LUTs) of skeletal muscle oxygenation as a function of capillary hematocrit and relative blood flow--------------------------71
Differences in venous oxygenation, muscle oxygenation, and perfusion heterogeneity can be predicted using a simulation of microvascular blood flow and oxygen transport ----------------------------------------------------------------------------------75
Sensitivity analysis reveals that the effects of varying vessel diameters are negligible relative to the effects of varying flow distributions------------------------77
Analysis of perfusion heterogeneity effects at a single bifurcation reveals the mechanisms by which perfusion heterogeneity interferes with oxygen transport-79
Outline of image processing steps used to identify capillary segments for software flow measurement------------------------------------------------------------------91
Outline of cross-correlation routine used to measure RBC flow velocity-------92
By-hand measurement techniques ----------------------------------------------95
Software and by-hand measurements detect the same capillary perfusion and RBC flux differences between chow-fed and HFD mice------------------------------------96
Additional measurements taken using software only----------------------------97
Visualization of HFD on capillary structure and perfusion--------------------98
Schema of established determinants of microvascular blood flow distribution based on the literature--------------------------------------------------------------108
Parameters used to create idealized arteriolar trees------------------------112
Influences of microvascular blood viscosity on microvascular perfusion heterogeneity in a simulated arteriolar tree----------------------------------------------115
Influences of microvascular blood viscosity on blood flow distribution at an idealized
capillary bifurcation-------------------------------------------------------------117
Effects of glycocalyx properties on the determinants of blood viscosity-----------119
Summary of findings---------------------------------------------------------------121
A perfusion-centric paradigm for inflammation-------------------------------------128
Proof of concept for clinical application of flow tracking and oxygen delivery algorithms------------------------------------------------------------------------136


1
CHAPTER I
EXERCISE, BLOOD FLOW, AND THE SKELETAL MUSCLE MICROCIRCULATION IN DIABETES MELLITUS
Co-authors: Timothy A. Bauer, Judith, G. Regensteiner, and Jane E.B. Reusch
1. Preface
This chapter serves as a topic introduction and overview of the literature concerning exercise capacity and its relationship to blood flow and the microcirculation in diabetes. Here I make the arguments for and against bulk blood flow as a limiting factor to exercise, both at the level of cardiac output and at the level of blood flow to the exercising muscle. Based on the current literature (as of early 2017), it is likely that both blood flow and its distribution play a role in reduced exercise capacity in diabetes. This chapter has been accepted for publication as a chapter in the upcoming book Diabetes and Exercise (Humana Press) and is currently in press.
2. Abstract
Aerobic exercise capacity is impaired in both Type 1 Diabetes (T1DM) and Type 2 Diabetes (T2DM), and this impairment is predictive of future morbidity and mortality. Although the precise etiology of impaired exercise capacity in diabetes remains unclear, several distinct lines of evidence indicate that reduced delivery of oxygen by the cardiovascular system plays a causal role. Cardiac output is often but not always reduced in diabetes. This change is sufficient but not necessary for reduced exercise capacity. Skeletal muscle blood flow is also often but not always reduced in diabetes. This change is also sufficient but not necessary for reduced exercise capacity. In addition, a growing number of animal and simulation studies show that heterogeneous distribution of blood flow within the


2
microcirculation contributes to oxygen delivery limitations in diabetes. Once again, this change is sufficient but not necessary for reduced exercise capacity. In this chapter, we discuss each of these changes in cardiovascular function and their likely causes, beginning with the heart and gradually progressing to capillary level. We then conclude our overview by interpreting the causality or lack thereof of each diabetes-related pathological change as it relates to reduced oxygen delivery to skeletal muscle.
3. Introduction
Any attempt to understand limitations in exercise function with diabetes would be incomplete without considering the influences of the cardiovascular system and blood flow regulation. Exercise capacity (VCtemax) is impaired both in Type 1 Diabetes (T1DM) and Type 2 Diabetes (T2DM) (1-6), and this impairment is predictive of mortality (7-11) and cardiovascular complications (12-15). These relationships suggest that exercise capacity is a sensitive measure of early changes in cardiovascular function with diabetes. This notion is further supported by the associations of cardiac output and skeletal muscle blood flow (SMBF) with VCEmax in healthy individuals (16-18). Limitations in both cardiac output and SMBF have been reported in diabetes (19-22), indicating that blood flow may be a component of exercise limitations in diabetes. In addition to reduction of total blood flow, increased heterogeneity of microvascular blood flow distribution (23-26), loss of capillary perfusion (27; 28), and reduced whole-body oxygen extraction (29) have also been reported, indicating that heterogeneous distribution of blood flow may also play a role in limiting aerobic capacity in diabetes. In this chapter, we will explore known changes to both blood flow and its distribution in diabetes, beginning with changes in cardiac function and progressing to the capillary level.


3
4. Cardiac Output
Both T1DM and T2DM are associated with left ventricular diastolic dysfunction (30; 31). Diastolic dysfunction in diabetes is associated with fibrotic remodeling of the myocardium (32), and eventually leads to heart failure. As fibrotic remodeling progresses, the contractile ability of the heart is not necessarily impaired, sometimes but not always allowing for maintenance of systolic function (33). The result of these changes is that stroke volume (SV) is reduced in diabetes (19), and cardiac output is impaired under conditions in which the heart cannot compensate with increased heart rate (HR), such as in maximal exercise (19; 21). It is not entirely clear from the literature whether the sub-clinical cardiac dysfunction observed in diabetes contributes to limitations in exercise capacity. Baldi et al found using an inert gas rebreathing technique that whole-body arteriovenous oxygen difference during exercise is reduced in diabetes and that reductions in oxygen extraction associated to reductions in VChmax (29), whereas Gusso et al found using a similar CO2 rebreathing technique that arteriovenous oxygen difference was not altered by diabetes, but that reductions in CO and V02max were associated (19). This juxtaposition of results suggests that differences in blood flow distribution (e.g. greater fractional perfusion of non-muscle tissues in diabetes) or mitochondrial demand are co-determinants of exercise capacity in diabetes along with CO. Given the substantial similarities between cardiac and skeletal muscle, it is probable that impaired cardiac function is itself a manifestation of muscle functional limitations in diabetes rather than their root cause.
Reduced cardiac output in diabetes is further compounded by the association between diabetes and hypertension (34; 35). In particular, large arteries become less responsive to vasomotor stimuli in diabetes, and this effect is particularly pronounced in assays of nitric


4
oxide (NO) mediated endothelium dependent vasodilation resulting from pharmacological agonists (36; 37) and fluid shear stress (38; 39). Reduction in vasomotor (and especially vasodilatory) function tends to occur in conjunction with fibrotic remodeling of large arteries (40), and vessel stiffness itself may contribute to reduced dynamism in vessel tone. In people without diabetes, hypertension causes impairments in both cardiac output and VO2 max (41). The rate of hypertension is significantly elevated in both T1DM and T2DM relative to the general population (36; 37), which suggests that hypertension and the associated reductions in cardiac output may contribute to population level differences in exercise capacity with diabetes. It is worth noting, however, that limitations in VChmax are observed in diabetes even in the absence of hypertension or any other overt cardiovascular disease state (2; 31). Although hypertension likely contributes to exercise limitations at the population level, hypertension alone cannot fully account for the exercise limitations observed in diabetes.
5. Skeletal Muscle Blood Flow
Independently of cardiac and large vessel function, oxygen delivery to skeletal muscle could be impaired by inappropriate distribution of blood flow among organs. As previously discussed, the vasomotor dynamism of large blood vessels is reduced in diabetes (37-40). It is likely that this effect would interfere with redistribution of blood flow from inactive organs (e.g. mesenteric blood flow) to active skeletal muscle, but this hypothesis has not yet been directly tested. In addition to possible differences in the dynamism of blood flow distribution, lean body mass as a fraction of total body mass is reduced in T2DM (42; 43). This effect would be expected to impair whole-body oxygen extraction by increasing blood flow to non-oxidative tissues as a fraction of total blood flow (44). However, body composition is not necessarily altered in T1DM (45; 46), and impaired VCtemax is found not


5
only when comparing of T2DM and lean, healthy individuals (1), but also in T1DM (3; 4) and even when comparing of T2DM and obese, sedentary individuals without diabetes (47).
It is therefore likely that the contributions of macrovascular blood flow distribution to reductions in aerobic capacity can occur independently of changes in body composition.
Regardless of whether these differences stem from cardiac dysfunction, vascular dysfunction, or something else entirely, SMBF measured at the whole-limb level is often but not always reduced in diabetes (20; 22; 48-51). Even if SMBF were to reach normal steady-state levels, the hyperemic response to exercise is often slowed in diabetes (48; 49; 52; 53). It is likely that slowed blood flow kinetics contribute to the increased discomfort the onset of exercise reported in diabetes (53-55). Interestingly, there are some studies in which steady-state SMBF is not reduced in diabetes and yet aerobic exercise capacity is still impaired (56-58). This juxtaposition of findings implies that organ-level (as opposed to whole-body) oxygen extraction is impaired in diabetes in addition to reductions in SMBF. True to form, human MRI studies by Zheng et al (59) and animal catheterization studies by Frisbee et al (60) show an impaired ability to increase skeletal muscle oxygen extraction fraction (SMOEF) following muscle contraction. This effect of diabetes does not appear to be unique to skeletal muscle, given that tissue-level oxygen extraction is also reduced in diabetic retinopathy and neuropathy (61; 62). As is also true of impaired cardiac function and hypertension, it appears likely that reduced SMBF contributes to but is not necessary for diabetic exercise dysfunction.
6. Microvascular Perfusion Heterogeneity
Impairment of oxygen delivery independently in T2DM of SMBF is likely caused by increased heterogeneity of microvascular perfusion. Frisbee et al have shown in the Obese Zucker Rat (OZR) model of T2DM that microvascular perfusion heterogeneity is increased


6
(25), that this perfusion heterogeneity contributes to peripheral vascular disease (23; 26), reversal of this perfusion heterogeneity with a cocktail of anti-adrenergic and endotheliumtargeting drugs acutely normalizes skeletal muscle function (24; 25; 60), and that these effects can be predicted from first principles in mass transport and anatomy (63).
Importantly, microvascular perfusion heterogeneity in the OZR model of Frisbee et al caused impaired muscle oxygenation in part through heterogeneous red blood cell (RBC) distribution at the capillary level (24; 63). This result is further recapitulated by the intravital microscopy results of Poole et al in the Goto-Kakizaki (GK) rat model of T2DM (28) and in the streptozotocin-treated model of T1DM (27). Not only are microvascular perfusion heterogeneity and a resulting impairment in oxygen availability observed in all these animal models of diabetes, it has also been shown that microvascular perfusion heterogeneity leads to impaired oxygen extraction independently of total blood flow on both theoretical (63-66) and empirical bases (26; 67; 68). The mechanism for this impairment (some capillaries are underperfused while others are overperfused and effectively saturate their capacity for oxygen delivery), is not tissue specific, consistent with observations of oxygen extraction limitations not only in skeletal muscle (59; 60), but also in other peripheral tissues (61; 62).
In addition to increases in the heterogeneity of microvascular perfusion, reduced capillary density is also observed in diabetes (69; 70), further reducing oxygen availability independently of SMBF. Combined, the effects of microvascular perfusion heterogeneity and reduced capillary density can account for discrepancies between SMBF and aerobic capacity (63).
Microvascular dysfunction and perfusion heterogeneity in diabetes may be caused by degradation of the endothelial glycocalyx. The endothelial glycocalyx is a semi-permeable,


7
space-filling layer of glycoproteins and glycosaminoglycans lining the luminal surface of the endothelium. Glycocalyx degradation has been reported in both T1DM and T2DM (71-73), and glycocalyx degradation is associated both in diabetes and in health with increased risk and early signs of future cardiovascular morbidities (74-78), as is exercise capacity (7-11). Degradation of the endothelial glycocalyx causes a similar redistribution of RBCs within the capillary network to that observed in diabetes whether glycocalyx degradation is achieved by enzymatic means (79; 80), or as a result of oxidative stress stemming from acute hyperglycemia or infusion of oxidized LDL (81; 82). Furthermore, physiologic glycocalyx degradation during sepsis or adenosine infusion has been shown to cause reductions in tissue oxygen extraction (83; 84). Although the connection between glycocalyx degradation and reduced aerobic capacity in diabetes has not yet been directly tested, it is noteworthy that glycocalyx degradation is involved in glomerular hyperfiltration (an early sign of diabetic nephropathy) (85; 86), glycocalyx degradation causes acute insulin resistance (87), and insulin resistance, glomerular hyperfiltration, and reduced aerobic capacity are all mutually correlated in diabetes (15; 88). Mass transport analysis reveals that insulin resistance and impaired exercise capacity can both be predicted from perfusion heterogeneity (66), and so it is likely that the perfusion effects of glycocalyx degradation contribute to these phenotypes. Simulation studies indicate that glycocalyx charge density (as a determinant of permeability) modulates the heterogeneity/homogeneity of microvascular perfusion (89), providing a plausible mechanism for increased microvascular perfusion heterogeneity in diabetes. Ongoing studies within our group seek to clarify the relationship between the endothelial glycocalyx and diabetic exercise dysfunction.


8
7. Considering Causality
Because blood flow and its distribution are altered at every level of the circulation in diabetes from the heart to the smallest capillaries, it is useful to consider precisely which effects might play a causal role in reducing aerobic capacity. Studies in which CO, SMBF, or both are normal in diabetes and yet aerobic capacity is reduced show that changes in macrovascular parameters are not necessary for diabetic exercise dysfunction (2; 31; 56-58). Although similar macrovascular changes are sufficient to reduce aerobic capacity in the general population (16-18), the fact that they are not necessary for reduced aerobic capacity in diabetes indicates that the root cause of diabetic exercise dysfunction may lie elsewhere. Antioxidant therapy normalizes many macrovascular parameters in diabetes (90; 91) and yet has not been shown to normalize exercise capacity (and would be expected to interfere with exercise training (92; 93)), whereas the microvascular dysfunction reported by Frisbee et al is acutely reversible and its reversal improves skeletal muscle function (25; 60). This combination of results indicates that microvascular dysfunction might play a causal role in diabetic exercise impairments. Given that the heart and the vessel walls of microvessels are themselves heavily vascularized and therefore subject to microcirculatory influence, it is plausible that cardiac and macrovascular dysfunction in diabetes are themselves caused by microvascular dysfunction.
In this overview of changes in blood flow and its distribution in diabetes, we sought to assess the possibility that exercise dysfunction in diabetes might be an early detector of impaired cardiovascular dysfunction. Blood flow is often (but not always) reduced at the whole-body level (19; 21), at the whole-limb level (20; 22; 48-51) and at the capillary level (27; 28) in diabetes. However, reduced aerobic capacity is sometimes observed even when


9
blood flow is maintained (2; 31; 56-58). This apparent discrepancy may be explained by microvascular alterations including increased perfusion heterogeneity (24; 25; 60) and reduced capillary density (69; 70). Microvascular perfusion heterogeneity is itself a plausible contributor cardiac and macrovascular dysfunction due to its effects on tissue oxygenation (64; 65; 67; 68), and can be recapitulated by glycocalyx degradation (79-82), which also occurs in T1DM (72), in T2DM (71), and more generally in states of acute nutrient stress (75; 81; 82). Further studies will be required to clarify the relationships between glycocalyx degradation, blood flow, and microvascular perfusion. It is even possible that the etiology of impaired exercise capacity varies from individual to individual, but it is clear that oxygen delivery limitations resulting from impairments in blood flow or its distribution play a central role.


10
CHAPTER II
DISSOCIATION OF LOCAL AND GLOBAL SKELETAL MUSCLE OXYGEN TRANSPORT METRICS IN TYPE 2 DIABETES
Co-authors: Timothy A. Bauer, Judith G. Regensteiner, Irene E. Schauer, Amy G.
Huebschmann, and Jane E. B. Reusch
1. Preface
This chapter provides an illustration of the disconnect between blood flow and tissue oxygenation that is thematic in this dissertation. If blood flow were the limiting factor to skeletal muscle oxygenation in diabetes, one would expect that: 1) blood flow is reduced, 2) blood flow is correlated to skeletal muscle oxygenation, and 3) blood flow is correlated to exercise capacity. In fact, none of these three statements is true in the diabetic patient population used for this study. Instead, there appears to be a general trend towards diabetes-related dissociation of local and global oxygen transport parameters that are correlated in overweight controls. Furthermore, measures of local skeletal muscle oxygenation and deoxygenation are more heterogeneous is diabetes, suggesting that heterogeneity of something may be related to the dissociation of blood flow and oxygen uptake in diabetic skeletal muscle. This chapter has been submitted to Diabetes Care for review.
2. Abstract
Aims: Exercise capacity is impaired in type 2 diabetes, and this impairment predicts excess morbidity and mortality. This defect appears to involve excess skeletal muscle deoxygenation, but the underlying mechanisms remain unclear. We hypothesized that reduced blood flow, reduced local recruitment of blood volume/hematocrit, or both contribute to excess skeletal muscle deoxygenation in type 2 diabetes. Methods: In patients with (n=23) and without (n=18) type 2 diabetes, we recorded maximal reactive hyperemic


11
leg blood flow, peak oxygen utilization during cycling ergometer exercise (V02peak), and near-infrared spectroscopy-derived measures of exercise-induced changes in skeletal muscle oxygenation and blood volume/hematocrit. Results: We observed a significant increase (p<0.05) in skeletal muscle deoxygenation in type 2 diabetes despite similar blood flow and recruitment of local blood volume/hematocrit. Within the control group skeletal muscle deoxygenation, local recruitment of microvascular blood volume/hematocrit, blood flow, and V02peak are all mutually correlated. None of these correlations were preserved in type 2 diabetes. Conclusions: These results suggest that in type 2 diabetes 1) skeletal muscle oxygenation is impaired, 2) this impairment may occur independently of bulk blood flow or local recruitment of blood volume/hematocrit, and 3) local and global metrics of oxygen transport are dissociated.
3. Introduction
According to CDC estimates, nearly half of American adults now have type 2 diabetes or prediabetes (94). The estimated lifetime risk of developing diabetes has risen to greater than 30% (95). People with type 2 diabetes suffer disproportionate cardiovascular and all-cause mortality, in addition to potentially disabling complications such as diabetic retinopathy and diabetic foot ulcer. The pathological mechanisms leading to excess morbidity and mortality in the diabetic population are not yet fully understood, but vascular and microvascular dysfunction are a common theme. Consistent with this observation, aerobic exercise capacity (VCtemax, a powerful clinical predictor of mortality (7; 11; 96)), is impaired in type 2 diabetes (47). Moreover, impaired aerobic exercise capacity is associated with diabetic complications (12) and insulin resistance (5), suggesting that the causes of impaired exercise capacity are intimately related to the broader pathology of type 2 diabetes. This


12
possibility mandates intensive investigation of the causes of impaired aerobic exercise capacity in type 2 diabetes.
Although the precise mechanisms by which VChmax is reduced in type 2 diabetes are not yet fully understood, impaired oxygen delivery is a likely contributor. Rodent studies reveal skeletal muscle hypoxia at the onset of exercise in rodents with diabetes (97), and these findings are corroborated by findings of increased skeletal muscle deoxygenation during exercise in humans with type 2 diabetes (52). There are several plausible mechanisms that may contribute to impaired oxygen delivery to skeletal muscle in type 2 diabetes, including reduced capillary density (70), reduced blood flow (51), loss of capillary perfusion (98), and heterogeneous distribution of blood flow (25; 63). Although each of these possible contributors has been previously noted, contradictory reports exist in the literature (especially with regards to bulk blood flow, e.g. (29)), and it remains unclear which specific parameters, if any, may limit oxygen delivery.
Oxygen delivery to peripheral tissues consists of convective (i.e. arrival of oxygenated blood) and diffusive (i.e. transport of oxygen from blood to mitochondria) steps. The convective step in oxygen delivery is primarily determined by bulk blood flow to the exercising muscle, and also determined to some extent by distribution of blood flow within the skeletal muscle circulation. In this manuscript, we report a metric of maximal leg blood flow as recorded during reactive hyperemia (RH) by venous occlusion plethysmography. The diffusive step of oxygen delivery has many determinants. One important component is the net recruitment of tissue hemoglobin content during exercise whether through microvascular recruitment, increased capillary hematocrit, or any other mechanism. Reductions in microvascular blood volume/hematocrit or its recruitment are widely reported in animal


13
models of type 2 diabetes (27; 70; 98), and these changes are likely to contribute to any defect in oxygen diffusion.
In this study, we used near-infrared spectroscopy (NIRS) to monitor skeletal muscle deoxygenation/oxygenation and local recruitment of microvascular blood volume/hematocrit in response to exercise. Interpretation of the NIRS signal is complex because tissue composition (e.g. adipose tissue thickness) influences NIRS results (99) and also because a majority of signal may come from myoglobin rather than hemoglobin (100). Furthermore, both hemoglobin and myoglobin can change in oxygenation status (i.e. oxygenated vs deoxygenated), but only hemoglobin content can acutely increase or decrease in the sampled tissue. Thus we interpreted changes in deoxygenation/oxygenation status (deoxy[hemoglobin +myoglobin], [HHb] and oxy[hemoglobin +myoglobin], [OHb]) as changes in muscle oxygenation, and interpreted changes in local signal intensity (total hemoglobin, [tHb]) as a change in a composite of local blood volume and microvascular hematocrit.
Our overarching hypothesis was that during exercise, either reduced blood flow, reduced local recruitment of microvascular blood volume/hematocrit, or both contribute to impaired oxygen delivery to skeletal muscle in type 2 diabetes. Based on our assessments of oxygen delivery and interpretations of NIRS signals, we formulated several sub-hypotheses to test the relationship of the convective and diffusive steps of oxygen delivery to reduced exercise capacity in type 2 diabetes: 1) RH leg blood flow correlates to VChpeak in both type 2 diabetes and in BMI-matched controls (i.e. blood flow limits oxygen uptake with or without type 2 diabetes), 2) increase in total hemoglobin correlates to VChpeak in type 2 diabetes but not in BMI-matched controls (i.e. local recruitment of microvascular blood volume/hematocrit limits oxygen diffusion only in type 2 diabetes), 3) RH leg blood flow


14
correlates to increase in total hemoglobin in both type 2 diabetes and in BMI-matched controls (i.e. blood flow and local recruitment of microvascular blood volume/hematocrit are coordinated), and 4) skeletal muscle deoxygenation correlates inversely with VCEpeak in type 2 diabetes but not in BMI-matched controls (i.e. skeletal muscle deoxygenation limits oxygen uptake in type 2 diabetes).
4. Methods
4.1 Source of data
The source of data analyzed in this manuscript is the INSITE study (Reusch (JEBR), Regensteiner (JGR) and Bauer (TAB), unpublished), which was designed to investigate differences in, and the effects of antioxidant treatment or exercise training on, exercise capacity and insulin sensitivity in overweight, middle aged men and premenopausal women. In this study, middle aged, overweight, and sedentary (defined as <1 hour of exercise per week) subjects either with (n=23) or without (n=18) type 2 diabetes underwent an incremental maximal exercise test on a cycling ergometer to assess V02peak by metabolic cart (Medgraphics CPX/D, Medical Graphics Corp., St. Paul, MN, USA) (JEB, JGR and TAB manuscript in progress). On a subsequent date, participants also performed two separate five-minute bouts of constant work rate cycling at 85% of lactate threshold, as determined by the V-slope method. Bouts were separated by a 10-minute rest period. Changes in muscle concentrations of [tHb], [OHb], and [HHb] were monitored by NIRS for the duration of the exercise protocol. Values for [tHb] and [HHb] used in this study were recorded in the vastus lateralis at rest and during constant work rate cycling at 85% of lactate threshold. In addition, maximal blood flow during reactive hyperemia (RH) was recorded using venous occlusion plethysmography.


15
4.2 NIRS data acquisition
Tissue total hemoglobin+myoglobin ([tHb]), deoxy[hemoglobin+myoglobin] ([HHb]), and oxy[hemoglobin+myoglobin] ([OHb]) were assessed by a frequency domain multi-distance NIRS monitor (Optiplex TS, ISS, Champaign, IL, USA) during each constant work rate exercise test. The NIRS monitor emits two wavelengths (690 and 830nm) and measures absorbance at distances of 2.0, 2.5, 3.0 and 3.5 cm. The NIRS data were sampled continuously and recorded at 50Hz. Upon export, data were down-sampled to 1 Hz using a running average of the higher-resolution 50 Hz data. During cycling exercise tests, the NIRS probe was positioned on the distal third of the vastus lateralis of the dominant limb, secured using a Velcro strap, and covered with a cloth bandage to exclude ambient light. The NIRS monitor was calibrated prior to each visit using a calibration phantom of known scattering and optical properties.
4.3 NIRS data analysis
Resting values of tissue [tHb], [HHb], and [OHb] were obtained by averaging the 30s prior to the onset of exercise. Exercise values of these parameters were obtained by averaging values between 270s and 300s following the onset of exercise. The absolute change in [tHb], [HHb], and [OHb] from rest to steady-state exercise were recorded as well. The change in [tHb] reflects local recruitment of microvascular blood volume/hematocrit (only the hemoglobin portion of the [tHb] signal can change acutely). The changes in [HHb] (deoxy[hemoglobin+myoglobin] accumulation) and [OHb] (oxy[hemoglobin+myoglobin] depletion) represent changes in local skeletal muscle oxygen availability and deoxygenation, respectively. Data which included negative values for concentrations of any hemoglobin


16
species were discarded. A visualization of the NIRS parameters employed in this analysis is
included in Figure 1.
NIRS Measures
Figure 1: Illustration of NIRS measures usedfor analysis. Following the onset of exercise, oxyhemoglobin (OHb) acutely decreases and then gradually returns partway to baseline. The depletion of oxy(hemoglobin+myoglobin) at steady-state was recorded as a metric of oxygen availability. Deoxy(hemoglobin+myoglobin) (HHb), meanwhile, increases gradually to steady state. The increase in [HHb] was recorded as a metric of muscle de oxygenation. Finally, total hemoglobin+myoglobin (tHb) comprises the sum of the [OHb] and [HHb] signals. The increase in [tHb] from rest to steady state exercise reflects the local increase in total hemoglobin concentration (i.e. local recruitment of microvascular blood volume hematocrit).
4.4 Inter-group comparisons
Values for [OHb] depletion (as assessed by change from rest to steady-state exercise) were compared between type 2 diabetes and control groups to assess the hypothesis that skeletal muscle oxygenation during exercise is impaired in type 2 diabetes. Values for [HHb] accumulation (as assessed by change from rest to steady-state exercise) were compared between type 2 diabetes and control groups to assess the hypothesis that skeletal muscle deoxygenation during exercise is increased in type 2 diabetes. Values for increase in total


17
hemoglobin were compared between type 2 diabetes and control groups to assess the hypothesis that the ability to recruit additional blood volume during exercise is impaired in type 2 diabetes. Values for RH blood flow were compared between type 2 diabetes and control groups, in order to assess the hypothesis that maximal blood flow is reduced in type 2 diabetes. Unpaired t-tests were used to compare groups for each of these comparisons, and a standard f-test was used to assess possible differences in group variance. Where significant differences in variance were found (p<0.05), Welchs correction (101) was applied to the original t-test to compare group means. Because each hypothesis was assessed independently (i.e. single comparison for each hypothesis), multiple comparisons corrections were not used in this analysis.
4.5 Intra-group comparisons
Increase in total hemoglobin and V02peak were compared within each group to assess the hypothesis that local recruitment of microvascular blood volume/hematocrit (as a determinant of oxygen diffusion) limits oxygen uptake in type 2 diabetes, but not in controls. Local deoxygenation and V02peak were compared within each group to assess the hypothesis that skeletal muscle deoxygenation limits oxygen uptake in type 2 diabetes, but not in controls. RH blood flow and V02peak were compared within each group to assess the hypothesis that that blood flow (as a determinant of oxygen convective delivery) limits oxygen uptake in both type 2 diabetes and controls. RH blood flow and increase in total hemoglobin were compared within each group to assess the hypothesis that whole limb blood flow and local recruitment of microvascular blood volume/hematocrit are associated in both type 2 diabetes and in controls. Pearsons R was used to assess the significance of correlations. Because each hypothesis was assessed independently (i.e. single comparison in


18
each group for each hypothesis), multiple comparisons corrections were not used in this analysis.
5. Results
5.1 Study subject characteristics
In this analysis, we used NIRS data from overweight, middle-aged subjects with (n=23) or without (n=18) type 2 diabetes. Study groups were well-matched in age (46.9+/-5.2 years type 2 diabetes vs 44.8+/-6.1 years control, p=NS), height (174.8+/-9.5 cm type 2 diabetes vs 173.1+/-9.9 cm control, p=NS), weight (92.6+/-17.5 kg type 2 diabetes vs 91.02+/-10.4 kg control, p=NS), and BMI (30.1+/-3.9 type 2 diabetes vs 30.4+/-2.7 control, p=NS). Study groups differed significantly (p<0.05) in HbAlc (6.9+/-0.8% type 2 diabetes vs 5.3+/-0.4% control) and sex balance (78% male type 2 diabetes vs 56% male control). Patients with type 2 diabetes had an average duration of diagnosis of 3.9+/-3.4 years. All group averages included above are expressed as mean+/-standard deviation.
5.2 Group differences in skeletal muscle deoxygenation/oxygenation and limb blood flow
Group comparisons of muscle deoxygenation/oxygenation, RH limb blood flow, and local recruitment of microvascular blood volume/hematocrit are shown in Figure 2. [HHb] accumulation from rest to steady-state exercise is significantly increased in type 2 diabetes relative to control (p=0.003, Figure 2A), consistent with our hypothesis of increased skeletal muscle deoxygenation in type 2 diabetes. [OHb] depletion from rest to steady-state exercise is significantly increased in type 2 diabetes relative to control (p=0.01, Figure 2B), consistent with our hypothesis of impaired skeletal muscle oxygen availability during exercise in type 2 diabetes. No significant difference (p=NS) was observed between type 2 diabetes and control


19
in increase in total hemoglobin, failing to support our hypothesis of impaired blood volume recruitment in type 2 diabetes. There was also no significant difference observed (p=NS) between type 2 diabetes and control in RH blood flow, failing to support our hypothesis of impaired limb blood flow in type 2 diabetes. Variance was significantly increased (p<0.05) in type 2 diabetes for all local NIRS metrics (increase in [tHb], [HHb] accumulation, and [OHb] depletion), but not for RH blood flow (p=NS).
Deoxy hemoglobin Accumulation
Oxyhemoglobin D ep letio n
mean -> p =0.003 variance -> p=0.Q02
s-i mean->p=NS
Figure 2: Impaired muscle oxygenation despite normal RH blood flow and increase in total hemoglobin in type 2 diabetes. (A) [HHb] accumulation is significantly (p=0.003) increased in type 2 diabetes, suggesting greater skeletal muscle deoxygenation during exercise. In addition, the variance of [HHb] accumulation is significantly (p=0.002) increased in type 2 diabetes, suggesting greater heterogeneity of skeletal muscle deoxygenation among type 2 diabetes subjects. (B) [OHb] depletion is significantly (p=0.01) increased in type 2 diabetes, suggesting reduced oxygen delivery. In addition, the variance of [OHb] depletion is significantly (p=0.003) increased in type 2 diabetes, suggesting greater heterogeneity of


20
skeletal muscle oxygenation. (C) Increase in total hemoglobin is not significantly different from controls in type 2 diabetes (p=NS), suggesting no difference in the increase in local recruitment of microvascular blood volume/hematocrit during moderate CWR exercise. However, the variance in the increase in total hemoglobin is significantly (p=0.006) increased in type 2 diabetes, suggesting greater heterogeneity of increase in total hemoglobin. (D) RH bloodflow does not significantly differ (p=NS) between Controls and type 2 diabetes, suggesting that total limb capacity for bloodflow is not limiting to muscle oxygenation.
5.3 Within-group comparisons of deoxygenation and increase in total hemoglobin to VChpeak
Comparisons within each group of deoxygenation and increase in total hemoglobin metrics to V02peak are shown in Figure 3 below. Increase in total hemoglobin significantly correlates (p=0.027) with V02peak in control subjects (Figure 3A) but not in type 2 diabetes subjects (Figure 3B, p=NS), failing to support our hypothesis that local recruitment of microvascular blood volume/hematocrit limits oxygen uptake in type 2 diabetes but not in controls. Local deoxygenation significantly correlates (p=0.0036) with V02peak in control subjects (Figure 3C) and does not correlate (p=NS) to V02peak in type 2 diabetes (Figure 3D), failing to support our hypothesis that skeletal muscle deoxygenation limits V02peak in type 2 diabetes but not in controls.


21
Control
Increase in Total Hemoglobin vs V02peak
Increase in Total Hemoglobin (JltHb], jiM)
C) Control
Deoxygenation vs V02peak
B)
40-1
E 30 55
20-
§
10-
D)
s
1
§
T2DM
Increase in Total Hemoglobin vs V02peak

R5=0.01
p=NS
i 10
i
15
20
Increase in Total Hemoglobin <-4ftH b], jiM)
T2DM
Deoxygenation vs V02peak
40 -|
30- *

20- *
10-
0-
R2=0.04
p=NS
i 10
i
15
20
i
25
[H H b] Accumulation ( |iM)
Figure 3: Increase in total hemoglobin and muscle deoxygenation do not correlate to V02peak in type 2 diabetes. (A) Increase in total hemoglobin and V02peak are significantly (p=0.027) correlated in control subjects. (B) Increase in total hemoglobin and V02peak are not correlated (p=NS) in type 2 diabetes. (C) Local deoxygenation and V02peak are significantly (p=0.0036) correlated in control subjects. (D) Local deoxygenation and V02peak are not correlated (p=NS) in type 2 diabetes.
5.4 Within-group comparisons of RH blood flow with VChpeak and increase in total hemoglobin
Comparisons of V02peak and increase in total hemoglobin are shown in Figure 4 below. RH blood flow correlates significantly (p=0.002) with V02peak in controls (Figure 4A) but not in type 2 diabetes (p=NS, Figure 4B), consistent with our hypothesis that blood flow limits oxygen uptake in controls, but failing to support our hypothesis that blood flow


22
limits oxygen uptake in type 2 diabetes. RH blood flow also correlates significantly
(p=0.018) with increase in total hemoglobin in controls (Figure 4C) but not in type 2 diabetes (p=NS, Figure 4D), consistent with our hypothesis that limb blood flow and local recruitment of microvascular blood volume/hematocrit are coordinated in controls, but failing to support
our hypothesis that the same is true in type 2 diabetes.
A)
40
31
*=
^ 20

ffl
& 10-
o
>
Control
Blood Flowvs VOjpeak
R2=0.54
o--
o
i--------1--------1------1------1
10 20 30 40 50
Plethy- Hypererric (rrl/dL'min)
B) T2DM
Blood Flow vs VOjpeak
Control
Blood Flow vs Increase in
Total Hemoglobin
!o
o
o
x
S
o
c
o

rs
Plethy- Hyperemic (ml/dUmin)
T2DM
Blood Flow vs Increase in
Figure 4: Mutual correlations between RH bloodflow, increase in total hemoglobin, and V02peak are abolished in type 2 diabetes. (A) RH bloodflow and V02peak are significantly correlated (p=0.002) in control subjects. (B) RH bloodflow and V02peak are not correlated (p=NS) in type 2 diabetes. (C) RH bloodflow and increase in total hemoglobin are significantly correlated (p=0.018) in control subjects. (D) RH bloodflow and increase in total hemoglobin are not correlated (p=NS) in type 2 diabetes.


23
6. Discussion
In this analysis, we found evidence that skeletal muscle oxygen availability is reduced and skeletal muscle deoxygenation is increased in exercise with type 2 diabetes as compared to a similarly overweight and sedentary control group, consistent with previous reports (52; 97; 102; 103). We also found that the increase in total hemoglobin during exercise is not reduced in type 2 diabetes. This finding is consistent with reports of preserved microvascular recruitment (measured using contrast enhanced ultrasound) during exercise with type 2 diabetes (104). We also found that RH blood flow was not impaired in type 2 diabetes. Previous studies have found conflicting results when investigating blood flow limitations in type 2 diabetes (29; 51), and the degree to which health status differs between disease and control groups may account for these inconsistencies. Finally, we found that variance of all local oxygenation metrics ([HHb] accumulation, [OHb] depletion, and increase in [tHb]) was increased in type 2 diabetes, despite no increase in variance of global or whole-limb oxygen transport metrics (VCtepeak and RH blood flow). Collectively, these results suggest that skeletal muscle oxygenation during exercise is impaired in type 2 diabetes, and that this impairment does not appear to be related to either decreased limb blood flow or local recruitment of microvascular blood volume/hematocrit.
Our correlation analysis returned unexpected results. As hypothesized, maximal RH blood flow and V02Peak were correlated in control subjects, but this correlation was lost in type 2 diabetes. Meanwhile, increase in total hemoglobin was correlated to V02Peak in control subjects but not in type 2 diabetes, contrary to our hypothesis that local recruitment of microvascular blood volume/hematocrit would limit exercise capacity only in type 2 diabetes. The recurring theme of correlation in health and dissociation in type 2 diabetes


24
continued with dissociation of skeletal muscle deoxygenation from VChpeakand dissociation of maximal RH blood flow from increase in total hemoglobin. Collectively, these findings suggest that in type 2 diabetes, local metrics of tissue oxygen transport do not reflect whole-limb or whole-body oxygen transport, blood flow is dissociated from local recruitment of microvascular blood volume/hematocrit, and factors besides bulk blood flow and local recruitment of microvascular blood volume/hematocrit are likely to limit oxygen uptake during exercise. Any plausible and complete explanation of these data must satisfy three observations: 1) skeletal muscle oxygenation is impaired in type 2 diabetes, 2) skeletal muscle oxygenation is more heterogeneous in type 2 diabetes, and 3) these changes appear to be unrelated to bulk blood flow or local recruitment of microvascular blood volume/hematocrit.
While our NIRS data do not allow us to draw definitive conclusions about the causes of impaired skeletal muscle oxygenation in type 2 diabetes, the literature does provide a plausible explanation that could satisfy all three of these requirements. Distribution of blood flow is more spatially heterogeneous in the obese Zucker rat model of type 2 diabetes, and this perfusion heterogeneity is associated with impaired oxygen uptake (25; 26; 60; 63). Simulation studies reveal that heterogeneous perfusion results in impaired muscle oxygenation on average, because over-perfused vessels cannot fully compensate for underperfused vessels (63-66). By definition, heterogeneous perfusion not resulting from heterogeneous tissue demand would result in flow/VCh mismatch. Moreover, these effects would influence skeletal muscle deoxygenation/oxygenation even in cases of normal bulk blood flow and local recruitment of microvascular blood volume/hematocrit. Although the findings referenced above have not yet been translated to human type 2 diabetes, it has been


25
shown in humans that increased perfusion heterogeneity correlates with reduced oxygen extraction, and perfusion heterogeneity decreases in response to endurance training (67; 68), supporting the plausibility of this explanation. Moreover, the effects of age on the vasculature, which parallel the influence of type 2 diabetes (70; 105), are themselves spatially heterogeneous (106). Visualizations of skeletal muscle perfusion in type 2 diabetes reported by Zheng et al appear more heterogeneous qualitatively (59).
It is worth noting, however, that the studies cited in the previous paragraph pertain to spatial heterogeneity within a single muscle, whereas the heterogeneity we observed in our NIRS was measured as population heterogeneity, and only one site per muscle was assessed. Although spatial heterogeneity could produce population heterogeneity in local but not global metrics as a statistical artifact if only one site were observed (which is exactly what our study design entailed), we cannot rule out the possibility that population heterogeneity was observed due to variance among study subjects rather than observed as an artifact of variance among locations. Future studies will be required to test the hypothesis that heterogeneous distribution of blood flow on both microvascular and macrovascular scales during exercise may contribute to reduced oxygen delivery to skeletal muscle in type 2 diabetes.
There are some limitations of our study design. First, leg blood flow, VChpeak, and skeletal muscle oxygenation measures were each recorded under separate conditions (following venous occlusion, during maximal cycling exercise, and during submaximal cycling exercise, respectively). It is possible that these discrepancies influenced our results. However, given that these disparate metrics were correlated in health but not in type 2 diabetes, the recurring theme of dissociation in type 2 diabetes remains relevant. The use of


26
NIRS to assess skeletal muscle deoxygenation/oxygenation may also introduce complexity to the interpretation of our data, given that there is controversy in the field about the relative contributions of hemoglobin (i.e. vascular) and myoglobin (i.e. muscular) contributions to the NIRS signal (100). Our analysis avoided this issue by reporting local deoxygenation/oxygenation without any attempt to distinguish vascular/intramuscular contributions and by reporting only the change in [tHb] from rest to exercise as a vascular-specific measure. This parameter must be vascular in origin, given that myoglobin does not acutely enter or leave the muscle upon contraction. Finally, the sex imbalance in our type 2 diabetes group might be expected to influence our results. If sex differences were driving the observed correlations in health and dissociations in type 2 diabetes, however, one would expect reduced variance in the type 2 diabetes group, and in fact we observed increased variance in type 2 diabetes. Thus, the dissociations in type 2 diabetes occurred despite a greater dynamic range over which correlations could be observed.
In summary, we found that skeletal muscle oxygenation during exercise is impaired in type 2 diabetes, and that this impairment can occur independently of changes in limb blood flow or local recruitment of microvascular blood volume/hematocrit. We also found that correlations between local and global oxygen transport metrics were abolished in type 2 diabetes, and that local muscle oxygenation is more heterogeneous. Although our data do not allow definitive conclusions as to the cause of these changes, it is plausible that heterogeneous blood flow distribution may account for dissociation of local and global oxygen transport in type 2 diabetes. Future studies will be required to more fully understand the heterogeneity and impaired coordination of skeletal muscle oxygenation in type 2 diabetes. In light of the previously discussed association of impaired exercise capacity with


27
premature mortality and excess morbidity (7; 11; 96), understanding the mechanisms leading to impaired oxygen transport in type 2 diabetes holds great translational potential.


