High altitude, natural selection, and birth weight

Material Information

High altitude, natural selection, and birth weight is small good or bad?
Wilson, Megan
Publication Date:
Physical Description:
xi, 218 leaves : ; 28 cm


Subjects / Keywords:
Altitude, Influence of ( lcsh )
Birth weight ( lcsh )
Natural selection ( lcsh )
Altitude, Influence of ( fast )
Birth weight ( fast )
Natural selection ( fast )
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )


Includes bibliographical references (leaves 211-218).
General Note:
Department of Health and Behavioral Sciences
Statement of Responsibility:
by Megan Jeannine Wilson.

Record Information

Source Institution:
|University of Colorado Denver
Holding Location:
|Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
318900973 ( OCLC )
LD1193.L566 2008d W54 ( lcc )

Full Text
Megan Jeannine Wilson
B.A., University of Colorado-Boulder, 2000
M.A., University of Colorado Denver, 2005
A thesis submitted to the
University of Colorado Denver
in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
Health and Behavioral Sciences

This thesis for the Doctor of Philosophy
degree by
Megan Jeannine Wilson
has been approved
David P. Tracer

Jill Norris

Wilson, Megan J. (Ph.D., Health and Behavioral Sciences)
High Altitude, Natural Selection, & Birth Weight: Is small good or bad?
Thesis directed by Professor Lorna G. Moore
Fetal growth is slowed at high altitude (HA) and preeclampsia more
common. Populations of multi-generational residence at HA show a
decrease in fetal growth at high altitude, but the decrease is less than that
seen in relative newcomers to HA. There is debate as to whether decreased
birth weight at HA has been selected for or against. Previously, direct tests of
the genetic adaptations in pregnancy to HA have been unavailable. We
explored the relationship between genes that show evidence of natural
selection and their effect on phenotype by 1) identifying hypoxia-related gene
regions that showed evidence of natural selection through the analysis of
genome-wide SNP microarray data, and 2) searching for association
between genotypes at candidate loci, circulating gene product (protein)
levels, uterine artery (UA) blood flow and other pregnancy phenotypes at HA.
In 50 multi-generational HA Andeans compared to low-altitude control
populations (Amerindians and Han Chinese), five hypoxia-related gene
regions showed differential expression of alleles in Andeans. In 55 multi-
generational HA pregnant Andeans, CDH1 gene region SNP genotypes were
associated with UA blood flow. Soluble CDH (sCDH1) correlated with CDH1
genotype and also with UA blood flow. Additionally, AMPKal genotype was
strongly associated with gestational age at birth and ARNT2 was associated
with birth weight. In all cases, alleles more frequently found in Andeans were
positively associated with UA blood flow, gestational age at birth, or birth
weight, suggesting evolution at HA may have selected for increased birth
weight, as opposed to small-but-healthy babies, in adapted populations.
This abstract accurately represents the content of the candidates thesis. I
recommend its publication.
Lorna G. Moore

Because 1) Its a sign of mediocrity when you demonstrate gratitude
with moderation (Roberto Benigni) and 2) a scientific work is never the work
of one person, this is no time for mediocrity.
I would like to first thank my advisor and mentor Dr. Lorna Moore for
her supervision, advice, and guidance. The opportunity to conduct this study
was available with the assistance of grants from the National Institutes of
Health (HL079647; HL060131; TW001188) and the National Science
Foundation (Graduate Research Fellowship).
I wish to thank my esteemed committee for their support and valuable
guidance. Moreover, David Tracer deserves individual thanks for sharing his
knowledge, guidance and perspective in things extra- and intracurricular.
We are forever indebted to our colleagues in Bolivia, most especially
Enrique Vargas for his patience, persistence, and fortitude. Additionally, it
was a pleasure to collaborate with the amazing people at PSU, particularly
Mark Shriver and Abby Bigham.
The support and fostering of ideas of the folks at the Altitude
Research Center, including Rob Roach and Ben Honigman especially was
invaluable. I would like to thank Abby Fitch, Jenny Hageman, Ruth Johnson
and Barbara Lommen for their indispensable help dealing with travel funds,
administration and bureaucratic matters during my journey. In reality, none
of us would get anything done without the imperative support of these
amazing people.
A shout out to my peeps in Cohort 10 (with honorary member Jennifer
Lloyd) for creating a stimulating and fun environment, sometimes rare in grad
school. Additionally, the other end of the brain tube, Colleen Julian,
deserves special gratitude for her provision of critical scientific discussion,
critically-important fun, and politically- critical protestation for the rights of
graduate students and uterine arteries. Heres to la futura!
I am so grateful for the eternal and enthusiastic support of my entire
extended family, with particular loving thanks to my grandparents and, of
course, to my mom for her love and patronage (both financial and emotional).
I appreciate your support and encouragement to do the best I can in all the
endeavors of my life.

To Ken, who enabled the completion of this tome1 through his
amazing love, support, well-timed distractions, and comprehension of my
hypnagogic ramblings, te§ekkur ederim so much. A person who can make
me laugh in the midst of this work is a gift I am so lucky to have. You are the
best combination of challenge and support for my mind, body, and soul. With
a partner who sustains you like that, who needs oxygen (see following 225
1 Besides referring to a big, heavy book; often an overly large and dense piece of writing",
tome can also mean good, excellent, cool, or a skinny British emo who is terrible at typing
according to, therefore I think it is a particularly appropriate descriptor
of this work. Except for the British part.

1. Introduction and Specific Aims...............................1
2 Background...................................................8
2.1 Challenges of Pregnancy......................................8
2.1.1 Mechanical Changes...........................................9
2.1.2 Cardiovascular Changes......................................11
2.1.3 Hormonal Mechanisms.........................................14
2.1.4 HIF Pathway.................................................16
2.1.5 Effect of Physiologic Responses on Maternal and Fetal Health.19
2.2 lUGRorSGA?................................................. 23
2.3 SGA.........................................................25
2.4 Causes of IUGR..............................................32
2.4.1 Malnutrition................................................32
2.4.2 Toxic insult................................................33
2.4.3 Socioeconomic Factors.......................................36
2.4.4 Hypoxia.....................................................39
2.5 High Altitude...............................................40
2.6 Basic Population Differences at High Altitude...............44
2.7 Hypoxic IUGR: Adaptive of Detrimental? The Debate...........49
2.7 1 Argument for lUGRs Harmful Effect at High Altitude.........49
2.7.2 Argument for lUGRs Beneficial Effect at High Altitude......51
2.8 Epidemiologic Evidence Regarding Mortality and Morbidity at
High Altitude...............................................52
2.8.1 Rocky Mountains.............................................52
2.8.2 Andes ......................................................55
2.8.3 Himalayas...................................................62
2.9 Role of the Uterine Artery..................................64
2.9.1 Pregnancy Physiology at Low Altitude........................66
2.9.2 Physiology at High Altitude.................................67

2.10 Genomic Approaches to Studying Natural Selection............72
2.10.1 Genes Provide Evidence of Evolution.........................75
3. Research Design and Methods.................................77
3.1 General Approach............................................77
3.2 Aim 1.......................................................79
3.2.1 Rationale...................................................80
3.2.2 Measurements................................................81
3.2.3 Confirming Self-assessed Ancestry...........................82
3.2.4 Genomic Scan................................................82
3.2.5 Measures of Population Genomic Heterogeneity................83 Wright Statistic............................................83 Locus-Specific Branch Lengths...............................85 Natural Log of the Ratio of Heterozygosities................87 Tajimas D..................................................88
3.3 Aim 2.......................................................91
3.3.1 Rationale...................................................92
3.3.2 Measurements................................................92 Genotyping..................................................93 Maternal and Fetal Phenotype Measurements...................95 Vascular Measurements.....................................96 Hematologic Measurements..................................98 Fetal Measurements........................................99 Linking Genotype With Phenotype.............................99 Search for Association Between Phenotype and Genotype.....102
3.4 Aim 3....................................................104
3.4.2 Rationale..................................................105
3.4.3 Measurements............................................. 105
4. Results....................................................108
4.1 Aim 1 Results..............................................108
4.2 Aim 2 Results..............................................113
4.2.1 Genotyping.................................................113
4.2.2 Vascular Measurements......................................118
4.2.3 Association of Genotype With Phenotype.....................121
4.2.4 Fetal/lnfant Characteristics...............................131
4.2.5 Gene Product Assays........................................135 Soluble CDH1...............................................135 Nitric Oxide Metabolites (NOx).............................140
5. Discussion.................................................141
5.1 Maternal Blood Flow........................................141
5.2 Association of Uterine Artery Measurements with Genotypes.142
5.2.1 CDH1.......................................................142

< CD
5.2.2 iNOS.......................................................149
5.3 Fetal Growth...............................................151
5.4 Fetal Brain Sparing........................................154
6. Summary & Conclusions......................................157
Schema of Gene Regions With SNPs.......................160
Description of Candidate Gene Function.....................165
References ........................................................186

1.1 Study Aims.....................................................4
2.1 Normal and Abnormal Placentation .............................16
2.2 Total Atmospheric Pressure and the Partial Pressure of
Oxygen Decrease with Increasing Altitude.....................43
2.3 Figure 2 (p. 211) from Beall (1981)...........................60
3.1 Schema Representing Branch Lengths Between Three
Different Populations........................................86
4.1 Graphic Representation of the False Discovery Rate...........128
4.2 Graphs showing significant associations between genotypes
and haplotypes in the CDH1 gene region with uterine artery
4.3 Graphs Showing Significant Associations Between
Genotype and Haplotype in the iNOS Gene Region with
Week 20 Uterine Artery Diameter.............................130
4.4 The Relationship Between ARNT2 and rs4778835 and
Baby Birth Weight Unadjusted................................132
4.5 ARNT2 rs4778835 Genotype Along with Maternal Height,
Fatness, Parity, Baby Sex, and Gestational Age at Birth
Predict Baby Birth Weight...................................133
4.6 Graphs Showing Significant Associations Between
Genotype (a, b) and Haplotype (c) in the AMPKal Gene
Region with Gestational Age at Delivery.....................134
4.7 Longitudinal Analysis Shows Circulating sCDH1 Increases
Postpartum Compared with Both Time Points in Pregnancy.....136
4.8 The Presence of the Allele Found More Often in Andeans
in rs8061932 is Positively Correlated with Circulating
sCDH1 at Week 36 and Tended to be Positively Associated
with CDH1 rs11864025....................................... 137
4.9 sCDH in Week 36 is Also Positively Correlated with UA Flow
in Week 20..................................................137

4.10 Three-dimensional Graph of Association Between
rs8061932 Genotype, Circulating Levels of sCDH1, and
UA Blood flow in Week 20................................138
4.11 CDH1 rs8061932 Genotype and sCDH1 at Week 36
Predict UA Flow at Week 20..............................139
A.1 AMPKal Chromosome Figure and Location Information
From UCSC Genome Browser on Human Mar.2006
A.2 ARNT2 Chromosome Figure and Location Information
From UCSC Genome Browser on Human Mar.2006
A.3 ATP1a1 Chromosome Figure and Location Information
From UCSC Genome Browser on Human Mar.2006
A.4 CDH1 Chromosome Figure and Location Information
From UCSC Genome Browser on Human Mar.2006
A. 5 iNOS Chromosome Figure and Location Information
From UCSC Genome Browser on Human Mar.2006
B. 1 AMPKal Information from GeneNote Version 2.1,
December 10, 2007......................................167
B.2 ATP1a1 Information from GeneNote Version 2.1,
December 10, 2007......................................171
B.3 ARNT2 Information from GeneNote Version 2.1,
December 10, 2007......................................175
B.4 CDH1 Information from GeneNote Version 2.1,
December 10, 2007......................................178
B.5 iNOS Information from GeneNote Version 2.1,
December 10, 2007......................................183

2.1 Review of the literature- adapted from Moore 2001................61
2.2 Previously reported comparisons of uterine artery flow
characteristics and birth weights at high altitude..............68
3.1 Study samples......................................................79
3.2 Maternal characteristics..........................................106
3.3 Birth and Fetal Characteristics...................................107
4.1 LSBL results identify four HIF-related gene regions in which
Andeans show different allele frequencies compared with low
altitude control populations................................110
4.2 LnRFI results identify ten HIF-related gene regions in which
Andeans show less heterozygosity compared with low altitude
control populations. Note: In LnRH measures, the more negative
the value, the less heterozygous the Andeans tend to be.....111
4.3 Tajimas D results identify 13 HIF-related gene regions in which
Andeans show less heterozygosity per segregating site
compared with low altitude control populations..............112
4.4 SNPs selected in or near gene regions.............................115
4.5 Primers for PCR...................................................116
4.6 Prevalence of haplotypes and genotypes for each gene region
tested ..........................................................117
4.7 Longitudinal changes in blood flow and hemodynamics...............120
4.8 Associations of genotypes with fetal and infant characteristics..125
4.9 Associations of genotypes with maternal characteristics...........126
4.10 Consideration of multiple comparisons through the False
Discovery Rate...................................................127

