Citation
Investigation of plasma membrane microdomains and their interactions with proteins using microscopic and analytical techniques

Material Information

Title:
Investigation of plasma membrane microdomains and their interactions with proteins using microscopic and analytical techniques
Creator:
Al-Junoori, Saif I.
Place of Publication:
Denver, CO
Publisher:
University of Colorado Denver
Publication Date:
Language:
English

Thesis/Dissertation Information

Degree:
Doctorate ( Doctor of philosophy)
Degree Grantor:
University of Colorado Denver
Degree Divisions:
College of Engineering and Applied Sciences, CU Denver
Degree Disciplines:
Engineering and applied science
Committee Chair:
Moldovan, Radu
Committee Members:
Lei, Tim
Benninger, Richard
Shepherd, Douglas P.
Dobrinskikh, Evgenia

Notes

Abstract:
Membranes are involved in almost every biological process and yet their organization and regulation are still not fully understood. Current membrane model, proposed by Simons and Van Meer in 1988, suggests that lipids form different microdomains within the bilayer and provide platforms for lipid-protein, protein-protein interactions and cell signaling processes. Since these membrane microdomains have small sizes – from 10 and up to 700 nm, consequently advanced microscopy techniques are necessary to investigate their properties and behaviors. In this dissertation, we used a combination of different spectroscopic and microscopic methods namely Generalized Polarization (GP), Fluorescence Correlation Spectroscopy (FCS), Fluorescence Lifetime Imaging Microscopy (FLIM), Förster Resonance Energy Transfer (FRET), and probabilistic GP-Lifetime (𝐺𝑃𝜏) to study microdomains fluidity effects on proteins dynamics and functions. Using Giant Unilamellar Vesicles (GUVs) made of native apical membranes, isolated from two regions of rat small intestine, we were able to resolve a mystery behind a function of an integrated membrane protein- sodium dependent phosphate co-transporter type 2b (NaPi2b), behavior. It was known for several years that this protein has more abundant expression in the jejunum, compare to the duodenum, but activity within these two regions was similar. Therefore, performing single point FCS measurements on both labeled NaPi2b and membrane microdomains in same location, we were able to show that jejunum has a unique NaPi2b protein cluster (pentamer), which resides in bigger microdomains compare to other clusters of NaPi2b. Measuring local fluidity of these microdomains, in contrast to mean fluidity of a whole membrane, we showed that they are more solid, compare to other microdomains, where NaPi2b reside, suggesting that these pentamers are inactive. Applying combination of FLIM-FRET techniques and fluidity measurements we were able to show that soluble Klotho binds specific lipid within rafts; disrupting them and shifting basolateral membrane in live cells towards more fluid state. Even though we could measure all of these parameters, visualization of the microdomains within the membranes was still challenging, given the resolution of the microscopy techniques. To improve membrane microdomains visualization, we developed a probabilistic 𝐺𝑃𝜏 method as a novel approach to enhance a contrast in intensity GP images.

Record Information

Source Institution:
University of Colorado Denver
Holding Location:
Auraria Library
Rights Management:
Copyright Saif I. Al-Juboori. Permission granted to University of Colorado Denver to digitize and display this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.

Downloads

This item is only available as the following downloads:


Full Text

PAGE 1

INVESTIGATION OF PLASMA MEMBRANE MICRODOMAINS AND THEIR INTERACTIONS WITH PROTEINS USING MICROSCOPIC AND ANALYTICAL TECHNIQUES by SAIF I. AL JUBOORI B.S., Nahrain University, 2008 M.S., University of Colorado Denver, 2013 A thesis submitted to the Faculty of the Graduate School of the Univer sity of Colorado in partial fulfi llment of the requirements for the degree of Doctor of Philosophy Engineering and Applied Science Program 201 8

PAGE 2

ii This thesis for the Doctor of Philosophy degree by Saif I. Al Juboori has been approved for the Engineering and Applied Science Program by Tim Lei, Advisor Ra du Moldovan , Chair Richard Benninger Douglas P. Shepherd Evgenia Dobrinskikh Date: May 12, 2018

PAGE 3

iii Al Juboori , Saif I. (Ph.D, Engineering and Applied Science Program ) Investigation of Plasma Membrane Microdomains and Their Interactions w ith Prote ins Using Microscopic a nd Analytical Techniques Thesis directed by Associate Professor Tim Lei ABSTRACT Membranes are involved in almost every biological process and yet their organization and regulation are still not fully understood. Current membrane model, proposed by Simons and Van Meer in 1988, suggests that lipids form different microd omains within the bilayer and provide platforms for lipid protein, protein protein interactions and cell signaling processes. Since these membrane microdomains have small sizes from 10 and up to 700 nm, consequently advanced microscopy techniques are nec essary to investigate their properties and behaviors. In this dissertation, we used a combination of different spectroscopic and microscopic methods namely Generalized Polarization (GP), Fluorescence Correlation Spectroscopy (FCS), Fluorescence Lifetime Imaging Microscopy (FLIM), Förster Resonance Energy Transfer (FRET), and probabilistic GP Lifetime ( ) to study microdomains fluidity effects on proteins dynamics and functions. Using Giant Unilamellar Vesicles (GUVs) made of native apical membranes, isolated from two regions of rat small intestine, we were able to resolve a mystery behind a function of an integrated membrane protein sodium dependent phosphate co transporter type 2b (NaPi2b), behavior. It was known for several years that this protein has more abundant expression in the jejunum, compare to the duodenum, but activity within these two regions was similar. Therefore, performing single point FCS measurements on both labeled NaPi2b and membrane microdomains in same location, we were able to show that jejunum has a unique NaPi2b

PAGE 4

i v protein cluster (pentamer), which resides in bigger microdomains compare to other clusters of NaPi2b. Measuring local fluidity of these microdomains, in contrast to mean fluidity of a whole membrane, we showed that the y are more solid, compare to other microdomains, where NaPi2b reside, suggesting that these pentamers are inactive. Applying combination of FLIM FRET techniques and fluidity measurements we were able to show that soluble Klotho binds specific lipid within rafts ; disrupting them and shifting basolateral membrane in live cells towards more fluid state. Even though we could measure all of these parameters, visualization of the microdomains within the membranes was still challenging, given the resolution of the microscopy techniques . To improve membrane microdomains visualization , we developed a probabilistic method as a novel approach to enhance a contrast in intensity GP images . The form and content of this abstract are approved. I recommend its publi cation. Approved: Tim Lei

PAGE 5

v I dedicate this work to my family.

PAGE 6

vi ACKNOWLEDGEMENTS I acknowledge the financial support by NIH/NIDDK 1K25DK095232 01A1 . I would like to thank Dr. Tim Lei and Dr. Evgenia Dobrinskikh for all their advising, consultation, encouragement and support that they have bestowed upon me throughout my Ph.D studies. Without their help, this work would not have been possible. Here, I would also like to thank Mieko Iwahashi, Veronica Hogg Cornejo, Eileen Sutherland, Dr. M oshe Levi, George Dalton , Sung Wan An , Nicole Nischan , Joonho Yoon , Donald W. Hilgemann , Jian Xie , Kate Luby Phelps , Jennifer J. Kohler , Lutz Birnbaumer , and Dr. Chou Long Huang for their collaboration. I am indeed grateful for having this wonderful opport unity to work with all of them. Furthermore, I extend my thanks to Dr. Radu Moldovan, Dr. Richard Benninger, and Dr. Douglas P. Shepherd for their valuable comments and feedback. Special thank you to Mr. Gregory Glazner for all his availability, dedication, technical training and helpful discussions. Finally, I love to express my deepest and most sincere gratitude towards my family, for if it was not for their big benevolent sacrifice, all the above would not have happene d.

PAGE 7

vii TABLE OF CONTENTS CHAPTER I. INTRODUCTION ................................ ................................ ................................ ........ 1 Biological Membranes ................................ ................................ ................................ .. 1 Main Hyp othesis and Aims ................................ ................................ ........................... 3 Ethics Statement and Animal Subjects ................................ ................................ ......... 4 Scientific Contributions ................................ ................................ ................................ 5 Dissertation Orientation ................................ ................................ ................................ 5 II. BACKGROUND ................................ ................................ ................................ .......... 7 Plasma Membrane Structure ................................ ................................ ......................... 7 Plasma Membrane Phases ................................ ................................ ............................. 7 Lipid Protein Interactions 10 Correlation between Protein Diffusion and Activity .. 12 Microdomains Vis 13 Microscopy Methods to Study Membrane Microdomains and T heir Interactions with Proteins ................................ ................................ ................................ ....................... 14 14 Fluorescence Lifetime Imaging Microscopy (FLIM) 18

PAGE 8

viii Fluorescence Lifetime Imaging Microscopy F rster Resonance T ransfer (FLIM 2 4 Fluorescence 2 8 3 2 III. TECHNICAL APPROACHES ................................ ................................ ................... 3 5 Brush Border Membranes Isolation ................................ ................................ ............ 35 Transport Activity Measurements ................................ ................................ ............... 36 Isolation of BBM Detergent resistant and Detergent sensitive Fractions .................. 36 Measurements of Cholesterol and Sphingomyelin in BBM s ................................ ...... 37 Western Blotting ................................ ................................ ................................ ......... 37 Immunohistochemistry ................................ ................................ ............................... 37 Giant Unilamellar Vesicles (GUVs) Electroformation ................................ ............... 38 Generalized Polarization (GP) Measurements ................................ ............................ 39 Spectroscopy ................................ ................................ ................................ ... 40 Microscopy for Isolated (Apical) Membrane Study (NaPi2b Study) ............. 4 1 Intensity Quantification 4 4 Microscopy for Intact (Basolateral) Membrane Study (sKL Study) .............. 4 4 Fluorescence Correlation Spectroscopy (FCS) and Photon Counting Histograms (PCH) Measurements and Analyses ................................ ................................ ........... 46

PAGE 9

ix Measurement of Membrane Order Using Fluorescence Lifetime Imaging Microscopy (FLIM) ................................ ................................ ................................ ........................ 5 1 Fluorescence Lifetime Imaging Microsco py (FLIM) Measurement of Förster Resonance Energy Transfer (FRET) Studies ................................ .............................. 5 2 Probabilistic GP Lifetime ( ) ................................ ................................ ................. 5 3 Noise Metrics and Measures 5 9 Detecti 61 6 5 6 6 Hypothesis 6 7 Lifetime GP Pixel to Pixel Statistical Analyses 6 8 Differences from the Mean and Two Dimensional (2D) Spatial Correlation.6 9 IV. EXPERIMENTAL RESULTS AND DATA ANALYSES ................................ ........ 72 Cell Plasma Membrane GP 3D ................................ ................................ .................. 73 Isolated Membrane Study ( Apical Membrane as an Example, NaPi2b Study) .......... 76 Impact ................................ ................................ ................................ ............. 76 Objective ................................ ................................ ................................ ......... 7 7 Results ................................ ................................ ................................ ............. 7 7

PAGE 10

x Despite NaPi2b Different Expression Levels in Rat Duodenum and Jejunum, the Co transporter Activity is Similar in Both Regions ...... 7 7 NaPi2b Resides in the Detergent Resistant Fractions in BBMs Isolated from Rat Duodenum and Jejunum ................................ ...................... 7 9 Jejunal Apical Membranes are More Fluid in Intact Native Rat Intestinal Tissues ................................ ................................ ................. 80 GUVs Made of Native BBMs Maintain The Properties of the Intact Small Intestines ................................ ................................ ................... 84 Jejunal BBMs Have More Distinct and Bigger Lipids Microdomains 85 Jejunum Has a Distinct Very Slowly Diffusing NaPi2b Subset ......... 8 8 NaPi2b Forms Larger Clusters in Jejunum ................................ ......... 91 Very Slow Diffusing NaPi2b Aggregates Reside in Unique More Solid Microdomains in Jejunal BBMs ................................ ......................... 94 Jejunal BBMs Form Bigger Microdomains ................................ ........ 9 7 Intact Membrane Study ( Basolateral Membrane as an Example, Klotho Study) ..... 100 Impact ................................ ................................ ................................ ........... 100 Objective ................................ ................................ ................................ ....... 10 0 Results ................................ ................................ ................................ ........... 101 sKL Binds Lipid Rafts and Modulates Lipid Organization within Rafts ................................ ................................ ................................ ........... 101

PAGE 11

xi Klotho Binds Lipid Rafts in Live Cells by Interacting with Raft Associated GM1 ................................ ................................ ................ 102 Klotho Decreases Membrane Order Analy zed by Generalized Polarization .. ... 105 Probabilistic GP Lifetime ( ) Method 106 Impact 106 Objective 10 7 Results .. 10 7 Intensity GP Measurements . 10 7 Single Channel Detection of di 4 ANEPPDHQ Dye Lifetime (TCSPC FLIM) 10 8 Single Channel Detection of di 4 ANEPPDHQ Dye Lifetime (Frequency domain FLIM) .. 10 9 Dual Channel Detection of di 4 ANE PPDHQ Dye Lifetime (T CSPC 1 11 Lifetime GP ( ) Method Contrast Enhancement . 1 14 Uncertainty Measurements .. 1 16 Lifetime GP ( ) Method Contrast Enhancement Revealed by 2D Spatial Correlation ....... 1 24

PAGE 12

xii V. DISCUSSION .. 12 9 VI. CONCLUSIONS AND FUTURE DIRECTIONS ... .. 1 34 REFERENCES ................................ ................................ ................................ ......... 1 37 APPENDI 14 6

PAGE 13

xiii LIST OF TABLES Table 2.1 Some of the fitting functions for FCS autocorrelation curves ... 9 3.1 Duodenal and jejunal diffusion coefficients recovered from five component fitting model for their corresponding autocorrelation curves. Diffusion coefficients D ar e expresse d in units 47 3. 2 Probabilistic GP Lifetime ( ) theoretical examples ... 5 8 4.1 Average diffusion coefficients and relative concentrations of NaPi2b for duodenal and jejunal GUVs measured using single point FCS. Diffusion coefficients D are expressed as mean ± SEM (a.u.). n is the number of data points 91 4.2 Photon Counting Histogram (PCH) measurements from duodenal and jejunal GUVs. The of NaPi2b diffusion components in the two regions. n is the number of data points ... 94 4.3 Average local area Generalized Polarization (GP) values for duodenal and jejunal GUVs. Mean local GP values were calculated from areas, where NaPi2b single poi nt FCS measurements taken from. Different NaPi2b diffusion species were correlated to their corresponding mean local area GP. n is the number of data points ... 9 7 4.4 Average diffusion coefficients of di 4 ANEPPDHQ for duodenal and jejunal GUVs measured using single point FCS. Diffusion coefficients D are expressed as mean ± SEM in .. 100

PAGE 14

xiv 4.5 Means and standard deviations of intensity GP, lifetime GP and lifetime GP difference ima ges. Data were presented as mean ± standard deviation. Standard deviation of lifetime GP method is twice as that of intensity GP 1 16

PAGE 15

xv LIST OF FIGURES Figure 2.1 Membrane phases and their characteristic properties. Figure was adapted from (1) .. ... . .. . 8 2.2 Lipid phase separation. Figure was adapted from (1) .. 9 2.3 Di 4 ANEPPDHQ emission intensity spectra. The dye fluoresces with a peak emission wavelength of ~560 nm (green) when residing in the liquid ordered phase, and ~620 nm (red) when residing in the liquid disordered phase. Figure was adapted fro m (2) 14 2.4 Fluorescence lifetime histograms for di 4 ANEPPDHQ in vesicles composed of DOPC and SM/Chol, 7:3. (Inset) Structure of the probe. Figure was adapted from (3) . ............. .. ... 16 2.5 Cholesterol modulation effects on membrane fluidity examined via Generalized Polarization (GP) measurements on GUVs made from isolated mouse kidney Endoplasmic Reticulum (ER) membranes. ER membranes have 0.413 (more fluid) and 0.00329 (more solid) mean GP values before and after cholesterol administration, respectively . 17 2.6 Cholesterol modulation effects on membrane fluidity examined v ia Fluorescence Lifetime Imaging Microscopy (FLIM) on GUVs made from DOPC at 37°C degrees using di 4 ANEPPDHQ. (A) Di 4 ANEPPDHQ fluorescence lifetime shifted to a longer lifetime (more solid) (purple circle on the corresponding phasor plot) after choleste rol administration. (B) Same conditions as (A) using lifetime fractional analysis .. 1 8 2.7 Jablonski diagram showing a molecule electronic states and transitions among them. and represent the ground state, the first excited state, the second excited state, and the first triplet state, respectively. Solid lines mark radiative transitions. Whereas, dashed

PAGE 16

xvi and dotted lines mark non radiative transitions and intersystem crossing, res pectively. Figure was adapted from (4) .. ... 1 9 2.8 Principle of TCSPC. Photons distribution is built up over time after the excitation pulses. Figure was adapted from (5) .. 2 1 2.9 Principles of Fluorescence Lifetime Imaging Microscopy (FLIM) techniques. (a)Time domain approach (b) Frequency domain approach. Figure was adapted from (6) 2 2 2.10 FLIM phasor plot representation in the polar coordinate system. and are the demodulation and the phase shift of the fluorescence emission signal, respectively 2 3 2.11 Illustration of FLIM FRET applications in studying membrane lipid domains. (A) The coexistence of lipid domains in domains formation and phase behavior experiments can be revealed by means of FRET of lipid probes that preferentially partition in the s ame domain. Similarly, colocalization of membrane interacting species (lipid lipid and lipid protein) in a specific lipid domain can be studied using FLIM FRET approach (B). Figure was adapted from (7) .. 2 6 2.12 Principles of Fluorescence Correlation Spectroscopy (FCS). (A) FCS confocal microscopy configuration. (B) Temporal fluctuations of the recorded fluorescence intensity signal. (C) Calculation of the autocorrelation function of the detected fluorescence signal. Figure was adapted from (8) 2 8 2.13 Timescales of different phenomena that can be monitored using fluorescence intensity autocorrelation analysis. Figure was adapted from (9) 30

PAGE 17

xvii 2.14 The concept of molecular brigh tness in PCH method. The larger the variance the bigger the molecules size for species with equal average intensity. Molecular brightness is the ratio of the fluorescence signal fluctuations to the average intensity of the signal. Figure was adapted from (10) ... 3 4 3.1 Microscopy methods illustration. (A) BBMs were isolated from the intestinal enterocytes were. For NaPi2b diffusion and membrane fluidity analysis GUVs were made from BBMs by electroformation. (B) GUVs were stained with di 4 ANEPPDHQ to analyze membrane microdomains. (C) GUVs were co stained with NaPi2b antibodies pre labeled with Alexa 647, and single point FCS measurements were taken to determine NaPi2b diffusion within the membranes ... 40 3.2 Probabilistic nature of di 4 ANEPPDHQ fluorescence emission spectra. GUVs, made from synthetic lipids, before (first row) and after (second row) cholesterol addition and HEK cells, before (third row) and after (fourth row) cholesterol depletion, were stained with di 4 ANEPPDHQ. For each channel number of photons was detected to test the hypothesis of different probability for photon detection emitting from membrane microdomains of different order (axis were kept the same for each condition). Di 4 ANEPPDHQ has a 60nm shift in the emis sion spectra when the dye resides in the two different phases (ordered phase and disordered phase); Ch1 and Ch2 correspond to these phases, respectively. Thus, we predicted that in channel 1 will be detected more photons from ordered microdomains; whereas channel 2 will have more photons from disordered microdomains. Cholesterol addition (more ordered membrane phases formation) led to the increase of photon detection in channel 1 (second row), whereas cholesterol depletion (fourth row) led to more photon de tected in channel 2, as was hypothesized 5 4

PAGE 18

xviii 3.3 Probability of detecting the same lifetimes in two detection channels is different. To demonstrate the dissimilarity of detecting the same fluorescence lifetimes in a two channel dete ction configuration, GUVs made of synthetic lipids DOPC before and after cholesterol addition were used. Frequency domain FLIM data were obtained, and datasets from the two different conditions were imported into the same phasor. Two cursors, blue and purp le, corresponding to two different fluorescence lifetimes on the phasor plot were selected to highlight pixels in both channels from the two datasets. We showed that pixels corresponding to the same lifetime were highlighted differently in each channel und er the two variant conditions ... 5 5 3.4 Probabilistic lifetime GP method rationale. Assumption is that most of the photons detected in Ch1 are photons from the ordered phase. While, most of the photons detected in Ch2 are ph otons from the disordered phase 56 3. 5 Noise metric measurements. Background and signal (membrane) Regions Of Interest (white boxes) were selected, from which the mean and the standard deviation (SD) for both regions were calculated .. 60 3. 6 Alexa 488 photon counts (left panel) and calculated histogram (right panel) in a solution 62 3. 7 Estimation of the fluorescence lifetime standard deviation using a dye with a known fluorescence lifetime in a solution. Alexa 488 dye in a solution was used. The dye fluorescence lifetime is 4.1 ns. Variations of the fluorescence lifetime as a function o f photon counts were estimated for two cases: higher photon counts and lower photon counts .. 64

PAGE 19

xix 3. 8 Verification of the variation in the fluorescence lifetime due to noise model. Alexa 488 dye in a solution was used. The dye fluorescence lifetime is 4.1 ns. The same two cases, higher photon counts and lower photon counts, of Figure 3.6 were considered. From the recovered lifetime histograms: =4.15 ns with =0.1 ns (compared to 0.098 ns estimated), and =4.18 ns with =0.15 ns (compared to 0.133 ns estimated), for higher photon counts and lower photon counts, respectively 6 5 3. 9 A schematic summary of lifetime GP random uncertainty generation procedure. A complete random uncertainty in the calculated lifetime GP images can be obtained by and pixel by pixel. The resultant matrices can then be added to and respective matrices causing variations in the recovered fluorescence lifetimes within 68 and 95 confidence intervals. Subsequently, images can be calculated by implementing the lifetime GP formula 6 7 3.10 Illustration of two dimensional (2D) spatial correlation concept. Lifetime GP method contrast enhancement implied through autocorrelation coefficient line profiles in a certa in direction 7 1 4.1 3D reconstruction of GP z stacks showing apical part of the cells. Control and treated with Klotho for various times HEK cells were stained with di 4 ANEPPDHQ. Intensity z stack images were collected and GP values were each condition. 3D reconstruction of not tresholded GP imag es was performed. Klotho treatment (middle panel) shifts apical cell membrane to more solid state, compare to control (left panel), and becomes closer to control after 1hr of treatment with Klotho (right panel) 74

PAGE 20

xx 4.2 3D reconstruction of GP z stacks showing basolateral part of the cells. Control and treated with Klotho for various times HEK cells were stained with di 4 ANEPPDHQ. Intensity z stack images were collected and GP values were each condition. 3D reconstruction of not tresholded GP images was performed. Klotho treatment (middle panel) shifts baso lateral cell membrane to more fluid state, compare to control (left panel), and this change is persistent after 1hr of treatment with Klotho (right panel) 74 4.3 Mean GP of the plasma membrane quantification of 3D reconstruction of GP z stacks. Mean GP for the membrane only over distance from coverslip was calculated. Control and treated with Klotho for various times HEK cells were stained with di 4 ANEPPDHQ. Intensity z stack i mages were collected and GP values were each condition. 3D reconstruction of not tresholded GP images was performed. Klotho treatment shifts apical cell membrane to more solid state, compare to control, and becomes closer to control after 1hr of treatment with Klotho. In contrast Klotho treatment shifts baso lateral cell membrane to more fluid state, compare to control, and this change is persistent after 1hr of treatment with Klotho 75 4.4 NaPi2b has similar activity in rat duodenum and jejunum despite different expression levels. Confocal images of rat duodenum (A) and jejunum (B) tissue sections showed higher signal for NaPi2b (green) in the microvilli of the jejunum. Microvilli are ou tlined with F Actin (red). (C). NaPi2b mean intensity was quantified in the microvilli of two intestinal regions. Jejunal mean intensity is significantly increased (p=0.0241), indicating higher level of NaPi2b expression. Comparative expression by WB (D) a nd corresponding densitometry (E) for NaPi2b levels in BBMs isolated from rat duodenums and jejunums showed that NaPi2b expression was significantly increased in the jejunal BBMs (p=0.0025). (F).