28
CHAPTER III
A CONCEPTUAL FRAMEWORK FOR PREDICTING AND ADDRESSING THE CONSEQUENCES OF DISEASE-RELATED MICROVSASCULAR DYSFUNCTION
Co-authors: Jefferson C. Frisbee and Jane E. B. Reusch
1. Preface
It is plausible that the diabetic skeletal muscle oxygenation heterogeneity described in the previous chapter is related to heterogeneous vascular delivery of oxygen. This chapter describes a series of sensitivity analyses drawing from established principles in mass transport and anatomy aiming to elucidate the roles of bulk blood flow, capillary density, perfusion heterogeneity to microvascular solute flux. These analyses apply to the exchange of not only oxygen, but of all small, blood-borne molecules. The results of this analysis provide a theoretical framework for understanding how blood flow and its distribution relate to clinical phenotypes such as impaired exercise capacity and insulin resistance, and provide a plausible explanation for why those two phenotypes are correlated. The key finding of this study was that the distribution of blood flow (number of perfused capillaries and flow distribution within these capillaries) can exert profound effects on tissue oxygenation and other, related mass transport processes independently of bulk blood flow. This chapter was published in Microcirculation in 2017.
2. Abstract
Objective: A growing body of evidence indicates that impaired microvascular perfusion plays a pathological role in a number of diseases. This manuscript aims to better define which aspects of microvascular perfusion are important, what mass transport processes (e.g. insulin action, tissue oxygenation) may be impacted, and what therapies might


29
reverse these pathologies. Methods: We derive a theory of microvascular perfusion and solute flux drawing from established relationships in mass transport and anatomy. We then apply this theory to predict relationships between microvascular perfusion parameters and microvascular solute flux. Results: For convection-limited exchange processes (e.g. pulmonary oxygen uptake), our model predicts that bulk blood flow is of primary importance. For diffusion-limited exchange processes (e.g. insulin action), our model predicts that perfused capillary density is of primary importance. For convection/diffusion co-limited exchange processes (e.g. tissue oxygenation), our model predicts that various microvascular perfusion parameters interact in a complex, context-specific manner. We further show that our model can predict established mass transport defects in disease (e.g. insulin resistance in diabetes). Conclusions: The contributions of microvascular perfusion parameters to tissue-level solute flux can be described using a minimal mathematical model. Our results hold promise for informing therapeutic interventions targeting microvascular perfusion.
3. Introduction
Research investigating microvascular perfusion has traditionally focused on the quantity of blood flow supplied by upstream arteries. The scientific literature increasingly shows that the distribution of this blood flow within the microcirculation is also physiologically important. Evidence of microvascular dysfunction contributing to tissue hypoxia has been reported in the metabolic syndrome (25; 26) and sepsis (107; 108). Similar microvascular defects have been reported in Alzheimers disease (109), inflammatory bowel disease (110), and hypertension and obesity (111). Hints of a mechanism linking impaired microvascular perfusion to tissue hypoxia can be found in simulation studies of ischemic


30
heart disease (65) and cerebrovascular disease (64). Tissue hypoxia is the most common focus of microvascular perfusion studies, but other solute exchange processes are also influenced by microvascular perfusion. For example, animal models of insulin resistance show that the primary barrier to glucose flux in the insulin resistant state lies in the extracellular (i.e. vascular to interstitial) step of glucose and insulin delivery (112). The sheer variety of diseases in which microvascular perfusion defects occur and the variety of consequences related to impaired perfusion suggests that microvascular perfusion is a critical physiological process in its own right.
This manuscript will discuss the importance of three distinct perfusion parameters: the amount of blood flowing through a microvascular network (bulk blood flow), its distribution at microvascular bifurcations (perfusion heterogeneity), and the number of capillaries accessible to flowing blood (perfused capillary density). It is unclear which solute exchange processes (e.g. oxygen delivery, lactate clearance) are impacted by which specific microvascular perfusion parameters and how perfusion abnormalities might be therapeutically targeted. The importance of bulk blood flow is illustrated by diseases that substantially decrease blood supply (e.g. heart failure, peripheral arterial disease, etc). Perfusion heterogeneity has been empirically demonstrated to modulate oxygen flux in skeletal muscle (25; 26), and is also thought to contribute to ischemic heart disease and cerebrovascular disease (64; 65). Perfused capillary density is modulated both by long-term changes in anatomical capillary density (density of capillaries present within the tissue, perfused or not) and by short-term changes in the fraction of capillaries present that are actually perfused. Long-term changes in anatomical capillary density can be stimulated by diseases such as type 2 diabetes mellitus (T2DM) (70) and physiological stressors such as


31
exercise training (113). Short-term changes in perfused capillary density can be caused by stimuli such as endothelial glycocalyx degradation, as is observed in sepsis and hyperglycemia (81; 114). Consideration of these parameters in combination rather than in isolation allows inference of their relative contextual importance.
In addition to developing a generalized theory of microvascular perfusion and solute flux, we will also attempt to persuade the reader of its utility in identifying therapeutic strategies. This is perhaps best illustrated by successful examples of therapies targeting perfused capillary density. For example, one major cause of reductions in perfused capillary density is plugging of capillaries by adherent leukocytes or microemboli (115; 116). Persistent reduction in perfused capillary density is a major determinant of organ failure and mortality in sepsis (117). As would be expected from these findings, therapies that reduce adhesive interactions in the microvasculature (and thus un-plug blocked capillaries) also improve patient outcomes (108). Anatomical capillary density and hemostatic status are also major determinants of insulin sensitivity in T2DM (118-122). Insulin sensitizing drugs such as metformin tend to also reduce adhesive interactions in the microvasculature (123), and this may be part of their mechanism of action. In both sepsis and T2DM, therapies that improve microvascular function served to treat pathologies that are not generally discussed as microvascular defects. In both cases, some element of trial and error was required to even identify the microcirculation as a potential issue, and again to determine which aspects of microvascular perfusion were most relevant. Trial and error was then required to identify therapies to exploit these parameters, and often this mechanism of action was discovered post-hoc. The analysis included in this manuscript aims to enable translational researchers to


32
bypass much of this confusion by accurately predicting which microvascular parameters are relevant to their research and how they might be targeted.
4.1 Governing Equations
We begin our analysis with a widely used formulation of microvascular solute flux derived from Ficks laws of diffusion (124):
Here Js is solute flux across the endothelium, Q is capillary blood flow, Ca is arterial concentration, CL is interstitial concentration, Pisa metric of permeability to the solute of interest, and A is capillary surface area. Note that S is used instead for capillary surface area in traditional representations of this formula. We elected to use A to denote surface area in this analysis so as not to conflict with subsequent use of S to denote sensitivity. For purposes of subsequent sensitivity analyses, we further specify the definition of permeability using the following equation:
Here V is capillary volume, Pisa diffusion rate constant, and Rd is the effective radius of diffusion. Using this definition of permeability, a more complete formulation of solute flux is:
4. Methods
(1)
(2)
(3)
For simplicity of presentation in subsequent analyses, we will also introduce the
dimensionless diffusion/convection matching parameter (3 (Equation 4). Values of /3 much


33
greater than 1 reflect an excess of diffusion capacity relative to bulk blood flow, while those much less than 1 reflect an excess of bulk blood flow relative to diffusion capacity.
P =
D*A*V
Rd*Q
(4)
Incorporating this simplified notation into Equations 1-3 yields:
Js = Q *(Ca~ Ct) (1 e~P) (5)
This governing equation will subsequently be applied to determine the effects of bulk blood flow (sections 5.1 and 5.2), diffusion capacity (sections 5.1 and 5.3), perfused capillary density (section 5.4), and perfusion heterogeneity (section 5.5), along with simulations of the consequences of experimentally defined phenotypes including alterations in each of these parameters on microvascular solute flux (section 5.6).
4.2 Fractional Equilibration and Diffusion/Convection Matching
Individual determinants of diffusion/convection matching (e.g. capillary blood volume, capillary surface area, diffusion capacity etc.) are difficult to measure. However, their physical consequences can be empirically observed using fractional equilibration (s) of the solute of interest between capillary blood and the interstitium, as defined by Eugene Renkin (125). Fractional equilibration is defined as the ratio of arteriovenous concentration difference to arterial-interstitial concentration difference (Equation 6, Figure 5A). Determining fractional equilibration requires measurement of solute concentration in three compartments- arterial blood (Ca), venous blood (Cv), and interstitial fluid (Q). Plausible reference values of Ca, Cv, 6), and e for compounds of physiologic interest are included in Figure 5B.
_ Cg~Cy
Ca~Ci
(6)


34
A) B)
Capillaries
Concentration
Interstitium
Example Compounds
Solute Ca C cv £
Oxygen (lung) 40 mniHg 105 mmHg 100 mmHg 92%
Oxygen (muscle) 100 mmHg 20 mmHg 40 mmHg 75%
Lactate (muscle) 3.2 mM 5 mM 4 mM 44%
Glucose (brain) 5mM 1 mM 4 mM 20%
Insulin (muscle) 500 pM 100 pM 490 pM 2.5%
Figure 5: Solutes vary in their degree fractional equilibration with the interstitium during capillary transit. (A) Visualization of the concept of equilibration. Complete equilibration implies that venous solute concentration is equal to interstitial solute concentration, while partial equilibration implies that venous concentration lies somewhere between arterial and interstitial concentrations. (B) Examples of physiologically relevant compounds and plausible values for their fractional equilibration.
Fractional equilibration is related to the diffusion/convection matching parameter (1 by following transformation:
E = l-e~P (7)
£ approaches 1 in cases where diffusion capacity exceeds convective delivery (e.g. pulmonary oxygen flux) and £ approaches zero in cases where convective delivery exceeds diffusion capacity (e.g. insulin delivery to skeletal muscle).
4.3 Blood Flow Distribution
To determine the influence of arteriolar perfusion heterogeneity on microvascular solute flux, we simulated blood flow distribution within an idealized arteriolar network as illustrated in Figure 6. Each arteriole bifurcates into two smaller daughter vessels (Figure 6A) for 15 consecutive vessel generations for a total of 215 simulated capillaries. Distribution of blood flow at each bifurcation is described by the parameter y, which is defined as the


35
fraction of parent vessel blood flow directed to the higher-flow daughter vessel (Figure 6B). Thus y = 0-5 represents homogenous perfusion, and increasing deviation from 0.5 represents increasingly heterogeneous perfusion.
Figure 6: Simulated distribution of blood flow in an idealized microvascular network. (A)
The simulated microvascular network consisted of a series of arterioles terminating in a symmetric bifurcation of two smaller arterioles. Fifteen vessel generations were simulated for a total of 215 individual capillaries; only four vessel generations are illustrated to preserve clarity. (B) Distribution of flow rate at each microvascidar bifurcation was defined by the parameter y, which is defined as the fraction ofparent vessel flow directed to the higher-flow daughter vessel. Figure recreated with permissions from Butcher et al 2013 (23).
5. Results
5.1 Effects of Blood Flow and Diffusion Capacity
The effects of bulk blood flow on microvascular solute flux are visualized in Figure 7A. The effects of diffusion capacity on microvascular solute flux are visualized in Figure 7B. These results reflect the predictions of Equation 8. Increases in blood flow and diffusion capacity both result in saturable increases in microvascular solute flux. The level at which increasing blood flow yields diminishing returns increases with increasing diffusion capacity, and vice versa. For diffusion-limited, low fractional equilibration molecules (e.g. insulin), microvascular solute flux increases robustly with increasing diffusion capacity but increases minimally with increasing blood flow. For convection-limited, high equilibration molecules


36
(e.g. oxygen in the lung), microvascular solute flux increases minimally with increasing diffusion capacity but increases robustly with increasing blood flow. For diffusion/convection co-limited, intermediate equilibration molecules (e.g. oxygen in skeletal muscle), both diffusion capacity and blood flow limit the rate of microvascular solute flux.
Figure 7: Effects of blood flow and diffusion capacity on microvascular solute flux. (A) Solute flax increases asymptotically with increasing flow. Increasing diffusion capacity increases the threshold of diminishing returns for increasing flow. (B) Solute flax increases asymptotically with increasing diffusion capacity. Increasing flow increases the threshold of diminishing returns for increasing diffusion capacity.
5.2 Sensitivity of Solute Flux to Blood Flow
To quantify the sensitivity of microvascular solute flux to blood flow, the modified convection/diffusion matching parameter /?' was defined as a function of fraction of baseline flow rate (/), which alters convection/diffusion matching solely through its interaction with blood flow at baseline (Q0):
P' =
D*A*V
Rd*Qi
D*A*V
Rd*Qo*f
Po
f
(8)
Combining Equations 8 and 5 while accounting for increased blood flow yields:


37
Is
f Qo (Q Q) * e f )
(9)
The sensitivity of microvascular solute flux to changes in blood flow (Sf) is defined as the partial derivative of microvascular solute flux (Js') with respect to fraction of baseline blood flow (/) normalized to baseline microvascular solute flux (Js,oY
Baseline blood flow (Q0) and arterial-interstitial concentration gradient (Ca Q) cancel out when performing this calculation, and so a transformation between fractional equilibration and convection/diffusion matching at baseline (/?0) can be used to solve for Sf as a function fractional equilibration at baseline (£0):
The results of this calculation are visualized in Figure 8. Note that fractional equilibration is also influenced by circumstances that change solute flux, and so £0 is not necessarily equal to s'. For convection-limited compounds which undergo near-complete equilibration with the interstitium (e.g. oxygen in the lung), flux increases or decreases near-proportionally with increasing/decreasing bulk blood flow. For diffusion-limited compounds which undergo minimal equilibration with the interstitium (e.g. insulin), flux changes minimally with changes in bulk blood flow. Flux of compounds at intermediate values of baseline fractional equilibration undergoes sub-proportional changes in response to changing
(10)

(11)
blood flow.


38
Sensitivity to Blood Flow
Fractional Equilibration at Baseline (e0)
Diffusion
limited
Convection
limited
Figure 8: Relationship between baseline fractional equilibration and sensitivity to blood flow. Convection-limited compounds, which equilibrate completely or near-completely with the interstitium during capillary transit, undergo a proportional increase decrease in flax in response to an increase decrease in bloodflow. Diffusion-limited compounds, which equilibrate minimally with the inter stitium during capillary transit, are minimally sensitive to changes in bloodflow.
5.3 Sensitivity of Solute Flux to Diffusion Rate Constant
To determine the sensitivity of microvascular solute flux to diffusion capacity, we defined the modified convection/diffusion matching parameter /?' as a function of fraction of baseline diffusion rate constant (k), which alters convection/diffusion matching solely
through its interaction with diffusion capacity at baseline (D0):
P' =
d'*s*v
Rd*Q
D0*k*S*V
Rd*Q
Po k
Combining Equations 12 and 5 yields:
Js =Q*{Ca-Ci)*{l-e(--P^)
(12)
(13)
The sensitivity of microvascular solute flux to changes in diffusion rate constant (Sk) is defined as the partial derivative of modified microvascular solute flux (Jf) with respect to


39
fraction of baseline diffusion rate constant (k) normalized to baseline microvascular solute
Baseline blood flow (Q0) and arterial-interstitial concentration gradient (Ca Q) cancel out when performing this calculation, and so the transformation between fractional equilibration and convection/diffusion matching parameter at baseline (/?0) can be used to solve for Sk as a function of fractional equilibration at baseline (f 0):
to(-d)*( l-e0)
sk = Vl~£;---------- (If
£0
The results of this calculation are visualized in Figure 9. For convection-limited compounds which undergo near-complete equilibration with the interstitium (e.g. oxygen in the lung), flux changes minimally with changes in diffusion rate constant. For diffusion-limited compounds which undergo minimal equilibration with the interstitium during capillary transit (e.g. insulin), flux increases or decreases near-proportionally with increasing/decreasing diffusion rate constant. Flux of compounds at intermediate values of baseline fractional equilibration undergoes sub-proportional changes in response to changes
flux (/s 0):
(14)
in diffusion rate constant.


40
Sensitivity to Permeability
Fractional Equilibration at Baseline (e0)
Diffusion
limited
Convection
limited
Figure 9: Relationship between baseline fractional equilibration and sensitivity to diffusion capacity. Convection-limited compounds, which equilibrate completely or near-completely with the interstitium during capillary transit, are minimally sensitive to changes in diffusion capacity. Diffusion-limited compounds, which equilibrate minimally with the inter stitium during capillary transit, undergo a proportional increase decrease in flax in response to an increase decrease in diffusion capacity.
5.4 Sensitivity of Solute Flux to Perfused Capillary Density
To determine the sensitivity of microvascular solute flux to perfused capillary density, we defined the modified convection/diffusion matching parameter /?' as a function of fraction of baseline perfused capillary density (d). Capillary density influences diffusion/convection matching through a variety of parameters. In this analysis, two modes of interaction with perfused capillary density were considered. In the first case (Equations 16-19), the concentration gradient within the interstitium is substantial and the solute of interest diffuses freely across the endothelium (e.g. lactate clearance). In this case, capillary density modulates the effective diffusion radius {Rd:0) along with capillary surface area (i40) and
blood volume (E0):


41
P' =
D*Ar*Vr D*A0*d*V0*d
Po d3
(16)
Rd*Q (Rd,o/d)*Q
Combining Equations 16 and 5 yields:
Js'= Q*(Ca-Cl)
(17)
The sensitivity of microvascular solute flux to relative change in perfused capillary density (Sd) is defined as the partial derivative of modified microvascular solute flux (Js') with respect to fraction of baseline perfused capillary density (d) normalized to baseline microvascular solute flux (JSoY
Blood flow (Q) and arterial-interstitial concentration gradient (Ca Q) cancel out when performing this calculation, and so the transformation between fractional equilibration and convection/diffusion matching at baseline (/?0) can be used to solve for Sd as a function of baseline fractional equilibration at baseline (£0):
Alternately, interstitial concentration gradients may be negligible relative to trans-endothelial concentration gradients (e.g. oxygen in skeletal muscle). In this case, the effective radius of diffusion (Rd) is not appreciably altered by perfused capillary density. Repeating the same procedure as in Equations 16-19 with this modification yields:
(18)
P =
D*Ar*Vr D*A0*d*V0*d
Po d2
(20)
Rd*Q Rd*Q
(21)
(22)


42
=

~£0
£0
(23)
The results of these calculations are visualized in Figure 10. Although most compounds will be co-limited by trans-endothelial and interstitial diffusion processes, the strictly endothelium-limited and strictly interstitium-limited results shown here comprise the lower/upper bounds of sensitivity to perfused capillary density. Flux of convection-limited compounds, which undergoes near-complete equilibration with the interstitium (e.g. oxygen in the lung), changes minimally with changes in perfused capillary density. Flux of diffusion-limited compounds which undergo minimal equilibration with the interstitium (e.g. insulin), will undergo increases or decreases between 2x and 3x larger than the corresponding incremental change in perfused capillary density. Flux of compounds at intermediate values of baseline fractional equilibration undergoes somewhat smaller but still supra-proportional changes in response to changes in diffusion rate constant.
Sensitivity to Capillary Density
Fractional Equilibration at Baseline (£0)
Diffusion
limited
Convection
limited
Figure 10: Relationship between baseline fractional equilibration and sensitivity to perfused capillary density. Convection-limited compounds, which equilibrate completely or near-


43
completely with the interstitium during capillary transit, are minimally sensitive to changes in perfused capillary density. Diffusion-limited compounds, which equilibrate minimally with the inter stitium during capillary transit, undergo a 2x-3x proportional increase/decrease in flux in response to an incremental increase/decrease in perfused capillary density. The magnitude of the effect ofperfused capillary density further depends upon whether the primary site of resistance to solute diffusion lies at the endothelium or within the inter stitium, interstitium-limited compounds being more sensitive to perfused capillary density than endothelium-limited compounds.
Note that the sensitivity of microvascular solute flux to perfused capillary density is geometrically identical to the sensitivity to diffusion rate constant, only scaled up 2x-3x.
Thus a 1% decrease in perfused capillary density could be fully compensated by a 2%-3% increase in diffusion rate constant, and vice versa.
5.5 Sensitivity of Solute Flux to Microvascular Perfusion Heterogeneity
In order to define the influence of perfusion heterogeneity on microvascular solute flux, we defined the total microvascular solute flux across a capillary network (Jtot) based on the mean capillary flow rate (Q) and the fraction of mean flow rate within each capillary (/, ):
Because the degree of perfusion heterogeneity (y) was found to influence the degree of sensitivity to perfusion heterogeneity (i.e. flux does not vary linearly with y), it is useful to discuss the sensitivity of microvascular solute flux to perfusion heterogeneity in terms of the
(25)
relative (%) change in solute flux between two representative perfusion states (see section 4.3
for discussion of perfusion distribution simulation):
%A/tot,l->2 100 *
Jtot, 1 Jtot,2
JtOt, 1
(26)
Mean capillary blood flow (Q) and arterial-interstitial concentration gradient (Ca
Ci) cancel out when performing this calculation, and so %A/tot can be expressed as a


44
function of mean solute convection/diffusion matching (/?) and blood flow distribution (as determined by y):
Z/ir
l-e
/?
/it
-Z/2f
l-e ?2i
1-2 100 *

P
l-e fn
(27)
For illustration purposes, we chose to simulate the differences in microvascular solute flux between Lean Zucker Rats (LZR, y = 0.52) and Obese Zucker Rats (OZR, animal model of T2DM, y = 0.59), as differences in arteriolar perfusion heterogeneity in these models has been extensively characterized in previous studies (23-26; 60). Figure 11 shows the predicted % decrease in solute flux in OZR relative to LZR as a function of fractional equilibration in the LZR. Compounds whose flux is entirely limited by diffusion or convection are not appreciably impacted by perfusion heterogeneity. However, intermediate-equilibration compounds whose flux is co-limited by diffusion and convection (e.g. oxygen in skeletal muscle) are sensitive to perfusion heterogeneity because over-perfused capillaries cannot fully compensate for under-perfused capillaries. For these molecules, losses in capillary perfusion are more impactful than are equivalent gains.


45
Sensitivity to Perfusion Heterogeneity
Fractional Equilibration at Baseline (£0)
Diffusion
limited
Convection
limited
Figure 11: Relationship between baseline fractional equilibration and sensitivity to arteriolar perfusion heterogeneity. Compounds whose flax is limited almost entirely by convection or almost entirely by diffusion are minimally affected by perfusion heterogeneity. Compounds whose flax is co-limited by diffusion and convection undergo reduced microvascular flax under conditions of microvascular perfusion heterogeneity. The range of fractional equilibration most affected by perfusion heterogeneity gradually shifts towards relatively diffusion-limited compounds with an increasing degree of perfusion heterogeneity (see Figure 12). The curve shown reflects the impacts of perfusion heterogeneity on solute flax in the OZR relative to the LZR.
5.6 Defining the Effects of Complex Phenotypes on Microvascular Solute Flux
The procedure for determining the effects of perfusion heterogeneity outlined in Equations 25-27 can also be used to determine the effects of a complex phenotype involving several alterations to microvascular perfusion. For this analysis we introduced parameters for fraction of healthy bulk blood flow through the entire microvascular network (fQ) and fraction of healthy perfused capillary density (fD):
' Po*fD3
1 e /*/£
hot Yjf Q Qo h (fla cd *
(28)


46
/ _£o\ / Po*fP3\
E Ai* 1-e fil -Z/Q*Ai* 1-e n*f2i
%AJtot, i->2 = 100 *-------------------^4----------^ (29)
SAi* 1-e M
For demonstration purposes, we characterized the effects of microvascular perfusion defects on solute flux in T2DM (OZR model), sepsis, and acute hyperglycemia or glycocalyx degradation (acute hyperglycemia causes glycocalyx degradation and related perfusion defects (81)). In the OZR model of T2DM, bulk blood flow is reduced by -20%, capillary density is reduced by -20%, and perfusion heterogeneity is increased from y = 0.52 to y = 0.59 (25). In sepsis, distribution of blood flow to meet metabolic demands is impaired, perfused capillary density is markedly reduced, and microvascular perfusion heterogeneity is readily visible under microscopic observation (84). To simulate a worst-case perfusion scenario, we assumed a 20% reduction in bulk blood flow, a 50% reduction in perfused capillary density, and an increase in y from 0.52 to 0.7. In cases of acute glycocalyx degradation or hyperglycemia, bulk blood flow is not significantly altered, perfused capillary density is reduced by -30%, and perfusion heterogeneity is increased, although the precise degree of this increase is unclear (114). For purposes of this analysis, perfusion heterogeneity during hyperglycemia was assumed be similar to that in the OZR. The results of this analysis are shown in Figure 12 below. Our model predicts that microvascular solute flux will be impaired in all three phenotypes, but that the precise etiology of this impairment varies.


47
A) T2DM fOZRl
Sepsis
Ra nge of i nterest for Range of i nterest for insuli n acti on exercise ca pacify
Range of interest for Range of interest for insulin action exercise capacity
Hyperglycemia or Glycocalyx Degradation
Range of i rite rest for Range of interest for
insulin action exercise capacity
Fractional Equilibration in Normoglycemia (£q)
Figure 12: Complex phenotypes influence solute flux through a variety of mechanisms. (A) Solute flax in diabetes is predicted to be affected primarily by perfused capillary density in the range of interest for insulin action, and by a complex mixture of bulk bloodflow, perfusion heterogeneity, and capillary density effects in the range of interest for exercise


48
capacity. (B) Solute flux in sepsis is predicted to be affected primarily by perfused capillary density and perfusion heterogeneity in the range of interest for both insulin action and exercise capacity. (C) Solute flux in hyperglycemia or gly cocalyx degradation is predicted to be affected primarily by perfused capillary density in the range of interest for insulin action, and by a combination of capillary density and perfusion heterogeneity in the range of interest for exercise capacity.
6. Discussion
In this manuscript, we derive a generalized theory of microvascular perfusion and solute flux in branching microvascular networks building from the single-capillary analysis of Eugene Renkin (125). This theory enables a number of physiologic predictions, which we will discuss beginning with associations between physiologic parameters, progressing to the implications of complex phenotypes, and eventually moving on to therapies effective for acute treatment of microvascular perfusion defects. Throughout our discussion, we will use exercise capacity (or tissue oxygenation where appropriate) and insulin action as model solute exchange processes. We encourage the reader to follow along with another, selfrelevant exchange process (e.g. drug delivery) in mind. Where the predictions we draw this model have been tested, both the causes and the solutions of impaired microvascular flux can typically be predicted from our model. In addition, we will outline the limitations of our model for the readers consideration when applying our theory.
6.1 Physiological associations
We begin with predictions of physiologic associations. The first is that pulmonary oxygen uptake (a convection-limited exchange process) should vary proportionally with bulk blood flow. Consistent with this prediction, cardiac output is the primary determinant of pulmonary oxygen uptake (126). Conversely, our model predicts that increasing bulk blood flow without correcting its distribution would do little to enhance insulin action in skeletal muscle (a diffusion-limited process). This prediction is plausible- animal models of insulin


49
resistance often display increased cardiac output (127) and regional blood flow diverges from regional glucose uptake (128). Further, our model predicts that capillary density (a determinant of diffusion capacity) would be a major determinant of insulin action. The observation that capillary density and insulin sensitivity are associated in both human and animal models support this prediction (118; 122). Our model also predicts that capillary density would be related, albeit weakly, to skeletal muscle oxygen uptake (a diffusion/convection co-limited process). This prediction is consistent with the observations that capillary density and V02max are correlated in peripheral arterial disease (129), and capillary density increases with aerobic exercise training (113). Finally, our model predicts that pulmonary capillary density (a determinant of diffusion capacity) would have little impact on pulmonary oxygen uptake (convection-limited) except in extreme cases. It is unclear that this prediction has been directly tested, but it is telling that canonical descriptions of diseases that interfere with pulmonary oxygen diffusion do not involve capillary rarefaction (130).
6.2 Disease phenotype predictions
We will now shift to discussion of specific microvascular perfusion phenotypes. For this purpose we will employ three physiologic states: 1) T2DM, 2) sepsis, and 2) acute glycocalyx degradation. Both T2DM (Figure 12A) and sepsis (Figure 12B) decrease perfused capillary density and increase perfusion heterogeneity. Our model therefore predicts insulin resistance in both disease states as a result of diffusion limitations. These same diffusion limitations would be expected to cause impaired exercise capacity in T2DM and impaired tissue oxygenation in sepsis. These predictions are consistent with literature reports in both disease states (47; 131-133). Assuming a causal role for reduced perfused capillary density


50
and microvascular perfusion heterogeneity in limiting diffusion capacity, one would predict that exercise capacity and insulin sensitivity in T2DM are correlated, as well as tissue oxygenation and insulin sensitivity in sepsis. Exercise capacity and insulin sensitivity in T2DM are indeed correlated (5). The possibility of a correlation between insulin resistance and tissue hypoxia in sepsis has not been directly tested, but insulin resistance, tissue hypoxia, and impaired capillary perfusion are all thought to contribute to mortality (134), suggesting that this correlation would be observed if were to be tested.
Mechanistically, our model predicts that the exercise impairment in T2DM results from a combination of impaired muscle oxygenation and impaired clearance of metabolic wastes such as lactate and CO2. Consistent with these hypotheses, muscle oxygenation during contractions is impaired in animal models of T2DM (97) and muscle pH decreases more robustly during exercise in human subjects with T2DM than in healthy controls (135). Similarly, our model predicts impaired clearance of metabolic wastes in sepsis. This prediction is consistent with decreased interstitial pH (both lactate and CO2 are acidic) and interstitial hypercapnia in sepsis (133; 136). Likewise, our model predicts that the mechanism of insulin resistance in T2DM would be impaired diffusion of insulin. Consistent with this hypothesis, the limiting stage in insulin action occurs at the extracellular (i.e. vascular-interstitial) levels (112), and diet-induced obesity (a precursor to T2DM) causes impaired insulin access to the interstitium (137). In sepsis, the influences of microvascular perfusion on solute diffusion are confounded by a robust increase in vascular permeability (138), which is not taken into account in our analysis.
In hyperglycemia- or enzymatic-induced glycocalyx degradation (Figure 12C), our model predicts both impaired exercise capacity and insulin resistance, again resulting from


51
diffusion limitations secondary to reduced perfused capillary density and heterogeneous perfusion. The interaction between glycocalyx degradation and exercise capacity has not been tested. As relates to insulin sensitivity, however, our model predictions are correct (87; 139). The mechanisms that have been previously investigated for hyperglycemia-induced insulin resistance often involve oxidative signaling processes rather than mass transport effects (140). Our model predicts that insulin resistance in these states is accounted for, at least in part, by impaired diffusion of glucose and insulin. These possibilities have not been tested.
6.3 Recommendations for scientific practice
We have observed and participated in a general trend in the medical literature that can be formulated as follows: 1) tissue-level flux of a molecule of interest is impaired in a disease of interest, 2) investigators suspect a vascular/microvascular contribution to this defect, 3) subsequent studies determine that bulk blood flow does not fully account for the defect, 4) subsequent studies determine that tissue-level production/demand does not fully account for the defect, and then finally, 5) subsequent studies demonstrate a contribution from microvascular perfusion. Many scholarly publications spanning many years of investigation are often required before research of putative microvascular therapies can even begin. In the interest of helping the reader to avoid this trap, we have created a flowchart (Figure 13 A) to guide investigations of this sort.
The rationale underlying our flowchart questions stem from basic qualitative attributes of our theory, and are reflected in the interpretations of flowchart results provided in Figure 13B. Microvascular perfusion-related transport defects of any etiology are predicted to have similar influences on the flux of molecules with similar equilibration


52
properties (see Figures 8-11). Thus, if a particular molecule is uniquely impacted by an observed flux defect, it is likely that the cause of the defect is specific to that molecule (e.g. tissue level demand, cell membrane transporters, etc.). If interstitial concentration of a molecule of interest equilibrates slowly with the bloodstream (e.g. insulin), any microvascular contribution to reduced flux of this molecule is likely to involve diffusion-limiting effects such as capillary dropout, capillary rarefaction, or perfusion heterogeneity. For rapidly-equilibrating molecules (e.g. oxygen), both diffusion- and convection-limiting effects may impact tissue-level flux. Thus if larger, more slowly equilibrating molecules are not also affected, reduced bulk blood flow (a convection-limiting effect) is a plausible culprit. Conversely, if tissue-level flux is impaired across a wide range of equilibration rates, diffusion-limiting perfusion defects such as perfusion heterogeneity and capillary dropout are plausible contributors.
Typical scientific practice in the past has been to bypass these considerations and assay bulk blood flow alone. Our theory predicts that this strategy will often generate confusion. Even for convection-limited or convection/diffusion co-limited transport processes such as tissue oxygenation, heterogeneous blood flow distribution and capillary dropout may recapitulate the effects of reduced blood flow.


53
Tissue-level flux of X is impaired, and a
vascular/microvascular cause is suspected.
Is the flux of molecules with similar equilibration properties also impaired?
B)
What to Look For Intent rotation
The observed defect is likely specific to X Reduced production, reduced consumption, reduced membrane transport etc., ofX Microvascular perfusion is probably NOT contributing to the defect
Assay blood flow. Reduced blood flow. If blood flow is reduced, thi3 change is a likely' cause for the defect If blood flow is NOT reduced, assay capillary' perfusion.
Assay capillary perfusion. Capillary dropout, perfusion heterogeneity, and reduced flow velocity. If perfusion is altered, this change is a likely cause for the defect. If perfusion is NOT altered, assay capillary perfusion AND density.
Assay capillary perfusion AND density. Capillary dropout, perfusion heterogeneity, reduced flow velocity' and or reduced capillary' density'. If perfusion is altered, this change is a likely cause for the defect. If capillary' density' is reduced, this change is a likely cause for the defect.


54
Figure 13: Flowchart of recommended steps to test for microvascular contributions to solute flux defects. (A) This flow chart is intended to help investigators determine an appropriate experimental design to assess the possibility that tissue-level flux of a molecule of interest (X) is impaired due to a vascular/microvascular defect. Direct observation of capillary perfusion is advisable in most cases. (B) Interpretation offlowchart destinations. In general, if a transport defect is solute-specific, it is probably not caused by a microcirculatory issue. Alterations in bloodflow, capillary density, and capillary bloodflow distribution are all probable causes of transport defects.
We will now provide a demonstration of these considerations using oxygen delivery to skeletal muscle in T2DM as an example. Like oxygen, blood glucose equilibrates near-fully with the interstitium during capillary transit (137). Glucose delivery to skeletal muscle during hyperinsulinemia is also impaired in T2DM (112), and the degree to which the capacity for glucose flux (i.e. insulin sensitivity) is reduced correlates to the degree to which the capacity for oxygen flux (i.e. VChmax) is reduced (5). These observations provide an answer of yes to the first flowchart question. Oxygen and glucose are both rapidly equilibrating molecules, thus providing an answer of yes to the second flowchart question. Delivery of insulin (a large, diffusion-limited molecule) to the muscle interstitium is also impaired in diet-induced obesity (a precursor to T2DM) (137), thus providing an answer of yes to the third flow chart question. The degree of skeletal hypoxia during exercise reported in T2DM is often severe (97), while the insulin transport defect is more subtle.
These observations provide an answer of lesser to the final flowchart question, yielding a recommendation of assay capillary perfusion True to form, studies of microvascular and capillary perfusion in multiple models of T2DM reveal substantial capillary dropout and perfusion heterogeneity (23-28), and we now know that reversal of these defects can acutely improve skeletal muscle oxygenation in at least one animal model (25; 60). With the benefit of hindsight and a theoretical understanding of microvascular solute flux, many years of trial and error could have been avoided using the approach recommended here.


55
6.4 Implications for therapy
We will now introduce interventions that have been shown to improve each of the microvascular perfusion parameters considered in our analysis. Acute reductions in perfused capillary density are often caused by adhesive interactions such as leukocyte adhesion or microembolism (115; 141). Certain anticoagulant drugs would therefore be expected to improve capillary perfusion, and this effect has indeed been observed (108). Microvascular perfusion heterogeneity is less well understood, but likely involves impaired vessel function within arteriolar networks. This defect could plausibly be mitigated by antioxidant treatment. Antioxidant treatment can also reduce adhesive interactions (thus potentially improving perfused capillary density) and arteriolar perfusion is mechanically coupled to capillary perfusion (i.e. capillary plugging causes perfusion heterogeneity), so antioxidants would be expected to acutely improve both perfusion heterogeneity and perfused capillary density.
Sure enough, acute antioxidant treatment does help to improve microvascular perfusion in certain contexts (108). Bulk blood flow, meanwhile, is controlled by vascular tone, and can be increased using vasodilators.
For diffusion-limited processes such as insulin action, our model predicts that the correct anticoagulants and antioxidants would be sufficient to acutely improve insulin sensitivity in insulin resistant individuals. As concerns antioxidants, this hypothesis has been repeatedly validated (142-144). The possibility that anticoagulants improve insulin sensitivity has not been directly tested, but its plausibility is supported by associations of a variety of hemostatic parameters with insulin resistance (119-121).
In sepsis, our simulations suggest that diffusion limitations caused by reduced perfused capillary density are a major cause of impaired substrate delivery and metabolite


56
clearance. More importantly, impaired capillary perfusion is thought to be a major cause of organ failure and death in sepsis (84; 107; 108; 145). Certain anticoagulants (e.g. activated protein C) and certain antioxidants (e.g. Vitamin E) markedly improve microvascular perfusion in sepsis (108). True to form, activated protein C substantially improves patient outcomes (146), as do certain antioxidants (147). These reports are consistent with our model predictions. Although the interactions of these therapies with microvascular solute flux has not been widely investigated, perfused capillary density and interstitial hypercapnia (a marker of impaired CO2 clearance) are inversely associated and interstitial hypercapnia can be acutely improved by capillary perfusion rescue (136).
For oxygen delivery in T2DM (see Figure 12A), our model predicts that a complex mixture of reduced bulk blood flow, reduced capillary density, and increased perfusion heterogeneity is required to explain the phenotype. Consequently, our model predicts that a drug combination consisting of anticoagulants, antioxidants, and vasodilators (or equivalent) would be necessary to reverse the oxygenation defect. In the OZR model of T2DM, these predictions have been tested in the perfused hindlimb precisely this combination of drugs is effective in acutely restoring normal oxygen flux (25).
Our model accurately predicts effective therapeutic targets for acute intervention in each case where model predictions have been explicitly tested, but except in the OZR model of T2DM, the proposed mechanisms have not yet been validated. Additionally, other compounds targeting microvascular perfusion may be more suitable for long-term use. It is tempting to speculate that the cardiovascular protective effects of aspirin (148), for example, relate to its anticoagulant and antioxidant properties, or that the apparent ability of metformin


57
to extend lifespan (149) relates to its microvascular perfusion benefits (123). Future studies will be required to investigate these possibilities.
6.5 Model limitations
Our model accurately predicts many physiological associations, predicts which physiological parameters and solute exchange processes will be impaired in a variety of disease states, and predicts the therapies that will normalize these phenotypes, yet there are key limitations. One limitation of our model is its simplicity. In reality, capillaries are heterogeneous (150), and we neglect this source of heterogeneity. Another consideration is that changes in endothelial or interstitial permeability were largely neglected. Our rationale for this simplification was twofold: 1) due to the selective permeability of the endothelium and interstitium, changes in permeability will be different for each molecule and cannot readily be generalized, and 2) capillary permeability is highly dynamic and does not lend itself to description by a single parameter. This limitation will thus require testing of specific molecules in different physiological settings to determine their individual properties. However, as evidenced by the fact that those model predictions which have previously been tested were correct, neglecting changes in capillary permeability does not compromise the general utility of the model. Based on the noted limitations, our model accurately reflects the contribution of microvascular perfusion changes to the disease phenotype, and that these contributions occur within the larger context of the local microenvironment.
6.6 Conclusions
The relationship between bulk blood flow and tissue-level solute flux is dissociated except in cases of convection-limited transport processes such as pulmonary oxygen uptake. Our simulations suggest that microvascular perfusion is adequately described by our model


58
to explain the observed dissociation. In instances where relevant experimental data exist, our model successfully predicts which solutes are most profoundly influenced by microvascular perfusion and which microvascular parameters are most strongly associated with solute flux. These results may account for literature reports in a variety of disease states and underscore the importance of a theoretical understanding of the microvascular parameters influencing microvascular exchange processes. Our model facilitates this theoretical understanding, and even allows some prediction of the therapies required to repair microvascular perfusion defects. Future work will be required both to further validate the predictive power of our model and to develop and test therapies targeting microvascular perfusion independently from bulk blood flow.


59
CHAPTER IV
IMPAIRED TISSUE OXYGENATION IN METABOLIC SYNDROME REQUIRES INCREASED MICRO VASCULAR PERFUSION HETEROGENEITY
Co-authors: Fan Wu, I. Mark Olfert, Christopher G. Ellis, Daniel Goldman, Jane E. B.
Reusch, and Jefferson C. Frisbee
1. Preface
As I developed a theoretical understanding of how heterogeneous blood flow distribution might impact skeletal muscle oxygenation in diabetes, I learned that the Jefferson Frisbee group had spent several years comprehensively characterizing perfusion heterogeneity and skeletal muscle oxygenation in diabetic rats. This chapter is documents the ensuing collaboration, which uses a more detailed version of the microvascular perfusion theory derived in the previous chapter to accurately predict the perfusion differences observed in diabetic rats. Not only do our simulations correctly predict increased perfusion heterogeneity in diabetic rats from observed limitations in oxygen diffusion, they also correctly predict partial recovery of blood flow distribution with anti-oxidant treatment, and the magnitude of the perfusion heterogeneity predicted is very close to that which is actually observed. This combined theoretical/empirical approach lends credibility to the possibility that heterogeneous distribution of microvascular blood flow contributes to impaired exercise capacity in diabetes. This chapter was published in the Journal of Cardiovascular Translational Research in 2017.
2. Abstract
Metabolic syndrome (MS) in obese Zucker rats (OZR) is associated with impaired skeletal muscle performance and blunted hyperemia. Studies suggest that reduced O2 diffusion capacity is required to explain compromised muscle performance, and that


60
heterogeneous microvascular perfusion distribution is critical. We modeled tissue oxygenation during muscle contraction in control and OZR skeletal muscle using physiologically-realistic relationships. Using a network model of Krogh cylinders with increasing perfusion asymmetry and increased plasma skimming, we predict increased perfusion heterogeneity and decreased muscle oxygenation in OZR, with partial recovery following therapy. Notably, increasing O2 delivery had less impact on VO2 than equivalent decreases in O2 delivery, providing a mechanism for previous empirical work associating perfusion heterogeneity and impaired O2 extraction. We demonstrate that increased skeletal muscle perfusion asymmetry is a defining characteristic of MS, and must be considered to effectively model and understand blood-tissue O2 exchange in this model of human disease.
3. Introduction
The growing incidence and prevalence of the metabolic syndrome and associated type 2 diabetes mellitus represent one of the greatest challenges to public health facing developed economies worldwide (151; 152). The metabolic syndrome is generally defined as the combined presentation of obesity, impaired glycemic control, atherogeneic dyslipidemia, and moderate hypertension, with the additional contributing conditions of pro-oxidant, -thrombotic and -inflammatory phenotypes (153-155). This clinical condition is present in an enormous and growing number of afflicted people worldwide, it has a powerful influence on reducing patient quality of life and life expectancy, and it causes enormous increases to the direct and indirect economic costs that must be borne by society as a result of its negative effects on health outcomes (151; 152). These considerations mandate detailed investigation into this multi-pathology state.


61
In the general population (96) and in specific elements of the metabolic syndrome such as type 2 diabetes mellitus (T2DM; Ref. (7)), aerobic exercise capacity as measured by maximal oxygen consumption (VChmax) is the strongest clinical predictor of mortality. Further, in both T2DM and metabolic syndrome, exercise capacity has been demonstrated to be reduced (156-158). While the precise mechanisms underlying impaired exercise capacity in these states remain unclear, there are distinctive mechanistic hints in the existing literature. In both human and animal studies of these states, hyperemic responses to exercising muscle are reduced with manifestation of the metabolic syndrome (20; 60), yet oxygen extraction can not only fail to increase to compensate for the ischemia, but can be reduced itself (29;
60). Given that mitochondrial capacity exceeds the oxygen delivering capacity of the vasculature (159), the coincidence of impaired blood flow and impaired oxygen extraction cannot be explained through reduced mitochondrial demand or function. Understanding these apparent phenotypic contradictions requires detailed mechanistic study of O2 transport and consumption comparing theoretical predictions to empirical data.
The obese Zucker rat (OZR; fa/fa) represents an excellent animal model of the metabolic syndrome with high utility in terms of comparing cardiovascular (dys)function to pathology in human subjects. OZR develop the metabolic syndrome due to chronic hyperphagia secondary to profound leptin resistance, and rapidly develop the systemic phenotypes listed above to comparable levels of severity with those identified in human subjects. Also similar to health outcomes in humans, OZR exhibiting the metabolic syndrome suffer from a progressive vasculopathy that ultimately develops into overt peripheral vascular disease (PVD; Ref. (160)), albeit one without the development of significant atherosclerotic lesions within macrovessels.


62
Recent experimental studies have revealed significant impairments to the fatigue resistance (i.e., the ability to maintain contractile performance over time) of in situ skeletal muscle of OZR as compared to that in lean Zucker rats (LZR) (60; 161). The extent of these impairments has been generally correlated with reductions in bulk blood flow and hyperemic responses to the elevated metabolic demand (60). However, the relationship between fatigue and reduced blood flow is not particularly robust and there is potential that factors beyond simple bulk perfusion could be responsible for the functional manifestation of PVD in metabolic syndrome.
We have previously published observations of reduced blood flow (Figure 14A), reduced oxygen extraction (Figure 14B), increased muscle fatigue associated with reduced diffusion capacity (Figure 14C), and increased perfusion heterogeneity (Figure 14E) in the OZR (25; 26; 60; 162). Perfusion heterogeneity was quantified using the parameter y (Figure 14D), defined as the fraction of total blood flow diverted to the higher-flow daughter vessel at a microvascular bifurcation. Pharmacological intervention (e.g. TEMPOL; l-Oxyl-2,2,6,6-tetramethyl-4-hydroxypiperidine, a cell permeable superoxide dismutase mimetic which acts as a powerful antioxidant) correcting this perfusion heterogeneity in the smallest arterioles (3a-5a) normalizes oxygen extraction, and intervention correcting both bulk blood flow and perfusion heterogeneity (using a combined treatment of TEMPOL, the ai/012 adrenoreceptor antagonist phentolamine and the TXA2/PGH2 receptor blocker SQ-29548) fully restores oxygen uptake in the OZR to LZR levels (25; 26; 60). Therefore, it is of interest to be able to quantitatively test the hypothesis that oxygen supply from blood to skeletal muscle is a major limiting factor for muscle contractile performance under conditions of elevated metabolic demand within OZR, and, if observed, to define factors limiting oxygen delivery.


63
D)
Parrot Arteriole
(l-r)
flEaigigg
Jt-
Arteriolar Diameter
Mm>
RBC Velocity (biid/i)
Daughter Arterioles
E)
1--1 LZR
oza EBa ozk T
Figure 14: Previous studies establish impaired oxygen diffusion and microvascular perfusion heterogeneity in the OZR. (A) Femoral artery bloodflow is significantly reduced in the OZR during 5 Hz contractions. This defect is not affected by treatment with TEMPOF (B) Oxygen extraction is significantly reduced in the OZR during 5 Hz contractions. This defect is reversed by treatment with TEMPOL. (C) Oxygen diffusion capacity, a determinant of oxygen extraction, is a determinant of muscle fatigue for both LZR and OZR across multiple pharmacological treatments. (D) Schematic illustrating the parameter y, used to quantify microvascular perfusion heterogeneity. (E) Microvascular perfusion heterogeneity in 4a and 5a arterials is significantly increased in the OZR. This defect is reversed by treatment with TEMPOL. Panels A-C recreated with permissions from Frisbee et al Exp Physiol 2011 (60). Panels D-E recreated with permissions from Frisbee et al Am J Physiol 2011 (25).
In this work, a computational model of oxygen transport within skeletal muscle was constructed based on the classical Krogh cylinder-type model (163), which has been further extended to account for capillary-wall transport barrier, myoglobin-facilitated diffusion, and Michaelis-Menten kinetics of oxygen consumption (164). The constructed computational model was then applied to analyze oxygenation levels and metabolic rates using experimental data on bulk tissue blood flow and blood oxygen tensions measured from in situ blood-


64
perfused contracting skeletal muscle of LZR and OZR and reported in a previous publication (60). The predictions of this model were compared to previously published empirical observations to determine the effects of perfusion heterogeneity in an idealized microvascular network taking into account phase separation (i.e. erythrocytes from plasma) effects at bifurcations (165).
4. Materials and Methods
4.1 Microvascular Network Model
We have previously reported increased microvascular blood flow heterogeneity in OZR compared with that in LZR (25; 26; 162). To test the hypothesis that the associated mismatch between oxygen supply and working capacity (60) is caused by the flow heterogeneity, an idealized microvascular network (shown in Figure 15 A) was applied to examine impacts of flow heterogeneity on oxygen transport on the microvascular level. Empirical data from a previous publication (60) was used for this analysis.
The idealized microvascular network employed in this analysis was based on Murrays cube law (166) and assumed a fully symmetric vascular tree (i.e. all vessels of the same generation are the same length and diameter). The resulting vascular tree is consistent with the L-system fractal formalism of Zamir (167), has recently been applied to assess the impact of the endothelial glycocalyx on microvascular blood flow distribution (89), and has been extensively validated as a reasonable approximation of microvascular branching architecture in vivo (168). The idealized microvascular network included in this analysis branched from a 192 pm parent vessel to 6 pm diameter capillaries over the course of 15 vessel generations.


65
At each microvascular bifurcation, the uneven distribution of blood flow into the daughter branches results in heterogeneity of downstream discharge hematocrits as described by the following equation (165):
Ye
1 + exp (0.018 +1.22 logit $.5 -1.042 (0.5 yB)))
(28)
where logit function is defined as
logit (x) = In
x
\-x
(29)
The ratios of bulk blood volume flow and erythrocyte volume flow are defined as:
Yb
Fb
fb
(30)
Ye
Yb,i ~Hctl
YBHct
(31)
respectively, with subscript 1 denoting daughter branch 1.
During the simulations, the ratio of bulk blood volume flow (yB) was varied from 0.5 to 0.7 (using notation consistent with our previously published perfusion heterogeneity analyses), and the discharge hematocrits in the daughter branches were calculated using Equations 29-31 for all vessel generations. In this work, arteriolar oxygen delivery is neglected, thus inlet PO2 to each capillary is assumed to be equal to arterial PO2, a reasonable assumption a reasonable assumption under conditions of high flow and high oxygen demand. Then the oxygen transport was calculated for each capillary using look-up table of Ctv values simulated from a range of representative discharge Hct and F values for LZR or OZR. Mitochondrial demand (Vmax) in all cases was adjusted to recreate whole-muscle VO2 for


66
LZR at each level of stimulation without perfusion heterogeneity, consistent with observations of minimal perfusion heterogeneity in the LZR (25; 26; 162).
4.2 Krogh Cylinder-Type Model
In the Krogh cylinder-type model, the tissue is assumed to consist of uniformly spaced cylinders, and each tissue cylinder is supplied by a capillary with perfusing blood flow within its volume. The oxygen carried by blood supports local metabolic demand through diffusion across the capillary wall and into the skeletal muscle cell. The geometry of the Krogh cylinder-type model is shown in Figure 15B.
Figure 15: Visualizations of simulations performed (A) A schematic illustration of the idealized microvascular network used in the present study. Note that 4 vessel generations are visualized, whereas 15 vessel generations were used in the simulation. Please see text for details. (B) A schematic illustration of the Krogh cylinder-type model usedfor simulation of single-capillary oxygen transport in this study.
4.3 Governing Equations
The Krogh cylinder-type model is described by two ordinary differential equations for radial and axial oxygen transport, respectively (164; 169).