1. Introduction and Specific Aims
Pregnancy is a crucial time in the human life cycle, with demands
being made on both the mother and fetus which challenge the limits of
human physiology, and profoundly affect reproductive success (Ellison,
2003; Rockwell et al., 2003). Birth weight (or other measures of fetal growth)
is an important pregnancy outcome that can be indicative of maternal as well
as fetal complications of pregnancy. Moreover, birth weight and fetal growth
have been shown to be subject to genetic control (e.g. Ong and Dunger,
2004; Little and Sing, 1987; Baker et al., 1993) and to influence morbidity
and mortality during infancy as well as later-in-life (see Background and
Significance for further explanation). Because birth weight is both heritable
and linked with differential fitness, it is possible that fetal growth is subject to
the forces of natural selection.
High altitude (>2500 m or 8000 ft) is an environmental stressor that
slows fetal growth, decreasing birth weight with increasing altitude in every
population studied to date. Thus, babies born at high altitude are more often
of low birth weight (<2500 gm or 5.5 lbs), intrauterine growth restricted
(IUGR) or small for gestational age (SGA, see Background and Significance

for further definition of terms) (Jensen and Moore, 1997). The effect of
altitude on birth weight is among the most powerful of all known
determinants, greater than maternal parity, age or amount of prenatal care
and comparable to smoking in terms of its influence on giving birth to a baby
born small for gestational age. About 140 million persons live at high altitude
(Niermeyer et al., 2001), making them the largest single group at risk of
reduction of fetal growth restriction (Krampl, 2002).
Yet, there is debate as to the consequences of altitude-associated
intrauterine growth restriction (IUGR). Some (e.g. Moore et al., 2001) posit
that restricted fetal growth due to hypoxia is associated with increased risk
for morbidity and mortality, just as is IUGR due to other causes (e.g.,
preeclampsia, poor nutrition, maternal smoking). If so, it would be expected
that low birth weight for gestational age would have been selected against in
long-resident high-altitude groups. Others (e.g. Krampl et al., 2001; Beall
and Goldstein, 1987) have argued that the expected selective disadvantage
of lower birth weight is reduced at high altitude, suggesting that, rather than
being disadvantageous for survival, lower birth weight is less
disadvantageous, and may even be advantageous, at high altitude.
The answer to the question as to whether lower birth weight is being
selected for or against at high altitude is currently unknown. Some data on
the effects of reduced fetal growth on intrauterine mortality and neonatal

morbidity have been examined through population surveys conducted at high
and low altitudes (e.g., Keyes et al., 2003; Unger et al., 1988). But little
actual data have been collected and no direct tests conducted to find if lower
birth weight is or is not adaptive at high altitude. This is due, in part, to the
inherent difficulties in achieving the requisite study design (e.g. finding
appropriate control groups, complete birth records, etc), due to altitude-
related differences in other factors affecting infant mortality such as
availability and quality of medical care (Moore, 2003).
Here I use a novel genomic and genetic approach for assessing the
adaptive value of reduced fetal growth at high altitude. This approach relies
on the examination of patterns of variation in allele frequency in order to find
evidence of natural selection. My hypothesis is that normal (by sea-level
standards) fetal growth has been selected for in Andean high-altitude
residents. I tested this hypothesis by examining long-term resident,
multigenerational, high-altitude population- the Andean Indians of Bolivia- for
evidence of positive selection in genes that are associated with fetal growth.
Specifically, my aims were:
1. Using a group of prospective candidate genes chosen for their implicated
role in oxygen transport and/or fetal growth, identify genes that show

evidence of directional selection in persons of indigenous, high-altitude
Andean ancestry by
A. examining genomic regions that differ significantly in high- versus low-
altitude populations of South America,
B. determining the most
frequent genotypes for
the genes of interest in
the high-altitude
populations, and
C. identifying the gene
families to which these
most-frequent genotypes
belong. Figure 1.1 Study Aims
2. Determine if the presence of the high-altitude alleles (1B above) is
associated with
A. normal fetal growth, as judged by fetal biometry and birth weight in
relation to gestational age,
B. greater pregnancy-associated increase in maternal uterine artery
C. greater near term, maternal uterine artery blood flow.

3. Test if increased fetal growth is or is not adaptive at high altitude by
determining if the number of Andean alleles (0, 1, or 2 for one locus, or up
to 2 x the number of loci for multiple loci) is
A. directly proportional to birth weight, other indices of fetal growth, or
uterine artery blood flow,
B. indirectly proportional to birth weight, other indices of fetal growth, or
uterine artery blood flow,
C. associated with intermediate fetal growth or uterine artery blood flow,
implicating balancing selection.
Summary of Results:
Results included the identification through multiple genomic tests of five
hypoxia-related gene regions that show differential expression of alleles in
Andeans compared with low altitude controls; namely AMP-activated kinase,
alpha-1 subunit (AMPKal), aryl-hydrocarbon receptor nuclear translocator 2
(ARNT2), ATPase, Na+/K+ transporting, alpha 1 polypeptide (ATP1A1),
cadherin 1, type 1, E-cadherin (epithelial) (CDH1), and inducible nitric oxide
synthase (iNOS) (AIM 1).
For Aim 3, it became clear that CDH1 showed a strong association
with uterine artery blood flow. Specifically, alleles found more frequently in
Andeans compared with low-altitude control populations were positively
correlated with uterine artery diameter, velocity, and volumetric flow during

pregnancy. Furthermore, measurements of one of the protein products of
the CDH1 gene- soluble CDH (sCDH1)- correlated with CDH1 gene region
SNP genotypes. This suggests the genotyped SNPs are linked to the
functional variant that codes for differential expression or function of CDH1
proteins. Additionally, the amount of circulating sCDH1 was positively
correlated with uterine artery blood flow. CDH1 is a cadherin that mediates
cell-cell adhesion and is implicated in successful trophoblast invasion and
subsequent remodeling of the spiral arteries. Thus, CDH1 is a strong
candidate for the mechanism by which Andeans demonstrate genetic
adaptation to high altitude through increased uterine artery blood flow.
Additionally, a strong positive relationship was found between
gestational age and the more characteristic Andean SNP genotypes within
the AMPKal gene region. AMPKal is a kinase activated by low energy
supplies (signaled by AMP), which, in turn, triggers the mTOR pathway.
mTOR increases angiogenesis and vascular cell proliferation in hypoxia
(Humar et al., 2002), as well as enabling blastocyst outgrowth and motility
(Van Winkle, 2001). AMPKal also plays a role in late gestation, regulating
the growth of fetal skeletal muscle (Zhu et al., 2006) and possible the
development of adipose tissue (Cho et al., 2004). Thus, these results
suggest that longer gestational age has been selected for in Andean

populations, perhaps via some genetic difference in the nutrient sensing
signaling molecule AMPKal.
Finally, birth weight (adjusted for factors known to affect birth weight
such as maternal height, fat, parity, baby sex, and gestational age at birth)
was found to be positively associated with alleles that were more frequent in
Andean populations in the ARNT2 gene region. ARNT2 is a HIF-associated
transcription factor that appears to activate transcription of downstream HIF-
targeted genes (Friedman and Fan, 2007), such as those leading to
angiogenesis, cell proliferation, and changes in metabolism (e.g. increased
glycolysis; Harris, 2002). In pregnancy, ARNT2 also influences angiogenesis
during the periimplation period (Daikotu et al., 2003). Thus, ARNT2 may
facilitate the increased birth weight seen in Andeans early in pregnancy via
increasing cell proliferation, trophoblast invasion, and angiogenesis in
vessels supplying the additional blood flow needs of pregnancy. Thus, it
appears that Andeans exhibit a unique genetic profile in the ARNT2 gene
region that leads to increased birth weight at high altitude, suggesting
evolution at high altitude has selected for increased birth weight, as opposed
to small but healthy babies, in adapted populations.

2. Background
2.1 Challenges of Pregnancy
For all known biological organisms, the process of reproduction
profoundly influences fitness. Humans have a long period of gestation to
accommodate for our relatively large body size and slow rate of maturation.
Therefore, pregnancy in humans is an especially crucial time for determining
reproductive success. The physical demands on the mother are great and
the energetic cost of pregnancy must be biologically balanced with an
offsetting benefit to reproductive fitness. Additionally, pregnancy challenges
the health of both mother and fetus acutely (e.g. preeclampsia, hemorrhage,
death2) and can also impact future health (e.g. continuation of gestational
diabetes, fetal programming of adult disease).
A mother supplies vast amounts of resources to a fetus that lives to
term, primarily in the form of nourishment and protection from the external
environment during gestation. Maternal biology must balance the
evolutionary benefits of reproductive success with the requirements to
maintain her own fitness insofar as a pregnancy that is detrimental to her
2 Globally, 1 in 92 women will die directly due to pregnancy (WHO, 2007) and 42 of every 1000
births will result in perinatal death (WHO, 2000).

health might impair her future ability to produce fit offspring, thereby reducing
her overall reproductive success (Tracer, 2002; Ellison, 2003). Thus, the
most adaptive phenotype would be one in which the maternal and fetal self-
interests overlapped to the greatest extent. Indeed, much of the time,
benefits to either the mother or fetus promote the fitness of the other, and the
relationship is reciprocally advantageous. Unfortunately, there are some
times when pregnancy creates too much of a demand on maternal
2.1.1 Mechanical Changes
Human mothers are particularly vulnerable during pregnancy because
of our unique biological commitment to bipedalism, big brains, and extended
length of gestation. With bipedalism, Homo sapiens shift about 25% of the
typical mammalian abdominal volume to the pelvic cavity, as the force of
gravity and restriction of the abdominal cavity (by the rectus abdominus and
abdominal wall muscles) places the support of abdominal organs onto the
pelvis, as opposed to the anterior abdominal wall which supports the organs
of mammalian quadrupeds (Abitbol, 1993). Humans have a larger fetal size
at full gestation relative to the size of the uterine cavity: Human uterine cavity
volume is 150% bigger than the abdominal cavity volume. Other mammals,
including chimpanzees, have a uterine cavity volume of less than 100% of

their abdominal cavity volume (Abitbol, 1993). Additionally, the fossil record
shows that the increased encephalization of hominids was accompanied by
an increase in size of the pelvic outlet and reshaping towards a platypelloid
shape (Lovejoy et al., 1973; Tague and Lovejoy, 1986).
These factors--pelvic reshaping with the adoption of bipedalism, large
uterine compared with abdominal cavity volume, and the gravitational
reorientation of the uterus and abdominal organs over the pelvis-- all
combine to compress abdominal organs and its vasculature, thus serving to
impede blood flow to the lower limbs (Abitbol, 1993). This is particularly
important in late gestation when fetal growth is at its greatest, given that the
drastic restriction of vena cava flow triggers a reduction in cardiac output in
the supine position, requiring significant redirection of the blood flow from
lower extremities to the uteroplacental vascular system via increased
vasodilatation and consequent fall in uteroplacental vascular resistance
(Rockwell et al., 2003; for further discussion, see below). Failure to
adequately redirect blood flow is then hypothesized to lead to conditions
such as preeclampsia. Thus, due to the effects of decreased cardiac output,
Homo sapiens have a much higher risk for preeclampsia compared with
other species3 (Rockwell et al., 2003).
3 Some suggest preecalmpsia occurs in one tenth of all human births. No instance of natural
spontaneous preeclampsia has been described in any nonprimate (Rockwell et al., 2003;
Robillard et al., 2003). There is some evidence of spontaneous preeclampsia in laboratory

2.1.2 Cardiovascular Changes
Early in pregnancy, when metabolic support of the placenta, uterus,
and fetus is high, the principal demands on the mother are primarily
molecular (e.g. nutritional), rather than mechanical (Ciliberto and Marx,
1998). The whole maternal cardiovascular system is challenged by
pregnancy-initiated vascular changes that are required to support the fetus.
To facilitate the supply of fetal demands early in pregnancy, maternal
cardiac output increases; by the fifth week of gestation, cardiac output has
already risen by about 10%, by the 16th week has increased 45% from the
output of prepregnancy, and cardiac output continues to rise until somewhere
between 24 (Ueland and Metcalfe, 1975) and 32 weeks (Robson et al., 1989;
Desai et al., 2004; Gilson et al., 1997; Chapman et al., 1998- left lateral
decubitus), after which it is generally, depending on posture, maintained
through the rest of normal pregnancy.
As mentioned above, the bipedal posture of humans influences
cardiac output; reducing the venous return of blood from the lower limbs
patas monkeys (Palmer et al., 1979; Ramsay etal., 1997) and one cited case of epilepsy
during delivery in a lowland gorilla (Baird, 1981; Robillard et al., 2003). In experimental
settings, preeclampsia has been induced surgically in sheep (Alexander, 2003; Ergaz et al.,
2005) and rhesus macaques (Combs et al., 1993), and genetically in rats (Sharkey et al.,
2001) and mice (Davisson et al., 2002). Even with these anecdotal and experimental
examples of nonhuman preeclampsia, preeclampsia is generally accepted as a
predominantly human disease (e.g. Robillard et al., 2003).

(Kerr, 1965; Ikard et al., 1971; Abitbol, 1993). This has been extensively
shown in the supine and seated posture (e.g. Bieniarz et al., 1969; Ikard et
al., 1971; Abitbol, 1976), but the pressure is relieved in the left lateral
decubitus posture4 (Ueland, 1976). Jeffreys et al. (2006) showed that
women in their third trimester have one-third less uterine artery blood flow
when supine then in the left-lateral decubitus position. Supine exercise
increased the amount of uterine artery blood flow, but flow never reached
that seen at rest in the left lateral decubitus position. Thus, maternal posture
can profoundly affect uteroplacental blood flow.
Less often demonstrated5, but possibly more pertinent to the human
context of evolutionary adaptedness, is the even more drastic venous
occlusion seen in the standing position, which obstructs femoral vein flow
from the limbs about 3.5 times that seen in the supine position in the third
trimester (Ikard etal., 1971). In other words, the human adoption of bipedal
posture has impeded venous return and, subsequently, cardiac output.
Thus, Rockwell et al. (2003) argue that the mechanical limitations of
bipedalisrn on cardiac output must be compensated for by extensive
4 Left lateral decubitus posture is lying down on the left side.
5 For measurement of blood flow erect (or standing) posture would be optimal as
it is more indicative of the predominant posture of the environment of evolutionary
adaptedness. Yet erect measurements done on standing women are not often performed
because of the difficulty involved in measuring standing patients. In the Ikerd etal. (1971)
study, subjects were required to balance on the opposite limb so that the dependent limb did
not bear any weight and was relaxed while the measurements were obtained. In this study,
similar conclusions were made regarding the effects of pregnancy on venous blood flow and
cardiac output, as mentioned above.

trophoblast invasion and maternal vascular remodeling to redirect the
maternal blood flow from the lower extremities to the uteroplacental system.
Additionally, systemic vascular resistance decreases early through a
decline in primarily diastolic, but also systemic blood pressures in pregnancy,
which occurs in turn by vasodilation, uteroplacental vascular growth, and
remodeling (Stock and Metcalfe, 1994). Decreased vascular resistance
permits or perhaps stimulates a rise in cardiac output and also facilitates the
distribution of the blood flow in pregnancy to the uteroplacental circulation
(Stock and Metcalfe, 1994). This rise in cardiac output is also facilitated by
an increase in plasma volume, which begins the first trimester primarily
through venodilation (Ciliberto and Marx, 1998). The rise in plasma volume
continues through pregnancy, whereas red cell mass increases primarily in
the third trimester (Stock and Metcalfe, 1994).
The extensive invasion of trophoblast cells into maternal uterine
vessels presents an additional challenge (Rockwell et al., 2003). Human
pregnancy depends on the hemochorial placenta to exchange nutrients
between the maternal and the fetal environments. With fetal trophoblast
cells incursion into the uterus upon implantation, a series of maternal
physiological changes are triggered. These include a fall of uteroplacental
vascular resistance through the loss of vasoreactivity, and an associated

vascular remodeling of maternal vasculature (Rockwell et al., 2003; see
figure 2.1).
The decrease in uteroplacental resistance contributes to the early as
well as continued rise in uterine artery blood flow (Stock and Metcalfe, 1994).
As a result, near term, between 12 (Thaler et al., 1990) and 25% (Ueland,
1976) of the cardiac output is directed to the uterine artery. Thus, measuring
the key changes in the uterine artery during pregnancy provides information
pertaining to the overall health or pathology of pregnancy.
2.1.3 Hormonal Mechanisms
At conception, a series of hormonal changes are triggered which
contribute to the cardiovascular and other physiologic changes associated
with pregnancy. Preovulation, both gonadotropins, luteinizing hormone (LH)
and follicle-stimulating hormone (FSH) are high. Additionally, estradiol
increases, causing an ovarian follicle to grow rapidly which then forms the
corpus luteum, the endometrium to proliferate and vascularize, and the
myometrium to thicken (Wood, 1994). At ovulation, estrogen levels peak and
begin to fall, while progesterone begins to increase from low levels (Wood,
1994). If conception occurs, the corpus luteum continues to secrete
progesterone, maintaining the endometrium. Additionally, estrogen and
progesterone regulate each other in a feedback loop. Estrogen production

increases the production of prostaglandins, which allow implantation. After
about two months of gestation, the placenta begins to secrete progesterone
to replace that produced by the corpus luteum. The trophoblast and, later,
the placenta, secrete molecules, such as progesterone and interleukin-10,
into the maternal system which, in turn, trigger maternal immunosuppression
(Cadet et al., 1995).
Additionally, the placenta secretes molecules such as vascular
endothelial growth factor (VEGF) and transforming growth factor-P (TGF-P),
which remodel maternal vessels, among other functions (Coultas et al.,
2005). Some suggest that preeclampsia is the result of inadequate signaling
by the placenta for immunosuppresion or vascular remodeling (Coultas et al.,
2005). Thus, the signaling of the placenta is important to the success of