PAGE 21

xxi s similar in both regions. Bar = 2 m ... 7 8 4.5 NaPi2b resides in the detergent resistant fractions in BBMs isolated from rat duodenum and jejunum. Detergent resistant and soluble fractions were separated using standard pr otocol. Comparative analysis of DR and DS by WB for (A) NaPi2b (top) and Flotillin 1 (raft marker, bottom) showed presence of these two proteins in the DR fraction of both regions, whereas signal was almost absent in the DS fraction. Densitometry for DR fr actions showed significantly higher abundance of NaPi2b (B) and Flotillin 1 (C) in the jejunum (p=0.036 and p=0.007 respectively) ...................................................... ............................... 80 4.6 Jejunal microvilli are more fluid in live rat intestinal tissue. Live tissues were stained with di 4 ANEPPDHQ (fluidity sensitive dye) and imaged within 15 min after dissection to preserve tissue viability and membrane properties. Representative GP ima ges for rat duodenum (A) and jejunum (B) are shown. (C) GP values from each pixel were quantified. Surprisingly jejunal mean GP had significantly higher negative value, indicating more fluid membrane (p=0.008). Membrane fluidity range was 0.9 (blue, fluid ) to 0.9 (red, solid) indicated by the color map. Magnified regions from GP images for duodenum (D) and jejunum (E) showed distribution of pixels with different fluidity within the microvilli. Duodenum had more solid microdomains (yellow regions, black arr ows), whereas jejunum had more and bigger fluid microdomains (blue regions, white arrows). BBMs isolated form duodenum had more cholesterol (F) and sphingomyelin (G) content in agreement with more solid duodenal microvilli. Bar = 5 m ... 83

PAGE 22

xxii 4.7 GUVs made of native membranes maintain the properties of the intact intestines. GUVs made of BBMs isolated from rat duodenal (A, D) and jejunal (B, E) intestinal segments were co stained with di 4 ANEPPDHQ (fluidity sensitive dye) and Alexa 647 N aPi2b. Representative GP (A, B) and confocal (D, E) images are shown. (C) Jejunal mean GP had significantly higher negative value, indicating more fluid membrane (p=0.003). (F) Mean intensity showed significantly increased signal for NaPi2b (red) in the j ejunal GUV (p=0.025), indicating presence of higher level of NaPi2b. Bar = 20 m ... 85 4.8 Jejunal BBMs have more discrete lipid microdomains. GP images for GUVs made of rat duodenal (A) and jejunal (B) BBMs are shown. Corresponding histograms for these GUVs showed narrow distribution of pixels with different fluidity for the duodenum (C), whereas jejunum (D) had broader GP histogram indicating more distinct and bigger lipids microdomains. Membrane fluidity had a range of 0.9 (blue, fluid) to 0.9 ( red, solid) indicated by the color map under the histogram. (E) Average GP histograms for duodenal (black) and jejunal (red) GUVs. (F) Histograms difference showed that jejunum had a number of pixels with more fluid and more solid GP values compared to duo denum. Bar = 20 m 8 7 4.9 Jejunum has a distinct very slowly diffusing NaPi2b subset. (A) Single point FCS measurements were taken for each GUV labeled with Alexa 647 NaPi2b. Autocorrelation fluorescence data were plott ed for both duodenum (black dotted line) and jejunum (red dotted line), and fitted with two and one components fitting models, respectively (solid lines). (B) Duodenum had two significantly different diffusion components: slow (D 3 =5.63± 2 /sec ) and fast (D 4 =10.23±0.25 2 /sec ) (C) Jejunum had three significantly different diffusion components: slow (D 3 =3.26± 2 /sec ), fast

PAGE 23

xxiii (D 4 =11.30±0.40 2 /sec ) and very slow (D 5 =0.33±0.09 2 /sec ). (D) NaPi2b had two similar diffusion components in duodenum and jejunum BBMs (D 3 and D 4 ), and a unique very slow component (D 5 ) in jejunum. (E) NaPi2b relative concentration corresponding to the different diffusion components was calculated from autocorrelation fitting models. Similar diffusion components (D 3 and D 4 i n both regions) did not have significantly different relative concentrations (C 3 and C 4 ) suggesting similar transporter activity for these subsets, in reference to NaPi2b activity shown in Figure 1F. The very slow component (D 5 ) in jejunum had different re lative concentration (C 5 ) (p=0.006), inferring that C 5 could be accounted for NaPi2b abundance discrepancy (shown in Figure 4.4E), indicating that it could also be inactive 90 4.10 NaPi2b forms bigger clusters in jejunum . Representative fluorescence intensity profiles for NaPi2b clusters in duodenum (A) and jejunum (B) corresponding to each diffusion components are shown. The intensity fluctuates strongly as the fluorescent clusters diffuse in and out of the FCS measured volume over a period of time. In duodenum, D 3 and D 4 have similar fluorescence intensity fluctuations over time compared to their jejunal counterparts (D 3 and D 4 ), whereas jejunal D 5 has a stronger and slower intensity fluctuations suggesting bigger NaPi2b clusters. Photon Counting Histograms (PCH) of slow (D 3 , +), fast (D 4 , +), and very slow (D 5 , +) diffusing Alexa 647 NaPi2b components (whose fluorescence intensity profiles are given in (A) and (B)) were plotted for duodenum (C) and jejunum (D). PCH data were fitted to a single species fitting model (solid lines). The average molecular brightness for the different diffusion species were calculated from the fit for duodenum (E) and jejunum (F). The average molecular brightness for D 3 ( =9.64±0.35) and D 4 ( =11.34±0.54) in duodenum is about twice that of freely diffusing Alexa 647 NaPi2b

PAGE 24

xxiv antibodies D s ( =4.56±0.14) obtained from solution measurements suggesting dimerization. Jejunum had D 3 ( =14.96±0.62), D 4 ( =10.33±0.67), and D 5 ( =25.20±3.10) suggesting trimerization, dimerization, and oligomerization, respectively 93 4.11 Very slow diffusing NaPi2b aggregates reside in unique more solid microdomains in jejunal BBMs. GUVs were co stained with Alexa 647 NaP i2b and di 4 ANEPPDHQ. Single point FCS measurements were taken on GUVs made from duodenal (A) and jejunal (B) BBMs for both dyes at the allocated point positions numbered 1 through 4. NaPi2b diffusion coefficients were calculated from each single point measurement and compared to its di 4 ANEPPDHQ diffusion counterparts, in addition to their local area GP values, point by point. The data were summarized in (C) for duodenum and (D) for jejunum. Duodenum fast diffusion coefficients were corresponding to more fluid local environmen ts, whereas, jejunum slow diffusions were mostly pertaining to a more solid local environment. Average local area GP values for the slow, fast, and very slow diffusion components were obtained from all GUVs measurements for both duodenum (E) and jejunum (F ). (G) D 3 and D 4 NaPi2b subsets reside in a similar microdomains environment in both duodenum and jejunum, while jejunal D 5 has a significantly (p<0.001) different local area GP, suggesting that NaPi2b species reside in a more solid microdomains. Bar = 2 0 m .. 96 4.12 Jejunal BBMs form bigger microdomains. (A) Single point FCS measurements were taken for each GUV stained with fluidity sensitive dye di 4 ANEPPDHQ. Autocorrelation fluorescence data were plotted for both duodenum (black dotted line) and jejunum (red dotted line ) and fitted with one component fitting model (solid lines). Diffusion coefficients for di 4 ANEPPDHQ were organized according to the Alexa 647 NaPi2b diffusions taken

PAGE 25

xxv from the same point for each GUV. D di 4 could be correlated with the domain size (ref) the faster diffusion would correspond to the smaller size. (B) In the duodenum both, slow and fast diffusing, NaPi2b subsets reside in microdomains with similar diffusion coefficients (D 3di 4 2 /sec and D 4di 4 =1.65±0 2 /sec, respectively), indicating comparable and small sizes. (C) In the jejunum slow and fast diffusing NaPi2b components reside in microdomains with slower, compare to the duodenum, but still similar diffusion coefficients (D 3di 4 2 /sec and D 4di 4 2 /sec, respectively), whereas very slow NaPi2b species reside in even slower diffusing lipids (D 5di 4 2 /sec), indicating overall bigger microdomains 99 4.13 Klotho binds lipid rafts and alters lipid o rganization. ( A) Cells were stained with BODIPY FL 505/510 C5 GM1 (donor) with or without CholEsteryl BODIPY 542/563 C 11 (acceptor), excited with a two photon laser at 900 nm, and emission was collected at 506 594 nm using FLIM. Merged intensity images (sh owing green BODIPY GM1 with some internalization) and pseudocolor lifetime images are shown. Experiments were performed four times wit h similar results. (Scale bars, B ) Phasor plot of fluorescence lifetime histogram from cells stained with GM1 onl y (Top) and the trajectory of F RET between GM1 and cholesterol (Bottom). Blue circle marks lifetime for donor only, purple circle for donor + acceptor, and green circle for background a utofluorescence (from unstained cells). Phasor plot analysis showed FRE T efficiency 25%, with fractional contribution of lifetimes 51% from quenched do nor, 44% from unquenched donor, and 5% from background. (11) ... 102 4.14 Klotho GM1 interaction live cell membranes analyzed by FLIM FRET. (A) HEK cells were stained with BODIPY FL 505/510 C5 GM1 (donor; 100 nM). Li fetime for GM1 alone has longer values (blue color cursor regions with longer lifetime values). Lifetime for

PAGE 26

xxvi GM1 shifted to the shorter values (purple color cursor) in the presence of fluorophore labeled sKL (acceptor; 300 pM) over 10 min, indicating que nching of GM1 by sKL and FLIM FRET occurrence between GM1 and klotho. (B) Phasor plot analysis of FLIM FRET data showed FRET efficiency 20%, with fractional contribution of lifetimes 36% from quenched donor, 55% from unquenched donor, and 9% from backgro und. In the trajectory, blue circle marks lifetime for donor plus unlabeled klotho, purple circle for donor plus labeled klotho, and green circle for background autofluorescence. (C and D) Cells were pretreated with d rafts. GM1 lifetime after addition of fluorophore labeled sKL was comparable to GM1 only, indicating little to no FRET occurrence between GM1 and klotho. Shown is representative of three separate experiments with similar findings. (E) Addition of unlabel ed klotho does not cause shift to shorter values (vs. A), indicating no quenching of BODIPY GM1 fluorescence by unlabeled klotho. (Scale (11) .. 104 4.15 Klotho decreases membrane order analyzed by using a polarity sensitive membrane probe. ( A) Cells were incubated with Di 4 at 37 °C before imaging (5 s per image consecutively for total 300 images over the subsequent 25 min). Line plots show mean GP values for 300 single images over time. (B) GP was calculated by ratiometric measurement of the fluorescence intensity recorded in two reversibility of GP (taken at 30 min) after washout of sKL. (11) 106 4.16 Conventional intensity GP measurements. Cont HEK cells were stained with di 4 ANEPPDHQ. Conventional Intensity GP were calculated

PAGE 27

xxvii the left compared to control, indicating more fluid phase ... 10 8 4.17 Single channel detection di 4 ANEPPDHQ lifetime (TCSPC). Control and cholesterol 4 ANEPPDHQ. Lifetime was extracted for lifetime histogram shifted towards shorter lifetimes compared to control indicating more fluid phase . 10 9 4.18 Single channel detection di 4 ANEPPDHQ lifetime (Frequency domain). Fractional analysis for cells and GUVs was carried out from their c orresponding phasors. Calculated normalized histograms showed that cholesterol enrichment shifted the histogram towards histogram towards more fluid microdomains compared t o control in both conditions ... 1 10 4.19 Probabilities of detecting liquid ordered and liquid disordered fluorescence lifetimes. Percentages of the photon counts in each channel relative to the total number of the photon counts detected in both channels were calculated at each pixel for control HEK cells and after cholesterol depletion ... . 11 treated cells. Histograms were fitted to a normal Gaussian fitting model ... 1 13 4.21 Contrast enhancement of lifetime GP ( ) method over intensity GP measurements. Control HEK cells were stained with di 4. TCSPC FLIM data in two channels was collected and lifetime for each channel was extracted. Conventional GP (top row), lifetime GP (middle row) and lifetime GP ( ) diffe rence (las t row) were calculated. Histo gram for each

PAGE 28

xxviii calculation was plotted. We have found that the standard deviations of lifetime GP ( ) measurements were as twice as that of intensity GP measurements ... 15 4.22 Simulation results of nois y lifetime GP image of homogeneous sample implementing error propagation analysis model. In the ideal world, noise free lifetime GP images would have a standard deviation of 0. However, due to detection noise and variations in the recovered fluorescence li fetimes in both channels, real world lifetime GP images have a non zero standard deviation. Pseducolor ranges were kept the same . 18 4.23 Actual Lifetime GP standard deviation is about four times greater than Lifetime GP variation due to noise. Pseducolor ranges were kept the same 19 4.24 Biased uncertainty in lifetime GP images in the 68% confidence interval. Morphological changes in the organization of membrane m icrodomains due to uncertainties in the recovered lifetimes can be examined. Very subtle, if any, changes in the organization of t hese microdomains were observed . 20 4.25 Biased uncertainty in lifetime GP images in the 95% confidence interval. Morphological changes in the organization of membrane microdomains due to uncertainties in the recovered lifetimes can be examined. Very subtle, if any, changes in the organization of t hese microdomains were observed . 21 4.26 Randomness simulations show the robustness of the Lifetime GP method in the sense of maintaining standard deviation ... 1 22 4.27 KS test. Two sample Kolmogorov Smirnov test was used to statistically compute how significantly different (5% significance level) each lifetime GP pixel to its left, right, up and

PAGE 29

xxix down adjacent neighbors. Null hypothesis is the two pixels are coming from the same d istribution. The test returns 1 (yellow) when the null hypothesis is rejected, and 0 (purple) otherwise .. 23 4.28 P values of KS test image in Figure IV.27. P values were all below 0.005 level .. . 1 24 4.29 Differences from the means. Intensity versus lifetime GP images and their corresponding histograms comparison are presented. Intensity GP has the exact same ranges of lifetime GP. Results showed histogram broadening of lifetime GP method as a demonstrative way of contrast enhancement ... 25 4.30 2D spatial correlation after 2 by 2 spatial filtering (1 st FOV). The comparison of conventional Intensity (left top panel) and Lifetime GP (right top panel) images . Line profiles for both methods (bottom row) sho w that in Lifetime GP objects can be resolved over ~ 3 pixels compare to ~ 8 pixels in the intensity GP . 26 4.31 2D spatial correlation after 2 by 2 spatial filtering (2 nd FOV). The comparison of conventional Intensity (left top panel) and Lifetime GP (right top panel) images . Line profiles for both methods (bottom row) show that in Lifetime GP objects can be resolved over ~ 3 pixels compare to ~ 8 pixels in the intensity GP . 2 7 4.32 2D spatial correlation after 2 by 2 spatial filtering (3 rd FOV). The comparison of conventional Intensity (left top panel) and Lifetime GP (right top panel) images . Line profiles for both methods (bottom row) show that in Lifetime GP objects can be resolved over ~ 3 pixels compare to ~ 8 pixels in the intensity GP . 2 7

PAGE 30

xxx 5.1 Comparison of membrane microdomains organization to some of the previously proposed models. Part of the f igure was adapted form (12) .. 1 33 6.1 Membrane lifetime GP ( ) comparison with theoretical calculat ions. A simple comparison of lifetime GP values in a representative lifetime GP image with our initial hypothetical calculations suggested that those lipid microdomains, which are designated by white arrows and having about 0.16 lifetime GP value, may contain 50% ordered and 50% disordered populations in their corresponding pixels, given the rest of the assumed parameters used in the theoretical lifetime GP calcul ations presented in Chapter III . 36

PAGE 31

xxxi LIST OF ABBREVIATION S GP Generalized Polarization FCS Fluorescence Correlation Spectroscopy FLIM Fluorescence Lifetime Imaging Microscopy FRET Förster Resonance Energy Transfer FLIM FRET Fluorescence Lifetime Imaging Microscopy Förster Resonance Energy Transfer GUVs Giant Unilamellar Vesicles NaPi2b Sodium Dependent Phosphate co transporter type 2b 3D Three Dimensional HEK Human Embryonic Kidney (cells) L d liquid disordered L o liquid ordered solid gel TMDs Transm embrane Domains methyl cyclodextrin DRMs Detergent Resistant Membranes DSMs Detergent Soluble Membranes

PAGE 32

xxxii GPI glycosylphosphatidylinositol RICS Raster Image Correlation spectroscopy NaPi2a sodium dependent phosphate cotransporter type 2a TIRFM Total Internal Reflection Fluorescence Microscopy GPCRs G protein coupled receptors UV ultraviolet STED Stimulated Emission Depletion ER Endoplasmic Reticulum DOPC Dipalmitoylphosphatidylcholine FOV Field Of View TD time domain FD frequency domain TCSPC Time Correlated Single Photon Counting 2D Two Dimensional SP FRET Steady State Probe Partitioning FRET FRAP Fluorescence Recovery After Photobleaching SMT Single Molecule Tracking

PAGE 33

xxxiii SPT Single Particle Tracking SM Sphingomyelin CHO Cholesterol DLPC 1,2 Dilauroyl sn glycero 3 phosphorylcholine DPPC 1,2 Dipalmitoyl sn glycero 3 phosphorylcholine DSPC 1,2 Distearoyl sn glycero 3 phosphorylcholine PCH Photon Counting Histogram PSF Point Spread Function BBMs Brush Border Membranes BBMVs Brush Border Membrane Vesicles DR detergent resistant DS detergent sensitive OCT optimal cutting media Pt Platinum PMTs Photo Multiplier Tubes SD standard deviation CNR Contrast to Noise Ratio

PAGE 34

xxxiv SNR Signal to Noise Ratio RV random variables KS Kolmogorov Smirnov CDFs Cumulative Distribution Functions

PAGE 35

1 CHAPTER I INTRODUCTION Biological Membranes It has been accepted that biological membranes are, in general, involved in almost all cellular activities and functions. In the preface of An Introduction To Biological Membranes (13) virtually any biological process. Cell membranes are composed of a lipid bi layer, containing proteins that span the bilayer and/or interact with the lipids on either side of the two leaflets. All the components of the membrane are in constant flux, which in addition to membranes very small size (<10 nm in width) makes studying membranes dynamics very challenging (13) . Biological membranes usually have both static and dynamic components. S tatic component refers to the composition of membranes and their constituents (lipids and proteins) , and the dynamic component describes the interactions of these constituents over time, which in turn translates to biologic al functions. The dynamic component is what makes membranes so essential. The m lipid composition) is heterogeneous hence different lipid species form distinct membrane micro domains, which could be between a few to several hundred nanometers in size (14,15) . The way these domains are organized and distributed across the membrane is not random . T hey are serving particular purposes and functions : such as provide interaction sites with the cytoskeleton to maintain the structure and integrity of the membrane and form platforms for cell signaling. Some proteins

PAGE 36

2 preferentially reside in certain microdomains whose characteristics are insightful in exa mining their functions. One fundamental characteristic of the membrane microdomains is membrane fluidity , which is crucial for proteins regulation and activity. Due to the notion that proteins might change their properties in a synthetic lipid environment, we are mostly interested in studying native membranes and proteins in their intrinsic location . The ultimate goal of this study is to apply advanced microscopy and data analysis methods on membranes in their intact form. However, it is not always possible due to the different technical challenges . So, the question becomes, how can we deploy microscopy and data analysis techniques to deduce information about the activity and function of proteins of interest, in conjunction with the fluidity of their microdomains of residence? Apical membrane of epithelial cells is very packed and dynamic, which makes it hard to st udy in the intact form. So one of the possibilities is to deal with it after the isolat ion of native membrane containing proteins of intere st, as in the case of transmembrane protein (NaPi2b) study in this thesis , to form more accessible Unilamellar Vesicles (GUVs) , which have been extensively used as a biomimetic systems (16) . Even though investigating intact cell plasma membrane, as in the case of soluble protein (Klotho) study, is the desired goal in this dissertation, visualizing membrane microdomains still imposing a challenge , despite less packed structure of the basolatera l part of cells .

PAGE 37

3 Special attention needs to be drawn to the local membrane fluidity pertain ing to the membrane proteins, rather than dealing with average membrane fluidity, as if those domains were one entity. The contrast provided by current (conventional and advanced) fluidity measurement techniques is not good enough to visualize and measure local membrane microdomains fluidity under different conditions . Thereby, the re is a need to develop a method that enhances the contrast of current flui dity measurements. Main Hypothesis and Aims Given the three dimensional (3D) structure of differentiated epithelial cells, their plasma membranes are cat egorized as apical and basolateral membranes separated by tight junctions . Apical versus basolateral membranes are d istinct in their lipid composition (17,18) which can be reflected in dissimilar membrane fluidity and different respond to the same treatment. There is also considerable amount of evidence that l ocal membrane composition (fluidity) could contribute to the lateral protein segregation (19,20) and therefore can play an important role for membrane function and protein s regulation. Here, our m ain hypothesis is that local membrane fluidity gives a more detailed understanding of membrane protein s dynamics and function than global average fluidity of the membrane . Pinning down proteins of interest in their local environment (local fluidity) via imaging techniques is a formidable task. It requires the combination of fluore scently labeled proteins, e nvironment sensitive dyes, and ultimately good resolution and contrast imaging for precise localization of these proteins in their local environments. Thus, developing methods that can provide enhanced contrast is of prime importance. To test our hypothesis, we will perform the following aims:

PAGE 38

4 Aim1. Determine apical and basolateral membrane fluidities (average and local) and the respective protein dynamics using microscopy and data analysis techniques. Aim1.1. Investigate membrane protein distribut ion into the different membrane microdomains and how these microdomains fluidity can influence membrane protein diffusion ( function ) in isolated native membranes (NaPi2b protein in the apical membrane of the enterocytes as an example). Aim1.2. Determine th e target and the effect on membrane fluidity for soluble protein in native intact membranes (Klotho effects on basolateral membrane of human embryonic kidney (HEK) epithelial cells as an example ). Aim2. Develop a P robabilistic GP L ifetime method to enhance contrast for membrane microdomains visualization . Ethics Statement and Animal Subjects All procedures were performed in accordance with protocols approved by the Institutional Animal Care and Use Committee at the University of Colorado, Denver. Male Sprague Dawley rats (200 250 g) were obtained from Harlan Laboratories (Madison, WI). Anima ls were kept on a diet with normal phosphate levels (0.6% Pi by weight) for two weeks. Chow was formulated by and obtained from Envigo (Madison, WI). Animals were studied for phosphate uptakes in the isolated BBMs, specific protein expression levels and lo calization in the small intestine (duodenal and jejunal regions), NaPi2b dynamics and membrane fluidity.

PAGE 39

5 Scientific Contributions Portions of this dissertation were previously published in : 1) NaPi2b diffusion and activity could be regulated by membrane fl uidity in the microvilli of the small intestine American Societies for Experimental Biology (FASEB) Journal. (First author) Al Juboori SI, Lei T, Hogg Cornejo V, Sutherland E, Levi M, Dobrinskikh E. NaPi2b diffusio n and activity could be regulated by membrane fluidity in the microvilli of the small intestine. FASEB J [Internet]. 2016 Apr 1;30(1 Supplement):861.3 861.3. Available from: http://www.fasebj.org/content/30/1_Supplement/861.3.abstract 2) Soluble klotho bind s monosialoganglioside to regulate membrane microdomains and growth factor signaling the National Academy of Sciences of the United States of America (PNAS) Journal. (The first two co authors and I have equally contri buted to this work) Dalton G, An SW, Al Juboori SI, Nischan N, Yoon J, Dobrinskikh E, et al. Soluble klotho binds monosialoganglioside to regulate membrane microdomains and growth factor signaling. Proc Natl Acad Sci U S A [Internet]. 2017;114(4):752 7. Av ailable from: http://www.ncbi.nlm.nih.gov/pubmed/28069944 Dissertation Orientation In the following chapters, a succinct background on plasma membrane structure, plasma membrane phases, and micros copy methods to study membrane microdomains and their interactions with proteins are given (Chapter II). Detailed technical approaches to investigate plasma membrane microdomains and their interactions with proteins, as well as,

PAGE 40

6 the Probabilistic GP Lifeti me ( ) method are presented in Chapter III. Empirical results and data analysis of plasma membrane fluidity of differentiated cells, apical membrane study (NaPi2b results), basolateral membrane study (Klotho results), and GP Lifetime ( ) results are shown in Chapter IV. In Chapter V, a comparison to different membrane organization models will be discussed. C onclusions and future directions are presented in Chapter VI.

PAGE 41

7 CHAPTER II BACKGROUND Plasma Membrane Structure Lipids are the major constituents of biological membranes. They outnumber proteins by ~40 to 1 up to ~200 to 1, and are responsible for giving membranes their structure and defining their environment (13) . The now putative concept of membrane structure is based on the Fluid Mosaic Model presented by Singer and Nicolson in 1972 (21) . This model was conceived and had been developed into a fully established form over many years (22) . It describes biological membranes as lipids bilayers : two amphipathic lipid monolayers situated back to bac k with proteins dissolved in them . They suggested that lipids and proteins can laterally diffuse within the membrane. Ten year s later in 1982 Karnovsky et al. (23) suggested that protein s do not freely diffuse in the membrane, but rather restrained within the lipid domains, which may have functional significance. In 1988 Simons and Van Meer (24) introduced the lipid raft concept for biological membranes, which later le t to develop principles of membrane subcompartmentalization into different microdomains (phases). Plasma Membrane Phases Membrane phase behavior is defined by membrane lipids diffusion (translational diffusion coefficient, ), and their order (the range of motion of their hydrophobic acyl chains, S ) (1) . The state of motion and order of the bilayer lipids is a function of temperature and composition (25) . There are three well known orders of membrane phases: liquid disordered (liquid crystalline) , liquid ordered , and solid gel . bi layers are composed of higher concentration of lipids with unsaturated acyl chains and have lower

PAGE 42

8 concent ration of cholesterol. Therefore, they are loosely pa cked and thin, whereas b ilayers contain mainly lipids with saturated acyl chains, and thus they are closely packed and form a thicker bilayer. b ilayers, on the other hand, are composed of a mix of lipids with saturated and unsaturated acyl chains with higher concentration of cholesterol , compared to L d . lipid b ilayers form what are so (1,25) . A d iagram for these three membrane phases is shown in Figure 2 .1. Figure 2 .1 Membrane phases and their characteristic properties. Figure was adapted from (1) . L ipid rafts are membrane domains, more ordered than the rest of the membrane , and presently defined as small heterogeneous nanodomains ( 10 200nm) that are cholesterol and sphingolipid enriched and compartmentalize cellular processes (14) . They are highly dynamic in both their lateral mobility and association dissociation (14,15) . Membrane rafts hav e the potential to form even bigger structures (>300nm) upon clustering induced by rafts

PAGE 43

9 components interactions (14,15) . Domains formation and rafts regulation can happen, in addition to hydrophobic match or mismatch, through lipid lipid and lipid proteins interactions and rafts connections to the cortical actin cytoskeleton (15) . The current model of the cell membranes proposes that different lipid phase s coexist within the membrane, segregating into different membrane microdomains (Figure 2 .2) that are unique in their composition and biophysical properties (15) . A cyl chains of membrane lipids vary in th eir length resulting in different hydrophobic tail lengths for individual lipids. In order to avoid exposure to aqueous environment, lipids divide according to their hydrophobic acyl chain length (15) . Consequently the long, saturated acyl chains of sphingolipids allow them to pack tight and their difference from the kinked, unsaturated acyl chains of bulk membrane phospholipids leads to phase separation (26) filling any gaps in sphingolipid packing (27) . Figure 2 .2 Lipid phase separation. Figure was adapted from (1)

PAGE 44

10 Lipid Protein I nteraction s Lipids within the rafts form functional platforms for lipid protein, protein protein interactions and cellular signaling processes regulation (15,28) . Most t ransmembrane proteins have hydrophobic sites, so called Transmembrane Domains (TMDs) , that span the lipid bilayer . There are many suggestions that activity of these proteins can be influenced by physical properties of the membrane, such as lipid composition and order (29) . Different proteins have distinctive length of TMDs, and therefore they favorably partition in lipid environments with matching acyl chain lengths. For example , proteins with long TMDs tend to reside in long saturated lipid environment (rafts). Convers e ly, proteins with short TMDs tend to reside in short unsaturated (non raft) lipid environment (15) . When mismatc hing happens, i.e., TMDs length differs from the acyl chain length of their surrounding lipids, protein protein interactions are more favored, causing protein clustering (15,30) . Consequent ly proteins could be separated being selectively included or excluded from the rafts (27) . In this way, localization within the raft s c an serve to facilitate or obstruct protein interactions (26,28) or act as a protein scaffold while allowing diffusion (31) . It is well known that raft residing proteins can interact with the surrounding lipids and these i nteractions induce conformational changes of the proteins and can influence their functions and activity (32,33) . Despite substantial improvement in methodology and techniques for lipid raft research, there are still some weaknesses in precise determination of their size, structure, and composition . The most important question , which still induces debate , is application of detergents (l ike Triton X 100 ) for isolation of lipid rafts and methyl cyclodextrin (M CD) for extraction of cholesterol from cell membranes which could lead to formation of artificial

PAGE 45

11 complexes not existing in the native environment. Regardless these debates , usage of Triton X 100 and M CD are still the standard way s of rafts isolation and disruption . The first class of proteins reported to be localized within the Deter gent Resistant Membranes (DRMs, rafts) were glycosylphosphatidylinositol (GPI) anchored proteins (17,34) . Subsequently, plenty of proteins were reported to be recovered in the DRMs . Among them were activated immune receptors, such as the Immunoglobulin E (IgE) receptor, the T cell receptor (15,35) , and the B cell receptor (15,36) . Whereas these receptors were found in Detergent Soluble Membranes (DSMs) when cells were at rest; suggesting that raft residency for them was essential for active signaling (15,37 39) . Not just rafts can regulate the proteins, but p roteins themselves play a significant role in rafts regulation. Thus the presenelin 1 can induce lipid raft formation and decrease the membrane fluidity (40) . Some proteins ( such as amyloid precursor protein (41) , chain of the T cell antigen receptor (42) ) have cholesterol binding motifs which determine their distribution across the membrane . Other proteins bind specifically to glycosphingolipids ( such as a synuclein (43) , and mannose binding protein (44) ) or sphingomyelin ( such as lysenin (45) , COPI machinery protein 24 (46) and chain of the T cell antigen receptor (42) ) , which facilitate their potential recruitment to raft like domains (15,46,47) . Another way proteins can regulate domains formation is via transbilayer interactions between immobilized lipid species with long acyl chain in the inner leaflet and long acyl chain containing lipid anchored proteins in the outer leaflet, in the presence of cholesterol. For example, the formation of ordered (raft like) domains as a result of phosphatidylserine and glycosylphosphatidylinositol (GPI) anchored proteins interactions. The immobilization

PAGE 46

12 of PS lipids, in this c ase, is due to either direct interactions with actin cytoskeleton, or mediated by adaptor proteins that have both phosphatidylserine and actin binding domains (15) . Correlation between Protein Diffusion and Activity Diffusion is a major concept in cell biology. The size of cells is limited by diffusion; by a universal constant which represents the solute diffusion rate in water (13) . Solute could be any of the cellular constituents. Few studies have demonstrated that translational proteins diffusion in the membrane can be an indicator of their activity. U sing Raster Image Correlation spectroscopy (RICS), Mikuni et al. (48) have shown that the average diffusion coefficient of EGFP labeled human Glucocorticoid Recepto r (EGFP hGR) strongly and negatively correlated with its function: the hGR diffusion coefficient decrease was reflected in the high affinity binding to DNA. Koopman et al. (49) determined the relationship between mitochondrial activity and matrix protein diffusion, measured by F luorescence C orrelation S pectroscopy (FCS): inhib ition of EYFP expressed mitochondrial Complex I (CI) by rotenone correlated with 2 fold increase in the diffusion rate of the CI in human skin fibroblasts (49) . Another example where a protein activity was examined in light of its diffusion is renal sodium dependent phosphate cotransporter type 2a (NaPi 2a) in dietary potassium deficiency fold transport activity under potassium deficiency (50) . Recently, Total Internal Reflection Fluorescence Microscopy (TIRFM) in conjunction with single molecule tracking analysis was used to assess metabotropic glutamate receptor 3

PAGE 47

13 (mGluR3), a G protein coupled receptor (GPCRs), activity. I t was shown that the activity of mGluR3 quantitatively correlated to its average diffusion coefficient under different ligand conditions. Also, this observation was confirmed for other GPCRs activity. These observations let to conclude that the diffusion c oefficient is a good index for estimating GPCRs activities indifferent of G protein coupling selectivity, chemical properties of the ligands, and the phylogenetic groups (51) . Microdomain s Visualization Extra source of controversy in lipid raft research , which still has to be solved , is determined by the lack of suitable detection techniques in living cells. Given the resolution of the current conventional diffraction limited microscopy techniques (~250nm), vi sualization of individual rafts, which could be between 10 and 700nm in size, is extremely challenging. In the last decade a number of new techniques helping to estimat e raft size and their visualization has been developed and applied . They include s ingle molecule spectroscopy and microscopy techniques such as F luorescence C orrelation S pectroscopy (FCS) , Fluorescence Lifetime Imaging Microscopy (FLIM) , Fluorescence Lifetime Imaging Microscopy rster Resonance Energy Transfer (FLIM FRET) , S timulated E mission D epletion (STED ) and others techniques. However, even using these techniques the subwavelength lipid domains have never been directly visualized (52) . Despite the limitations these new techniques were able to confirm the existence of nanoscale cholesterol based assemblies of lipids and proteins in the membranes of living cells .