67
Oxygen Movement in the Radial Direction: Based on Ficks law of diffusion and mass transport, the radial oxygen transport is described by:
r d
dC
V dr J
= VOn
(32)
where D denotes the molecular diffusivity of oxygen, C* denotes the modified tissue oxygen concentration accounting for myoglobin-facilitated diffusion, and VO2 denotes oxygen consumption rate in the tissue.
The modified tissue oxygen concentration C* is defined as:
C*=C
D
Mb .fi V V
^Mb y m 0 Mb
D
(33)
where DMb denotes the molecular diffusivity of myoglobin, CMb denotes the tissue myoglobin concentration, Vm denotes the molecular volume of oxygen, and Stub denotes the myoglobin saturation level.
The myoglobin saturation level is calculated from:
$Mb = r r (34)
^ + ^50 ,Mb
where Cso.Mb denotes half-maximal myoglobin saturation O2 concentration.
The oxygen consumption rate (M) is assumed to follow Michaelis-Menten kinetics:
M(r,z) = (35)
L+Km
where Vmax denotes the maximal mitochondrial oxygen consumption rate and Km denotes the Michaelis-Menten constant of oxygen consumption in the skeletal muscle.
Assuming that no oxygen exchange occurs across the outer boundary of the tissue cylinder (i.e. the outer boundary of the cylinder represents a local minimum of PO2), a no-flux boundary condition is obtained:


68
dC
dr
= 0
(36)
r=R2
At the interface between capillary and tissue, an oxygen permeability barrier is assumed:
D
dC
dr
= P,
\c,-cf'-L)
(37)
where D denotes the molecular diffusivity of oxygen in tissue, pw denotes cross-capillary-wall permeability of oxygen, and R2 denotes the outer radius of the tissue cylinder.
Oxygen Movement in the Axial Direction: Since the axial oxygen gradient is much less than the radial oxygen gradient, the axial oxygen transport is assumed to be dominated by convection and is described as follows:
dCT(z)_ V02avg(z)
dz
F Ptissue L
(38)
where Ct denotes the total oxygen concentration in blood, V02avg denotes the oxygen consumption rate averaged along the radial direction, F denotes the flow volumetric rate normalized to tissue mass, ptissue denotes the tissue mass density, L denotes the capillary length.
The average oxygen consumption rate (VChavg) is defined as:
J*r=R 2 f>r=R 2
2m V02 (r, z)-dr r- V02 (r, z) dr
r=Rl Jr=Rl
V02avg(z) =
Jr=R2
27ir-dr
r=Rl
{Rl-Rl)!2
(39)
The total oxygen concentration (Ct) is defined as:
CT =Cf-{\-Hct)+CRBC Hct+Hct-Cm-Sj
Hb
(40)


69
where Cf denotes free oxygen concentration in plasma, Crbc denotes free oxygen concentration in red blood cells (RBCs), Hct denotes discharge hematocrit, CHb denotes concentrations of oxygen binding sites in RBCs, and SHb denotes hemoglobin-oxygen saturation level.
The hemoglobin saturation level is calculated from:
SHb =

P H _|_ P
r02 r ^ 50, Hb
(41)
where Pso.iib denotes half-maximal hemoglobin saturation O2 tension and //// denotes the Hill coefficient for hemoglobin-oxygen binding.
The free oxygen concentration in plasma and red blood cells (RBCs) can be related to oxygen tension (PO2) as:
r =PO -a
1KJ1 u plasma
(42)
Crbc ~ PC 2 aRBC
(43)
where Otpiasma and 0.m- are oxygen solubility in plasma and RBCs, respectively.
4.4 Look-Up Tables of Venous and Muscle Oxygenation
The Krogh cylinder-type single-capillary model discussed in the previous section was applied recursively at each level of muscle stimulation (changes mitochondrial demand and blood flow parameters), for each animal model (LZR vs OZR, changes capillary density and blood flow parameters), and for blood flow values ranging from zero to 200x average. These values were then compiled into look-up tables (LUTs) of venous and muscle oxygenation (see Figures 16 and 17). These LUTs were combined with the network perfusion model discussed above to rapidly obtain an array of single-capillary values for venous oxygen


70
content and mean muscle oxygenation. The oxygen concentration of the mixed venous effluent was then calculated as:
CTv
Y.Qi*ct
SQi
(44)
Finally, mean muscle Pm02 was calculated from the mean of all single-capillary Pm02 values.
hterrutacrt Hmatacrit
0.2 0,4 0 0.8 02 0.4 OS 0.8
Hwnatocnt Hwnatocnt
i
E
c
::
0
U
c.
I
U
o
V*
3
0
c
s
Figure 16: Heatmaps visualizing look-up tables (LUTs) of venous oxygenation as a function of capillary hematocrit and relative bloodflow. (A) LUT of simulated venous oxygenation in the LZR at rest. (B) LUT of simulated venous oxygenation in the LZR during 5 Hz contractions. Note that venous oxygenation is substantially decreased (i.e. oxygen extraction is increased) relative to rest given similar flow and hematocrit. (C) LUT of simulated venous oxygenation in the OZR at rest. (D) LUT of simulated venous oxygenation in the OZR during 5 Hz contractions. Note that venous oxygenation is substantially less than at rest and marginally less than in the OZR given similar flow and hematocrit.


71
Figure 17: Heatmaps visualizing look-up tables (LUTs) of skeletal muscle oxygenation as a function of capillary hematocrit and relative bloodflow. (A) LUT of simulated muscle oxygenation in the LZR at rest. (B) LUT of simulated muscle oxygenation in the LZR during 5 Hz contractions. Note that muscle oxygenation is substantially decreased relative to rest given similar flow and hematocrit. (C) LUT of simulated muscle oxygenation in the OZR at rest. (D) LUT of simulated muscle oxygenation in the OZR during 5 Hz contractions. Note that muscle oxygenation is substantially less than at rest and marginally less than in the OZR given similar flow and hematocrit.
4.5 Sensitivity Analysis Methods
Microvascular perfusion heterogeneity results from a complex combination of microfluidic mechanical changes (e.g. glycocalyx degradation, capillary plugging by adherent leukocytes) and changes in vessel tone/diameter. Vessel diameter influences distribution of RBCs independently of flow distribution (165), and it is plausible that these effects, which were not considered in our muscle oxygenation simulations, are significant in vivo. To assess this possibility, we performed a sensitivity analysis on the simulated effects


72
of diameter randomization independent of perfusion heterogeneity. The maximum degree of flow asymmetry considered in this manuscript (y=0.7) can be produced by a ~9% difference in downstream vessel diameters, assuming laminar flow. Accordingly, we simulated CvC>2, Pm02, and VO2 with up to 10% randomization of vessel diameter in each vessel segment. 100 simulations at each degree of diameter randomization were performed, and the coefficient of variance (CV = SD/mean) of simulation results was used as a metric of the effect size of diameter randomization.
4.6 Model Parameters
The parameters used in this model are taken from either the present experimental results or from validated literature values and are summarized in Table 1.


73
Table 1: Summary of model parameters used in the present study.
Parameter Description Blood parameters Value Reference
Pm Oxygen tension in inlet capillary flow 91 to 95 mmHg Experimental data
Pout Oxygen tension in outlet capillary flow - Model outputs
Flow (F) Blood flow per tissue mass 7.1e-3 to 3.55e-2 mL sec"1 (mL tissue)'1 Experimental data
Hct Hematocrit 0.45 Default value
nH Hill coefficient for Hb-oxygen binding 2.7 (McGuire & Secomb 2004)
P 02,Hb Half-maximal Hb saturation 02 tension 38.8 mmHg (Kiwull-Schone et al. 1987)
SHb Hemoglobin saturation level - -
plasma Oxygen solubility in plasma 1.3e-6 M mmHg'1 (Beard et al. 2003)
Frbc Oxygen solubility in red blood cells 1.53e-6 M mmHg"1 (Beard et al. 2003)
pw Permeability for O2 across the capillary wall Tissue parameters 250 1 m sec'1 (Beard et al. 2003)
D Oxygen diffusivity in tissue 2410 1 lm2/sec (Goldman & Popel 1999)
DMb Myoglobin diffusivity in tissue 17.3 m2/sec (Goldman & Popel 1999)
CsO.Mb Half-maximal Mb saturation O2 concentration 9.22e-6 M (Goldman & Popel 1999)
tissue Oxygen solubility in muscle tissue 1.74e-6 M mmHg'1 (Beard et al. 2003)
Km Michaelis-Menten constant for O2 consumption 1.74e-6 M (McGuire & Secomb 2004)
V max Mitochondrial oxygen demand (rest) 0.039 ml/g/min Model optimization
Mitochondrial oxygen demand (5Hz) Geometric parameters 0.185 ml/g/min Model optimization
Ri Capillary radius 2.98 Hm (LZR) 3.07 Im(OZR) (Skalaketal. 1986)
r2 Tissue cylinder outer radius 19.9 m (LZR) 23.0 m(OZR) Experimental data
L Capillary length 1012 1 Im (Honig et al. 1977)
The measured blood flow (F) and arterial oxygen tensions (PaCh) were used as inputs to the computational model to simulate venous oxygen tensions (PvCh). The model


74
parameters were set up using values listed in Table 1, except that the oxygen consumption demand (VO2) at rest and at 5Hz were generated via model optimization by minimized differences between model-predicted and experimentally measured Pv02.
The Krogh cylinder-type model was implemented on a standard desktop personal computer using MATLAB (Version R2013b, MathWorks, Natick, MA). The optimization was performed using the fmincon function included in MATLAB Optimization Toolbox.
5. Results
5.1 Effects of perfusion heterogeneity on oxygen transport in microvascular networks
An idealized microvascular network (Figure 15 A) was used to examine the effects of microvascular flow heterogeneity on oxygen transport. The simulations were conducted by assuming that the mitochondrial demand (Vmax) in LZR and OZR are equal, and sufficient to produce measured values of VO2 with homogeneous perfusion in the LZR. Empirical data from a previous publication (60) was used for this analysis. The results of this simulation are shown in Figure 18. Model simulations of Pv02 using varying flow distributions reveal that simulated Pv02 matches measured PvOi at values of y = 0.6164 for untreated OZR, y = 0.5562 for TEMPOL-treated OZR, and at 7 = 0.5121 for LZR at 5 HZ (Figure 18A), indicating heterogeneous perfusion in the OZR that can be partially corrected by TEMPOL treatment. Using the model-predicted values of y = 0.5121, 0.5562 and 0.6164 for LZR, TEMPOL-treated OZR, and untreated OZR, respectively, our model predicts a mean muscle Pm02 of 30.4 mmHg in the LZR, 18.9 mmHg in untreated OZR, and 23.48 mmHg in TEMPOL-treated OZR (Figure 18B) at 5 Hz, indicating muscle hypoxia during exercise in the OZR that can be partially corrected by TEMPOL treatment. Our simulations also predict that the contributions of perfusion heterogeneity account for more than half of the total


75
decrease in VCEin the OZR relative to LZR (Figure 18C), a majority of which is due to perfusion heterogeneity in the small 3a-5a arterioles where TEMPOL most profoundly influences perfusion heterogeneity (25).
A)
Venous Oxygenation
5 13
r
I
g ID
y
i
C)"
3
0.12
}0.11 01
& 0 03
1 ora £
1 0 07
B DM
QOS
o.t
DIR StfixiUtad
- Lift CvO?
- DIR CvO-OZR+TEMPQi &*Q,
066 0.6 0.66 0.7
Micro^Kcular Ptrfti&cn H4!ognafty >vi
Oxygen Uptake
B)
02 30
| 20 26 tu
1 27
Si 20
s9 2 5 I
14
Muscle Oxygenation
X: 0.5421
Y: 30.4
---LZR 3mul**d
OZR
---LZR r
---QZRt
---OZR+TEMPOL-,
X 0 6582
Y:
X 0 6164
V: 10.0
S 0 65 0.6 066 0.7
kfeiCHtteJ* FsrfMM i
GSIWMOn a HwM (w end (fif'iyyjwo.
IA-2A
MrW CGfWMiCfl (A mcmvadv

LZR 02 ft SmJabad LZR VO, 02ft VO. QZft+TEMPQL VO, QZft wfa h*l VO.
NOTE: Horizontal and vertical lines represent single-variable measures (e.g. VO: and y were not simultaneously measured). They do NOT represent independence of axes.
0 66 0 6 0.66 MfcrovKoJar Prfuvc Figure 18: Differences in venous oxygenation, muscle oxygenation, and perfusion heterogeneity can be predicted using a simulation of microvascidar blood flow and oxygen transport. (A) Venous oxygenation increases with increasing perfusion heterogeneity in both the LZR and OZR. The intersection between simulated venous oxygenation (light, dashed lines) and empirically measured venous oxygenation (bold lines) can be used to predict the degree of perfusion heterogeneity. Our model correctly predicts greater perfusion heterogeneity in the OZR and partial correction ofperfusion heterogeneity with TEMPOL treatment. (B) The intersection between simulated muscle oxygenation (light, dashed lines) and model-predicted perfusion heterogeneity (bold lines) can be used to predict muscle oxygenation. Our model correctly predicts reduced muscle oxygenation in the OZR and also predicts partial correction of muscle oxygenation with TEMPOL treatment. (C) The intersections of simulated oxygen uptake (light, dashed lines) with empirically measured oxygen uptake (bold lines) and with model-predicted oxygen uptake without perfusion heterogeneity in the OZR (light, dotted line) can be used to predict the relative contributions of various oxygen transport parameters to the observed oxygen uptake defect. Our model predicts that perfusion heterogeneity in small (3a-5a) arterioles plays a major role in reduced oxygen uptake in the OZR.


76
5.2 Results from Sensitivity Analysis
The results of our sensitivity analysis are shown in Figure 19. Results were largely consistent among the results for CvC>2 (Panel A), PmCh (Panel B) and VO2 (Panel C). In all cases, the CV of the oxygenation metric of interest increases quadratically with increasing diameter randomization. The influence of diameter randomization independent of perfusion heterogeneity is several orders of magnitude smaller than that of perfusion heterogeneity independent of diameter randomization. These results suggest than our muscle oxygenation analysis with varying degrees of perfusion heterogeneity was negligibly influenced by our choice to neglect possible effects from vessel tone independent of blood flow and its distribution.


77
A)
25
Diameter Sensitivity of CvO? Results
x id
B)
2.5
Diafneter Sensitivity of PmO? Results
x 10
1 5 1DSin*i Ui ttuUZR-OZR
1.7' 10 imidM ran 2 pr#d tl*d LRZ-OZR drffritnt*

0 2 4 6 a 10
% Diameter RandormzHQan
0 2 4 6 B 10
% Dtamater Randonuzmon
q Diameter Sensitivity of VO, Results
x 10 *
1.2
1-
UQ'tO* b KtuU LZft-DZR difference
SMl
g O.B
0.4 0 2-

2 4 o a io
T4 DiifTwttr Rwidonuzmon
Figure 19: Sensitivity analysis reveals that the effects of varying vessel diameters are negligible relative to the effects of varying flow distributions. (A) Variance in simulated Cv02 increases quadratically with increasing variance of vessel diameters. At the level of diameter variance (10%) required to explain the maximum degree of perfusion heterogeneity considered in this manuscript (y=0.7), the effects of diameter variance without flow heterogeneity are still several orders of magnitude less than the effects offlow heterogeneity without diameter variance. (B) Variance in simulatedPm02 increases quadratically with increasing variance of vessel diameters. This effect is several orders of magnitude smaller than the effects of flow heterogeneity. (C) Variance in simulated V02 increases quadratically with increasing variance of vessel diameters. This effect is several orders of magnitude smaller than the effects of flow heterogeneity.
5.3 Mechanisms underlying the effects of perfusion heterogeneity
To determine the mechanism(s) by which microvascular perfusion heterogeneity would theoretically reduce oxygen diffusion capacity, we compared the effects of equivalent increases or decreases in flow or hematocrit on oxygen uptake in single capillaries (Figure 20), using LZR 5Hz values (60) in a single-cylinder version of our model for illustration


78
purposes. Increasing capillary blood flow increases oxygen uptake, but not to the same degree that an equivalent decrease in capillary blood flow decreases oxygen uptake. The mean oxygen uptake of two capillaries with a fixed total blood flow supply thus decreases with increasing flow disparity between the two capillaries (Figure 20A). Similarly, increasing capillary hematocrit increases oxygen uptake, but not to the same degree that an equivalent decrease in capillary hematocrit decreases oxygen uptake. The mean oxygen uptake of two capillaries with a fixed total RBC flux thus decreases with increasing hematocrit disparity between the two capillaries (Figure 20B). Due to the plasma-skimming effect at microvascular bifurcations (165), capillary flow and discharge hematocrit are positively correlated (see Figure 20C, uses simulated y = 0.7 distribution for illustration purposes). The effects of flow and hematocrit disparities thus synergize to produce the cumulative effects of perfusion heterogeneity.


79
c)
s
I
X
i
&
O.fl 0,7 O.fl 0.5 0.4
03 0.2 0.1
0 4 .j i i
10 10 10 10 10 Fraction of Mean Capilary Fkwr
Flow vs Hematocrit
i
Figure 20: Analysis of perfusion heterogeneity effects at a single bifurcation reveals the mechanisms by which perfusion heterogeneity interferes with oxygen transport. (A) Increasing flow through one capillary while subtracting the same amount offlow from another capillary results in a greater decrease in oxygen uptake by the under-perfused capillary than the corresponding increase in oxygen uptake by the over-perfused capillary. This effect occurs independently of hematocrit effects. (B) Increasing hematocrit in one capillary while subtracting the same number of RBCs from another capillary results in a greater decrease in oxygen uptake by the low-hematocrit capillary than the corresponding increase in oxygen uptake by the high-hematocrit capillary. This effect occurs independently offlow effects. (C) Capillary flow and hematocrit are correlated at the single-capillary level, andflow heterogeneity results in hematocrit heterogeneity. These consequences of the plasma-skimming effect cause the effects of flow heterogeneity and hematocrit heterogeneity to synergize in reducing oxygen uptake under conditions of mi cr ovascular perfusion heterogeneity.
6. Discussion
It has been clearly demonstrated by our group (60; 161) and by others (170) that the development of the metabolic syndrome is associated with alterations to skeletal muscle arterial and arteriolar function. These changes are correlated with impaired hyperemic


80
responses to elevated metabolic demand, constrained oxygen uptake (VO2) across skeletal muscle and impaired fatigue resistance of in situ blood perfused skeletal muscle. However, beyond a restricted number of studies, there has been much less effort dedicated to determining how these key parameters relate to one another. The purpose of this study was to take the first steps into developing a computational model for microvascular perfusion under the conditions of the metabolic syndrome and to model, based on established literature values and original data from the OZR model of the metabolic syndrome, how microvascular function might be altered based on the underlying principles of mass transport and exchange. This is intended to be a first step in a larger effort to develop predictive models and biosimulations of sufficient veracity as to be useful for hypothesis development and informing experimental design. These models will be refined with the ultimate goal of application to clinical settings to inform a better understanding of conditions affecting microvascular function, cardiorespiratory fitness and response to clinical intervention.
A major initial interpretation of the results from the present simulations is that the striking similarity determined in the heatmaps (Figures 16 and 17) for oxygen transport and exchange between LZR and OZR is conceptually in contrast to all of the functional (i.e., muscle performance) data collected to date. If oxygen transport is minimally affected in between LZR and OZR using flow characteristics, then differences in perfusion and/or distribution are required to explain the discrepancy in outcomes. This highlights the importance of introducing perfusion heterogeneity into our simulations of microvascular perfusion in the OZR manifesting metabolic syndrome in order to more accurately understanding the system function in this challenged state.


81
Our present simulations show that perfusion heterogeneity is mathematically sufficient to account for our previously published observations of increased venous oxygenation (Figure 18A) in the OZR, and sufficient to explain why venous oxygenation is partially corrected by TEMPOL treatment (60). Perfusion heterogeneity is also mathematically sufficient to account for literature reports of muscle hypoxia (Figure 18B) in the OZR (103). Our model predicts that this muscle hypoxia is partially relieved by TEMPOL treatment. Although this prediction has not yet been directly tested, it does provide a plausible explanation for partial recovery of oxygen uptake with TEMPOL treatment (60). Finally, our model predicts that perfusion heterogeneity accounts for a majority of the oxygen uptake (Figure 18C) defect in during 5 Hz contractions the OZR, consistent with experimental recovery of a majority of the oxygen uptake defect during 5 Hz contractions when treated with a drug cocktail that fully reverses microvascular perfusion heterogeneity (25; 26; 60). The degree of perfusion heterogeneity required to mathematically account for oxygen transport differences between OZR and LZR in our simulation is not only qualitatively, but also quantitatively similar to experimentally determined values of y in LZR and OZR (25; 26). Finally, our simulations provide a mechanism by which perfusion heterogeneity results in impaired oxygen transport (Figure 20). Namely, an individual capillarys oxygen uptake is more sensitive to decreases in oxygen supply than to increases of similar magnitude. Similar to ventilation-perfusion mismatch in the lung, over-perfused vessels cannot fully compensate for under-perfused vessels.
Our simulations show that microvascular perfusion heterogeneity is mathematically sufficient to account for oxygen transport defects in the OZR not caused by reduced bulk blood flow. We have also previously published experiments showing that perfusion


82
heterogeneity is empirically necessary to account for these same defects (25; 26; 162). Given the theoretical sufficiency and empirical necessity of perfusion heterogeneity to account for the OZR phenotype, it is very likely that 1) our model describes tissue-level oxygen transport to a reasonable degree of accuracy, and 2) perfusion heterogeneity is a key pathological feature of the metabolic syndrome.
In contrast with this manuscript, the bulk of the scientific literature discussing exercise capacity in the metabolic syndrome and T2DM focuses on the total mount blood flow delivered to exercising muscle. Given the previously demonstrated impairments to vascular/arteriolar function in the skeletal muscle of OZR (161), the progressive rarefaction of the microvascular networks (171) and the reduction to bulk perfusion to the skeletal muscle across metabolic intensities (60), impaired oxygen transport is not particularly surprising. However, combining our present model results with our previous findings of heterogeneous microvascular blood flow distribution using both direct microvascular visualization (25; 26; 63) and tracer washout kinetics (162), an alternate interpretation of literature data suggests that increased muscle fatigue in OZR may also reflect an increasingly heterogeneous distribution of perfusion within microvascular networks. Microvascular perfusion heterogeneity contributes to impaired muscle function by causing reduced O2 extraction and an elevated PvOi. Our simulations and previously published experiments support the hypothesis that perfusion heterogeneity, along with the resulting phase separation of plasma from RBCs (165), results in failure to compensate for reduced blood flow in OZR with increased oxygen extraction.
Collectively, these echoes of previously published data in our simulations lend credibility to the notion that microvascular perfusion heterogeneity is necessary and


83
sufficient to account for experimentally observed barriers to oxygen transport in OZR. Although there are assumptions within our model that require further refinement (e.g. neglecting axial O2 diffusion, neglecting heterogeneous non-Kroghian capillary morphology, neglecting microscopic variances in blood O2 content, etc.), the broad consistency between our computational results, our previously published experimental results, and the broader published literature suggests that our simulations were sufficient to predict observed oxygen transport phenotypes from first principles. These findings suggest that clinical strategies aiming to improve bulk blood flow without also improving perfusion may have limited impact in the metabolic syndrome. Microvascular perfusion heterogeneity represents an unaddressed therapeutic target for improving tissue oxygenation and muscle function.


84
CHAPTER V
FULLY AUTOMATED SOFTWARE FOR COMPREHENSIVE QUANTIFICATION OF CAPILLARY STRUCTURE AND PERFUSION
Co-authors: Ian M. Williams, Sara E. Hull, David H. Wasserman, and Jane E. B. Reusch
1. Preface
In the obese Zucker rat model of T2DM, microvascular blood flow and its distribution have been adequately characterized to enable detailed engineering analysis quantitatively relating microvascular perfusion to tissue oxygenation. This perfusion characterization required many years and spanned several scientific publications. In this chapter, I show that the perfusion heterogeneity observed in the obese Zucker rat is also observed in high fat-fed mice (another animal model of insulin resistance), and allude to literature reports from nearly twenty years ago documenting very similar perfusion phenotypes is the GK rat (another T2DM model) and streptozotocin-treated rats (an animal model of type 1 diabetes). Furthermore, the technique developed in this study reduces the time required to achieve detailed characterization of microvascular perfusion heterogeneity from several weeks to several hours. This technique, combined with the microvascular perfusion theory described in Chapter III, holds promise for enabling detailed mechanistic analyses such as that described in Chapter IV for virtually any organ in virtually any disease state. Future experiments will be required to more completely validate this software technique for use with a variety of microscopy techniques in a variety of organs.
2. Abstract
Changes in microvascular perfusion have been shown to contribute to a variety of disease states, and recent advances in intravital microscopy hold promise for more detailed


85
study of microvascular perfusion. However, current methods for quantifying microvascular perfusion parameters are largely inconsistent and often provide incomplete characterization of microvascular perfusion phenotypes. In this manuscript, we report an automated (i.e. no user inputs required) software system for simultaneous measurement of plasma-perfused capillary density, RBC-perfused capillary density, % absence of RBC flux, mean capillary flow velocity, and heterogeneity of capillary blood flow distribution from intravital microscopy videos. This software was tested on videos of the gastrocnemius muscle microcirculation in high fat-fed (HFD, n=4, model of insulin resistance) and chow-fed (n=6, insulin-sensitive control) mice during hyperinsulinemia. Software measurements of plasma-perfused capillary density, % absence of RBC flux, and RBC-perfused capillary density were compared to by-hand measures of these same parameters. In each case, no significant difference was observed between software and by-hand measurements in either HFD or chow-fed mice (all p=NS). Software and by-hand measures both revealed significantly decreased plasma-perfused capillary density, significantly increased % absence of RBC flux, and significantly decreased RBC-perfused capillary density in HFD mice (all p<0.01). These results are consistent with previous literature reports of reduced capillary density and loss of RBC flux in T2DM (an insulin resistant state). In addition, this software revealed significantly increased heterogeneity of capillary flow velocity (as assessed by coefficient of variance) in HFD mice (p=<0.001), consistent with previous reports of increased microvascular perfusion heterogeneity in T2DM. Collectively, these results demonstrate that capillary perfusion can be quantified in an accurate, high-throughput, user-independent, and comprehensive fashion using software designed specifically for this purpose.


86
3. Introduction
Delivery of nutrients, oxygen, medications, or any other molecule of interest to peripheral tissues requires trans-capillary efflux. Capillary structure and perfusion are major determinants of this efflux. For example, variations in skeletal muscle capillary density have been shown to associate with both changes in glucose transport (i.e. insulin sensitivity) and changes in oxygen transport (i.e. aerobic exercise capacity) (113; 118; 122; 129). Reduced blood flow also has well-documented detrimental effects on end-organ function in a variety of contexts (172-174). Distribution of blood flow within microvascular networks is critical to maintaining normal trans-capillary exchange beyond bulk blood flow and capillary density. Heterogeneous spatial distribution of microvascular blood flow has been shown to contribute to impaired tissue oxygenation in the metabolic syndrome and sepsis (23-26; 107; 108; 117; 136; 175), and it is also thought to contribute to impaired tissue oxygenation in ischemic heart disease and cerebrovascular disease (64; 65). Collectively, perfused capillary density, bulk blood flow, and distribution of blood flow within capillary networks are each critical parameters in the physiology and pathophysiology of many disease states.
Despite hundreds of papers reporting microvascular perfusion measurements, there is a paucity of perfusion data suitable for engineering analysis estimating the contributions of microvascular perfusion to trans-capillary exchange of physiologically important molecules. To assess the extent of this issue in the microcirculation literature, we reviewed the first 50 primary research articles returned by PubMed searches for capillary perfusion and microvascular perfusion (search performed on 1-25-2017, results sorted by relevance). We recorded the total number of articles quantifying microvessel density (or an equivalent measure such as blood volume measured through MRI or contrast-enhanced ultrasound),


87
flow velocity and/or regional bulk blood flow, and spatial heterogeneity of perfusion. We also took note of whether these measurements were taken by hand (e.g. counting capillaries in a histological cross section) or using software (e.g. Doppler flow measurements). The results of this literature analysis are summarized in Table 2.
Table 2: Summary of microvascular perfusion measures in the primary literature.
Of 100 manuscripts reviewed, the number reporting... By-Hand Measurement Software Measurement
Microvessel Density 39 15
Flow Velocity and/or Bulk Blood Flow 19 29
Perfusion Heterogeneity 1 4

All of the Above 3
Two of the Above 28
One of the Above 38
None of the Above 31
In this sample of 100 primary research articles, a majority (69%) quantified no more than one of the three key perfusion parameters, and a sizable minority (31%) did not quantify microvascular perfusion. Examples of studies with no quantification of key perfusion parameters (31% of the total) include methods such as qualitative scoring of capillary perfusion or simulation rather than measurement of perfusion. Furthermore, the vast majority of studies reviewed (97%) did not report quantitation of all three perfusion parameters. This is a key limitation, as mass transport analysis reveals that all three parameters are needed to fully account for the contributions of microvascular perfusion to transcapillary solute flux (66). The most commonly neglected parameter is heterogeneity of microvascular perfusion,


88
which can be locally quantified by recording flow velocity at the capillary level (including a majority of the 29 manuscripts we found reporting software measurements of flow velocity). Flow heterogeneity has been demonstrated both theoretically (64-66) and empirically (23;
25; 26; 108; 117; 136; 175) to exert substantial independent influence on tissue oxygenation.
The limitations of previous microvascular perfusion data outlined above are primarily a function of available methods. Measurements taken by hand remain more common than software-assisted measurements. By-hand measurements of capillary perfusion are often very laborious, thus limiting sample size and measured endpoints. Aside from by-hand measurements, several techniques for software-assisted perfusion quantification have previously been published (176-178). The most impactful microvascular perfusion measurement software to date has been the Cap-Image software produced by Zeintl et al (176), which accounts for a majority of the manuscripts in our survey of the literature that used software for quantification of capillary perfusion. Cap-Image has been valuable in advancing the field. However, it does have the limitation that it requires user inputs to select vessels for measurement. As a result, use of this software can be very time consuming (177; 179), and the potential exists for human error or bias. There is a need for software that permits standardized quantification of microvascular perfusion that is less labor intensive and subject to investigator bias. Even with methods that have the potential to avoid these pitfalls, perfusion is rarely characterized comprehensively enough to make valid predictions about the relationships between microvascular perfusion and solute flux.
We therefore set out to create a fully automated software system for simultaneous quantification of perfused capillary density, capillary flow velocity, and microvascular perfusion heterogeneity. Our metrics for defining success were 1) software measurements


89
must agree with by-hand measurements when by-hand measurements are available, 2) software measurements must agree with literature results in an established phenotype, 3) software measurements must be taken without any user inputs (to reduce the possibility of biased results), and 4) software measurements must be comprehensive enough to plausibly enable future engineering analyses (e.g. (66)) relating microvascular perfusion parameters to transport of physiological molecules (e.g. tissue oxygenation). Methods and an example application for a software technique meeting all of these criteria will be described. Any readers who may be interested in using this technique are encouraged to contact us directly.
4. Methods
4.1 Animals
Following a 3 week weaning period, of either gender C57B16/J mice fed standard chow were group housed in micro-isolator cages. Mice were housed at 23C on a 0700-1900 light cycle, and allowed free access to H20 and food. Mice were fed a standard chow or high fat. The high fat diet contains 60, 24, and 16% of kcal/g as fat, carbohydrate, and protein respectively. Jugular catheterization (Ayala et al. Diabetes 2006) was performed at 12 -14 weeks of age. Experiments were carried out upon restoration of body weight (6-7 days postoperative). All procedures were approved by the Vanderbilt Institutional Animal Care and Use Committee.
4.2 Intravital microscopy
We used a Zeiss LSM 5 Live slit-confocal system for high resolution imaging of capillary perfusion in the gastrocnemius -100 pm beneath the tissue surface at the high speeds necessary to resolve blood flow dynamics. Mice were implanted with a jugular vein catheter prior to the procedure. The venous catheter was then used for infusion of insulin (4


Full Text
xml version 1.0 encoding ISO-8859-1
DISS_submission publishing_option 0 embargo_code third_party_search Y
DISS_authorship
DISS_author type primary
DISS_name
DISS_surname McClatchey
DISS_fname Penn
DISS_middle Mason
DISS_suffix
DISS_affiliation University of Colorado at Denver
DISS_contact current
DISS_contact_effdt 04/21/2017
DISS_address
DISS_addrline 12625 E Pacific Cir Unit D
DISS_city Aurora
DISS_st CO
DISS_pcode 80014
DISS_country US
DISS_email pmasonmcclatchey@gmail.com
future
04/21/2017
12625 E Pacific Cir Unit D
Aurora
CO
80014
US
pmasonmcclatchey@gmail.com
DISS_citizenship US
DISS_description page_count 170 doctoral external_id http:dissertations.umi.comucdenver:10843 apply_for_copyright no
DISS_title Heterogeneous Distribution of Microvascular Blood Flow Contributes to Impaired Skeletal Muscle Oxygenation in Diabetes
DISS_dates
DISS_comp_date 2017
DISS_accept_date 01/01/2017
DISS_degree Ph.D.
DISS_institution
DISS_inst_code 0765
DISS_inst_name University of Colorado Denver
DISS_inst_contact Bioengineering
DISS_processing_code D
DISS_advisor
Reusch
Jane
EB
DISS_cmte_member
Hunter
Kendall
S
Benninger
Richard
KP
Bauer
Timothy
A
Frisbee
Jefferson
C
Weiser-Evans
Mary
CM
Schmidt
Eric
P
DISS_categorization
DISS_category
DISS_cat_code 0786
DISS_cat_desc Biophysics
0719
Physiology
DISS_keyword exercise, image processing, inflammation, mass transport, microvascular perfusion, type 2 diabetes
DISS_language en
DISS_content
DISS_abstract
DISS_para People with type 2 diabetes mellitus (T2DM) suffer excess morbidity and mortality. The strongest clinical predictor of morbidity and mortality in the general population is reduced aerobic exercise capacity. T2DM causes impaired exercise capacity, and traditional explanations for this impairment involve reduced blood flow and/or reduced mitochondrial capacity. However, recent studies indicate that exercise capacity may be impaired independently from either tissue demand or bulk blood flow. The most likely explanation for this disconnect is heterogeneous distribution of microvascular blood flow. Under conditions of heterogeneous blood flow, some capillaries are over-perfused and thus saturate their capacity for oxygen delivery, while others are under-perfused and thus cannot support local tissue demand. Local measures reveal excess skeletal muscle deoxygenation during exercise in T2DM despite normal limb blood flow, and skeletal muscle deoxygenation is more heterogeneous in T2DM than in overweight controls. Sensitivity analyses building from established principles in mass transport reveal that heterogeneous blood flow alone is sufficient to cause both impaired skeletal muscle oxygenation and insulin resistance in T2DM. A more detailed version of this model was applied to oxygen transport in the obese Zucker rat (OZR, a common animal model of T2DM), and accurately predicted the degree of perfusion heterogeneity observed in the OZR. A novel software technique for quantifying capillary blood flow and its distribution reveals that high fat feeding (an experimental model of insulin resistance) causes heterogeneous capillary blood flow in mice. Finally, a combined analysis drawing from both first principles in microfluidics and empirical measurements of blood viscosity reveals that diabetes-induced degradation of the endothelial glycocalyx (a gel-like layer of macromolecules lining the interior surface of blood vessels) can account for heterogeneous distribution of microvascular blood flow. Collectively, these findings help to explain impaired oxygen extraction despite reduced blood flow in T2DM, and also offer a potential explanation as to why exercise capacity would predict morbidity and mortality: the proposed mechanism could plausibly apply to all tissues, not just to skeletal muscle
DISS_supp_abstract
DISS_binary PDF McClatchey_ucdenver_0765D_10843.pdf
DISS_restriction
DISS_repository
DISS_version 2011-11-08 15:37:33
DISS_agreement_decision_date 2017-04-21 13:12:27
DISS_acceptance 1
DISS_delayed_release
DISS_access_option



PAGE 1

HETEROGENEOUS DISTRIBUTION OF MICROVASCULAR BLOOD FLOW CONTRIBUTES TO IMPAIRED SKELETAL MUSCLE OXYGENATION IN DIABETES by PENN MASON MCCLATCHEY JR M.S., University of Colorado Denver, 2015 B.S., Georgia Institute of Technology, 2013 A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirements for the degree of Doctor of Philosophy Bioengineering Program 2017

PAGE 2

ii This thesis for the Doctor of Philosophy degree by Penn Mason McClatchey, Jr. has been approved for the Bioengineering Program by Kendall S. Hunter Chair Jane E. B. Reusch Advisor Richard K. P. Benninger Timothy A. Bauer Jefferson C. Frisbee Mary C. M. Weiser Evans Eric P. Schmidt Date: May 13 2017

PAGE 3

iii McClatchey Jr. Penn Mason ( Ph D. Bioengineering) Heterogeneous Distribution of Microvascular Blood Flow Contributes to Impaired Skeletal Muscle Oxygenation in Diabetes Thesis directed by Pr ofessor Jane E.B. Reusch ABSTRACT People with type 2 diabetes mellitus (T2DM) suffer excess morbidity and mortality The strongest clinical predictor of morbidity and mortality in the general population is reduced aerobic exercise capacity T2DM causes impaired e xercise capacity, and traditional explanations for this impairment involve reduced blood flow and/or reduced mitochondrial capacity However, recent studies indicate that exercise capacity may be impaired independently from either tissue demand or bulk blood flow. The most likely explanation for this disconnect is heterogeneous distribution of microvascular blood flow Under conditions of heterogeneous blood flow, some capillaries are ove r perfused and thus saturate the ir capacity for oxygen delivery, while others are under perfused and thus cannot support local tissue demand L ocal measures reveal excess skeletal muscle deoxygenation d uring exercise in T2DM despite normal limb blood flow, and skeletal muscle de oxygenation is more heterogen eous in T2DM than in overweight controls. S ensitivity analyses building from established princip les in mass transport reveal that heterogeneous blood flow alone is sufficient to cause both impaired skeletal muscle oxygenation and insulin resistance in T2DM A more detailed version of this model was applied to oxygen transport in the obese Zucker rat (OZR, a common animal model of T2DM) and accurately predicted the degree of perfusion heterogeneity observed in the OZR A novel software technique for quantifying capillary blood flow and its distribution reveals that high fat fe e d ing ( an experimental model

PAGE 4

iv of insulin resistance ) causes heterogeneous capil lary blood flow in mice Finally, a combined analysis drawing from both first principles in microfluidics and empirical measurements of blood viscosity reveals that diabetes induced degradation of the endothelial glycocalyx (a gel like layer of macromolecu les lining the interior surface of blood vessels) can account for heterogeneous distribution of microvascular blood flow Collectively, these findings help to explain impaired oxygen extraction despite reduced blood flow in T2DM and also offer a potential explanation as to why exercise capacity would p redict morbidity and mortality: the proposed mechanism could plausibly apply to all ti ssues, not just to skeletal muscle The form and content of this thesis are approved. I recommend its publication. A pprov ed: Jane E. B. Reusch

PAGE 5

v ACKNOWLEDGEMENT S I would like to acknowledge each of my thesis committee members for their contributions to this project and for the mentorship they provided me during my pursuit of this degree : Jane Reusch, Richard Benninger, Kendall Hunter, Timothy Bauer, Jefferson Frisbee, Mary Weiser Evans, and Eric Schmidt were each invaluable mentors in their own ways. I would also like to thank my collaborators and co authors outside of my thesis committee who helped to produce the resear c h included in this dissertation, including Judith Regensteiner, Irene Schauer, Amy Huebschmann Fan Wu, Mark Olfert, Ch ristopher Ellis, Daniel Goldman, Sara Hull, Ian Williams, David Wasserman, and Michal Schafer. In addition, I would like to thank the ma ny people not mentioned above who provided feedback and guidance in the course of this project, including Pete Watson, Robert Roach, Andrew Subudhi, Rebecca Scalzo, Leslie Knaub, and Amy Keller. Finally, I would like to thank the friends and family to help ed in me in innumerable ways particularly my mother, Anne McClatchey (who homeschooled me and taught me to rec ognize the inherent mathematical relationships in natural systems ) and my wife, Clio McClatchey (who proofread many of m y papers and helped me tr anslate my idea s into lay English ). This dissertation bears my name, but it required intensive efforts from dozens of others to come to fruition.

PAGE 6

vi TABLE OF CONTENTS C HAPTER I. EXERCISE, BLOOD FLOW, AND THE SKELETAL MUSCLE MICROCIRCULATION IN DIABETES MELLITUS -------------------------------------------------------------------------------------1 1. Preface ----------------------------------------------------------------------------------------------1 2. Abstract ----------------------------------------------------------------------------------------------1 3. Introduction -----------------------------------------------------------------------------------------2 4. Cardiac Output -------------------------------------------------------------------------------------3 5. Skeletal Muscle Blood Flow -----------------------------------------------------------------------4 6. Microvascular Perfusion Heterogeneity --------------------------------------------------------5 7. Considering Causality ----------------------------------------------------------------------------8 II. DISSOCIATION OF LOCAL AND GLOBAL SKELETAL MUSCLE OXYGEN TRANSPORT METRICS IN TYPE 2 DIABETES -----------------------------------------------------------------------10 1. Preface ----------------------------------------------------------------------------------------------10 2. Abstract -------------------------------------------------------------------------------------------10 3. Introduction ----------------------------------------------------------------------------------------11 4. Methods -------------------------------------------------------------------------------------------14 5. Results ----------------------------------------------------------------------------------------------18 6. Discussion -----------------------------------------------------------------------------------------23 III. A CONCEPTUAL FRAMEWORK FOR PREDICTING AND ADDRESSING THE CONSEQUENCES OF DISEASE RELATED MICROVASUCLAR DYSFUNCTION -----------28 1. Preface ----------------------------------------------------------------------------------------------28 2. Abstra ct --------------------------------------------------------------------------------------------28 3. Introduction ----------------------------------------------------------------------------------------29 4. Methods --------------------------------------------------------------------------------------------32 5. Results ----------------------------------------------------------------------------------------------35

PAGE 7

vii 6. Discussion ----------------------------------------------------------------------------------------48 IV. IMPAIRED TISSUE OXYGENATION IN METABOLIC SYNDROME REQUIRES INCREASED MICROVASCULAR PERFUSION HETEROGENEITY ---------------------------59 1. Preface --------------------------------------------------------------------------------------------59 2. Abstract -------------------------------------------------------------------------------------------59 3. Introduction ---------------------------------------------------------------------------------------60 4. M aterials and m ethods --------------------------------------------------------------------------64 5. Results ----------------------------------------------------------------------------------------------74 6. Discussion -------------------------------------------------------------------------------------------79 V. FULLY AUTOMATED SOFTWARE FOR COMPREHENSIVE QUANTITATION OF CAPILLARY STRUCTURE AND PERFUSION ---------------------------------------------------------84 1. Preface ---------------------------------------------------------------------------------------------84 2. Abstract -------------------------------------------------------------------------------------------84 3. Introduction --------------------------------------------------------------------------------------86 4. Methods -------------------------------------------------------------------------------------------89 5. Results ---------------------------------------------------------------------------------------------95 6. Discussion -----------------------------------------------------------------------------------------98 VI. THE ENDOTHELIAL GLYCOCALYX PROMOTES HOMOGENOUS BLOOD FLOW DISTRIBUTION WITHIN THE MICROVASCUL ATURE -------------------------------------------104 1. Preface --------------------------------------------------------------------------------------------104 2. Abstract -----------------------------------------------------------------------------------------104 3. Introduction -------------------------------------------------------------------------------------105 4. Methods ------------------------------------------------------------------------------------------109 5. Results --------------------------------------------------------------------------------------------114 6. Discussion ----------------------------------------------------------------------------------------119 VII. CONCLUSIONS AND FUTURE DIRECTION S -----------------------------------------------126

PAGE 8

viii 1. Abstract -------------------------------------------------------------------------------------------126 2. A perfusion centric paradigm for inflammatio n ---------------------------------------------127 3. Clinical and laboratory translation of microvascular perfus ion findings ------------------1 31 4. Potential microvascular perfusion therapies --------------------------------------------------137 5. Conclusions --------------------------------------------------------------------------------------139 REFERENCES ----------------------------------------------------------------------------------------------140

PAGE 9

ix LIST OF TABLES TABLE 1. Summary of model parameters used in the present study -------------------------------73 2. Summary of microvascular perfusion measures in the primary literature -------------87

PAGE 10

x LIST OF FIGURES FIGURE 1. Illustration of NIRS measures used for analysis ----------------------------------------1 6 2. Impaired muscle oxygenation despite normal blood flow and hemoglobin reserve in T2DM ----------------------------------------------------------------------------------------19 3. Hemoglobin reserve and muscle deoxygenation do not correlate to VO2peak in T2DM --------------------------------------------------------------------------------------------------21 4. Mutual correlations between blood flow, hemoglobin reserve, and VO2peak are abolished in T2DM -------------------------------------------------------------------------22 5. Solutes vary in their degree fractional equilibration with the interstitium during capillary transit -----------------------------------------------------------------------------34 6. Simulated distribution of blood flow i n an idealized microvascular network -------35 7. Effects of blood flow and diffusion capacity on microvascular solute flux ----------36 8. Relationship between baseline fractional equilibration and sensitivity to blood flow 38 9. Relationship between bas eline fractional equilibration and sensitivity to diffusion capacity --------------------------------------------------------------------------------------40 10. Relationship between baseline fractional equilibration and sensitivity to perfused capillary density ---------------------------------------------------------------------------42 11. Relationship between baseline fractional equilibration and sensitivity to arteriolar perfusion heterogeneity --------------------------------------------------------------------45 12. Complex phenotypes influence solute flux through a variety of mechanisms -------47 13. Flowchart of recommended steps to test for microvascular contributions to solute flux defects ---------------------------------------------------------------------------------------53 14. Previous studies establish impaired oxygen diffusion and microvascular perfusion heterogeneity in the OZR -----------------------------------------------------------------63 15. Visualizations of simulations performed -----------------------------------------------66 16. Heatmaps visualizing look up tables (LUTs) of venous oxygenation as a function of capillary hematocrit and relative blood flow -------------------------------------------70

PAGE 11

xi 17. Heatmaps visualizing look up tables (LUTs) of skeletal muscle oxygenation as a function of capillary hematocrit and relative blood flow -------------------------------71 18. Differences in venous oxygenation, muscle o xygenation, and perfusion heterogeneity can be predicted using a simulation of microvascular blood flow and oxygen transport -------------------------------------------------------------------------------------------------75 19. Sensitivity analysis revea ls that the effects of varying vessel diameters are negligible relative to the effects of varying flow distributions ------------------------------------77 20. Analysis of perfusion heterogeneity effects at a single bifurcation reveals the mechanisms by w hich perfusion heterogeneity interferes with oxygen transport ---79 21. Outline of image processing steps used to identify capillary segments for software flow measurement --------------------------------------------------------------------------------91 22. Outline of cross correlation routine used to measure RBC flow velocity -----------92 23. By hand measurement techniques ------------------------------------------------------95 24. Software and by hand measurements detect the same capillary perfusion and RBC flux differences between chow fed and HFD mice ------------------------------------------96 25. Additional measurements taken using software only -----------------------------------97 26. Visualization of HFD on capillary structure and perfusion ----------------------------98 27. Schema of established determinants of microvascular blood flow distribution based on the literature -------------------------------------------------------------------------------108 28. Parameters used to create idealiz ed arteriolar trees -----------------------------------11 2 29. Influences of microvascular blood viscosity on microvascular perfusion heterogeneity in a simulated arteriolar tree ------------------------------------------------------------115 30. Influe nces of microvascular blood viscosity on blood flow distribution at an idealized capillary bifurcation ----------------------------------------------------------------------117 31. Effects of glycocalyx properties on the determinants of blood viscosity -----------119 32. Summary of findings ---------------------------------------------------------------------121 33. A perfusion centric paradigm for inflammation ---------------------------------------128 34. Proof of concept for clinical application of flow tracking and oxygen delivery algorithms -----------------------------------------------------------------------------------136

PAGE 12

1 CHAPTER I EXERCISE, BLOOD FLOW, AND THE SKELETAL MUSCLE MICROC IRCULATION IN DIABETES MELLITUS Co authors: Timothy A. Bauer, Judith, G. Regensteiner, and Jane E.B. Reusch 1. Preface This chapter serves as a topic introduction and overview of the literature concerning exercise capacity and its r elationship to blood flow and the mic rocirculation in diabetes. Here I make the arguments for and against bulk blood flow as a limiting factor to exercise, both at the level of cardiac output and at the level of blood flow to the exercising muscle. Based on the current literature (as of early 2017), it is likely that both blood flow and its distribution play a role in reduced exercise capacity in diabetes. T his chapter has been accepted for publication as a chapte r in the upcoming book Diabetes and Exercise (Humana Press) and is currently in press. 2. Abstract Aerobic exercise capacity is impaired in both Type 1 Diabetes (T1DM) and Type 2 Diabetes (T2DM), and this impairment is predictive of future morbidity and mortality. Although the precise etiology of impaired exercise capacity in diabetes remains unclear, several distinct lines of evidence indicate that reduced delivery of oxygen by the cardiovascular system plays a causal role. Cardiac output is often but no t always reduced in diabetes. This change is sufficient but not necessary for reduced exercise capacity. Skeletal muscle blood flow is also often but not always reduced in diabetes. This change is also sufficient but not necessary for reduced exercise capa city. In addition, a growing number of animal and simulation studies show that heterogeneous distribution of blood flow within the

PAGE 13

2 microcirculation contributes to oxygen delivery limitations in diabetes. Once again, this change is sufficient but not necess ary for reduced exercise capacity. In this chapter, we discuss each of these changes in cardiovascular function and their likely causes, beginning with the heart and gradually progressing to capillary level. We then conclude our overview by interpreting th e causality or lack thereof of each diabetes related pathological change as it relates to reduced oxyge n delivery to skeletal muscle. 3. Introduction Any attempt to understand limitations in exercise function with diabetes would be incomplete without con sidering the influences of the cardiovascular system and blood flow regulation. Exercise capacity (VO 2max ) is impaired both in Type 1 Diabetes (T1DM) and Type 2 Diabetes (T2DM) (1 6) and this impairment is predictive of mortality (7 11) and cardiovascular complications (12 15) These relationships suggest that exercise capacity is a sensitive measure of early changes in cardiovascular function with diabetes. This notion is further supported by the associations of cardiac output and skeletal m uscle blood flow (SMBF) with VO 2max in healthy individuals (16 18) Limitations in both cardiac output and SM BF have been reported in diabetes (19 22) indicating that blood flow may be a component of exercise limitations in diabetes. In addition to reduction of total blood flow, increased heterogeneity of microvascular blood flow distribution (23 26) loss of capillary perfusion (27; 28) and reduced whole body oxygen extraction (29) have also been reported, indicating that h eterogeneous distribution of blood flow may also play a role in limiting aerobic capacity in diabetes. In this chapter, we will explore known changes to both blood flow and its distribution in diabetes, beginning with changes in cardiac function and progre ssing to the capillary level.