Normal placentation
Abnormal placentation

Figure 2.1 Normal and Abnormal Placentation. Figure from Redman and Sargent
(2005) depicting (A) normal placentation with extensive invasion of
cytotrophoblast cells and maternal vascular remodeling of the spiral arteries, and
(B) abnormal placentation (which often leads to preeclampsia or restricted fetal
growth) with cytotrophoblast invasion inhibited and less remodeling of the
maternal spiral arteries.
2.1.4 HIF Pathway
Interestingly, many of the molecules secreted by the placenta can be
controlled by the presence (or absence) of oxygen via its effects on a
transcription factor. The primary transcription factor induced by hypoxia, the
hypoxia-inducible factor (HIF), has garnered much attention for its role in
cancer, pregnancy, and a host of other biologic responses. HIF is made up
of 2 subunits (alpha and beta) that form a heterodimer that function as a

transcription factor for several genes involved in oxygen homeostasis. HIF
has been identified as a key part of oxygen sensing and signaling of cellular
responses to hypoxia.
Under normoxic conditions, HIF is largely degraded, but under hypoxic
conditions, HIF is both transcriptionally upregulated (Gorlach and Kietzmann,
2007) and protected from degradation by both decreased activity of HIF-
degrading proteins (e.g. prolyl hydroxylase enzymes). Thus, HIF stability is
increased through the absence of a hydroxylation, preventing the binding
with the von Hippel Lindau protein and the stabilization of HIF by reactive
oxygen species (Gorlach and Kietzmann, 2007).
HIF is an active transcriptional factor for several genes, including ones
involved in pH regulation (e.g. carbonic anhydrase), cell proliferation (e.g.
p21, IGF-2), erythropoiesis or iron metabolism (e.g. erythropoietin,
transferrin), glucose metabolism (e.g. glucose transporters), and regulation of
angiogenesis and vascular tone (e.g. VEGF, NOS 2, heme oxygenase 1, aiB-
adrenergic receptor) (Freeman and Barone, 2005).
HIF and the HIF pathway appear to be extremely important in
pregnancy. Many (e.g. Rajakumar et al., 2007; Sweda et al., 2002)
hypothesize that HIF is key to successful trophoblast invasion and
subsequent remodeling of the maternal vessels in order to redirect blood to
the uteroplacental system. In preeclamptic pregnancies, both placental HIF-

1a and HIF-2a concentrations are reduced (Rajakumar et al., 2001;
Rajakumar et al., 2007b), likely due to both increased production and
decreased degradation (Rajakumar et al., 2007b). It is unknown at this point
whether this difference in HIF and its products are the cause or effect of the
preeclamptic phenotype (Rajakumar et al., 2007). Rajakumar et al. (2007)
suggest that the increased HIF activity probably exacerbates preeclampsia
by provoking responses in the maternal endothelium.
Interestingly, placentae of pregnancies characterized by IUGR at low
altitude do not show differences in HIF expression compared with normal
pregnancies (Rajakumar et al., 2007). Thus, the resulting growth restriction
is either not due to hypoxic placentae or the response of the IUGR placentae
to hypoxia is impaired. Placentae from normal pregnancies at high altitude
showed an increase in HIF-1a and von Hippel-Lindau protein compared with
placentae from pregnancies at low or moderate altitude (Zamudio et al.,
2007). While the relationship between concentration of placental HIF-1a and
birth weight was not discussed, Zamudio et al. (2007) did find that babies
born small relative to their placental size had increased expression of HIF-1a
in the placenta, thus may be under greater hypoxic stress. Thus, it appears
that with the hypoxia of high altitude, IUGR placentas may have also have
increased HIF concentration, resembling low-altitude preeclamptic placentae.
If increased HIF does cause maternal vascular disease as Rajakumar et al.

(2007) suggest, then an increase in high altitude IUGR placental HIF may
have an effect much wider than just reduced fetal growth.
With the implied roles of HIF in both hypoxia and pregnancy, the HIF
pathway and its targets are a logical choice for prospective study of
candidate genes pregnancy at high altitude.
2.1.5 Effect of Physiologic Responses on Maternal and Fetal Health
As described above, in order to maintain proper and sufficient
development of the fetus, pregnancy demands extensive maternal
physiological changes. These changes exemplify the delicate balance
between fetal and maternal interests. Immediate maternal reproductive
success in the form of a healthy fetus must be balanced against the
physiologic costs which could threaten her own fitness and future
reproductive possibilities (and overall reproductive fitness). If the demands
of pregnancy are too great the maternal ability to survive may be
endangered, and, thus, also compromise the fetus existence.
Adequate nourishment in the form of oxygen, glucose, lactate and
amino acids (McMillen et al., 2001), as well as a protective uterine
environment, are necessary for normal fetal growth. Without these, fetal
growth will likely be impaired and place the fetus at risk for IUGR.

Growth-restricted fetuses are at increased risk of intrauterine and
neonatal (from birth to 28 days after birth) mortality (Williams et al., 1982), as
well as increased infant morbidity (Starfield et al., 1982) (see section 2.3).
Additionally, low birth weight due to being born small for gestational age
(SGA) has been associated with increased risk for noninsulin-dependent
diabetes mellitus (e.g. Barker, 1998; Godfrey and Barker, 2000; Harding,
2001; Ravelli et al., 1998), hypertension (e.g. Barker, 2002; Barker etal.,
2002; Roseboom et al., 1999, coronary heart disease (e.g. Barker, 2001;
Eriksson et al., 2001; Forsen et al., 1999; Huxley et al., 2000; Stein et al.,
1996), and obesity, which aggravates many of the above conditions (e.g.
Oken and Gillman, 2003). These later-in-life diseases have been found to be
associated with reduced fetal growth resulting from a myriad of causes, such
as toxic chemical exposure (for instance, to nicotine (Slotkin, 1998; Power
and Jefferis, 2002), cocaine (Slotkin, 1998), and some pesticides (Cory-
Slechta et al., 2003)), twinning (Christensen et al., 2001; Ozanne and Hales,
2002; Johansson-Kark et al., 2002), low socioeconomic status (Kramer,
1998; Giussani et al., 2001), and emotional stress (Mulder et al., 2002).
Although, at this time, the association between altitude-associated
intrauterine growth restriction (IUGR) and adult disease has not been
established, because of the similarities in fetal coping mechanisms when
faced with growth restriction triggered by one of multiple causes (such as

stress or hypoxia), it is reasonable to expect there may be an association
between hypoxia-related IUGR and adult health and illness. Thus, adequate
nourishment of the fetus during gestation profoundly affects its future health
and fitness as a neonate and may affect its health throughout the lifespan.
Besides fetal health, pregnancy may also affect maternal health. In
order for a mother to maximize her own fitness, she needs to produce a
healthy offspring with each pregnancy, but not at the cost of future offspring6.
Thus, from the maternal perspective, in an optimal pregnancy, the maternal
system supports the health and development of her fetus, but not at the
expense of her own long-term reproductive health.
Pregnancies in which fetal and maternal fitness requirements are
discordant (e.g. when maintenance of fetal growth can only continue at the
cost of maternal health) can trigger maternal health problems, including
diabetes, edema, malnutrition, anemia, bleeding, puerperal endometritis, and
preeclampsia/eclampsia. Many of these pathologies can be understood as
the result of greater fetal demand on maternal resources than that which she
can support. Preeclampsia, too, may be seen in this fashion, as
6 According to life history theory, as maternal age and parity increase, the mother should
begin to accept more risk to her health with each pregnancy to produce healthy offspring.
Indeed, several studies have shown that some populations exhibit a decline in maternal
condition, as measured by maternal energy reserves and subsequent infant ponderal index,
after pregnancy with increasing parity (Tracer, 2002).

preeclampsia is more common under a range of conditions7, all of which
share an increased fetal demand relative to the maternal capacity for supply.
Increase in fetal demand beyond maternal supply is associated with
increased maternal peripheral vascular resistance, and, in turn, elevated
maternal blood pressure (Haig, 1993; Odent, 2001), and hence threatening
maternal health. Thus, preeclampsia can be seen as a result of the struggle
between the conflicting interests of mother and fetus (Haig, 1993).
For optimization of fitness, a balance must be struck between these
often complimentary, but occasionally conflicting maternal (physiological self-
maintenance) and fetal self-interests. Fitness may be increased through
adaptations, or variable traits that confer increased fitness for an organism
(Mayr, 1988; Dobzhansky, 1956). Not all adaptations, though, are heritable
and, thus, subject to change through biological evolution. Specific
adaptations may thus vary in their mode of expression (physiological,
behavioral, developmental, etc.) and not all adaptations are produced
through gene expression (many are acquired or environmentally influenced).
Only the frequency of genetically heritable traits may be affected by
biological evolution through natural selection within a population. Therefore,
it is important in discussions of adaptation to distinguish genetic from non-
7 Conditions such as twinning and hydatiform moles (Moore et al., 1982) which are both
associated with increased placental tissue.

genetic aspects of individual phenotypes. In this thesis, we are primarily
concerned with heritable adaptations, as these are subject to natural
selection, and, thus, biological evolution. In this perspective, the most
adaptive pregnancy/fetal growth phenotype would be one in which maternal
and fetal fitness were maximized, minimizing detrimental phenotypes, such
as pathological growth restriction.
2.2 lUGRorSGA?
Birth weight is used as an indicator of fetal growth because often more
direct measures of organ growth (e.g. head circumference, abdominal
circumference, maturity of organs, fat deposition, etc.) are not available or as
reliable. Thus, birth weight is the most accurate and most easily obtained
indicator of fetal growth when considered in relation to gestational age at
birth (Haggarty et al., 2004). Although it may be the best measure available
in many cases, it is by no means optimal. Birth weight takes into account
body size, rather than the relative growth of the brain versus the trunk.
Additionally, birth weight may measure hydration rather than tissue volume,
although measures of hydration may still be indicative of organ function and
the health of the newborn (Bielecka-Winnicka, 1966; Roy and Sinclair, 1975).
Thus, babies may be born small for their gestational age due to
nonpathological causes such as small maternal stature or parity (Mamelle et

al., 2001) rather than as a result of asymmetrical growth (indicated by slower
abdominal growth compared with head growth). This is important because
some SGA babies would be classified as growth restricted when they do not
share the risk for higher morbidity and mortality with truly growth-restricted
Small for gestational age (SGA) is defined as an infant born in the
lowest 10th percentile of birth weight for gestational age and sex (Kliegman
and Das, 2002). The tenth percentile is used as a cutoff because there is an
exponential rise in perinatal, neonatal, and infant mortality and morbidity
below the tenth percentile (Williams et al., 1982). More recently, the 5th and
3rd percentiles have been used to identify the portion of infants at the
greatest risk for adverse outcomes (Chatelain, 2002). The measurement of
SGA is imperfect as it does not assess asymmetrical growth. Not assessing
asymmetrical growth results in the failure to include all growth-restricted
fetuses but includes constitutionally-small fetuses. Yet, this assessment is
still widely used because the inclusion of healthy but small babies is relatively
rare. One must simply be mindful of cases in which constitutionally-low birth
weight babies may substantially contribute to SGA-classified infants.
In the current literature, IUGR is defined as low birth weight for
gestational age, as the result of some pathology that does not allow the fetus
to reach its growth potential. In comparison, SGA is defined also as low birth

weight for gestation age, but often lacking pathology (Chatelain, 2000;
Bakketeig, 1998). Often, these terms are incorrectly used interchangeably.
Many use the term IUGR because of its implication of increased importance
concerning the effects on health. Yet, without measurement of pathology or
solid etiologic reasoning, this term should not be used. SGA is more
appropriate until actual pathology is identified and substantiated.
Clinically, when the presence of a pathology causing decreased fetal
growth cannot be assessed, SGA is measured in place of IUGR.
Classification of reduced fetal growth as IUGR or SGA is problematic,
particularly in babies born small at high altitude. First, the actual contribution
of hypoxia on reduced fetal growth is difficult to assess. Second, some
would argue that hypoxia may not even cause pathology (e.g. Krampl, 2000).
Thus, unless referring to a specific case where the growth potential of a fetus
has been pathologically impeded (which will be called IUGR), this study will
identify and use the term SGA when no pathology has been clearly
2.3 SGA
The large size of the human fetus at term presents a considerable
challenge to maternal physiology. For instance, maternal physiology may not
be able to adequately shield the fetus from a variety environmental stresses

and insults, such as toxic chemical exposure, malnutrition, or emotional
stress. When the fetus is incompletely protected it may either die, escape
the uterine environment early (premature birth), or slow its growth to
accommodate conditions of the insufficient uterine environment. This
decelerated growth results in babies that are born with lower birth weights
compared with other normal infants born at the same gestational age. This
effect of slowed fetal growth due to a pathological problem is termed
intrauterine growth restriction (IUGR).
Clearly, intrauterine death has the most detrimental effect on fitness.
Premature birth has a rather immediate and severe effect on survival; about
85% of neonatal deaths in Great Britain are due to premature birth (Rush et
al., 1976). Since preterm births are more prevalent in developing countries
(Villar et al., 2004 suggest two times more prevalent compared with
developed countries), it is likely that prematurity has an even greater public
health impact in poorer nations. IUGR has a lesser effect compared with
prematurity on immediate infant mortality, but is still significant. In India, for
example, the relative risk of perinatal mortality of babies born with IUGR was
2.6 greater than those of normal gestational age and birth weight babies.
Babies born prematurely had a 21.1-fold increased risk of perinatal mortality
relative to normal gestational age and birth weight babies (Mavalankar et al.,
1991). Part of this differential can be attributed to the fact that some severely

growth-restricted fetuses die in utero and would not be included in this kind of
epidemiological survey. Yet, this cannot account for all of the difference in
risk: premature infants are much more likely to die soon after birth than are
IUGR babies (Tambyraja and Ratnam, 1982; Wilcox, 2001).
The effect of IUGR is longer-lasting than that of prematurity.
Premature birth has a severe short-term effect on ophthalmologic (Kramer,
2003), neurological, psychomotor developmental, and neonatal and
childhood pulmonary health. IUGR on the other hand, is associated with
increased risk of metabolic imbalances (e.g. neonatal hypoglycemia,
hypocalcaemia, polycythemia), growth irregularities during early childhood,
neurocognitive problems (Kramer, 2003), and an increased risk for
cardiovascular disease, non-insulin dependent diabetes, stroke, hypertension
(Barker 1998), and possibly obesity in adult life (Ravelli et al., 1999).
Although premature and IUGR infants may weigh the same at birth,
their varied morbidities illustrate differences in the etiology of prematurity and
IUGR. While premature infants have bodies and limbs that are appropriate
for their gestational age, babies that have been growth-restricted show a
distinctive body shape. When a fetus is growth restricted, it is due to some
sort of fetal deprivation, as mentioned above, be it of nutrients (calories,
vitamins, or even oxygen) or stress (exogenous chemical or maternal), which
are thought to alter nutrient uptake and metabolism (Lesage et al., 2004).