PAGE 48

14 In the next secti ons of this chapter, the underlying principles of the current microscopy techniques that have been used to investigate membrane f luidity and protein dynamics throughout the course of this dissertation are reviewed. Microscopy Methods to Study Membrane Microdomains and Their Interactions with Proteins Generalized Polarization (GP) Membrane fluidity is the measure of membrane phase behavior. Membrane fluidity can be measured using special polarity sensitive probes such as Laurdan (6 lauryl 2 dimethylamino napthalene) and di 4 ANEPPDHQ , which have a blue shift in emission when they re side in membrane liquid ord ered phase relative to membrane liquid disordered phase (2 ) (Figure 2 . 3 ) . Figure 2 . 3 Di 4 ANEPPDHQ emission intensity spectra. The dye fluoresces with a peak emission wavelength of ~560 nm (green) when residing in the liquid ordered phase, and ~620 nm (red) when residing in the liquid disordered phase . Figure was adapted from (2) .

PAGE 49

15 Di 4 ANEPPDHQ has been the dye of choice throughout our studies, for it does not require multiphoton excitation, thus imaging can be achieved using standard confocal microscopy. Di 4 ANEPPDHQ is more photostable and less susceptible to photobleaching, and therefore more suitable for time lapse imaging (2) , and less phototoxic to cells compared to Laurdan which requires UV excitation. Utilizing the spectral shift property of the dye emission spectrum , membrane order can be quantitatively assessed and reported using a ratiometric formula as Generalized Polarization (GP) value. The GP function has a similar form to the fluore scence polari zation function (53) , that is: w here and are the blue shifted and red shifted emission intensity maxima for the fluorescent dye used (di 4 ANEPPDHQ), respectively. More details about GP measurements will be presented in the technical approach section. In addition to the aforementioned membrane fluidity measurements based on the spectral properties (fluorescence intensity) of di 4 ANEPPDHQ, the dye fluorescence lifetime can also be used to measure membrane order. di 4 ANEPPDHQ has a 1700 ps lifetime shift between liquid disordered and liquid orde red phases. The fluorescence lifetime histograms peak at 1850 ps and 3550 ps when residing in disordered phase and ordered phase, respectively (3) (Figure 2 .4 )

PAGE 50

16 Figure 2 .4 Fluorescence lifetime histograms for di 4 ANEPPDHQ in vesicles composed of DOPC and SM/Chol, 7:3. (Inset ) Structure of the probe. Figure was adapted from (3) . It is well known that adding cholesterol to vesicles made of synthet ic lipids decreases their membrane fluid ity (makes membranes more solid ) (54,55) Since we are working with native membra nes, which have more biological relevance, we therefore checked if our model is correct and can be modulated by adding the cholesterol to the GUVs made of native Endoplasmic Reticulum (ER) membranes isolated from mouse kidneys in addition to GUVs made of synthetic lipids ( Dipalmitoylphosphatidylcholine = DOPC) served as a control. Membrane fluidity decrease is represented by a shift of the corresponding GP histogram to the right (more solid), and a shift to a longer lifetime on the phasor plot for FLIM measurements. Cholesterol effect on membrane fluidity using GP measurements and FLIM is illustrated in Figure 2 . 5 and Figure 2 . 6 , correspondingly .

PAGE 51

17 Figure 2 . 5 Cholesterol modulation effects on membrane fluidity examined via Generalized Polarization (GP) measurements on GUVs made from isolated mouse kidney Endoplasmic Reticulum (ER) membranes. ER membranes have 0.413 (more fluid) and 0.00329 (more solid) mean GP values before and after cholesterol administration, respectively. The dipole moment of the dye aligns alongside membrane bilayer acyl chains. Since the excitation laser light used was linearly polarized (p polarized), the dye was preferentially excited at the equator of GUVs than their poles. Maximum excitation occurs when light polarization is parallel to the dipole moment of the dye. Bar = 20 m

PAGE 52

18 Figure 2 . 6 Cholesterol modulation effects on membrane fluidity examined via Fluorescence Lifetime Imaging Microscopy (FLIM) on GUVs made from DOPC at 37 °C degrees using di 4 ANEPPDHQ. (A) Di 4 ANEPPDHQ fluorescence lifetime shifted to a longer lifetime (more solid) (purple circle on the corresponding phasor plot) after cholesterol administration. (B) Same conditions as (A) using lifetime fractional analysis. The reason why the poles and equator of GUVs look very different was due to photo select ion phenomenon described previously in the figure legend of Figure 2.5 . Bar = 20 m Fluorescence Lifetime Imaging Microscopy (FLIM) Fluorescence Lifetime Imaging Microscopy (FLIM) techniques hinge on the statistical nature of photons emission of a fluorophore after being excited. Fluorescence lifetime is a characteristic of the transition of the excited state of a fluorophore to the ground state. This transition yields emission of photons whose temporal arrival follow exponential decay law, which co uld be either monoexponential or multiex ponential. This radiative transition is depicted on the following Jablonski diagram as ( ) (Figure 2 . 7 ).

PAGE 53

19 Figure 2 . 7 Jablonski diagram showing a molecule electronic states and transitions among them. and represent the ground state, the first excited state, the second excited state, and the first triplet state, respectively. Solid lines mark radiative transitions. Whereas, dashed and dotted lines mark non radiative transitions and intersystem crossing, respectively. Figure was adapted from (4) Despite the fact that fluorescence lifetime is an intrinsic property of a fluorophore, decay times vary due to the instability of the excited state caused by temperature variation, interaction with surrounding environment, and the possibility of non radiat ive energy transfer to an acceptor species. The good temporal resolution of FLIM measurements, typically in the order of nanoseconds, enables studying quick dynamics of biological processes. Two main approaches can be used to measure fluorescence lifetime s; time domain approach and frequency domain approach. In time domain approach, fluorescence lifetime is extracted from fitting fluorescence decay curves. These curves are generated by firs t excite the fluorophore with ultra short pulse s from a light sourc e, and then record fluorescence intensity in time gated mode; in which fluorescence intensity is detected in distinct time bins after certain delays from the excitation pulse s . This requires illuminating sections of the

PAGE 54

20 photosensitive area in case of a Cha rged Coupled Device (CCD) camera detection , or multiple photomultipliers at various delay times (7) . Thus, the former imposes limitation in the number of pixels available to register the fluorescence intensity signal, and thereby the total Field Of View (FOV) size. Furthermore, this technique demands a strong intensity signal, as only few number of photo ns are detected in the assigned time bin, i. e. low detection efficiency. Distortion of fluorescence intensity decay curves might arise from photobleaching of fluorophores hence, photostability of fluorophores and dyes is required (7) . Unlike the aforementioned time gated m ethod of time domain approach, where fluorescence intensity is integrated over the width of a certain time bin, Time Correlated Single Photon Counting (TCSPC) method does not integrate fluorescence intensity in distinct time bins after delays. Instead, it measures the differential delay times between the light source pulse s and the very arrival of fluorescent photons. Single photon arrival time is a very essential concept in TCSPC measurements. It is performed by time measuring electronics. It requires star t stop signals. Early TCSPC applications, the pulsed light source used to trigger the time measurement (start), while the signal from the detector in response to photon arrival used to end the time measurement (stop). With the recent high pulse repetitio n rates light sources, there has been a need to reverse this mode; now detector signal starts the measurement, and laser source ends it. That way electronics do not have to operate at the higher repetition rates of the pulsed source (5) . Those single photons and their arrival times are stored and binned, and decay curves can be generated from which fluorescence l ifetimes will be recovered (Figure 2 . 8 ).

PAGE 55

21 Figure 2 . 8 Principle of TCSPC. Photons distribution is built up over time after the excitation pulses. Figure was adapted from (5) . Since TCSPC is a single photon detection method, it is very sensitive offers high detection efficiency and well suited for samples with weak fluorescence intensity signal (7) . For old TCSPC systems, count rates were limited by pile up effects (distortion in the number of pho ton counts due to low pulse repetition rate for electronics used), and low photon count

PAGE 56

22 rates were achieved in one detection cycle. In modern system, however, photon count rates are rather limited by the photostability of samples (5) . Therefore, long acquisition times might be required to attain reliable fluorescence decay curves. The underlying principle of frequency domain lifetime imaging is the modulation of the excitation source. This source could be either in the form of a conti n uous sinusoidal waveform, or a stream of delta pulses (7) . Based on the fact that the delayed fluorescence signal will be demodulated and phase shifted compared to the original excitation s ource signal, fluorescence lifetimes can be calculated from the measured phase shift and signal demodulation. Figure 2 . 9 summarizes FLIM two different approaches. Figure 2 . 9 Principles of Fluorescence Lifetime Imaging Microscopy (FLIM) techniques . (a) Time domain approach (b) Frequency domain approach. Figure was adapted from (6) .

PAGE 57

23 Phasor plot approach is an alternative way of frequency domain lifetime imaging data analysis . The phasor plot transformation and analysis w ere first introduced by Digman et al (56) . T he modulation and phase differences compared to the excitation source are used to gener ate a two dimensional (2 D) polar plot (phasor plot). In the polar coordinate system, points are represented using two values; radial and angular. In a polar plot, modulation (m i,j ) determines the radial coordinate and phase ( i,j ) determines angular coordinate; which can be later used to determine G and S values according to the following equations: where i and j indexes represen t a pixel location in the image (Figure 2 .10). Figure 2 .10 FLIM p hasor plot r epresentation in the polar coordinate system. and are the demodulation and the phase shift of the fluorescence emission signal, respectively.

PAGE 58

24 A single exponential fluorescence decay corresponds to a point on the semi circle , while more complex fluorescence decays are represented by points inside the semi circle . Each pixel of the FLIM image gives rise to a single point in the phasor plot, and when used in reciprocal mode, enables each point of the phasor plot to be mapped back to the respective pixel of the FLIM image . Because phasors follow simple vector algebra, it is possible to determine the fractional contribution of two or more independent molecular species coexisting in the same pixel (56) . Fractional analysi s of the order ed versus disorder e d phase in each pixel can be determined based on fluorescence lifetime correspond ing to 4 ns and 1.8 ns, respectively (3) . As noted earlier, variations in fluorescence lifetimes could be associated with interaction with different surrounding environments. Therefore, this property can be exploite d to investigate membrane microenvironments with the aid of fluorescent dyes that change their fluorescence lifetimes accordingly. Fluorescence Lifetime Imaging Microscopy Förster Resonance Energy Transfer (FLIM FRET) Giant Unilamellar Vesicles (GUVs) are conv enient biomimetic systems of biological membrane s that have been increasingly used to quantitatively address biophysical and biochemical processes related to lipid and protein dynamics and function s (16) . More details concerning GUVs formation will be presented in the technical approach sec tion. Membrane domains ( phases ) can be either microscopic or nanoscopic. While lipid microdomains have been primarily and extensively studied by means of GUVs which have

PAGE 59

25 shown vivid evidence of the existence of lipid phases (57) , lipid nanodomains are typically characterized using Förster Resonance Energy Transfer (FRET) (1) . Due to the fact that membrane domains have unique biological properties, fluorescently labeled phospholipids analogs mostly preferentially partition in one domain over the other. These lipid domains can be resolved as separate entities as lo ng as their sizes are bigger than the optical resolution of the microscopy technique used; direct visualization. However, indirect visualization is still possible by means of FRET. FRET measurements require a FRET pair of spectrally overlapping probes a donor and an acceptor. The energy transfer happens when these two spectrally overlapping species become in ~10 nm proximity , following the rate constant for FRET, , equation given by (58) where is the intermolecular separation. is the F rster distance typically in the order of a few nanometers. lifetime. FRET efficiency is a measure of the energy transfer efficiency between the donor and the acceptor. While FRET efficiency measurements can be either intensity based or lifetime based, the latter is more advantageous since fluorescence lifetime me asurements are probe concentration independent.

PAGE 60

26 FLIM FRET efficiency (E) can be calculated from the following equation where and are the fluorescence lifetime of the donor in the presence of the acceptor, and the fluorescence lifetime of the donor in the absence of the acceptor, respectively (7) . FLIM FRET has been extensively used in membrane domain formation and phase behavior experiments, moreover, in localization of membrane probes experiments (7,59) . The concepts of these two main types of experiment are illustrated in Figure 2 . 1 1 . Figure 2 . 1 1 Illustration of FLIM FRET applications in studying membrane lipid domains. (A) The coexistence of lipid domains in domains formation and phase behavior experiments can be revealed by means of FRET of lipid probes that preferentially partition in the same domain. Similarly, colocalization of membrane interacting species (lipid lipid and lipid protein) in a specific lip id domain can be studied using FLIM FRET approach (B). Figure was adapted from (7) .

PAGE 61

27 FRET can be used to measure the partitioning of lipid dyes in membrane bilayers (60) . St eady State Probe Partitioning FRET (SP FRET) was used to show the coexistence of membrane phases in cuvette based binary mixture vesicles in suspension experiments. A FRET pair was used in a pure single phase and coexisting phases measurements. When the pr obes were added to the single phase suspension, an intermediate FRET signal intensity was observed; compared to a higher FRET signal when the probes prefer the same phase (one out of the binary mixture), and a lower FRET signal when the probes preferentially partitioned in the two different phases of the binary mixture (60) . It is noteworthy that FRET efficien cy depends on domain size, and the self quenching properties of the FRET pair (7) . Membrane lipid lipid and lipid protein interactions in certain lipid environments (domains) have been verified via FLIM FRET measurements (7,59,60) . These measurements were done in biological samples such as cells plasma membrane where identifying individual lipid domains is not possible due to limitation in fluorescence microscopy resolution. For these experiments, two membrane probes (FRET pair) are used to fluorescently label the species under investigation, one of which is well known to preferentially partition in the local lipid domain of interest. Similarly, proteins and their membrane receptors can be identified in their specific (local) lipi d domains using this valu a ble approach. In order to investigate membrane domains in live cells below the resolution of diffraction limited optical microscopy, a viable superresolution microscopy alternative needs to be considered. In a recent article by Se zgin et al, polarity sensitive probes for superresolution Stimulated Emission Depletion (STED) microscopy have been tested (61) . Among the polarity sensitive dyes (di 4 ANEPPDHQ, di 4 AN(F)EPPTEA, and NR12S)

PAGE 62

28 whose performance w as tested . di 4 ANEPPDHQ had the lowest photostability (61) , so it was not feasi ble to use it to get reliable STED images. Fluorescence Correlation Spectroscopy (FCS) The concept of Fluorescence Correlation Spectroscopy (FCS) was first introduced in 1972 by Magde et al (62) . This technique is based on the statistical analysis of the temporal fluctuations of fluorescence intensity signal of fluorophores in a very small volume (~fL) (8) . FCS is a single molecule detection method. Compared to Fluorescence Recovery After Photobleaching (FRAP), a common technique that has been widely used to study membrane diffusion (8,63,64) , FCS measurements require fluorophore concentration and laser powers orders of magnitude lower (8,65) . Compared to other Single Molecule Tracking ( SMT) and Single Particle Tracking (SPT) techniques, FCS offers a relatively immediate e xperimental readout, without the need for more time consuming off line data analysis. One drawback, though, is the loss of single molecule behavioral information such as specific trajectories and temporary confinement (8) . Figure 2 . 1 2 shows FCS setup of a confocal microscope. Figure 2. 1 2 Principles of Fluorescence Correlation Spectroscopy (FCS). (A) FCS confocal microscopy configuration. (B) Temporal fluctuations of the recorded fluorescence intensity signal . (C) Calculation of the autocorrelation function of the detected fluorescence signal. Figure was adapted from (8) .

PAGE 63

29 The microscope objective focuses the excitation light to a diffraction limited spot. The pinhole inserted in emission optical path provides tight axial confinement resulting in a detection volume in the order of femtoliters (8) . F luorophores coming in and out of the detection volume results in fluctuation in the recorded emission fluorescence intensity signal. The autocorrelation function which describes the similari t y of the inten sity signal over time is given by the following equation; Here denotes the time average, and where is time average of the recorded intensity signal, and is the lag time of the correlation . Autocorrelation curves can be calculated either on line or off line, by hardware correlator cards or by fast software computation (66) . The diffusion coefficient D and the relative concentration C of fluorophores can be extracted from fitting the respective experimental autocorrelation curves with a mathematical model function. Some of these fitting functions are given in Table 2 .1 Table 2 .1 Some of the fitting functions for FCS auto correlation curves. Type of diffusion Model function Three dimensional (3D) Brownian diffusion Two dimensional (2D) membrane diffusion Two dimensional (2D) membrane diffusion with triplet

PAGE 64

30 Here the number of particles in the detection volume is given by ( ; in case of two dimensional (2D) Gaussian detection area) , under nanomolar concentrations, where and the form factor S . S defines the aspect ratio of the elliptical detection volume. T is the fraction of fluorophores in the triplet state within the observation volume. The equation that relates the diffusion time wit h the diffusion coefficient D derived from Sto kes Einstein diffusion equation is as follows: The laser beamwaist also denoted as can be empirically determined by a calibration measurement of a dye with a known diffusion coefficient (8) . Not only do fluctuations in the detected fluorescence intensity signal arise from fluorophores traversing the detection volume, but also from some photophysical and photochemical p henomena, such as antibunching, rotational fluctuations, triplet and blinking. These phenomena span various timescales as shown in Figure 2 .1 3 . Figure 2 .1 3 Timescales of different phenomena that can be monitored using fluorescence intensity autocorrelation analysis. Figure was adapted from (9) .

PAGE 65

31 Fluorescence Correlation Spectroscopy (FCS), in conjunction with laser scanning microscopy, has been successfully used to investigate lipid and protein dynamics (67,68) . GUVs made of either pure, or a mixture of binary or ternary lipid species have been employed to examine me mbrane phase behaviors. On these GUVs, lipid translational diffusion has been studied using FCS. Diffusion coefficients can be calculated from the autocorrelation curves of the fluorescence temporal fluctuations of lipophilic probes incorporated in membran e bilayers. This translational diffusion can be modelled as two dimensional Brownian motion (67) . Membrane lipids have dissimilar diffusion rates when they reside in different membrane phases. The translational diffusion coefficient of solid gel phase (~ is about a 1000 times slower than liquid disordered and liquid ordered phases (25) . Despite the fact that liquid ordered phases have higher order compared to liquid disordered phases, their translational diffusion coefficients are similar (~1 (25) . It has been shown that membrane domains sizes are correlated to and can be estimated from their lipid di ffusion coefficients; the bigger the domain the slower its lipids diffuse (69) . Several studies have shown chol esterol modulation effects on membrane phase behaviors (54,67,70) . Cholesterol was found to have effects (54,55,70) . Increasing cholesterol concentration in binary mixtures of low phase transition temperature lipids ( (such as DOPC and DLPC)/sterols ) decreased the lateral diffusion coefficients of these lipids as measured by FCS on GUVs made from those mixtures, indicating a (54,55) . Whereas, when cholesterol concentration was increased in binary mixtures of high phase transition temperature lipids (such as SM,

PAGE 66

32 DPPC and DSPC)/sterols, the diffusion coefficients increased, indicating cholesterol (54,70) . Technical difficulties accompanying the reconstitution of membrane proteins in their functional forms into control systems have been the main reason for our limited understanding of membrane proteins compared to their soluble counterparts (68) . FCS has been used widely to study protein dynamics and protein/lipid interactions in model membranes and in living cells. Once membrane proteins are suc cessfully fluorescently labelled without affecting their functions, FCS can be applied as a powerful tool to estimate protein diffusion coefficients, concentrations and aggregations, and their molecular interactions with other species (68) . Photon Counting Histogram (PCH) From the same FCS datasets, Photon Counting Histogram (PCH) analysis can be performed. PCH is a complementary tool to FCS. While the relative concentration of fluorophores in the observation volume can be estimated in FCS measurements, PCH was developed to determine the average number of fluorophores in the diffusing species (71) . FCS tec hnique is not sensitive enough to detect relatively subtle variations in the size of the diffusing molecules. Since the diffusion rate of a molecule is inversely proportional to the cube root of its volume, the volume must increase about eightfold before t wofold reduction in the diffusion rate can be detected (72) . In PCH approach, the fluorescence signal is utilized to generate the so called PCH histograms. Th ese histograms are plots of the frequency of photon counts per bin time. For a single molecular species, the corresponding PCH histogram can be fully characterized by two parameters, namely, the average number of molecules in the detection volume and the

PAGE 67

33 m olecular brightness (i. e., photon counts per molecule per sampling time.) On the other hand, for multiple molecular species, the resultant PCH histogram is the result of convolving the PCH histogram for each individual species with the others. The shape of those PCH histograms exhibit a super Poissonian behavior. This extra broadening to a Poisson distribution is due to fluctuations in the fluorescence intensity signal (71) . These fluctuations are noise, and fluctuation s in the number of molecules in the observation volume (71,73) . Using a nonlinear least square fitting model, such as Levenberg Marquardt or Gauss Newton, the recorded PCH histogram is fitted to the following variance ( ) equation: where p(k) is the probability of the detected photon counts in a bin, and M is the total number of measur e ments (73) . This variance, in conjunction with the average intensity of the recorded fluorescence signal , can then be used to calculate the molecular brightness of the species from the following equation: where is the geometric shape of the PSF (73,74) . Hence, in a solution, while a [2x] of a monomer (one fluorescent dye) and an [x] of a dimer (two fluorescent dyes) would have the same average fluorescent intensity, the molecular brightness of the dimer would be twice that of the monomer because of the larger variance (73) . This PCH underlying principle is depicted in Figure 2 .1 4 .

PAGE 68

34 Figure 2 .1 4 The concept of molecular brightness in PCH method. The larger the variance the bigger the molecules size for species with equal average intensity. Molecular brightness is the ratio of the fluorescence signal fluctuations to the average intensity of the signal. Figure was adapted from (10) . In PCH method, oligomerization state of fluorescently labeled species can be examined. Providing that these diffusing species are homogenous and do not contain unlabeled molecules, their clustering state is deduced by comparing the measured molecular brigh tness to a control (monomer or dimer) (73) . In addition to its application to in vitro studies (75,76) , PCH technique has been used to characterize the molecular brightness of autofluoresc ence molecules and the Enhanced Green Fluorescence Protein (EGFP), in the cytoplasm and the nucleus of HeLa cells under in vivo conditions (77) . Furthermore, sodium dependent phosphate cotransporter type 2a (NaPi2a) protein aggregation/clustering was studied in control and potassium deficient apical membranes reconstituted into GUVs, by combining FCS and PCH approaches (50) .