PAGE 14

3 4. Cardiac Output Both T1DM and T2DM are associated with left ventricular diastolic dysfunction (30; 31) Diastolic dysfunction in diabetes is associated with fibrotic remodeling of the myocardium (32) and e ventually leads to heart failure. As fibrotic remodeling progresses, the contractile ability of the heart is not necessarily impaired, sometimes but not always allowing for maintenance of systolic function (33) The result of these changes is that stroke volume (SV) is reduced in diabetes (19) and cardiac output is impaired under conditions in which the heart cannot compensate with increased heart rate (HR), such as in maximal exercise (19; 21) It is not entirely clear from the literature whether the sub clinical cardiac dysfunction observed in diabetes contributes to limitations in exercise capacity. Baldi et al found using an inert gas rebreathing technique that whole body arteriovenous oxygen difference during exercise is reduced in diabetes and that reductions in oxygen extraction associated to reduct ions in VO 2max (29) whereas Gusso et al found using a similar CO 2 rebreathing technique that arteriovenous oxygen difference was not altered by diabetes, but that reductions in CO and VO 2max were a ssociated (19) This juxtaposition of results suggests that differences in blood flow distribution (e.g. greater fractional perfusion of non muscle tissues in diabetes) or mitochondrial deman d are co determinants of exercise capacity in diabetes along with CO. Given the substantial similarities between cardiac and skeletal muscle, it is probable that impaired cardiac function is itself a manifestation of muscle functional limitations in diabet es rather than their root cause. Reduced cardiac output in diabetes is further compounded by the association between diabetes and hypertension (34; 35) In particular, large arteries become less responsive to vasomotor stimuli in diabetes, and this effect is particularly pronounced in assays of nitric

PAGE 15

4 oxide (NO) mediated endothelium dependent vasodilation resulting from pharmacological agonists (36; 37) and fluid shear stress (38; 39) Reduction in vasomotor (and especially vasodilatory) function tends to occur in conjunction wit h fibrotic remodeling of large arteries (40) and vessel stiffness itself may contribute to reduced dynamism in v essel tone. In people without diabetes, hypertension causes impairments in both cardiac output and VO 2max (41) The rate of hypertension is significantly elevated in both T1DM and T2DM relative to the general population (36; 37) which suggests that hypertension and the associated reductions in cardiac output may contribute to population level differences in exercise capacity with diabetes. It is worth noting, however, that limitations in VO 2max are observed in diabetes even in the absence of hypertension or any other overt cardiovascular disease state (2; 31) Although hypertension likely contributes to exercise limitations at the population level, h ypertension alone cannot fully account for the exercise li mitations observed in diabetes. 5. Skeletal Muscle Blood Flow Independently of cardiac and large vessel function, oxygen delivery to skeletal muscle could be impaired by inappropriate distribution of blood flow among organs. As previously discussed, the vasomotor dynamism of large blood vessels is reduced in diabetes (37 40) It is likely that this effect would interfere with redistribution of blood flow from inactive organs (e.g. mesenteric blood flow) to active skeletal muscle, but this hypothesis has not yet been directly tested. In addition to possible differences in the d ynamism of blood flow distribution, lean body mass as a fraction of total body mass is reduced in T2DM (42; 43) This effect would be expected to impair whole body ox ygen extraction by increasing blood flow to non oxidative tissues as a fraction of total blood flow (44) However, body composition is not necessarily altered in T1DM (45; 46) and impaired VO 2max is found not

PAGE 16

5 only when comparing of T2DM and lean, healthy individuals (1) but also in T1DM (3; 4) and even when comparing of T2DM and obese, sedentary individuals without diabetes (47) It is therefore likely that the contributions of macrovascular blood flow dist ribution to reductions in aerobic capacity can occur independently of changes in body composition. Regardless of whether these differences stem from cardiac dysfunction, vascular dysfunction, or something else entirely, SMBF measured at the whole limb leve l is often but not always reduced in diabetes (20; 22; 48 51) Even if SMBF were to reach normal steady state levels, the hyperemic response to exercise is often slowed in diabetes (48; 49; 52; 53) It is likely that slowed blood flow kinetics con tribute to the increased discomfort the onset of exercise reported in diabetes (53 55) Interestingly, there are some studies in which steady state SMBF is not reduced in diabetes and yet aerobic exercise capacity is still impaired (56 58) This juxtaposition of findings implies that organ level (as opposed to whole body) oxygen extraction is impaired in diabetes in addition to reductions in SMBF. True to form, human MRI studies by Zheng et al (59) and animal catheterization studies by Frisbee et al (60) show an impaired ability to increase skeletal muscle oxygen extraction fraction (SMOEF) following muscle contr action. This effect of diabetes does not appear to be unique to skeletal muscle, given that tissue level oxygen extraction is also reduced in diabetic retinopathy and neuropathy (61; 62) As is also true of impaired cardiac function and hypertension, it appears likely that reduced SMBF contributes to but is not necessary for diabetic exercise dysfunction. 6. Microvascular Perfusion Heterogeneity Impairment of oxygen delivery independently in T2DM of SMBF is likely caused by increased heterogeneity of microvascular perfusion. Frisbee et al have shown in the Obese Zucker Rat (OZR) model of T2DM that microvascular perfusion heterogeneit y is increased

PAGE 17

6 (25) that this perfusion heterogeneity contributes to peripheral vascular disease (23; 26) reversal of this perfusion heterogeneity with a cocktail of anti adrenergic and endothelium targeting drugs ac utely normalizes skeletal muscle function (24; 25; 60) and that these effects can be predicted from first principles in mass transport and anatomy (63) Importantly, microvascular perfusion heterogeneity in the OZR model of Frisbee et al caused impaired muscle oxygenation in part through heterogeneous red blood cell (RBC) distribution at the capillary level (24; 63) This result is further recapitulated by the intravital microscopy results of Poole et al in the Goto Kakizaki (GK) rat model of T2DM (28) and in the str eptozotocin treated model of T1DM (27) Not only are microvascular perfusion heterogeneity and a resulting impairment in oxygen availability observ ed in all these animal models of diabetes, it has also been shown that microvascular perfusion heterogeneity leads to impaired oxygen extraction independently of total blood flow on both theoretical (63 66) and empirical bases (26; 67; 68) The mechanism for this impairment (some capillaries are underperfused while others are overperfused and effectively saturate their capacity for oxygen delivery), is not tissue specific, consistent with observations of oxygen extraction limitations not on ly in skeletal muscle (59; 60) but also in other peripheral tissues (61; 62) In addition to increases in the heterogeneity of microvascu lar perfusion, reduced capillary density is also observed in diabetes (69; 70) further reducing oxygen availability independently of SMBF. Combined, the effects of microvascular perfusion heterogeneity and reduced capillary density can account for discrepancies between SMBF and aerobic capacity (63) Microvascular dysfunction and perfusion heterogeneity in diabetes may be caused by deg radation of the endothelial glycocalyx. The endothelial glycocalyx is a semi permeable,

PAGE 18

7 space filling layer of glycoproteins and glycosaminoglycans lining the luminal surface of the endothelium. Glycocalyx degradation has been reported in both T1DM and T2D M (71 73) and glycocalyx degradation is associated both in diabetes and in health with increased risk and early signs of future cardiovascular morbidities (74 78) as is exercise capacity (7 11) Degradation of the endothelial glycocalyx causes a similar redistribution of RBCs within the capillary network to that observed in diabetes whether glycocalyx degradation is achieved by enzymatic means (79; 80) or as a result of oxidative stress stemming from acute hyperglycemia or infusion of oxidized LDL (81; 82) Furthermore, physiologic glycocalyx degradation during sepsis or adenosine infusion has been shown to cause reductions in tissue oxygen extraction (83; 84) Although the connection between glycocalyx degradation and reduced aerobic capacity in diabetes has not yet been directly tested, it is noteworthy that glycocalyx degradation is involved in glomerular hyperfiltration (an early sign of diabetic nephropath y) (85; 86) glycocalyx degradation causes acute insulin resistance (87) and insulin resistance, glomerular hyperfiltration, and reduced aerobic capacity are all mutually correlated in diabetes (15; 88) Mass transport analysis reveals that insulin resistance and impaired exercise capacity can both be predicted from perfusion heterogeneity (66) and so it is likely that the perfusion effects of glycocalyx degradation contribute to these phenotypes. Simulation studies indicate that glycocalyx c harge density (as a determinant of permeability) modulates the heterogeneity/homogeneity of microvascular perfusion (89) providing a plausible mechanism for increased microvascular perfusion heterogeneity in diabetes. Ongoing studies within our group seek to clarify the relationship between the endothelial glycocalyx and diabetic exercise dysfunction.

PAGE 19

8 7. Considering Causality Because blood flow and its distribution are altered at every level of the circulation in diabetes from the heart to the smallest capillaries, it is useful to consider precisely which e ffects might play a causal role in reducing aerobic capacity. Studies in which CO, SMBF, or both are normal in diabetes and yet aerobic capacity is reduced show that changes in macrovascular parameters are not necessary for diabetic exercise dysfunction (2; 31; 56 58) Although similar macrovascular changes are sufficient to reduce aerobic capacity in the general population (16 18) the fact that they are not necessary for reduced aerobic capacity in diabetes indicates that the root cause of diabetic exercise dysfunction may lie else where. Antioxidant therapy normalizes many macrovascular parameters in diabetes (90; 91) and yet h as not been shown to normalize exercise capacity (and would be expected to interfere with exercise training (92; 93) ), whereas the microvascular dysfunction reported by Frisbee et al is acutely reversible and its reversal improves skeletal muscle function (25; 60) This combination of results indicates that microvascular dysfunction might play a causal role in diabetic exercise impairments. Given that the heart and the vessel walls of microvessels are themselve s heavily vascularized and therefore subject to microcirculatory influence, it is plausible that cardiac and macrovascular dysfunction in diabetes are themselves caused by microvascular dysfunction. In this overview of changes in blood flow and its distrib ution in diabetes, we sought to assess the possibility that exercise dysfunction in diabetes might be an early detector of impaired cardiovascular dysfunction. Blood flow is often (but not always) reduced at the whole body level (19; 21) at the whole limb level (20; 22; 48 51) and at the capillary level (27; 28) in diabetes. However, reduced aerobic capacity is sometimes observed even when

PAGE 20

9 blood flow is maintained (2; 31; 56 58) This apparent discrepancy may be explained by microvascular alterations including increased perfusion heterogeneity (24; 25; 60) and reduced capillary density (69; 70) Microvascular perfusion heterogeneity is itself a plausible contributor cardiac and macrovascular dysfunction due to its effects on tissue oxygenation (64; 65; 67; 68) and can be recapitulated by glycocalyx degradation (79 82 ) which also occurs in T1DM (72) in T2DM (71) and more generally in states of acute nutrient stress (75; 81; 82) Further studies will be required to clarify the relationships between glycocalyx degradation, blood flow, and microvascular perfusion. It is even po ssible that the etiology of impaired exercise capacity varies from individual to individual, but it is clear that oxygen delivery limitations resulting from impairments in blood flow or its distribution play a central role.

PAGE 21

10 CHAPTER II DISSOCIATION OF LOCAL AND GLOBAL SKELETAL MUSCLE OXYGEN TRANSPORT METRICS IN TYPE 2 DIABETES Co authors: Timothy A. Bauer, Judith G. Regensteiner, Irene E. Schauer, Amy G. Huebschmann, and Jane E. B. Reusch 1. Preface This chapter provides an illustration of the disconne ct between blood flow and tissue oxygenation that is thematic in this dissertation. If blood flow were the limiting factor to skeletal muscle oxygenation in diabetes, one would expect that : 1) blood flow is reduced, 2) blood flow is correlated to skeletal muscle oxygenation, and 3) blood flow is correlated to exercise capacity. In fact, none of these three statements is true in the diabetic patient population used for this study Instead, t here appears to be a general trend towards diabetes related dissocia tion of local and global oxygen transport parameters that are correlated in overweight controls Furthermore, measures of lo cal skeletal muscle oxygenation and deoxygenation are more heterogeneous is diabetes, suggesting that heterogeneity of something may be related to the dissociation of blood flow and oxygen uptake in diabetic skeletal muscle. This chapte r has been submitted to Diabetes Care for review. 2. Abstract Aims: Exercise capacity is impaired in type 2 diabetes, and this impairment predicts excess morbidity and mortality. This defect appears to involve excess skeletal muscle deoxygenation, but the underlying mechanisms remain unclear. We hypothesized that reduced blood flow, reduced local recruitment of blood volume/hematocrit, or both contribute to excess skeletal muscle deoxygenation in type 2 diabetes. Methods: In patients with (n=23) and without (n=18) type 2 diabetes, we recorded maximal reactive hyperemic

PAGE 22

11 leg blood flow, peak oxygen utilization during cycling ergometer exercise (VO2peak), and near infrared spectroscopy derived measures of exercise induced changes in skeletal muscle oxygenation and blood volume/hematocrit. Results: We observed a significant increase (p<0.05) in skeletal muscle deoxygenation in type 2 diabetes despite similar blood flow and recruitment of local blood volume/hematocrit. Within the control group skeletal muscle deoxygenation, local recruitment of micro vascular blood volume/hematocrit, blood flow, and VO2peak are all mutually correlated. None of these correlations were preserved in type 2 diabetes. Conclusions: These results suggest that in type 2 diabetes 1) skeletal muscle oxygenation is impaired, 2) t his impairment may occur independently of bulk blood flow or local recruitment of blood volume/hematocrit, and 3) local and global metrics of oxygen transport are dissociated. 3. Introduction According to CDC estimates, nearly half of American adults now have type 2 diabetes or prediabetes (94) The estimated lifetime risk of developing diabetes has risen to greater than 30% (95) People with type 2 diabetes suffer disproportionate cardiovascular and all cause mortality, in addition to potentially disabling complications such as diabetic retinopathy and diabetic foot ulcer. The pathological mechanisms leading to excess morbidity and mortality in the diabetic population are not yet fully understood, but vascular and microvascular dysfunction are a common theme. Consistent with this observation, aerobic exercise capacity (VO 2max a powerful clinical predictor of mortality (7; 11; 96) ), is impaired in type 2 diabetes (47) Moreover, impaired aerobic exercise capacity is associated with diabetic complications (12) and insulin resistance (5) suggesting that the causes of impaired exercise capacity are intimately related t o the broader pathology of type 2 diabetes. This

PAGE 23

12 possibility mandates intensive investigation of the causes of impaired aerobic exercise capacity in type 2 diabetes. Although the precise mechanisms by which VO 2max is reduced in type 2 diabetes are not yet fully understood, impaired oxygen delivery is a likely contributor. Rodent studies reveal skeletal muscle hypoxia at the onset of exercise in rodents with diabetes (97) and these findings are corroborated by findings of increased skeletal muscle deoxygenation during exercise in humans with type 2 diabetes (52) There are several plausible mechanisms that may contribute to impaire d oxygen delivery to skeletal muscle in type 2 diabetes, including reduced capillary density (70) reduced blood flow (51) loss of capillary perfusion (98) and heterogeneous distribution of blood flow (25; 63) Altho ugh each of these possible contributors has been previously noted, contradictory reports exist in the literature (especially with regards to bulk blood flow, e.g. (29) ), and it remains unclear which specific parameters, if any, may limit oxygen delivery. Oxygen delivery to peripheral tissues consists of convective (i.e. arrival of oxygenated blood) and diffusive (i.e. transport of oxygen from blood to mitochondria) steps. The convective step in oxyg en delivery is primarily determined by bulk blood flow to the exercising muscle, and also determined to some extent by distribution of blood flow within the skeletal muscle circulation. In this manuscript, we report a metric of maximal leg blood flow as r ecorded during reactive hyperemia (RH) by venous occlusion plethysmography. The diffusive step of oxygen delivery has many determinants. One important component is the net recruitment of tissue hemoglobin content during exercise whether through microvascu lar recruitment, increased capillary hematocrit, or any other mechanism. Reductions in microvascular blood volume/hematocrit or its recruitment are widely reported in animal

PAGE 24

13 models of type 2 diabetes (27; 70; 98) and these changes are likely to contribute to any defect in oxygen diffusion. In t his study, we used near infrared spectroscopy (NIRS) to monitor skeletal muscle deoxygenation/oxygenation and local recruitment of microvascular blood volume/hematocrit in response to exercise. Interpretation of the NIRS signal is complex because tissue co mposition (e.g. adipose tissue thickness) influences NIRS results (99) and also because a majority of signal may come from myoglobin rather than hemoglobin (100) Furthermore, both hemoglobin and myoglobin can change in oxygenation status (i.e. oxygenated vs deoxygenated), but only hemoglobin content can acutely increase or decrease in the sampled tissue. Thus we interpreted changes in deoxygenation/oxygenation st atus (deoxy[hemoglobin +myoglobin], [HHb] and oxy[hemoglobin +myoglobin], [OHb]) as changes in muscle oxygenation, and interpreted changes in local signal intensity (total hemoglobin, [tHb]) as a change in a composite of local blood volume and microvascula r hematocrit. Our overarching hypothesis was that during exercise, either reduced blood flow, reduced local recruitment of microvascular blood volume/hematocrit, or both contribute to impaired oxygen delivery to skeletal muscle in type 2 diabetes. Based on our assessments of oxygen delivery and interpretations of NIRS signals, we formulated several sub hypotheses to test the relationship of the convective and diffusive steps of oxygen delivery to reduced exercise capacity in type 2 diabetes: 1) RH leg bloo d flow correlates to VO 2peak in both type 2 diabetes and in BMI matched controls (i.e. blood flow limits oxygen uptake with or without type 2 diabetes), 2) increase in total hemoglobin correlates to VO 2peak in type 2 diabetes but not in BMI matched control s (i.e. local recruitment of microvascular blood volume/hematocrit limits oxygen diffusion only in type 2 diabetes), 3) RH leg blood flow

PAGE 25

14 correlates to increase in total hemoglobin in both type 2 diabetes and in BMI matched controls (i.e. blood flow and lo cal recruitment of microvascular blood volume/hematocrit are coordinated), and 4) skeletal muscle deoxygenation correlates inversely with VO 2peak in type 2 diabetes but not in BMI matched controls (i.e. skeletal muscle deoxygenation limits oxygen uptake in type 2 diabetes). 4. Methods 4.1 Source of data The source of data analyzed in this manuscript is the INSITE study (Reusch (JEBR), Regensteiner (JGR) and Bauer (TAB), unpublished), which was designed to investigate differences in, and the effects of anti oxidant treatment or exercise training on, exercise capacity and insulin sensitivity in overweight, middle aged men and premenopausal women. In this study, middle aged, overweight, and sedentary (defined as <1 hour of exercise per week) subjects either wit h (n=23) or without (n=18) type 2 diabetes underwent an incremental maximal exercise test on a cycling ergometer to assess VO2peak by metabolic cart (Medgraphics CPX/D, Medical Graphics Corp., St. Paul, MN, USA) (JEB, JGR and TAB manuscript in progress). O n a subsequent date, participants also performed two separate five minute bouts of constant work rate cycling at 85% of lactate threshold, as determined by the V slope method. Bouts were separated by a 10 minute rest period. Changes in muscle concentration s of [tHb], [OHb], and [HHb] were monitored by NIRS for the duration of the exercise protocol. Values for [tHb] and [HHb] used in this study were recorded in the vastus lateralis at rest and during constant work rate cycling at 85% of lactate threshold. In addition, maximal blood flow during reactive hyperemia (RH) was recorded using venous occlusion plethysmography.

PAGE 26

15 4.2 NIRS data acquisition Tissue total hemoglobin+myoglobin ([tHb]), deoxy[hemoglobin+myoglobin] ([HHb]), and oxy[hemoglobin+myoglobin] ([OHb]) were assessed by a frequency domain multi distance NIRS monitor (Optiplex TS, ISS, Champaign, IL, USA) during each constant work rate exer cise test. The NIRS monitor emits two wavelengths (690 and 830nm) and measures absorbance at distances of 2.0, 2.5, 3.0 and 3.5 cm. The NIRS data were sampled continuously and recorded at 50Hz. Upon export, data were down sampled to 1 Hz using a running average of the higher resolution 50 Hz data. During cycling exercise tests, the NIRS probe was positioned on the distal third of the vastus lateralis of the dominant limb, secured using a Velcro strap, and covered with a cloth bandage to exclude ambient li ght. The NIRS monitor was calibrated prior to each visit using a calibration phantom of known scattering and optical properties. 4.3 NIRS data analysis Resting values of tissue [tHb], [HHb], and [OHb] were obtained by averaging the 30s prior to the o nset of exercise. Exercise values of these parameters were obtained by averaging values between 270s and 300s following the onset of exercise. The absolute change in [tHb], [HHb], and [OHb] from rest to steady state exercise were recorded as well. The chan ge in [tHb] reflects local recruitment of microvascular blood volume/hematocrit (only the hemoglobin portion of the [tHb] signal can change acutely). The changes in [HHb] (deoxy[hemoglobin+myoglobin] accumulation) and [OHb] (oxy[hemoglobin+myoglobin] deple tion) represent changes in local skeletal muscle oxygen availability and deoxygenation, respectively. Data which included negative values for concentrations of any hemoglobin

PAGE 27

16 species were discarded. A visualization of the NIRS parameters employed in this a nalysis is included in Figure 1. Figure 1: Illustration of NIRS measures used for analysis. Following the onset of exercise, oxyhemoglobin (OHb) acutely decreases and then gradually returns partway to baseline. The depletion of oxy(hemoglobin+myoglobin) at steady state was recorded as a metric of oxygen availability. Deoxy(hemoglobin+myoglobin) (HHb), meanwhile, increases gradually to steady state. The increase in [HHb] was recorded as a metric of muscle deoxygenation. Finally, total hemoglobin+myoglobin (tHb) comprises the sum of the [OHb] and [HHb] signals. The increase in [tHb] from rest to steady state exercise reflects the local increase in total hemoglobin concentration (i.e. local recruitment of microvascular blood volume/hematocrit). 4.4 Inter gro up comparisons Values for [OHb] depletion (as assessed by change from rest to steady state exercise) were compared between type 2 diabetes and control groups to assess the hypothesis that skeletal muscle oxygenation during exercise is impaired in type 2 diabetes. Values for [HHb] accumulation (as assessed by change from rest to steady state exercise) were compared between type 2 diabetes and control groups to assess the hypothesis that skeletal muscle deoxygenation during exercise is increased in type 2 diabetes. Values f or increase in total

PAGE 28

17 hemoglobin were compared between type 2 diabetes and control groups to assess the hypothesis that the ability to recruit additional blood volume during exercise is impaired in type 2 diabetes. Values for RH blood flow were compared bet ween type 2 diabetes and control groups, in order to assess the hypothesis that maximal blood flow is reduced in type 2 diabetes. Unpaired t tests were used to compare groups for each of these comparisons, and a standard f test was used to assess possible differences in group variance. Where significant (101) was applied to the original t test to compare group means. Because each hypothesis was assessed independently (i.e. single comparison for each hypothesis), multiple comparisons corrections were not used in this analysis. 4.5 Intra group comparisons Increase in total hemoglobin and VO2peak were compared within each group to assess the hypothesis that local recruitment of microvascular blood volume/hematocrit (as a determinant of oxygen diffusion) limits oxygen uptake in type 2 diabetes, but not in con trols. Local deoxygenation and VO2peak were compared within each group to assess the hypothesis that skeletal muscle deoxygenation limits oxygen uptake in type 2 diabetes, but not in controls. RH blood flow and VO2peak were compared within each group to as sess the hypothesis that that blood flow (as a determinant of oxygen convective delivery) limits oxygen uptake in both type 2 diabetes and controls. RH blood flow and increase in total hemoglobin were compared within each group to assess the hypothesis tha t whole limb blood flow and local recruitment of microvascular blood volume/hematocrit are associated in both correlations. Because each hypothesis was assessed independent ly (i.e. single comparison in

PAGE 29

18 each group for each hypothesis), multiple comparisons corrections were not used in this analysis. 5. Results 5.1 Study subject characteristics In this analysis, we used NIRS data from overweight, middle aged subjects with (n= 23) or without (n=18) type 2 diabetes. Study groups were well matched in age (46.9+/ 5.2 years type 2 diabetes vs 44.8+/ 6.1 years control, p=NS), height (174.8+/ 9.5 cm type 2 diabetes vs 173.1+/ 9.9 cm control, p=NS), weight (92.6+/ 17.5 kg type 2 diabet es vs 91.02+/ 10.4 kg control, p=NS), and BMI (30.1+/ 3.9 type 2 diabetes vs 30.4+/ 2.7 control, p=NS). Study groups differed significantly (p<0.05) in HbA1c (6.9+/ 0.8% type 2 diabetes vs 5.3+/ 0.4% control) and sex balance (78% male type 2 diabetes vs 56 % male control). Patients with type 2 diabetes had an average duration of diagnosis of 3.9+/ 3.4 years. All group averages included above are expressed as mean+/ standard deviation. 5.2 Group differences in skeletal muscle deoxygenation/ oxygenation and limb blood flow Group comparisons of muscle deoxygenation/oxygenation, RH limb blood flow, and local recruitment of microvascular blood volume/hematocrit are shown in Figure 2. [HHb] accumulation from rest to steady state exercise is significantly increas ed in type 2 diabetes relative to control (p=0.003, Figure 2A), consistent with our hypothesis of increased skeletal muscle deoxygenation in type 2 diabetes. [OHb] depletion from rest to steady state exercise is significantly increased in type 2 diabetes r elative to control (p=0.01, Figure 2B), consistent with our hypothesis of impaired skeletal muscle oxygen availability during exercise in type 2 diabetes. No significant difference (p=NS) was observed between type 2 diabetes and control

PAGE 30

19 in increase in tota l hemoglobin, failing to support our hypothesis of impaired blood volume recruitment in type 2 diabetes. There was also no significant difference observed (p=NS) between type 2 diabetes and control in RH blood flow, failing to support our hypothesis of imp aired limb blood flow in type 2 diabetes. Variance was significantly increased (p<0.05) in type 2 diabetes for all local NIRS metrics (increase in [tHb], [HHb] accumulation, and [OHb] depletion), but not for RH blood flow (p=NS). Figure 2: Impaired muscl e oxygenation despite normal RH blood flow and increase in total hemoglobin in type 2 diabetes. ( A ) [HHb] accumulation is significantly (p=0.003) increased in type 2 diabetes, suggesting greater skeletal muscle deoxygenation during exercise. In addition, the variance of [HHb] accumulation is significantly (p=0.002) increased in type 2 diabetes, suggest ing greater heterogeneity of skeletal muscle deoxygenation among type 2 diabetes subjects. ( B ) [OHb] depletion is significantly (p=0.01) increased in type 2 diabetes, suggesting reduced oxygen delivery. In addition, the variance of [OHb] depletion is signi ficantly (p=0.003) increased in type 2 diabetes, suggesting greater heterogeneity of

PAGE 31

20 skeletal muscle oxygenation. ( C ) Increase in total hemoglobin is not significantly different from controls in type 2 diabetes (p=NS), suggesting no difference in the incre ase in local recruitment of microvascular blood volume/hematocrit during moderate CWR exercise. However, the variance in the increase in total hemoglobin is significantly (p=0.006) increased in type 2 diabetes, suggesting greater heterogeneity of increase in total hemoglobin. ( D ) RH blood flow does not significantly differ (p=NS) between Controls and type 2 diabetes, suggesting that total limb capacity for blood flow is not limiting to muscle oxygenation. 5.3 Within group comparisons of deoxygenation and i ncrease in total hemoglobin to VO 2 peak Comparisons within each group of deoxygenation and increase in total hemoglobin metrics to VO2peak are shown in Figure 3 below. Increase in total hemoglobin significantly correlates (p=0.027) with VO2peak in control subjects (Figure 3A) but not in type 2 diabetes subjects (Figure 3B, p=NS), failing to support our hypothesis that local recruitment of microvascular blood volume/hematocrit limits oxygen uptake in type 2 diabetes but not in controls. Local deoxygenation s ignificantly correlates (p=0.0036) with VO2peak in control subjects (Figure 3C) and does not correlate (p=NS) to VO2peak in type 2 diabetes (Figure 3D), failing to support our hypothesis that skeletal muscle deoxygenation limits VO2peak in type 2 diabetes but not in controls.

PAGE 32

21 Figure 3: Increase in total hemoglobin and muscle deoxygenation do not correlate to VO2peak in type 2 diabetes. ( A ) Increase in total hemoglobin and VO2peak are significantly (p=0.027) correlated in control subjects. ( B ) Increase in total hemoglobin and VO2peak are not correlated (p=NS) in type 2 diabetes. ( C ) Local deoxygenation and VO2peak are significantly (p=0.0036) correlated in control subjects. ( D ) Local deoxygenation and VO2peak are not correlated (p=NS) in type 2 diabetes. 5.4 Within group comparisons of RH blood flow with VO 2 peak and increase in total hemoglobin Comparisons of VO2peak and increase in total hemoglobin are shown in Figure 4 below. RH blood flow correlates significantly (p=0.002) with VO2peak in controls (Figure 4A) but not in type 2 diabetes (p=NS, Figure 4B), consistent with our hypothesis that blo od flow limits oxygen uptake in controls, but failing to support our hypothesis that blood flow

PAGE 33

22 limits oxygen uptake in type 2 diabetes. RH blood flow also correlates significantly (p=0.018) with increase in total hemoglobin in controls (Figure 4C) but not in type 2 diabetes (p=NS, Figure 4D), consistent with our hypothesis that limb blood flow and local recruitment of microvascular blood volume/hematocrit are coordinated in controls, but failing to support our hypothesis that the same is true in type 2 dia betes. Figure 4: Mutual correlations between RH blood flow, increase in total hemoglobin, and VO2peak are abolished in type 2 diabetes. ( A ) RH blood flow and VO2peak are significantly correlated (p=0.002) in control subjects. ( B ) RH blood flow and VO2peak are not correlated (p=NS) in type 2 diabetes. ( C ) RH blood flow and increase in total hemoglobin are significantly correlated (p=0.018) in control subjects. ( D ) RH blood flow and increase in total hemoglobin are not correlated ( p=NS) in type 2 diabetes.

PAGE 34

23 6. Discussion In this analysis, we found evidence that skeletal muscle oxygen availability is reduced and skeletal muscle deoxygenation is increased in exercise with type 2 diabetes as compared to a similarly overweight and sedentary control group, consistent with previ ous reports (52; 97; 102; 103) We also found that the increase in total hemoglobin during exercise is not reduced in type 2 diabetes. This finding is consistent with reports of pr eserved microvascular recruitment (measured using contrast enhanced ultrasound) during exercise with type 2 diabetes (104) We also found that RH blood flow was not impaired in type 2 diabetes. Previous studies have found conflicting results when investigating blood flow l imitations in type 2 diabetes (29; 51) and the degree to which health status differs between disease and control groups may account for these inconsistencies. Finally, we found that variance of all local oxygenation metrics ([HHb] accumulation, [OHb] depletion, and increase in [tHb]) was increased in type 2 diabetes, despite no increase in variance of global or whol e limb oxygen transport metrics (VO 2peak and RH blood flow). Collectively, these results suggest that skeletal muscle oxygenation during exercise is impaired in type 2 diabetes, and that this impairment does not appear to be related to either decreased lim b blood flow or local recruitment of microvascular blood volume/hematocrit. Our correlation analysis returned unexpected results. As hypothesized, maximal RH blood flow and VO 2peak were correlated in control subjects, but this correlation was lost in type 2 diabetes. Meanwhile, increase in total hemoglobin was correlated to VO 2peak in control subjects but not in type 2 diabetes, contrary to our hypothesis that local recruitment of microvascular blood volume/hematocrit would limit exercise capacity only in type 2 diabetes. The recurring theme of correlation in health and dissociation in type 2 diabetes

PAGE 35

24 continued with dissociation of skeletal muscle deoxygenation from VO 2peak and dissociation of maximal RH blood flow from increase in total hemoglobin. Collect ively, these findings suggest that in type 2 diabetes, local metrics of tissue oxygen transport do not reflect whole limb or whole body oxygen transport, blood flow is dissociated from local recruitment of microvascular blood volume/hematocrit, and factors besides bulk blood flow and local recruitment of microvascular blood volume/hematocrit are likely to limit oxygen uptake during exercise. Any plausible and complete explanation of these data must satisfy three observations: 1) skeletal muscle oxygenation is impaired in type 2 diabetes, 2) skeletal muscle oxygenation is more heterogeneous in type 2 diabetes, and 3) these changes appear to be unrelated to bulk blood flow or local recruitment of microvascular blood volume/hematocrit. While our NIRS data do n ot allow us to draw definitive conclusions about the causes of impaired skeletal muscle oxygenation in type 2 diabetes, the literature does provide a plausible explanation that could satisfy all three of these requirements. Distribution of blood flow is mo re spatially heterogeneous in the obese Zucker rat model of type 2 diabetes, and this perfusion heterogeneity is associated with impaired oxygen uptake (25; 26; 60; 63) Simulation studies reveal that heterogeneous perfusion results in impaired muscle oxygenation on average, because over perfused vessels cannot fully compensate for under perfused vessels (63 66) By definition, heterogeneous perfusion not resulting from heterogeneous tissue demand would result in flow/VO 2 mismatch. Moreover, these effects would influence skeletal muscle deoxygenati on/oxygenation even in cases of normal bulk blood flow and local recruitment of microvascular blood volume/hematocrit. Although the findings referenced above have not yet been translated to human type 2 diabetes, it has been

PAGE 36

25 shown in humans that increased perfusion heterogeneity correlates with reduced oxygen extraction, and perfusion heterogeneity decreases in response to endurance training (67; 68) supporting the plausibility of this explanation. Moreover, the effects of age on the vasculature, which parallel the influence of type 2 diabetes (70; 105) are themselves spatially heterogeneous (106) Visualizations of skeletal muscle perfusion in type 2 diabetes reported by Zheng et al appear more heterogeneous qualitatively ( 59) It is worth noting, however, that the studies cited in the previous paragraph pertain to spatial heterogeneity within a single muscle, whereas the heterogeneity we observed in our NIRS was measured as population heterogeneity, and only one site per muscle was assessed. Although spatial heterogeneity could produce population heterogeneity in local but not global metrics as a statistical artifact if only one site were observed (which is exactly what our study design entailed), we cannot rule out the possibility that population heterogeneity was observed due to variance among study subjects rather than o bserved as an artifact of variance among locations. Future studies will be required to test the hypothesis that heterogeneous distribution of blood flow on both microvascular and macrovascular scales during exercise may contribute to reduced oxygen deliver y to skeletal muscle in type 2 diabetes. There are some limitations of our study design. First, leg blood flow, VO 2peak and skeletal muscle oxygenation measures were each recorded under separate conditions (following venous occlusion, during maximal cycli ng exercise, and during submaximal cycling exercise, respectively). It is possible that these discrepancies influenced our results. However, given that these disparate metrics were correlated in health but not in type 2 diabetes, the recurring theme of dis sociation in type 2 diabetes remains relevant. The use of

PAGE 37

26 NIRS to assess skeletal muscle deoxygenation/oxygenation may also introduce complexity to the interpretation of our data, given that there is controversy in the field about the relative contribution s of hemoglobin (i.e. vascular) and myoglobin (i.e. muscular) contributions to the NIRS signal (100) Our analysis avoided this issue by reporting local deoxygenation/oxygenation without any attempt to distinguish vascular/intramuscular contributions and by reporting only the change in [tHb] from rest to exercise as a vascular specific measure. This parameter must be vascular in origin, given that myoglobin does not acutely enter or leave the muscle upon contractio n. Finally, the sex imbalance in our type 2 diabetes group might be expected to influence our results. If sex differences were driving the observed correlations in health and dissociations in type 2 diabetes, however, one would expect reduced variance in t he type 2 diabetes group, and in fact we observed increased variance in type 2 diabetes. Thus, the dissociations in type 2 diabetes occurred despite a greater dynamic range over which correlations could be observed. In summary, we found that skeletal musc le oxygenation during exercise is impaired in type 2 diabetes, and that this impairment can occur independently of changes in limb blood flow or local recruitment of microvascular blood volume/hematocrit. We also found that correlations between local and g lobal oxygen transport metrics were abolished in type 2 diabetes, and that local muscle oxygenation is more heterogeneous. Although our data do not allow definitive conclusions as to the cause of these changes, it is plausible that heterogeneous blood flow distribution may account for dissociation of local and global oxygen transport in type 2 diabetes. Future studies will be required to more fully understand the heterogeneity and impaired coordination of skeletal muscle oxygenation in type 2 diabetes. In l ight of the previously discussed association of impaired exercise capacity with

PAGE 38

27 premature mortality and excess morbidity (7; 11; 96) understanding the mechanisms leading to impaired oxygen transport in type 2 diabetes holds great translational potential.

PAGE 39

28 CHAPTER III A CONCEPTUAL FRAMEWORK FOR PREDICTING AND ADDRESSING THE CONSEQUENCES OF D ISEASE REL ATED MICROVSASCULAR DYSFUNCTION Co authors: Jefferson C. Frisbee and Jane E. B. Reusch 1. Preface It is plausible that the diabetic skeletal muscle oxygenation heterogeneity described in the previous chapter is related to heterogeneous vascular delivery of oxygen. This chapter describes a series of sensitivity analyses drawing from established principles in mass transport and anatomy aiming to elucidate the roles of bulk blood flow, capillary density, perfusion heterogeneity to microvascular solu te flux. These analyses apply to the exchange of not only oxygen but of all small, blood borne molecules. The results of this analysis provide a theoretical framework for understanding how blood flow and its distribution relate to clinical phenotypes such as impaired exercise capacity and insulin resistance, and provide a plausible explanation for why those two phenotypes are correlated. The key finding of this study was that the distribution of blood flow (number of perfused capillaries and flow distribut ion within these capillaries) can exert profound effects on tissue oxygenation and other, related mass transport processes independently of bulk blood flow. This chapter was published in Microcirculation in 2017. 2. Abstract Objective: A growing body of evidence indicates that impaired microvascular perfusion plays a pathological role in a number of diseases. This manuscript aims to better define which aspects of microvascular perfusion are important, what mass transport processes (e.g. insulin action, tissue oxygenation) may be impacted, and what therapies might

PAGE 40

29 reverse these pathologies. Methods: We derive a theory of microvascular perfusion and solute flux drawing from established relationships in mass transport and anatomy. We then ap ply this theory to predict relationships between microvascular perfusion parameters and microvascular solute flux. Results: For convection limited exchange processes (e.g. pulmonary oxygen uptake), our model predicts that bulk blood flow is of primary impo rtance. For diffusion limited exchange processes (e.g. insulin action), our model predicts that perfused capillary density is of primary importance. For convection/diffusion co limited exchange processes (e.g. tissue oxygenation), our model predicts that v arious microvascular perfusion parameters interact in a complex, context specific manner. We further show that our model can predict established mass transport defects in disease (e.g. in sulin resistance in diabetes). Conclusions: The contributions of micr ovascular perfusion parameters to tissue level solute flux can be described using a minimal mathematical model. Our results hold promise for informing therapeutic interventions tar geting microvascular perfusion. 3 Introduction Research investigating microvascular perfusion has traditionally focused on the quantity of blood flow supplied by upstream arteries. The scientific literature increasingly shows that the distribution of this blood flow within the microcirculation is also physiologically important. Evidence of microvascular dysfunction contributing to tissue hypoxia has been reported in the metabolic syndrome (25; 26) and sepsis (107; 108) Similar mic (109) inflammatory bowel disease (110) and hypertension and obesity (111) Hints of a mechanism linking impaired microvascular perfusion to tissue h ypoxia can be found in simulation studies of ischemic

PAGE 41

30 heart disease (65) and cerebrovascular disease (64) Tissu e hypoxia is the most common focus of microvascular perfusion studies, but other solute exchange processes are also influenced by microvascular perfusion. For example, animal models of insulin resistance show that the primary barrier to glucose flux in the insulin resistant state lies in the extracellular (i.e. vascular to interstitial) step of glucose and insulin delivery (112) The sheer variety of diseases in which microvascular perfusion defects occur and the variety of consequences related to impaired perfusion suggests that microvascular perfusion is a critical physiological process in its own right. This manuscript will discuss the importance of three distinct perfusion parameters: the amount of blood flowing through a microvascular network (bulk blood flow), its distribution at microvascular bifu rcations (perfusion heterogeneity), and the number of capillaries accessible to flowing blood (perfused capillary density). It is unclear which solute exchange processes (e.g. oxygen delivery, lactate clearance) are impacted by which specific microvascular perfusion parameters and how perfusion abnormalities might be therapeutically targeted. The importance of bulk blood flow is illustrated by diseases that substantially decrease blood supply (e.g. heart failure, peripheral arterial disease, etc). Perfusion heterogeneity has been empirically demonstrated to modulate oxygen flux in skeletal muscle (25; 26) and is also thought to contribute to ischemic heart disease and cerebrovas cular disease (64; 65) Perfused capillary density is modulated both by long t erm changes in anatomical capillary density (density of capillaries present within the tissue, perfused or not) and by short term changes in the fraction of capillaries present that are actually perfused. Long term changes in anatomical capillary density c an be stimulated by diseases such as type 2 diabetes mellitus (T2DM) (70) and physiological stressors such as

PAGE 42

31 exercise training (113) Short term changes in perfused capillary density can be caused by stimuli such as endothelial glycocalyx degradation, as is observed in sepsis and hyperglycemia (81; 114) Consideration of these parameters in combination rather than in isolation allows inference of their relative contextual importance. In addition to developing a generalized theory of microvascular perfusion and solute flux, we will also attempt to persuade the reader of its utility in identifying therapeutic strategies. This is perhaps best illustrated by successful examples of therapies targeting perfused capillary density. For example, one major cause of reductions in perfused capillary density is plugging of capillaries by adherent leukocytes or microemboli (115; 116) Pers istent reduction in perfused capillary density is a major determinant of organ failure and mortality in sepsis (117) As would be expected from these findings, therapies that reduce adhesive interactio ns in the microvasculature (and thus un plug blocked capillaries) also improve patient outcomes (108) Anatomical capillary density and hemostatic status are also major determinants of insulin sensitivity in T2DM (118 122) Insulin sensitizing drugs such as metformin tend to also reduce adhesive interactions in the microvasculature (123) and this may be part of their mechanism of action. In both sepsis and T2DM, therapies that improve microvascular function served to treat pathologies that are not generally discussed as microvascular defects. In both cases, some element of trial and error was required to even identify the microcirculation as a potential issue, and again to determine which aspects of microvascular perfu sion were most relevant. Trial and error was then required to identify therapies to exploit these parameters, and often this mechanism of action was discovered post hoc. The analysis included in this manuscript aims to enable translational researchers to

PAGE 43

32 b ypass much of this confusion by accurately predicting which microvascular parameters are relevant to their research and how they might be targeted. 4 Methods 4 .1 Governing Equations We begin our analysis with a widely used formulation of microvascular solute flux (124) : (1) Here is solute flux across the endothelium, is capillary blood flow, is arterial concentration, is interstitial concentration is a metric of permeability to the solute of interest, and is capillary surface area. Note that is used instead for capillary surface area in traditional representations of this formula. We elected to use to denote surface area in this analysis so as not to conflict with subsequent use of to denote sensitivity. For purposes of subsequent sensitivity analyses, we further specify the definition of permeabilit y using the following equation: (2) Here is cap illary volume, is a diffusion rate constant, and is the effective radius of diffusion. Using this definition of permeability, a more complete formulation of solute flux is: (3) For simplicity of presentation in subsequent analyses, we will also introduce the dimensionless diffusion/convection matching parameter (Equation 4). Values of much

PAGE 44

33 greater than 1 reflect an excess of diffusion capacity relative to bulk blood flow, w hile those much less than 1 reflect an excess of bulk blood flow relative to diffusion capacity. (4) Incorporating this simplified nota tion into Equations 1 3 yields: (5) This governing equation will subsequently be applied to determine the effect s of bulk blood flow (sections 5.1 and 5 .2) diffusion capacity (sections 5.1 and 5 .3), perfu sed capillary density (section 5 .4), and perfusion he terogeneity (section 5 .5), along with s imulations of the consequences of experimentally defined phenotypes including alterations in each of these parameters on micr ovascular solute flux (section 5 .6). 4 .2 Fractional Equilibration and Diffusion/Convection Matching Individual determinants of diffusion/convection matching (e.g. capillary blood volume, capillary surface area, diffusion capacity etc.) are difficult to measure. However, their physical consequences can be empirically observed using fractional equilibration ( ) of the solute of int erest between capillary blood and the interstitium, as defined by Eugene Renkin (125) Fractional equilibration is defined as the ratio of arteriovenous concentration difference to arterial interstitial concentration difference (Equation 6, Figure 5 A). Determining fractional equilibration requires measurement of solute c oncentration in three compartments arterial blood ( ), venous blood ( ), and interstitial fluid ( ). Plausible reference values of and for compounds of physiologic i nterest are included in Figure 5 B. (6)

PAGE 45

34 Figure 5 : Solutes vary in their degree fractional equilibration with the interstitium during capillary transit. ( A ) Visualization of the concept of equilibration. Complete equilibration implies that venous solute concentration is equal to interstitial solute concentration, while partial equilibration implies that venous concentration lies somewhere between arterial and interstitial concentrations. ( B ) Examples of physiologically relevant compounds and plausible values for their fractional equilibration. Fractional equilibration is related to the diffusion/convection matching parameter by following transformation: (7) approaches 1 in cases where diffusion capacity exceeds convective delivery (e.g. pulmonary oxygen flux) and approaches zero in cases where convective delivery exceeds diffusion capacity (e.g. insulin delivery to skeletal muscle). 4 .3 Blood Flow Distribution To determine the influence of arteriolar perfusion heterogeneity on microvascular solute flux, we simulated blood flow distribution within an idealized arteriolar ne twork as illustrated in Figure 6 Each arteriole bifurcates into two sm aller daughter vessels (Figure 6 A) for 15 consecutive vessel generations for a total of 2 15 simulated capillaries. Distribution of blood flow at each bifurcation is described by the parameter which is defined as the

PAGE 46

35 fraction of parent vessel blood flow directed to the highe r flow daughter vessel (Figure 6 B). Thus represents homogenous perfusion, and increasing deviation from 0.5 represents increasingly heterogeneous perfusion. Figure 6 : Simulated distribution of blood flow in an idealized microvascular network. ( A ) The simulated microvascular network consisted of a series of arterioles terminating in a symmetric bifurcation of two smaller arterioles. Fifteen vessel generations were simulated for a total of 2 15 individual capillaries; only four vessel generations are illustrated to preserve clarity. ( B ) Distribution of flow rate at each microvascular bifurcation was defined by the parameter which is defined as the fraction of parent vessel flow directed to the higher flow daughter vessel. Figure recreated with permissions from Butcher et al 2013 (23) 5 Results 5 .1 Effects of Blood Flow and Diffusion Capacity The effects of bulk blood flow on microvascular solute flux are visualized in Figure 7 A. The effects of diffusion capacity on microvascular solute flux are visualized in Figure 7 B. These results reflect the predictions of Equation 8. Increases in blood flow and diffusion capacity both result in saturable increases in microvascular solute flux. The level at which increasing blood flow yields diminish ing returns increases with increasing diffusion capacity, and vice versa. For diffusion limited, low fractional equilibration molecules (e.g. insulin), microvascular solute flux increases robustly with increasing diffusion capacity but increases minimally with increasing blood flow. For convection limited, high equilibration molecules

PAGE 47

36 (e.g. oxygen in the lung), microvascular solute flux increases minimally with increasing diffusion capacity but increases robustly with increasing blood flow. For diffusion/co nvection co limited, intermediate equilibration molecules (e.g. oxygen in skeletal muscle), both diffusion capacity and blood flow limit the rate of microvascular solute flux. Figure 7: Effects of blood flow and diffusion capacity on microvascular solute flux. ( A ) Solute flux increases asymptotically with increasing flow. Increasing diffusion capacity increases the threshold of diminishing returns for increasing flow. ( B ) Solute flux incr eases asymptotically with increasing diffusion capacity. Increasing flow increases the threshold of diminishing returns for increasing diffusion capacity. 5 .2 Sensitivity of Solute Flux to Blood Flow To quantify the sensitivity of microvascular solute fl ux to blood flow, the modified convection/diffusion matching parameter was defined as a function of fraction of baseline flow rate ( ), which alters convection/diffusion matching solely through its interaction with blood flow at baseline ( ): (8) Combining Equations 8 and 5 while accounting f or increased blood flow yields:

PAGE 48

37 (9) The sensitivity of microvascular solute flux to changes in blood flow ( ) is defined as the partial derivative of microvascular solute flux ( ) with respect to fraction of baseline blood flow ( ) normalized to baseline microvascular solute flux ( ): (10) Baseline blood flow ( ) and arterial interstitial concentration gradient ( ) cancel out when performing this calculation, and so a transformation between fractional equilibration and convection/diffusion matching at baseline ( ) can be used to solve for as a function fractional equilibration at baseline ( ): (11) The results of this calcul ation are visualized in Figure 8 Note that fractional equilibration is also influenced by circumstances that change solute flux, and so is not necessarily equal to For convection limited compounds which undergo near complete equilibration with the interstitium (e.g. oxygen in t he lung), flux increases or decreases near proportionally with increasing/decreasing bulk blood flow. For diffusion limited compounds which undergo minimal equilibration with the interstitium (e.g. insulin), flux changes minimally with changes in bulk bloo d flow. Flux of compounds at intermediate values of baseline fractional equilibration undergoes sub proportional changes in response to changing blood flow.