For example, when faced with deficiencies, the fetus directs resources
preferentially to the brain through the diversion of blood flow from the body
and other organ systems to the fetal head, thus preserving its development.
This ensures the fetus immediate survival, but may compromise its later
health. When resources are scarce, then, trunk growth and fat deposition
are decreased. Thus, growth restricted fetuses show asymmetric growth
restriction with abdominal organs being smaller relative to the brain.
Therefore, growth-restricted fetuses are often described as distinctively long
and lean when compared with their normally-grown counterparts.
This is not just a problem of reduced organ size. Although there may
be a different effect depending on the timing of growth restriction8,
pronounced fetal growth restriction is most evident in the third trimester
because this is when the majority of fetal growth occurs. Thus, while the
antecedents to IUGR may occur earlier, most of what is known about the
effects of IUGR on infant and fetal health is based on third trimester
A good example of the detrimental effects of asymmetrical organ
growth is evident in the kidney. Reduced cell replication slows the growth of
8 This is debated; some suggest that IUGR is the result of early events in pregnancy which
establish growth restriction early in development and others suggest IUGR occurs late
because measurable differences primarily appear in the third trimester, when the majority of
fetal growth occurs. While most agree there is likely a different effect depending on time of
the origin of fetal growth restriction, it is not entirely known what exactly this differential effect
is (Barker, 1998; Rockwell et al., 2003).

the kidney, which results in congenital oligonephropathy (Barker, 1998).
Oligonephropathy is characterized by a decrease in total kidney cells,
including beta cells and renal glomeruli, and organ shape which is small, thin,
and long.
The number of nephrons in the human kidney shows great variation
between 300,000 to 1,000,000, with an average of 600,000 (Nyengaard,
1992). This wide variation in nephron number is created during fetal
development, as no renal cell differentiation occurs after birth. Thus, slowing
of kidney growth in the third trimester permanently affects lifelong kidney
function. Hinchliffe et al. (1992) found babies with IUGR have 35% fewer
nephrons and increased glomerular volume compared with normal birth
weight babies. As a result, there is hyperfiltration that leads to
glomerulusclerosis and hypertension (Lumbers, 2001). This fetal
programming of kidney function is exacerbated by rapid childhood growth,
obesity, and additional hypertension, which all demand increased
hyperfiltration and hypertension. Thus, the already challenged kidney is
further stressed, leading to conditions that continue to worsen in life.
The hypothalamic-pituitary-adrenal axis is also established largely in
late gestation. As described by Tortortiello (2002), if the fetus is stressed late
in gestation, levels of corticotrophin-releasing hormones (CRH) increase.
CRH binds corticotrophin-releasing hormone receptors on the pituitary and

triggers the release adrenocorticotrophic hormone (ACTH) and increased
basal and stress-induced glucocorticoid secretion. When this occurs during
gestation, the CRH-responsiveness that persists throughout life is
programmed, and reduces the plasticity of the pituitary (Tortoriello, 2002).
People who were growth-restricted in utero show significantly higher levels of
cortisol, which is associated (although the mechanism is not entirely clear-
see Whitworth et al., 2000, Panarelli, 1998) with increased hypertension risk
in adult life. Corticoids in the third trimester signal to the fetus that the
mother is experiencing some kind of stress that may render the maternal
environment less advantageous. Thus, corticoids signal for quick maturation
and curtailing of fetal organ growth (such as the oligonephropathy of the
kidney described above). Taking advantage of this, treatment with
glucocorticoids is often employed when preterm labor occurs in order to
hasten lung maturation. Additionally, increased exposure of glucocorticoids
during gestation also interferes with the development of the sympathetic
nervous system, although the exact mechanisms remain unclear (Young,
2002). Thus, IUGR as characterized by asymmetrical growth of the brain
compared with other fetal organs raises infant, and even adult, morbidity and
mortality. Hence, the best evaluation of intrauterine growth restriction would
be to measure fetal or newborn organs and compare growth of the brain to
other body organs, such as the kidneys, liver, heart, muscle (such as that of

the limbs), as well as fat deposition (Haggarty et al., 2004), but this is
extremely difficult unless the baby has been delivered stillborn.
Recently, ultrasound techniques have been used to measure
subcutaneous fat in the upper limbs of fetuses to assess total fat deposition
(e.g. Padoan et al., 2004). Also, fetal abdominal circumference relative to
head circumference as assessed by ultrasound, or immediately at birth, can
provide a comparison of the growth of the abdominal organs versus the brain
(e.g. Padoan et al., 2004). Fetal body fat measurements can also be reliably
and easily done, yet are not likely to be adopted in developing countries
because of the lack of time or resources to learn and utilize an additional
Another index of fetal growth and SGA is provided by the ponderal
index, commonly employed in many hospitals worldwide. This index uses
the weight of the neonate and length9, both of which are routinely measured,
so as to indicate whether an infant has low body weight for length. While
theoretically sound, in reality the ponderal index is not as good at predicting
body fat content, compared with the abdominal circumference or the ratio
between abdominal circumference and head circumference (Haggarty et al.,
9Ponderal index is equal to birth weight (in kilograms) divided by birth length (in meters)
cubed or:
PI = birth weight
(birth length)3

2004), likely due to difficulties in consistently measuring the length of a
squirmy infant. Consequently, birth weight adjusted for gestational age is
often the most accurate indicator of fetal growth available.
2.4 Causes of IUGR
2.4.1 Malnutrition
IUGR has been attributed to many different causes. Perhaps the most
studied is malnutrition; more specifically, malnutrition due to caloric
deprivation during pregnancy. Caloric undernutrition can be compensated
for by reduced physical activity, reduced metabolic rate, or increased
utilization of fat deposits (Henriksen, 1999), but occasionally the maternal
system cannot buffer her fetus adequately from undernutrition, and nutrition-
related IUGR results.
Nutrition-related IUGR has been associated with increased neonatal
mortality (e.g. Ceesay et al., 1997; Rush, 1989; Kramer and Kakuma, 2003),
infant mortality especially through sudden infant death syndrome (Oyen et
al., 1995) and childhood mortality (McCormick, 1985). Morbidity also
increases with malnutrition-related IUGR, as it is associated with increased
risk for perinatal asphyxia (Neerhor, 1995), childhood cognitive delays and
neurologic impairment (Mervis et al., 1995). Metaanalyses (Kramer and
Kakuma, 2003; Villar et al., 2004) have shown that nutritional interventions

that were actually able to change the maternal diet do improve fetal growth,
but the actual, realized public health benefit gained through advice-only
interventions is debated (Kramer and Kakuma, 2003).
Additionally, IUGR related to malnutrition has been found to have
consequences that extend beyond childhood. Nutrition-related IUGR
increases risk for many chronic diseases of adulthood ranging from
noninsulin-dependent diabetes mellitus (e.g. Barker, 1998), hypertension
(e.g. Roseboom et al., 1999), coronary heart disease (Forsen et al., 1999;
Huxley et al., 2000), and the propensity for obesity, which aggravates many
of the above conditions (Oken and Gillman, 2003). Thus, IUGR due to
malnutrition has extensive lifelong detrimental consequences.
2.4.2 Toxic Insult
Exposure to a variety of toxic chemicals in utero is associated with
fetal growth restriction. For instance, maternal cigarette smoking during
pregnancy causes placental insufficiency (Kramer, 1987), which causes
drastic growth restriction. Smoking during pregnancy has been shown not
only to cause IUGR, but also increase the incidence of premature birth and
the risk for sudden intrauterine unexplained death or SIUD (Froen et al.,

IUGR caused by maternal smoking has been related to increased
perinatal and neonatal death (Froen et al., 2004; Rush and Kass, 1972).
Additionally, APGAR scores (a measure of infant health at birth based on
muscle activity, pulse, grimacing, skin color appearance, and respiration) are
lower and lung function impaired in babies with smoking-related IUGR
(Greenough et al., 2004). Children who were growth restricted in utero have
impaired bone growth and less bone mass (Jones et al., 1999). Studies
linking smoking-related IUGR to later-in-life diseases have not been
performed, likely due to the absence of verifiable links between maternal
smoking with IUGR in birth records. Animal studies, though, suggest that
prenatal nicotine exposure impairs cardiac function in young rats (Zhu et al.,
1997). Other studies suggest rats exposed to nicotine show other signs that
suggest that adult cardiovascular disease may be programmed in utero
through fetal growth restriction (Zhang, 2005).
Moderate (two drinks per day or more) exposure to alcohol
(specifically ethanol) during gestation is also associated with IUGR (Cogswell
and Yip, 1995). Alcohol-associated IUGR is linked with increased perinatal
and infant mortality, reduced mental capacity throughout life, and increased
incidence of cerebral palsy, as well as many other morbidities (Olegard et al.,
1979). Again, human retrospective or longitudinal studies have not been
done to examine the possible association between alcohol-related IUGR and

adult cardiovascular disease, but studies in rats have shown marked change
in heart myocyte function and apoptosis, which suggests fetal programming
may occur (Zhang, 2005). With chronic prenatal ethanol exposure,
cardiomyocytes show multinucleation and changes in myofilament
organization (Adickes et al., 1993), Ca2+ responsiveness is decreased, likely
due to Ca2+ overloading (Ren et al., 2002), and myocyte apoptosis increases,
as shown by increased caspase-3 activation (Ren et al., 2002), which all
contribute to reduced cardiac contractile function.
Cocaine use during pregnancy is associated with increased incidence
of IUGR (Zhang, 2005; Calhoun and Watson, 1991), and perinatal mortality
(Calhoun and Watson, 1991). APGAR scores are lower for cocaine-related
growth restricted infants (Calhoun and Watson, 1991) and the infants also
have increased neurobehavioral problems, increased risk for both
cardiovascular and developmental disorders (Zhang, 2005). These
alterations of normal heart function persist throughout adulthood (Bae et al.,
There has also been IUGR found associated with exposure during
pregnancy to pesticides (e.g. Siddiqui et al., 2003), yet the effect on morbidity
and mortality has not been shown because of the low incidence of
demonstrable pesticide exposure in the community.

2.4.3 Socioeconomic Factors
Education, income, and prenatal care have been shown to be
associated with increased rates of IUGR, independent of any factors
mentioned thus far (Collins and Martin, 1998). Although the effect of
socioeconomic elements may be independent of the previous factors, it
should be noted that low socioeconomic status is often intricately linked with
malnutrition and a host of toxic chemical exposure risks (Kramer et al.,
It has been suggested that, besides access to health care, the main
link from socioeconomic status to fetal growth may be by increasing maternal
stress10 and impairing psychosocial well-being11 (Kramer et al., 2000),
although this is less well-studied. Maternal stress during pregnancy is
thought to restrict fetal growth through increased glucocorticoid stimulus
(Lesage et al., 2004). The exact mechanism by which excessive
glucocorticoids affect fetal development is not known, but some suggest
reduced fetal adrenal growth and activity along with reduced corticotrophin-
releasing hormones probably play a role (Lesage et al., 2004). Additionally,
maternal increases in catecholamine release decreases placental nutrient
10 Maternal stress increases both maternal cortisol levels and placental secretion of
corticotrophin-releasing hormone (CRH), which can both induce premature labor, increase
maternal blood pressure, and alter fetal HPA development (Kramer et al,. 2000).
11 Decreased social support networks, increased acute stressful life events, increased
chronic stress, abuse and depression contribute to decreased psychosocial well-being,
which can all be linked to CRH production (Kramer et al., 2000).

perfusion and can trigger both preterm delivery and IUGR (Copper et al.,
1996; Rondo et al., 2003). Thus, through the factors of access to health
care, stress and feelings of wellbeing, economics can have a significant
effect on pregnancy.
Bolivia, as part of the periphery/developing world, is extremely poor,
exhibiting major instability in economic, political, and health arenas. Within
the country, there are notable social and economic stratification and power
differentials, and these divisions often occur along ethnic lines (typically
between European, mestizo, and indigenous peoples) leaving indigenous
Andeans little agency. These factors all greatly impact the etiology of
intrauterine growth restriction (Kramer, 1998; Giussani et al., 2001; Mulder et
al., 2002).
Bolivia is one of the poorest countries of Latin America with a gross
national income of $900 per capita in 2002, a health development index
indicator of .548 (UN, 2002), and the World Bank reporting 29.4% of the
population lives in extreme poverty (below $1/day, 1999 UN Human
Development Report). When income is broken down by type of worker, the
disparities between the rich and poor in Bolivia show that a great many
working Bolivians live on very little money12 (Thiele and Piazolo, 2002).
12 Smallholders, who make up about 40% of all Bolivian workers, averaged 244 Bolivianos or
about US$44 per month in 1998 (Thiele and Piazolo, 2002). Urban informals (about 27% of all

Consideration of ethnicity adds an extra dimension to this
discouraging picture. Fifty-seven percent of the population of Bolivia is of
indigenous descent (1992 census cited by WHO, 1996), with 36% of the
urban population (EIH, 1993) and, possibly, 92% of the rural population
made up of indigenous Bolivians (WHO, 1996). In La Paz, where this study
was conducted, almost haif (49.5%) of the population is of indigenous
descent (Censo Nacional Poblacion y Vivienda, 2001). Indigenous people,
though, carry much more of the burden of poverty in Bolivia, with 68% in
poverty, compared with nonindigenous population for whom 36% were in
poverty (WHO, 1996).
As mentioned above, socioeconomic status is a proven environmental
and psychological stressor which has been found to exacerbate IUGR
(Kramer, 1998); (Giussani et al., 2001). Thus, it could be expected that
indigenous Bolivians may have a drastically increased risk for IUGR
compared with their richer, more advantaged compatriots. Intriguingly,
however, the indigenous populations of Bolivia have a lower incidence of
IUGR compared with nonindigenous groups (Haas et al., 1987), despite their
increased rates of poverty and other non-hypoxia related risk factors. This
protection from IUGR despite the increase in other risk factors may be due to
multi-generational biological adaptation to the hypoxia of high altitude (e.g.
Bolivian workers) averaged 415 Bolivianos or US$75 per month. Each of these types of workers
were likely the primary wage earners in their household (Thiele and Piazolo, 2002).