PAGE 69

35 CHAPTER III TECHNICAL APPROACHES In t his chapter, a detailed description of the technical approaches used to study isolated and intact membranes is presented. The GP Lifetime ( ) method which intends to enhance contrast compared to intensity Generalized Polarization (GP) measurements is introduced at the end of this chapter. Brush Border Membrane Isolation Rats were anesthetized via an intraperitoneal injection of 100 mg/kg pentobarbital sodium (Pentothal, Abbott). Duodenal and Jejunal BBMs were isolated by double Mg 2+ precipitation technique as described before (50,78) with slight variations. Intestinal mucosal scrapes from each rat were in 15 ml isolation buffer consisting of 50 mM Mannitol, 2 mM Hepes/NaOH (pH 7.1), and Complete protease inhibitor (Roche Diagnostics, Germany). Intestinal mucosa samples were homogenized with a Potter Elvehjem homogenizer. BBMV s were prepared by a double ser ial Mg 2+ precipitation procedure. First, MgCl 2 was added to the homogenates (13 mM final concentration), incubated on ice for 20 min and centrifuged at 3,000 g for 15 min . The supernatant was centrifuged at 38,000 g 4 o C for 40 min. A s econd Mg 2+ precipitat ion step was performed by resuspending the membrane pellet in 7.5 ml of solution B (300 mM Mannitol, 0.1 mM MgSO4, 20 mM Hepes/NaOH pH=7.1 and Complete protease inhibitor) with a 20Ga needle. The samples were centrifuged at 6,000 g for 30 min and the super natant was transferred to a clean tube. Final centrifugation at 38,000 g, 4 o C for 40 min resulted in a pellet containing duodenal or jejunal BBMs that was resuspended in a final buffer (300 mM Mannitol, 16 mM Hepes/Tris pH=7.5 and Complete protease inhibit or)

PAGE 70

36 by passing through a 25Ga needle and aliquoted for (a) measurement of total protein concentration, (b) Na/Pi cotransport activity measurement, (c) protein electrophoresis and Western blotting, (d) DR and DS fractions separation, and (e) preparation of giant unilamellar vesicles (GUVs) for FCS measurements. BBM s protein concentration was determined by BCA protein assay (Pierce). Transport Activity Measurements Transport activity measurements were performed in fresh isolated BBM vesicles by radiotracer uptake followed by rapid Millipore filtration ( 79,80) . To measure sodium gradient dependent 32 Pi uptake (Na/Pi cotransport), 10 u1 of BBM or membrane fragments preloaded in an intravesicular buffer (in mM) of 300 mannitol, 16 HEPES, and 10 Tris, pH 7.5, was vortex mixed at 25°C with 40 ul of an extr a vesicular uptake buffer of 150 mM NaCl, 100 uM K 2 H 32 PO 4 (PerkinElmer Life Sciences), 16 mM HEPES, and 10 mM Tris, pH 7.5. Uptake was terminated after 30 s (representing the initial linear rate) by an ice cold stop solution. All uptake measurements were p erformed in triplicate, and uptake was calculated on the basis of specific activity determined in each experiment and expressed as pmol Pi /(1min)/(mg of BBM protein). Isolation of BBM Detergent resistant and Detergent sensitive Fractions For isolation of detergent resistant (DR) and detergent sensitive (DS) BBM fractions, the BBM sample was first incubated for 30 min on ice in TNET buffer (50 mM Tris HCl, pH 7.4, 150 mM NaCl, 5 mM EDTA, and a complete protease inhibitor mixtu re) containing 1% Triton X 100 (50) . Following centrifugation at 100,000 g at 4°C for 1 h, the DR (pellet) and

PAGE 71

37 DS (supernatant) fractions were collected and analyzed for (a) total protein by BCA protein assay (Pierce) and (b) Western blots f or NaPi2b and Flotilin 1. Measurements of C holesterol and S phingomyelin in BBMs Total cholesterol was determined enzymatically using Amplex Red (ThermoFisher, Waltham, MA ) and Sphingomyelin was determined with a fluorescence assay, using reagents supplied in the Amplex Red Sphingomyelinase kit (ThermoFisher, Waltham, MA) and following the standard protocol, provided by the supplier . Western B lotting Rat intestine duodenal and jejunal BBMs protein lysates (n=3 for each region) were separated on a 12% mini ge l, transferred onto nitrocellulose membranes (Bio Rad, Hercules, CA) and subjected to Western blot analysis. After blocking in 5% non fat dry milk, membranes were incubated in primary antibody (NaPi2b (1:1000; Genzyme) ; Flotillin 1 (1:1000; Becton Dickinso n; Franklin Lakes, NJ)), washed in blot buffer (150 mM NaCl, 10 mM Na 2 HPO4, 5 mM EDTA, 1% Triton X 100; pH 7.4), incubated in horseradish peroxidase conjugated secondary antibodies (1:10,000 dilution; Jackson ImmunoResearch, West Grove, PA) and washed in b lot buffer. The antibody complexes were detected using enhanced chemiluminescence (Pierce, Rockford, IL) and captured using a photodocumentation system (UVP, Upland, CA). Immunohistochemistry For immunohistochemistry, multiple blocks from fixed intestinal regions (~5 mm pieces) were infused with 5% (2 hr), 10% (2 hr) and 25% (overnight) sucrose, frozen in

PAGE 72

38 liquid nitrogen in OCT compound (Fisher, Houston, TX) and cryosectioned (5 m). Intestinal sections were blocked (10% normal goat serum and 1% BSA in PBS) and incubated overnight at 4 °C with primary antibody (NaPi2b 1:100; Genzyme). After being washed, the sections were incubated (30 min, room temperature) with an appropriate mix of Atto 647N conjugated goat anti rabbit IgG (1:800; Thermo Fisher; Waltham, MA), and Phalloidin Alexa 594 conjugated (1:1000; Thermo Fisher; Waltham, MA), washed with PBS and mounted in Fluromount G (Fisher; Hampton, NH). Giant Unilamellar Vesicles (GUVs) Electroform ation GUVs were prepared by means of an electroformation method developed by Angelova and Dimitrov (81) , in a temperature controlled chamber as previously described in (82 84) . Briefly, a custom made Teflon chamber was cleaned by sonicating in Tween 20 detergent, soaking in 80% ethanol 20% water, and sonicating in NanoPure water for one hour each, respectively. The chamber was then dried out und er nitrogen flow. BBMs from the duodenal and jejunal enterocytes were isolated using Mg 2+ precipitation method. BBM samples were diluted to 1 µg/µL in 10 µL which were then spread over two Platinum (Pt) wires (5 µL on each wire) and le ft to dry under the nitrogen flow to form thin lipid films. Next the two P t wires were glued (using J B Kwik epoxy) together with a coverslip to the bottom of the chamber . After th e adhesive had hardened , the chamber was mounted onto the microscope stage and 50 0 µL of Tris buffer (pH 7.4) w as added to it for the lipid hydration. An AC sinusoidal wave (3 , 10 Hz) was applied to both Pt wires. Electroformation was performed over 30 minutes at (above the liquid ordered to liquid disordered phase transition temperature) by connecting the chamber to a circulating water bath. When the electroformation was done, the function generator power was shut off, and the temperature

PAGE 73

39 was decreased to . Upon reaching that temperature, 10 µL of the Alexa Fluor 647 (Lif e Technologies) pre labeled NaPi2b Ab (Genzyme) were immediately added . Next 1 µL of di 4 ANEPPDHQ (2mM) was added and incubated for 30 minutes to give 4 µM final concentration. At this point GUVs were ready to be imaged for both fluidity (GP) and NaPi2b d iffusion (FCS) measurements (Figure 3 .1 ). Generalized Polarization (GP) Measurements The generalized polarization (GP) function has a similar form to the fluorescence polarization function (53) , that is: Where and are the blue shifted and red shifted emission intensity maxima for the fluorescent dye used (which i s di 4 ANEPPDHQ), respectively. The dye fluoresces with a peak emission wavelength of ~560 nm (green) when residing in the ordered phase, and ~620 nm (red) when residing in the disordered phase (2) . Generalized Polarization (GP) measurements were obtained through spectroscopy and microscopy techniques (Figure 3 .1 B).

PAGE 74

40 Figure 3 . 1 Microscopy m ethods illustration. (A) BBMs were isolated from the intestinal enterocytes were. For NaPi2b diffusion and membrane fluidity analysis GUVs were made from BBMs by electroformation. ( B ) GUVs were stained with di 4 ANEPPDHQ to analyze membrane mi crodomains . ( C ) GUVs were co stained with NaPi2b antibodies pre labeled with Alexa 647, and single point FCS measurements were taken to determine NaPi2b diffusion within the membranes. Spectroscopy An ISS K2 spectrofluorometer (Champaign, Illinois, USA) was used to obtain GP measurements. The di 4 ANEPPDHQ emission spectrum was first acquired from the diluted Brush Border Membranes (BBMs) Vesicles (BBMVs) (1 µg/µL) in 200 µL Tris buffer (pH 7.4) insi de a glass microcuvette placed in the spectrofluorometer sample holder (which was

PAGE 75

41 connected to a temperature controlled water bath), over an emission 500 800 nm range at 37 . The spectrofluorometer light source was Xenon Arc Lamp. The sample was excited with 488 nm wavelength by the spectrofluorometer excitation monochromator, and the emitted light was detected by Photo Multiplier Tubes (PMTs) in a photon counting mode. The ordered and disordered channels, 560 nm and 620 nm respectively, were detected se parately via the spectrofluorometer emission monochromator having 0.2 mm width . And their ratio and GP values (applying ) were obtained and averaged out over ten different iterations at 37 . Microscopy for Isolated ( Apical ) Membrane Study (N aPi2b Study) For GP measurements, di 4 ANEPPDHQ was excited using a s ingle photon 488 nm Argon laser , and the emission that correspond ed to each lipid order phase was detected through two channels; Ch1: 505 605 nm and Ch2: 655 755 nm, corresponding to lipid ordered and lipid disordered phases, respectively, using a pseudo photon counting detection mode on an Olympus FV1000 laser scanning confocal microscope . UPLSAPO 60X (NA: 1.2) water immersion object ive was used, with a 280 µm working distance. 1024 [ pixel] x 1024 [pixel] (211.761 [µm] x 211.761 [µm]) images were taken with a sampling speed of 10 µs/pixel. Due to the differences in collection efficiency in these two channels caused by, among other factors, the use of different Photo Multiplier Tube (PM T) gains between experiments, the need for a correction or calibration factor (G Factor) to compensate for those differences arises (2) . This also helps in maintaining the fluorescence ratio in the two channels to be the same across different sets of experiments. The GP function now becomes:

PAGE 76

42 The G Factor was calculated from the following formula: where GP ref is the GP value of di 4 ANEPPDHQ in a solution measured by a spectrofluorometer, and I 505 605 and I 655 755 are the mean intensity values of Ch1 and Ch2 images (taken from their corresponding histograms) of the dye in a solution scanned by the microscope using the same microscope settings used while taking GUVs images. Two measurements were taken, and their corresponding G Factors were calcu lated, using , and averaged out. The resultant G Factor was calculated to be 1.150. This value was used in calculating the Mean GP from the original raw data using the following procedure: 1) The raw data images were im ported through LOCI, a Bio Formats Importer plug in, using ImageJ software (NIH; Bethesda, Maryland). 2) Ch1 and Ch2 images were split. 3) Each pixel intensity value in Ch2 images, , was multiplied by the G Factor. 4) Then was carried out to form the Mean GP images. 5) Before saving and exporting the Mean GP images in their final format, a thresholding was performed to limit the GP values of the images to from ( 0.9) to (0.9); given that the domain of is from ( 1) to (1). The reason for excluding

PAGE 77

43 these two intensity values from the images, and thereby from the histograms, was the prior knowledge that they correspond to pixels with low intensity values, i.e., noi se, compared to the actual high intensity value s fluorescence signal. Ex Vivo rat intestinal tissues GP images were obtained using a Zeiss 780 laser scanning confocal/multiphoton excitation fluorescence microscope with a 34 channel GaAsP QUASAR Detection Unit and nondescanned detectors for two photon fluorescence (Zeiss, Tho rnwood, NY). di 4 ANEPPDHQ was excited using a 488 nm Argon laser, and the dye emission was detected through two spectral channels; ChS1: 508 588 nm and ChS2: 615 695 nm, corresponding to lipid ordered and lipid disordered phases, respectively. Fluores cence images were captured with a Zeiss C Apochromat 40X (NA: 1.2) W Korr FCS M27 water immersion objective. 512 [pixel] x 512 [pixel] (58.40 [µm] x 58.40 [µm]) images were taken with a 1.58 µs pixel dwell time. Ex Vivo GP images were analyzed in ImageJ software (NIH; Bethesda, Maryland). The two channels were thresholded and made into binary masks, and then multiplied by each corresponding channel to account for membranes only. The thresholded Ch2 was then multip lied by a calibration factor, G Factor. At each pixel, GP values were calculated using the equation (2) : , where I 508 588 and I 615 695 are intensity values from the ordered (Ch1) and disordered (Ch2) channels, respectively. The G Factor was calculated from the following formula: where is the GP value of the di 4 ANEPPDHQ in a solution measured by the spectrofluorometer, and I 508 588 and I 615 695 are the mean intensity values of Ch1 and Ch2 images (taken from their correspon ding histograms) of the dye in a solution taken by the microscope using the same microscope settings used while

PAGE 78

44 taking the time series images. The average G factor value (= 1.106; determined from 3 separate measurements) was used to calculate the mean GP f rom the original raw data in this study. Mean GP values were further limited to 0.9 to 0.9 range. Intensity Quantification Images were imported using ImageJ software (NIH; Bethesda, Maryland) via LOCI plugin. A region from the actin channel was selected. NaPi2b channel was thresholded for background removal. And the selection was transferred to NaPi2b channel. The total pixel number in the selected area was obtained from the selected area histogram. The selected area from actin channel was transferred to N aPi2b channel. Intensity profile was obtained from "plot profile" command, and values were saved in individual .txt file. Microscopy for Intact ( Basolateral ) Membrane Study (sKL Study) HEK293 cells were grown onto 35 mm glass bottom dishes (MatTek Corpora tion, Ashland, MA) to 70% confluency. Experimental cells were placed to the phenol free media (DMEM/F 12 media with 10% fetal bovine serum, 15 mM HEPES, 2.5 mM L glutamine, 100 U/ml penicillin, 100 U/ml streptomycin) 1hr before imaging. d i 4 ANEPPDHQ (2 µM final concentration) was added to the media and incubated for 5 minutes at 37 o C. Glucose oxidase ( 200 U/ml ) from Aspergillus niger (MP Biomedical, Santa Ana, CA), catalase from bovine liver ( 0.04 mg/ml ) (Fisher, Pittsburgh, PA), were added to the media r ight before imaging to reduce the phototoxic effect of the dye. Live cell images were obtained using Zeiss 780 laser scanning confocal/multiphoton excitation fluorescence microscope with a 34 channel GaAsP QUASAR Detection Unit and nondescanned detectors f or two photon fluorescence (Zeiss, Thornwood, NY). The imaging settings were kept constant for all

PAGE 79

45 measurements for comparative imaging and results. Time series of 30 mins with no time interval between consecutive frames (focusing at the basal part of the cells) were captured with a Zeiss C Apochromat 40x/1.2NA Korr FCS M27 water immersion lens objective. The definite focus mechanism was implemented to maintain the same focal plane over the entire image acquisition period. A 30 mW Argon laser set at 1% pow er was used for excitation at 488 nm, and emission signals corresponding to the ordered and disordered phases were detected through two spectral channels; Ch1: 508 588 nm (ordered) and Ch2: 615 695 nm (disordered), simultaneously. Image acquisition was p erformed using Zeiss ZEN 2012 software. The series of images were analyzed in ImageJ software (NIH; Bethesda, Maryland). The two channels were thresholded and made into binary masks, and then multiplied by each corresponding channel to account for membrane s only. The thresholded Ch2 was then multiplied by a calibration factor, G Factor. At each pixel, GP values were cal culated using the equation (2) : , where I 508 588 and I 615 695 are intensity values from the ordered (Ch1) and disordered (Ch2) channels, respectively. The G Factor was calculated from the following formula: where is the GP value of the di 4 ANEPPDHQ in a solution measured by the spectrofluorometer, and I 508 588 and I 615 695 are the mean intensity values of Ch1 and Ch2 images (taken from their corresponding histograms) of the dye in a solution taken by the microscope using the same microscope settings used while taking the time series images. The average G factor value (1.106 ) determined from 3 separate measurements) was used to calculate the mean GP from the original raw data in this study. Mean GP valu es were further limited to 0.9 to 0.9 range.

PAGE 80

46 Fluoresc ence Correlation Spectroscopy (FCS) and Photon Counting Histograms (PCH) Measurements and Analyses NaPi2b diffusion and di 4 ANEPPDHQ dye diffusion were measured using single point FCS measurements. 635 nm and 488 nm single photon excitation was used to ex cite Alexa 647 NaPi2b and di 4 ANEPPDHQ, respectively. (BA655 755) nm and (BA505 605) nm emission filters were used to detect the fluorescence intensity. A total of 32766 points (with a pixel residence time of 10 µs) were taken for each single point FCS measurement. For diffusion measurements, the excitation beam waist was calibrated using 100 nM Rodamine 110 dye, which has a 430 diffusion coefficient, before every single experiment ( =0.25±0. 0 1 µm). Single point FCS measurements were taken on the equator of GUVs made of isolated BBMs from duodenum and jejunum, at , and the emitted signal was detected with the pseudo photon counting mode of the confocal microscope (Olympus FV 1000) ( Figure 3 .1 ). Al exa 647 NaPi2b was excited using a single photon 635nm wavelength. Globals SimFCS Software (Laboratory for Fluorescence Dynamics, Irvine) was used to export the raw data, which were then imported to Igor 6 (WaveMetrics, Inc., OR) for fitting. The calibrati on FCS raw data were fitted using a three dimensional one component fitting model (since the dye is freely diffusing in a solution in a Brownian motion fashion) , while two dimensional one component or two component fitting models were used for the NaPi2b p rotein FCS measurements in GUVs . The average number of particles, diffusion times and diffusion coefficients for the labeled NaPi2b protein were obtained from those fitting models using the calibrated beam waist .

PAGE 81

47 Duodenum and jejunum western blot resul ts showed three major bands suggesting three different NaPi2b subsets in the two intestinal regions. These subsets may represent monomer, glycosylated, and dimmer forms of the protein corresponding to 75, 110, and 150 kDa, respectively. There is a rational e for having five variant diffusing species for single point FCS measurements. Those species are Alexa 647 freely diffusing dye, Alexa 647 conjugated NaPi2b protein antibody (unbound to the membrane), Alexa 647 conjugated NaPi2b protein antibody bound to N aPi2b protein subset 1, Alexa 647 conjugated NaPi2b protein antibody bound to NaPi2b prtein subset 2, and Alexa 647 conjugated NaPi2b protein antibody bound to NaPi2b protein subset 3. Reduced chi square criterion is essential for evaluating the goodness o f fit of the autocorrelation function. Although it is anticipated that a five component fitting model will result in a reduced chi square value, the retrieved diffusion coefficients might not be of biological relevance. For instance, we fitted some duodenal and jejunal autocorrelation curves to a five component fitting model complemented with the exponential term and baseline (B), and extracted corresponding diffusion coefficients . A representative example is given in Table 3.1 . Table 3.1 Duodenal a nd jejunal diffusion coefficients recovered from five component fitting model for their corresponding autocorrelation curves. Diffusion D 1 D 2 D 3 D 4 D 5 Chi square Duodenum 7.66 7.66 7.59 7.74 3993.17 2.47 10 5 Jejunum 0.17 0.17 0.17 0.17 0.17 7.48 10 6 In addition to the replicated diffusion components recovered from both regions, it is noted that duodenal D 5 is of no biological relevance (i.e., it cannot represent NaPi2b protein diffusion knowing that the fastest diffusing species is Alexa 647 free dye whose diffusion coefficient was measured to be ~2 6 0 µm2/s.)

PAGE 82

48 The probability of having Alexa 647 freely diffu sing dye in the FCS observation volume is very low given the high binding efficiency of the dye to NaPi2b protein antibodies (IgGs) and multiple washing steps during labeling procedure. We, therefore, decided to exclude this component. NaPi2b subsets 1 3 do not have a huge molecular weight difference. It is known that a separation of two dissimilar diffusing species requires a minimum difference of their molecular weight ratio of 5 8 (85) . Hence, those three subsets can be represented by one component in the fitting model. The other component to account for in the measurement volume is Alexa 647 conjugated NaPi2b antibodies, the probability of which relies on NaPi2b protein labeling effici ency and NaPi2b protein expression level in the two intestinal regions. Taken all together, we proposed a two component fitting model. Duodenal Alexa 647 NaPi2 b autocorrelation data were fitted using two dimensional two components fitting model giving b y the following equation: The two translational components were complemented by an exponential term to account for the incomplete relaxations of slowly diffusing species (50) , in addition to the baseline term (B). Jejunal autocorrelation curves looked markedly different. Thus, we decided to examine a one component fitting model in addition to the two component fitting model previously proposed. The reason behind this notion is the higher expression level of NaPi2b protein in the jejunum, and hence the probability of having unbound Alexa 647 conjugated NaPi2b protein antibodies is less under efficient conditions . Nineteen jejunal FCS

PAGE 83

49 measurement dat asets were, consequently, examined to show the validity of using either fitting model. The test criteria consisted of calculating and comparing the chi square of each one and two component fitting model, as well as evaluating the biological relevance of t he extracted diffusion coefficients. For example, a reduced chi square values for a two component fitting do not necessarily favor the model, especially when the recovered diffusion coefficients are of no biological relevance. Following the above judging c riteria, thoroughly, 11 out of 19 (~58%) datasets fell under the one component fitting model category. Therefore, we decided to favor the one component fitting model for jejunal FCS measurements. J ejunal Alexa 647 NaPi2 b autocorrelation data were fitted u sing two dimensional one component fitting model giving by the following equation: The diffusion coefficients (in /s) were calculated from according to the following equation: We have experimentally determined the freely diffusing dye as well as NaPi2b antibodies unbound to membrane diffusion coefficients in a solution , and those were (263.85±23.70) (mean±SD) µm 2 /s and (46.21±27.78) (mean±SD) µm 2 /s, respectively. R etrieved diffu sion coefficients , whose means fell within two standard deviations from the mean of the freely diffusing dye diffusion coefficient, were assigned D 1 . R etrieved diffusion

PAGE 84

50 coefficients , whose means fell within one standard deviation from the mean of NaPi2b antibodies unbound to membrane diffusion coefficient, were assigned D 2 . Recovered diffusion coefficients outside these ranges were sorted out, by observation, to slow (~5 µ m 2 /s), fast (~10 µm 2 /s ), and very slow (~0.3 µm 2 /s ) diffusion coefficients, and thes e were assigned D 3 , D 4 , and D 5 , respectively. For di 4 ANEPPDHQ single point FCS measurements, the dye was excited with a single photon 488nm wavelength. The fluorescence intensity fluctuations of the dye were recorded and detected in Ch1: 505 605nm, using Olympus FV1000 laser scanning confocal microscope. Duodenal and jejunal di 4 ANEPPDHQ autocorrelation data were best fitted using the two dimensional one component fitting model similar to that of jejunum above. The PCH analysis depends on the probability distribution of finding molecules in the observation volume as measured by FCS measurements (50,86) . The concentration and the molecular brightness of free molecules, or molecules that are bound together and travel as one unit, can be calculated using PCH analysis (50,77,86,87) . Oligomerization (dimers, trimer interest to the intrinsic molecular brightness of their monomeric forms; such that 2 fold increment in the molecular brightness indicates dimerization, and so forth. Alexa 647 NaPi2b Abs in the Tris buffer was used to determine the monomeric molecular brightness of NaPi2b protein. It is know that IgGs are normally in a monomeric state (88) . Duodenal and jejunal PCH data were binned by 8 with a sampling frequency of 100 kHz, and fitted with a single component fitting model by using Globals SimFCS Software (Laboratory fo r Fluorescence Dynamics, Irvine).

PAGE 85

51 Measurement of M embrane O rder U sing F luorescence L ifetime I maging M icroscopy (FLIM) FLIM was performed using a Zeiss 780 laser scanning confocal/multiphoton excitation fluorescence microscope with a 34 channel GaAsP QUASAR detection unit and nondescanned detectors for two photon fluorescence (Zeiss, Thornwood, NY) equipped with a n ISS A320 FastFLIM box (ISS, Champaign, IL) and a titanium:sapphire Chameleon Ultra II (Coherent, Santa Clara, CA). A dichroic filter (488 nm, Di02 R488 25 D, Semrock Inc, Rochester, NY) was used to separate the fluorescence signal . For the acquisition of FLIM images ( for both FD and TCSPC) , fluorescence signal was detected by a photon counting PMT detector (H7422p 40; Hamamatsu), after passing through bandpass Ch1: 506 594nm and Ch2: 604 679nm (Semrock) emission filter s, for ordered phase and disordered phase, respectively . The 2 photon excitation was blocked by a 2 photon emission f ilter. Images of the baso lateral membrane were obtained with VistaVision software by ISS in the 256x256 format with a pixel dwell time of 6.3 µs/pixel (FD) or 12.61 µs/pixel ( TCSPC) and averaging over 30 frames. Cells, stained with d i 4 ANEPPDHQ (0.25 µM final concentration) for 5 minutes were exited at 900nm by a tunable infrared Coherent Chameleon Ultra II laser (680 1080 nm). An average power of about 10 mW was used to excite the live cells. For frequency domain FLIM imaging calibr ation of the system was performed by measuring the known lifetime of the Alexa 488 with a single exponentia l decay of 4.1 ns (89) . Calibration and d i 4 ANEPPDHQ imaging of the live cells were done using the same acquisition parameters. The p hasor transformation and data analysis for FD were performed using the Global s SimFCS software (Laboratory for Fluorescence Dynamics, University of California, Irvine) as described previously (56,90) . Briefly, the fluorescence decay in each pixel of the FLIM

PAGE 86

52 line scan was transformed into the sine and cosine components, which were then represented in a 2D polar plot (ph asor plot) (56) . Lifetime fitting for TCSPC FLIM data in each channel were fitted to single exponential decays pixel by pixel using ISS VistaVision software. Fluorescence L ifetime I maging M icroscopy (FLIM) M easurement of Förster R esonance Energy Transfer (FRET) Studies Cells were stained (at room temperature for 5 min) with BODIPY FL 505/510 C5 Ganglioside GM1 (donor; 100nM) and with or without CholEsteryl BODIPY® 542/563 C11 (acceptor; 400nM). Basic FLIM microscopy was done as described above. Fo r FLIM FRET studies, a dichroic filter (488 nm, Di02 R488 25 D, Semrock Inc, Rochester, NY) was used to separate the fluorescence signal, which was then detected by a 506 594 nm (Semrock Inc) emission filter that was placed in front of the detector (GM1 C holesterol FLIM FRET). A dichroic filter (594 nm, Di02 R594 25 D, Semrock Inc, Rochester, NY) was used to reflect the fluorescence signal into a 506 594 nm (Semrock Inc) filter that was placed in front of the detector. (GM1 Klotho FLIM FRET). The 2 photon excitation was blocked by a 2 photon emission filter. Images of the basolateral membrane were obtained in the 256x256 format with a pixel dwell time of 12.61 µs/pixel and averaging over 30 frames. VistaVision software by ISS was used for the acquisition an d analysis of FLIM images. A digital frequency domain setup measured the modulation and phase at each pixel within an image. For analysis using the phasor diagram, the modulation and phase determined the radial and angular coordinate of the phasor in a pol ar plot, respectively. The phasor of the unquenched donor was determined as the phasor of cells stained only with BODIPY GM1. The phasor of the

PAGE 87

53 background was determined as the phasor of the autofluorescence signal from unstained cells. To examine FRET occ urrences, shifts of the donor acceptor versus donor only phasors were determined. The trajectory of variable FRET efficiencies was drawn in the plot. The position that donor was quenched in the presence of acceptor yields the value of efficiency associated with the FRET interaction. Probabilistic GP Lifetime ( ) Owen et al, have shown that Fluorescence Lifetime Imaging (FLIM) of Di 4 ANEPPDHQ enhanced contrast compared to spectral techniques, when imaging membrane orders in model membranes and live cel ls (3) . The dye had a 1700 ps shift in lifetime between liquid ordered and liquid disordered phases providing more pronounced contrast as to the 60 nm spectral shift. Thus, we hypothesize that FLIM will more likely detect small variations in membrane orders. Nonetheless, the above mentioned techniques are yet still incapable of imaging individual microdomains below the diffraction limit of confocal microscopes, due to their tens of nanometers sizes. Therefore, the fluorescence signal will be emanating from both ordered and disordered phases, assuming these phases coexisted in the diffraction limited imaged volume. Hence, the GP value represents an average measure of membrane order in each pixel. The method proposed aims to optimize GP separation of membrane ordered and disordered phases by replacing spectral intensity terms in the main GP formula with the corresponding contrast enhancing fluorescence lifetimes. Furthermore, this method will be developed to exploit the probabilistic nature of di 4 ANEPPDHQ fluorescence emission spectra (Figure 3 . 2 ) and its corresponding fluorescence lifetimes (Figure 3 . 3 ) when the dye resides in liquid ordered and liquid disordered phases in order to estimate the relative percentage of coexistence of these two phases at a pixel level.

PAGE 88

54 Figure 3 .2 Probabilistic nature of di 4 ANEPPDHQ fluorescence emission spectra. GUVs, made from synthetic lipids, before (first row) and after (second row) cholesterol addition and HEK cells, before (third row) and after (fo u rth row) cholesterol depletion , were stained with di 4 ANEPPDH Q. For e ach channel number of photons w as detected to test the hypothesis of different probability for photon detection emitting from membrane microdomains of different order (axis were kept the same for each condition) . Di 4 ANEPPDHQ has a 60nm shift in t he emission spectra when the dye resides in the two different phases (ordered phase and disordered phase); Ch1 and Ch2 correspond to these phases, respectively. Thus, we predicted that in channel 1 will be detected more photons from ordered microdomains; w hereas channel 2 will have more photons from dis ordered microdomains. Cholesterol addition (more ordered membrane phases formation) led to the increase of photon detection in channel 1 (second row), whereas cholesterol depletion (fo u rth row) led to more ph oton detected in channel 2 , as was hypothesized. The reason why the poles and equator of GUVs look very different was due to photo selection phenomenon described previously in the figure legend of Figure 2.5. for cells images.

PAGE 89

55 Figure 3 .3 Probability of detecting the same lifetimes in two detection channels is different. To demonstrate the dissimilarity of detecting the same fluorescence lifetimes in a two channel detection configuration, GUVs made of synthetic lipids DOPC before and after cholesterol addition were used. Frequency domain FLIM data were obtained , and datase ts from the two different conditions were imported into the same phasor. Two cursors, blue and purple, corresponding to two different fluorescence lifetimes on the phasor plot were selected to highlight pixels in both channels from the two datasets. We sho wed that pixels corresponding to the same lifetime were highlighted differently in each channel under the two variant conditions. The reason why the poles and equator of GUVs look very different was due to photo selection phenomenon described previously in the figure legend of Figure 2.5. Bar = 20 m Unlike the conventional lifetime imaging of d i 4 ANEPPDHQ, in which only one detect or is used, probabilistic GP Lifetime ( method will use two spectral ranges for each lipid order (Ch1: 506 594nm and Ch2: 604 679nm.) Instead of using a full detection range for the dye emission spectrum (500 nm 800 nm, for example), the emission will be detected through two different channels corresponding to each individual phase. The probability of measuring fluorescence lifetimes belonging to membrane phases will be different in each channel. Figure 3.4 depicts probabilistic lifetime GP method rationale.