PAGE 49

38 Figure 8 : Relationship between baseline fractional equilibration and sensitivity to blood flow. C onvection limited compounds, which equilibrate completely or near completely with the interstitium during capillary transit, undergo a proportional increase/decrease in flux in response to an increase/decrease in blood flow. Diffusion limited compounds, wh ich equilibrate minimally with the interstitium during capillary transit, are minimally sens itive to changes in blood flow. 5 .3 Sensitivity of Solute Flux to Diffusion Rate Constant To determine the sensitivity of microvascular solute flux to diffusion capacity, we defined the modified convection/diffusion matching parameter as a function of fraction of baseline diffusion rate constant ( ), which alters convection/diffusion match ing solely through its interaction with diffusion capacity at baseline ( ): (12) Combi ning Equations 12 and 5 yields: (13) The sensitivity of microvascular solute flux to changes in diffusion rate constant ( ) is defined as the partial derivative of modified microvascular solute flux ( ) with respect to

PAGE 50

39 fraction of baseline diffusion rate constant ( ) normalized to baselin e microvascular solute flux ( ): (14) Baseline blood flow ( ) and arterial interstitial concentration gradient ( ) cancel out when performing this calculation, and so the transformation between fractional equilibration and convection/diffusion matching parameter at baseline ( ) can be used to solve for as a function of fractional equilibration at baseline ( ): (15) The results of this calcul ation are visualized in Figure 9 For convection limited compounds which undergo near complete equilibration with the interstitium (e.g. oxygen in the lung), flux changes minimally with changes in diffusion rate constant. For diffusion limited compou nds which undergo minimal equilibration with the interstitium during capillary transit (e.g. insulin), flux increases or decreases near proportionally with increasing/decreasing diffusion rate constant. Flux of compounds at intermediate values of baseline fractional equilibration undergoes sub proportional changes in response to changes in diffusion rate constant.

PAGE 51

40 Figure 9 : Relationship between baseline fractional equilibration and sensitivity to diffusion capacity. Convection limited compounds, which equilibrate completely or near completely with the interstitium during capillary transit, are minimally sensitive to changes in diffusion capacity. Diffusion limited compounds, which equilibrate minimally with the interstitium during capillary transit, undergo a proportional increase/decrease in flux in response to an increase/ decrease in diffusion capacity. 5 .4 Sensitivity of Solute Flux to Perfused Capillary Density To determine the sensitivity of microvascular solute flux to perfused capillary density, we defined the modified convection/diffusion matching parameter as a function of fraction of baseline perfused capillary density ( ). Capillary density influences diffusion/convection matching through a variety of parameters. In this analysis, two modes of interaction with perfused capillary density were considered. In the first case (Equations 16 19), the concentration gradient within the interstitium is substantial and the solute of interest diffuses freely across the endothelium (e.g. lactate clearance). In this case, capillary density modulates the effective diffusion radius ( ) along with capillary surface area ( ) and blood volume ( ):

PAGE 52

41 (16) Combi ning Equations 16 and 5 yields: (17) The sensitivity of microvascular solute flux to relative change in perfused capillary density ( ) is defined as the partial derivative of modified microvascular solute flux ( ) with respect to fraction of baseline perfused capillary density ( ) normalized to baseline microvascular solute flux ( ): (18) Blood flow ( ) and arterial interstitial concentration gradient ( ) cancel out when performing this calculation, and so the transformation between fractional equilibration and convection/diffusion matching at baseline ( ) can be used to solve for as a function of baseline fractional equilibration at baseline ( ): (19) Alternately, interstitial concentration gradients may be negligible relative to trans endothelial concentration gradients (e.g. oxygen in skeletal muscle). In this case, the effective radius of diffusion ( ) is not appreciably altered by perfused capillary density. Repeating the same procedure as in Equations 16 19 with this modification yields: (20) (21) (22)

PAGE 53

42 (23) The results of these calculations are v isualized in Figure 10 Although most compounds will be co limited by trans endothelial and interstitial diffusion processes, the strictly endothelium limited and strictly interstitium limited results shown here compr ise the lower/upper bounds of sensitivity to perfused capillary density. Flux of convection limited compounds, which undergoes near complete equilibration with the interstitium (e.g. oxygen in the lung), changes minimally with changes in perfused capillary density. Flux of diffusion limited compounds which undergo minimal equilibration with the interstitium (e.g. insulin), will undergo increases or decreases between 2x and 3x larger than the corresponding incremental change in perfused capillary density. Fl ux of compounds at intermediate values of baseline fractional equilibration undergoes somewhat smaller but still supra proportional changes in response to changes in diffusion rate constant. Figure 10 : Relationship between baseline fractional equilibration and sensitivity to perfused capillary density. Convection limited compounds, which equilibrate completely or near

PAGE 54

43 completely with the interstitium during capillary transit, are minimally sensitive to changes in perfused capillary density. Diffusion limited compounds, which equilibrate minimally with the interstitium during capillary transit, undergo a 2x 3x proportional increase/decrease in flux in response to an incremental increase/decrease in perfus ed capillary density. The magnitude of the effect of perfused capillary density further depends upon whether the primary site of resistance to solute diffusion lies at the endothelium or within the interstitium, interstitium limited compounds being more se nsitive to perfused capillary density than endothelium limited compounds. Note that the sensitivity of microvascular solute flux to perfused capillary density is geometrically identical to the sensitivity to diffusion rate constant, only scaled up 2x 3x. Thus a 1% decrease in perfused capillary density could be fully compensated by a 2% 3% increase in diffusion rate constant, and vice versa. 5 .5 Sensitivity of Solute Flux to Microvascular Perfusion Heterogeneity In order to define the influence of perfusion heterogeneity on microvascular solute flux, we defined the total microvascular solute flux across a capillary network ( ) based on the mean capillary flow rate ( ) and the fraction of mean flow rate within each capillary ( ): (25) Because the degree of perfusion heterogeneity ( ) was found to influence the degree of sensitivity to perfusion heterogeneity (i.e. flux does not vary linearly with ) it is useful to discuss the sensitivity of microvascular solute flux to perfusion heterogeneity in terms of the relative (%) change in solute flux between two representative perfusion states (see section 4 .3 for discussion of perf usion distribution simulation): (26) Mean capillary blood flow ( ) and arterial interstitial concentration gradient ( ) cancel out when performing this calculation, and so can be expressed as a

PAGE 55

44 function of mean solute convection/diffusion matching ( ) and blood flow distribution (as determined by ): (27) For illustration purposes, we chose to simulate the differences in microvascular solute flux between Lean Zucker Rats (LZR, ) and Obese Zucker Rats (OZR, animal model of T2DM, ), as differences in arteriolar perfusion heterogeneity in these m odels has been extensively characterized in previous studies (23 26; 60) Figure 11 shows th e predicted % decrease in solute flux in OZR relative to LZR as a function of fractional equilibration in the LZR. Compounds whose flux is entirely limited by diffusion or convection are not appreciably impacted by perfusion heterogeneity. However, interme diate equilibration compounds whose flux is co limited by diffusion and convection (e.g. oxygen in skeletal muscle) are sensitive to perfusion heterogeneity because over perfused capillaries cannot fully compensate for under perfused capillaries. For these molecules, losses in capillary perfusion are more impactful than are equivalent gains.

PAGE 56

45 Figure 11: Relationship between baseline fractional equilibration and sensitivity to arteriolar perfusion heterogeneity. Compounds whose flux is limited almost entirely by convection or almost entirely by diffusion are minimally affected by perfusion heterogeneity. Compounds whose flux is co limited by diffusion and convection undergo reduced microvascular flux under conditions of microvascular perfusion heterogeneity. The range of fractional equilibration most affected by perfusion heterogeneity gradually shifts tow ards relatively diffusion limited compounds with an increasing degree of perfu sion heterogeneity (see Figure 12 ). The curve shown reflects the impacts of perfusion heterogeneity on solute flux in the OZR relative to the LZR. 5 .6 Defining the Effects of Co mplex Phenotypes on Microvascular Solute Flux The procedure for determining the effects of perfusion heterogeneity outlined in Equations 25 27 can also be used to determine the effects of a complex phenotype involving several alterations to microvascular perfusion. For this analysis we introduced parameters for fraction of healthy bulk blood flow through the entire microvascular network ( ) and fraction of healthy perfused capillary density ( ): (28)

PAGE 57

46 (29) For demonstration purposes, we characterized the effects of microvascular perfusion defects on solute flux in T2DM (OZR model), sepsis, and acute hyperglycemia or glycocalyx degradation (acute hyperglycemia causes glycocalyx degradation and related perfusi on defects (81) ). In the OZR model of T2DM, bulk blood flow is reduced by ~20%, capillary density is reduced by ~20%, and perfusion heterogeneity is increase d from to (25) In sepsi s, distribution of blood flow to meet metabolic demands is impaired, perfused capillary density is markedly reduced, and microvascular perfusion heterogeneity is readily visible under microscopic observation (84) To simulate a worst case perfusion scenario, we assumed a 20% reduction in bulk blood flow, a 50% reduction in perfused capillary density, and an increase in from 0.52 to 0.7. In cases of acute glycocalyx degradation or hyperglycemia, bulk blood flow is not significantly altered, perfused capillary density is reduced by ~30%, and perfusion het erogeneity is increased, although the precise degree of this increase is unclear (114) For purposes of this analysis, perfusion heterogeneity during hyperglycemia was assumed be similar to that in the OZR. The results of thi s analysis are shown in Figure 12 below. Our model predicts that microvascular solute flux will be impaired in all three phenotypes, but that the precise etiology of this impairment varies.

PAGE 58

47 Figure 12 : Complex phenotypes influence solute flux through a variety of mechanisms. ( A ) Solute flux in diabetes is predicted to be affected primarily by perfused capillary density in the range of interest for insulin action, and by a complex mixture of bulk blood flow, perfusion heterogeneity, and capillary density effects in the range of in terest for exercise

PAGE 59

48 capacity. ( B ) Solute flux in sepsis is predicted to be affected primarily by perfused capillary density and perfusion heterogeneity in the range of interest for both insulin action and exercise capacity. ( C ) Solute flux in hyperglycemia or glycocalyx degradation is predicted to be affected primarily by perfused capillary density in the range of interest for insulin action, and by a combination of capillary density and perfusion heterogeneity in the range of interest for exercise capacity 6 Discussion In this manuscript, we derive a generalized theory of microvascular perfusion and solute flux in branching microvascular networks building from the single capillary analysis of Eugene Renkin (125) This theory enables a number of physiologic predictions, which we will discuss beginning with associations between physiologic parameters, progressing to the implications of complex phenotypes, and eventually moving on to therapies effective for acute treatment of microvascular perfusion defects. Throughout our discussion, we will use exercise capa city (or tissue oxygenation where appropriate) and insulin action as model solute exchange processes. We encourage the reader to follow along with another, self relevant exchange process (e.g. drug delivery) in mind. Where the predictions we draw this mode l have been tested, both the causes and the solutions of impaired microvascular flux can typically be predicted from our model. In addition, we will outline the limitations of our 6 .1 Physiolog ical associations We begin with predictions of physiologic associations. The first is that pulmonary oxygen uptake (a convection limited exchange process) should vary proportionally with bulk blood flow. Consistent with this prediction, cardiac output is the primary determinant of pulmonary oxygen uptake (126) Conversely, our model predicts that increasing bulk blood flow without correcting its distribution would do little to enhance insulin action in skeletal muscle (a diffusion limited process). This prediction is plausible animal models of insulin

PAGE 60

49 resista nce often display increased cardiac output (127) and regional blood flow diverges from regional glucose uptake (128) Further, our model predicts that capillary density (a determinant of diffusion c apacity) would be a major determinant of insulin action. The observation that capillary density and insulin sensitivity are associated in both human and animal models support this prediction (118; 122) Our model also predicts that capillary density would be related, albeit weakly, to skeletal muscle oxygen uptake (a diffusion/convection co limited process). This prediction is consistent with the observations that capillary density and VO 2 max are correlated in periphe ral arterial disease (129) and capillary density increases with aerobic exercise training (113) Finally, our model predicts that pulmonary capillary density (a determinant of diffusion capacity) would have little impact on pulmonary oxygen uptake (convection limited) except in extreme cases. It is unclear that this prediction has been directly tested, but it is telling that canonical descriptions of diseases that interfere with pulmonary oxygen diffusion do not involve capillary rarefaction (130) 6 .2 Disease phenotype predictions We will now shift to discussion of specific microvascular perfusion phenotypes. For this purpose we will employ three physiologic states: 1) T2DM, 2) sepsis, and 2) acute glycocalyx degradation. Both T2DM (Figure 12A) and sepsis (Figure 12 B) decrease perf used capillary density and increase perfusion heterogeneity. Our model therefore predicts insulin resistance in both disease states as a result of diffusion limitations. These same diffusion limitations would be expected to cause impaired exercise capacity in T2DM and impaired tissue oxygenation in sepsis. These predictions are consistent with literature reports in both disease states (47; 131 133) Assuming a causal role for reduced perfused capillary density

PAGE 61

50 and microvascular perf usion heterogeneity in limiting diffusion capacity, one would predict that exercise capacity and insulin sensitivity in T2DM are correlated, as well as tissue oxygenation and insulin sensitivity in sepsis. Exercise capacity and insulin sensitivity in T2DM are indeed correlated (5) The possibility of a correlatio n between insulin resistance and tissue hypoxia in sepsis has not been directly tested, but insulin resistance, tissue hypoxia, and impaired capillary perfusion are all thought to contribute to mortality (134) suggesting that this correlation would be observed if were to be tested. Mechanistically, our model predicts that the exerci se impairment in T2DM results from a combination of impaired muscle oxygenation and impaired clearance of metabolic wastes such as lactate and CO 2 Consistent with these hypotheses, muscle oxygenation during contractions is impaired in animal models of T2D M (97) and muscle pH decreases more robustly during exercise in human subjects with T2DM than in healthy controls (135) Similarly, our model predicts impaired clearance of metabolic wastes in sepsis. This prediction is consistent with decreased interstitial pH (both lactate and CO 2 are acidic) and interstitial hypercapnia in sepsis (133; 136) Likewise, our model predicts that the mechanism of insulin resistance in T2DM would be impaired diffusion of insulin. Consistent with this hypothesis, the limi ting stage in insulin action occurs at the extracellular (i.e. vascular interstitial) levels (112) and diet induced obesity (a precursor to T2DM) causes impaired insulin access to the interstitium (137) In sepsis, the infl uences of microvascular perfusion on solute diffusion are confounded by a robust increase in vascular permeability (138) which is not taken into account in our analysis. In hype rglycemia or enzymatic induced glycocalyx degradation (Figure 12 C), our model predicts both impaired exercise capacity and insulin resistance, again resulting from

PAGE 62

51 diffusion limitations secondary to reduced perfused capillary density and heterogeneous per fusion. The interaction between glycocalyx degradation and exercise capacity has not been tested. As relates to insulin sensitivity, however, our model predictions are correct (87; 139) The mechanisms that hav e been previously investigated for hyperglycemia induced insulin resistance often involve oxidative signaling processes rather than mass transport effects (140) Our model predicts that insulin resistance in these states is accounted for, at least in part, by impaired diffusion of glucose and insulin. These poss ibilities have not been tested. 6 .3 Recommendations for scientific practice We have observed and participated in a general trend in the medical literature that can be formulated as follows: 1) tissue level flux of a molecule of interest is impaired in a disease of interest, 2) investigators suspect a vascular/microvascular contribution to this defect, 3) subsequent studies determine that bulk blood flow does not fully account for the defect, 4) subsequent studies determine that tissue level production/demand does not fully accou nt for the defect, and then finally, 5) subsequent studies demonstrate a contribution from microvascular perfusion. Many scholarly publications spanning many years of investigation are often required before research of putative microvascular therapies can even begin. In the interest of helping the reader to avoid this trap, we ha ve created a flowchart (Figure 13 A) to guide investigations of this sort. The rationale underlying our flowchart questions stem from basic qualitative attributes of our theory, and are reflected in the interpretations of flowchart resul ts provided in Figure 13 B. Microvascular perfusion related transport defects of any etiology are predicted to have similar influences on the flux of molecules with similar equilib ration

PAGE 63

52 properties (see Figures 8 11 ). Thus, if a particular molecule is uniquely impacted by an observed flux defect, it is likely that the cause of the defect is specific to that molecule (e.g. tissue level demand, cell membrane transporters, etc.). If interstitial concent ration of a molecule of interest equilibrates slowly with the bloodstream (e.g. insulin), any microvascular contribution to reduced flux of this molecule is likely to involve diffusion limiting effects such as capillary dropout, capillary rarefaction, or p erfusion heterogeneity. For rapidly equilibrating molecules (e.g. oxygen), both diffusion and convection limiting effects may impact tissue level flux. Thus if larger, more slowly equilibrating molecules are not also affected, reduced bulk blood flow (a c onvection limiting effect) is a plausible culprit. Conversely, if tissue level flux is impaired across a wide range of equilibration rates, diffusion limiting perfusion defects such as perfusion heterogeneity and capillary dropout are plausible contributor s. Typical scientific practice in the past has been to bypass these considerations and assay bulk blood flow alone. Our theory predicts that this strategy will often generate confusion. Even for convection limited or convection/diffusion co limited transport processes such as tissue oxygenation, heterogeneous blood flow distribution and capillary dropout may recapitulate the effects of reduced blood flow.

PAGE 64

53

PAGE 65

54 Figure 13 : Flowchart of recommended steps to test for microvascular contributions to solute flux defects. ( A ) This flow chart is intended to help investigators determine an appropriate experimental design to assess the possibility that tissue level flux of a molecule of interest (X) is impaired due to a vascular/microvascular defect. Direct observation of capillar y perfusion is advisable in most cases. ( B ) Interpretation of flowchart destinations. In general, if a transport defect is solute specific, it is probably not caused by a microcirculatory issue. Alterations in blood flow, capillary density, and capillary b lood flow distribution are all probab le causes of transport defects. We will now provide a demonstration of these considerations using oxygen delivery to skeletal muscle in T2DM as an example. Like oxygen, blood glucose equilibrates near fully with the in terstitium during capillary transit (137) Glucose delivery to skeletal muscle during hyperinsulinemia is also impaired in T2DM (112) and the degree to which the capacity for glucose flux (i.e. insulin sensitivity) is reduc ed correlates to the degree to which the capacity for oxygen flux (i.e. VO 2 max) is reduced (5) These observations provide an a Delivery of insulin (a large, diffusion limited molecule) to the muscle interst itium is also impaired in diet induced obesity (a precursor to T2DM) (137) thus providing an answer of ring exercise reported in T2DM is often severe (97) while the insulin transport defect is more subtle. These observations provide assay capillary perfusion capillary perfusion in multiple models of T2DM reveal substantial capillary dropout and perfusion he terogeneity (23 28) and we now know that reversal of these defects can acutely improve skeletal muscle oxygenation in at least one animal model (25; 60) With the benefit of hindsight and a theoretical understanding of microvascular solute flux, many years of trial and error could have been avoided using the approach recommended here.

PAGE 66

55 6 .4 Implications for therapy We will now introduce interventions that have been shown to improve each of the microvascular perfusion parameters considered in our analysis. Acute reductions in perfused capi llary density are often caused by adhesive interactions such as leukocyte adhesion or microembolism (115; 141) Certain anticoagulant drugs would therefore be expected to improve capillary perfusion, and this effect has indeed been observed (108) Microvascular perfusion heterogeneity is less well understood, but likely involves impai red vessel function within arteriolar networks. This defect could plausibly be mitigated by antioxidant treatment. Antioxidant treatment can also reduce adhesive interactions (thus potentially improving perfused capillary density) and arteriolar perfusion is mechanically coupled to capillary perfusion (i.e. capillary plugging causes perfusion heterogeneity), so antioxidants would be expected to acutely improve both perfusion heterogeneity and perfused capillary density. Sure enough, acute antioxidant treatm ent does help to improve microvascular perfusion in certain contexts (108) Bulk blood flow, meanwhile, is controlled by vascular tone, and can be increased using vasodilators. For diffusion limited processes such as insulin action, our model predicts that the correct anticoagulants and antioxidants would be sufficient to acutely improve insulin sensitivity in insulin resistant individuals. As concerns antioxidants, this hypothesis has been repeatedly validated (142 144) The possibility that anticoagulant s improve insulin sensitivity has not been directly tested, but its plausibility is supported by associations of a variety of hemostatic parameters with insulin resistance (119 121) In sepsis, our simulations suggest that diffusion limitations caused by reduced perfused capillary density are a major cause of impaired substrate delivery and metabolite

PAGE 67

56 clearance. More importantly, impaired capillary perfusion is thought to be a major cause of organ failure and death in sepsis (84; 107; 108; 145) Certain anticoagulants (e.g. activated protein C) and certain antioxidants (e.g. Vitamin E) markedly improve microvascular perfusion in sepsis (108) True to form, act ivated protein C substantially improves patient outcomes (146) as do certain antioxidants (147) These reports are consistent with our model predictions. Although the interactions of these therapies with microvascular solute flux has not been widely investigated, perfused capillary density and interstitial hypercapnia (a marker of impaired CO 2 clearance) are inversely associated and interstitial hypercapnia can be acutely improved by capillary perfusion rescue (136) For oxygen delivery in T2DM (see Figure 12 A), our model predicts that a complex mixture of reduced bulk blood flow, reduced capillary density, a nd increased perfusion heterogeneity is required to explain the phenotype. Consequently, our model predicts that a drug combination consisting of anticoagulants, antioxidants, and vasodilators (or equivalent) would be necessary to reverse the oxygenation d efect. In the OZR model of T2DM, these predictions have been tested in the perfused hindlimb precisely this combination of drugs is effective in acutely restoring normal oxygen flux (25) Our model accurately predicts effective therapeutic targets for acute intervention in each case where model predictions have been explicitly tested, but except in the OZR model of T2DM, the proposed mechanisms have not yet been validated. Additionally, other compounds targeting microvascular perfusion may be more suitable for long term use. It is tempting to sp eculate that the cardiovascular protective effects of aspirin (148) for example, relate to its anticoagulant and antioxidant properties, or that the apparent ability of metformin

PAGE 68

57 to extend lifespan (149) relates to its microvascular perfusion benefits (123) Future studies will be required to i nvestigate these possibilities. 6 .5 Model limitations Our model accurately predicts many physiological associations, predicts which physiological parameters and solute exchange processes will be impaired in a variety of disease states, and predicts the therapies that will normalize these phenotyp es, yet there are key limitations. One limitation of our model is its simplicity. In reality, capillaries are heterogeneous (150) and we neglect this source of heterogeneity. Another consideration is that changes in endothelial or interstitial permeability were largely neglected. Our rationale for this simplification wa s twofold: 1) due to the selective permeability of the endothelium and interstitium, changes in permeability will be different for each molecule and cannot readily be generalized, and 2) capillary permeability is highly dynamic and does not lend itself to description by a single parameter. This limitation will thus require testing of specific molecules in different physiological settings to determine their individual properties. However, as evidenced by the fact that those model predictions which have previ ously been tested were correct, neglecting changes in capillary permeability does not compromise the general utility of the model. Based on the noted limitations, our model accurately reflects the contribution of microvascular perfusion changes to the dise ase phenotype, and that these contributions occur within the larger context of the local microenvironment. 6 .6 Conclusions The relationship between bulk blood flow and tissue level solute flux is dissociated except in cases of convection limited transpor t processes such as pulmonary oxygen uptake. Our simulations suggest that microvascular perfusion is adequately described by our model

PAGE 69

58 to explain the observed dissociation. In instances where relevant experimental data exist, our model successfully predic ts which solutes are most profoundly influenced by microvascular perfusion and which microvascular parameters are most strongly associated with solute flux. These results may account for literature reports in a variety of disease states and underscore the importance of a theoretical understanding of the microvascular parameters influencing microvascular exchange processes. Our model facilitates this theoretical understanding, and even allows some prediction of the therapies required to repair microvascular perfusion defects. Future work will be required both to further validate the predictive power of our model and to develop and test therapies targeting microvascular perfusion independently from bulk blood flow.

PAGE 70

59 CHAPTER IV IMPAIRED TISSUE OXYGENATION IN METABOLIC SYNDROME REQUIRES INCREASED MICROV ASCULAR PERFUSION HETEROGENEITY Co authors: Fan Wu, I. Mark Olfert, Christopher G. Ellis, Daniel Goldman, Jane E. B. Reusch, and Jefferson C. Frisbee 1. Preface As I developed a theoretical understanding of how heterogeneous blood flow distribution might impact skeletal muscle oxygenation in diabetes, I learned that the Jefferson Frisbee group had spent several years comprehensively characterizing perfusion heterogeneity and skeletal muscle oxygenation in diabeti c rats. This chapter is documents the ensu ing collaboration, which use s a more detailed version of the microvascular perfusion theory derived in the previous chapter to accurately predict the perfusion differences observed in diabetic rats. Not only do our simulations correctly predict increased perfusion heterogeneity in diabetic rats from observed limitations in oxygen diffusion they also correctly predict partial recovery of blood flow distribution with anti oxidant treatment, and the magnitude of the perfusion heterogeneity predicted is very close to that which is actually observed. This combined theoretical/empirical approach lends credibility to the possibility that heterogeneous distribution of microvascular blood flow contributes to impaired exercise capacity in diabetes. This chapter was published in the Journal of Cardiovascular Translational Research in 2017. 2. Abstrac t Metabolic syndrome (MS) in obese Zucker rats (OZR) is associated with impaired skeletal muscle performance and blunted hyperemia. Studies suggest that reduced O 2 diffusion capacity is required to explain compromised muscle performance, and that

PAGE 71

60 heteroge neous microvascular perfusion distribution is critical. We modeled tissue oxygenation during muscle contraction in control and OZR skeletal muscle using ph ysiologically realistic relationships. Using a network model of Krogh cylinders with increasing per fusion asymmetry and increased plasma skimming, we predict increased perfusion heterogeneity and decreased muscle oxygenation in OZR, with partial recovery following therapy. Notably, increasing O 2 delivery had less impact on VO 2 than equivalent decreases in O 2 delivery, providing a mechanism for previous empirical work associating perfusion heterogeneity and impaired O 2 extraction. We demonstrate that increased skeletal muscle perfusion asymmetry is a defining characteristic of MS, and must be considered to effectively model and understand blood tissue O 2 exchange in this model of human disease. 3. Introduction The growing incidence and prevalence of the metabolic syndrome and associated type 2 diabetes mellitus represent one of the greatest challenges to public health facing developed economies worldwide (151; 152) The metabol ic syndrome is generally defined as the combined presentation of obesity, impaired glycemic control, atherogeneic dyslipidemia, and moderate hypertension, with the additional contributing conditions of pro oxidant, thrombotic and inflammatory phenotypes (153 155) This clinical condition is present in an enormous and growing number of afflicted people worldwide, it has a powerful influence on reducing patient quality of life and life expectancy, and it causes enormous increases to the direct and indirect economic costs that must be borne by society as a result of its negative effects on health outcomes (151; 152) These considerations mandate detailed investigation into this mult i pathology state.

PAGE 72

61 In the general population (96) and in specific elements of the metabolic syndrome such as type 2 diabetes mellitus (T2DM; Ref. (7) ), aerobic exercise capacity as measured by maximal oxygen consumption (VO 2max ) is the strongest clinical predictor of mortality. Further, in both T2DM and metabolic syndrome, exercise capacity has been demonstrated to be reduced (156 158) While the precise mechan isms underlying impaired exercise capacity in these states remain unclear, there are distinctive mechanistic hints in the existing literature. In both human and animal studies of these states, hyperemic responses to exercising muscle are reduced with mani festation of the metabolic syndrome (20; 60) yet oxygen extractio n can not only fail to increase to compensate for the ischemia, but can be reduced itself (29; 60) Given that mitochondrial capacity exceeds the oxygen delivering capacity of the vasculature (159) the coincidence of impaired blood flow and impaired oxygen extraction cannot be explained through reduced mitochondrial demand or function. Understandi ng these apparent phenotypic contradictions requires detailed mechanistic study of O 2 transport and consumption comparing theoretical predictions to empirical data. The obese Zucker rat (OZR; fa/fa ) represents an excellent animal model of the metabolic sy ndrome with high utility in terms of comparing cardiovascular (dys)function to pathology in human subjects. OZR develop the metabolic syndrome due to chronic hyperphagia secondary to profound leptin resistance, and rapidly develop the systemic phenotypes listed above to comparable levels of severity with those identified in human subjects. Also similar to health outcomes in humans, OZR exhibiting the metabolic syndrome suffer from a progressive vasculopathy that ultimately develops into overt periphera l v ascular disease (PVD; Ref. (160) ), albeit one without the development of significant atherosclerotic lesions within macrovessels.

PAGE 73

62 Recent experimental studies have revealed significant impairments to the fatigue resistance (i.e., the ability to maintain contractile performance over time) of in situ skeletal muscle of OZR as co mpared to that in lean Zucker rats (LZR) (60; 161) The extent of these impairments has been generally correlated with reductions in bulk blood flow and hyperemic responses to th e elevated metabolic demand (60) However, the relationship between fatigue and reduced blood flow is not particularly robust and there is potential that factors beyond simple bulk perfusion could be responsible for the functional manifestation of PVD in metabolic syndrome. We have previously published observatio ns of reduced blood flow (Figure 1 4 A), reduced oxygen extraction (Figure 1 4 B), increased muscle fatigue associated with reduced diffusion capacity (Figure 1 4 C), and increased perfusion heterogeneity (Figure 1 4 E) in the OZR (25; 26; 60; 162) 1 4 D), defined as the fraction of total blood flow diverted to the higher flow daughter vessel at a microvascular bifurcation. Pharmacological intervention (e.g. TEMPOL; 1 O xyl 2,2,6,6 tetramethyl 4 hydroxypiperidine, a cell permeable superoxide dismutase mimetic which acts as a powerful antioxidant ) correcting this perf usion heterogeneity in the smallest arterioles (3a 5a) normalizes oxygen extraction, and intervention corre cting both bulk blood flow and perfusion heterogeneity (using a com 1 2 adrenoreceptor antagonist phentolamine and the TxA 2 /PGH 2 receptor blocker SQ 29548) fully restores oxygen uptake in the OZR to LZR levels (25; 26; 60) Therefore, it is of interest to be able to quantitatively test the hypothesis that oxygen supply from blood to skeletal muscle is a major limiting factor for muscle contractile performance under conditions of elevated metabolic demand within OZR, and, if o bserved, to define factors limiting oxygen delivery.

PAGE 74

63 Figure 1 4 : Previous studies establish impaired oxygen diffusion and microvascular perfusion heterogeneity in the OZR. ( A ) Femoral artery blood flow is significantly reduced in the OZR during 5 Hz contractions. This defect is not affected by treatment with TEMPOL. ( B ) Oxygen extraction is significantly reduced in the OZR during 5 Hz contractions. This defect is reversed by treatment with TEMPOL. ( C ) Oxygen diffusion capacity, a determinant of oxygen extraction, is a determinant of muscle fatigue for both LZR and OZR across multiple pharmacological treatments. ( D quantify microvascular perfusion heterogeneity. ( E ) Microvascular perfusion heterogeneity i n 4a and 5a arterials is significantly increased in the OZR. This defect is reversed by treatment with TEMPOL. Panels A C recreated with permissions from Frisbee et al Exp Physiol 2011 (60) Panels D E recreated with permissions from Frisbee et al Am J Physiol 2011 (25) In this work, a computational model of oxygen transport within skeletal muscle was constructed based on the classical Krogh cylinder type model (163) which has been further extended to account for capillary wall transport barrier, myoglobin fa cilitated diffusion, and Michaelis Menten kinetics of oxygen consumption (164) The constructed computational model was then applied to analyze oxygenation levels and metabolic rates using experimental data on bulk tissue blood flow and blood oxygen tensions measured from in situ blood

PAGE 75

64 perfused contracting skeletal muscle of LZR and OZR and reporte d in a previous publication (60) The predictions of this model were compared to previously published empirical observations to determine the effects of perfusion heterogeneity in an idealized microvascular network taking into account phase separation (i.e. erythrocytes from plas ma) eff ects at bifurcations (165) 4. Materials and Methods 4.1 Microvascular Network Model We have previously reported increased microvascular blood flow heterogeneity in OZR compared with that in L ZR (25; 26; 162) To test the hypothesis that the associated mismatch between oxygen sup ply and working capacity (60) is caused by the flow heterogeneity, an idealized microvas cular network (shown in Figure 15 A) was applied to examine impacts of flow heterogeneity on oxygen transport on the microvascular level. Empirical data from a previous publication (60) was used for this analysis. The idealized microvascular network employed in this analysis was based on (166) and assumed a fully symmetric vascular tree (i.e. all vessels of the same generation are the same length and diameter). The resulting vascular tree is consistent with the L system fractal formalism of Zamir (167) has recently been applied to assess the impact of the endothelial glycocalyx on microvascular blood flow distribution (89) and has been extensively validated as a reasonable approximation of microvascular branching architecture in vivo (168) The idealized microvascular network included in this analysis r the course of 15 vessel generations.

PAGE 76

65 At each microvascular bifurcation, the uneven distribution of blood flow into the daughter branches results in heterogeneity of downstream discharge hematocrits as described by the following equation (165) : (28 ) where logit function is defined as (2 9 ) The ratios of bulk blood volume flow and erythrocyte volume flow are defined as: (3 0 ) (31 ) respectively, with subscript 1 denoting daughter branch 1. During the simulations, the ratio of bulk blood volume flow ( B ) was varied from 0.5 to 0.7 (using notation consistent with our previously published perfusion heterogeneity analyses), and the discharge hematocrits in the daughter branches were calculated using Equations 2 9 31 for all vessel generations. In this work, arteriolar oxygen delivery is neglected, thus inlet PO 2 to each capillary is assumed to be equal to arterial PO 2 a reasonable assumption a reasonable assumption under conditions of high flow and high oxygen demand. Th en the oxygen transport was calculate d for each capillary using look up table of C TV values simulated from a range of representative discharge Hct and F values for LZR or OZR. Mitochondrial demand (V max ) in all cases was adjusted to recreate whole muscle VO 2 for

PAGE 77

66 LZR at each level of stimulati on without perfusion heterogeneity, consistent with observations of minimal perfusion heterogeneity in the LZR (25; 26; 162) 4.2 Krogh Cylinder Type Model In the Krogh cylinder type model, the tissue is assumed to consist of uniformly spaced cylinders, and each tissue cylinder is supplied by a capillary with perf using blood flow within its volume. The oxygen carried by blood supports local metabolic demand through diffusion across the capillary wall and into the skeletal muscle cell. The geometry of the Krogh cylinder type model is shown in Figure 15 B. Figure 15 : Visualizations of simulations performed ( A ) A schematic illustration of the idealized microvascular network used in the present study. Note that 4 vessel generations are visualized, whereas 15 vessel generations were used in the simulation. Please se e text for details. ( B ) A schematic illustration of the Krogh cylinder type model used for simulation of single capillary oxygen transport in this study. 4.3 Governing Equations The Krogh cylinder type model is described by two ordinary differential equat ions for radial and axial oxygen transport, respectively (164; 169)

PAGE 78

67 Oxygen Movement in the Radial Direction: transport, the radial oxygen transport is described by: (32 ) where D denotes the molecular diffusivity of oxygen, C denotes the modified tissue oxygen concentration accounting for myoglobin facilitated diffusion, and VO 2 denotes oxygen consumption rate in the tissue. The modified tissue oxygen concentration C is defined as: (33 ) where D Mb denotes the molecular diffusivity of myoglobin, C Mb denotes the tissue myoglobin concentration, V m denotes the molecular volume of oxygen, and S Mb denotes the myoglobin saturation level. The myoglobin saturation level is calculated from: (34 ) where C 50,Mb denotes half maximal myoglobin saturation O 2 concentration. The oxygen consumption rate (M) is assumed to follow Michaelis Menten kinetics: (35 ) where V max denotes the maximal mitochondrial oxygen consumption rate and K M denotes the Michaelis Menten constant of oxygen consumption in the skeletal muscle. Assuming that no oxygen exchange occurs across the outer boundary of the tissue cylinder (i.e. the outer boundary of the cylinder represents a local minimum of PO 2 ), a no flux boundary condition is obtained:

PAGE 79

68 (36 ) At the interface between capillary and tissue, an oxygen permeability barrier is assumed: (37 ) where D denotes the molecular diffusivity of oxygen in tissue, p w denotes cross capillary wall permeability of oxygen, and R 2 denotes the outer radius of the tissue cylinder. Oxygen Movement in the Axial Direction: Since the axial oxygen gradient is much less than the radial oxygen gradient, the axial oxygen transport is assumed to be dominated by convection and is described as follows: (38 ) where C T denotes the total oxygen concentration in blood, VO 2avg denotes the oxygen consumption rate averaged along the radial direction, F denotes the flow volumetric rate normalized to tissue mass, tissue denotes the tissue mass density, L denotes the capillary length. The average oxygen consumption rate (VO 2avg ) is defined as: (39 ) The total oxygen concentration (C T ) is defined as: ( 40 )

PAGE 80

69 where C f denotes free oxygen concentration in plasma, C RBC denotes free oxygen concentration in red blood cells (RBCs), Hct denotes discharge hematocrit, C Hb denotes concentrations of oxygen binding sites in RBCs, and S Hb denotes hemoglobin oxygen saturation level. The hemoglobin saturation level is calculated from: (41 ) where P 50,Hb denotes half maximal hemoglobin saturation O 2 tension and n H denotes the Hill coefficient for hemoglobin oxygen binding. The free oxygen concentration in plasma and red blood cells (RBCs) can be related to oxygen tension (PO 2 ) as: (42 ) (43 ) where and are oxygen solubility in plasma and RBCs, respectively. 4.4 Look Up Tables of Venous and Muscle Oxygenation The Krogh cylinder type single capillary model discussed in the previous section was applied recursively at each level of muscle stimulation (changes mitochondrial demand and blood flow parameters), for each animal model (LZR vs OZR, changes capillary dens ity and blood flow parameters), and for blood flow values ranging from zero to 200x average. These values were then compiled into look up tables (LUTs) of venous and mu scle oxygenation (see Figures 1 6 and 1 7 ). These LUTs were combined with the network perf usion model discussed above to rapidly obtain an array of single capillary values for venous oxygen

PAGE 81

70 content and mean muscle oxygenation. The oxygen concentration of the mixed venous effluent was then calculated as: ( 44 ) Finally, mean muscle PmO 2 was calculated from the mean of all single capillary PmO 2 values. Figure 16: Heatmaps visualizing look up tables (LUTs) of venous oxygenation as a function of capillary hematocrit and relative blood flow. ( A ) LUT of simulated venous oxygenation in the LZR at rest. ( B ) LUT of simulated venous oxygenation in the LZR during 5 Hz contractions. Note that venous oxygenation is substantially decreased (i.e. oxygen extraction is increased) relative to rest given simil ar flow and hematocrit. ( C ) LUT of simulated venous oxygenation in the OZR at rest. ( D ) LUT of simulated venous oxygenation in the OZR during 5 Hz contractions. Note that venous oxygenation is substantially less than at rest and marginally less than in the OZR given similar flow and hematocrit.

PAGE 82

71 Figure 17: Heatmaps visualizing look up tables (LUTs) of skeletal muscle oxygenation as a function of capillary hematocrit and relative blood flow. ( A ) LUT of simulated muscle oxygenation in the LZR at rest. ( B ) LUT of simulated muscle oxygenation in the LZR during 5 Hz contractions. Note that muscle oxygenation is substantially decreased relative to rest given similar flow and hematocrit. ( C ) LUT of simulated muscle oxygenation in the OZR at rest. ( D ) LUT of sim ulated muscle oxygenation in the OZR during 5 Hz contractions. Note that muscle oxygenation is substantially less than at rest and marginally less than in the OZR given similar flow and hematocrit. 4.5 Sensitivity Analysis Methods Microvascular perfusion heterogeneity results from a complex combination of microfluidic mechanical changes (e.g. glycocalyx degradation, capillary plugging by adherent leukocytes) and changes in vessel tone/diameter. Vessel diameter influences distribution of RBCs independently of f low distribution (165) and it is plausible that these effects, which were not considered in our muscle oxygenation simulations, are significant in vivo. To assess this possibility, we per formed a sensitivity analysis on the simulated effects

PAGE 83

72 of diameter randomization independent of perfusion heterogeneity. The maximum degree of in downstream vessel dia meters, assuming laminar flow. Accordingly, we simulated CvO 2 PmO 2 and VO 2 with up to 10% randomization of vessel diameter in each vessel segment. 100 simulations at each degree of diameter randomization were performed, and the coefficient of variance (C V = SD/mean) of simulation results was used as a metric of the effect size of diameter randomization. 4.6 Model Parameters The parameters used in this model are taken from either the present experimental results or from validated literature values and are summarized in Table 1

PAGE 84

73 Table 1 : Summary of model parameters used in the present study. Parameter Description Value Reference Blood parameters P in Oxygen tension in inlet capillary flow 91 to 95 mmHg Experimental data P out Oxygen tension in outlet capillary flow Model outputs Flow (F) Blood flow per tissue mass 7.1e 3 to 3.55e 2 mL sec 1 (mL tissue) 1 Experimental data Hct Hematocrit 0.45 Default value n H Hill coefficient for Hb oxygen binding 2.7 (McGuire & Secomb 2004) P O2,Hb Half maximal Hb saturation O 2 tension 38.8 mmHg (Kiwull Schone et al. 1987) S Hb Hemoglobin saturation level plasma Oxygen solubility in plasma 1.3e 6 M mmHg 1 (Beard et al. 2003) RBC Oxygen solubility in red blood cells 1.53e 6 M mmHg 1 (Beard et al. 2003) p w Permeability for O 2 across the capillary wall 250 m sec 1 (Beard et al. 2003) Tissue parameters D Oxygen diffusivity in tissue 2410 m 2 /sec (Goldman & Popel 1999) D Mb Myoglobin diffusivity in tissue 17.3 m 2 /sec (Goldman & Popel 1999) C 50,Mb Half maximal Mb saturation O 2 concentration 9.22e 6 M (Goldman & Popel 1999) tissue Oxygen solubility in muscle tissue 1.74e 6 M mmHg 1 (Beard et al. 2003) K M Michaelis Menten constant for O 2 consumption 1.74e 6 M (McGuire & Secomb 2004) V max Mitochondrial oxygen demand (rest) 0.039 ml/g/min Model optimization Mitochondrial oxygen demand (5Hz) 0.185 ml/g/min Model optimization Geometric parameters R 1 Capillary radius 2.98 m (LZR) 3.07 m (OZR) (Skalak et al. 1986) R 2 Tissue cylinder outer radius 19.9 m (LZR) 23.0 m (OZR) Experimental data L Capillary length 1012 m (Honig et al. 1977) The measured blood flow (F) and arterial oxygen tensions (P a O 2 ) were used as inputs to the computational model to simulate venous oxygen tensions (P v O 2 ). The model

PAGE 85

74 parameters were set up using values listed in Table 1 except that the oxygen consumption demand (VO 2 ) at rest and at 5Hz were g enerated via model optimization by minimized differences between model predicted and experimentally measured P v O 2 The Krogh cy linder type model was implemented on a standard desktop personal computer using MATLAB (Version R2013b, MathWorks, Natick, MA). The optimization was performed using the fmincon function included in MATLAB Optimization Toolbox. 5. Results 5.1 Effects of perfusion heterogeneity on oxygen transport in microvascular networks An idealized microvascular network (Figure 15 A) was used to examine the effects of microvascular flow heterogeneity on oxygen transport. The simulations were conducted by assuming tha t the mitochondrial demand ( V max ) in LZR and OZR are equal, and sufficient to produce measured values of VO 2 with homogeneous perfusion in the LZR. Empirical data from a previous publication (60) was used for this analysis. The results of this simulation are shown in Figure 18 Model simulations of PvO 2 using varying flow distributions reveal that simulated PvO 2 matches measured PvO 2 at values of = 0.6164 for untreated OZR, = 0.5562 for TEMPOL treated OZR, and at = 0 .5121 for LZR at 5 HZ (Figure 18 A), indicating heterogeneous perfusion in the OZR that can be partially corrected by TEMPOL treatment. Using the model predicted values of = 0.5121, 0.5562 and 0.6164 for LZR, TEMPOL treated OZR, and untreated OZR, respectively, our model predicts a mean muscle PmO 2 of 30.4 mmHg in the LZR, 18.9 mmHg in untreated OZR, and 23.48 mmH g in TEMPOL treated OZR (Figure 18 B) at 5 Hz, indicating muscle hypoxia during exercise in the OZR that can be partially corrected by TEMPOL treatment. Our simulations also predict that the contributions of perfusion heterogene ity account for more than half of the total

PAGE 86

75 decrease in VO 2 in the OZR relative to LZR ( Figure 18 C), a majority of which is due to perfusion heterogeneity in the small 3a 5a arterioles where TEMPOL most profoundly influen ces perfusion heterogeneity (25) Figure 18 : Differences in venous oxygenation, muscle oxygenation, and perfusion heterogeneity can be predicted using a simulation of microvascular blood flow and oxygen transport. ( A ) Venous oxygenation increases with increasing perfusion heterogeneity in both the L ZR and OZR. The intersection between simulated venous oxygenation (light, dashed lines) and empirically measured venous oxygenation (bold lines) can be used to predict the degree of perfusion heterogeneity. Our model correctly predicts greater perfusion he terogeneity in the OZR and partial correction of perfusion heterogeneity with TEMPOL treatment. ( B ) The intersection between simulated muscle oxygenation (light, dashed lines) and model predicted perfusion heterogeneity (bold lines) can be used to predict muscle oxygenation. Our model correctly predicts reduced muscle oxygenation in the OZR, and also predicts partial correction of muscle oxygenation with TEMPOL treatment. ( C ) The intersections of simulated oxygen uptake (light, dashed lines) with empiricall y measured oxygen uptake (bold lines) and with model predicted oxygen uptake without perfusion heterogeneity in the OZR (light, dotted line) can be used to predict the relative contributions of various oxygen transport parameters to the observed oxygen upt ake defect. Our model predicts that perfusion heterogeneity in small (3a 5a) arterioles plays a major role in reduced oxygen uptake in the OZR.