Jensen and Moore, 1997; Brutstaert et al., 2000). Thus, the likely
confounding social factors for this study are likely to be negative
confounders, increasing the likelihood of seeing less of the effect of hypoxic
SGA in indigenous Andeans.
In juxtaposition to the Andean communities are the high altitude
populations of Colorado. In one study (Unger et al., 1988), women living at
high altitude had higher socioeconomic indicators (such as education, fewer
teenage mothers, etc.), as well as better access and utilization of health care
compared with their low-altitude counterparts (living in Denver and the
Colorado plains), yet still gave birth to babies of smaller birth weight.
2.4.4 Hypoxia
High altitude also has a pronounced effect on fetal growth, having an
effect size second only to gestational age and greater than the effects of
socioeconomic status (Giussani, 2001), normal variation in parity, smoking,
or prenatal care (Jensen and Moore, 1997). Birth weights have been shown
to be inversely correlated with altitude during pregnancy; above 2000 meters
in Colorado (Moore, 2003), for every 1000 meter increase in altitude, there is
a 102 gram decrease in average birth weight (rate adjusted for potential
confounders; Jensen and Moore, 1997). In the Andes, Mortola et al. (2000)
showed a 122 gram decrease in birth weight for every 1000 meter increase

in altitude above 2000 meters. This decrease is independent of other IUGR
risk factors, including parity, smoking status, weight gain, prenatal visits,
hypertension, or fetal sex (Jensen and Moore, 1997). Thus, this reduction in
birth weight is attributed to the hypoxic (low oxygen) environment found at
high altitudes (see section 2.5 below for further explanation).
As mentioned above, fetal demand for oxygen may exceed maternal
oxygen supply, causing significant stress on maternal physiology. Thus, the
hypoxic environment found at high altitude, presents a challenge for both
maternal health and fetal growth and health during pregnancy.
2.5 High Altitude
Atmospheric pressure decreases nonlinearly with altitude. At sea
level, the partial pressure of oxygen is 21.18 kPa (20.9% O2 *101.325 kPa),
and humans generally have an arterial oxygen saturation of 98-100%. At
3000 meters, the partial pressure of oxygen drops to 14.65 kPa, or about 2/3
the partial pressure of oxygen at sea level (see Figure 2.1). Consequently, at
low altitude, average human arterial oxygen saturation at 3000 meters drops
to 90-95%. At 3800 meters (the altitude of La Paz, Bolivia), the partial
pressure of atmospheric oxygen is 13.22 kPa, which is about 62% that of sea
level. But, it has been shown that arterial oxygen saturation does not
uniformly drop with falling atmospheric oxygen partial pressure. For

instance, people of Andean descent at 3800 have been shown to have at
least 1% higher arterial oxygen saturation at rest13 than people of European
descent at the same altitude and activity level (Beall et al., 1999; Beall, 2000;
Brutsaert et al., 2000; Brutsaert, 2001; Winslow et al., 1989). Additionally,
Andeans maintain their normal arterial oxygen saturation with exercise to a
greater extent than do low altitude natives at high altitude (Favier et al.,
1995), likely due to more efficient diffusion of inspired oxygen to the blood
through greater lung diffusing capacity. This is possibly the result of both
environmental and genetic factors: Increased lung volume and diffusing
capacity is evident in humans (Frisancho, 1969) and beagles (Johnson et al.,
1985; Grover et al., 1988) that spend their childhood at high altitude. Lung
capacity is also influenced by chest size, a primarily genetic trait, which
varies directly with length of population residence rather than with
environmental exposure (Frisancho and Greksa, 1989).
The high-altitude environment consists of other factors besides
hypoxia. Due to adiabatic lapse, air cools with increasing altitude, dropping
in temperature at a rate of about 10 degrees centigrade/ 1000 meters. Thus,
at a certain point in the ascent of air (pushed over mountain ranges by the jet
stream or weather systems), the dew point is likely to be reached, causing
water vapor to condense and precipitate out, leaving the ascending air much
13 There is a statistical difference in both raw Sa02 and Sa02 after adjusting for oxygen consumption

dryer. Additionally, at high altitude, less ozone (O3) protects the earths
surface from the sun. Ozone absorbs a fair amount of ultraviolet radiation
from the sun. Therefore, with decreasing ozone protection there is an
increase in ultraviolet radiation (about a 4% increase in UVR per 300
meters). Therefore, the combination of low humidity, less nocturnal ozone
insulation, and increased daytime solar radiation found in high altitude allows
for more diurnal temperature variation (Niermeyer et al., 2001; Moore et al.,
Nonhypoxic stressors found at high altitude may be relatively more
easily mediated through cultural adaptations, such as clothing and
agricultural innovations, than is hypoxia. However, these additional factors
have been found in the absence of hypoxia to have little impact on fetal
growth or other major phenotypes of high altitude. Thus, it seems likely that
the altitude-specific IUGR is likely attributable to hypoxia rather than other
environmental factors of high altitude.
Unlike many other environmental stressors, hypoxia is not easily
modified by human culture. At high altitudes, hypoxia may be avoided only
by descending or using recently-developed technologies such as oxygen
supplementation. The effects of hypoxia may be alleviated slightly by labor
minimization or labor delegation to younger children who require less oxygen
due to smaller body sizes compared with adults for a similar task (Thomas,

1976). Some people do utilize temporary emigration from high altitude, but
both emigration and labor adjustment are limited in scope and use. Because
cultural adaptation does relatively little to relieve chronic hypoxia, humans
who have lived at high altitude for multiple generations may have acquired
genetic biological adaptations to hypoxic environments.
Figure 2.2 Total Atmospheric Pressure and the Partial Pressure of Oxygen
Decrease with Increasing Altitude.

2.6 Basic Population Differences at High Altitude
Not all high-altitude populations, though, are similarly affected by
hypoxia. Populations with longer residence at high altitude (such as
Andeans or Tibetans) have a lesser reduction in birth weight compared with
relatively newer immigrant populations (Haas et al., 1980; Moore, 1990;
Zamudio et al., 1993; Moore et al., 2001), such as European (less than 500
years at high altitude) or Han (less than 60 years at high altitude) people
who have recently moved to high altitude. This suggests that adaptations to
hypoxia have been acquired in longer-resident populations, but the specific
adaptations involved are unknown.
The Tibetan plateau is estimated to have been occupied as early as
25,000-50,000 years ago (Sensui, 1981; Zhimin, 1982; Aldenderfer, 2003;
Zhang and Li, 2002). Thus, there has been ample time for adaptations to
high altitude to have accumulated in this population given the intense
selective pressure exerted by hypoxia. Indeed, Tibetans, compared with
Europeans or Han show less altitude-associated decrease in birth weight
(Moore et al., 2001; Zamudio et al., 1993). With respect to vascular
response, the Tibetans showed a greater increase in uterine artery flow
velocity, and a subsequent increase in the UA/CI flow velocity ratio, likely

indicating a greater redistribution of blood flow to the uterine artery in
Tibetans (Moore et al., 2001).
Andeans have inhabited high altitude for an extended period of time
intermediate to long-resident Tibetans and the short-resident Han and
Europeans. Yet, similar to Tibetans, the extended length, spanning
hundreds of generations, of high-altitude residence by the Andeans likely has
been long enough for natural selection to have affected this population.
Current evidence shows that humans were in the Americas at least 14,000
years ago14 (Gilbert et al., 2008) and the human occupation of South
America to have started at least 10,000 years ago (Denell et al., 1988), but
cultural innovations likely did not make the Andean Altiplano inhabitable until
4,000-6,000 years ago (Niermeyer et al., 2001; MacNeish and Berger, 1970;
Lynch, 1978; Nunez, 1983). Andeans have a lesser altitude-associated
decrease in birth weight compared with people of European descent (Haas et
al., 1980; Julian et al., 2006). This could possibly be attributed to the greater
increase in uterine artery blood flow observed in Andeans compared with
Europeans (Wilson et al., 2007), as discussed below.
14 Gilbert et al. (2008) recently found human coprolites in Oregon dating about 14,300 years old,
predating the previous evidence for Clovis occupation by about 1,000 years.

2.7 Hypoxic IUGR: Adaptive or Detrimental? The Debate
This debate centers around the link between hypoxia related IUGR
and infant mortality or morbidity. Whereas associations between nutrition-
related IUGR and increased perinatal, infant, and adult mortalities and
morbidities have been demonstrated, the relationship between hypoxia-
related IUGR and mortality and morbidity throughout the lifecycle has not
been as clearly shown.
2.7.1 Argument for lUGRs Harmful Effect at High Altitude
Since all pathological causes of IUGR studied thus far (see above
discussion) have been shown to have a negative effect (or at least no
positive effect) on infant health and survivorship, many (e.g. Howard et al.,
1957; McCullough et al., 1977; Unger et al., 1988; Moore et al., 2001; Keyes
et al., 2003) have suggested that hypoxia-related IUGR is likely similarly
detrimental (further discussion below). These infants are born with the signs
of pathological, asymmetrical growth restriction (Keyes et al., 2003), as
opposed to regular, constitutional small body size for gestational age. Thus,
it is likely that these infants likely suffer from thoracic organ hypotrophy, as
well as impairment of HPA-axis, just as infants born growth-restricted due to
other causes. Furthermore, infants at high altitude experience increased risk
for a variety of morbidities, including early arterial oxygen desaturation

(Niermeyer, 2003), acute respiratory infection (Choudhuri et al., 2006), and
persistent pulmonary hypertension of the newborn (Niermeyer, 1998).
Additionally, longer-resident high-altitude populations, such as
Tibetans and Andeans, show less of a reduction in birth weight in full term
babies with altitude than do non-adapted populations, such as Europeans or
Han (Zamudio et al., 1993; Moore et al., 2001). This would suggest that
heavier birth weight has been selected for at high altitude. Although the
specific mechanisms for greater intrauterine growth are not fully known, it is
suspected to occur through increased maternal oxygen transport to the
placenta through increased blood flow in Andeans (Wilson et al., 2007) and
Tibetans (Moore et al., 2001).
The hypothesized adaptation of increased oxygen transport is not
sufficient, though, to completely protect long-resident populations. Among
Tibetans and Andeans, both populations who have been suggested to have
lived long enough at high altitude to have acquired adaptations, there is still a
decrease in term birth weights with increasing altitude (Niermeyer et al.,
2001). The continued reduction in birth weight may reflect physiological
constraints upon the human system. Adaptations utilizing changes in blood
flow and blood oxygen content are limited by the constraints on the human
biological system like extended gestation and bipedal posture (dictated by
phylogenetic precursors; see above discussion regarding cardiac output).

Constraints involved in oxygen transport, such as the viscosity and clotting
potential of increased red blood cells, such as those seen in high-altitude
newcomers, can lead to complications such as stroke. Thus, adaptive
strategies to address one environmental challenge can be limited (due to
physiological and mechanical constraints) and detrimental if too extensively
utilized. Hence, if hypoxia-related IUGR is a threat to maternal and fetal
survival and, thus, reproductive success, selective pressure would be placed
on populations to increase fetal growth.
2.7.2 Argument for lUGRs Beneficial Effect at High Altitude
While the lesser decrease in term birth weight in long-resident high
altitude populations suggests that fetal growth restriction has been at least
partially overcome, some (Krampl, 2002; Beall, 1987) point to the persistent
decrease in term birth weight with increasing altitude, despite the multiple
generations that these populations have had to adapt to the hypoxia of high
altitude. These researchers use this persistent growth restriction with
increasing altitude across all populations to suggest that hypoxia-related
IUGR is not indicative of an insult but is rather a developmental adaptation to
hypoxic conditions.
It is argued that small body size at birth could be adaptive at high
altitude and small body size in utero is constitutive to the fetus throughout its

lifespan, persisting through infancy, childhood, and into adulthood (Brodsky
and Christou, 2004). Both oxygen and nutritional requirements can be
reduced with a smaller body size at high altitude (Thomas, 1976; Frisancho
et al., 1973; Haas, 1976). In this way, IUGR could be adaptive for the infant
living in hypoxia.
This argument has been made largely on the basis of a few
epidemiologic studies which have shown that babies born with low birth
weight associated with high altitude do not show as great an increased risk
for neonatal or infant mortality and morbidity as do babies born with IUGR
due to other causes. On the other hand, proponents of the hypothesis that
hypoxia-related IUGR is harmful point to still other epidemiologic studies that
show there is an increased risk for neonatal and infant mortality and
morbidity at high altitude. Therefore, it is worth reviewing the studies cited in
support of these opposing hypotheses to determine if there is conclusive
evidence for either viewpoint.
2.8 Epidemiologic Evidence Regarding Mortality and Morbidity at High
2.8.1 Rocky Mountains
The Rocky Mountains have been continuously occupied year-round
for less than 500 years. Most of the population studied in Colorado (the state

with the highest elevation at which persons live in the Rocky Mountains) has
lived at high altitude for less than that, suggesting that extensive genetic
adaptation to the hypoxic conditions is unlikely. The first study of fetal growth
at altitude (Lichty et al., 1957) examined births in Leadville, Colorado (3100
meters) and found that babies born at high altitude were about 1.8 times as
likely to die in the neonatal period compared to babies born at 1600 meters
(41.6 versus 23.4 deaths/1000 live births respectively). Unfortunately, babies
born small due to constitutional factor or prematurity were not excluded,
individual birth weights were not adjusted for gestational age, and morbidity
data were not reported.
McCullough et al. (1977) found that there was no difference in mean
gestational age at high (>2740 m) versus low (Denver, Colorado; <2130 m)
altitude, but high altitude neonates had 1.6 times the risk of mortality
compared with low altitude neonates. Interestingly, the mortality rates for
both high-altitude Colorado (18.5 deaths/1000 live births) and Denver (11.9
deaths/ 1000 live births) were noticeably lower than those reported twenty
years earlier. This reduction was speculated as due to increased availability
of medical technology for improving the survival of low birth weight babies.
When the analysis was restricted to term births only, babies born at high
altitude had 1.5 times the risk for mortality, but the sample size was small