PAGE 90

56 Figure 3.4 Probabilistic lifetime GP method rationale. Assumption is that most of the photons detected in Ch1 are photons from the ordered phase. While, most of the photons detected in Ch2 are photons from the disordered phase. Since, at each pixel, the mean average lifetime is given by: Where are the relative abundances of each fluorophore with a lifetime in each channel will be decomposed as follows: C h 1: C h 2: Where and are the modified average fluorescence lifetime coefficients for liquid ordered and liquid disordered phases, correspondingly. and are the probabilities of

PAGE 91

57 detecting liquid ordered and liquid disorder ed fluorescence lifetimes in channel 1 and channel 2, respectively. These probabilities will be assumed to equal the percentage of the photon counts detected in each lipid order detection channel divided by the total number of photon counts detected in bot h channels. Hence, Probabilistic GP Lifetime ( ) will have the following formula: The relative contribution of liquid ordered and liquid disordered species in one pixel can be estimated from the following difference equation: To demonstrate the above rational argument, one might consider the following examples, Given: = 3.5 ns and = 1.8 ns the probabilities of detecting liquid ordered and liquid disordered fluorescence lifetimes are = 75% in Channel 1 and =75% in Channel 2 .

PAGE 92

58 Now, if in one pixel we have, 50% ordered population and 50% disordered population, then = (75%).(50%).(3.5 ns) + (25%).(50%).(1.8 ns)= 1.5375 ns = (75%).(50%).(1.8 ns) + (25%).(50%).(3.5 ns)= 1.1125 ns = 0.16, d= 0.425 If, 75% ordered population and 25% disordered population, = 2.081 ns = 0.993 ns = 0.35, d= 1.088 If, 25% ordered popula tion and 75% disordered population, = 0.993 ns = 1.231 ns = 0.107, d= 0.238 These examples, and others, are summarized in Table 3 . 2 Table 3 .2 Probabilistic GP Lifetime ( ) theoretical examples. Population Difference (d) 0% ordered + 100% disordered 0.5 0.9 10% ordered + 90% disordered 0.322 0.635 25% ordered + 75% disordered 0.107 0.238 50% ordered + 50% disordered 0.16 0.425 75% ordered + 25% disordered 0.35 1.088 90% ordered + 10% disordered 0.445 1.485 100% ordered + 0% disordered 0.5 1.75

PAGE 93

59 From the above examples, it can be observed that the more negative value of the difference (d) the greater the contribution of the disordered population. Similarly, the more positive value of the difference (d) the greater the contribution of the ordered population. Consequently, d images, that are psuedocolored, can simply show the relative contribution of liquid ordered and liquid disordered species at one pixel level. Noise Metrics and Measures In order to verify that t he imaging settings and acquisition parameters (such as, laser compared to background, the below noise metrics (91) were carried out. These metrics are essentially relying on the number of photons counts (N) per pixel. The main purpose of these measures is to assert that the number of photons detected is sufficient to create a high contrast to differentiate membrane signal from background signal. The mean and the standard deviation for both background and signal counts w ere calculated as shown in Figure 3 . 5 .

PAGE 94

60 Figure 3 . 5 Noise metric measurements. Background and signal (membrane) Regions Of Interest ( white boxes ) were selected, from which the mean and the standard deviation (SD) for both regions were calculated . Bar = 7.5 m From calculated mean photon counts and SD we can calculate Contrast to Noise Ratio (CNR) or Signal to Noise Ratio (SNR) from the equations shown below, to compare level of desired signal (membrane staining) to the level of the background noi se. where and are the mean counts for signal and background, and are the standard deviations for signal and background, and where is the number of counts in pixel I , respectively. CNR = 64.18 and SNR = 4928.17 were calculated for the example on figure 3 .4, which indicates a high contrast image (91) .

PAGE 95

61 The above Signal to Noise Ratio equation was modified into Average Signal to Noise Ratio ( ); to look like the traditional SNR with the background noise subtracted from each pixel. = 70.40 Calculating noise statistics of the background also further help s in choosing the correct value of thresholding. Now that the proper thresholding values can be established, the Coefficient of Variation (COV) (Relative noise %) (91) in the remaining pixels after thresholding is Where Thus, when the image will be threshold ed with N=700 counts number of photons, the Relative Noise % is ~3.78% , and if we will increase this amount to N=900 counts, the Relative Nosie % will be about ~3.3% . Detection Noise To account for detection noise (shot noise, for example), empi rical measurement on Alexa 488 dye in a solution was performed. Microscopy settings were kept the same as those used to acquire HEK cells membranes images. The image was thresholded by 100 photon counts, and its photon counts histogram was plotted, as illu strated in Figure 3 . 6 . At 100

PAGE 96

62 photon counts threshold level, Bayesian fitting does not offer any advantage over the conventional fitting models (92) . Figure 3 . 6 Alexa 488 photon counts (left panel) and calculated histogram (right panel) in a solution . From histogram we can obtain = 130.83 counts and = 10.67 counts and can calculate theoretical Therefore, detection noise (fluctuation in the number of photon counts) is approximated by: For estimators, probability distributions are called sampling distributions (93) . Since the standard deviation of the sa mpling distribution is the Random Variable (RV), and n is the number of measurements (samples)), is lower than the theoretical limit .

PAGE 97

63 Given its formula, uncertainty in lifetime GP values is basically stem med from uncertainties in the recovered lifetimes from the two detection channels (i.e. and ) The variation in lifetime due to noise is the reciprocal square root of the number of photons, i.e., the relative standard deviation of the lifetime is can be calculated for each pixel (5,94,95) . Where i and j designate the location of the pixel in the image along the x and y coordinates, and N is the number of photon counts in that pixel. For simulation purposes, the last equation was modified to account for detection noise. Since we empirically showed that our measurements are shot noise limited (Figure 3 .5), the standard deviation of the recovered lifetime as a function of photon counts can be rewritten as follows: The negative sign in the denominator of the last equation is meant to increase the variation in the fluorescence lifetime to accommodate for the largest standard deviation. This lifetime variation model was experimentally tested using Alexa 488 dye in a s olution. The dye has a lifetime of 4.1 ns. Given this known lifetime and the average number of photon counts , the standard deviation of the lifetime due to noise can be estimated, and

PAGE 98

64 compared to the empirically calculated from the recovered lifetime hi stogram (Figure 3 . 7 and Figure 3 . 8 ). Figure 3 . 7 Estimation of the fluorescence lifetime standard deviation using a dye with a known fluorescence lifetime in a solution. Alexa 488 dye in a solution was used. The dye fluorescence lifetime is 4.1 ns. Variations of the fluorescence lifetime as a function of photon counts were estimated for two cases: higher photon counts and lower photon counts.

PAGE 99

65 Figure 3 . 8 Verification of the variation in the fluorescence lifetime due to noise model. Alexa 488 dye in a solution was used. The dye fluorescence lifetime is 4.1 ns. The same two cases, higher photon counts and lower photon counts, of Figure 3 .6 were considered. From the recovered lifetime histograms: =4.15 ns with =0.1 ns (compared to 0.098 ns estimated), and =4.18 ns with =0.15 ns (compared to 0.133 ns estimated), for higher photon counts and lower photon counts, respectively. Error Propagation Model At each pixel, due to the uncertainty in the recovered lifetimes from the two detection channels, and can be dealt with as two random variables owning their means and standard deviations. Consequently, lifetime GP value becomes a random variable itself ( ) ; N is the numerator RV, and D is the denominator RV

PAGE 100

66 Calculating the statistics of this new random variable is not a straightforward process. While the means and the standard deviations of N and D can be readily calculated from the following equations, the statistics of their ratio ( ) require certain approximations. and and The mean and variance of a ratio of two random variables can be approximated by means of first and second order Taylor expansions (96,97) as follows where , , , , and are the mean of N, the mean of D, the variance of N , the variance of D , and the covariance of N and D , respectively. Random Uncertainty (Unbiased) This method was devised to test the robustness of lifetime GP formula to variations in the fluorescence lifetimes around the means. It was also used to show the similarity of membrane microdomains organization when lifetime GP membrane images were calculated from the recovered and , or from and . The random uncertainty method is schematically summarized in Figure 3 . 9 .

PAGE 101

67 Figure 3 . 9 A schematic summary of lifetime GP random uncertainty generation procedure. A complete random uncertainty in the calculated lifetime GP images can be them by and pixel by pixel. The resultant matrices can then be added to and respective matrices causing variations in the recovered fluorescence lifetimes within 68 and 95 confidence intervals. Subsequently, images can be calculated by implementing the lifetime GP formula. Hypotheses Testing Single sample z test and two sample t test were deployed to test hypotheses, and decide whether to accept or reject null hypotheses, . Single sample z test has the following test statistic, z where and are the population mean and population standard deviation, respectively (93) . is the sample mean, n is the sample size.

PAGE 102

68 Whereas, two s ample t test (also known as student t test) has the following test statistic, t w here and are the sample mean and sample standard deviation, respectively (93) . n is the sample size. Lifetime GP Pixel to Pixel Statistical Analyses On lifetime GP images, two sample Kolmogorov Smirnov (KS) tests were performed to examine the statistical significance of lifetime GP values at each pixel compared to its very contiguous neighbors. The two sample Kolmogorov Smirnov test is a nonparametric hypothesis test which evaluates the difference between the Cumulative Distribution Functions (CDFs) of the distributions of two data sets. U tilizing lifetime variation error model, each pixel in lifetime GP images now has a mean and a standard deviation. Thus, at each pixel, CDFs can be generated. Pixel to pixel statistical significance tests were carried out using kstest2 matlab function. Th e function has a syntax [h,p]=kstest2(pixel1distribution, pixel2distribution) , that returns the test decision for the null hypothesis in variable h, and the p value in variable p. At the 5% significance level, the result h is 1 if the test rejects the null hypothesis, and 0 otherwise.

PAGE 103

69 Differences from the Mean and Two Dimensional (2D) Spatial Correlation Differences from the mean analyses were done to provide a vivid visualization of contrast enhancement of lifetime GP method relative to intensity GP conventional method. The analyses were performed on the same exact FOVs. The means of intensity GP and life time GP images were calculated and subtracted from each pixel. The resulting images and their corresponding histograms can, therefore, be juxtaposed for direct visual and quantitative comparison of contrast between the two approaches pseducolor and histog ram ranges were kept the same. For 2D spatial correlation analysis, new 2 by 2 pixels averaged versions of the intensity and lifetime GP images were first created. The 2x2 averaging window was set up using fspecial matlab function, and then moved around the respective images using imfilter matlab function. After that, means of the resulting images were calculated and subtracted from each pixel. 2D spatial correlation was implemented by virtue of xcorr2 matlab function. Later, autocorrelation matrices were further normalized by their maximum correlation value. 2D spatial correlation analyses were used to elicit contrast enhancement from intensity and lifetime GP images. These images were autocorrelated to show the extension (size) of lipid microdomains wit h similar fluidity in a certain direction. In other words, an estimate average of the number of pixels over which microdomains with similar fluidity extend. The autocorrelation process is defined by the sum of the products of the overlapping pixels as a fu nction of displacement, given by the autocorrelation coefficient, c. The autocorrelation coefficient could have a positive or a negative value, as well as zero, based on the resultant sum of the products. The sum of multiplications of similar sign values ( either

PAGE 104

70 positive or negative) yields a positive autocorrelation coefficient. Otherwise, the autocorrelation coefficient has a negative value. Pixels of intensity and lifetime GP images have positive and/or negative values indicating relative phase fluidity (ordered and disordered.) Therefore, multiplication of similar phases (microdomains) results in a positive autocorrelation coefficient value. Conversly, dissimilar microdomains multiplication results in a negative autocorrelation coefficient val ue. Moreove r, the transition of autocorrelation coefficient values from positive to negative, and vice versa, implies an overall transition in the microdomains phases along the chosen autocorrelation coefficient line profiles direction. Consequently, the more c=0 cro ssing of the autocorrelation coefficient line profiles, the higher the sensitivity for detecting microdomains phase transitions which was our metric to evaluate the contrast enhancement of lifetime GP method over the conventional intensity GP measurements . More importantly, we would like to emphasize that the exact autocorrelation coefficient values are inconclusive, but rather the patern and distance between consecutive peaks of the autocorrelation coefficient line profiles. The above argument is depicted in Figure 3.10.

PAGE 105

71 Figure 3.10 Illustration of two dimensional (2D) spatial correlation concept. Lifetime GP method c ontrast enhancement implied through autocorrelation coefficient line p rofiles in a certain direction. Statistical analysis Data are presented as mean ± s.e.m, unless stated otherwise. Statistical comparison between two groups of data was made using two test. Multiple comparisons were determined using one way analysis of variance (ANOVA) followed by Tukey's m ultiple comparison tests.

PAGE 106

72 CHAPTER IV EXPERIMENTAL RESULTS AND DATA ANALYSES Membranes fluidity is essential to proteins regulation and function, and membrane proteins of interest reside in either basolateral or apical membranes. Now the question is, how can we deploy microscopy and data analysis techniques to deduce information about the activity and function of proteins of interest activity and function, in conjunction with the fluidity of their microdomains of residence? Due to the notion that pro teins might change their properties and function in native membranes and proteins in their native environment. The ultimate goal of this study is to apply advanced microscopy and analytic al methods on membranes in their intact form. However, this is not always possible due to technical challenges innate to the biological system at hand, and provisions need to be adopted to circumvent these limitations. One stipulation is to deal with the i solated form of native membranes containing proteins of interest, as in the case of NaPi2b study in this thesis. Even though investigating intact membranes, as in the case of Klotho study in this, is the desired goal in this thesis, visualizing their mic rodomains imposing a challenge. The contrast provided by current conventional fluidity measurement techniques is not good enough to localize precisely proteins of interest in their microdomains local environment for better understanding of their regulation and function. Thereby, the need to develop a method that enhances the contrast of current fluidity measurements was conceived.

PAGE 107

73 In this work, Generalized Polarization (GP) from fluorescence lifetime measurements ( ) method was developed to serve tha t very contrast enhancement purpose of membrane fluidity measurements (i.e. membranes lipid microdomains.) The following results sections showcase two different studies which we performed to examine membrane proteins of interest functions in their local m icroenvironment: in isolated (NaPi2b study), and in intact (Klotho study) membranes. Finally, GP Lifetime results aiming to enhance contrast conclude this chapter. Cell Plasma Membrane GP 3D 70% Confluent HEK cells were stained wit h di 4 ANEPPDHQ for 5 min in phenol free media. Control or Klotho (300pM added right or an hour before imaging) treated cells were imaged on the Zeiss 780 confocal micros cope . Signals from the ordered and disordered phases wer e collected. GP analysis was done as described pre viously in Chapter III . For each condition, a z stack of 2D confocal GP images was obtained, which was later reconstructed using Icy software ( an open community plat form for bioimage informatics, created by the Bio Images Analysis Unit at Institute Pasteur ) as shown in Figure 4 .1 and Figure 4 .2 , for apical and basolateral parts of the cells , respectively .

PAGE 108

74 Figure 4 . 1 3D reconstruction of GP z stacks showing apical part of the cells . Control and treated with Klotho for various times HEK cells were stained with di 4 ANEPPDHQ. Intensity z stack images were collected and GP values were each condition. 3D reconstruction of no n tresholded GP images was performed. Klotho treatment (middle panel) shifts apical cell membrane to more solid state , compare to control (left panel) , and becomes closer to control after 1hr of treatment with Klotho (right panel) . Figure 4 . 2 3D reconstruction of GP z stacks showing basolateral part of the cells Control and treated with Klotho for various times HEK cells were stained with di 4 ANEPPDHQ. Intensity z stack images were collected and GP values were each condition. 3D reconstruction of no n tresholded GP images was performed. Klotho treatment (middle p anel) shifts basolateral cell membrane to more fluid state, compare to control (left panel) , and this change is persistent after 1hr of treatment with Klotho (right panel) .

PAGE 109

75 For these three conditions (control, Klotho, and 1 hr Klotho) plasma memb rane mean GP was calculated from each respective z stacks, and plotted as a function of the distance from coverslip, as shown in Figure 4 . 3. Interestingly, the difference between basolateral and apical membranes fluidity is quite pronounced; which indicate s the discrepancy in their corresponding lipid composition. We found that basolateral membranes are more fluid, whereas apical membranes are more solid, and they responded different ly to the same treatment. Figure 4 . 3 Mean GP of the plasma membrane quantification of 3D reconstruction of GP z stacks. Mean GP for the membrane only over distance from coverslip was calculated. Control and treated with Klotho for various times HEK cells were stained with di 4 ANEPPDHQ. Intensity z stack images were collected and GP values were each condition. 3D reconstruction of no n tresholded GP images was performed. Klotho treatment shifts apical cell membrane to more solid state, compare to co ntrol, and becomes closer to control after 1hr of treatment with Klotho. In contrast Klotho treatment shifts basolateral cell membrane to more fluid state, compare to control, and this change is persistent after 1hr of treatment with Klotho.

PAGE 110

76 Isolated Memb rane Study ( Apical Membrane as an E xample, NaPi2b Study) Impact Inorganic phosphate (Pi) plays an important role in growth, development, bone formation and cellular metabolism. A physiological phosphate balance is maintained through the multi organ communi cation among intestine, kidneys and bones. Dysregulation of phosphate balance can induce many disorders ranging from osteoporosis to cardiovascular calcification. Pi cannot easily cross the cell membrane barrier and needs a special cell transporter to carr y it into the cell. Sodium dependent phosphate (NaPi) co transporters play a key role in the regulation of phosphate metabolism by mediating phosphate absorption and reabsorption in the small intestine and kidney. Despite having important biological functi ons, the molecular regulation of the intestinal sodium dependent phosphate co transport is not fully understood. Even though NaPi2b expression levels are lower in the duodenal than jejunal Brush Border Membranes (BBMs), phosphate uptake is similar in both regions; indicating similar activity of the co transporters (98) . Knowledge of the mechanisms of NaPi2b activity regulation in the intestinal epithelial cells will open new therapeutic pathways to treat phosphate balance disorders. In addition, this research will shed light on the lipid protein organization of the intest inal membrane, which is necessary for NaPi2b function within the apical membrane of the enterocytes. Since it is very hard to study proteins dynamics in the microvilli because of their packed structure and movement, Giant Unilamellar Vesicles (GUVs) are pr oposed as convenient biomimetic systems of the membrane that have been increasingly used to

PAGE 111

77 quantitatively address biophysical and biochemical processes related to protein function s (16) . Objective Investigate NaPi2b distribution into the different microdomains and how the fluidity of these membran e microdomains can influence NaPi2b diffusion ( function ) in isolated native membranes. Results Despite NaPi2b D ifferent E xpression L evels in R at D uodenum and J ejunum, the C o transporter A cti vity is Similar in Both Regions In agreement with previous publications (90,99) , we have demonstrated by western blotting and immunofluorescent staining that NaPi2b abundance was increased in the jejunal Brush Border Membranes (BBMs) compared to the duodenal (Figure 4 .4 ). Rat duodenum and jejunum tissues were fixed, cut onto glass slides and stained against NaPi2b. Microvilli were outlined with F Actin (Figure 4 .4 , A and B). Microvilli areas on the confocal images taken for the two intestinal regions were selected, and the corresponding NaPi2b fluorescence intensity signal was quantified. Jejunal NaPi2b mean intensity was significantly greater, in dicating higher level of NaPi2b expression ( 17.60±1.73 a.u. in the jejunum and 11.42±1.89 a.u. in the duodenum, p=0.024; Figure 4 .4 C ). Western blotting was done on BBMs isolated using Mg 2+ precipitation method from rat duodenums and jejunums to show NaPi2b relative abundance (Figure 4 .4 D ). NaPi2b relative expression was significantly increased in the jejunal BBMs (Figure 4 .4 E ). To examine the co transporter activity in both intestinal regions, phosphate uptake s were measured in duodenal and jejunal BBM V s. NaPi2b activity

PAGE 112

78 was similar in both regions (Figure 4 .4 F ), albeit the co transporter being in higher abundance in the jejunum. Figure 4 . 4 NaPi2b has similar activity in rat duodenum and jejunum despite different expression levels. Confocal images of rat duodenum (A) and jejunum (B) tissue sections showed higher signal for NaPi2b (green) in the microvilli of the jejunum. Microvilli are outlin ed with F Actin (red). (C). NaPi2b mean intensity was quantified in the microvilli of two intestinal regions. Jejunal mean intensity is significantly increased (p=0.0241), indicating higher level of NaPi2b expression. Comparative expression by WB (D) and c orresponding densitometry (E) for NaPi2b levels in BBMs isolated from rat duodenums and jejunums showed that NaPi2b expression was significantly increased in the jejunal BBMs (p=0.0025). milar in both regions. Bar = 2 m

PAGE 113

79 Na P i2b Resides i n t he Detergent Resistant Fractions in BBM s Isolate d from Rat Duodenum and Jejunum It was shown that in kidneys NaPi co transporter was partitioning in cholesterol , sphingomyelin , and glycosphingolipids enriched membrane microdomains (i.e., lipid rafts) (50) . Lipid rafts are known to be det ergent resistant membrane domains. Detergent Resistance (DR) and Detergent Sensitive (DS) fractions were isolated by single step high speed centrifugation in the presence of 1% Triton X 100 as previously described in (50) . We have found that NaPi2b resides in the DR fractions of BBMs isolated from rat duodenum and jejunum (Figure 4 .5 ). Western Blotting for NaPi2b and Flotillin 1 (lipid raft marker (100,101) ) was done on DR and DS fractions of both regions (Figure 4 .5 A ). Results showed that both of these two proteins reside mainly in the DR fractions, whereas they were almost absent in the DS fractions of the two regions. NaPi2b and Flotillin 1 were significantly more abundant in jejunal DR fractions compare to the duodenal DR, as shown by their corresponding relative expression quantification (Figure 4 .5 B ) and (Figure 4 .5 C ), respectively.

PAGE 114

8 0 Figure 4 . 5 NaPi2b resides in the detergent resistant fractions in BBMs isolated from rat duodenum and jejunum. Detergent resistant and soluble fractions were separated using standard protocol. Comparative analysis of DR and DS by WB for (A) NaPi2b (top) and Flotillin 1 (raft marker, bottom) showed presence of these two proteins in the DR fraction of both regions, whereas signal was almost absent in the DS fraction. Densitometry for DR fractions showed significantly higher abundance of NaPi2b (B) and Flotillin 1 (C) in the jejunum (p=0.036 and p=0.007 respectively). Jejunal Apical Membranes are More Fluid i n Intac t Native Rat Intestinal Tissues Protein activity can be regulated by the membrane microenvironment, which can be measured by membrane fluidity. Therefore, in order to investigate NaPi2b similar activity regardless of the co transporter expression level discrepancy in both intestinal regions, we

PAGE 115

81 have decided to measure membrane fluidity. Membrane fluidity can be measured by using special probes such as di 4 ANEP PDHQ, which has a blue shift in emission for the membrane liquid ordered phase relative to membranes in liquid disordered phase (2) . Live intestinal tissues (ex vivo) were stained for 5 min with the di 4 ANEPPDHQ, and imaged in two channels corresponding to ordered and disordered phases using Zeiss 780 confocal microscope. In order to p reserve tissues viability and membrane properties, images were taken within 15 minutes after dissection. GP images were calculated as described in the methods section. Duodenal and jejunal representative GP mages of apical membranes (microvilli) are shown in Figure 4 .6 A and Figure 4 .6 B , respectively. Given NaPi2b higher abundance in jejunum, and the co transporter greatest presence in jejunal DR fractions, we expected jejunal microvilli to be more solid. Nonetheless, to our surprise, jejunal microvilli were markedly more fluid, as quantitatively measured from GP images ( 0.066±0.007 for the jejunum and 0.028±0.008 for the duodenum, p=0.0077; Figure 4 .6 C ). Magnified regions from the duodenal and jejunal representative GP images showed a remarkable distr ibution of pixels with different fluidity within microvilli of the two regions (Figure 4 .6 , D and E). Duodenum had more solid uniformly mixed microdomains, whereas jejunum had bigger and more fluid microdomains, as indicated by the arrows in Figure 4 .6 , D and E, respectively (white arrows indicate more fluid microdomains and black arrows more solid) . To verify that jejunal microvilli are more fluid, cholesterol and sphingomyelin contents in BBMs isolated from duodenum and jejunum were measured and quantif ied (Figure 4 .6 , F and G, respectively). Jejunal BBMs had less cholesterol and sphingomyelin content (not significant probably due to variability of assays) , which was in agreement with more fluid jejunal microvilli obtained from GP measurements. Average m embrane fluidity for two small

PAGE 116

82 intestinal regions could not help us to determine the difference in NaPi2b activity. Protein diffusion and clustering are another factors, which can regulate protein function. Microvilli have very packed structure and in addi tion to the structural difficulties they move, which makes it very hard to study protein dynamics in them. Therefore, we decided to measure membrane fluidity and NaPi2b diffusion and clustering in GUVs, which are suitable systems of biological membranes th at are progressively used to quantitatively study lipid and protein dynamics and functions .

PAGE 117

83 F igure 4 . 6 Jejunal microvilli are more fluid in live rat intestinal tissue. Live tissues were stained with di 4 ANEPPDHQ (fluidity sensitive dye) and imaged within 15 min after dissection to preserve tissue viability and membrane properties. Representative GP images for rat duodenum (A) and jejunum (B) are shown. (C) GP values fro m each pixel were quantified. Surprisingly jejunal mean GP had significantly higher negative value, indicating more fluid membrane (p=0.0077). Membrane fluidity range was 0.9 (blue, fluid) to 0.9 (red, solid) indicated by the color map. Magnified regions from GP images for duodenum (D) and jejunum (E) showed distribution of pixels with different fluidity within the microvilli. Duodenum had more solid microdomains (yellow regions, black arrows), whereas jejunum had more and bigger fluid microdomains (blue r egions, white arrows). BBMs isolated form duodenum had more cholesterol (F) and sphingomyelin (G) content in agreement with more solid duodenal microvilli. Bar = 5 m

PAGE 118

84 G UV s Made o f Native B BM s Maintain The Properties of t he Intact Small Intestines In order to prove that GUVs are valid membrane models to study NaPi2b dynamics, we repeated membrane fluidity and NaPi2b fluorescence intensity measurements on GUVs made of native BBMs isolated from duodenum and jejunum (Figure 4 .7 ). GUVs were made using e lectro formation method as described in the methods section. Formed GUVs were co stained with the fluidity sensitive dye, di 4 ANEPPDHQ, and Alexa 647 NaPi2b. di 4 ANEPPDHQ fluorescence intensity was detected in two channels corresponding to the ordered an d disordered phases using FV 1000 Olympus confocal microscope. Average mean GP were calculated from all GUVs measurements for both intestinal regions. Representative GUVs GP images for duodenum and jejunum are shown in Figure 4 .7 A and Figure 4 .7 B , respec tively. We have found that GUVs made of jejunal BBMs were significantly more fluid compare to the GUVs made of duodenal BBMs ( 0.551±0.015 in the jejunum and 0.472±0.018 in the duodenum, p=0.003; Figure 4 .7 C ). NaPi2b was fluorescently labeled with Alexa 647 conjugated NaPi2b antibodies, and the fluorescence mean intensity from all GUVs made of duodenal (representative image, Figure 4 .7 D ) and jejunal (representative image, Figure 4 .7 E ) BBMs was quantified. Jejunal fluorescence mean intensity was signific antly greater, indicating higher NaPi2b presence in GUVs made of jejunal native BBMs ( 18.58±2.14 a.u. for the jejunum and 12.81±0.10 a.u. for the duodenum, p=0.02 5 ; Figure 4 .7 F ).