PAGE 87

76 5.2 Re sults from Sensitivity Analysis The results of our sensitivit y analysis are shown in Figure 19 Results were largely consistent among the results for CvO 2 (Panel A), PmO 2 (Panel B) and VO 2 (Panel C). In all cases, the CV of the oxygenation metric of interest increases quadratically with increasing diameter randomization. The influence of diameter randomizati on independent of perfusion heterogeneity is several orders of magnitude smaller than that of perfusion heterogeneity independent of diameter randomization. These results suggest than our muscle oxygenation analysis with varying degrees of perfusion hetero geneity was negligibly influenced by our choice to neglect possible effects from vessel tone independent of blood flow and its distribution.

PAGE 88

77 Figure 19 : Sensitivity analysis reveals that the effects of varying vessel diameters are negligible relative to the effects of varying flow distributions. ( A ) Variance in simulated CvO 2 increases quadratically with increasing variance of vessel diameters. At the level of diameter variance (10%) required to explain the maximum degree of perfusion heterogeneity flow heterogeneity are still several orders of magnitude less than the effects of flow heterogeneity without diameter variance. ( B ) Variance in simulated PmO 2 increases quadratically with increasing variance of vessel diameters. This effect is several ord ers of magnitude smaller than the effects of flow heterogeneity. ( C ) Variance in simulated VO 2 increases quadratically with increasing variance of vessel diameters. This effect is several orders of magnitude smaller than the effects of flow heterogeneity. 5.3 Mechanisms underlying the effects of perfusion heterogeneity To determine the mechanism(s) by which microvascular perfusion heterogeneity would theoretically reduce oxygen diffusion capacity, we compared the effects of equivalent increases or decreases in flow or hematocrit on oxygen uptake in single capillaries (Figure 20), using LZR 5Hz values (60) in a single cylinder version of our model for illustration

PAGE 89

78 purposes. Increasing capillary blood flow increases oxygen uptake, but not to the same degree that an equivalent decrease in capillary blood flow decreases oxygen uptake. The mean oxygen uptake of two capillaries with a fixed total blood flow supply thus decreases with increasing flow disparity betwe en the two capillaries (Figure 20 A). Similarly, increasing capillary hematocrit increases oxygen uptake, but not to the same degree tha t an equivalent decrease in capillary hematocrit decreases oxygen uptake. The mean oxygen uptake of two capillaries with a fixed total RBC flux thus decreases with increasing hematocrit disparity betwe en the two capillaries (Figure 20 B). Due to the plasma skimming effect at microvascul ar bifurcations (165) capillary flow and discharge hematocrit are pos itively correlated (see Figure 20 C, uses simulated = 0.7 distribution for illustration purposes). The effects of flow and hematocrit disparities thus synergize to produce the cumulative effects of perfusion heterogeneity.

PAGE 90

79 Figure 20 : Analysis of perfusion heterogeneity effects at a single bifurcation reveals the mechanisms by which perfusion heterogeneity interferes with oxygen transport. ( A ) Increasing flow through one capillary while subtracting the same amount of flow from another capillary results in a greater decrease in oxygen uptake by the under perfused ca pillary than the corresponding increase in oxygen uptake by the over perfused capillary. This effect occurs independently of hematocrit effects. ( B ) Increasing hematocrit in one capillary while subtracting the same number of RBCs from another capillary res ults in a greater decrease in oxygen uptake by the low hematocrit capillary than the corresponding increase in oxygen uptake by the high hematocrit capillary. This effect occurs independently of flow effects. ( C ) Capillary flow and hematocrit are correlate d at the single capillary level, and flow heterogeneity results in hematocrit heterogeneity. These consequences of the plasma skimming effect cause the effects of flow heterogeneity and hematocrit heterogeneity to synergize in reducing oxygen uptake under conditions of microvascular perfusion heterogeneity. 6. Discussion It has been clearly demonstrated by our group (60; 161) and by others (170) that the development of the metabolic syndrome is associated with alterations to skeletal mu scle arterial and arteriolar function. These changes are correlated with impaired hyperemic

PAGE 91

80 responses to elevated metabolic demand, constrained oxygen uptake (VO 2 ) across skeletal muscle and impaired fatigue resistance of in situ blood perfused skeletal mu scle. However, beyond a restricted number of studies, there has been much less effort dedicated to determining how these key parameters relate to one another. The purpose of this study was to take the first steps into developing a computational model for microvascular perfusion under the conditions of the metabolic syndrome and to model, based on established literature values and original data from the OZR model of the metabolic syndrome, how microvascular function might be altered based on the underlying principles of mass transport and exchange. This is intended to be a first step in a larger effort to develop predictive models and biosimulations of sufficient veracity as to be useful for hypothesis development and informing experimental design. These m odels will be refined with the ultimate goal of application to clinical settings to inform a better understanding of conditions affecting microvascular function, cardiorespiratory fitness and response to clinical intervention. A major initial interpretatio n of the results from the present simulations is that the striking similarity deter mined in the heatmaps (Figures 16 and 17 ) for oxygen transport and exchange between LZR and OZR is conceptually in contrast to all of the functional (i.e., muscle performanc e) data collected to date. If oxygen transport is minimally affected in between LZR and OZR using flow characteristics, then differences in perfusion and/or distribution are required to explain the discrepancy in outcomes. This highlights the importance of introducing perfusion heterogeneity into our simulations of microvascular perfusion in the OZR manifesting metabolic syndrome in order to more accurately understanding the system function in this challenged state.

PAGE 92

81 Our present simulations show that per fusion heterogeneity is mathematically sufficient to account for our previously published observations of increased venous oxygenatio n (Figure 18 A) in the OZR, and sufficient to explain why venous oxygenation is partially co rrected by TEMPOL treatment (60) Perfusion heterogeneity is also mathematically sufficient to account for literature rep or ts of muscle hypoxia (Figure 18 B) in the OZR (103) Our model predicts that this muscle hypoxia is partially relieved by TEMPOL treatment. Although this prediction has not yet been directly tested, it does pr ovide a plausible explanation for partial recovery of oxygen u ptake with TEMPOL treatment (60) Finally, our model predicts that perfusion heterogeneity accounts for a majorit y of the oxygen uptake (Figure 18 C) defect in during 5 Hz contractions the OZR, consistent with experimental recovery of a majority of the oxygen uptake defect during 5 Hz cont ractions when treated with a drug cocktail that fully reverses microvascular perfusion heterogeneity (25; 26; 60) The degree of perfusion heterogeneity required to mathematically account for oxygen transport differences between OZR and LZR in our simulati on is not only qualitatively, but also quantitatively similar to experimentally determined val and OZR (25; 26) Finally, our simulations provide a mechanism by which perfusion heterogeneity results in imp aired oxygen transport (Figure 20 ). Namely, an individual of similar m agnitude. Similar to ventilation perfusion mismatch in the lung, over perfused vessels cannot fully compensate for under perfused vessels. Our simulations show that microvascular perfusion heterogeneity is mathematically sufficient to account for oxygen tr ansport defects in the OZR not caused by reduced bulk blood flow. We have also previously published experiments showing that perfusion

PAGE 93

82 heterogeneity is empirically necessary to account for the se same defects (25; 26; 162) Given the theoretical sufficiency and empirical necessity of perfusion heterogeneity to account for the OZR phenotype, i t is very likely that 1) our model describes tissue level oxygen transport to a reasonable degree of accuracy, and 2) perfusion heterogeneity is a key pathological feature of the metabolic syndrome. In contrast with this manuscript, the bulk of the scient ific literature discussing exercise capacity in the metabolic syndrome and T2DM focuses on the total mount blood flow delivered to exercising muscle. Given the previously demonstrated impairments to vascular/arteriolar function in the skeletal muscle of OZ R (161) the progressive rarefaction of t he microvascular networks (171) and the reduction to bulk perfusion to the skeletal muscle a cross metabolic intensities (60) impaired oxygen transport is not particu larly surprising. However, combining our present model results with our previous findings of heterogeneous microvascular blood flow distribution using both direct microvas cular visualization (25; 26; 63) and tracer washout kineti cs (162) an alternate interpretation of literature data suggests that increased muscle fatigue in OZR may also reflect an increasingly heterogeneous distribution of p erfusion within microvascular networks. Microvascular perfusion heterogeneity contributes to impaired muscle function by causing reduced O 2 extraction and an elevated PvO 2 Our simulations and previously published experiments support the hypothesis that perfusion heterogeneity, along with the resulting phase sep aration of plasma from RBCs (165) results in failure to compensate for reduced blood flow in OZR with increased oxygen extraction. Collectively, these echoes of previously published data in our simulations lend credibility to the notion that microvascular perfusion heterogeneity is necessary and

PAGE 94

83 sufficient to account for experimentally observed barriers to oxygen t ransport in OZR. Although there are assumptions within our model that require further refinement (e.g. neglecting axial O 2 diffusion, neglecting heterogeneous non Kroghian capillary morphology, neglecting microscopic variances in blood O 2 content, etc.), t he broad consistency between our computational results, our previously published experimental results, and the broader published literature suggests that our simulations were sufficient to predict observed oxygen transport phenotypes from first principles. These findings suggest that clinical strategies aiming to improve bulk blood flow without also improving perfusion may have limited impact in the metabolic syndrome. Microvascular perfusion heterogeneity represents an unaddressed therapeutic target for im proving tissue oxygenation and muscle function.

PAGE 95

84 CHAPTER V FULLY AUTOMATED SOFTWARE FOR COMPREHENSIVE QUANTIFICATION OF CA PILLARY STRUCTURE AND PERFUSION Co authors: Ian M. Williams, Sara E. Hull, David H. Wasserman, and Jane E. B. Reusch 1. Preface In the obese Zucker rat model of T2DM, microvascular blood flow and its distribution have been adequately characterized to enable detailed engineering analysis quantitatively relating microvascular perfusion to tissue oxygenation. This perfusion characteri zation required many years and spanned several scientific publications. In this chapter, I show that the perfusion heterogeneity observed in the obese Zucker rat is also observed in high fat fed mice (another animal model of insulin resistance ), and allude to literature reports from nearly twenty years ago documenting very similar perfusion phenotypes is the GK rat (another T2DM model) and streptozotocin treated rats (an animal model of type 1 diabetes). Furthermore, the technique developed in this study re duces the time required to achieve detailed characterization of microvascular perfusion heterogeneity from several weeks to several hours. This technique, combined with the microvascular perfusi on theory described in Chapter III holds promise for enabling detailed mechanistic analyses such as that described in C hapter IV for virtually any organ in virtually any disease state. Future experiments will be required to more completely validate this software technique for use with a variety of microscopy techniq ues in a variety of organs. 2. Abstract Changes in microvascular perfusion have been shown to contribute to a variety of disease states, and recent advances in intravital microscopy hold promise for more detailed

PAGE 96

85 study of microvascular perfusion. However, current methods for quantifying microvascular perfusion parameters are largely inconsistent and often provide incomplete characterization of microvascular perfusion phenotypes. In this manuscript, we report an automated (i.e. no user inputs required) softw are system for simultaneous measurement of plasma perfused capillary density, RBC perfused capillary density, % absence of RBC flux, mean capillary flow velocity, and heterogeneity of capillary blood flow distribution from intravital microscopy videos. Thi s software was tested on videos of the gastrocnemius muscle microcirculation in high fat fed (HFD, n=4, model of insulin resistance) and chow fed (n=6, insulin sensitive control) mice during hyperinsulinemia. Software measurements of plasma perfused capill ary density, % absence of RBC flux, and RBC perfused capillary density were compared to by hand measures of these same parameters. In each case, no significant difference was observed between software and by hand measurements in either HFD or chow fed mice (all p=NS). Software and by hand measures both revealed significantly decreased plasma perfused capillary density, significantly increased % absence of RBC flux, and significantly decreased RBC perfused capillary density in HFD mice (all p<0.01). These re sults are consistent with previous literature reports of reduced capillary density and loss of RBC flux in T2DM (an insulin resistant state). In addition, this software revealed significantly increased heterogeneity of capillary flow velocity (as assessed by coefficient of variance) in HFD mice (p=<0.001), consistent with previous reports of increased microvascular perfusion heterogeneity in T2DM. Collectively, these results demonstrate that capillary perfusion can be quantified in an accurate, high through put, user independent, and comprehensive fashion using software designed specifically for this purpose.

PAGE 97

86 3. Introduction Delivery of nutrients, oxygen, medications, or any other molecule of interest to peripheral tissues requires trans capillary efflux. Capillary structure and perfusion are major determinants of this efflux. For example, variations in skeletal muscle capilla ry density have been shown to associate with both changes in glucose transport (i.e. insulin sensitivity) and changes in oxygen transport (i.e. aerobic exercise capacity) (113; 118; 122; 129) Reduced blood flow also has well documented detrimental effects on end organ function in a variety of contexts (172 174) Distribution of blood flow within microvascular networks is critical to maintaining normal trans capillary exchange beyond bulk blood flow and capillary density. Heterogeneous spatial distribution of microvascular blood flow has been shown to contribute to impaired tissue oxygenation in the metabolic syndrome and sepsis (23 26; 107; 108; 117; 136; 175) and it is also thought to contribute to impaired tissue oxygenation in ischemic heart disease and cerebrovascular disease (64; 65) Collectively, perfused capillary density, bulk blood flow, and distribution of blood flow within capillary networks are each critical parameters in the physiology and pathophysiology of many disease states. Despite hundreds of papers reporting microvascular perfusion measurements, there is a paucity of perfusion data suitable for engineering a nalysis estimating the contributions of microvascular perfusion to trans capillary exchange of physiologically important molecules. To assess the extent of this issue in the microcirculation literature, we reviewed the first 50 primary research articles re 25 2017, results sorted by relevance). We recorded the total number of articles quantifying microvessel density (or an equivalent measure such as blood volume measured through MRI or contrast enhanced ultrasound),

PAGE 98

87 flow velocity and/or regional bulk blood flow, and spatial heterogeneity of perfusion. We also took note of whether these measurements were taken by hand (e.g. counting capillaries in a histolo gical cross section) or using software (e.g. Doppler flow measurements). The results of this literature an alysis are summarized in Table 2 Table 2: Summary of microvascular perfusion measures in the primary literature Of 100 manuscripts reviewed, the number reporting By Hand Measurement Software Measurement Microvessel Density 39 15 Flow Velocity and/or Bulk Blood Flow 19 29 Perfusion Heterogeneity 1 4 All of the Above 3 Two of the Above 28 One of the Above 38 None of the Above 31 In this sample of 100 primary research articles, a majority (69%) quantified no more than one of the three key perfusion parameters, and a sizable minority (31%) did not quantify microvascular perfusion. Examples of studies with no quantification of key pe rfusion parameters (31% of the total) include methods such as qualitative scoring of capillary perfusion or simulation rather than measurement of perfusion. Furthermore, the vast majority of studies reviewed (97%) did not report quantitation of all three p erfusion parameters. This is a key limitation, as mass transport analysis reveals that all three parameters are needed to fully account for the contributions of microvascular perfusion to transcapillary solute flux (66) The most commonly neglected parameter is heterogeneity of microvascular perfusion,

PAGE 99

88 which can be locally quantified by recording flow velocity at the capillary level (including a majority of the 29 manuscripts we found reporting software measurements of flow velocity). Flow heterogeneity has been demonstrated both theoretically (64 66) and empirically (23; 25; 26; 108; 117; 136; 175) to exert substantial independent influence on tissue oxygenation. The limitations of previous mic rovascular perfusion data outlined above are primarily a function of available methods. Measurements taken by hand remain more common than software assisted measurements. By hand measurements of capillary perfusion are often very laborious, thus limiting s ample size and measured endpoints. Aside from by hand measurements, several techniques for software assisted perfusion quantification have previously been published (176 178) The most impactful microvascular perfusion measurement software to date has been the Cap Image software produced by Zeintl et al (176) w hich accounts for a majority of the manuscripts in our survey of the literature that used software for quantification of capillary perfusion. Cap Image has been valuable in advancing the field. However, it does have the limitation that it requires user inp uts to select vessels for measurement. As a result, use of this software can be very time consuming (177; 179) and the potential exists for human error or bias. There is a need for software that permits standardized quantification of microvascular perfusion that is less labor intensive and subject to investigator bias. Even with meth ods that have the potential to avoid these pitfalls, perfusion is rarely characterized comprehensively enough to make valid predictions about the relationships between microvascular perfusion and solute flux. We therefore set out to create a fully automat ed software system for simultaneous quantification of perfused capillary density, capillary flow velocity, and microvascular perfusion heterogeneity. Our metrics for defining success were 1) software measurements

PAGE 100

89 must agree with by hand measurements when b y hand measurements are available, 2) software measurements must agree with literature results in an established phenotype, 3) software measurements must be taken without any user inputs (to reduce the possibility of biased results), and 4) software measur ements must be comprehensive enough to plausibly enable future engineering analyses (e.g. (66) ) relating microvascular perfusion parameters to transport of physiological molecules (e.g. tissue oxygenation). Methods and an example application for a software technique meeting all of these criteria will be described. Any readers who may be interested in using this technique are encouraged to contact us direc tly. 4 Methods 4 .1 Animals Following a 3 w ee k weaning period, of either gender C57Bl6/J mice fed standard chow were group housed in micro isolator cages. Mice were housed at 23C on a 0700 1900 light cycle, and allowed free access to H2O and food. Mice were fed a standard chow or high fat. The high fat diet contains 60, 24, and 16% of kcal/g as fat, carbohydrate, and protein respectively. J ugular catheterization (Ayala et al. Diabetes 2006) was performed at 12 14 w ee ks of age. Experiments were carried o ut upon restoration of body weight (6 7 days postoperative). All procedures were approved by the Vanderbilt Institutional Animal Care and Use Committee. 4 .2 Intravital microscopy We used a Zeiss LSM 5 Live slit confocal system for high resolution imaging of capillary perfusion in the gastrocnemius ~ 1 00 m beneath the tissue surface at the high speeds necessary to res olve blood flow dynamics. Mice we re implanted with a jugular vein catheter prior to the procedure. The venou s cathete r was then used for infusion of i nsulin (4

PAGE 101

90 U/kg) L saline immediately prior to microscopic observation Prior to each experiment, mice were fasted for 6 h and then injected with RBCs pre labeled with the fluorescent dye sulfa rhodamine B and fluorescein conju gated 70 kDa dextran as a plasma contrast agent Plasma fluorescence, and not RBC fluorescence was used for all subsequent software measurements (see Section 2.3). For all videos used in this analysis, a 512x512 pixel field Pinhole size was adjusted to ensure resolution 4 .3 Software measurements of ca pillary structure and perfusion The first 100 frames of each intravital microscopy video were exported in AVI format and imported into Matlab 2013 (Mathworks, Inc). In raw microscope videos, motion artifacts due to breathing are readily apparent. Our software stabilizes imported microscope videos by comparing successive frames using 2 dimensional c ross correlation and shifting each frame to maximize image agreement between successive frames. This routine was found to satisfactorily remove motion artifacts for all videos tested (see supplemental data). Image processing steps used to identify capillary segments for software flow measurement are outlined in Figure 2 1. First, the time average of plasma fluorescence 2 1A). Otsu thresholding (180) is then applied to an image array produced by difference between local pixel intensity and mean pixel intensity in a 10 pixel neighborhood. This scheme serves to detect in focus, plasma perfused capillaries (Figure 2 1B). Matlab image processing toolbox functions are then used to identify branch points (e.g. capillary bifurcations) in the vessel map (Figure 2 1C). Regions within a 10 pixel neighborhood of branch points are then removed to leave only straight vessel segments with uniform flow velocity (Figure 2 1D). Very short or very wide vessel

PAGE 102

91 segments (in which RBCs may not be pres ent for two successive frames) were removed by removing data from vessel segments for which aspect ratio < 5. Figure 2 1: Outline of image processing steps used to identify capillary segments for software flow measurement. ( A ) Time averaged plasma fluorescence is used as the starting image. B ) An adaptive thresholding algorithm is used to identify in focus vessels. ( C ) Branch points (e.g. capillary bifurcations and intersections) are identified and removed to ensure a uniform velocity within each capillary segment. ( D ) Remaining vessel segments are used for flow velocity measurement. The routine used to measure flow velocity is outlined in Figure 2 2. Pixel intensities along the center line path of each capillary segment (Figure 2 2A) are measured. The deviation from time averaged pixel intensity at each pixel position along the vessel path is compared in successive frames (Figure 2 2B). Cross correlation of consecutive frame intensity deviation profil es is then used to determine the distance that image features moved between successive frames (Figure 22 C). This procedure is repeated for each capillary segment in each combination of successive frames and the time averaged result (n=100

PAGE 103

92 frames) is used t frame rate and pixel size. Figure 2 2: Outline of cross correlation routine used to measure RBC flow velocity. ( A ) The fluorescence intensity profile of the center line of each capillary segment is used to track ( B ) Comparison of center line fluorescence intensity profiles in successive frames reveals image features shifted longitudinally. ( C ) Cross correlation is used to detect the longitudinal movement of image features between successive frames. The offset of image features between successive frames is proportional to capillary flow velocity in pixels/frame. In addition to mean capillary fl ow velocity, heterogeneity of capillary flow velocity (assessed by coefficient of variance, CV = Std/Mean) and loss of RBC flux (% of capillary segments with mean flow velocity = 0 during 3 5s of imaging) are recorded by the software for each video. Mean c apillary plasma perfused diameter ( ) is calculated from total vessel map edge length ( Lt ), total vessel map projected area ( At ), and total number of vessel segments ( #t ) using Equations 45 47 This calculation assumes long, thin vessel segments of approximately cylindrical shape

PAGE 104

93 (45) (46) (47) Capillary density ( CD ) is then calculated from mean capillary diame ter ( ), projected area of the vascular network in the field of view ( Av ), and total projected area of the field of view ( At ) using equations 4 8 49 This calculation assumes that capillary spacing perpendicular to the image plane is similar to capillary spacing within the image plane. (48) (49) 4 .4 By hand measurements of capillary structure and perfusion Plasma perfused capillary density, loss of RBC flux, and RBC perfused capillary density were also measured by hand using the process outlined in Figure 2 3. Although computerized analysis could readily detect motion of image features between successive fram es (see Figure 2 2B), the video was not visually clear enough in many vessels to permit accurate by hand measurements of capillary flow velocity. However, presence/absence of RBC flux was usually visibly apparent even when consistent by hand velocity measur ements were not possible. We therefore chose to validate our software measurements using metrics of capillary density and the presence or absence of RBC flux in each vessel. For each microscope video, the first frame was exported as an image and software analyzed capillary segments were outlined for technician reference (Figure 2 3A). A laboratory technician (SEH) performed by hand measurements (illustrated in Figure 2 3B)

PAGE 105

94 while blinded to study design and software results in order to prevent bias towards a greement with software in sensitive by hand measures. To assess plasma perfused capillary density, the total number of vessel crossings (including vessels not analyzed by our program) in each of 3 image cross sections was recorded. The average number of ve ssel crossings in these 3 cross sections was then corrected for vessel orientation (acute angles between the image cross section and the typical capillary artificially reduce the number of crossings) and used to calculate capillary density. In addition, ea ch outlined vessel was marked as either obviously flowing (F, including intermittent flow), obviously not flowing (NF), or not obviously flowing (?). Hand labelled images produced in this analysis are included in the supplemental data. The total number of non flowing vessels (NF+?) in each image was then used to calculate % vessels lacking RBC flux and RBC perfused capillary density. Capillary density ( CD ) was calculated from orientation corrected mean number of vessels in an image cross section ( defined in Figure 2 3B), image edge length in pixels ( L ), and pixel size ( S ) using Equation 50 This calculation assumes that capillary spacing perpendicular to the image plane is similar to capillary spacing within the image plane. (50)

PAGE 106

95 Figure 2 3: By hand measurement techniques. ( A ) A raw microscope image with software analyzed capillary segments outlined was used as a reference for by hand measures. ( B ) Measurements taken by hand include RBC flow status for each software analyzed capillary number of visible capillaries in each of 3 image cross sections. The average number of capillaries in an image cross sectio n is then corrected for vessel orientation ( ) and used to calculate capillary density. 5 Results 5 .1 Comparison of software and by hand measurements Side by side comparison of software and by hand measurements of capillary structure and perfusion are shown in Figure 24. Plasma perfused capillary density is significantly reduced in HFD mice relative to chow fed mice using both software (Figure 24A) and by hand (Figure 24D) measurement techniques (both p<0.01). No significant difference was observed between software and by hand measures of plasma perfused capillary density for either chow fed or HFD mice (both p>0.1). The fraction of plasma perfused capil laries without detectable RBC flux is significantly increased in HFD mice relative to chow fed mice using both software (Figure 24B) and by hand (Figure 24E) measurement techniques

PAGE 107

96 (both p<0.01). No significant difference was observed between software and by hand measures of RBC flux status for either chow fed or HFD mice (both p>0.1). RBC perfused capillary density is significantly reduced in HFD mice relative to chow fed mice using both software (Figure 24C) and by hand (Figure 24F) measurement techniques (both p<0.01). No significant difference was observed between software and by hand measures of RBC flux status for either chow fed or HFD mice (both p>0.1). Collectively, these results suggest 1) impaired microvascular perfusion in diet induced obesity, a nd 2) equivalence of software and by hand measures of capillary structure and perfusion. Figure 2 4: Software and by hand measurements detect the same capillary perfusion and RBC flux differences between chow fed and HFD mice. A ) Plasma perfused capillary density measured by our program is significantly lower for HFD than for chow diet (p=0.004). B ) Absence of capillary RBC flux detected by our program is significantly more prevalent in HFD than in chow diet (p=0.003). C ) RBC pe rfused capillary density measured by our program is significantly lower for HFD than for chow diet (p=0.09). D ) Measurement by hand closely agrees with software measurement for plasma perfused capillary density. E ) Measurement by hand closely agrees with s oftware measurement for absence of RBC flux. F ) Measurement by hand closely agrees with software measurement for RBC perfused capillary density.

PAGE 108

97 5 .2 Software only measures of capillary structure and perfusion Our software also records mean capillary dia meter, mean capillary flow velocity, and heterogeneity of capillary flow velocity (assessed using coefficient of variance, CV =Std/Mean). Comparison of HFD and chow fed mice in these metrics is shown in Figure 5. No significant difference in mean capillary diameter (Figure 5A) was observed between HFD and chow fed mice (p=NS). No significant difference in mean capillary flow velocity (Figure 5B) was observed between HFD and chow fed mice (p=NS). Perfusion heterogeneity (Figure 5C), on the other hand, was si gnificantly increased in HFD mice relative to controls (p<0.001), consistent with previous reports of heterogeneous microvascular perfusion in diet induced obesity (24 26) Figure 2 5: Additional measurements taken using software only. A ) Capillary diameter does not differ significantly between HFD and control (p=NS). B ) Mean capillary flow velocity does not differ significantly between HFD and control (p=NS). C ) Heterogeneity of capillary flow velocity is significantly greater in HFD than in control (p<0.01). In addition to quantitative results, our software also allows visualization of capillary perfusion in a still image suitable manuscripts use. An example of this visualization is demonstra ted in Figure 26 using data from a representative each from the HFD and chow fed groups. Visual comparison of detected vessels (left) indicates somewhat reduced capillary density in HFD (bottom) relative to chow fed control (top). Visual comparison of inte nsity maps of capillary flow velocity (right) reveals substantial capillary dropout and flow

PAGE 109

98 heterogeneity in HFD (bottom) relative to chow fed control (top). These observations agree with visual inspection of the corresponding videos and also agree with o ur quantitative measures of capillary structure and perfusion. All vessel maps and perfusion intensity maps generated are included in the supplemental materials. Figure 2 6: Visualization of HFD on capillary structure and perfusion. For a representative i mage each from the HFD (bottom) and chow diet (top) groups, plasma perfused capillaries detected by the software are shown on the left and an intensity map of RBC flow velocities is shown on the right. Absence of RBC flux in many HFD capillary segments is visibly apparent. 6 Discussion In this manuscript, a novel software technique for quantifying capillary structure and perfusion from intravital microscopy videos employing fluorescently labelled plasma as a contrast agent is described. We apply this method to a test case of capillary pe rfusion in HFD (insulin resistant) and chow fed (insulin sensitive) mice during hyperinsulinemia, and validate software measurements by comparison to by hand measurements. By hand

PAGE 110

99 measurements and software measurements both revealed significantly decreased density of both plasma perfused and RBC perfused capillaries in HFD mice, along with significantly increased prevalence of plasma perfused capillaries without detectable RBC flux. Measurements taken by hand and with software did not differ for any metric. In addition to the aforementioned measures, we also demonstrate quantification of mean capillary diameter, mean capillary flow velocity, and heterogeneity of capillary perfusion (which was increased in HFD mice) using our software. 6 .1 Va lidity of softwar e measurements Previous approaches to quantifying capillary perfusion using intravital microscopy videos have primarily used by hand measurements. Depending on the study, these by hand measurements may range from easily measured parameters such as RBC per fused capillary density to more laborious measurements of RBC flow velocity. Our experience when performing by hand measurements for validation of software results was that interpretation of intravital microscopy videos was often ambiguous, and this ambigu ity resulted in some degree of subjectivity. To mitigate the impacts of this subjectivity, we assigned a laboratory technician to take fully blinded by hand measurements so that any measurement error in our by hand measures would occur independent of inves tigator bias. With these limitations of by hand intravital microscopy measurements in mind, we used by hand measurements as a standard for validation our software results, drawing from the observation that most previous papers have used by hand measurement s. Comparisons of software measurements to by hand measurements and/or literature data suggest that our software measures capillary structure and perfusion accurately. In all cases, software and by hand measurements yielded equivalent results for HFD vs c how fed

PAGE 111

100 differences. No significant differences were observed between software and by hand measurements for either group in any metric. Furthermore, statistical significance of our observations was equal or greater using software measurements in all cases, suggesting that the inherent consistency of software measurement may actually improve measurement error relative to by hand measurement. Finally, software measured values for capillary density in mouse gastrocnemius (~1000 capillaries/mm 2 ), perfused capil values (27; 28; 181 183) Collectively, these results indicate that our software provides reliable, quantitative measurements of key parameters relating to capillary structure and perfusion. 6 .2 Interpretation of capillary perfusion diff erences in diet induced obesity Our so ftware analysis indicates several important differences in hyperinsulinemic capillary perfusion between HFD and control mice. First, plasma perfused capillary density is significantly reduced in HFD mice relative to chow fed controls (463/mm 2 vs 924/mm 2 r espectively). This is consistent with previous reports of capillary rarefaction in more severe models of metabolic disease (122) although it should be noted that this difference is not always observed (28) Reduced capillary density has previously been shown to correlate with insulin resistance (118) and it is likely that reduced capillary surface area for insulin efflux mediates this correlation (184) Our software also reports a substantial number of capillaries without detectable RBC flux in HFD mice (35% vs 5%, respectively). This effect size may seem unre alistically large at first glance, but 30% 40% capillary dropout has previously been reported not only in multiple animal models of diabetes (an insulin resistant state) (27; 28) but also in acute hyperglycemia (81) This loss of RBC flux in some capillaries compounds

PAGE 112

101 the already reduced capillary density in HFD mice to result in substantially reduced RBC perfused capillary density (290/mm 2 vs 882/mm 2 respecti vely), and likely contributing to well documented skeletal muscle oxygenation defects in T2DM (26; 52; 60; 97) Mean capillary flow velocity is not significantly alte red in HFD mice relative to chow fed controls muscle blood flow. This is consistent with previous reports of reduced skeletal muscle blood flow in T2DM (20; 22; 28; 51; 59) Finally, heterogeneity of capillary perfusion was substantially increased in HFD mice re lative to chow fed controls. This is consistent with previous reports of microvascular perfusion heterogeneity in metabolic disease (23 26) which has been shown to contribute to impaired muscle oxygenation (63) Despite the small sample size used in this pilot study (n=6 chow fed, n=4 HFD) reduced capillary density, loss of RBC f lux, reduced blood flow, and heterogeneous capillary perfusion are observed. These would each be expected to cause impaired exchange of soluble molecules (e.g. insulin and glucose) between blood and the interstitium (66) One would expect, then, that microvascular dysfunction contributes to insulin resistance. This prediction is consistent with the relationship between capillary perfusion and insulin action i n humans (184) consistent with impaired insulin access to the interstitium in diet induced obesity (137) and consistent with data indicating that vascular glucose delivery is rate limiting to insulin action in HFD mice (112) Although it is probable that perfusion d efects such as those we observed are mechanistically linked to insulin resistance in HFD mice, it is less clear which perfusion parameter(s) is/are most critical to changes in insulin and glucose transport. Our recently published theory of microvascular pe rfusion and solute exchange (66) provides a mathematical framework for predicting which parameters are critical drawing from

PAGE 113

102 knowledge of capillary flow dis tribution. Future studies are warranted to fully characterize the relationships of capillary density, bulk blood flow, and perfusion heterogeneity with insulin action in T2DM. 6 .3 Using this software in your research It is our hope that investigators will apply this technique in their own investigations of microvascular perfusion. This software can potentially be used to assemble a comprehensive database of capillary flow distributions in a wide variety of tissues, physiological contexts, disease states, a nd pharmacological interventions, and to use this database for engineering analyses relating microvascular perfusion parameters to trans capillary flux of small molecules (e.g. oxygen, insulin). While the computational approach to capillary perfusion measu rement developed in this manuscript is broadly applicable, there are many parameters in our algorithms (e.g. size of pixel neighborhood for brightness normalization) that are likely to require adjustment with different microscopes or contrast agents. Inves tigators that would like to use this software technique in their research are encouraged to contact us with the details of their specific microscope setup. Several video attributes that are essential for this analy sis technique are listed below: Rapid fram e rate (>30 fps recommended) Many capillaries in field of view (>10 recommended) Sufficient image quality for RBC flux to be visibly apparent Although this next stipulation is not absolutely necessary, we strongly re commend using a fluorescent plasma label. The velocity measurement technique developed in this manuscript is, in principle, applicable to fluorescently labelled RBCs, light microscopy, and any other

PAGE 114

103 technique in which RBCs are distinguishable from plasma. However, without a plasma specific contrast mechanism, vessels without RBC flux will not be detectable. 6 4 Conclusions In this manuscript, we describe a novel software technique for automated quantitation of cap illary structure and perfusion. Videos of capillary perfusion in the gastrocnemius muscle of HFD and chow fed mice during hyperinsulinemia were used for a pilot study of this software technique. Our software was able to detect reduced capillary density, loss of RBC flux, and heterogeneous capillar y perfusion in HFD mice. These differences were entirely consistent with by hand measures and previously published literature results, and provide a plausible mechanism for impaired vascular delivery of insulin and glucose in insulin resistant skeletal mus cle. Using our software, capillary structure and perfusion can be measured in a high throughput, automated fashion, with minimal investigator bias, and yielding a wide variety of consistent, quantitative metrics that can be reported in a standardized fash ion. This technique can be combined with existing theories relating the distribution of capillary flow to the transport of physiologically relevant molecules such as oxygen and glucose (66) Using this combination, we hope to advance the research of microvascular perfusion defects in disease from the domain of largely binary, observational results (i.e. perfusion is or is not altered) to the domain of detaile d, mechanistic results that help to define specific elements for pathology and measurable targets for therapeutic intervention. Future studies employing videos from a variety of sources in a wide variety of tissues and conditions will be required to fully realize this potential.

PAGE 115

104 CHAPTER VI THE ENDOTHELIAL GLYCOCALYX PROMOTES HOMOGENOUS BLOOD FLOW DISTRIBUT ION WITHIN THE MICROVASCULATURE Co authors: Michal Schafer, Kendall S. Hunter, and Jane E. B. Reusch 1. Preface Up to this point, I have discussed 1) the relevance of blood flow and its distribution to exercise capacity in diabetes, 2) evidence that reduced blood flow alone cannot account for impaired muscle oxygenation in diabetes, 3) a theory relating blood flow distribution to tissue oxygenation, 4) empirical validation of this theory, and 5) additional experimental evidence for perfusion heterogeneity in insulin resistant states The reader may be left wondering why microvascular perfusion heterogeneity would occur in diabetes to begin with. In this chapter, I make a first foray into the underlying mechanisms leading to microvascular perfusion heterogenei ty. The endothelial glycocalyx, a gel like layer coating the in t erior surface of blood vessels, is degraded in diabetes. Analyses building from both empirical and theoretical observations both converge upon the conclusion that the glycocalyx promotes homogenous blood flow distribution in the microcirculation through pa ssive mechanical interactions with circulating cells. Thus, under conditions of glycocalyx degradation (such as in diabetes and other inflammatory disease states), microvascular perfusion heterogeneity and impaired tissue oxygenation are to be expected. Th is chapter was published in the American Journal of Physiology Heart and Circulatory Physiology in 2016. 2. Abstract Many common diseases involve impaired tissue perfusion, and heterogeneous distribution of blood flow in the microvasculature contributes to this pathology. The

PAGE 116

105 physiological mechanisms regulating homogeneity/heterogeneity of microvascular perfusion are presently unknown. Using established empirical formulations for blood viscosity modelling in vivo (blood vessels) and in vitro (glass tubes) we show that the in vivo formulation predicts more homogenous perfusion of microvascular networks at the arteriolar and capillary levels. Next we show that the more homogeneous blood flow under simulated in vivo conditions can be explained by changes in red blood cell interaction with the vessel wall. Finally, we demonstrate that the presence of a space filling, semi permeable layer (such as the endothelial glycocalyx) at the vessel wall can account for the changes red blood cell interactions with the ves sel wall that promote homogenous microvascular perfusion. Collectively, our results indicate that the mechanical properties of the endothelial glycocalyx promote homogeneous microvascular perfusion. Preservation or restoration of normal glycocalyx properti es may be a viable strategy for improving tissue perf usion in a variety of diseases. 3 Introduction Microvascular perfusion heterogeneity (unequal distribution of microvascular blood flow within target tissues) has been observed in a wide variety of disease states, including sepsis (185) aging (186) type 2 diabetes (25) and chronic he art failure (187) In each of these disease st ates, capacity for aerobic exercise is reduced (47; 188 190) Many mechanisms have been proposed to contribute to exercise impairment in these conditions, including loss of muscle mass (189) reduced nutritive blood flow (190) abdominal obesity (191) increased circulating inflammatory markers (192) and endothelial dysfunction (193) It is reasonable to suggest that microvascular perfusion heterogeneity itself might contribute to reductions whole body oxygen uptake, based upon both empirical (26; 68) and theoretical

PAGE 117

106 (64; 65) data demonstrating that microvascular perfusion heterogenei ty limits oxygen extraction independently of total blood flow. Perfusion heterogeneity is a plausible mechanism for impaired oxygenation in virtually any tissue, given that the mechanism of impairment (namely, that some capillaries are underperfused and de liver less oxygen while others are overperfused and effectively saturate their capacity for oxygen delivery) is not tissue specific. One potential cause of disease related microvascular perfusion heterogeneity is degradation of the endothelial glycocalyx and subsequent changes in the determinants of effective blood viscosity. The glycocalyx is a space permeable layer of glycosaminoglycans and proteoglycans lining the luminal surface of blood vessels. Experimental glycocalyx degra dation causes microvascular perfusion heterogeneity, thus reducing the number of capillaries supporting red blood cell (RBC) flux (79 81) and decreasing oxygen utilization (83; 84) Precisely these same derangements in microvascular perfusion are observed in sepsis (185) type 1 diabetes (27) type 2 diabetes (28) and heart failure (194) along with findings of glycocalyx perturbation (71; 195 197) and oxygen utilization limitations (47; 188 190) It seems li kely, then, that glycocalyx degradation could cause impairments in tissue oxygenation secondary to microvascular perfusion heterogeneity. The link between disease mediated glycocalyx degradation and microvascular perfusion heterogeneity has not been fully explained. Hints of a unifying mechanism can be found in rheological studies of blood viscosity. Pries et al observed that microvascular blood viscosity in vivo (measurements performed in blood vessels) is far greater than that observed in vitro (measureme nts performed in glass tubes) (198; 199) particularly at the pre capillary

PAGE 118

107 and capillary levels. This discrepancy in viscosity can be mathematically accounted for by the mechanical effects of a space filling glycocalyx at the endothelial surface (200) Several distinct lines of historical data inform our understanding of the relationship between glycocalyx properties and microvascular perfusion independently of to tal blood flow (Schema: Figure 27 ). Empirical formulations of blood viscosity (198; 1 99) and their application microvascular perfusion simulations (201) indicate that distribution of blood flow at microvascular bifurcations is determined by both blood viscosity effects and differences in downstream resistance. The interactions between vessel diameter and effective blood viscosity (dec reasing diameter decreases viscosity in larger microvessels and increases viscosity in smaller microvessels) and between hematocrit and viscosity (increasing hematocrit increases viscosity) introduce additional determinants of flow distribution within the microcirculation as compared to a simple Newtonian fluid (Figure 27 A). Vessel diameter has a complex interaction with effective blood viscosity in microvessels due the parti culate nature of blood (Figure 27 iameter), decreasing vessel diameter is associated with decreasing viscosity as RBC flow gradually shifts from multi file to single file (i.e. shear thinning) (202) In smaller vessels (i.e. less than deformation incr eases with decreasing vessel diameter. The shift from shear thinning to RBC deformation occurs at a larger vessel diameter in vivo relative to in vitro. Increasing hematocrit (fraction of vessel lumen volume filled by RBCs) is associated with increased vi scosity regardless of scale, and this effect is most pronounced in the smallest blood vessels. Hematocrit is increased in the higher flow branch of a microvascular bifurcation and decreased in the lower flow branch (165) (Figure 27 C), providing a negative feedback

PAGE 119

108 mechanism to regulate flow partition (increased flow increases hematocrit, thus increasing viscosity and limiting the increase in flow). The dependence of blood viscosity on hematocrit is nearly an order of magnitude greater in vivo than in vitr o (Figure 27 D), indicating that unequal distribution of RBCs would incur a greater ener gy cost in vivo than in vitro. Figure 27 : Schema of established determinants of mi crovascular blood flow distribution based on the literature. ( A ) Relationships between parameters influencing microvascular blood flow distribution during the transition from multi file to single file flow (left) and during RBC deformation (right). Causal increases (i.e. increases in A cause increases in B) are indicated by solid arrows while causal decreases are indicated by dashed arrows. Here B ) Blood viscosity (y axis) changes w ith changing vessel diameter (x axis) both in vivo and in vitro. In vitro, increasing vessel diameter increases microvascular blood viscosity at all physiologically relevant vessel diameters. In vivo, increasing vessel diameter decreases blood viscosity at the capillary and pre capillary levels, but otherwise increases blood viscosity. ( C ) Ratio of daughter vessel hematocrit to parent vessel hematocrit (y axis) as a function of fractional of parent vessel flow rate received by the daughter vessel (x axis). An idealized bifurcation in this graph. ( D ) Derivative of microvascular blood viscosity (y axis) with respect to vessel hematocrit (x axis). Although increases in hematocrit increase blood viscosity both in vivo and in vitro, this effect is much stronger in vivo (198; 199)

PAGE 120

109 In this manuscript we employ mathematical models of blood viscosity and microvascular rheology to simulate the impact of differences in microvascular structure, blood viscosity behaviors, and glycocalyx properties on distribution of blood flow within the microcirculation. We use this simulation method to predict the impact of glycocalyx modification (common in agi ng and disease) upon tissue perfusion. These data support the hypothesis that the glycocalyx is a central regulator of tissue perfusion and oxygenation independently of blood flow redistribution to match metabolic demand. This finding has important implic ations for tissue perfusion in health and disease, as the glycocalyx is highly dynamic and presents a novel therapeutic target. 4 Methods 4 .1 Simulation of Blood Flow and Red Blood Cell Distribution The methods of Pries et al (201) were used for simulation of blood flow and RBC distribution. In this technique, distribution of RBCs and w hole blood (including both RBCs and plasma) at microvascular bifurcations are determined according to an empirical formulation relating RBC distribution to vessel diameters, flow rates, and hematocrit (165) These relation ships are shown in Equations 51 55 Here FQ E is fractional RBC flow, FQ B is fractional blood flow, logit is a logistic function defined in Equation 5 1, X 0 is minimal fractional blood flow to support RBC flux, D 0 is parent vessel diameter, D 1 is diameter of daughter vessel 1, D 2 is diameter of daughter vessel 2, H is parent vessel hematocrit, and A and B are empirically defined scaling parameters. ( 5 1) ( 5 2)

PAGE 121

110 ( 5 3) ( 5 4) ( 5 5) Flow distribution was optimized at each bifurcation to ensure total pressure drops across parallel vascular routes were equal to a relative error less than 10 6 Pressure drops were assessed using established empirical formulations of microvascular blood viscosity in vitro (glass tubes) (198) a nd in vivo (blood vessels) (199) to assess the contributions of the endothelium to regulation of blood flow dis tribution in vivo. These empirical microvascular blood viscosity relati onships are shown in Equations 56 58 (in vivo) and 56, 59, and 60 (in vitro). Here is in vivo blood viscosity at hematocrit H is plasma viscosity, is in v ivo blood viscosity at a hematocrit of 0.45, C is an empirical scaling factor that modulates the interplay between vessel diameter and hematocrit sensitivity of blood viscosity, D is in vitro blood viscosity at hemato crit H and is in vitro blood vis cosity at a hematocrit of 0.45. ( 5 6) ( 5 7) ( 5 8) ( 5 9) (6 0) Because heterogeneity of trans it times (time spent within the microvasculature) has been shown to be the basis for impaired oxygen transport with heterogeneity of

PAGE 122

111 microvascular perfusion (64; 65) we used transit time heterogeneity (as assessed by standard deviation to mean ratio) at both the terminal arteriole (consistent with theoretical analyses (64; 65) ) and whole network (c onsistent with animal studies using radiolabeled plasma (24; 25) ) levels to assess microvascular blood flow di stribution in our simulations. 4 .2 Idealized Arteriolar Branching Structure for Network Perfusion Simulations An idealized arteriolar tree consisting of 7 successive vessel generations was employed for all network perfusion simulations. A ve vessel generation was used for all simulations, resulting in a range of vessel diameters consistent with pre vivo and in vitro blood viscosity behav iors begin to div erge at this scale (see Figure 27 B). Terminal vessel diameters varied according to the parameter values used, but were centered pa rameters illustrated i n Figure 28 The parameter was used to define asymmetry in (166) was used to determine the scale of downstream vessels at each bifurcation, and the resulting vascular tree was consistent with the L system fractal formalism of Zamir et al (167) Although more sophisticated formulations of microvascular branching structure have been developed (203 205) the mathematic al rules employed in this study have been validated as reasonable approximations of microvascular structure in a variety of tissues (167; 168) and allow for straightforward assessment of the interactions between blood viscosity formulations, microvascular structure, and the resulting distribution of microvascular blood flow.