enough that the power to find a significant change was lost. Once again, no
morbidity data were reported.
Finally, Unger et al. (1988) studied Colorado high- altitude (>2740m)
and compared it to Colorado low-altitude (<2130) births and found that for all
births, without excluding preterm deliveries, there was no change in neonatal
or infant mortality. Babies born with low birth weight at high altitude had less
risk of mortality than their low-altitude counterparts, but the high-altitude low
birth weight babies still had much higher mortality compared with normal birth
weight babies (no specific mortality rate, odd ratios, or standard errors were
given but the relationship was shown graphically). Thus, the authors argue
there is no survival advantage at high altitude for low birth weight, but
hypoxia-related IUGR may be less severe than other IUGR related factors in
terms of the impact on neonatal or infant survival. Interestingly, within a 10
year period (1971-1981) infant mortality for all births above about 2700
meters was reduced to a third of the previous rate, and low birth weight births
resulted in a reduction in infant mortality to a quarter of its previous rate. The
authors suggested this was due to the increase in actual births at high
altitude, but this seems unlikely to change the mortality rates. The more
likely argument the authors present is the increased access to health care
that occurred through the 1970s improved infant mortality risk. Although the
authors do not suggest it, it may be that infant mortality at high altitude may

have been attenuated by access to health care, but there may still be
detrimental health consequences to the infant at high altitude, not reported in
these data, as morbidity data are not available from vital statistic records.
2.8.2 Andes
As mentioned above, evidence shows human inhabitation of the
Andean Altiplano was likely possible 4,000-6,000 years ago (Niermeyer et
al., 2001). This is likely enough time for genetic adaptation to have occurred
if the selective pressure of high altitude is high enough. Some mortality data
suggest a significant pressure exists. Mazess (1965) studied Peruvian
populations at high (Nunoa; 3030 meters) and low altitude (338 meters), and
found neonates born at high altitude had 1.8 times the risk of dying
compared with neonates born at low altitude (high altitude mortality rate of
52.8 deaths/ 1000 live births compared with low altitude mortality rate of 28.6
deaths/1000 live births). Gestational age was not taken into account in this
analysis, so these mortality rates included preterm and growth-restricted
Beall (1981) also studied neonatal mortality in Peru, looking at births
in the cities of Puno (3860 m) and Tacna (600 m). Beall concluded that the
low-altitude (Tacna) high rate of infant mortality among low birth weight
babies was much less than the rate of infant mortality among low birth weight

babies born at high altitude (Puno). Thus, Beall concluded, low birth weight
enhanced neonatal fitness at high altitude. There are four main problems
with this conclusion. First, Bealls data are imprecisely presented: No
specific mortality numbers were presented, but only a comparison of
mortality rates across birth weight groups between the two cities in graphic
form was shown (see figure 2.3), thus making comparison to other literature
extremely difficult. The author uses this graph to conclude that growth
restriction enhances fitness at high altitude. While simply reporting mortality
data graphically is not a methodological problem, it does obscure the
interpretation of the data. For instance, variance is not given, nor results for
tests of significant differences.
Second, the calculation of mortality here may be in error because the
numerator and denominator of the mortality equation seem to be
inappropriate. Infant mortality here is the number of deaths among infants
admitted to or receiving healthcare at a hospital divided by the number of live
births occurring in the same hospital. This could be confounded by not
including deaths occurring during delivery, loss of follow-up for babies whose
mothers changed health care location, or loss of follow-up for babies whose
mothers who did not return to the hospital to report a an infant death.
Third, the mortality rate for Tacna is higher across almost all of the
birth weight categories. This is interesting when combined with Bealls report

that many more infant deaths occur in the hospital in Tacna compared with
Pufio (67% and 43% of infant deaths occurring in a hospital, respectively),
suggesting there may be some etiologic difference across all infant deaths.
This could possibly include an unexplained insult that drastically increases
infant mortality across all birth weights in Tacna, or differential use of health
care in the two cities.
Fourth, there are more deaths in Puno in the >4500 g category but
this is likely based on a very small sample size: The Puno sample consisted
of 72 infant deaths for all birth weight categories, so it is likely the >4500 g
spike in infant mortality was caused by one infant death.
Finally, and most importantly, no adjustment was made for length of
gestation. In this setting, the mortality impact of prematurity is likely much
greater than IUGR. Thus, if Tacna infant deaths are due to premature birth
(which is more severe than any kind of IUGR), then it is likely that there
would be more deaths at low birth weights. Thus, it would not be appropriate
to make conclusions based on the data presented by Beall about the
adaptiveness of low birth weight at high altitude because infant mortality in
the control group was influenced by some other unmentioned factor.

Haas (1987) compared early neonatal mortality in Mexico City15 to that
of Santa Cruz, Bolivia (300 m). Births unadjusted for gestational age in
Mexico City had 0.7 times the risk of early neonatal death compared with
births in Santa Cruz. This study was good in that it also examined term
births and divided these up into small for gestational ages categories, and
even further divided this category into symmetrical growth restriction (likely
due to constitutional factors) and asymmetrical growth restriction. Among
asymmetric growth-restricted births, Mexico City infants were at 0.6 times the
risk of early neonatal death compared with Santa Cruz infants (statistical
significance of differences between groups not given). Both cities showed
that infants with disproportional growth restriction had an increased risk for
early neonatal death compared with appropriate for gestational age term
births. There are two main problems with this study; the first is that, unlike
the studies discussed above wherein neonatal mortality is calculated as
deaths within the first 28 days of life, this study only measured deaths during
the first 2 days of life (likely due to the difficulties in following neonates after
they leave the hospital). This cannot comprehensively estimate survival
across all of the vulnerable period of infancy. Therefore, these mortality data
are not true neonatal mortality estimates. Secondly, the altitude of Mexico
15 This study considered Mexico City high altitude, yet Mexico City is usually cited at an
altitude of 2240 m and is located on the Central Plateau of Mexico, which ranges from 1500
m to a maximum of 2440 m, none of which reaches the 2500 m which is considered high

City (2240m) may not be high enough for the effect of hypoxia on fetal
growth, and subsequent mortality, to be evident. Finally, this study compares
two very distinct populations separated ethnically, economically, and
geographically. Therefore, it is hard to say whether the fetal growth and
mortality differences can be attributed solely to altitude differences.
Another study in Bolivia (Keyes et al., 2003) found that IUGR was
much higher at high altitude (16.8% at high altitude compared with 5.9% at
low altitude), and a corresponding increased risk for newborn respiratory
distress. Additionally, the intrauterine mortality rate at high altitude was
about 12 versus 7 births per 1000 births at low altitude (Keyes et al., 2003),
but neonatal mortality rate was not calculated.
Bennett et al. (in press) further analyzed the Keyes data set and
showed that the degree of Andean ancestry (as identified by maternal and
paternal surnames) was positively correlated with birth weight, with paternal
parentage contributing more to the birth weight effect, likely though
imprinting. Babies of European ancestry were three times as likely to be
born SGA compared with babies of Andean ancestry. Bennett et al. also
suggest that the similarity in gravidity but reduced parity found in Europeans
compared with Andean women may indicate that there is more fetal death in
European pregnancies.

niRTHWFiaMT (rtMQl
Figure 2.3 Figure 2 (p. 211) from Beall (1981).

Table 2.1 Review of the literature (adapted from Moore, 2001).
Altitude Difference Birthweight Difference Neonatal/lnfant Mortality Rate Difference Neonatal/lnfant Mortality Risk Neonatal/lnfant Morbidity Difference
Rocky Mtns
Lichty et al., 1957 1500m -380g 18.2 1.8 ?
McCullough et al., 1977 >610m -204g 6.6 1.6 ?
Unger et al., 1988 >610m -177g 0.5 1.1 ?

Mazess, 1965 2692m ? 0.9 ?
Haas et al., 1987 1800m* ? -0.7 0.8 ?
Beall, 1981 3260m -223 ? ? ?
Zamudio et al., 2007
Andeans 3200m -236g ? ? ?
Europeans 3200m -419g ? ? ?

Wiley, 1994 Not available, but HA mortality rate 144,
or more than 5x any LA pop reported in this table

2.8.3 Himalayas
As mentioned above, the Tibetan plateau is estimated to have been
occupied as early as 25,000 years ago (Sensui, 1981; Zhimin, 1982; Zhang
and Li 2002; Aldendorfer, 2003). Thus, there has been ample time for
adaptations to high altitude to have accumulated in this population. Yet,
there have not been many studies to examine infant mortality related to
IUGR in this region. Wiley (1994) studied a population in Ladakh, India
(3500 m) and found that the neonatal mortality rate there was 144 deaths for
every 1000 live births. While there was no comparison with a low altitude
population, this neonatal mortality rate is extremely high, or about 3.5 times
that of the highest reviewed here thus far (Lichty, 1957: high altitude
population) or over 5 times that of the largest mortality rate of any low altitude
population reported here (Mazess, 1965). These birth weight data were not
adjusted for gestational age, which would help distinguish mortality due to
prematurity versus restricted fetal growth (nor were the birth weights adjusted
for gestational age in Lichty, 1957 nor Mazess, 1965), but Wiley did show
that both neonatal and infant mortality were inversely linked with fatness
(skin folds) and body growth (calf circumference). Thus, it would seem that
at least some of the mortality is likely due to growth restriction in utero. A low
altitude Ladakhi-like comparison population would be desirable to control for

socioeconomic and/or nutritional differences that could be affecting this high
mortality rate.
Wileys study does address infant morbidity at high altitude. As
mentioned above, with increased access to medical technology, mortality
data tell less and less about the unmitigated effect of IUGR on health. To
find evidence of stressors, it is necessary to examine health outcomes that
do not result in death. Accordingly, Wiley (1994) showed that high altitude
babies had an increased risk for respiratory disease. Unfortunately, it was
not possible to test an association of respiratory disease with birth weight
because the health and birth records were unlinked.
As shown, the epidemiologic studies done at high altitude have
shortcomings in that they often do not account for gestational age, do not
have appropriate control comparison groups, do not draw unbiased
conclusions from the data, and do not present morbidity data across the
lifespan. Most of the weaknesses in these studies were unavoidable due to
the limited nature of the data available at the time, particularly in the case
sparse or unlinked health records. Thus, studies on hypoxia-induced IUGR
done up to this point do not provide a conclusive answer to the question of
whether growth restriction is adaptive or maladaptive at high altitude.
Fortunately, health records have improved in many of the possible study
populations, knowledge has expanded regarding the important factors to

consider, and technology has been developed to aid in the study of fetal
growth at high altitude. Therefore, I propose to use an alternate approach
which takes advantage of relatively new innovations in genetics, engineering,
and statistics, to look for signs of natural selection.
2.9 Role of the Uterine Artery
Proper fetal growth requires the efficient delivery of nutrients including
oxygen, glucose, lactate and amino acids (McMillen et al., 2001). Without
these, fetal growth is impaired. Environments low in oxygen, such as the
hypoxic environment found at high altitude, present a challenge for both
maternal and fetal health during pregnancy. In order to maintain proper and
sufficient development of the fetus, pregnancy demands extensive maternal
physiological changes. Many of these physiological changes are triggered
by the implantation of the trophoblast and subsequent placentation. Primate
pregnancy is unique in the extensive invasion of trophoblast cells into the
maternal vessels in the placenta (hemochorial placentation) (Rockwell et al.,
2003). Uteroplacental vascular resistance falls as the result of the physical
enlargement of the end-terminal vessels (spiral arteries), as well as the
erosion of the vascular smooth muscle surrounding those vessels, thus
causing a loss of vasoreactivity (Pijnenborg et al., 2006). Upstream from the
spiral arteries, the radial arteries, uterine arteries and veins (those vessels

involved in uterine circulation) show substantial vessel-wall growth (Keyes et
al., 1997), reduction in vasoconstriction response (Nelson et al., 1995), and
increase in vasodilatory response to flow (Mateev et al., 2006).
This change in uteroplacental resistance is largely responsible for the
increase in uterine artery blood flow early in pregnancy, which continues to
rise throughout pregnancy (Stock and Metcalfe, 1994). The uterine artery
represents a significant part of physiologic changes with pregnancy, thus, it is
a good indicator of both the healthy adaptation of the mother and indicates
nutritional supply to the fetus, enabling fetal growth potential to be fulfilled.
The maternal system is challenged globally by pregnancy-initiated
vascular changes, which are required to support the fetus. Early in
pregnancy, when metabolic support of the placenta, uterus, and fetus are
high, the principal demands on the mother are primarily molecular, rather
than mechanical (Ciliberto and Marx, 1998). Blood volume begins to
increase in the first trimester primarily through an increase in plasma volume
that also continues to rise through pregnancy (Ciliberto and Marx, 1998),
rather than red cell mass, which increases primarily in the third trimester
(Stock and Metcalfe, 1994).
By the fifth week of gestation, cardiac output has already risen by
about 10%, and continues to rise until about week 32, after which it is either

maintained through the rest of normal pregnancy (Robson et al., 1989) or
falls somewhat, depending on posture (Ueland, 1976).
2.9.1 Pregnancy Physiology at Low Altitude
In general, in normal conditions at low altitude pregnancy provokes an
early and dramatic change in maternal physiology. In normal low-altitude
pregnancies, oxygen delivery to the fetus is facilitated by an early doubling of
the uterine artery diameter and consequent four-fold increase in cross-
sectional area (by week 21) and gradual increase in uterine artery flow
velocity throughout gestation (Palmer et al., 1992). Similarly, Bernstein et al.
(2002) found that uterine artery blood flow increases over seven-fold by week
12 of gestation.
Additionally, at low altitude, maternal vascular adjustments to
pregnancy enable an increase in uterine artery blood flow which, in turn,
fuels fetal growth. The correlation of uterine artery flow with fetal growth by
week 36 is supported by other studies of low-altitude pregnancy. For
instance, Konje (2003) found that mothers at high risk for IUGR who
subsequently give birth to lower birth weight babies, had significantly smaller
uterine artery diameters by at least week 24 and lower volumetric flows by
week 20 compared with a group at low risk for IUGR. Interestingly, lower

birth weights are correlated with lesser uterine artery flow velocity at low but
not high altitude among individuals in Colorado (Zamudio et al., 1993).
Therefore, at low altitude, pregnancy increases uterine artery flow
early in gestation through a rise in both velocity and vessel diameter.
Moreover, uterine artery flow and flow velocity have been shown to be
associated with IUGR at low altitude.
2.9.2 Physiology at High Altitude
At high altitude, where hypoxia presents a challenge for delivering
sufficient oxygen to the fetus during gestation, studies in permanent
Colorado residents also showed that pregnancy increases uterine artery
diameter, flow velocity, and, thus, volumetric (measured in Leadville,
Colorado; Zamudio et al., 1995). This general trend of increase in pregnancy
is similar to that found at low altitude (see Table 2.1), yet contrastingly, at
high altitude compared with low altitude, there is a lesser increase in vessel
diameter, greater increase in flow velocity, which results in a lesser increase
in volumetric flow (Zamudio et al., 1995). Pregnancies with hypertensive
complications at high altitude showed increased uterine artery flow velocity
early, with no further increase through the rest of gestation, which resulted in
no increase in volumetric flow in the third trimester (from week 24 to 36;
Zamudio et al., 1995).