PAGE 119

85 Figure 4 . 7 GUVs made of native membranes maintain the properties of the intact intestines. GUVs made of BBMs isolated from rat duodenal (A, D) and jejunal (B, E) intestinal segments were co stained with di 4 ANEPPDHQ (fluidity sensitive dye) and Alexa 647 NaPi2b. Re presentative GP (A, B) and confocal (D, E) images are shown. (C) Jejunal mean GP had significantly higher negative value, indicating more fluid membrane (p=0.003). (F) Mean intensity showed significantly increased signal for NaPi2b (red) in the jejunal GUV (p=0.025), indicating presence of higher level of NaPi2b. Bar = 20 m Jejunal BBMs Have More Distinct a nd Bigger Lipids Microdomains Native membranes are well known to be heterogeneous. This heterogeneity can manifest itself in the shape of GP values histograms. Membrane fluidity, represented by GP values, covered a range from 0.9 (fluid) to 0.9 (solid) indicated by the color map under the

PAGE 120

86 histograms. GP images were also pseudocolored using the same color map. Therefore, in order to examine intestinal BBMs heterogeneity, we have calculated GP images, and plotted the corresponding GP values histograms for all GUVs made from duodenal and jejunal BBMs. Representative GP images of GUVs from duodenal and jejunal BBMs are shown in Figure 4 .8 A and Figure 4 .8 B , respectively. Their corresponding GUVs GP values histograms are given in Figure 4 .8 C (duodenum) and Figure 4. 8 D (jejunum). Duodenal GUVs GP histograms were narrow, and pixels with different fluidity were distributed closely around the mean. On the other hand, jejunal GUVs GP histograms were broader, and pixels spanned a wider range of fluidity around the mean, indic ating more distinct and bigger lipid microdomains, which are either more fluid or more solid than their duodenal counterparts. Some jejunal histograms had two discrete peaks (Figure 4 .8 D ) suggesting a striking difference in membrane orders of jejunal micr odomains, as illustrated by the two distinct layers of membrane fluidity in the representative jejunal GUV GP image (Figure 4 .8 B , arrows). For each intestinal region, all calculated GUVs GP histograms were combined and averaged. Average GP histograms of G UVs made of duodenal and jejunal isolated BBMs were normalized to values calculated to duodenum, and are shown in Figure 4 .8 E . To show the difference in the distribution of pixels having the same lipid order between the two intestinal regions, duodenal GU Vs average GP histogram was subtracted form jejunal GUVs average GP histogram, and the difference is illustrated in Figure 4 .8 F . Histogram difference s showed that jejunal BBMs had a greater number of pixels having more fluid phase compared to duodenal BBM s, in addition to fewer increased number of pixels having more solid phase.

PAGE 121

87 Figure 4 . 8 Jejunal BBMs have more discrete lipid microdomains. GP images for GUVs made of rat duodenal (A) and jejunal (B) BBMs are shown. Corresponding histograms for these GUVs showed narrow distribution of pixels with different fluidity for the duodenum (C), whereas jejunum (D) had broader GP histogram indicating more distinct and bigger lipid microdomains. Membrane fluidity had a range of 0.9 (blue, fluid) to 0.9 (red, solid) indicated by the color map under the histogram. (E) Average GP histograms for duodenal (black) and jejunal (red) GUVs. (F) Histograms difference showed that jejunum had a number of pixels with more fluid and more solid GP values compared to duodenum. Bar = 20 m

PAGE 122

88 Jejunum Has a Di stinct Very Slowly Diffusing NaP i2b Subset Single point Fluorescence Correlation Spectroscopy (FCS) measurements were taken for Alexa 647 NaPi2b from multiple locations at the equatorial region of each GUV made of duodenal and jejun al BMMs. Fluorescence intensity fluctuations were recorded, and autocorrelation curves were calculated and plotted against time. Autocorrelation data from duodenum were best fit with a two dimensional two components fitting model, whereas jejunum data were best fit with a two dimensional one component fitting model. Representative duodenal and jejunal autocorrelation curves and their corresponding fits are shown in Figure 4 .9 A . From fitting all data, we have obtained five different diffusion coefficients (diffusion components) represented by correspond ing to the diffusion coefficients of Alexa 647 freely diffusing dye (D 1 ) , unbound Alexa 647 NaPi2b antibodies measured out side of GUVs (D 2 ) , slow (duodenal and jejunal) NaPi2b diffusing species (D 3 ) , fast (duodenal and jejunal) NaPi2b diffusing species (D 4 ) , and very slow (jejunal, only) NaPi2b diffusing species (D 5 ) , respectively , measured in ( ). Duodenum had two significantly different diffusion components: slow (D 3 =5.63±0.30 2 /sec) and fast (D 4 =10.23±0.25 2 /sec) (Figure 4 .9 B ). Jejunum, on the other hand, had three significantly different diffusion components: slow (D 3 =3.26±0.52 2 /sec), fast (D 4 =11.30±0.40 2 /sec) and very slow (D 5 =0.33±0.09 2 /sec) (Figure 4 .9 C ). Statistical analysis on , , and NaPi2b diffusion components in both intestinal regions revealed that the slow and fast duodenal diffusion components ( and ) were of no significant difference compared to their jejunal counterparts, while jejunal unique very slow component ( ) was significantly different once compared against and in the two regions (Figure 4 .9 D ). NaPi2b relative concentration corresponding to th e different diffusion components was calculated from

PAGE 123

89 autocorrelation fitting models ( ), and is given by , , and . Similar diffusion components ( and in both regions) did not have significantly different relative concentrations ( and ) (Figure 4 .9 E ), suggesting that these components might have similar co transporter activity for these subsets in the two intestinal regions (see NaPi2b activity shown in Figure 4 .4 F ). The very slow component ( ) in jejunum had significantly different relative concentration ( ) (Figure 4 .9 E ), inferring that could be accounted for NaPi2b abundance discrepancy (shown in Figure 4 .4 E ), indicating that it could also be inactive species. NaPi2b diffusion components and relative concentrations are summarized and listed in Table 4 .1 .

PAGE 124

90 Figure 4 . 9 Jejunum has a distinct very slowly diffusing NaPi2b subset. (A) Single point FCS measurements were taken for each GUV labeled with Alexa 647 NaPi2b. Autocorrelation fluorescence data were plotted for both duodenum (black dotted line) and jejunum (red dotted line), and fitted with two and one components fitting models, respectively (sol id lines). (B) Duodenum had two significantly different diffusion components: slow (D 3 =5.63± 2 /sec ) and fast (D 4 =10.23±0.25 2 /sec ) (C) Jejunum had three significantly different diffusion components: slow (D 3 =3.26± 2 /sec ), fast (D 4 =11.30±0.40 2 /sec ) and very slow (D 5 =0.33±0.09 2 /sec ). (D) NaPi2b had two similar diffusion components in duodenum and jejunum BBMs (D 3 and D 4 ), and a unique very slow component (D 5 ) in jejunum. (E) NaPi2b relative concentration corresponding to the different diff usion components was calculated from autocorrelation fitting models. Similar diffusion components (D 3 and D 4 in both regions) did not have significantly different relative concentrations (C 3 and C 4 ) suggesting similar transporter activity for these subsets, in reference to NaPi2b activity shown in Figure 1F. The very slow component (D 5 ) in jejunum had different relative concentration (C 5 ) (p=0.00 6 ), inferring that C 5 could be accounted for NaPi2b abundance discrepancy (shown in Figure 1E), indicating that it could also be inactive.

PAGE 125

91 Table 4 .1 Average diffusion coefficients and relative concentrations of NaPi2b for duodenal and jejunal GUVs measured using single point FCS. Diffusion coefficients D e concentrations C are expressed as mean ± SEM (a.u.). n is the number of data points. Duodenum Jejunum D 1 (free Alexa 647) 305.40 ± 59.36 (n=2) D 2 (free antibodies) 47.76 ± 6.17 (n=12) 44.18 ± 12.63 (n=6) D 3 5.63 ± 0.30 (n=16) 3.65 ± 0.60 (n=8) D 4 10.23 ± 0.25 (n=20) 11.30 ± 0.40 (n=6) D 5 0.33 ± 0.09 (n=8) C 3 197.79 ± 24.10 (n=16) 50.01 ± 8.30 (n=8) C 4 125.99 ± 17.03 (n=20) 170.22 ± 13.97 (n=6) C 5 48.38 ± 23.95 (n=8) NaPi2b Forms Larger Clusters i n Jejunum It has been shown that proteins diffuse slower when they are in aggregates (50) . In order to examine whether NaPi2b different diffusion is mainly due to aggregation and clustering, we performed intensity profil es and Photon Counting Histogram (PCH) analyses from the single point FCS measurements on GUVs made of duodenal and jejunal BBMs (Figure 4 .10 ). GUVs were stained with Alexa 647 NaPi2b. The fluorescence intensity fluctuates strongly as NaPi2b labeled cluste rs diffuse in and out of the FCS measured volume over a period of time. In duodenum, D 3 and D 4 had similar fluorescence intensity fluctuations over time compared to their jejunal counterparts (D 3 and D 4 ), whereas jejunal D 5 had a stronger and slowly varying intensity fluctuations suggesting bigger NaPi2b clusters. Representative fluorescence intensity profiles for NaPi2b clusters in duodenum and jejunum that are corresponding to each diffusion components are shown in Figure 4 .10 A and Figure 4 .10 B , respectively. PCH takes advantage of the brightness of a single molecule, the labeled NaPi2b antibody in a solution in our case, to extrapolate the number of molecules in an aggregate. PCH data were fitted to a single species fitt ing model using Globals SimFCS

PAGE 126

92 Software (Laboratory for Fluorescence Dynamics, Irvine) . PCH of slow D 3 , fast D 4 , and very slow D 5 diffusing Alexa 647 NaPi2b components (their fluorescence intensity profiles are given in Figure 4 .10 , A and B) were plotted f or duodenum (Figure 4 .10 C ) and jejunum (Figure 4 .10 D ). The average molecular brightness for the different diffusion species was calculated from the fit for duodenum (Figure 4 .10 E ) and jejunum (Figure 4 .10 F ). The average molecular brightness for ( =9.64±0.35) and ( =11.34±0.54) in duodenum was about twice that of freely diffus ing Alexa 647 NaPi2b antibodies ( =4.56±0.14) obtained from solution measurements suggesting dimerization. Jejunal diffusion components had average molecul ar brightness of =14.96±0.62, =10.33±0.67, and =25.20±3.10 suggesting trimerization, dimerization, and oligomerization, respectively, for the slow ( ), fast ( ) and very slow ( ) diffusing species. PCH data are summarized and listed in Table 4 .2 .

PAGE 127

93 Figure 4 . 10 NaPi2b forms bigger clusters in jejunum. Representative fluorescence intensity profiles for NaPi2b clusters in duodenum (A) and jejunum (B) corresponding to each diffusion components are shown. The intensity fluctuates strongly as the fluorescent clusters diffuse in and out of the FCS measured volume over a period of time. In duodenum, D 3 and D 4 have similar fluorescence intensity fluctuations over time compared to their jej unal counterparts (D 3 and D 4 ), whereas jejunal D 5 has a stronger and slower intensity fluctuations suggesting bigger NaPi2b clusters. Photon Counting Histograms (PCH) of slow (D 3 , +), fast (D 4 , +), and very slow (D 5 , +) diffusing Alexa 647 NaPi2b componen ts (whose fluorescence intensity profiles are given in (A) and (B)) were plotted for duodenum (C) and jejunum (D). PCH data were fitted to a single species fitting model (solid lines). The average molecular brightness for the different diffusion species we re calculated from the fit for duodenum (E) and jejunum (F). The average molecular brightness for D 3 ( 3 =9.64±0.35) and D 4 ( 4 =11.34±0.54) in duodenum is about twice that of freely diffusing Alexa 647 NaPi2b antibodies D s ( s =4.56±0.14) obtained from s olution measurements suggesting dimerization. Jejunum had D 3 ( 3 =14.96±0.62), D 4 ( 4 =10.33±0.67), and D 5 ( 5 =25.20±3.10) suggesting trimerization, dimerization, and oligomerization, respectively.

PAGE 128

94 Table 4 .2 Photon Counting Histogram (PCH) measurements from duodenal and jejunal GUVs . dwell time of NaPi2b diffusion components in the two regions. n is the number of data point s. Region Duodenum 9.64 ±0.34 (n=14) 11.34 ±0.54 (n=13) Jejunum 14.96 ±0.72 (n=7) 10.33 ±0.67 (n=5) 25.20 ±3.10 (n=7) Very Slow Diffusing NaPi2b Aggregates Reside i n Unique More Sol id Microdomains in Jejunal BBMs Local membrane microenvironment fluidity can impact membrane protein diffusion in addition to protein size, clustering and aggregation and interaction with non membrane proteins (15,30) . Therefo re, we decided to investigate NaPi2b dynamics as a function of their local environment. To achieve that, we correlated NaPi2b various diffusion components in both intestinal regions to the fluidity of their corresponding microenvironment (Figure 4 .11 ). GUVs made of duodenal and jejunal BBMs were co stained with Alexa 647 NaPi2b and di 4 ANEPPDHQ. Single point FCS measurements were taken from multiple locations at the equatorial region of each GUV for both dyes. For each GUV GP image an average fluidity o f the area from which single point FCS measurements were taken was calculated. Representative overlaid images of GP images and their corresponding Alexa 647 NaPi2b for both duodenal and jejunal GUVs are shown in Figure 4 .11 A and Figure 4 .11 B , respectivel y. NaPi2b diffusion coefficients were calculated from each single point measurement and compared to its di 4 ANEPPDHQ diffusion counterparts, in addition to their local area GP values, point by point (Figure 4 .11 , C and D). Duodenum fast diffusion coeffici ents were corresponding to more fluid local environments, whereas, jejunum slow diffusions were

PAGE 129

95 mostly pertaining to a more solid local environment. Average local area GP values for the slow, fast, and very slow diffusion components were obtained from all GUVs measurements for both duodenum and jejunum, as presented in Figure 4 .11 E and Figure 4 .11 F , respectively. Mean local area GP values for the duodenal slow ( , = 0.457±0.009) and fast ( , = 0.477±0.007) NaPi2b diffusion components were combined and statistically compared against their jejunal counterparts ( , = 0.435±0.013 and , = 0.491±0.013), in addition to jejunal very slow diffusing component ( , = 0.362±0.030) (Figure 4 .11 G ). Results showed that and N aPi2b subsets resided in a similar microdomains environment in both duodenum and jejunum, while jejunal NaPi2b species resided in a significantly more solid microdomains. Average Local Area GP values for duodenal and jejunal NaPi2b different diffusion s ubsets are summarized and listed in Table 4 .3 .

PAGE 130

96 Figure 4 . 11 Very slow diffusing NaPi2b aggregates reside in unique more solid microdomains in jejunal BBMs. GUVs were co stained with Alexa 647 NaPi2b and di 4 ANEPPDHQ. Single point FCS measurements were taken on GUVs made from duodenal (A) and jejunal (B) BBMs for both dyes at the allocated point positions numbered 1 through 4. NaPi2b diffusion coefficients were calculated from each single point measurement and compared to its di 4 ANEPPDHQ diffusion counterparts, in addition to their local area GP values, point by point. The data were summarized in (C) for duodenum and (D) for jejunum. Duodenum fast diffusion coefficients were corresponding to more fluid local environmen ts, whereas, jejunum slow diffusions were mostly pertaining to a more solid local environment. Average local area GP values for the slow, fast, and very slow diffusion components were obtained from all GUVs measurements for both duodenum (E) and jejunum (F ). (G) D 3 and D 4 NaPi2b subsets reside in a similar microdomains environment in both duodenum and jejunum, while jejunal D 5 has a significantly (p<0.001) different local area GP, suggesting that NaPi2b species reside in a more solid microdomains. Bar = 2 0 m

PAGE 131

97 Table 4 .3 Average local area Generalized Polarization (GP) values for duodenal and jejunal GUVs. Mean local GP values were calculated from areas, where NaPi2b single point FCS measurements taken from. Different NaPi2b diffusion species were correlated to their corresponding mean local area GP. n is the number of data points. Region Duodenum 0.457 ±0.008 (n=13) 0.477 ±0.007 (n=12) Jejunum 0.435 ±0.013 (n=6) 0.491 ±0.013 (n=5) 0.362 ±0.029 (n=9) Jejunal BBMs Form Bigger M icrodomains We have found that distinct subsets of NaPi2b reside in special microdomain s with different fluidity. Not just protein aggregation and microdomains fluidity, but also the size of these microdomains can play a role in the protein activity . In order to estimate the size of the microdomains where NaPi2b resides, we decided to exploit the temporal fluctuation of the fluidity sensitive dye di 4 ANEPPDHQ intensity to extract the diffusion time for those microdomains, by using FCS measurements. Ow en et al have found that the size of lipid microdomains is inversely proportional to their translational diffusion, and can be estimated accordingly. In other words, the faster the diffusion, the smaller the size of the microdomains (102) . Therefore, GUVs made of duodenal and jejunal isolated BBMs were co stained with Alexa 647 NaPi2b and di 4 A NEPPDHQ. Single point FCS measurements were taken from multiple locations at the equatorial region of each GUV for both dyes. di 4 ANEPPDHQ autocorrelation data were fitted with a two dimensional one component fitting model. Representative di 4 ANEPPDHQ au tocorrelation curves and their corresponding fits for both duodenum and jejunum are shown in Figure 4 .12 A . Since Alexa 647 NaPi2b and di 4 ANEPPDHQ single point FCS measurements were taken from the same location on GUVs, di 4 ANEPPDHQ diffusion coefficients were organized according to the Alexa 647 NaPi2b different diffusion components determined previously a nd assigned analogous nomenclature

PAGE 132

98 (i.e., , , and for the slow, fast, and very slow diffusion species in both intestinal regions). We found that, in duodenum, both the slow and fast diffusing NaPi2b subsets reside in microdomains with si milar diffusion coefficients ( =1.33±0.18 2 /sec and =1.65±0.12 2 /sec, respectively), indicating comparable and small sizes of microdomains (Figure 4 .12 B ). In the jejunum, however, slow and fast diffusing NaPi2b components reside in microdomains with slower and yet still similar diffusion coefficients ( =0.54±0.21 2 /sec and =0.87±0.44 2 /sec, respectively), whereas the very slow NaPi2b species resided in very slow diffusing lipid microdomains ( =0.15±0.05 2 /sec); indicating overall bigger microdomains for jejunum (Figure 4 .12 C ). Average diffusion coefficients of di 4 ANEPPDHQ for duodenal and jejunal GUVs are summarized and listed in Table 4 .4 .

PAGE 133

99 Figure 4 . 12 Jejunal BBMs form bigger microdomains. (A) Single point FCS measurements were taken for each GUV stained with fluidity sensitive dye di 4 ANEPPDHQ. Autocorrelation fluorescence data were plotted for both duodenum (black dotted line) and jejunum (red dotted line), and fitted with one component fitting mod el (solid lines). Diffusion coefficients for di 4 ANEPPDHQ were organized according to the Alexa 647 NaPi2b diffusions taken from the same point for each GUV. D di 4 could be correlated with the domain size (ref) the faster diffusion would correspond to t he smaller size. (B) In the duodenum both, slow and fast diffusing, NaPi2b subsets reside in microdomains with similar diffusion coefficients (D 3di 4 =1.33±0. 1 9 2 /sec and D 4di 4 2 /sec, respectively), indicating comparable and small sizes. (C) In the jejunum slow and fast diffusing NaPi2b components reside in microdomains with slower, compare to the duodenum, but still similar diffusion coefficients (D 3di 4 2 /sec and D 4di 4 2 /sec, respectively), whereas very slow NaPi2b s pecies reside in even slower diffusing lipids (D 5di 4 2 /sec), indicating overall bigger microdomains.

PAGE 134

100 Table 4 . 4 Average diffusion coefficients of di 4 ANEPPDHQ for duodenal and jejunal GUVs measured using single point FCS. Diffusion coefficie nts D are expressed as mean ± SEM in units of . n is the number of data points. Region D 3di 4 D 4di 4 D 5di 4 Duodenum 1.33 ±0.19 (n=12) 1.65 ±0.12 (n=12) Jejunum 0.54 ±0.21 (n=5) 0.87 ±0.44 (n=4) 0.15 ±0.05 (n=6) Intact Membrane Study ( Basolateral Membrane as an Example, Klotho Study) Impact Soluble klotho is the shed ectod omain of the antiaging membrane klotho that exhibits pleiotropic actions, including down regulation of growt h factor driven PI3K signaling, contributing to l ifespan prolongati on, cardioprotection, and tumor inhibition. Whether membr ane receptors exist for soluble klotho is unknown. We identify lipid rafts as receptors for soluble klotho. We show klotho bind s specific sialic acid residues of gangliosides highly enriched in the outer leaflet of lipid rafts. Klotho binding to gangliosid es modulates lipid organization and inhibits lipid raft dep endent PI3K signaling. In vivo, klotho deficient mouse hearts have heightened raft dependent PI3K signaling vs. wild type hearts. We reveal a novel physiological regulator of lipid raft s tructure and function, and open new research to understand pleiot ropic effects of klotho in cell signaling and metabolism (11) . Objective Determine the target and an effect on membrane fluidity for soluble Klotho in native intact basolateral membranes of HEK cells.

PAGE 135

101 Results sKL Binds Lipid Rafts an d Modulates Lipid Organization w ithin Rafts The resolution of confocal microscop y is limited by the diffraction of light to approximately 250 nm depending on wavelength . To support t hat klotho targets lipid rafts, we conducted Förster resonance energy transfer (FRET) studies. FRET detects molecular interactions on a scale of 1 10 nm, smaller than the lowest size limit of rafts (10 200 nm). FRET was measured by fluoresce nce lifetime im aging microscopy (FLIM) and analyzed by the pha sor approach (56) . As a control, lifetime for GM1 (donor) in the absence of ch olesterol (acceptor) has longer values, and sKL treatment did no t affect lifetime for GM1 alone (Figure 4 .13 A , Top; blue co lor corresponds to longer lifetime values). Lifetime for GM1 shifted to the shorter val ues (purple) in the presence of cholesterol, indicating quenching of GM1 by cholesterol (i.e., FRET occurrence b etween GM1 and cholesterol; Figure 4 .13 A , Middle ). Klotho treatment decreased FRET between GM1 and cholesterol lifetime for GM1 sh ifted back to the longer values (Figure 4 .13 A , Bottom). Phasor plot ana lysis supported the validity of FLIM FRET data (Figure 4 .13 B ). T he results indicate that klotho tre atment decreased FL IM FRET occurrences between GM1 and cholesterol, supporting an ef fect of klotho binding on lipid organization within rafts. Lipid rafts feature a high degree of membrane order. sKL decreases membrane order ( next section and Figure 4 . 15 ), further supporting that i t binds lipid rafts to modulate lipid organization. (11) .

PAGE 136

102 Figure 4 . 13 Klotho binds lipid rafts and alters lipid organization. ( A) Cells were stained with BODIPY FL 505/510 C5 GM1 (donor) with or without CholEsteryl BODIPY 542/563 C 11 (acceptor), excited with a two photon laser at 900 nm, and emission was collected at 506 594 nm using FLIM. Merged intensity images (showing green BODIPY GM1 with some internalization) and pseudocolor lifetime images are shown. Experiments were performed four times wit h similar results. (Scale bars, B ) Phasor plot of fluorescence lifetime histogram from cells stained with GM1 only (Top) and the trajectory of F RET between GM1 and cholesterol (Bottom). Blue circle marks lifetime for donor only, purple circle for donor + acceptor, and green circle for background a utofluorescence (from unstained cells). Phasor plot analysis showed FRET efficiency 25%, with fractional contribution of lifetimes 51% from quenched do nor, 44% from unquenched donor, and 5% from backgrou nd. (11) . Klotho Binds Lipid Rafts in Live Cells by Interacting with Raft Associated GM1 To further support that klotho indeed binds to monosialogangliosides cluste red in lipid rafts, we examined klotho interaction with GM1 in live cells by studying FRET between

PAGE 137

103 fluorophore labeled klotho and BODIPY GM1. Addition of fluorophore labeled klotho causes fluorescence que nching of cell membrane BODI PY GM1 analyzed by FLIM FRET in 10 min (Figure 4 . 14 A and B), ind icating klotho and GM1 interact within a 10 nm distance. In cells pretreated w CD to disrupt lipid rafts, BODIPY GM1 remained p artitioned into cell membranes, but addition of fluor ophore labeled klotho failed to quench GM1 fluorescence (Figure 4 . 14 C and D ). As a control, addition of unlabeled klotho to normal c ells did not cau se quenching of BODIPY GM1 (Figure 4 . 14 E). The res ults provide direct evidence to support that klotho targets raft associated GM1. (11) .

PAGE 138

104 Figure 4 . 14 Klotho GM1 interaction live cell membranes analyzed by FLIM FRET. (A ) HEK cells were stained with BODIPY FL 505/510 C5 GM1 (donor; 100 nM). Lifetime for GM1 alone has longer values (blue color cursor regions with longer lifetime values). Lifetime for GM 1 shifted to the shorter values (purple color cursor) in the presence of fluorophore labeled sKL (acceptor; 300 pM) over 10 min, indicating q uench ing of GM1 by sKL and FLIM FRET occurrence between GM1 and klotho. (B) Ph asor plot analysis of FLIM FRET data showed FRET efficiency 20%, with fractional contribution of lifetime s 36% from quenched donor, 55% fro m unquenched donor, and 9% from backg round. In the trajectory, blue circle marks li fetime for donor plus unlabeled klotho, purple circle for donor plus l abeled klotho, and green circle for background autofluorescence. (C and D ) Cells were pretreated with lipid raf ts. GM1 lifetime after addition of fluorophore labeled sKL was comparable to GM1 only, indicating little to no FRET occ urrence between GM1 and klotho. Shown is representative of three separate experiments with similar findings. (E ) Addition of unl abeled klotho does not cause shift to shorter values (vs. A ), indicating no quenching of BODIPY GM1 fluorescence by unlabeled klotho. (Scale bars, . (11) .

PAGE 139

105 Klotho Decreases Membrane Order Analyzed by Generalized Polarizat ion The biophysical hallmark of lipid raf ts is a high degree of membrane order. We examined the effect of k lotho on membrane order using a polarity sensitive membrane p robe that undergoes an emission spectral shift when residing in th e liquid ordered vs. disordered phase (2) . Generalized polarization (GP ), the ratio of fluorescence intensity recorded in two spec tral channels, becomes increasingly more negative within 5 10 min of klotho t reatment, indicating decreasing membrane ord er (Figure 4 .15 A and B ).The effect of klotho on GP is reversib le after klotho washout (Figure 4 .15 C ). Together, the above results provide compelling sup port for the notion that klotho binds to lipid rafts a nd that binding modulates lipid organization within rafts. (11) .

PAGE 140

106 Figure 4 .15 Klotho decreases membrane order analyzed by using a polarity sensitive membrane probe. (A) Cells were incubated with Di 4 klotho (300 pM) for 5 min at 37 °C before imaging (5 s per image con secutively for total 300 images over the subsequent 25 min). Line plots show mean GP values for 300 single images over time. (B) GP was calculated by ratiometric measurement of the fluorescence i ntensity graph shows reversibility of GP (taken at 30 min) after washout of sKL. (11) . Probabilistic GP Lifetime ( ) Method Impact Visualization of individual rafts is very difficult due to the resolution of the current conventional diffraction limited microscopy techniques (~250nm) and very small raft size.