PAGE 123

112 Figure 2 8 : Parameters used to create idealized arteriolar t rees. Each parent vessel terminates at a bifurcation of two smaller vessels. At each bifurcation, the parent vessel is denoted by subscript 0, the smaller daughter vessel is denoted by subscript 1, and the larger daughter vessel is denoted by subscript 2. Bifurc terms of the relative diameters of the daughter vessels. Length diameter ratio is held constant (166) was used to define the scaling of successive vessel generations. Here D is v essel diameter, L is vessel length, and su bscripts denote vessel segment. 4 .3 Simulation of Glycocalyx Influences on Luminal Plasma Velocity Profiles Whereas the analyses described in sections 4.1 4 .2 aim to determine differences perfusion predicted from in vivo and in vitro measurements, the analysis in this section aims to determine whether these perfusion differences can be accounted for by the mechanical influences of the endothelial glycocalyx. The mechanical influences of the endothelial glycocalyx derive from both its space filling and semi permeable nature. In order to assess the effects such a layer at the vessel wall on the sensitivity of blood viscosity to hematocrit, flow velocity profiles ere developed in vessels containing an idealized glyco calyx were simulated using a radial, 1 dimensional finite di fference approach (see section 5 .4 for examples). Simulations were performed in Matlab 2013. The glycocalyx was assumed to

PAGE 124

113 have a fixed width of w K (in Darcy units). A fixed lift distance of 100 nm between the glycocalyx and RBC membrane was used for these simulations. Fully developed flow and a no slip boundary condition were assumed for all simulations. Velocities at each radial node were adjusted to achieve a uniform (relative error of less than 10 10 ), non zero pressure drop across a length normalized capillary segme nt as assessed using Equations 61 62 A radial resolution of 12.5 nm was used for all simulations. Each simulation was repeated twice : once ass uming plasma only (Equation 61 ), and once assuming an RBC in the vessel lumen (Equation 62 ). Here r is radial position within the capillary, R is the radius of the capillary, and R C is the radius of the RBC. These physical parameters were closely based on those employed in previous studies simulating the effect of the glycocalyx on luminal flow profiles (206 208) (61 ) (62 ) Empirical studies of microvascular rheology demonstrate that effective microvascular blood viscosity depends heavily upon hematocrit and diameter, and that the relative importance of these sensitivities is different in vivo and in vitro (198; 199) Drawing from the results of Secomb et al (207) the sensitivity of microvascular blood viscosity to hematocrit can be characterized for discha rge hematocrits (hematocrit of effluent blood in a microvessel) less than ~0.6 based on differences in flow resistance with and without an RBC in the vesse l

PAGE 125

114 lumen as shown in Equation 63 Here is effective blood viscosity, is plasma viscosity, Hct is hematocrit, and K T is the hematocrit sensitivity of blood viscosity. (63 ) 5 Results 5 .1 Flow Distribution in Branching Arteriolar Networks Observation of arteriolar branching architecture in vivo reveals that arterioles often bifurcate into a larger vessel (usually serving a larger downstream capillary bed) and a smaller vessel (usually serving a smaller downstream capillary bed) (167; 168) In order to assess the effects of asymmetric bifurcations in the arteriolar tree on tissue perfusion, we first modeled flow in idealized arteriolar networks with 7 successive vessel generations, co nstant length to diameter ratio, and a fixed ratio of smaller daughter vessel diameter to larger da ughter vessel diameter (Figure 29A) as detailed in section 4 .2 Increasing bifurcation asymmetry results in increased heterogeneity of vascular transit time s (as assessed by standard deviation to mean ratio) at both the whole vasculature (Figure 29B) and single vessel (29 C) levels. In both cases, the increase in transit time heterogeneity is far more pronounced using an empirical in vitro blood viscosity form ulation than using an empirical in vivo blood viscosity formulation. These results are in contrast to the effects of bifurcation asymm etry on flow asymmetry (Figure 29 D, assessed as standard deviation of fractional flow rate at each bifurcation) and terminal vessel he matocrit heterogeneity (Figure 29 E, assessed as standard deviation of hematocrit). According to these metrics, increases in flow and hematocrit heterogeneity are more pronounced using an empirical in vivo blood viscosity formulation than using an empirical in vitro blood viscosity formulation. The apparent divergence of transit time and flow heterogeneity results stems

PAGE 126

115 from the fact that larger upstream vessels feed a larger volume of downst ream blood vessels (see Figure 29 A). Because tran sit time is determined by the ratio of luminal volume to flow rate, a more pronounced asymmetry of flow distribution at the single bifurcation level with i n vivo blood viscosity (Figure 29 D) would be expected to result in reduced heterogeneity of vascular transit times und er the same conditions (Figure 29 B). These results indicate that the microvascular perfusion heterogeneity caused by vessel diameter effects at the arteriolar level would be more pronounced in glass tubes than in blood vessels, and that th is effect is a result of the increased effective blood viscosity in pre capill ary arterioles in vivo (Figure 27 B). Figure 29 : Influences of microvascular blood viscosity on microvascular perfusion heterogeneity in a simulated arteriolar tree. ( A ) Schema tic of the idealized arteriolar networks simulated consisted of 7 successive vessel generations with fractal vessel architecture. ( B ) Heterogeneity (SD/mean) of whole n etwork transit time is reduced in vivo

PAGE 127

116 C ) Heterogeneity (SD/mean) of terminal vessel transit time is reduced in vivo compared to in vitro for all D ) Standard deviation o f fractional daughter vessel flow rate is subtly increased E ) Standard deviation of terminal vessel hematocrit is subtly increased in vivo compared t 5 .2 Flow Distribution at an Idealized Capillary Bifurcation Observation of capillary networks in vivo reveals that variations in capillary diameter within a given tissue are typically minor compared to variations in capillary length, glycocalyx dimensions, and other diameter i ndependent determinants of capillary blood flow. (150; 209; 210) To assess the differential effects of empirical blood viscosity formulations based on m easurements taken in vivo (blood vessels) and in vitro (glass tubes) on capillary perfusion, we simulated blood flow distribution at an idealized capillary vessels with 1 2 (Figure 30 A). Downstream resistances were multiplied by blood viscosity such that increases in microvessel hematocrit induced increases in downstream resistance. As a result of the increased hematocrit dependence of microvascular blood viscosity in vivo, fractional flowrate at an idealized capillary bifurcation is less dependent upon downstream resistances in vivo than in vitro (Figure 30 B). As a result of this effect and the relationship between fractional flowrate and downstream hematocrit (Figure 27 C), our simulations predict that capillary hematocrits downstream of an idealized bifurcation would be more different using an empirical in vitro blood viscosity formulation than an empirical in vivo blood viscosity form ulation (Figure 30 C). These results indicate that an increased hematocrit depende nce of blood viscosity (Figure 27 D) limits microvascular perfusion heterogeneity at the capillary level in blood vessels relative to glass tubes.

PAGE 128

117 Figur e 30 : Influences of mi crovascular blood viscosity on blood flow distribution at an idealized capillary bifurcation. ( A ) Schematic of idealized capillary bifurcation. Arrows indicate the direction of flow. Both daughter vessels and the parent vessel are all of identical diameter 1 2 ). ( B ) Fraction of parent vessel flow rate received by daughter vessel 1 as a function of 1 2 ). In vivo blood viscosity behaviors provide a more 1 2 ( C ) Fraction of parent vessel hematocrit in 1 2 ). In vivo blood viscosity behaviors result in less downstrea m capillary hematocrit heterogeneity for all 1 2 5 .4 Influence of Glycocalyx Properties on the Determinants of Blood Viscosity In order clarify the contributions of the endothelial glycocalyx specifically (rather than predicting differences in blood flow distribution in blood vessels as opposed to glass tubes) we used a 1 dimensional finite difference approach to simulate the influences of glycocalyx permeability and width on the sensitivity of blood viscosity to hematocrit and the

PAGE 129

118 effective viscosity of blood at the capillary level in isolation of other influences. Drawing from previous studies (206 208) we elected to test the effects of glycocalyx permeability and width. Flow profiles at near physiologic values of glycocalyx permeability (K 0 =10 10 Darcy) those report ed in previous studies (Figure 31 A). Note that some of the studies referenced above report hydraulic resistivity (the inverse of permeability) or water permeability (permeability to water) rather than Darcy permeability per se. Literature val ues for glycocalyx permeability are largely consistent when presented in similar units. The sensitivity of microvascular blood viscosity to hematocrit decreases with increasing glycocalyx permeability regardless of glycocalyx width, and this effect is mos t pronounced at near physiologic values (206 208 ; 211; 212) for g lycocalyx permeability (Figure 31 B). The sensitivity of blood viscosity to hematocrit varies relatively little with glycocalyx width within the range of literature reported values (200; 212 215) The effective viscosity of blood (Hct=0.45) at the capillary level decreases with increasing glycocalyx permeability, and this effec t is further modulated by glycocalyx width (Figure 31 C). In total, our results indicate that the sensitivity of blood viscosity to hematocrit is primarily determined primarily by glycocalyx permeability. The effective viscosity of blood in the microcircula tion is strongly modulated by both glycocalyx permeability and glycocalyx width.

PAGE 130

119 Figure 31 : Effects of glycocalyx properties on the determinants of blood viscosity. ( A ) Representative velocity profiles within the vessel with and without an RBC in the lumen. Shaded gray areas represent the glycocalyx. Compare to the results of Secomb et al 2001 (207) ( B ) Sensitivity of mi crovascular blood viscosity to hematocrit decreases with increasing glycocalyx permeability (Darcy permeability). This effect is influenced little by glycocalyx width. ( C ) Effective blood viscosity at the capillary level increases with decreasing glycocal yx permeability. This effect is further compounded by th e effects of glycocalyx width. 6 Discussion The simulations outlined in this manuscript were performed in order to clarify the contribution of the endothelial glycocalyx to regulation of microvascular blood flow distribution. Collectively, our results indicate that the endothelial glycocalyx increas es the homogeneity of microvascular perfusion by enhancing interactions between RBCs and the vessel wall within the microcirculation. By comparing flow distributions predicted with

PAGE 131

120 empirical in vivo (blood vessels) and in vitro (glass tubes) formulations o f blood viscosity in idealized arteriolar trees with unequal downstream diameters at each bifurcation, we show that blood vessels promote homogenous perfusion in branching microvascular networks. By comparing flow distributions predicted with empirical in vivo and in vitro formulations of blood viscosity at an idealized capillary bifurcation, we show that the increasing marginal energy costs associated with cellular flow promotes homogenous perfusion at the capillary level. Finally, by comparing luminal pla sma velocity profiles and pressure gradients with and without an RBC in the vessel lumen across a wide range of values for glycocalyx width and permeability, we show that the sensitivity of microvascular blood viscosity to hematocrit at a given vessel diam eter is determined primarily by glycocalyx permeability, while the increased effective viscosity of blood is strongly influenced by both the width and permeability of the glycocalyx. In total, our simulations suggest that the presence of a semi permeable, space filling glycocalyx layer in the vessel lumen is sufficient to promote homogeneity of microvascular perfusion by enhancing RBC interactions with the endothelium. A non technical summary of our findings is included in Figure 32 below. Our simulations begin at the pre capillary arteriolar level (diameter < 50 capillary level. Because the majority of the pressure drop in the arterial circulation has already occurred at this point, the redistribution of blood flow to meet metabolic dema nds occurs upstream of the microvessels considered in this analysis. Oxygen delivery (as opposed to blood flow) may still be modulated by the homogeneity/heterogeneity of blood flow and RBC distribution. At the precapillary arteriolar level, daughter vesse l diameters at any given bifurcation may be unequal, making both the diameter and hematocrit dependencies of blood

PAGE 132

121 viscosity (see Figure 27 ) important to arteriolar blood flow distribution. These re sults are discussed in section 5 .1. At the capillary level daughter vessel diameters are likely to be similar at any given bifurcation, making the hematocrit dependency of blood viscosity an important determinant of capillary blood flow distribution. These re sults are discussed in section 5 .2. Endothelial glycoc alyx width influences the diameter the diameter dependency of blood viscosity (primarily important at the arteriolar level), while glycocalyx permeability influences the hematocrit sensitivity of blood viscosity (primarily important at the capillary level) These re sults are discussed in Section 5 .3. Collectively, our data are consistent with perturbed capillary perfusion under conditions that increase glycocalyx permeability (79 81) consistent with the increased width of the glycocalyx at the arteriolar level relative to the capillary level (214) and suggest that the endothelial glycocalyx promotes homogen ous distribution of blood flow and RB Cs within the microvasculature. Figure 32 : Summary of findings. Our simulations concern distribution of blood flow and RBCs within precapillary arterioles and capillaries. At this level of the circulation, total flow rate has already been determined by upstream arteries. At the arteriolar level (see section 5.1 and Figure 29 ), both diameter and hematocrit dependencies of blood viscosity influence

PAGE 133

122 flow distribution. At th e capillary level (see section 5.2 and Figure 30 ) the hematocrit dependency of blood viscosity modulates flow distribution. Based on the influences of glycocalyx properties on the diameter and hematocrit dependencies o f blood viscosity (see section 5.3 and Figure 31 ), this suggests that both glycocalyx width and permeability modulate flow distribution at the arteriolar level whereas glycocalyx permeability is the primary determinant at the capillary level. The physiological implications of our simulation results are best understood in light of the knowl edge that heterogeneous distribution of oxygenated blood within the microvasculature limits effective tissue oxygenation (26; 64; 65; 68) Enzymatic degradation of the glycocalyx by hyaluronidase or hep arinase has been reported to acutely redistribute flow to a smaller number of capillaries at a higher hematocrit (79; 80) In addition, hyaluronidase treatment has been shown to markedly increase the heterogeneity of flow velocities in both arterioles and venules (see Figure 1 in Reference (79) ). These results support our finding that glycocalyx permeability is an important determinant of perfusion homogeneity/heterogeneity, given that the hyaluronan component of the glycocalyx is a major determinant of glycocalyx permeability (216) Furthermore, glycocalyx shedding during adenosine infusion (83) or systemic inflammation (84) has been shown to be associated with reduced oxygen extraction, as would be expected in the case of increased perfusion heterogeneity (26; 64; 65; 68) Our f inding that the glycocalyx is a crucial regulator of microvascular perfusion and tissue oxygenation independently from total blood flow is consistent with these clinical and experimental observations. Reports of impaired microvascular perfusion in diabete s further support the clinical relevance of our simulation results. Direct measurement endothelial glycocalyx properties in human diabetes reveals decreased glycocalyx width and increased glycocalyx permeability, which can be partially reversed by diabetic medications Metformin and Sulodexide (71; 217) The microvascular effects reported with either enzymatic glycocalyx degradation or diabetes

PAGE 134

123 are also recapitulated by short term hyperglycemia, which acutely increases vascular permeability and increases heterogeneity of capillary perfusion (81) The results of Frisbee et al are of particular interest, showing divergence of arterial perfusion and skeletal muscle fatigue re sistance in the Obese Zucker Rat (a common animal model of Type 2 Diabetes) owing to impaired oxygen extraction (60) This effect was determined to stem from increased heterogeneity of microvascular perfusion (25) which also caused an increase in capillary hematocrit heterogeneity (consistent with the effects of enzymatic glycocalyx degradation) (24) and has been determined to contribute to peripheral vascular disease in this model (26) Reduction in the fraction of capillaries supporting RBC flux has been reported in animal models of both Type 1 and Type 2 diabetes (27; 28) Given the striking similarity between these profiles of altered capillary hematocrit and given that perturbed ox ygen extraction during exercise has been demonstrated in human diabetes (29; 59) these data are once again consistent with our finding that the sensitivity of blood viscosity to hematocrit imparted by the glycocalyx is a crucial regulator of microvascular perfusion and tissue oxyg enation independent from total blood flow. The protective effects of the endothelial glycocalyx (76) may include promoting homogenous microvascular perfusion and thus improving tissue oxygenation. This is of clinical relevance as microvascular perfusion heterogeneity is found i n sepsis (185) a ging (186) diabetes (25) and chronic heart failure (187) peripheral arterial disease (218) cancer (219) and pulmonary hypertension (220) among others. Our findings of microvascular perfusion heterogeneity with glycocalyx degradation are not entirely novel (79; 80) The simulations in this paper are, however, the first to explicitly define a mechanism for increased microvascular perfusion heterogeneity under conditions of glycocalyx degradation

PAGE 135

124 (marginal energy costs of increasing hematocrit being inversely related to glycocalyx permeability ). In addition, this study is also the first to our knowledge to mechanistically link glycocalyx degradation to impaired tissue oxygenation. There are certain limitations of the analyses included in this study. The first is that the formulation of arteriol ar branching employed here does not necessarily represent real microvascular architecture (203 205) (166) may influence our results. However, simulations using alternative power scaling laws with powers both greater than and less than 3 yielded qualitatively similar results to those reported here, and the formulations used here have been validated elsewhere as ap proximations of arteriolar branching structure in a variety of tissues (167; 168) Moreover, our simplified vascular architecture lends itself to straightforward analysis of the complex interplay be tween blood viscosity and microvascular architecture, the results of which can be applied to tissue specific properties in future studies. In addition, our analysis of flow partitioning at an idealized capillary bifurcation could conceivably fail to genera lize to an entire capillary network. There is ample empirical evidence, however, that the predicted alterations in capillary perfusion do in fact occur if the glycocalyx is degraded (79 81) Finally, the empirical formulation of in vivo blood viscosity used in this study was derived from measurements in the rat mesentery (199) and may not generalize to other species or vascular beds. Based on our finding that the semi permeable nature the glycocalyx is sufficient to recapitulate key differences between the referenced in vivo and in vitro blood viscosity formulations, and given the similarity of our results to previous simulation resul ts (206 208) it appears likely that our findings represent a general property of the endothe lial glycocalyx and not an organ specific or species specific effect. Our finding that the endothelial

PAGE 136

125 glycocalyx limits microvascular perfusion heterogeneity by increasing the influence of RBC vessel wall interactions holds great promise for future mechan istic studies in a variety of disease states known to involve both glycocalyx degradation and tissue oxygenation defects.

PAGE 137

126 CHAPTER VII CO NCLUSIONS AND FUTURE DIRECTIONS 1. Abstract Based on the results and analyses discussed in previous chapters, heterogeneous blood flow distribution in the microcirculation likely accounts for 1) impaired tissue oxygenation independently of bulk blood flow, 2) impaired insulin delivery to skeletal muscle, an d 3) accumulation of metabolic wastes in peripheral tissues. In this chapter, I discuss future directions for development of these concepts. First, I propose a perfusion centric paradigm for inflammation linking these concepts and drawing connections to ot her, nominally unrelate d inflammatory disease states. In addition, f uture investigations will be necessary to further demonstrate the relevance of this paradigm to insulin action and to human diabetes, since all of my perfusion heterogeneity studies were p erformed in rodent models or in silico, and all focused primarily on oxygen transport. I also propose creation of a database of microvascular perfusion phenotypes suitable for engineering analysis of effects on mass transport (e.g. Chapter IV), and propose translation of the microvascular perfusion methods described in Chapter V for clinical use and show an early proof of concept In addition, I discuss potential targets for future microvascular perfusion therapies, including restoration of glycocalyx mech anical properties, existing perfusion targeting drugs, and the potential to re purpose existing drugs for use as microvascular therapies. In each case, understanding the potential future directions for this project requires thinking of inflammation in term s of functional/mechanical processes in the local tissue microenvironment rather than in terms of biochemical signaling cascades. Although the future directions proposed in this chapter are far reaching and unlikely to be performed by a

PAGE 138

127 single investigator they are all scientifically plausible and hold great potential for future innovations 2. A Perfusion Centric Paradigm for Inflammation T he key findings of this dissertation project have broad implications for not just diabetes, but for virtually inflammatory disease states G lycocalyx degradation is an inflammatory process that enables extravasation of inflammatory cells and macromolecules in response to injury or infection (221; 222) and the associated perfusion defects are a well documented microvascular response to inflammation (223) The theories developed in the course of my dissertation work were heavily informed by previous findings in sepsis and critical care Particularly enlightening were the clinical observations in critical care settings that 1) reduction in p erfused capill ary density predicts mortality (117) 2) biomarkers of glycocalyx degradation predict mortality (195) 3) hypoxia induced organ failure can occur despite maintenance of norma l bulk blood flow (107) and 4) interventions that improve microvascular perfusion may also improve patient outcomes (108) Collectively, these observations suggest that the glycocalyx degradation and impaired capillary perfusion observed in critical care patients repre sent s a more extreme version of the same phenomenon that occurs in skeletal muscle in diabetes. This inference is further supported by theoretical considerations discussed in Chapter III (e.g. compare Figure 12A to Figure 12B). Not only does inflammation itself cause impaired microvascular perfus ion and glycocalyx degradation (84; 197; 223) there are several mechanisms by which these microvascular effects in turn promote inflammation (summarized in Figure 33). First, tissue hypoxia can re sult from perfusion disr uption (63; 66; 84) and hypoxia triggers inflammatory signaling cascades (224) The same perfusion defects leading to tissue hypoxia can also cause

PAGE 139

128 impaired clearance of metabolic wastes and tissue acidosis (136) thus promoting inflammation through yet another mechanism (225) In addition, enzymatic glycocalyx degradation enables adhesion of inflammatory cell s to the capillary endothelium (221) which in turn triggers an oxidative burst and associated inflammatory signaling (226; 227) Finally, impaired microvascular perfusion would b e expected on both theoretical (66) and experimental (228) bases to cause insulin resistance, thus promoting inflammation and oxidative stress through yet another route (229) Notably, all four of these feed forward mediators (tissue hypoxia, accumulation of metabolic wastes, leukocyte adhesion, and insulin resistance) have been documented in both diabetes and sepsis (84; 103; 132; 135; 136; 145; 174; 230) albeit with different degrees of severity. Figure 33: A perfusion centric paradigm for inflammation.

PAGE 140

129 Glycocalyx degradation and microvascular perfusion heterogeneity participate in several feed forward inflammatory processes. Microvascular perfusion recovery represents a plausible strategy for the resolution of inflammation. Restoring normal microvascul ar perfusion could plausibly promote the resolution of inflammation by interfering with several feed forward inflammatory processes. This concept is not entirely novel (108; 213; 223) and the feed forward processes described in the previous paragraph are well established However, the overwhelming majority of ideas suggest ed by mainstream scientists to promote the resolution of inflammation focus on immunological cell signaling processes (231) while therapies that act in part through the microcirculation are often discovered inadvertently through strategies intended to target other par ameters (e.g. (71; 123) ) A perfusion centric mental paradigm for thinking about inflammation thus h as the potential to inspire novel therapeutic approaches to the resolution of inflammation that would b e unlikely to result from traditional immunology centric paradigms. It is worth noting that this paradigm is not on ly applicable to just the diseases discuss ed in detail here Beyond diabetes and sepsis, m icrovascular perfusion defects have also been reported in (110) lupus (232) cancer (233) ischemia reperfusion (115) and (109) to name a few Not only are microvascular perfusion defects seemingly ubiquitous in disease, there are indications in the literature t hat strategies for improving perfusion with or without targeting standard immunological cell signaling pathways c ould be effective in controlling inflammation For example, activation of the CB2 receptor (which acts in part through the vascular endothelium (234; 235) ) has been shown impr ove perfused capillary density (236; 237) Agonists of the CB2 receptor have also been reported limit inflammatory processes in the retina (238) kidneys (239) liver (240) and

PAGE 141

130 intestines (241) CB2 agonists have also been reported to arrest the onset of type 1 diabetes (242) and even appear to prevent several processes contributing diabetic complications (243 245) Similarly, activated Protein C not only improves mic rovascular perfusion in sepsis (108) but may also improve patient survival (146) although this effect is not always observed (246) Activated protein C also protects against diabetic nephropathy (247) and improv e s survival of pancreatic islet transplant s (248) For both CB2 agonists and activated protein C targets outside of the usual immunological cell signaling pathway s appear to improve capillary perfusion and may produce surprisingly broad anti inflammatory benefits P revious investigations into these and other related pathways have focused on biochemistry and cell signaling, but it is likely that the mass transport effect s of improved microvascular perfusion (e.g. more efficient delivery and clearance of small molecules) also contribute to their benefit s In light of the multiple feed forward mechanisms linking inflammation to impaired microvascular perfusion discuss ed above, these reports should not be entirely surprising. I contend that the scarcity of clinically available microvascular perfusion targeting drugs at present results from 1) a neglect of microcirculatory parameters in mainstream immunology, and 2) a la ck of cl arity on what would constitute success in microvascular perfusion therapy and how to measure it Beyond characterizing a contributing cause to exercise dysfunction in T2DM, t his dissertation also define s success for microvascular perfusion therapies (Chapter III) and enable s comprehensive characterization of their effectiveness (Chapter V). These developments pave the way for more direct future investigation s of potential therapies targeting the microcirculation

PAGE 142

131 3. Clinical and Laboratory Translation of Microvascular Pe rfusion Findings 3.1 Establish ing relevance to insulin resistance The theory of microvascular perfusion and solute flux derived in Chapter III predicts that the same microvascular perfusion defects that interfere with skeletal muscle oxygenation in diabetes would also cause insulin resistance. True to form, insulin sensitivity and exercise capacity are correlated in type 1 diabetes, type 2 diabetes, and in health (5; 249) Additionally, tracer studies of glucose transport reveal that glucose delivery to skeletal muscle limits i nsulin action in diabetic mice (112 ) and insulin action can be improved by targeting the microcirculation (250) Finally insulin access to the interstit ium is impaired in diet induced obesity (137) consistent with impaired microvascular delivery of insulin and glucose even before the onset of overt T2DM T hese findings collect ively make a strong case that microvascular perfusion defects may play a ca usal role in insulin resistance as well as reduced exercise capacity However, there are a lternative explanations for insulin resistance most of which are center ed on cell signali ng within skeletal muscle myocytes. For example, the same inflammatory signaling processes that lead to glycocalyx degradation and heterogeneous microvascular perfusion also promote insulin re sistance at the cellular level (139) Insulin action is reduced in insulin resistant humans to an even greater degree th an is insul in delivery to skeletal muscle (251) sugg esting that this cellular level insulin resistance is itself clinically significant. It is not implausible that the observed correlations between insulin sen sitivity and exercise capacity (5; 249) are spurious correlations produced by upstream inflammation rather than upstream perfusion defect, a lthough this explanation would

PAGE 143

132 con tradict our published theories (66) and analyses of trace r studies from Wasserman et al (112; 228; 250) Further studies are warranted to distinguish the relative contributions of perfusion level and cellular level impediments to glucose uptake. In order to establish the relevance of this paradigm to insulin resistance, several findings would be required. First, a cute disruption of microvascular perfusion in healthy muscle must cause impaired transport of both oxygen and glucose. Reduced insulin action following glycocalyx degradation suggests that this may indeed be the case, at least wit h re spect to glucose transport (87) Studies of oxygen transport following acute glycocalyx degradation ar e warranted. Second, a cute microvascular perfusion rescue must improve the transport of both oxygen and glucose in insulin resistant muscle. This claim is well supported wi th respect to oxygen transport (25; 60; 63; 162) and preliminary results suggest that the same may be true for glucose trans port (250) Further studies of insulin sensitivity with microvascular perfusion therapy are warranted. Finally, t he degree of microvascular perfusion heterogeneity m ust correlate to the degree of oxygen/glucose transport disruption. Although this prediction is consistent with the correlation of exercise capacity and insulin resistance (5; 249) it has not yet been directly tested. 3.2 Establish ing relevance to human diabetes In this dissertation, I discuss findings of microvascular perfusi on heterogeneity in the GK rat (28) the obese Zucker rat (23 25; 162) and high fat fe d mice (Chapter V) as models of T2DM and/or insulin resistance and streptozotocin treated rats as a model of type 1 diabetes (27) I also show tha t glycocalyx degradation is theoretically sufficient to cause microvascular perfusion heterogeneity (Chapter VI), and discuss experiments showing precisely this effect (79 81) in addition to studies showing glycocalyx degradation in both

PAGE 144

133 di abetes and acute hype rglycemia (71 73; 75; 81; 217) Furthermore, the disconnect between blood flow and tissue oxygenation discussed at length in Chapters I and II would be expected if perfusion heterogeneity contributes to exercise impairment in T2DM. These observations collectively make a strong cas e for the presence of microvascular perfusion abnormalities in human T2DM. However, there remains the possibility that microvascular perfusion heterogeneity is unique to untreated diabetes, or to diabetes with complications. M ost people with T2DM are treated with metformin, which par tially restores the glycocalyx (217) Metformin also improves microvascular perfusion, and these effects may be p art of its mechanism of action (123) Moreover, studies monitoring microvascular perfusion during exercise in humans using indirect measures have reported effects from diabetes with microvascular complications, but not from uncomplicated diabetes (104) Even th ough reduced oxygen extraction (29; 59) and discrepancies between blood flow and skeletal muscle oxygenation (Chapter II) do occur in T2DM, these observations do not necessita te microvascular perfusion heterogeneity. Heterogeneous distribution of blood flow in macrovessels would have a qualitative ly similar effect, for example (68; 252) Although current evidence indicates that microvascular perfusion heterogeneity is likely to contribute to reduced exercise capacity in human T2DM, this proposition has not yet been conclusively proven. Demonstrating that microvascular perfusion heterogeneity occurs in dia betic human skeletal muscle likely to prove difficult, as capillary blood flow in human skeletal muscle cannot be directly observed. There are, however, several plausible way s to indirectly test this hypothesis. First, the sort of microvascular dysfunction discussed here occur s not only in skeletal muscle but also in the kidney (15) in the retina (62) and in the skin of the foot (61)

PAGE 145

134 These tissues have very little in common, so diabetes related microvascular dysfunction is likely to be sys temic. Direct observation of capillary blood flow and its distribution is possible in th e human sublingual circulation (253) thus providing an alternative site for perfusion measurement s In addition, perfusion heterogeneity can be measured in rats through analy sis of tracer washout kinetics (162) This technique could be translated for use in human studies Although skeletal muscle capillary blood flow cannot be directly observed in human subjects, some combination of tracer studies and observation of other capillary beds could be used to establish the relevance of microvascular perfusion heterogeneity to tissue oxygenation in human T2DM. 3.3 Creating a database of microvascular perfusion phenotypes Despite hundreds of published papers discussing m icrovascular/capillary perfusion, there are very few cases in which perfusion has been adequately characterized so as to allow fully constrained engineering analysis of perfusion effects on solute flux processes as in Chapter IV. Thus, while microvascular perfusion defects have been described in many different disease states, it is difficult to determine their quantitative importance. By advertising the software developed in Chapter V to potential collaborator s I hope to assemble a database comprehensively characterizing capillary blood flow and its distribution in a wide variety of disease states, tissues, animal models, and interventions. This database will represent an unprecedentedly detailed picture of when and where microvascular perfusion may contrib ute to pathology and what drugs may fix it. 3.4 Perfusion quantitation as a diagnostic tool The perfusion quantitation technique developed in Chapter V has the potential not only to validate the possibility of perfusion heterogeneity in humans, but also t o be used as a

PAGE 146

135 diagnostic tool in the clinic. The most immediate applications would be in critical care. Loss of perfused sublingual capillary density is predictive of mortal ity in this patient population (117) sublingual microscopes for clinical use are a lready available on the market (253) and yet sublingual capillary den sity is not routinely monitored. This is likely due in large part due to the difficulty of measurement. In principle, translation of the capillary perfusion meas urement techniques developed in this dissertation for use with side darkfield (SDF) microscopy could enable real time, automated monitoring of perfused capillary density in the clinic. I have established a business relationship with Microvision Medical (A msterdam, Netherlands) to combine the perfusion measurement and oxygen delivery calculation algorithms of chapters V and III, respectively, with their existing SDF microscopy system. Preliminary results using Microvision provided sample videos of healthy a nd unhealthy circulation (shown in Figure 34 below) indicate that this approach to measuring oxygen delivery in humans is sufficient to distinguish health vs disease. Future studies will be performed to validate these preliminary results in a more controll ed fashion.

PAGE 147

136 Figure 34: Proof of concept for clinical application of flow tracking and oxygen delivery algorithms. ( A ) Comparison of mean flow velocity in sample videos of healthy and unhealthy circulation. Mean flow velocity is significantly reduced in the unhealthy group (p=0.016). ( B ) Comparison of Krogh diffusion radius (a metric of capillary density) in sample videos of healthy and unhealthy circulation. Diffusion radius is significantly increased (i.e. capillary density is significantly decreased) in the unhealthy group (p=0.016) ( C ) Comparison of oxygen conductance (scales with oxygen delivery given fixed interstitial PO 2 ) in sample videos of healthy and unhealthy circulation. Oxygen conductance is significantly reduced in the unhealthy group (p<0. 0001), and this metric appears to better distinguish health vs disease than either flow velocity or diffusion radius alone. Although the diagnostic uses outside of critical care are more speculative, there is potential for at least two additional uses: 1 ) detection of systemic inflammation, and 2) assessment of response to therapies targeting the microcirculation. Current inflammation assays used in clinical practice typically involve measurement of specific biomarkers (e.g. C

PAGE 148

137 peptide) Given that there a re many, often r edundant inflammatory pathways it is possible that functional endpoints such as capillary perfusion would be able to detect systemic inflammation that current biochemical methods miss entirely. Returning to the perfusion centric paradigm f or inflammation discussed in section 2, it is also likely that microcirculation targeted drugs will reach market in the near future. Development of microvascular therapies has alre ady begun and is well underway (108; 213; 223) With tools in hand to directly measure the microcirculation in humans (methods of Chapter V adapted for SDF microscopy as shown in Figure 34 ) and theories for understanding what parameters matter in which contexts (Chapter III), it is likely that development of microcirculation targeted drugs could be substantially accelerated. 4. P otential Microvascular Perfusion Therapies 4.1 Mechanical restoration of endothelial glycocalyx Degradation of the endothelial glycocalyx appears to play a role in microvascular perfusion heterogeneity in diabetes, suggesting that restoration of the glycocalyx may help to improve insulin sensitivity and tissue oxygenation. Outside of diabetes, glycoc alyx degradation has been shown to confer worse patient outcomes in a variety of contexts (85; 195; 254) which c an be understood in terms of the paradigm suggested in section 2. A considerable amount of work has been dedicated to pursuing glycocalyx restoration therapies (213) So far, attempts at glycocalyx restoration has seen mixed success as a therapeutic strategy, but investigators in the field remain optimistic that a suitable str ategy for glycocalyx restoration would represent a powerful therapeutic advance Whereas previous strategies for glycocalyx restoration have f ocused on chemical reconstitution (213) the results discussed in Chapter VI suggest that restoring the mechanical

PAGE 149

138 properties of the glycocalyx may be a more productive approach. The semi permeable, high charge density nature of the glycocalyx confer its perfusion normalizing and anti adhesive effects (89; 221) and these alone could be very useful in clinical practice. Whereas many chemical constituents of the glycocalyx are also expressed elsewhere and thus may have significant off target effects it is plausible that a large, negatively charged polymer could be functionalized so as to bind specifically to the endothelium, thus restoring a negatively charged gel layer to the endothelial surface without ever accessing the extravascular space. Given the discovery of tissue sp (255) it is even possible that mechanical glycocalyx restoration could be perform ed in an organ specific manner. 4.2 Biochemical targets for perfusion therapy There are several classes of drugs already available that may improve microvascular perfusion. In particular, nitric oxide (NO) signaling is impaired in diabetes (256) this impairment contributes to microvascular dysfunction (257) and NO signaling can be a cutely target ed using nitrites (258) In addition, glucagon like peptide 1 (GLP 1) signaling is also altered in diabetes (259) also appears to inf luence microvascular perfusion (260) and can be targeted using a variety of GLP 1 analogues or agonists already approved for use in T2DM (261) GLP 1 effects on microvascular perfusion are at least partially independent of NO signaling (262) thus allowing two pathways to plausibly target microvascular perfusion with already approved drugs. Beyond strategies directly targeting the microcirculation, there is also the potential for therapies developed for other reasons to improve microvascular perfusion. For example, metformin im proves microvascular perfusion (123) par tially restores the glycocalyx (217) and reduces leukocyte adhesion in vivo (263) These effects are now thought to be part of

PAGE 150

139 m (123) Metformin sensitivity and although it slightly decreases exercise capacity in healthy ind ividuals (264) this effect is due to its effects on mitochondrial capacity (265) which likely mask any underlying improvements in oxygen delivery. Given the difficulty associated with measuring microcirculatory endpoints, it is likely that other drugs thought to operate through mechanisms outside the microcirculation could also have clinically significant effects on microvascular perfusion. 5. Conclusion Microvascular perfusion heterogeneity contributes to impaired tissue oxygenation in rodent models of T2DM. This finding is also probably relevant to insulin sensitivity, to human T2DM, and to a variety of solutes besides oxygen and insulin in a variety of disease states besides T 2DM. Futures studies will be required to better understand when, where, and why disruption of microvascular perfusion may play a causal role in disease, and how the pathology of impaired perfusion may be addressed. Continued development of new tools for as sessing microvascular perfusion in humans will substantially enable these future studies. Collectively, the findings discussed in this dissertation reframe the issue of exercise impairment with T2DM away from bulk blood flow or dysfunction of the muscle it self, and instead towards the question of distribution of blood flow within the muscle. Further validation of the theories developed for this project is warranted, but if nothing else, it is now clear that impaired microvascular perfusion can contribute to pathological mass transport issues such as impaired tissue oxygenation and insulin resistance.

PAGE 151

140 REFERENCES 1. Awotidebe TO, Adedoyin RA, Yusuf AO, Mbada CE, Opiyo R, Maseko FC. 2014. Comparative functional exercise capacity of patients w ith type 2 diabetes and healthy controls: a case control study. Pan Afr Med J 19:257 2. Grdal A, Kasikcioglu E, Yakal S, Bugra Z. 2015. Impact of diabetes and diastolic dysfunction on exercise capacity in normotensive patients without coronary artery dise ase. Diab Vasc Dis Res 12:181 8 3. Komatsu WR, Gabbay MA, Castro ML, Saraiva GL, Chacra AR, et al. 2005. Aerobic exercise capacity in normal adolescents and those with type 1 diabetes mellitus. Pediatr Diabetes 6:145 9 4. Komatsu WR, Barros Neto TL, Chacra AR, Dib SA. 2010. Aerobic exercise capacity and pulmonary function in athletes with and without type 1 diabetes. Diabetes Care 33:2555 7 5. Nadeau KJ, Zeitler PS, Bauer TA, Brown MS, Dorosz JL, et al. 2009. Insulin resistance in adolescents with type 2 di abetes is associated with impaired exercise capacity. J Clin Endocrinol Metab 94:3687 95 6. Regensteiner JG, Bauer TA, Reusch JE. 2005. Rosiglitazone improves exercise capacity in individuals with type 2 diabetes. Diabetes Care 28:2877 83 7. Church TS, Che ng YJ, Earnest CP, Barlow CE, Gibbons LW, et al. 2004. Exercise capacity and body composition as predictors of mortality among men with diabetes. Diabetes Care 27:83 8 8. Lyerly GW, Sui X, Lavie CJ, Church TS, Hand GA, Blair SN. 2009. The association betwe en cardiorespiratory fitness and risk of all cause mortality among women with impaired fasting glucose or undiagnosed diabetes mellitus. Mayo Clin Proc 84:780 6 9. McAuley PA, Myers JN, Abella JP, Tan SY, Froelicher VF. 2007. Exercise capacity and body mas s as predictors of mortality among male veterans with type 2 diabetes. Diabetes Care 30:1539 43 10. Kokkinos P, Myers J, Nylen E, Panagiotakos DB, Manolis A, et al. 2009. Exercise capacity and all cause mortality in African American and Caucasian men with type 2 diabetes. Diabetes Care 32:623 8 11. Nylen ES, Kokkinos P, Myers J, Faselis C. 2010. Prognostic effect of exercise capacity on mortality in older adults with diabetes mellitus. J Am Geriatr Soc 58:1850 4 12. Estacio RO, Regensteiner JG, Wolfel EE, J effers B, Dickenson M, Schrier RW. 1998. The association between diabetic complications and exercise capacity in NIDDM patients. Diabetes Care 21:291 5 13. Seyoum B, Estacio RO, Berhanu P, Schrier RW. 2006. Exercise capacity is a predictor of cardiovascular events in patients with type 2 diabetes mellitus. Diab Vasc Dis Res 3:197 201

PAGE 152

141 14. Yu CM, Lau CP, Cheung BM, Fong YM, Ho YY, et al. 2000. Clinical predicto rs of morbidity and mortality in patients with myocardial infarction or revascularization who underwent cardiac rehabilitation, and importance of diabetes mellitus and exercise capacity. Am J Cardiol 85:344 9 15. Bjornstad P, Cree Green M, Baumgartner A, M aahs DM, Cherney DZ, et al. 2015. Renal function is associated with peak exercise capacity in adolescents with type 1 diabetes. Diabetes Care 38:126 31 16. Montero D, Daz Caestro C. 2015. Endurance training and maximal oxygen consumption with ageing: Rol e of maximal cardiac output and oxygen extraction. Eur J Prev Cardiol 17. Montero D, Diaz Role of Maximal Cardiac Output and Oxygen Extraction. Medicine and science in sports and exercise 47:202 4 33 18. Wagner PD. 2006. Counterpoint: in health and in normoxic environment VO2max is limited primarily by cardiac output and locomotor muscle blood flow. J Appl Physiol (1985) 100:745 7; discussion 7 8 19. Gusso S, Hofman P, Lalande S, Cutfield W, Robinson E, Baldi JC. 2008. Impaired stroke volume and aerobic capacity in female adolescents with type 1 and type 2 diabetes mellitus. Diabetologia 51:1317 20 20. Lalande S, Gusso S, Hofman PL, Baldi JC. 2008. Red uced leg blood flow during submaximal exercise in type 2 diabetes. Med Sci Sports Exerc 40:612 7 21. Pinto TE, Gusso S, Hofman PL, Derraik JG, Hornung TS, et al. 2014. Systolic and diastolic abnormalities reduce the cardiac response to exercise in adolesce nts with type 2 diabetes. Diabetes Care 37:1439 46 22. Kingwell BA, Formosa M, Muhlmann M, Bradley SJ, McConell GK. 2003. Type 2 diabetic individuals have impaired leg blood flow responses to exercise: role of endothelium dependent vasodilation. Diabetes C are 26:899 904 23. Butcher JT, Goodwill AG, Stanley SC, Frisbee JC. 2013. Blunted temporal activity of microvascular perfusion heterogeneity in metabolic syndrome: a new attractor for peripheral vascular disease? Am J Physiol Heart Circ Physiol 304:H547 58 24. Butcher JT, Stanley SC, Brooks SD, Chantler PD, Wu F, Frisbee JC. 2014. Impact of increased intramuscular perfusion heterogeneity on skeletal muscle microvascular hematocrit in the metabolic syndrome. Microcirculation 21:677 87 25. Frisbee JC, Wu F, G oodwill AG, Butcher JT, Beard DA. 2011. Spatial heterogeneity in skeletal muscle microvascular blood flow distribution is increased in the metabolic syndrome. Am J Physiol Regul Integr Comp Physiol 301:R975 86 26. Frisbee JC, Goodwill AG, Frisbee SJ, Butch er JT, Wu F, Chantler PD. 2014. Microvascular perfusion heterogeneity contributes to peripheral vascular disease in metabolic syndrome. J Physiol

PAGE 153

142 27. Kindig CA, Sexton WL, Fedde MR, Poole DC. 1998. Skeletal muscle microcirculatory structure and hemodynami cs in diabetes. Respir Physiol 111:163 75 28. Padilla DJ, McDonough P, Behnke BJ, Kano Y, Hageman KS, et al. 2006. Effects of Type II diabetes on capillary hemodynamics in skeletal muscle. Am J Physiol Heart Circ Physiol 291:H2439 44 29. Baldi JC, Aoina JL Oxenham HC, Bagg W, Doughty RN. 2003. Reduced exercise arteriovenous O2 difference in Type 2 diabetes. J Appl Physiol (1985) 94:1033 8 30. Grandi AM, Piantanida E, Franzetti I, Bernasconi M, Maresca A, et al. 2006. Effect of glycemic control on left vent ricular diastolic function in type 1 diabetes mellitus. Am J Cardiol 97:71 6 31. Zabalgoitia M, Ismaeil MF, Anderson L, Maklady FA. 2001. Prevalence of diastolic dysfunction in normotensive, asymptomatic patients with well controlled type 2 diabetes mellit us. Am J Cardiol 87:320 3 32. Asbun J, Villarreal FJ. 2006. The pathogenesis of myocardial fibrosis in the setting of diabetic cardiomyopathy. J Am Coll Cardiol 47:693 700 33. Boonman de Winter LJ, Hoes AW, Cramer MJ, de Jongh G, Janssen RR, Rutten FH. 201 5. Prognosis of screen detected heart failure with reduced and preserved ejection fraction in patients with type 2 diabetes. Int J Cardiol 185:162 4 34. Lago RM, Singh PP, Nesto RW. 2007. Diabetes and hypertension. Nat Clin Pract Endocrinol Metab 3:667 35. Maahs DM, Kinney GL, Wadwa P, Snell Bergeon JK, Dabelea D, et al. 2005. Hypertension prevalence, awareness, treatment, and control in an adult type 1 diabetes population and a comparable general population. Diabetes Care 28:301 6 36. Khan F, Elhadd TA, Gr eene SA, Belch JJ. 2000. Impaired skin microvascular function in children, adolescents, and young adults with type 1 diabetes. Diabetes Care 23:215 20 37. Caballero AE, Arora S, Saouaf R, Lim SC, Smakowski P, et al. 1999. Microvascular and macrovascular re activity is reduced in subjects at risk for type 2 diabetes. Diabetes 48:1856 62 38. Jrvisalo MJ, Raitakari M, Toikka JO, Putto Laurila A, Rontu R, et al. 2004. Endothelial dysfunction and increased arterial intima media thickness in children with type 1 diabetes. Circulation 109:1750 5 39. Anderson RA, Evans ML, Ellis GR, Graham J, Morris K, et al. 2001. The relationships between post prandial lipaemia, endothelial function and oxidative stress in healthy individuals and patients with type 2 diabetes. Ath erosclerosis 154:475 83 40. Stehouwer CD, Henry RM, Ferreira I. 2008. Arterial stiffness in diabetes and the metabolic syndrome: a pathway to cardiovascular disease. Diabetologia 51:527 39