Table 2.2 Previously reported comparisons of uterine artery flow
characteristics and birth weights at high altitude
Altitude (m) UA Diameter UA Flow velocity UA Volumetric Flow UA impedance to flow (Rl, PI. S/D) Birth Weight Difference (low- high altitude) Citation
USA Low attitude residents compared with high altitude residents 1600 vs 3100 Low alt increase > Hiah alt. increase Low alt. increase < Hiqh alt. increase Low alt increase > High alt increase not reported 279g Zamudio et al. 1995a
Peruvian Andes. Low altitude mestizo residents compared with high altitude mestizo 4300 vs. 100 not reported not reported not reported Low alt Rl and PI > High alt. Rl and PI 371g Krampl et al. 2001
Tibetans compared with Han. both high altitude residents 3600 not reported Hiaher not reported not reported 635g Moore et al 2001
Bolivian Andes: High attitude Andean natives compared with high altitude-resident Europeans 3600 Andean > European Andean > European Andean > European Andean Rl. PI > European Rl. PI: S/D no difference 209g, after adjusting for gestational age, maternal height, and parity Wilson et al. 2007
Bolivian Andes High altitude Andean natives compared with high attitude-resident Europeans 3600 Andean > European Andean < European Andean > European not reported 323g after adjusting for maternal age. parity, infant sex. and gestational age) Zamudio et al. 2007

Indeed, high-altitude Tibetans, compared with high-altitude resident
Europeans or Han show higher uterine artery flow velocity, and greater
UA/CI flow velocity ratio (Table 2.1), likely indicating a greater redistribution
of blood flow to the uterine artery (Moore et al., 2001). Unfortunately, due to
the lack of imaging technology, these Tibetan studies were only able to
describe velocity, and not vessel diameter. Thus, volumetric flow data are
not available for the Tibetan studies.
Among Andean populations, Krampl et al. (2001) compared high- to
low-altitude pregnant women and found that Doppler-derived indices of
uteroplacental vascular resistance were lower at high altitude compared with
at low altitude. Additionally, we (Wilson et al., 2007) found Andean
compared with European high-altitude residents had almost 60% greater
uterine artery blood flow, resulting 1.6 times more uteroplacental oxygen
delivery at 36 weeks of gestation.
Contrastingly, Zamudio et al (2007) found no difference in oxygen
delivery between women of Andean ancestry and women of European
ancestry at high altitude at week 38 of gestation, but Andean women did
show greater uterine artery blood flow compared with women of European-
descent at high altitude. Because the Zamudio paper does not give absolute
values, but presents only tables, it is difficult to compare the flows between
papers, but it appears that both Zamudio and Wilson demonstrate an

ancestry effect on uterine artery blood flow at high altitude, with high -altitude
Andeans showing a significantly greater uterine artery blood flow in late
pregnancy. The difference in findings on oxygen delivery (determined using
uterine artery blood flow, arterial oxygen saturation, and hemoglobin
measurements) appears to be due to hemoglobin measurements, as arterial
oxygen saturation was not different between ancestry groups in both papers.
Hemoglobin15, on the other hand, was found by Wilson et al. (and Vargas et
al., 2007) not to differ between ancestry groups at high altitude, while
Zamudio et al. appear to have found hemoglobin to be increased in
Europeans compared with Andeans (no p-value is discussed for the
comparison of ancestry groups at high altitude only, but the mean values with
S.E M. suggest a significant difference). Thus, the difference in conclusions
between the two papers with regards to oxygen delivery seems to be entirely
based on differences in hemoglobin data.
Interestingly, while Zamudio et al. did seem to find a difference in
hemoglobin, they found no difference in hematocrit between ancestry groups.
This suggests either Andeans appear to have lower mean corpuscular
hemoglobin (an indicator of microcytic anemia such as that caused by iron
insufficiency) or there may have been a methodological problem in analyzing
hemoglobin. If the difference in hemoglobin between ancestry groups at high
15 Hemoglobin, hematocrit, and blood volume all were found not to differ between ancestry groups
(Vargas et al., 2007).

altitude is real, it would be interesting to look at iron status and prenatal
vitamin use between these two groups to see if the difference in hemoglobin
is the result of an endogenous factor such as ancestry, or exogenous factors
such as vitamin intake, diet, SES, etc.. Unfortunately, the issue of
hemoglobin and hematocrit differences was not discussed in this paper.
In Zamudio et al.s (2007) comparison of high-altitude compared with
low-altitude groups at 38 weeks, altitude decreased uterine artery flow and
oxygen delivery in both ancestry groups. Our group (Julian et al. in press)
has found the opposite: Altitude nearly doubles uterine artery flow in
Andeans at high altitude compare with Andeans at low altitude at week 36.
Additionally, while Zamudio et al. found no correlation between uterine artery
measurements and birth weight in any group except for high-altitude
European descendents, Julian et al. find a correlation between uterine artery
blood flow and birth weight in both ancestry groups at high altitude
These differences between our and Zamudio et al.s studies deserve
further examination. The difference could be due to several factors, but the
most likely is probably the time at which measurements were taken. While
Julian et al. have longitudinal measurements through the last half of
pregnancy, Zamudio et al. studied women about one week before delivery, at
38 weeks, which may be too late to find precursors or causal factors of
reduced fetal growth. Reduced uterine artery blood flow at 38 weeks could

be indicative of reduced fetal demand due to small size. Alternatively, the
difference could be due to sampling variation. Zamudio et al. (2007), unlike
previously-published studies (e.g. Zamudio et al., 1993; Moore et al., 2001),
found that Europeans had smaller babies at all altitudes compared with
Andeans. Due to the large-scale record review methods of the previous
studies compared with the smaller sample sizes found in physiologic studies
such as in Zamudio et al. (2007), it is possible that the low-altitude group in
Zamudio et al. represented a different sub-section (different SES, nutritional
status, stress level) of the population.
In summary, while there is debate as to the predictive value of the
uterine artery for healthy pregnancy, the majority of studies suggest that the
uterine arterys ability to supply nutrients to the placenta and the fetus is
imperative for fetal growth. Thus, I hypothesize that increased uterine artery
blood flow is necessary for the preservation of fetal growth at high altitude.
2.10 Genomic Approaches to Studying Natural Selection
The advent of sexual reproduction prompted an explosion of
phenotypic variation and enhanced the opportunities for evolution through
natural selection. In such a process, environmental pressures and limited
resources lead to differential reproductive success, with certain phenotypes

possessing selective advantage. The phenotype of an organism is the result
of the organisms genotype, environmental influences, and the genotype-by-
environment interactions. While evolution acts through the differential
selection of phenotypes, only the genotypic contribution to the phenotype
(not the environment) is heritable. Advantageous genotypes (the heritable
portion of variation) that contribute to advantageous phenotypes are
propagated through natural selection These genotypes become more
prevalent in subsequent generations. Thus, in order to find the effects of
natural selection on phenotype, the challenge lies in teasing out genetic from
environmental effects.
Environmental conditions are not constant across the globe. Thus,
within a species, phenotypic variation is generally distributed clinally due to
the typically gradual change in environmental conditions over space. In
humans, such geographic differences between populations exist, but
because of the extreme mobility and interbreeding of human populations,
more variation exists within rather than between populations (Bamshad et al.,
2004). Nonetheless, there are some environmental challenges or selective
pressures which are strong enough to cause high levels of evolution in
certain adaptive traits (according to the definition of evolution as a change in
allelic frequency in a population over time).

Phenotypic variation among individuals can be utilized to find
information about the genes responsible for a specific phenotype. There are
two ways of doing so: association and linkage disequilibrium studies.
Linkage disequilibrium resulting from three possible sources is
generally used in genetic analyses. The first source, called background
linkage disequilibrium refers to the tendency of two loci to be inherited
together because of their close physical proximity in the genome (Montana
and Pritchard, 2004). This is also referred to as hitchhiking.
The second source, called mixture linkage disequilibrium, refers to the
fact that the genome of an individual will reflect a small part of the
populational gene pool from which it came (Falush et al., 2003) Thus, all of
the alleles that exist in an individual's genome must have existed somewhere
within the population (except for the few cases where mutations have
occurred). Mixture linkage disequilibrium is often utilized in the assessment
of individual ancestry (e.g. Pritchard et al., 2000). Mixture linkage
disequilibrium is not generally informative for gene mapping because it
describes markers across the entire genome My study will utilize this type of
linkage disequilibrium to confirm population ancestry.
The third and final source of linkage disequilibrium is called admixture-
linkage disequilibrium. This occurs when there are segments of DNA with
marker alleles recognizable as having been inherited from a certain ancestral

population (Falush et al., 2003). In population mapping, one looks for
association between ancestry-specific markers and the phenotypic trait of
interest. It is expected that the trait should most often be found together with
markers specific to the ancestral population in which the trait has the highest
incidence (Hoggart et al., 2004).
Hence, variation propagated through natural selection has aided in the
effort to map genes, and gene mapping has furthered our understanding of
evolution and the process of natural selection.
2.10.1 Genes Provide Evidence of Evolution
Genes and, more specifically, their population frequencies can be
used to show evidence of evolution. As mentioned above, the unit upon
which natural selection acts directly is the holo-organism or the individual.
Yet, the individual and his/her resultant fitness are the product of both the
genotype and environment, with selection acting on those individuals whose
alleles enhance fitness. Thus, although natural selection does not act
directly upon genes, it affects the frequencies of alleles in a population.
Cavalli-Sforza (1966) originally suggested that natural selection may
become evident by investigating the amount of population variance and allele
frequency. Lewontin and Krakauer (1973) refined this idea, suggesting that
natural selection through the selection of specific advantageous alleles,

differs from other causes of allelic frequency change in populations
(mutation, gene flow, and drift) because only natural selection is directional.
Currently, several tests for evidence of natural selection at the
genomic level have been developed. The neutral theory of molecular
evolution posits that most mutations are selectively neutral (e.g synonymous
mutations, in noncoding regions, etc.). Most of the tests for selection, then,
typically use the neutral theory of molecular evolution to predict the null
hypothesis (Nielsen 2001). One type of test for neutrality takes into
consideration the allelic distribution and levels of population variability, as
suggested by Cavalli-Sforza (1966); such tests include the Wright statistic or
Fst (Wright 1951), LSBL (Shriver, 2004), InRH (Kauer et al., 2003;
Schlotterer and Dieringer, 2004), Tajimas D (1989), Fu and Lis D and F
(1993), and Fay and Wus H (2000) (see section 4.2 for further discussion).
Hence, gene-searching tools and mapping have been used to inform
our knowledge of evolution and the process of natural selection.
Reciprocally, variation propagated through natural selection has aided in the
gene-mapping endeavor. Both of these burgeoning areas have created
much interest and promise to challenge scientists in biology, informatics, and
anthropology to exciting new collaborations.

3. Research Design and Methods
3.1 General Approach
This study consists of three main parts: 1) a genomic scan to identify
genes that show evidence of natural selection by comparing high- with low-
altitude natives, 2) a test of associations between genotype at naturally-
selected genes and fetal growth phenotype, and 3) the interpretation of
results in relation to the controversy over the adaptive consequences of
hypoxia-related fetal growth restriction.
Multigenerational high-altitude residents in Bolivia provide a valuable
study opportunity, having lived at high altitude long enough for selection to
have occurred (see section 2.4), being relatively accessible, and including
groups of similar ancestry residing at both high and low altitude.
Additionally, the opportunity to conduct this study was available with
the assistance of grants for an already-existing large study funded by the
National Institutes of Health (HL079647; HL060131; TW001188; Principal
investigator Dr. Lorna G. Moore) and the National Science Foundation
(Graduate Research Fellowship). The NIH grants provided funding for the
extensive physiological studies and genomic analysis done by collaborators
at Pennsylvania State University (Dr. Mark Shriver and Abigail Bigham)

These collaborators provided the results of the SNP-chips and computed the
test statistics described. All subsequent steps involving testing for phenotype
associations of physiologic traits were done by me.
The study design for the genetic analyses was comparative, and
comparative and longitudinal for the physiologic aspects. The genetic
analyses (aim 1) were comparative studies in subjects of Andean ancestry
and a selection of low-altitude ancestry control groups. Pregnancy
phenotype data were longitudinal (aim 2), obtained by studying women at
weeks 20, 30, 36 of pregnancy and 3 months postpartum for a measurement
in the non-pregnant state. Finally, testing for the adaptive fetal growth
phenotype (aim 3) was conducted through comparisons between Andean
genotypes and birth-weight phenotypes and/or uterine artery blood flow
Subjects for aim 1 were recruited from the South American Andean
Altiplano in both Peru and Bolivia. These were compared to low-altitude
control groups chosen to represent the range of possible source populations
for the native Andeans, including a Native American group (Nahua and
Maya), Asian (Han Chinese), European, Indian, and African American
groups. Subjects for aims 2 and 3 were pregnant Andean women recruited
from La Paz, Bolivia, which is situated in the Andes of South America. The
residential areas of La Paz range from about 3,600 to 4,000 meters.