PAGE 141

107 In the last decade a number of new techniques helping to visualiz e rafts has been developed and applied . They include single molecule spectroscopy and advanced microscopy techniques such as FCS , FLIM, FLIM FRET and others . But , even using these techniques the nanoscale lipid domains have never been directly visualized (52) . Despite t he limitations these techniques were able to confirm the existence of nanoscale cholesterol based assemblies of lipids and proteins in the membranes of living cells . Thus, enhancement to Generalized Polarization (GP) measurement could be a great tool for better separation and characterization of liquid ordered and liquid disordered phases. This method might provide a better m embrane organization visualiz ation i n addition to its potential to estimate the relative percentage of coexistence of these two phases at a pixel level . More details can be found in the technical approach section. Objective Develop a Probabilistic GP Lifetime method for existing environment sensitive dye to enhance co ntrast for membrane microdomains visualization. Results Intensity GP Measurements treated (5mM for 3 hours) cells were stained with di 4 ANEPPDHQ dye, and imaged in two separate detection channels, corresponding to ordered and disordered phases, simultaneously. GP images were calculated from the conventional intensi ty GP formula (53)

PAGE 142

108 Where and are the blue shifted and red shifted emission intensity maxima for the fluorescent dye used (which is di 4 ANEPPDHQ), respectively. images, and their corresponding histograms, are depicted in Figure 4 .16. Figure 4 .16 Conventional intensity GP measurements. Control and cholesterol depleted ( ) HEK cells were stained with di 4 ANEPPDHQ. Conventional Intensity GP were calculated for each condition. Calculated histograms showed that shifted more to the left compared to control, indicating m ore fluid phase. Single Channel Detection of di 4 ANEPPDHQ Dye Lifetime ( TCSPC FLIM) treated (5mM for 3 hours) cells were stained with di 4 ANEPPDHQ dye. Emission signal was det ected in a single channel using Time Correlated

PAGE 143

109 Single Photon Counting detection mode. ISS VistaVission software was used to fit di 4 ANEPPDHQ fluorescence lifetimes over the whole image frame. Data were fitted to a single component decay model, and fluore scence lifetimes were extracted for each pixel. Chi square criterion was used to show the goodness of the fit. Lifetimes maps images, for control and treated cells, in conjunction with their corresponding chi square and lifetime histograms are shown i n Figure 4 .17. Figure 4 .17 Single channel detection di 4 ANEPPDHQ lifetime ( TCSPC ) . Control and ere stained with di 4 ANEPPDHQ. Control and ifetime s were extracted for each condition. counts and 1020 counts, respectively. Calculated histograms showed histogram shifted towards shorter lifetimes compared to control indica ting more fluid phase. Single Channel Detection of di 4 ANEPPDHQ Dye Lifetime (Frequency domain FLIM) HEK treated (5mM for 3 hours) cells, as well as GUVs made of DOPC, were stained with di 4

PAGE 144

110 fluid, adding cholesterol to DOPC GUVs makes them more solid. Emission signal was detected in a singl e channel using FastFLIM box A320 (ISS company) . Fractional analysis from the phasor approach of FLIM data was done using simFCS software, as shown in Figure 4 .18. From these results, it can be observed that membrane microdomains are hard to resolve. Fig ure 4 .18 Single channel detection di 4 ANEPPDHQ lifetime (Frequency domain). Fractional analysis for cells and GUVs was carried out from their corresponding phasors. Calculated normalized histograms showed that cholesterol enrichment shifted the histogram the histogram towards more fluid microdomains compared to control in both conditions . The reason why the poles and equator of GUVs look very different was due to photo selection phenomenon described previously in the figure legend of Figure 2.5.

PAGE 145

111 Dual Channel Detection of di 4 ANEPPDHQ Dye Lifetime ( TCSPC FLIM) Instead of detecting di 4 ANEPPDHQ emission signal in a single channel, as we saw previously, two detection channels corresponding to each lipid order phase were used. At a pixel level, we are interested in the probabilities of de tecting fluorescence lifetimes belonging to either liquid ordered or liquid disordered phases in the two designated detection channels. and equations mentioned in chapter III can be applied at each pixel, and images of the percentage of photons distributions in both channels, in addition to their corresponding histograms can be obtained. These calculations were performed on HEK treated cells, and the results a re shown in Figure 4 .19. Figure 4 .19 Probabilities of detecting liquid ordered and liquid disordered fluorescence lifetimes. Percentages of the photon counts in each channel relative to the total number of the photon counts detected in both channels wer e calculated at each pixel for control HEK cells and after cholesterol depletion .

PAGE 146

112 As mentioned earlier, intact live cells plasma membranes are more packed and tend to exhibit a higher order. Therefore, and according to fluorescence lifetime GP rationale, it was predicted to detect more photon counts in the ordered phase channel (i.e., C h1). Percentages of photon counts histogram of HEK control cells in Ch1 confirmed our prediction (Figure 4 .19) with a mean centered around 60% relative to about 40% in the disordered channel (i.e., Ch2). id by extracting cholesterol. Thus, it was predicted that the probability of detecting photons in the ordered phase channel to be less. From Figure 4 treated cells did indeed decrease to about 50 % compared to HEK control cells, as indicated by the corresponding percentages of photon counts histogram. The above percentages were assumed to represent probabilities of detecting liquid ordered and liquid disordered lifetimes in the respective channel. We hypothesized that the shift of percentages of photon counts (henceforth treated cells in Ch1 relative to control was due tests, namely single sample z test and two sample t test, were performed. Percentages of photon counts histograms in Ch1 of both conditions were fitted to a least square Gaussian fitting model to extract their means and standard deviations (Figure 4 .20).

PAGE 147

113 Figure 4 .20 Histograms fitting of percentages of photon counts in Ch1 for HEK control treated cells. Histograms were fitted to a normal Gaussian fitting model. From the fit, means were 0.562 and 0.524, and standard deviations were 0.019 and treated cells, respectively. For single sample z test, the null hypothesis : There is no shift in photon counts % in Ch1 under M CD treatment compared to control . Applying the z test statistic formula introduced in Chapter III, ; P(z< 57.689)=0 and P(z>57.689)=0 . Hence, was rejected . For two sample t test, the null hypothesis : There is no shift in photon counts % in Ch1 under M CD treatment compared to control . Applying the t test statistic formula introduced in Chapter III, . At the 1% level of significance, =0.01, the critical value, = =2.326. Therefore, the rejection region is [2.326, ). Hence, was rejected .

PAGE 148

114 Lifetime GP ( ) Method Contrast Enhancement To test and verify the contrast enhancement of lifetime GP ( ) method against the conventional intensity GP method, measurements on the same exact Field Of Views (FOVs) and taken with the same exact imaging settings were needed for direct comparison p urposes. HEK control cells were stained with di 4 ANEPPDHQ dye. GP measurements were taken on Zeiss 780 confocal microscope with TCSPC card. The environment sensitive dye was excited using 900 nm two photon laser. Fluorescence intensity and lifetime images were detected simultaneously in Ch1:506 594nm and Ch2:604 679nm corresponding to liquid ordered and liquid disordered phases, respectively. To eliminate fluorescence signals coming from the cytoplasm and intracellular membranes, images were thresholded by 900 photons/pixel using ISS VistaVision software, to account for cells plasma membranes predominantly. With this 900 counts/pixel thresholding, the relative noise percentage per pixel was calculated to be ~3% ( see Chapter III for more details ; Noise Me trics and Measures section.) Intensity GP images were calculated by implementing the conventional GP formula on fluorescent signals from Ch1 and Ch2 (Figure 4 .21). Images of TCSPC FLIM data in each channel were fitted to single exponential decays pixel by pixel using ISS VistaVision software. Representative lifetime maps and their corresponding chi square images for both channels are shown in Figure 4 .21. Subsequently, lifetime GP images, as well as lifetime GP difference images (d), were calculated by appl ying their proposed formulas on the recovered lifetimes from each channel (Figure 4 .21) .

PAGE 149

115 Figure 4 .21 Contrast enhancement of lifetime GP ( ) method over intensity GP measurements. Control HEK cells were stained with di 4 . TCSPC FLIM data in two channels was collected . Ch1 and Ch2 images were thresholded by 900 counts before lifetime s for each channel were extracted. The average number of photon counts for Ch1 and Ch2 images were 1368 counts and 1176 counts, respectively. Conventional GP (top row) , lifetime GP (middle row) and lifetime GP ( ) difference (last row) were calculated. Histogram for each calculation was plotted. We have found that the standard deviations of lifetime GP ( ) measurements were as twice as that of intensity GP measurements. Bars As shown in Figure 4 .21, histograms of intensity GP, lifetime GP, and lifetime GP difference images were generated, from which means and standard deviations were obtained. These mean plasma membrane outlined by white rectangle (Figure 4 .21), were summarized and listed in Table 4 .5

PAGE 150

116 Table 4 .5 Means and standard deviations of intensity GP, lifetime GP and lifetime GP difference images. Data were presented as mean ± standard deviation. Standard deviation of lifetime GP method is twice as that of intensity GP. Condition Intensity GP Lifetime GP Lifetime GP difference Cell Membrane Cell Membrane Cell Membrane Control 0.08 ± 0.05 0.09 ± 0.04 0.05 ± 0.08 0.06 ± 0.07 0.26 ± 0.42 0.28 ± 0.37 We have found that the standard deviations of lifetime GP ( ) measurements were as twice as that of intensity GP measurements. Standard deviation is the square root of the variance. It is known that variance is a measure of contrast (103) . In order to conclude that histograms broadening was in fact due to contrast enhancement p ostulated by this method, uncertainty measurements of lifetime GP method were used . Uncertainty Measurements In order to verify the contrast enhancement of lifetime GP ( ) method, the subsequent uncertainty measures were conducted. Uncertainty in l ifetime GP of h omogeneous s ample . The statistically rigorous error propagation analysis model ( Chapter III) was used to simulate the uncertainty in fluorescence lifetime GP values of a homogeneous sample. For simulation purposes, the following assumption were made: Given

PAGE 151

117 Taken the empirically proven shot noise limited detection noise of our measurements into account, the following Random Variables (RVs) were recovered for statistical comparison purposes. RVs below were expressed as means±standard deviation. propagation model) were generated using matlab (Figure 4 .22).

PAGE 152

118 Figure 4 .22 Simulation results of noisy lifetime GP image of homogeneous sample implementing error propagation analys is model. In the ideal world, noise free lifetime GP images would have a standard deviation of 0. However, due to detection noise and variations in the recovered fluorescence lifetimes in both channels, real world lifetime GP images have a non zero standard deviatio n. Pseducolor ranges were kept the same. These images provided a visual appreciation of the variations of lifetime GP of a hypothetical single (homogeneous) phase order due to uncertainties, paving the way for a more realistic comparison to actual lifet ime GP measurements of cells plasma membranes, where more than one phase order exists (Figure 4 .23).

PAGE 153

119 Figure 4 . 2 3 Actual Lifetime GP standard deviation is about four times greater than Lifetime GP variation due to noise. Pseducolor ranges were kept the same. We found that lifetime GP standard deviation of our live cells membrane measurements were about four times greater than that of lifetime GP variation as a result of noise obtained from the error propagation analysis model. Uncertainty in lifetime G P ( ) membrane organization due to deviations of fluorescence lifetimes from their recovered means. Lifetime GP images of cells plasma membrane showed a unique organization of membrane phases. To look for morphological changes in membrane organization as a result of variations in the recovered lifetimes from both channels, two uncertainty measurement app uncertainty approach, images were calculated by skewing recovered fluorescence lifetime means in both channels by either one standard deviation (within 68% confidence

PAGE 154

120 interval) or two standard deviations (95% con fidence interval) in the positive or negative direction (Figure 4 .24 and Figure 4 .25). The directionality of this deviation from the mean was applied to the pixels in both channels with equal probability; i.e., one(two) standard deviation(s) either added o r subtracted from fluorescence lifetime means of all pixels. This mean of recovered lifetimes happened in one direction to all pixels involved. Figure 4 .24 Biased uncertainty in lifetime GP images in the 68% confidence interval. Morphological changes in the organization of membrane microdomains due to uncertainties in the recovered lifetimes can be examined. Very subtle, if any, changes in the organizatio n of these microdomains were observed .

PAGE 155

121 Figure 4 .25 Biased uncertainty in lifetime GP images in the 95% confidence interval. Morphological changes in the organization of membrane microdomains due to uncertainties in the recovered lifetimes can be examined. Very subtle, if any, changes in the organization of these microdomains were observed. Another, more applicable alternative was the random (un biased) approach. Using random unc ertainty method described in details in Chapter III, variations within the 68% or 95% confidence intervals from the recovered lifetime means were applied to each pixel in the two detection channels, independently, in a complete random fashion. This latte r approach allowed us to investigate the statistical robustness of lifetime GP ( ) method as a function of complete randomness in fluorescence lifetime variations in both channels (Figure 4 .26).

PAGE 156

122 Figure 4 .26 Randomness simulations show the robustness of the Lifetime GP method in the sense of maintaining standard deviation. Lifetime GP ( ) pixel to pixel statistical significance. To calculate probability that each pixel is different from its n eighbors piece wise, two sample Kolmogorov Smirnov (KS) test was performed on lifetime GP representative image. While this probability difference might arise from directionality of how images of samples were taken, our results proved insensitive (Figure 4 . 27 and Figure 4 .28).

PAGE 157

123 Figure 4 .27 KS test . Two sample Kolmogorov Smirnov test was used to statistically compute how significantly different (5% significance level) each lifetime GP pixel to its left, right, up and down adjacent neighbors. Null hypothesis is the two pixels are coming from the same distribution. The test returns 1 (yellow) when the null hypothesis is rejected, and 0 (purple) otherwise.

PAGE 158

124 Figure 4 .28 P values of KS test image in Figure IV.27. P values were all below 0.005 level. These results suggested that we had equal confidence in different pixels based on direction. Lifetime GP ( ) Method Contrast Enhancement Revealed by 2D Spatial Correlation We have used earlier the variance argument and histograms broadening analysi s as tools to demonstrate and verify lifetime GP method contrast enhancement over the conventional intensity GP measurements. We showed that the standard deviation of lifetime GP method for cells and membranes fluidity measurements was twice that of intens ity based (Table 4 .5).

PAGE 159

125 Alternatively, a two dimensional (2D) spatial correlation analysis approach was used to manifest method contrast enhancement. As described previously in Chapter III, means of intensity and lifetime GP images were subtracted from their respective originals (Figure 4 .29), and the resulting images were autocorrelated in the x and y directions. Figure 4 .29 Differences from the means. Intensity versus lifetime GP images and their corresponding histograms comparison are presented. Intensity GP has the exact same ranges of lifetime GP. Results showed h istogram broadening of lifetime GP method as a demonstr ative way of contrast enhancement. Albeit this approach lacks the specificity and direct mapping of membrane microdomains and the determination of their sizes and distribution, it has given us a general idea and approximate estimation of the smallest microdomains sizes that these intensity and lifetime GP methods can reveal.

PAGE 160

126 Nyquist sampling criterion states that in order to reconstruct an object in the spatial frequency domain, the object has to be sampled twice. A pixel is the sampling element in a n imaging system. Since we have used 250 nm pixel size, the smallest object that can be resolved is about 500 nm. Three different membrane Field Of Views (FOVs) from three separate measurements were analyzed by means of 2D spatial correlation analysis, a fter a 2 by 2 moving average processed, to avoid any aliasing due to a non Nyquist imaging acquisition. Normalized 2D autocorrelation coefficient heatmaps and line profiles plots for representative intensity and lifetime GP images are shown in Figure 4 .30, Figure 4 .31, and Figure 4 .32. Figure 4 .30 2D spatial correlation after 2 by 2 spatial filtering (1 st FOV) . The comparison of conventional Intensity (left top panel) and Lifetime GP (right top panel) images . Line profiles for both methods (bottom row) show that in Lifetime GP objects can be resolved over ~ 3 pixels compare to ~ 8 pixels in the intensity GP.

PAGE 161

127 Figure 4 .31 2D spatial correlation after 2 by 2 spatial fi l tering (2 nd FOV) . The comparison of conventional Intensity (left top panel) and Lifetime GP (right top panel) images . Line profiles for both methods (bottom row) show that in Lifetime GP objects can be resolved over ~ 3 pixels compare to ~ 8 pixels in the intensity GP. Figure 4 .32 2D spatial correlation after 2 by 2 spatial filtering (3 rd FOV) . The comparison of conventional Intensity (left top panel) and Lifetime GP (right top panel) images . Line profiles for both methods (bottom row) show that in Lifetime GP objects can be resol ved over ~ 3 pixels compare to ~ 8 pixels in the intensity GP.

PAGE 162

128 The 2D autocorrelation coefficient line profiles have revealed that, on average, the smallest microdomains sizes that can be resolved extend over ~8 pixels and ~3 pixels in intensity GP and Lifetime GP images, respectively, along the directions these line profiles were chose n . After 2 by 2 pixels spatial filtering (averaging) of intensity GP and lifetime GP i mages, standard deviation of lifetime GP is still about twice that of intensity GP (Intensity GP: 0.097±0.023, lifetime GP: 0.056±0.036) .

PAGE 163

129 CHAPTER V DISCUSSION In this work, the fundamental role of local membrane fluidity of lipid microdomains, in con trast to global average membrane fluidity, was investigated through a problem which a lot of investigators have long been trying to solve . The question is regarding the activity of a sodium dependent phosphate co transporter protein in the small intestine. . The sodium dependent phosphate co transporter type 2 b (NaPi2b) protein is a transmembrane protein with seven Tr ansmembrane Domains (TMDs) . Studies have shown that this protein is disproportionally expressed in the duodenum and jejunum regions of the sma ll intestine. Despite the fact that the protein had a higher expression levels in the jejunum, its activity was similar to that of the duodenum, which suggests that there might be some mechanisms that keep certain proportion of the transmembrane protein in active. The question was how to identify these underlying mechanisms. It is a well established fact that lipid protein interactions play a significant role in proteins regulation, activity, and function. For transmembrane proteins, there are many suggest ions that their activity can be influenced by the physical properties of their membrane microdomains of residence . Some of these properties are lipid composition and order. The hydrophobic sites of transmembrane proteins, TMDs, are believed to influence the partitioning of these proteins in certain lipid environment based on matching mismatching ngth with the acyl chain lengths of their lipids of residence. As we have introduced earlier (Chapter II), raft concept plays an essential role in membrane cell biology. We also mentioned that membrane proteins could reside in either

PAGE 164

130 raft like or non raf t lipid environment which affects their activity. While some membrane proteins were active in rafts, others were not. Present NaPi2b apical membrane study was inspired by a research done on sodium dependent phosphate co transporter type 2 a (NaPi2a) prot ein in the kidney. In spite of NaPi2a increased abundance in the apical membranes under potassium deficiency dietary, protein transport activity decreased relative to control diet. Inoue and colleagues (50) hav e shown that NaPi2a protein predominantly resides in cholesterol , sphingomyelin , and glycosphingolipid enriched membrane domains (raft like) which are more solid, and its diffusion in potassium deficiency was 2 fold less compared to control, which could be associated with the decreased cotransport activity. Following a similar line of reasoning, we therefore decided to measure duodenal and jejunal membrane fluidity, and the corresponding NaPi2b protein diffusi o n therein. Fluidity measurements on GUVs ma de of isolated native Brush Border Membranes (BBMs) from both intestinal regions were performed using Intensity Generalized Polarization (GP) method . In addition, NaPi2b diffusion coefficients were determined by single point Fluorescence Correlation Spectr oscopy (FCS) measurements. We identified two different diffusing species of NaPi2b in the duodenum. In the jejunum, however, three dissimilar NaPi2b diffusing species were identified, two of which had similar diffusion rates to those of duodenum. The third jejunal diffusing species , however, revealed a very slow diffusion component compared to the rest. Therefore, we expected jejunal membranes to be more solid. Yet, to our surprise, total average membrane fluidity measurements yielded the opposite. Jejunum membranes were more fluid as revealed by average average GP measurements on GUVs made from the jejunal BBMs. Therefore , we decided to correlate

PAGE 165

131 single point FCS diffusion measurements to local microenvironments from which these FCS measurements were taken. We found that t he NaPi2b species with common diffusion rates in both regions reside in a similar microenvironment, whereas the jejunal very slow components reside in a distinctly more solid microenvironment. According to previous work on the NaPi2a protein, these unique jejunal species could be inactive and thus give rise to the similar transport activity of NaPi2b protein in both intestinal regions , despite of its increased expression levels in the jejunum. Therefore, we emphasize that global average membrane fluidity could be misleading, and protein localization, diffusion, activity needs to be correlat ed to the local membrane microenvironments. As stated earlier, not only do membrane lipid microdomains interact with membrane proteins to influence their activity and function, but also proteins themselves play a key role in microdomains formation and regulation via specialized mechanisms. For instance, some proteins bind cholesterol, sphingomyelin, and glycosphingolipi ds and induce raft formation (15,41 47) . While other proteins form rafts by means of transbilayer interactions with immobilized lipids in the inner leaflet. Raft association and disassociation processes are very important because they trigger cascades of downstream signaling pathways (11) . Studying these cellular processes in live cells is an ultimate goal for researchers. Direct visualization of membrane microdomains formation and regulation imposes great challenges, given the optical resolution of the current advanced microscopy techniques and the lack of good enough contrast provided by these techniques. Nonetheless, indirect imaging methods still can be used to infer information about microdomains dynamics and

PAGE 166

132 lipid protein interactions; however, at the expense of more cumbersome experiments, and sometimes the need to manipulate membrane fluidity to extract answers to the biological questions at h and (11) . For example, in order to identify soluble Klotho (sKL) protein membrane targets and its effects on membrane fluidity of HEK cells, we had to use FLIM disrupt rafts a step that might have non desir able effects. Contrast enhancement promises a better localization of fluorescently labeled proteins in their native microenvironment, thereby better understanding of their activity and function. Furthermore, it will facilitate the direct visualization of lipid microdomains formation in membranes of intact cells, circumventing issues that might arise from indirect visualization considerations. We have, therefore, developed lifetime GP ( ) method to enhance contrast provided by the conventional intensi ty GP fluidity measurements. Lifetime GP ( ) method is a novel tool that can further allow us to investigate membrane microdomains organization models, such as the ones proposed in (12) . These models argued the possibilities of having discontinuous liquid ordered microdomains within continuous liquid disordered, and vise versa. Our results on HEK control cells have suggested that the former scenario may be true (Figure 5 .1).

PAGE 167

133 Figure 5 .1 Comparison of membrane microdomains organization to some of the previously proposed models. Part of the f igure was adapted form (12) . Additionally , we are interested in determining the membrane organization and range of different lipid microdomains sizes. We have attempted to quantitatively estimate the membrane microdomains sizes by virtue of 2D spatial correlation analyses and normalized autocorre lation coefficients line profiles . The way line profiles were utilized to quantify microdomains sizes has a drawback of taking only a single line of pixels in one direction (orientation) into account. In our results, line profiles were chosen such that the y traverse along the membrane to account for maximum pixels displacements possible over which normalized autocorrelation coefficients were calculated. However, this is not quite ideal, and different quantitative methods should be considered .

PAGE 168

134 CHAPTER VI CONCLUSIONS AND FUTURE DIRECTIONS Advance microscopy and data analysis techniques have been used to investigate plasma membrane microdomains and their interactions with proteins. G lobal average and local fluidity of apical and basolateral membranes were determined. NaPi2b protein in the apical membranes of enterocytes and soluble Klotho (sKL) protein in the basolateral membranes of Human Embryonic Kidney (HEK) cells were used to test the efficacy of local versus global membrane fluidity hypothesis in isol ated and intact native plasma membranes, respectively. We studied NaPi2b protein diffusion (function) in light of its microdomains of residence in GUVs made of the duodenal and jejunal region s of the small intestine . Global average membrane fluidity fail ed to explain the very slow diffusing NaPi2b species in the jejunum given that jejunal membranes were more fluid. Collectively, Fluorescence Correlation Spectroscopy (FCS) and Photon Counting Histogram (PCH) measurements were efficiently correlated with Ge neralized Polarization (GP) measurements to induce information about this unique jejunal species. We found that these species form pentamers and reside in bigger and more solid membrane lipid microdomains. We have, thus, demonstrated t he role local fluidit y of membrane microdomains played in understanding NaPi2b diffusion, and thereby potential function in contrast to global average membrane fluidity. By virtue of Fluorescence Lifetime Imaging Microscopy Forster Resonance Energy Transfer (FLIM FRET), we identified soluble Klotho (sKL) protein membrane receptor, and

PAGE 169

135 its effects on membrane fluidity. We found that sKL targets GM1 in the raft causing rafts disassociation, leading to a more fluid state of the basolateral membr anes. Lack of good enough contrast led to more cumbersome experiments, and the need to manipulate membrane fluidity in the intact native membranes study (sKL study). Out of the necessity for better direct visualization of membrane microdomains, and to ci rcumvent any issues that might arise from special provisions su ch as membrane manipulation, we developed lifetime GP ( ) method to enhance the contrast of mem brane microdomains imaging . results showed an improved contrast of fluidity measuremen ts over the conventional intensity GP measurements. Nevertheless, this method and the proposed quantitative analyses lack the precise estimation of membrane microdomains sizes, and their relative distribution across the plasma membrane. Hence, more analytical techniques to better quantify and interpret 2D spatial correlation results, and map them back on respective membrane images , will be necessary. Moreover, c alculation of the relative contributions of ordered and disordered species within a pixel (i.e., and calculations) can be performed by so lving a system of two equations corresponding to the recovered fluorescence lifetimes means from the two detection channels (see Figure 6 .1).

PAGE 170

136 Figure 6 .1 Membrane lifetime GP ( ) comparison with theoretical calculations. A simple comparison of lifet ime GP values in a representative lifetime GP image with our initial hypothetical calculations suggested that those lipid microdomains, which are designated by white arrows and having about 0.16 lifetime GP value, may contain 50% ordered and 50% disordered populations in their corresponding pixels, given the rest of the assumed parameters used in the theoretical lifetime GP calculations presented in Chapter III. Lastly, another, and yet very important future consideration is to perform m ore experiments and imaging with ~125nm/pixel instead of 250nm to enhance the contrast even better by making full use of the optical resolution provided by diffraction limited co nfocal microscopy.