PAGE 154

143 41. Goodman JM, McLaughlin PR, Plyley MJ, Holloway RM, Fell D, et a l. 1992. Impaired cardiopulmonary response to exercise in moderate hypertension. Can J Cardiol 8:363 71 42. Park SW, Goodpaster BH, Strotmeyer ES, de Rekeneire N, Harris TB, et al. 2006. Decreased muscle strength and quality in older adults with type 2 dia betes: the health, aging, and body composition study. Diabetes 55:1813 8 43. Han TS, Feskens EJ, Lean ME, Seidell JC. 1998. Associations of body composition with type 2 diabetes mellitus. Diabet Med 15:129 35 44. Clark MG, Rattigan S, Clerk LH, Vincent MA, Clark AD, et al. 2000. Nutritive and non nutritive blood flow: rest and exercise. Acta Physiol Scand 168:519 30 45. Ingberg CM, Srnblad S, Palmr M, Schvarcz E, Berne C, Aman J. 2003. Body composition in adolescent girls with type 1 diabetes. Diabet Med 20:1005 11 46. Ingberg CM, Palmr M, Aman J, Arvidsson B, Schvarcz E, Berne C. 2004. Body composition and bone mineral density in long standing type 1 diabetes. J Intern Med 255:392 8 47. Reusch JE, Bridenstine M, Regensteiner JG. 2013. Type 2 diabetes mellitus and exercise impairment. Rev Endocr Metab Disord 14:77 86 48. Poitras VJ, Bentley RF, Hopkins Rosseel DH, LaHaye SA, Tschakovsky ME. 2015. Independent effect of type 2 diabetes beyon d characteristic comorbidities and medications on immediate but not continued knee extensor exercise hyperemia. J Appl Physiol (1985) 119:202 12 49. Kiely C, O'Connor E, O'Shea D, Green S, Egaa M. 2014. Hemodynamic responses during graded and constant loa d plantar flexion exercise in middle aged men and women with type 2 diabetes. J Appl Physiol (1985) 117:755 64 50. Rissanen AP, Tikkanen HO, Koponen AS, Aho JM, Peltonen JE. 2015. Central and peripheral cardiovascular impairments limit VO(2peak) in type 1 diabetes. Med Sci Sports Exerc 47:223 30 51. Kingwell BA, Formosa M, Muhlmann M, Bradley SJ, McConell GK. 2003. Type 2 Diabetic Individuals Have Impaired Leg Blood Flow Responses to Exercise. 52. Bauer TA, Reusch JE, Levi M, Regensteiner JG. 2007. Skeleta l muscle deoxygenation after the onset of moderate exercise suggests slowed microvascular blood flow kinetics in type 2 diabetes. Diabetes Care 30:2880 5 53. Huebschmann AG, Reis EN, Emsermann C, Dickinson LM, Reusch JE, et al. 2009. Women with type 2 diab etes perceive harder effort during exercise than nondiabetic women. Appl Physiol Nutr Metab 34:851 7 54. Saleh F, Mumu SJ, Ara F, Hafez MA, Ali L. 2014. Non adherence to self care practices & medication and health related quality of life among patients wit h type 2 diabetes: a cross sectional study. BMC Public Health 14:431

PAGE 155

144 55. Egan AM, Mahmood WA, Fenton R, Redziniak N, Kyaw Tun T, et al. 2013. Barriers to exercise in obese patients with type 2 diabetes. QJM 106:635 8 56. Skyrme Jones RA, Berry KL, O'Brien RC, Meredith IT. 2000. Basal and exercise induced skeletal muscle blood flow is augmented in type I diabetes mellitus. Clin Sci (Lond) 98:111 20 57. Skyrme Jones RA, O'Brien RC, Meredith IT. 2000. Vasodilator prostanoids, but not nitric oxide, may account for skeletal muscle hyperaemia in Type I diabetes mellitus. Clin Sci (Lond) 99:383 92 58. Regensteiner JG, Bauer TA, Huebschmann AG, Herlache L, Weinberger HD, et al. 2015. Sex differences in the effects of type 2 diabetes on exercise performance. Med Sci Sports Exerc 47:58 65 59. Zheng J, Hasting MK, Zhang X, Coggan A, An H, et al. 2014. A pilot study of regional perfusion and oxygenation in calf muscles of individuals with diabetes with a noninvasive measure. J Vasc Surg 59:419 26 60. Frisbee JC, Goodwill AG, Butcher JT, Olfert IM. 2011. Divergence between arterial perfusion and fatigue resistance in skeletal muscle in the metabolic syndrome. Exp Physiol 96:369 83 61. Boulton AJ, Scarpello JH, Ward JD. 1982. Venous oxygenation in the diabetic neuropathic f oot: evidence of arteriovenous shunting? Diabetologia 22:6 8 62. Hammer M, Vilser W, Riemer T, Mandecka A, Schweitzer D, et al. 2009. Diabetic patients with retinopathy show increased retinal venous oxygen saturation. Graefes Arch Clin Exp Ophthalmol 247:1 025 30 63. McClatchey PM, Wu F, Olfert IM, Goldman E, Reusch JEB, Frisbee JC. 2017. Impaired Tissue Oxygenation in Metabolic Syndrome Requires Increased Microvascular Perfusion Heterogeneity. Journal of Cardiovascular Translational Research In Press 64. Je spersen SN, stergaard L. 2012. The roles of cerebral blood flow, capillary transit time heterogeneity, and oxygen tension in brain oxygenation and metabolism. J Cereb Blood Flow Metab 32:264 77 65. Ostergaard L, Kristiansen SB, Angleys H, Frkir J, Micha el Hasenkam J, et al. 2014. The role of capillary transit time heterogeneity in myocardial oxygenation and ischemic heart disease. Basic Res Cardiol 109:409 66. McClatchey PM, Frisbee JC, Reusch JEB. 2017. A Conceptual Framework for Predicting and Addressi ng the Consequences of Disease Related Microvascular Dysfunction. Microcirculation In Press 67. Kalliokoski KK, Oikonen V, Takala TO, Sipil H, Knuuti J, Nuutila P. 2001. Enhanced oxygen extraction and reduced flow heterogeneity in exercising muscle in end urance trained men. Am J Physiol Endocrinol Metab 280:E1015 21

PAGE 156

145 68. Kalliokoski KK, Knuuti J, Nuutila P. 2004. Blood transit time heterogeneity is associated to oxygen extraction in exercising human skeletal muscle. Microvasc Res 67:125 32 69. Mrin P, Ande rsson B, Krotkiewski M, Bjrntorp P. 1994. Muscle fiber composition and capillary density in women and men with NIDDM. Diabetes Care 17:382 6 70. Groen BB, Hamer HM, Snijders T, van Kranenburg J, Frijns D, et al. 2014. Skeletal muscle capillary density and microvascular function are compromised with aging and type 2 diabetes. J Appl Physiol (1985) 116:998 1005 71. Broekhuizen LN, Lemkes BA, Mooij HL, Meuwese MC, Verberne H, et al. 2010. Effect of sulodexide on endothelial glycocalyx and vascular permeabilit y in patients with type 2 diabetes mellitus. Diabetologia 53:2646 55 72. Nieuwdorp M, Mooij HL, Kroon J, Atasever B, Spaan JA, et al. 2006. Endothelial glycocalyx damage coincides with microalbuminuria in type 1 diabetes. Diabetes 55:1127 32 73. Perrin RM, Harper SJ, Bates DO. 2007. A role for the endothelial glycocalyx in regulating microvascular permeability in diabetes mellitus. Cell Biochem Biophys 49:65 72 74. Gouverneur M, Berg B, Nieuwdorp M, Stroes E, Vink H. 2006. Vasculoprotective proper ties of the endothelial glycocalyx: effects of fluid shear stress. J Intern Med 259:393 400 75. Nieuwdorp M, van Haeften TW, Gouverneur MC, Mooij HL, van Lieshout MH, et al. 2006. Loss of endothelial glycocalyx during acute hyperglycemia coincides with end othelial dysfunction and coagulation activation in vivo. Diabetes 55:480 6 76. Van Teeffelen JW, Brands J, Stroes ES, Vink H. 2007. Endothelial glycocalyx: sweet shield of blood vessels. Trends Cardiovasc Med 17:101 5 77. Nieuwdorp M, Meuwese MC, Vink H, H oekstra JB, Kastelein JJ, Stroes ES. 2005. The endothelial glycocalyx: a potential barrier between health and vascular disease. Curr Opin Lipidol 16:507 11 78. Noble MI, Drake Holland AJ, Vink H. 2008. Hypothesis: arterial glycocalyx dysfunction is the fir st step in the atherothrombotic process. QJM 101:513 8 79. Cabrales P, Vazquez BY, Tsai AG, Intaglietta M. 2007. Microvascular and capillary perfusion following glycocalyx degradation. J Appl Physiol (1985) 102:2251 9 80. Desjardins C, Duling BR. 1990. Hep arinase treatment suggests a role for the endothelial cell glycocalyx in regulation of capillary hematocrit. Am J Physiol 258:H647 54 81. Zuurbier CJ, Demirci C, Koeman A, Vink H, Ince C. 2005. Short term hyperglycemia increases endothelial glycocalyx perm eability and acutely decreases lineal density of capillaries with flowing red blood cells. J Appl Physiol (1985) 99:1471 6

PAGE 157

146 82. Constantinescu AA, Vink H, Spaan JA. 2001. Elevated capillary tube hematocrit reflects degradation of endothelial cell glycocalyx by oxidized LDL. Am J Physiol Heart Circ Physiol 280:H1051 7 83. Brands J, Spaan JA, Van den Berg BM, Vink H, VanTeeffelen JW. 2010. Acute attenuation of glycocalyx barrier properties increases coronary blood volume independently of coronary flow reserve. Am J Physiol Heart Circ Physiol 298:H515 23 84. Chappell D, Westphal M, Jacob M. 2009. The impact of the glycocalyx on microcirculatory oxygen distribution in critical illness. Curr Opin Anaesthesiol 22:155 62 85. Svennevig K, Kolset SO, Bangstad HJ. 2006 Increased syndecan 1 in serum is related to early nephropathy in type 1 diabetes mellitus patients. Diabetologia 49:2214 6 86. Zhang C, Meng Y, Liu Q, Xuan M, Zhang L, et al. 2014. Injury to the endothelial surface layer induces glomerular hyperfiltratio n rats with early stage diabetes. J Diabetes Res 2014:953740 87. Eskens BJ, Mooij HL, Cleutjens JP, Roos JM, Cobelens JE, et al. 2013. Rapid insulin mediated increase in microvascular glycocalyx accessibility in skeletal muscle may contribute to insulin me diated glucose disposal in rats. PLoS One 8:e55399 88. Bjornstad P, Snell Bergeon JK, Rewers M, Jalal D, Chonchol MB, et al. 2013. Early diabetic nephropathy: a complication of reduced insulin sensitivity in type 1 diabetes. Diabetes Care 36:3678 83 89. Mc Clatchey PM, Schafer M, Hunter KS, Reusch JE. 2016. The endothelial glycocalyx promotes homogenous blood flow distribution within the microvasculature. Am J Physiol Heart Circ Physiol 311:H168 76 90. Beckman JA, Goldfine AB, Gordon MB, Garrett LA, Keaney J F, Creager MA. 2003. Oral antioxidant therapy improves endothelial function in Type 1 but not Type 2 diabetes mellitus. Am J Physiol Heart Circ Physiol 285:H2392 8 91. Mullan BA, Young IS, Fee H, McCance DR. 2002. Ascorbic acid reduces blood pressure and a rterial stiffness in type 2 diabetes. Hypertension 40:804 9 92. Strobel NA, Peake JM, Matsumoto A, Marsh SA, Coombes JS, Wadley GD. 2011. Antioxidant supplementation reduces skeletal muscle mitochondrial biogenesis. Med Sci Sports Exerc 43:1017 24 93. Rist ow M, Zarse K, Oberbach A, Klting N, Birringer M, et al. 2009. Antioxidants prevent health promoting effects of physical exercise in humans. Proc Natl Acad Sci U S A 106:8665 70 94. 2014. National Diabetes Statistics Report: Estimates of Diabetes and Its Burden in the United States, 2014. Atlanta, GA: U.S. Department of Health and Human Services, Centers For Disease Control And Prevention

PAGE 158

147 95. Narayan KM, Boyle JP, Thompson TJ, Sorensen SW, Williamson DF. 2003. Lifetime risk for diabetes mellitus in the Un ited States. JAMA 290:1884 90 96. Myers J, Prakash M, Froelicher V, Do D, Partington S, Atwood JE. 2002. Exercise capacity and mortality among men referred for exercise testing. N Engl J Med 346:793 801 97. Behnke BJ, Kindig CA, McDonough P, Poole DC, Sext on WL. 2002. Dynamics of microvascular oxygen pressure during rest contraction transition in skeletal muscle of diabetic rats. Am J Physiol Heart Circ Physiol 283:H926 32 98. Padilla DJ, McDonough P, Behnke BJ, Kano Y, Hageman KS, et al. 2006. Effects of T ype II diabetes on capillary hemodynamics in skeletal muscle. 99. van Beekvelt MC, Borghuis MS, van Engelen BG, Wevers RA, Colier WN. 2001. Adipose tissue thickness affects in vivo quantitative near IR spectroscopy in human skeletal muscle. Clin Sci (Lond ) 101:21 8 100. Davis ML, Barstow TJ. 2013. Estimated contribution of hemoglobin and myoglobin to near infrared spectroscopy. Respir Physiol Neurobiol 186:180 7 101. Welch BL. 1947. The generalization ofstudent's' problem when several different population variances are involved. Biometrika 34:28 35 102. Tagougui S, Leclair E, Fontaine P, Matran R, Marais G, et al. 2015. Muscle oxygen supply impairment during exercise i n poorly controlled type 1 diabetes. Med Sci Sports Exerc 47:231 9 103. Frisbee JC. 2001. Impaired dilation of skeletal muscle microvessels to reduced oxygen tension in diabetic obese Zucker rats. Am J Physiol Heart Circ Physiol 281:H1568 74 104. Womack L, Peters D, Barrett EJ, Kaul S, Price W, Lindner JR. 2009. Abnormal skeletal muscle capillary recruitment during exercise in patients with type 2 diabetes mellitus and microvascular complications. J Am Coll Cardiol 53:2175 83 105. Stansberry KB, Hi ll MA, Shapiro SA, McNitt PM, Bhatt BA, Vinik AI. 1997. Impairment of peripheral blood flow responses in diabetes resembles an enhanced aging effect. Diabetes Care 20:1711 6 106. Barton M, Cosentino F, Brandes RP, Moreau P, Shaw S, Lscher TF. 1997. Anatom ic heterogeneity of vascular aging: role of nitric oxide and endothelin. Hypertension 30:817 24 107. Langenberg C, Bellomo R, May C, Wan L, Egi M, Morgera S. 2005. Renal blood flow in sepsis. Crit Care 9:R363 74 108. De Backer D, Donadello K, Taccone FS, O spina Tascon G, Salgado D, Vincent JL. 2011. Microcirculatory alterations: potential mechanisms and implications for therapy. Ann Intensive Care 1:27

PAGE 159

148 109. Farkas E, Luiten PG. 2001. Cerebral microvascular pathology in aging and Alzheimer's disease. Prog Ne urobiol 64:575 611 110. Hatoum OA, Binion DG, Otterson MF, Gutterman DD. 2003. Acquired microvascular dysfunction in inflammatory bowel disease: Loss of nitric oxide mediated vasodilation. Gastroenterology 125:58 69 111. Levy BI, Schiffrin EL, Mourad JJ, A gostini D, Vicaut E, et al. 2008. Impaired tissue perfusion: a pathology common to hypertension, obesity, and diabetes mellitus. Circulation 118:968 76 112. Fueger PT, Bracy DP, Malabanan CM, Pencek RR, Granner DK, Wasserman DH. 2004. Hexokinase II overexp ression improves exercise stimulated but not insulin stimulated muscle glucose uptake in high fat fed C57BL/6J mice. Diabetes 53:306 14 113. Andersen P, Henriksson J. 1977. Capillary supply of the quadriceps femoris muscle of man: adaptive response to exer cise. J Physiol 270:677 90 114. Cabrales P, Vzquez BY, Tsai AG, Intaglietta M. 2007. Microvascular and capillary perfusion following glycocalyx degradation. J Appl Physiol (1985) 102:2251 9 115. Engler RL, Schmid Schnbein GW, Pavelec RS. 1983. Leukocyte capillary plugging in myocardial ischemia and reperfusion in the dog. Am J Pathol 111:98 111 116. Schwartz RS, Burke A, Farb A, Kaye D, Lesser JR, et al. 2009. Microemboli and microvascular obstruction in acute coronary thrombosis and sudden coronary death : relation to epicardial plaque histopathology. J Am Coll Cardiol 54:2167 73 117. Sakr Y, Dubois MJ, De Backer D, Creteur J, Vincent JL. 2004. Persistent microcirculatory alterations are associated with organ failure and death in patients with septic shock Crit Care Med 32:1825 31 118. Lillioja S, Young AA, Culter CL, Ivy JL, Abbott WG, et al. 1987. Skeletal muscle capillary density and fiber type are possible determinants of in vivo insulin resistance in man. J Clin Invest 80:415 24 119. Godsland IF, Croo k D, Proudler AJ, Stevenson JC. 2005. Hemostatic risk factors and insulin sensitivity, regional body fat distribution, and the metabolic syndrome. J Clin Endocrinol Metab 90:190 7 120. Hori Y, Gabazza EC, Yano Y, Katsuki A, Suzuki K, et al. 2002. Insulin r esistance is associated with increased circulating level of thrombin activatable fibrinolysis inhibitor in type 2 diabetic patients. J Clin Endocrinol Metab 87:660 5 121. Juhan Vague I, Thompson SG, Jespersen J. 1993. Involvement of the hemostatic system i n the insulin resistance syndrome. A study of 1500 patients with angina pectoris. The ECAT Angina Pectoris Study Group. Arterioscler Thromb 13:1865 73 122. Frisbee JC. 2005. Hypertension independent microvascular rarefaction in the obese Zucker rat model o f the metabolic syndrome. Microcirculation 12:383 92

PAGE 160

149 123. Wiernsperger NF, Bouskela E. 2003. Microcirculation in insulin resistance and diabetes: more than just a complication. Diabetes Metab 29:6S77 87 124. Renkin EM, Michel CC, Geiger SR. 1984. Microcirc ulation American Physiological Society 125. Renkin EM. 1968. Transcapillary exchange in relation to capillary circulation. J Gen Physiol 52:96 108 126. Bassett DR, Howley ET. 2000. Limiting factors for maximum oxygen uptake and determinants of endurance p erformance. Med Sci Sports Exerc 32:70 84 127. Buchanan J, Mazumder PK, Hu P, Chakrabarti G, Roberts MW, et al. 2005. Reduced cardiac efficiency and altered substrate metabolism precedes the onset of hyperglycemia and contractile dysfunction in two mouse models of insulin resistance and obesity. Endocr inology 146:5341 9 128. Iozzo P, Chareonthaitawee P, Rimoldi O, Betteridge DJ, Camici PG, Ferrannini E. 2002. Mismatch between insulin mediated glucose uptake and blood flow in the heart of patients with Type II diabetes. Diabetologia 45:1404 9 129. Robbin s JL, Jones WS, Duscha BD, Allen JD, Kraus WE, et al. 2011. Relationship between leg muscle capillary density and peak hyperemic blood flow with endurance capacity in peripheral artery disease. J Appl Physiol (1985) 111:81 6 130. AUSTRIAN R, McCLEMENT JH, RENZETTI AD, DONALD KW, RILEY RL, COURNAND A. 1951. Clinical and physiologic features of some types of pulmonary diseases with impairment of alveolar capillary diffusion; the syndrome of "alveolar capillary block". Am J Med 11:667 85 131. Kahn SE, Hull RL, Utzschneider KM. 2006. Mechanisms linking obesity to insulin resistance and type 2 diabetes. Nature 444:840 6 132. Carlson GL. 2003. Insulin resistance in sepsis. Br J Surg 90:259 60 133. Gutierrez G, Palizas F, Doglio G, Wainsztein N, Gallesio A, et al. 1992. Gastric intramucosal pH as a therapeutic index of tissue oxygenation in critically ill patients. Lancet 339:195 9 134. Vincent JL, Abraham E, Annane D, Bernard G, Rivers E, Van den Berghe G. 2002. Reducing mortality in sepsis: new directions. Crit Ca re 6 Suppl 3:S1 18 135. Scheuermann Freestone M, Madsen PL, Manners D, Blamire AM, Buckingham RE, et al. 2003. Abnormal cardiac and skeletal muscle energy metabolism in patients with type 2 diabetes. Circulation 107:3040 6 136. Creteur J, De Backer D, Sakr Y, Koch M, Vincent JL. 2006. Sublingual capnometry tracks microcirculatory changes in septic patients. Intensive Care Med 32:516 23 137. Broussard JL, Castro AV, Iyer M, Paszkiewicz RL, Bediako IA, et al. 2016. Insulin access to skeletal muscle is impaire d during the early stages of diet induced obesity. Obesity (Silver Spring) 24:1922 8

PAGE 161

150 138. Chinkes D, deMelo E, Zhang XJ, Traber DL, Wolfe RR. 1995. Increased plasma glucose clearance in sepsis is due to increased exchange between plasma and interstitial fl uid. Shock 4:356 60 139. Vuorinen Markkola H, Koivisto VA, Yki Jarvinen H. 1992. Mechanisms of hyperglycemia induced insulin resistance in whole body and skeletal muscle of type I diabetic patients. Diabetes 41:571 80 140. Haber CA, Lam TK, Yu Z, Gupta N, Goh T, et al. 2003. N acetylcysteine and taurine prevent hyperglycemia induced insulin resistance in vivo: possible role of oxidative stress. Am J Physiol Endocrinol Metab 285:E744 53 141. Ito H. 2006. No reflow phenomenon and prognosis in patients with ac ute myocardial infarction. Nat Clin Pract Cardiovasc Med 3:499 506 142. Hirashima O, Kawano H, Motoyama T, Hirai N, Ohgushi M, et al. 2000. Improvement of endothelial function and insulin sensitivity with vitamin C in patients with coronary spastic angina: possible role of reactive oxygen species. J Am Coll Cardiol 35:1860 6 143. Evans JL, Goldfine ID. 2000. Alpha lipoic acid: a multifunctional antioxidant that improves insulin sensitivity in patients with type 2 diabetes. Diabetes Technol Ther 2:401 13 144 Faure P, Rossini E, Lafond JL, Richard MJ, Favier A, Halimi S. 1997. Vitamin E improves the free radical defense system potential and insulin sensitivity of rats fed high fructose diets. J Nutr 127:103 7 145. Sair M, Etherington PJ, Peter Winlove C, Evan s TW. 2001. Tissue oxygenation and perfusion in patients with systemic sepsis. Crit Care Med 29:1343 9 146. Bernard GR, Vincent JL, Laterre PF, LaRosa SP, Dhainaut JF, et al. 2001. Efficacy and safety of recombinant human activated protein C for severe sep sis. N Engl J Med 344:699 709 147. Berger MM, Chiolro RL. 2007. Antioxidant supplementation in sepsis and systemic inflammatory response syndrome. Crit Care Med 35:S584 90 148. Xie M, Shan Z, Zhang Y, Chen S, Yang W, et al. 2014. Aspirin for primary prevention of cardiovascular events: meta analysis of randomized controlled trials and subgroup analysis by sex and diabetes status. PLoS One 9:e90286 149. Martin Montalvo A, Mercken EM, Mitchell SJ, Palacios HH, Mote PL, et al. 2013. Metformin improves healthspan and lifespan in mice. Nat Commun 4:2192 150. Poole DC, Copp SW, Ferguson SK, Musch TI. 2013. Skeletal muscle capillary function: contemporary observations and novel hypothese s. Exp Physiol 98:1645 58 151. Boudreau DM, Malone DC, Raebel MA, Fishman PA, Nichols GA, et al. 2009. Health care utilization and costs by metabolic syndrome risk factors. Metab Syndr Relat Disord 7:305 14

PAGE 162

151 152. Shamseddeen H, Getty JZ, Hamdallah IN, Ali M R. 2011. Epidemiology and economic impact of obesity and type 2 diabetes. Surg Clin North Am 91:1163 72, vii 153. Aleixandre de Artiano A, Miguel Castro M. 2009. Experimental rat models to study the metabolic syndrome. Br J Nutr 102:1246 53 154. Fellmann L, Nascimento AR, Tibiria E, Bousquet P. 2013. Murine models for pharmacological studies of the metabolic syndrome. Pharmacol Ther 137:331 40 155. Tofovic SP, Jackson EK. 2003. Rat models of the metabolic syndrome. Methods Mol Med 86:29 46 156. F ang ZY, Sharman J, Prins JB, Marwick TH. 2005. Determinants of exercise capacity in patients with type 2 diabetes. Diabetes Care 28:1643 8 157. Regensteiner JG, Sippel J, McFarling ET, Wolfel EE, Hiatt WR. 1995. Effects of non insulin dependent diabetes on oxygen consumption during treadmill exercise. Med Sci Sports Exerc 27:875 81 158. Wong CY, O'Moore Sullivan T, Fang ZY, Haluska B, Leano R, Marwick TH. 2005. Myocardial and vascular dysfunction and exercise capacity in the metabolic syndrome. Am J Cardiol 96:1686 91 159. Boushel R, Gnaiger E, Calbet JA, Gonzalez Alonso J, Wright Paradis C, et al. 2011. Muscle mitochondrial capacity exceeds maximal oxygen delivery in humans. Mitochondrion 11:303 7 160. Frisbee JC, Delp MD. 2006. Vascular function in the met abolic syndrome and the effects on skeletal muscle perfusion: lessons from the obese Zucker rat. Essays Biochem 42:145 61 161. Frisbee JC, Butcher JT, Frisbee SJ, Olfert IM, Chantler PD, et al. 2016. Increased peripheral vascular disease risk progressively constrains perfusion adaptability in the skeletal muscle microcirculation. Am J Physiol Heart Circ Physiol 310:H488 504 162. Wu F, Beard DA, Frisbee JC. 2011. Computational analyses of intravascular tracer washout reveal altered capillary level flow distr ibutions in obese Zucker rats. J Physiol 589:4527 43 163. Krogh A. 1919. The rate of diffusion of gases through animal tissues, with some remarks on the coefficient of invasion. J Physiol 52:391 408 164. McGuire BJ, Secomb TW. 2004. Theoretical predictions of maximal oxygen consumption in hypoxia: effects of transport limitations. Respir Physiol Neurobiol 143:87 97 165. Pries AR, Ley K, Claassen M, Gaehtgens P. 1989. Red cell distribution at microvascular bifurcations. Microvasc Res 38:81 101 166. Murray CD 1926. THE PHYSIOLOGICAL PRINCIPLE OF MINIMUM WORK APPLIED TO THE ANGLE OF BRANCHING OF ARTERIES. J Gen Physiol 9:835 41

PAGE 163

152 167. Zamir M. 2001. Arterial branching within the confines of fractal L system formalism. J Gen Physiol 118:267 76 168. Zamir M, Medei ros JA. 1982. Arterial branching in man and monkey. J Gen Physiol 79:353 60 169. Beard DA, Schenkman KA, Feigl EO. 2003. Myocardial oxygenation in isolated hearts predicted by an anatomically realistic microvascular transport model. Am J Physiol Heart Circ Physiol 285:H1826 36 170. Prieto D, Contreras C, Snchez A. 2014. Endothelial dysfunction, obesity and insulin resistance. Curr Vasc Pharmacol 12:412 26 171. Frisbee JC, Goodwill AG, Frisbee SJ, Butcher JT, Brock RW, et al. 2014. Distinct temporal phases of microvascular rarefaction in skeletal muscle of obese Zucker rats. Am J Physiol Heart Circ Physiol 307:H1714 28 172. Gonzlez Alonso J, Calbet JA. 2003. Reductions in systemic and skeletal muscle blood flow and oxygen delivery limit maximal aerobic capa city in humans. Circulation 107:824 30 173. Flammer J, Orgl S, Costa VP, Orzalesi N, Krieglstein GK, et al. 2002. The impact of ocular blood flow in glaucoma. Prog Retin Eye Res 21:359 93 174. Baron AD, Laakso M, Brechtel G, Edelman SV. 1991. Mechanism of insulin resistance in insulin dependent diabetes mellitus: a major role for reduced skeletal muscle blood flow. J Clin Endocrinol Metab 73:637 43 175. De Backer D, Verdant C, Chierego M, Koch M, Gullo A, Vincent JL. 2006. Effects of drotrecogin alfa activ ated on microcirculatory alterations in patients with severe sepsis. Crit Care Med 34:1918 24 176. Klyscz T, Jnger M, Jung F, Zeintl H. 1997. [Cap image -a new kind of computer assisted video image analysis system for dynamic capillary microscopy]. Biomed Tech (Berl) 42:168 75 177. Demir SU, Hakimzadeh R, Hargraves RH, Ward KR, Myer EV, Najarian K. 2012. An automated method for analysis of microcirculation videos for accurate assessment of tissue perfusion. BMC Med Imaging 12:37 178. Dobbe JG, Streekstra G J, Atasever B, van Zijderveld R, Ince C. 2008. Measurement of functional microcirculatory geometry and velocity distributions using automated image analysis. Med Biol Eng Comput 46:659 70 179. Bauer A, Kofler S, Thiel M, Eifert S, Christ F. 2007. Monitorin g of the sublingual microcirculation in cardiac surgery using orthogonal polarization spectral imaging: preliminary results. Anesthesiology 107:939 45 180. Otsu N. 1975. A threshold selection method from gray level histograms. Automatica 11:23 7

PAGE 164

153 181. Kindig CA, Richardson TE, Poole DC. 2002. Skeletal muscle capillary hemodynamics from rest to contractions: implications for oxygen transfer. J Appl Physiol (1985) 92:2513 20 182. Rufaihah AJ, Huang NF, Jam S, Lee JC, Nguyen HN, et al. 2011. Endothelial c ells derived from human iPSCS increase capillary density and improve perfusion in a mouse model of peripheral arterial disease. Arterioscler Thromb Vasc Biol 31:e72 9 183. SCHMIDT NIELSEN K, PENNYCUIK P. 1961. Capillary density in mammals in relation to bo dy size and oxygen consumption. Am J Physiol 200:746 50 184. Barrett EJ, Eggleston EM, Inyard AC, Wang H, Li G, et al. 2009. The vascular actions of insulin control its delivery to muscle and regulate the rate limiting step in skeletal muscle insulin actio n. Diabetologia 52:752 64 185. Lam C, Tyml K, Martin C, Sibbald W. 1994. Microvascular perfusion is impaired in a rat model of normotensive sepsis. J Clin Invest 94:2077 83 186. Levin DL, Buxton RB, Spiess JP, Arai T, Balouch J, Hopkins SR. 2007. Effects o f age on pulmonary perfusion heterogeneity measured by magnetic resonance imaging. J Appl Physiol (1985) 102:2064 70 187. Quitter F, Figulla HR, Ferrari M, Pernow J, Jung C. 2013. Increased arginase levels in heart failure represent a therapeutic target to rescue microvascular perfusion. Clin Hemorheol Microcirc 54:75 85 188. Borges RC, Carvalho CR, Colombo AS, da Silva Borges MP, Soriano FG. 2015. Physical activity, muscle strength, and exercise capacity 3 months after severe sepsis and septic shock. Inten sive Care Med 41:1433 44 189. Fleg JL, Lakatta EG. 1988. Role of muscle loss in the age associated reduction in VO2 max. J Appl Physiol (1985) 65:1147 51 190. Wilson JR, Martin JL, Schwartz D, Ferraro N. 1984. Exercise intolerance in patients with chronic heart failure: role of impaired nutritive flow to skeletal muscle. Circulation 69:1079 87 191. O'Leary VB, Marchetti CM, Krishnan RK, Stetzer BP, Gonzalez F, Kirwan JP. 2006. Exercise induced reversal of insulin resistance in obese elderly is associated wi th reduced visceral fat. J Appl Physiol (1985) 100:1584 9 192. Kondo N, Nomura M, Nakaya Y, Ito S, Ohguro T. 2005. Association of inflammatory marker and highly sensitive C reactive protein with aerobic exercise capacity, maximum oxygen uptake and insulin resistance in healthy middle aged volunteers. Circ J 69:452 7 1 93. Hambrecht R, Fiehn E, Weigl C, Gielen S, Hamann C, et al. 1998. Regular physical exercise corrects endothelial dysfunction and improves exercise capacity in patients with chronic heart failure. Circulation 98:2709 15

PAGE 165

154 194. Richardson TE, Kindig CA, Musc h TI, Poole DC. 2003. Effects of chronic heart failure on skeletal muscle capillary hemodynamics at rest and during contractions. J Appl Physiol (1985) 95:1055 62 195. Neves FM, Meneses GC, Sousa NE, Menezes RR, Parahyba MC, et al. 2015. Syndecan 1 in Acut e Decompensated Heart Failure -Association With Renal Function and Mortality. Circ J 79:1511 9 196. Williamson KA, Hamilton A, Reynolds JA, Sipos P, Crocker I, et al. 2013. Age related impairment of endothelial progenitor cell migration correlates with str uctural alterations of heparan sulfate proteoglycans. Aging Cell 12:139 47 197. Burke Gaffney A, Evans TW. 2012. Lest we forget the endothelial glycocalyx in sepsis. Crit Care 16:121 198. Pries AR, Neuhaus D, Gaehtgens P. 1992. Blood viscosity in tube flow : dependence on diameter and hematocrit. Am J Physiol 263:H1770 8 199. Pries AR, Secomb TW, Gessner T, Sperandio MB, Gross JF, Gaehtgens P. 1994. Resistance to blood flow in microvessels in vivo. Circ Res 75:904 15 200. Pries AR, Secomb TW. 2005. Microvasc ular blood viscosity in vivo and the endothelial surface layer. Am J Physiol Heart Circ Physiol 289:H2657 64 201. Pries AR, Secomb TW, Gaehtgens P, Gross JF. 1990. Blood flow in microvascular networks. Experiments and simulation. Circ Res 67:826 34 202. Pr ies AR, Secomb TW. 2003. Rheology of the microcirculation. Clin Hemorheol Microcirc 29:143 8 203. Binder KW, Murfee WL, Song J, Laughlin MH, Price RJ. 2007. Computational Network Model Prediction of Hemodynamic Alterations Due to Arteriolar Remodeling in I nterval Sprint Trained Skeletal Muscle. Microcirculation 14:181 92 204. Huo Y, Kassab GS. 2012. Intraspecific scaling laws of vascular trees. J R Soc Interface 9:190 200 205. Sriram K, Intaglietta M, Tartakovsky DM. 2014. Hematocrit dispersion in asymmetri cally bifurcating vascular networks. Am J Physiol Heart Circ Physiol 307:H1576 86 206. Feng J, Weinbaum S. 2000. Lubrication theory in highly compressible porous media: the mechanics of skiing, from red cells to humans. Journal of Fluid Mechanics 422:281 3 17 207. Secomb TW, Hsu R, Pries AR. 1998. A model for red blood cell motion in glycocalyx lined capillaries. Am J Physiol 274:H1016 22 208. Secomb TW, Hsu R, Pries AR. 2001. Motion of red blood cells in a capillary with an endothelial surface layer: effect of flow velocity. Am J Physiol Heart Circ Physiol 281:H629 36

PAGE 166

155 209. Mathieu Costello O, Hoppeler H, Weibel ER. 1989. Capillary tortuosity in skeletal muscles of mammals depends on muscle contraction. J Appl Physiol (1985) 66:1436 42 210. Poole DC, Mathieu Costello O, West JB. 1989. Capillary tortuosity in rat soleus muscle is not affected by endurance training. American Journal of Physiology H eart and Circulatory Physiology 256:H1110 H6 211. Secomb TW, Pries AR. 2013. Blood viscosity in microvessels: experiment and theory. C R Phys 14:470 8 212. Weinbaum S, Tarbell JM, Damiano ER. 2007. The structure and function of the endothelial glycocalyx l ayer. Annu Rev Biomed Eng 9:121 67 213. Becker BF, Chappell D, Bruegger D, Annecke T, Jacob M. 2010. Therapeutic strategies targeting the endothelial glycocalyx: acute deficits, but great potential. Cardiovasc Res 87:300 10 214. Kim S, Kong RL, Popel AS, I ntaglietta M, Johnson PC. 2007. Temporal and spatial variations of cell free layer width in arterioles. Am J Physiol Heart Circ Physiol 293:H1526 35 215. Reitsma S, Slaaf DW, Vink H, van Zandvoort MA, oude Egbrink MG. 2007. The endothelial glycocalyx: comp osition, functions, and visualization. Pflugers Arch 454:345 59 216. Henry CB, Duling BR. 1999. Permeation of the luminal capillary glycocalyx is determined by hyaluronan. Am J Physiol 277:H508 14 217. Eskens BJ, Zuurbier CJ, van Haare J, Vink H, van Teeff elen JW. 2013. Effects of two weeks of metformin treatment on whole body glycocalyx barrier properties in db/db mice. Cardiovasc Diabetol 12:175 218. Bauer TA, Brass EP, Hiatt WR. 2004. Impaired muscle oxygen use at onset of exercise in peripheral arterial disease. J Vasc Surg 40:488 93 219. Maeda H, Wu J, Sawa T, Matsumura Y, Hori K. 2000. Tumor vascular permeability and the EPR effect in macromolecular therapeutics: a review. J Control Release 65:271 84 220. Mohsenifar Z, Jasper AC, Koerner SK. 1988. Rela tionship between oxygen uptake and oxygen delivery in patients with pulmonary hypertension. Am Rev Respir Dis 138:69 73 221. Constantinescu AA, Vink H, Spaan JA. 2003. Endothelial cell glycocalyx modulates immobilization of leukocytes at the endothelial su rface. Arterioscler Thromb Vasc Biol 23:1541 7 222. Vink H, Duling BR. 2000. Capillary endothelial surface layer selectively reduces plasma solute distribution volume. Am J Physiol Heart Circ Physiol 278:H285 9

PAGE 167

156 223. Granger DN, Senchenkova E. Inflammation and the Microcirculation. Proc. Colloquium Series on Integrated Systems Physiology: From Molecule to Function 2010 2:1 87: Morgan & Claypool Life Sciences 224. Eltzschig HK, Carmeliet P. 2011. Hypoxia and inflammation. N Engl J Med 364:656 65 225. Steen KH, Steen AE, Kreysel HW, Reeh PW. 1996. Inflammatory mediators potentiate pain induced by experimental tissue acidosis. Pain 66:163 70 226. Waddell TK, Fialkow L, Chan CK, Kishimoto TK, Downey GP. 1994. Potentiation of the oxidative burst of human neutrop hils. A signaling role for L selectin. J Biol Chem 269:18485 91 227. Detmers PA, Zhou D, Polizzi E, Thieringer R, Hanlon WA, et al. 1998. Role of stress activated mitogen activated protein kinase (p38) in beta 2 integrin dependent neutrophil adhesion and t he adhesion dependent oxidative burst. J Immunol 161:1921 9 228. Wasserman DH, Kang L, Ayala JE, Fueger PT, Lee Young RS. 2011. The physiological regulation of glucose flux into muscle in vivo. J Exp Biol 214:254 62 229. Esposito K, Nappo F, Marfella R, Gi ugliano G, Giugliano F, et al. 2002. Inflammatory cytokine concentrations are acutely increased by hyperglycemia in humans: role of oxidative stress. Circulation 106:2067 72 230. Barouch FC, Miyamoto K, Allport JR, Fujita K, Bursell SE, et al. 2000. Integr in mediated neutrophil adhesion and retinal leukostasis in diabetes. Invest Ophthalmol Vis Sci 41:1153 8 231. Serhan CN, Savill J. 2005. Resolution of inflammation: the beginning programs the end. Nat Immunol 6:1191 7 232. Plazak W, Pasowicz M, Kostkiewicz M, Podolec J, Tomkiewicz Pajak L, et al. 2011. Influence of chronic inflammation and autoimmunity on coronary calcifications and myocardial perfusion defects in systemic lupus erythematosus patients. Inflamm Res 60:973 80 233. Baxter LT, Jain RK. 1990. Tr ansport of fluid and macromolecules in tumors. II. Role of heterogeneous perfusion and lymphatics. Microvasc Res 40:246 63 234. Zoratti C, Kipmen Korgun D, Osibow K, Malli R, Graier WF. 2003. Anandamide initiates Ca(2+) signaling via CB2 receptor linked to phospholipase C in calf pulmonary endothelial cells. Br J Pharmacol 140:1351 62 235. Rajesh M, Mukhopadhyay P, Btkai S, Hask G, Liaudet L, et al. 2007. CB2 receptor stimulation attenuates TNF induced human endothelial cell activation, transendothelial migration of monocytes, and monocyte endothelial adhesion. American Journal of Physio logy Heart and Circulatory Physiology 293:H2210 H8 236. Toguri JT, Moxsom R, Szczesniak AM, Zhou J, Kelly ME, Lehmann C. 2015. Cannabinoid 2 receptor activation reduces leukocyte adhesion and improves capillary

PAGE 168

157 perfusion in the iridial microvasculature dur ing systemic inflammation. Clin Hemorheol Microcirc 61:237 49 237. Kianian M, Al Banna NA, Kelly ME, Lehmann C. 2013. Inhibition of endocannabinoid degradation in experimental endotoxemia reduces leukocyte adhesion and improves capillary perfusion in the g ut. J Basic Clin Physiol Pharmacol 24:27 33 238. Xu H, Cheng CL, Chen M, Manivannan A, Cabay L, et al. 2007. Anti inflammatory property of the cannabinoid receptor 2 selective agonist JWH 133 in a rodent model of autoimmune uveoretinitis. J Leukoc Biol 82: 532 41 239. Mukhopadhyay P, Rajesh M, Pan H, Patel V, Mukhopadhyay B, et al. 2010. Cannabinoid 2 receptor limits inflammation, oxidative/nitrosative stress, and cell death in nephropathy. Free Radic Biol Med 48:457 67 240. Lotersztajn S, Teixeira Clerc F, Julien B, Deveaux V, Ichigotani Y, et al. 2008. CB2 receptors as new therapeutic targets for liver diseases. Br J Pharmacol 153:286 9 241. Wright KL, Duncan M, Sharkey KA. 2008. Cannabinoid CB2 receptors in the gastrointestinal tract: a regulatory system i n states of inflammation. Br J Pharmacol 153:263 70 242. Weiss L, Zeira M, Reich S, Slavin S, Raz I, et al. 2008. Cannabidiol arrests onset of autoimmune diabetes in NOD mice. Neuropharmacology 54:244 9 243. Rajesh M, Mukhopadhyay P, Btkai S, Hask G, Lia udet L, et al. 2007. Cannabidiol attenuates high glucose induced endothelial cell inflammatory response and barrier disruption. Am J Physiol Heart Circ Physiol 293:H610 9 244. Rajesh M, Mukhopadhyay P, Btkai S, Patel V, Saito K, et al. 2010. Cannabidiol a ttenuates cardiac dysfunction, oxidative stress, fibrosis, and inflammatory and cell death signaling pathways in diabetic cardiomyopathy. J Am Coll Cardiol 56:2115 25 245. Horvth B, Mukhopadhyay P, Hask G, Pacher P. 2012. The endocannabinoid system and p lant derived cannabinoids in diabetes and diabetic complications. Am J Pathol 180:432 42 246. Abraham E, Laterre PF, Garg R, Levy H, Talwar D, et al. 2005. Drotrecogin alfa (activated) for adults with severe sepsis and a low risk of death. N Engl J Med 353 :1332 41 247. Isermann B, Vinnikov IA, Madhusudhan T, Herzog S, Kashif M, et al. 2007. Activated protein C protects against diabetic nephropathy by inhibiting endothelial and podocyte apoptosis. Nat Med 13:1349 58 248. Contreras JL, Eckstein C, Smyth CA, B ilbao G, Vilatoba M, et al. 2004. Activated protein C preserves functional islet mass after intraportal transplantation: a novel link between endothelial cell activation, thrombosis, inflammation, and islet cell death. Diabetes 53:2804 14

PAGE 169

158 249. Nadeau KJ, R egensteiner JG, Bauer TA, Brown MS, Dorosz JL, et al. 2010. Insulin resistance in adolescents with type 1 diabetes and its relationship to cardiovascular function. J Clin Endocrinol Metab 95:513 21 250. Ayala JE, Bracy DP, Julien BM, Rottman JN, Fueger PT, Wasserman DH. 2007. Chronic treatment with sildenafil improves energy balance and insulin action in high fat fed conscious mice. Diabetes 56:1025 33 251. Castillo C, Bogardus C, Bergman R, Thuillez P, Lillioja S. 1994. Interstitial insulin concentrations determine glucose uptake rates but not insulin resistance in lean and obese men. Journal of Clinical Investigation 93:10 252. Cano I, Roca J, Wagner PD. 2015. Effects of lung ventilation perfusion and muscle metabolism perfusion heterogeneities on maximal O2 transport and utilization. J Physiol 593:1841 56 253. Goedhart PT, Khalilzada M, Bezemer R, Merza J, Ince C. 2007. Sidestream Dark Field (SDF) imaging: a novel stroboscopic LED ring based imaging modality for clinical assessment of the microcirculation. Opt Express 15:15101 14 254. Vlahu CA, Lemkes BA, Struijk DG, Koopman MG, Krediet RT, Vink H. 2012. Damage of the endothelial glycocalyx in dialysis patients. J Am Soc Nephrol 23:1900 8 255. Renkonen J, Tynninen O, Hyry P, Paavonen T, Renkonen R. 2002. G lycosylation might provide endothelial zip codes for organ specific leukocyte traffic into inflammatory sites. Am J Pathol 161:543 50 256. Bucala R, Tracey KJ, Cerami A. 1991. Advanced glycosylation products quench nitric oxide and mediate defective endoth elium dependent vasodilatation in experimental diabetes. J Clin Invest 87:432 8 257. Frisbee JC. 2005. Reduced nitric oxide bioavailability contributes to skeletal muscle microvessel rarefaction in the metabolic syndrome. Am J Physiol Regul Integr Comp Phy siol 289:R307 R16 258. Lundberg JO, Gladwin MT, Ahluwalia A, Benjamin N, Bryan NS, et al. 2009. Nitrate and nitrite in biology, nutrition and therapeutics. Nat Chem Biol 5:865 9 259. Nauck MA, Vardarli I, Deacon CF, Holst JJ, Meier JJ. 2011. Secretion of g lucagon like peptide 1 (GLP 1) in type 2 diabetes: what is up, what is down? Diabetologia 54:10 8 260. Subaran SC, Sauder MA, Chai W, Jahn LA, Fowler DE, et al. 2014. GLP 1 at physiological concentrations recruits skeletal and cardiac muscle microvasculatu re in healthy humans. Clin Sci (Lond) 127:163 70 261. Buse JB, Rosenstock J, Sesti G, Schmidt WE, Montanya E, et al. 2009. Liraglutide once a day versus exenatide twice a day for type 2 diabetes: a 26 week randomised, parallel group, multinational, open la bel trial (LEAD 6). Lancet 374:39 47

PAGE 170

159 262. Nystrm T, Gonon AT, Sjholm A, Pernow J. 2005. Glucagon like peptide 1 relaxes rat conduit arteries via an endothelium independent mechanism. Regul Pept 125:173 7 263. Mamputu JC, Wiernsperger NF, Renier G. 2003. Antiatherogenic properties of metformin: the experimental evidence. Diabetes Metab 29:6S71 6 264. Braun B, Eze P, Stephens BR, Hagobian TA, Sharoff CG, et al. 2008. Impact of metformin on peak aerobic capacity. Appl Physiol Nutr Metab 33:61 7 265. Andrzejewski S, Gravel SP, Pollak M, St Pierre J. 2014. Metformin directly acts on mitochondria to alter cellular bioenergetics. Cancer Metab 2:12