Table 3.1 Study samples
Region of Ancestry Population Altitude Name of Group Sample Sizes
Aim 1
South America Amerindian High Peruvian Quechua 71
Amerindian High Bolivian Andean 94
Amerindian Low Peruvian Quechua 71
Amerindian Low Bolivian Aymara 94
Central America Amerindian Low Guatemalan Maya 40
Amerindian Low Mexican Nahua 29
Amerindian Low Trinidad Amerindian 50
Asia Chinese Low Han Chinese 94
Africa West African Low West African, living in US 50
Europe European Low European, living in US 50
Aim 2 and 3
South America Amerindian High Bolivian Andean 45
3.2 Aim 1
Out of a group of prospective candidate genes chosen for their implicated
role in oxygen transport and/or fetal growth, identify genes that show
evidence of directional selection in persons of indigenous, high-altitude
Andean ancestry by
A. examining genomic regions that differ significantly in high- versus low-
altitude populations of both Bolivia and Peru (done by collaborators),
B. determining the most frequent genotypes in the high-altitude
populations, and

C. identifying the gene families to which these most-freguent genotypes
3.2.1 Rationale
Previous studies have shown that Andeans residing at high altitude
are at a reduced risk for having SGA babies compared to relative high-
altitude newcomers (Moore et al., 2001). Additionally, Andeans have
different pregnancy phenotypes with respect to uterine blood flow (Wilson et
al., 2007) and different expression of gene products that affect endothelial
cell proliferation, vasodilation and constriction, such as VEGF, Flt-1, and NO
(preliminary results). Each of these gene products is at least partially
regulated by the HIF pathway. The population differences in SGA and
uterine artery flow during pregnancy suggest genomic mapping may be
useful for identifying genes that are associated with hypoxia-related
alterations in fetal growth and uterine artery blood flow. Furthermore, the
involvement of the HIF pathway in responding to both hypoxia and
pregnancy suggests that is regulatory and targeted genes are good
prospective candidate genes for use in sifting genome-wide SNP data (see
section 2.1).
Using measures of population genomic heterogeneity (e g. Wrights
fixation index (Fst), LSBL, Tajimas D, and InRH), FIIF-targeted and

regulatory genes were identified as candidates of natural selection at high
altitude. Prospective candidates were designated as candidates for natural
selection if at least one test of population structure and one test of genetic
diversity (see below for descriptions) showed significant evidence for natural
selection Based on preliminary analyses, an alpha of 0.05 was used to
identify candidate genes.
This genomic scan resulted in identifying a subgroup of prospective
candidate genes that show evidence of natural selection and also identify an
allele at each locus more frequently found in high-altitude adapted Andeans.
3.2.2 Measurements
Genomic DNA was extracted from leukocytes in samples listed in
Table 3.1. High- altitude Andean residents included Peruvians, primarily self-
identified as Quechuan16, high-altitude residents who were able to establish
a lineage at high altitude and were also confirmed to have a low degree of
European admixture (about 2%; Wilson et al., 2007; Brutstaert et al., 2004;
see below), and Bolivians, primarily self-identified as Aymara.
16 Aymara and Quechua populations are two distinct ethnic groups, but we pooled them here
based on the geographical overlap, linguistic group connection, and mitochondrial DNA
haplogroup similarities (Bert et al., 2001).

3.2.3 Confirming Self-assessed Ancestry
Genomic DNA was used to assess genetic admixture as measured
by the respective contributions of 100 highly polymorphic short tandem
repeat genetic markers (STR) characteristic of European, African and Native
American populations17. Each SNP was amplified using polymerase chain
reaction (PCR), digested using restriction enzymes to create restriction
fragment length polymorphisms (RFLP), which were scored using a melting-
curve assay (Hybaid DASH machine) (Brutsaert et al., 2004; Brutsaert et al.,
2003). These results confirmed the self-assessed native ancestry of these
individuals, as self-reported ancestry coincided with AIM ancestry estimates.
3.2.4 Genomic Scan
Blood samples were collected from Andeans, primarily identified as
Quechuan or Aymara, high-altitude residents who were confirmed as having
a low degree of European admixture (about 2%; Wilson et al., 2007).
Individuals selected for their high degree of Andean ancestry
(determined by AIMs) were genotyped for a large scale genomic-scan using
17 The original 22-marker panel was expanded to 100 markers. Of the original 22 STRs used were.
MID-575, TSC1102055, WI-11153, MID-52, SGC30610, WI-17163, Wl-9231, WI-4019, Wl-
11909, D11S429, TYR-192, DRD2 TaqD, DRD2 Bcl\, WI-14319, CYP19, PV92. WI-7423,
CKM, MID-161, MID-93, FY, and F13B. Of these, 20 markers had highly divergent
frequencies between European and Native American populations (>30%) and two markers
were unique to African populations For a more detailed description of the expanded panel,
see Bigham et al. in press.

an Affymetrix 500k chip. If the chip contained SNPs evenly distributed
(Nicolae et al., 2006) throughout the genome (in reality some regions, such
as telomeric regions, were underrepresented), an average gene of about 10
to 15kb (Strachan and Read ,1999) would contain at least one locus
genotyped by the chip. In reality, the chip is typically skewed with more
SNPs in coding regions, increasing the likelihood that a locus within each
gene was typed. Thereby, with total genome coverage, many loci other than
the HIF prospective candidates can be screened for linkage.
3.2.5 Measures of Population Genomic Heterogeneity Wright Statistic
Genetic variation of populations can be generally compared using the
Wright statistic, Fst (Wright, 1951). The Wright statistic was developed for
comparing two subpopulations within a given population, but can also be
used for any within-species comparison because frequencies within the
whole species may be used as the total population value. Fst is calculated
using the heterozygosity of the sub- (Hs) and total- populations (HT)
according to the equation:
Fst = (Ht-Hs) (3.1)

This statistic can be alternatively calculated by using allele frequencies in
place of heterozygosities (Bamshad et al., 2004). A Fst of 0 means that
there is no difference between allele frequencies in the two subpopulations
and an Fst of 1 means that the alleles in the two subpopulations are mutually
If Fst is governed by drift or gene flow alone, then all loci will be
affected equally (Lewontin and Krakauer, 1973; Akey et al., 2002) and in a
predictable manner18 (Cavalli-Sforza, 1966, Akey et al., 2002). Alternatively,
if Fst is governed by natural selection (which we will say here selects
unidirectionally for the sake of simplicity), then the locus under selection will
have a different Fst compared with other loci (unlinked) in the genome not
under selection (Akey et al., 2002).
This application of genetics to anthropological study has heretofore
been a pipedream, discussed theoretically but deemed unfeasible (e.g.
Cavalli-Sforza, 1966; Lewontin and Krakauer, 1973). With the recent
construction of dense marker maps for the human genome and methods for
assessing genotypes for many markers relatively quickly, using the Wright
statistic to find naturally selected genes has become possible. To do this,
the Fst statistic for many alleles can be calculated and compared across the
18 For instance, if both populations being compared have similar rates of genetic drift, Fst will
remain the same (close to 0) or if there is gene flow into one population, Fst will increase
similarly across all loci

genome and between populations. Those that deviate from the normal
distribution of Fst values are then, by definition, candidates as possible
targets of natural selection (Goldstein and Chikhi, 2002).
As an example of such an approach, Akey et al. (2002) did a genome-
wide scan using three populations and identified 174 candidate genes, 17 of
which are known disease genes which negatively affect fitness.
Contrastingly, Wilson et al. (2001) showed how this technology could be
applied by the pharmaceutical industry for examining the allelic frequency
variations in loci related to drug metabolism. In this study, a panel of 35-45
markers was genotyped in eight different populations from all over the world.
Four of six of the candidate genes had significantly different Fst values.
Thus, it has become evident that using the Wright statistic can provide a
means for substantiating (or refuting) previous assertions regarding the
mechanisms by which natural selection functions in humans, as well as
providing new genome targets for study. Locus-Specific Branch Lengths
The Wright statistic, Fst, is limited by the need to compare one
subpopulation to a larger, encompassing population. This method of
calculating genetic divergence is a bit deceptive in that the total population is
made up of several subpopulations, each of which also has natural selective

pressures acting upon their genomes. Thus, Fst may be misleading because
a high Fst for the population of interest may simply be indicative of a high
degree of divergence by one or more the other subpopulations (Shriver,
2004). In other words, a group used as an outgroup comparison for the
population of interest may have evolved its own suite of adaptive genes and,
therefore, cause erroneous identification of naturally selected regions.
In order to describe divergence in the subpopulation of interest, a
locus-specific branch lengths (LSBLs) measure has been developed by our
collaborator, Dr. Mark Shriver (Akey et at., 2001; Shriver et al., 2004). The
LSBL method uses the pairwise difference in Fst at a particular locus, as
demonstrated below:
Population A
Population B Population C
Figure 3.1 Schema representing branch lengths between three different
populations (figure from Shriver et al., 2004)

dAB = the Fst pairwise distance of populations A and B
dAc = the Fst pairwise distance of populations A and C
dBc = the Fst pairwise distance of populations B and C
x= (dAB + dAc dec)/2
y= (dAB + dBc dAc)/2
z= (dAc + dec dAB)/2
where x, y, and z represent the branch lengths of population A, B, and C,
respectively (Akey et al., 2001; Shriver et al., 2004).
This calculation treats each designated subpopulation individually with
respect to the other populations, and gives an LSBL value for each
subpopulation. Thus, by comparing branch lengths, one may discern
whether the population of interest is indeed the population undergoing
natural selection. Natural Log of the Ratio of Fleterozygosities
While LSBL identifies candidate gene regions for natural selection by
searching for differences in allele frequencies between adapted and
nonadapted populations, the natural log of the ratio of heterozygosities
(InRH) identifies candidate genes through searching for evidence of selective
sweeps within a population (Kauer et al., 2003, Schlotterer and Dieringer,
2004). That is, genomic regions with low heterozygosity compared with the
rest of the genome. Natural selection should reduce heterozygosity, or

variation, at loci that affect reproductive success, while loci not under
selective pressures should remain relatively heterozygous, or variable
(making the neutral expectation).
Because low heterozygosity may be due to population bottlenecks,
drift, low levels of mutation, or natural selection, InRH uses the whole
genome to control for expected levels of heterozygosity at nonselected sites
(usually silent loci). In order to control for locus-specific mutation rates, the
heterozygosity of the population of interest is divided by heterozygosities of a
similar population. This gives an equation for InRH of:
The empirical distribution of the InRH for all loci is normally distributed. Thus,
loci that are significantly less heterozygous than the average locus (the
negative tail of the distribution) are considered candidates for natural
Where H = 1
(Ohta and Kimura, 1973)
Wl + 20,
88 Tajimas D20
Tajimas D test (Tajima, 1989) compares 020 21 (Watterson, 1975)
expectations based on average number of nucleotide differences (or per
nucleotide heterozygosity) and 0 calculated from the number of segregating
sites (or number of nucleotide positions at which a polymorphism is found).
The formula for calculating Tajimas D is as follows:
D= 9n 0k (3.3)
Son 0k
Where 0^ is calculated based on number of pairwise differences or
heterozygosity, 0k is calculated based on number of segregating sites22, and
Sen is the standard error of 0^.
Therefore, under neutral variation (null hypothesis), one would expect
that Tajimas D should be 0 because expectations based on heterozygosity
Tajimas D uses the infinite allele model of mutation, which assumes new mutations
always occur at a different site. This assumption is only appropriate if the two sequences
being compared are closely related enough to expect mutations at different locations
21AIso called the Watterson coefficient, 0 = 4Nep or 4*effective population size*mutation
rate per generation. In Wattersons (1975,p. 258) words, it is the mean number of new
mutant sites for the population, per generation, which can be used to find either the
mutation rate at each site or the proportion of mutating sites in a population
22 0k can also be defined as k/an, where k is the number of segregating sites and an is
I (l/k).

versus expectations based on number of segregating sites should be the
same. In, or shortly after, a population bottleneck, Tajimas D would be
greater than 0 because the heterozygosity measure Gjt would not change
(moderate to high frequency alleles that make the greatest contribution to
heterozygosity are not likely to be reduced unless the bottleneck is severe
and prolonged), while the number of polymorphic positions would decrease,
thereby reducing 0k. Under both purifying selection and population
expansion, one would expect Tajimas D to be less than 0. In purifying
selection, heterozygosity, and thereby would be reduced. In situations of
population expansion, Tajimas D would also be less than 0 because the
number of segregating sites would not be lost (larger 0k), but mutations
would not be common enough to increase heterozygosity (smaller 0^).
A disadvantage of Tajimas D is that it relies heavily on the
assumption of no population structure and constant population size. Thus, it
is difficult to reject the null hypothesis even if Tajimas D is statistically
significant from 0. Therefore, Tajimas D is rarely used alone, but, rather,
used to triangulate selected genes.
These measures of population genomic heterogeneity were calculated
for the most complete identification of genomic regions that are likely
candidates for natural selection. This genomic scan resulted in identifying
SNPs within or close to the HIF- prospective candidate genes with differing

allele frequencies in Andeans.
Candidate genes were chosen that showed significant signs of natural
selection by at least one test of population structure (LSBL or InRH) and the
test for genetic diversity (Tajimas D). Thus, regions were identified that
show evidence of both significant difference in allele frequency and in genetic
linkage, while controlling for the bottlenecking Andean Amerindians have
likely experienced23.
3.3 Aim 2
Determine if the presence of the high-altitude alleles (1B above) is
associated with
A. normal fetal growth, as judged by fetal biometry and birth weight in
relation to gestational age,
B. greater pregnancy-associated increase in maternal uterine artery
C. greater near term, maternal uterine artery blood flow.
23 Andeans, as are all Native American groups, are estimated to have experienced extensive founder
effects with the migrations into the Americas (which similarly affects genomic heterogeneity as
population bottlenecks) and at least one major bottleneck after European contact and subsequent
Native American population decimation through disease (Cavalli-Sforza et al., 1994; Ramenenovsky,
1996; Novick et al. 1998; Silva et ah, 2002).

3.3.1 Rationale
Previous studies have shown that Andeans have a population-specific
response to hypoxia in pregnancy, as evidenced by higher-birth weight
babies (Moore et al., 2001), and preliminary studies showed that Andeans
exhibit greater uterine artery flow compared with European newcomers to
high altitude (Wilson et al., 2007). Aim 1 resulted in identification of a set of
loci that have Andean-specific allelic frequency profiles. Through aim 2, I
tested if characteristically-Andean genotypes are related to characteristically-
Andean phenotypes.
3.3.2 Measurements
Andean high-altitude residents of Bolivia (Table 3.1, aim 2 and 3),
were studied serially at weeks 23 (23.4 .6), 31 (30.8 2), and week 37
(36.6 .6) of pregnancy, and three months postpartum for a measurement in
the nonpregnant state in order to assess their physiologic responses to
pregnancy. The 52 subjects self-identified as being of Andean ancestry
living in La Paz, Bolivia (3800 meters) and, in aim 1, this ancestry was
confirmed using AIMs. Informed consent was obtained and study protocol
approved by the Human Subjects Review Committees of the Colorado
Multiple Institutions Review Board and the Colegio Medico, the institutional
review counterpart at the Instituto Boliviano de Biologia Altura in La Paz.