PAGE 171

137 REFERENCES 1. Dick RA, Vogt VM. Membrane interaction of retroviral Gag proteins. Vol. 5, Frontiers in Microbiology. 2014. 2. Owen DM, Rentero C, Magenau A, Abu Siniyeh A, Gaus K. Quantitative imaging of membrane lipid order in cells and organisms. Nat Protoc [Internet]. 2012;7(1):24 35. Available from: http://www.ncbi.nlm.nih.gov/pubmed/22157973 3. Owen DM, Lanigan PMP, Dunsby C, Munro I, Grant D, Neil MAA, et al. Fluorescence lif etime imaging provides enhanced contrast when imaging the phase sensitive dye di 4 ANEPPDHQ in model membranes and live cells. Biophys J [Internet]. 2006;90(11):L80 2. Available from: http://www.cell.com/article/S0006349506725632/fulltext 4. Borst JW, Vis ser AJWG. Fluorescence lifetime imaging microscopy in life sciences. Meas Sci Technol [Internet]. 2010;21(10):102002. Available from: http://stacks.iop.org/0957 0233/21/i=10/a=102002?key=crossref.ec7427511a63a7d4629b41599a2815b4 5. Becker W, editor. Advan ced Time Correlated Single Photon Counting Applications. Springer International Publishing; 2015. XXIII, 624. 6. Lakowicz JR, editor. Topics in Fluorescence Spectroscopy, Volume 1: Techniques. 1st ed. New York: Plenum Press; 1991. XIV, 454. 7. Stöckl M T, Herrmann a. Detection of lipid domains in model and cell membranes by fluorescence lifetime imaging microscopy. Biochim Biophys Acta [Internet]. 2010;1798(7):1444 56. Available from: http://dx.doi.org/10.1016/j.bbamem.2009.12.015 8. Chiantia S, Ries J , Schwille P. Fluorescence correlation spectroscopy in membrane structure elucidation. Biochim Biophys Acta [Internet]. 2009;1788(1):225 33. Available from: http://www.ncbi.nlm.nih.gov/pubmed/18801332 9. Schwille P, Haustein E. Fluorescence correlation sp ectroscopy. An introduction to its concepts and applications. Biophys Textb Online [Internet]. 2001;1 33. Available from: http://scholar.google.com/scholar?hl=en&btnG=Search&q=intitle:Fluorescence+Corre lation+Spectroscopy+An+Introduction+to+its+Concepts+an d+Applications#0%5Cnht tp://scholar.google.com/scholar?hl=en&btnG=Search&q=intitle:Fluorescence+correlat ion+spectroscopy.+ 10. Laboratory for Fluorescence Dynamics (LFD) Workshop. In 2015. 11. Dalton G, An SW, Al Juboori SI, Nischan N, Yoon J, Dobrinskik h E, et al. Soluble klotho binds monosialoganglioside to regulate membrane microdomains and growth factor signaling. Proc Natl Acad Sci U S A [Internet]. 2017;114(4):752 7. Available

PAGE 172

138 from: http://www.ncbi.nlm.nih.gov/pubmed/28069944 12. Owen DM, Williamso n DJ, Magenau A, Gaus K. Sub resolution lipid domains exist in the plasma membrane and regulate protein diffusion and distribution. Nat Commun. 2012;3. 13. Elsevier/Academic P ress; 2013. 367 p. 14. Pike LJ. Rafts defined: a report on the Keystone symposium on lipid rafts and cell function. J Lipid Res [Internet]. 2006;47(7):1597 8. Available from: http://www.jlr.org/lookup/doi/10.1194/jlr.E600002 JLR200 15. Sezgin E, Leventa l I, Mayor S, Eggeling C. The mystery of membrane organization: Composition, regulation and roles of lipid rafts. Vol. 18, Nature Reviews Molecular Cell Biology. 2017. p. 361 74. 16. Bagatolli LA. To see or not to see: Lateral organization of biological membranes and fluorescence microscopy. Vol. 1758, Biochimica et Biophysica Acta Biomembranes. 2006. p. 1541 56. 17. Brown DA, Rose JK. Sorting of GPI anchored proteins to glycolipid enriched membrane subdomains during transport to the apical cell surfa ce. Cell. 1992;68(3):533 44. 18. Owen DM, Magenau A, Majumdar A, Gaus K. Imaging membrane lipid order in whole, living vertebrate organisms. Biophys J. 2010;99(1). 19. Andersen OS, Koeppe RE. Bilayer Thickness and Membrane Protein Function: An Energeti c Perspective. Annu Rev Biophys Biomol Struct [Internet]. 2007;36(1):107 30. Available from: http://www.annualreviews.org/doi/10.1146/annurev.biophys.36.040306.132643 20. Phillips R, Ursell T, Wiggins P, Sens P. Emerging roles for lipids in shaping membra ne protein function. Vol. 459, Nature. 2009. p. 379 85. 21. Singer SJ, Nicolson GL. The Fluid Mosiac Model of the Structure of Cell Membranes. Science (80 ). 1972;175:720 31. 22. Ling GN. Life at the cell and below fundamental revolution in biology. Original. New York: Pacific Press; 2001. x, 373 . 23. Karnovsky MJ, Kleinfeld AM, Hoover RL, Klausner RD. The concept of lipid domains in membranes. J Cell Biol. 1982;94(1):1 6. 24. Simons K, Van Meer G. Lipid Sortin g in Epithelial Cells. Biochemistry. 1988;27(17):6197 202. 25. van Meer G, Voelker DR, Feigenson GW. Membrane lipids: where they are and how they behave. Nat Rev Mol Cell Biol [Internet]. 2008;9(2):112 24. Available from:

PAGE 173

139 http://www.nature.com/doifinder/ 10.1038/nrm2330 26. Brown DA, London E. Structure and function of sphingolipid and cholesterol rich membrane rafts. J Biol Chem. 2000;275(23):17221 4. 27. Simons K, Ikonen E. Functional rafts in cell membranes. Vol. 387, Nature. 1997. p. 569 72. 28. Lingwood D, Simons K. Lipid rafts as a membrane organizing principle. Vol. 327, Science. 2010. p. 46 50. 29. Sharpe HJ, Stevens TJ, Munro S. A Comprehensive Comparison of Transmembrane Domains Reveals Organelle Specific Properties. Cell. 2010;142(1):158 69. 30. Kaiser H J, Orlowski A, Rog T, Nyholm TKM, Chai W, Feizi T, et al. Lateral sorting in model membranes by cholesterol mediated hydrophobic matching. Proc Natl Acad Sci [Internet]. 2011;108(40):16628 33. Available from: http://www.pnas.org/cgi/doi/ 10.1073/pnas.1103742108 31. Maxfield FR, Tabas I. Role of cholesterol and lipid organization in disease. Vol. 438, Nature. 2005. p. 612 21. 32. Lingwood D, Binnington B, Róg T, Vattulainen I, Grzybek M, Coskun Ü, et al. Cholesterol modulates glycolipid conformation and receptor activity. Nat Chem Biol. 2011;7(5):260 2. 33. Laganowsky A, Reading E, Allison TM, Ulmschneider MB, Degiacomi MT, Baldwin AJ, et al. Membrane proteins bind lipids selectively to modulate their structure and function. Nature. 201 4;510(7503):172 5. 34. Hooper NM, Turner AJ. Ectoenzymes of the kidney microvillar membrane. Differential solubilization by detergents can predict a glycosyl phosphatidylinositol membrane anchor. Biochem J. 1988;250(3):865 9. 35. Dinic J, Riehl A, Adle r J, Parmryd I. The T cell receptor resides in ordered plasma membrane nanodomains that aggregate upon patching of the receptor. Sci Rep. 2015;5. 36. Gupta N. Visualizing Lipid Raft Dynamics and Early Signaling Events during Antigen Receptor mediated B L ymphocyte Activation. Mol Biol Cell [Internet]. 2003;14(2):432 44. Available from: http://www.molbiolcell.org/cgi/doi/10.1091/mbc.02 05 0078 37. Beck García K, Beck García E, Bohler S, Zorzin C, Sezgin E, Levental I, et al. Nanoclusters of the resting T c ell antigen receptor (TCR) localize to non raft domains. Biochim Biophys Acta Mol Cell Res. 2015;1853(4):802 9. 38. Sproul TW, Malapati S, Kim J, Pierce SK. Cutting edge: B cell antigen receptor signaling occurs outside lipid rafts in immature B cells. J Immunol. 2000;165:6020 3.

PAGE 174

140 39. phase like domains supports emergent signaling function in b lymphocyte plasma membranes. Elife. 2017;6. 40. Eckert GP, Müller WE. Presenilin 1 modifies lipid raft composition of neuronal membranes. Biochem Biophys Res Commun. 2009;382(4):673 7. 41. Beel AJ, Sakakura M, Barrett PJ, Sanders CR. Direct binding of cholesterol to the amyloid precursor protein: An important interaction in lipid Alz relationships? Vol. 1801, Biochimica et Biophysica Acta Molecular and Cell Biology of Lipids. 2010. p. 975 82. 42. Molnár E, Swamy M, Holzer M, Beck García K, Worch R, Thiele C, et al. Cholesterol and sphingomyelin drive ligand indepen dent T cell antigen receptor nanoclustering. J Biol Chem. 2012;287(51):42664 74. 43. Fantini J, Yahi N. Molecular basis for the glycosphingolipid binding specificity of alpha synuclein: Key role of tyrosine 39 in membrane insertion. J Mol Biol [Internet] . 2011;408(4):654 9. Available from: http://dx.doi.org/10.1016/j.jmb.2011.03.009 44. Kyogashima M, Krivan HC, Schweinle JE, Ginsburg V, Holt GD. Glycosphingolipid binding specificity of the mannose binding protein from human sera. Arch Biochem Biophys. 19 90;283(1):217 22. 45. Yamaji A, Sekizawa Y, Emoto K, Sakuraba H, Inoue K, Kobayashi H, et al. Lysenin, a novel sphingomyelin specific binding protein. J Biol Chem. 1998;273(9):5300 6. 46. Contreras FX, Ernst AM, Haberkant P, Björkholm P, Lindahl E, Gönen B, et al. domain. Nature. 2012;481(7382):525 9. 47. Coskun U, Grzybek M, Drechsel D, Simons K. Regulati on of human EGF receptor by lipids. Proc Natl Acad Sci [Internet]. 2011;108(22):9044 8. Available from: http://www.pnas.org/cgi/doi/10.1073/pnas.1105666108 48. Mikuni S, Yamamoto J, Horio T, Kinjo M. Negative correlation between the diffusion coefficient and transcriptional activity of the glucocorticoid receptor. Int J Mol Sci. 2017;18(9). 49. Koopman WJH, Hink MA, Verkaart S, Visch HJ, Smeitink JAM, Willems PHGM. Partial complex I inhibition decreases mitochondrial motility and increases matrix protein diffusion as revealed by fluorescence correlation spectroscopy. Biochim Biophys Acta Bioenerg. 2007;1767(7):940 7. 50. Inoue M, Digman MA, Cheng M, Breusegem SY, Halaihel N, Sorribas V, et al. Partitioning of NaPi cotransporter in cholesterol , sphing omyelin , and glycosphingolipid enriched membrane domains modulates NaPi protein diffusion, clustering, and activity. J Biol Chem. 2004;279(47):49160 71. 51. Yanagawa M, Hiroshima M, Togashi Y, Yamashita T, Shichida Y, Murata M, et al.

PAGE 175

141 Single molecule di ffusion based estimation of GPCR activity. doi.org [Internet]. 2017;205161. Available from: https://www.biorxiv.org/content/early/2017/10/18/205161 52. Eggeling C, Ringemann C, Medda R, Schwarzmann G, Sandhoff K, Polyakova S, et al. Direct observation of the nanoscale dynamics of membrane lipids in a living cell. Nature. 2009;457(7233):1159 62. 53. Brewer J, Bernardino de la Serna J, Wagner K, Bagatolli L a. Multiphoton excitation fluorescence microscopy in planar membrane systems. Biochim Biophys Acta [ Internet]. 2010;1798(7):1301 8. Available from: http://www.ncbi.nlm.nih.gov/pubmed/20226161 54. Kahya N, Scherfeld D, Bacia K, Poolman B, Schwille P. Probing lipid mobility of raft exhibiting model membranes by fluorescence correlation spectroscopy. J Bio l Chem. 2003;278(30):28109 15. 55. Demel RA, De Kruyff B. The function of sterols in membranes. Biochim Biophys Acta [Internet]. 1976;457(2):109 32. Available from: http://www.ncbi.nlm.nih.gov/pubmed/184844 56. Digman MA, Caiolfa VR, Zamai M, Gratton E. The phasor approach to fluorescence lifetime imaging analysis. Biophys J [Internet]. 2008;94(2):L14 6. Available from: http://www.ncbi.nlm.nih.gov/pubmed/17981902 57. Zhao J, Wu J, Heberle FA, Mills TT, Klawitter P, Huang G, et al. Phase studies of model biomembranes: Complex behavior of DSPC/DOPC/Cholesterol. Biochim Biophys Acta Biomembr. 2007;1768(11):2764 76. 58. Mertz J. Introduction to Optical Microscopy. Roberts and Company Publishers; 2010. 59. Loura LMS, Prieto M. FRET in membrane biophysic s: An overview. Front Physiol. 2011;2 NOV. 60. Buboltz JT. Steady state probe partitioning fluorescence resonance energy transfer: A simple and robust tool for the study of membrane phase behavior. Phys Rev E Stat Nonlinear, Soft Matter Phys. 2007;76(2 ). 61. Sensitive Probes for Superresolution Stimulated Emission Depletion Microscopy. Biophys J. 2017;113(6):1321 30. 62. Magde D, Elson E, Webb WW. Thermodynamic Fluctuat ions in a Reacting System Measurement by Fluorescence Correlation Spectroscopy. Phys Rev Lett [Internet]. 1972;29(11):705 8. Available from: http://link.aps.org/doi/10.1103/PhysRevLett.29.705%5Cnhttp://prl.aps.org/abstract/PR L/v29/i11/p705_1%5Cnhttp://prl. aps.org/pdf/PRL/v29/i11/p705_1 63. Reits EAJ, Neefjes JJ. From fixed to FRAP: Measuring protein mobility and activity in living cells. Vol. 3, Nature Cell Biology. 2001.

PAGE 176

142 64. Axelrod D, Koppel DE, Schlessinger J, Elson E, Webb WW. Mobility measurement by analysis of fluorescence photobleaching recovery kinetics. Biophys J. 1976;16(9):1055 69. 65. Bacia K, Scherfeld D, Kahya N, Schwille P. Fluorescence correlation spectroscopy relates rafts in model and native membranes. Biophys J. 2004;87(2):1034 43. 6 6. Bacia K, Schwille P. A dynamic view of cellular processes by in vivo fluorescence auto and cross correlation spectroscopy. Vol. 29, Methods. 2003. p. 74 85. 67. Kahya N, Scherfeld D, Bacia K, Schwille P. Lipid domain formation and dynamics in giant unilamellar vesicles explored by fluorescence correlation spectroscopy. Vol. 147, Journal of Structural Biology. 2004. p. 77 89. 68. García Sáez AJ, Schwille P. Fluorescence correlation spectroscopy for the study of membrane dynamics and protein/lipid in teractions. Methods. 2008;46(2):116 22. 69. Wit G De, Danial JSH, Kukura P, Wallace MI. Dynamic label free imaging of lipid nanodomains. Proc Natl Acad Sci [Internet]. 2015;2015(24):3 7. Available from: http://m.pnas.org/content/112/40/12299.abstract%5Cn http://www.pnas.org/lookup/doi /10.1073/pnas.1508483112 70. Scherfeld D, Kahya N, Schwille P. Lipid dynamics and domain formation in model membranes composed of ternary mixtures of unsaturated and saturated phosphatidylcholines and cholesterol. Biophys J [ Internet]. 2003;85(6):3758 68. Available from: http://dx.doi.org/10.1016/S0006 3495(03)74791 2 71. Chen Y, Müller JD, So PTC, Gratton E. The photon counting histogram in fluorescence fluctuation spectroscopy. Biophys J. 1999;77(1):553 67. 72. Meseth U, Wohland T, Rigler R, Vogel H. Resolution of fluorescence correlation measurements. Biophys J. 1999;76(3):1619 31. 73. Youker RT, Teng H. Measuring protein dynamics in live cells: protocols and practical considerations for fluorescence fluctuation microsc opy. J Biomed Opt [Internet]. 2014;19(9):90801. Available from: http://biomedicaloptics.spiedigitallibrary.org/article.aspx?doi=10.1117/1.JBO.19.9.09 0801 74. Thompson NL. Fluorescence Correlation Spectroscopy. In: Topics in Fluorescence Spectroscopy. 1991 . p. 337 378. 75. Margeat E, Poujol N, Boulahtouf A, Chen Y, Müller JD, Gratton E, et al. The human 1 coactivator molecule with an affinity dictated by agonist structure. J Mol Biol. 2001;306(3):433 42. 76. Van Rompaey E, Chen Y, Müller JD, Gratton E, Van Craenenbroeck E, Engelborghs Y, et al. Fluorescence fluctuation analysis for the study of interactions between oligonucleotides and polycationic polymers. Biol Chem. 2001;382(3):379 86.

PAGE 177

143 77. Chen Y, Müller JD, Ruan Q, Gratton E. Molecular brightness characterization of EGFP in vivo by fluorescence fluctuation spectroscopy. Biophys J [Internet]. 2002;82(1 Pt 1):133 44. Available from: http://www.ncbi.nlm.nih.gov/pubmed/11751302%5Cnhttp://www.pubmedcentral.nih . gov/articlerender.fcgi?artid=PMC1302455 78. Biber J, Stieger B, Haase W, Murer H. A high yield preparation for rat kidney brush border membranes Different behaviour of lysosomal markers. BBA Biomembr. 1981;647(2):169 76. 79. Zajicek HK, Wang H, Putta parthi K, Halaihel N, Markovich D, Shayman J, et al. Glycosphingolipids modulate renal phosphate transport in potassium deficiency. Kidney Int. 2001;60(2):694 704. 80. Levi M, Molitoris BA, Burke TJ, Schrier RW, Simon FR. Effects of vitamin D induced chr onic hypercalcemia on rat renal cortical plasma membranes and mitochondria. Am J Physiol [Internet]. 1987;252(2 Pt 2):F267 75. Available from: http://www.ncbi.nlm.nih.gov/htbin post/Entrez/query?db=m&form=6&dopt=r&uid=3812741 81. Angelova MI, Dimitrov DS. Liposome Electro formation. Faraday Discuss Chem SOC. 1986;81:303 11. 82. Bagatolli L a, Gratton E. Two photon fluorescence microscopy observation of shape changes at the phase transition in phospholipid giant unilamellar vesicles. Biophys J [Internet]. 1999;77(4):2090 101. Available from: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1300490&tool=pmcentrez &rendertype=abstract 83. Bagatolli LA, Gratton E. Two Photon Fluorescence Microscopy of Coexisting Lipid Domains in Giant Unilamellar Ves icles of Binary Phospholipid Mixtures. Biophys J [Internet]. 2000;78(1):290 305. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0006349500765921 84. Dietrich C, Bagatolli LA, Volovyk ZN, Thompson NL, Levi M, Jacobson K, et al. Lipid Rafts Re constituted in Model Membranes. Biophys J [Internet]. 2001;80(3):1417 28. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0006349501761140 85. Müller JD, Chen Y, Gratton E. Fluorescence correlation spectroscopy. Methods Enzymol. 2003;361:69 9 2. 86. Chen Y, Müller JD, So PTC, Gratton E. The Photon Counting Histogram in Fluorescence Fluctuation Spectroscopy. Biophys J [Internet]. 1999;77(1):553 67. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0006349599769122 87. Müller JD, Chen Y, Gratton E. Resolving heterogeneity on the single molecular level with the photon counting histogram. Biophys J [Internet]. 2000;78(1):474 86. Available from:

PAGE 178

144 http://www.ncbi.nlm.nih.gov/pubmed/10620311%5Cnhttp://www.pubmedcentral.nih. gov /articlerender.fcgi?artid=PMC1300655 88. IMMUNOGLOBULINS STRUCTURE AND FUNCTION [Internet]. [cited 2018 Apr 22]. Available from: http://www.microbiologybook.org/mobile/m.immuno 4.htm 89. o fluorescent probes and labeling technologies [Internet]. 11th ed. Carlsbad, CA: Live Technologies Corporation; 2010. xiv, 1160 . Available from: http://probes.invitrogen.com/handbook/ 90. Giral H, Cranston D, Lanzano L, Caldas Y, Sutherland E, Rachelson J, et al. NHE3 Regulatory Factor 1 (NHERF1) modulates intestinal sodium dependent phosphate transporter (NaPi 2b) expression in apical microvilli. J Biol Chem. 2012;287(42):35047 56. 91. Bushberg JT. The essential physics of medical imaging. 3rd ed. Phi ladelphia: Wolters Kluwer Health/Lippincott Williams & Wilkins; 2012. xii, 1030 . 92. Rowley MI, Barber PR, Coolen ACC, Vojnovic B. Bayesian analysis of fluorescence lifetime imaging data. In: Progress in Biomedical Optics and Imaging Proceedings of SP IE. 2011. 93. Shahbaba B. Biostatistics with R: An introduction to statistics through biological data. Biostatistics with R: An Introduction to Statistics Through Biological Data. 2012. 1 352 p. 94. Köllner M, Wolfrum J. How many photons are necessary for fluorescence lifetime measurements? Chem Phys Lett. 1992;200(1 2):199 204. 95. Marcu L, French PMW, Elson DS, editors. Fluorescence Lifetime Spectroscopy and Imaging: Principles and Applications in Biomedical Diagnostics. CRC Press; 2015. 96. Stuar 97. Johnson NL, Elandt Johnson RC. Survival Models and Data Analysis. John Wiley & Sons NY; 1980. 98. Al Juboori SI, Lei T, Hogg Cornejo V, Sutherland E, Levi M, Dobrinskikh E. NaPi2b d iffusion and activity could be regulated by membrane fluidity in the microvilli of the small intestine. FASEB J [Internet]. 2016 Apr 1;30(1 Supplement):861.3 861.3. Available from: http://www.fasebj.org/content/30/1_Supplement/861.3.abstract 99. Giral H, Caldas Y, Sutherland E, Wilson P, Breusegem S, Barry N, et al. Regulation of rat intestinal Na dependent phosphate transporters by dietary phosphate. Am J Physiol Renal Physiol. 2009;297(5):F1466 75. 100. Kokubo H, Helms JB, Ohno Iwashita Y, Shimada Y, Horikoshi Y, Yamaguchi H. Ultrastructural localization of flotillin 1 to cholesterol rich membrane microdomains, rafts, in rat brain tissue. Brain Res. 2003;965(1 2):83 90.

PAGE 179

145 101. Volonte D, Galbiati F, Li S, Nishiyama K, Okamoto T, Lisanti MP. Flotillins/ cavatellins are differentially expressed in cells and tissues and form a hetero oligomeric complex with caveolins in vivo. Characterization and epitope mapping of a novel flotillin 1 monoclonal antibody probe. J Biol Chem [Internet]. 1999 Apr 30 [cited 201 8 Mar 24];274(18):12702 9. Available from: http://www.ncbi.nlm.nih.gov/pubmed/10212252 102. Owen DM, Williamson DJ, Magenau A, Gaus K. Sub resolution lipid domains exist in the plasma membrane and regulate protein diffusion and distribution. Nat Commun [I nternet]. 2012;3:1256. Available from: http://www.ncbi.nlm.nih.gov/pubmed/23212385%5Cnhttp://www.nature.com/ncomms /journal/v3/n12/pdf/ncomms2273.pdf 103. Gonzalez RC, Woods RE, Eddins SL. Digital Image processing using MATLAB. 2nd ed. S.I.: Gatesmark Pub. ; 2009. xviii, 826 .

PAGE 180

146 APPENDIX % Lifetime GP % Noise Metrics LifetimeGPMean = zeros([28 18]); % preallocate 2 D array LifetimeGPMean (:,:) = 0.102; % homogenious GP area, another value can be entered here! b=0.02; % standard deviation LifetimeGPNoise = LifetimeGPMean + b*randn(28, 18); imagesc(LifetimeGPNoise) % Four random generators R1 = binornd(1,0.5,28,18); for i=1:28; for j=1:18; if R1(i,j)==0 R1(i,j)=R1(i,j) 1; end end end R2 = binornd(1,0.5,28 ,18); for i=1:28; for j=1:18; if R2(i,j)==0 R2(i,j)=R2(i,j) 1; end end end R3 = binornd(1,0.5,28,18); for i=1:28; for j=1:18; if R3(i,j)==0 R3(i,j)=R3(i,j) 1; end end end R4 = binornd( 1,0.5,28,18); for i=1:28; for j=1:18; if R4(i,j)==0 R4(i,j)=R4(i,j) 1; end end end GP1=(TauOrdered TauDisordered)./(TauOrdered+TauDisordered); %Lifetime GP Mean A=TauOrdered+(R1.*Vordered); B=TauDisordered+(R3.*Vdisorder ed); GP2=(A B)./(A+B); %Lifetime GP One Standard Deviation C=TauOrdered+(R2.*Vordered2); D=TauDisordered+(R4.*Vdisordered2); GP3=(C D)./(C+D); %Lifetime GP Two Standard Deviation

PAGE 181

147 imagesc(GP1);figure;imagesc(GP2);figure;imagesc(GP3) %LEFT for i=1:28; for j=2:18; if isnan(Mu(i,j)) continue %Pass control to next iteration of for or while loop elseif isnan(Mu(i,j 1)) continue %Pass control to next iteration of for or while loop end [hL(i,j),LEFT(i ,j)]=kstest2(normrnd(Mu(i,j),Sigma(i,j),[1 100000]),normrnd(Mu(i,j 1),Sigma(i,j 1),[1 100000])); end end %RIGHT for i=1:28; for j=1:17; if isnan(Mu(i,j)) continue %Pass control to next iteration of for or while loop elseif isnan(Mu(i,j+1)) continue %Pass control to next iteration of for or while loop end [hR(i,j),RIGHT(i,j)]=kstest2(normrnd(Mu(i,j),Sigma(i,j),[1 100000]),normrnd(Mu(i,j+1),Sigma(i,j+1),[1 100000])); end end %DOWN for i=1:27; for j=1:18; if isnan(Mu(i,j)) continue %Pass control to next iteration of for or while loop elseif isnan(Mu(i+1,j)) continue %Pass control to next iteration of for or while loop end [hD(i, j),DOWN(i,j)]=kstest2(normrnd(Mu(i,j),Sigma(i,j),[1 100000]),normrnd(Mu(i+1,j),Sigma(i+1,j),[1 100000])); end end %UP for i=2:28; for j=1:18; if isnan(Mu(i,j)) continue %Pass control to next iteration of for or while loop elseif isnan(Mu(i 1,j)) continue %Pass control to next iteration of for or while loop end [hU(i,j),UP(i,j)]=kstest2(normrnd(Mu(i,j),Sigma(i,j),[1 100000]),normrnd(Mu(i 1,j),Sigma(i 1,j),[1 100000]));

PAGE 182

148 end end % 2D Spatial Correlation Intensity thresholding IntensityGP=IntensityGPMatrix; for i=1:28; for j=1:18; if IntensityGP(i,j)<0 IntensityGP(i,j)=0; %converting anything less than 0 to 0s end end end for i=1:28; for j=1:18; if IntensityGP(i,j)>0.5 IntensityGP(i,j)=0; %converting anything greater than 0.5 to 0s end end end %2D Spatial Correlation Normalization GP difference from the mean NotaNumber=IntensityGP; %to make sure not to alternate L ifetimeGP values for i=1:28; for j=1:18; if NotaNumber(i,j)==0 NotaNumber(i,j)=NaN; %converting 0s to NaN for mean calculation purpose end end end M=0.087963906; %mean of lifetime GP calculated on a different Excel sheet difference=NotaNumber M; %difference of lifetime GP for each pixel from the mean for i=1:28; for j=1:18; if isnan(difference(i,j)) difference(i,j)=0; %converting NaNs back to 0s before xcorr2 operation end end end ID=xcorr2(difference); %2D spatial correlation of the difference PointMaxI=max(max(ID)); %finding the first max in the autocorrelation function ID=ID./PointMaxI; %normalization to the first max in the autocorrelation function % 2D Spatial Correlation Lifetime LifetimeGP=LifetimeGPMatrix; for i=1:28; for j=1:18; if LifetimeGP(i,j)==1 LifetimeGP(i,j)=0; %converting 1s to 0s end end end

PAGE 183

149 for i=1:28; for j=1:18; if LifetimeGP(i,j)== 1 LifetimeGP(i,j)=0; %converting 1s to 0s end end end %2D Spatial Correlation Normalization GP difference from the mean NotaNumber=LifetimeGP; %to make sure not to alternate LifetimeGP values for i=1:2 8; for j=1:18; if NotaNumber(i,j)==0 NotaNumber(i,j)=NaN; %converting 0s to NaN for mean calculation purpose end end end M= 0.055277158; %mean of lifetime GP calculated on a different Excel sheet difference= NotaNumber M; %difference of lifetime GP for each pixel from the mean for i=1:28; for j=1:18; if isnan(difference(i,j)) difference(i,j)=0; %converting NaNs back to 0s before xcorr2 operation end end end LD=xcorr2( difference); %2D spatial correlation of the difference PointMaxTau=max(max(LD)); %finding the first max in the autocorrelation function LD=LD./PointMaxTau; %normalization to the first max in the autocorrelation function