The Childhood Plasma Proteome: Discovering Its Applications in Public Health Advisor: Dr

THE CHILDHOOD PLASMA PROTEOME:

DISCOVERING ITS APPLICATIONS IN PUBLIC HEALTH

By Sun Eun Lee, M.S.

A dissertation submitted to Johns Hopkins University in conformity with the requirements for the degree of Doctor of Philosophy

Baltimore, Maryland March, 2015

ABSTRACT

Background: Child health is shaped by cumulative interactions with environments even before birth. However, our understanding of the underlying biological mechanisms remains far from complete. Plasma proteomics may offer unique opportunities to understand underpinning biological processes that respond to early nutritional exposure, reflect ongoing health conditions, or mediate health consequences. The overall goal of this thesis is to evaluate applications of plasma proteomics in discovering new bio- signatures or generating hypotheses with regard to prenatal micronutrient (MN) supplementation and childhood inflammation and cognitive function.

Methods: In 1999-2001, a double-blind randomized trial of antenatal micronutrient supplementation was conducted in the rural District of Sarlahi, Nepal. Pregnant women received either vitamin A alone as the control, or with folic acid, iron-folic acid, iron- folic acid-zinc, or a multiple micronutrient supplement containing all three plus 11 other vitamins and minerals from early pregnancy to 3 months postpartum. From 2006-2007, children born during this trial were followed up at the age of 6-8 years for plasma specimen collection and a year later for the assessment of cognitive function by psychological tests. We applied quantitative proteomics to identify proteins in plasma of

500 Nepalese children that co-vary with a plasma inflammation biomarker, alpha-1-acid glycoprotein (AGP) and general intellectual function, measured by an immunoradial diffusion assay and the Universal Nonverbal Intelligence Test (UNIT), respectively. We evaluated the effects of antenatal micronutrient supplementation by examining differentially abundant proteins and enriched gene sets by maternal MN intervention

(each compared to the control group). A subset of children (n=249) who had both

ii proteomics and psychological test outcomes were included for the analysis of child cognitive outcome.

Results: Among 982 proteins quantified in >10% of total samples, 99 were strongly associated with AGP at a family-wise error rate of 0.1%. Positively associated proteins include known positive acute phase proteins and numerous unexpected intracellular signaling proteins. Negatively associated proteins were secretory hepatic proteins involved in transporting lipids, micronutrients, growth factors and sex hormones, and extra-hepatic proteins regulating extracellular matrix metabolism. Among 751 proteins quantified in >10% of the sub-samples, 9 and 13 proteins were positively and negatively associated with the UNIT score, passing a false discovery rate (FDR) threshold of 5%, respectively. In fully adjusted models, associations of 7 proteins involved in subclinical inflammation remained significant, explaining an additional 5~9 % of variance in the

UNIT score. Lastly, there were no overall effects of antenatal micronutrient supplementation on plasma protein profiles of children. In sex-stratified analyses, maternal folic acid supplementation increased the relative abundance of nesprin-1 by 50.9

(95% CI: 24.7, 82.8) % among boys and positively enriched 4 gene sets related to cytoskeleton and organ development among girls (all passing FDR threshold of 5%).

Conclusions: Our findings suggest that a vast plasma proteome reflects homeostatic control of inflammation, low-degree chronic inflammation is an important risk factor for child development, and there was no prominent long-term effects of prenatal micronutrient supplementation on childhood plasma proteome. Further studies should be followed to evaluate future public health use of plasma biomarkers of chronic

iii inflammation and potential health consequences of subtle but enduring gene-specific and functional changes by maternal folic acid supplementation in undernourished populations.

Advisor: Dr. Keith West

Thesis Readers: Dr. Pierre Coulombe, Dr. Parul Christian, and Dr. Ingo Ruczinski, and

Dr. Yiwu He

ACKNOWLEDGEMENTS

From 2009 fall to 2015 spring, all my doctoral work at Hopkins would have been impossible without countless help and support by many people. First and foremost, I would like to offer my sincere thanks to all study participants, community field workers, and the Nepal Nutrition Intervention Project Sarlahi (NNIPS) team in the field. I had the privilege to conduct a secondary data analysis using high quality data of NNIPS-3 and its two child follow-up studies. When I visited Sarlahi District in Nepal in 2013 summer,

I realized that the high quality of NNIPS data could not be achieved without incredible amount of field work, training, and efforts made by the field workers and NNIPS team.

Without their dedication to community work and participation of community people, this study could not exist and I would not have this wonderful research opportunity. I would like to express special thank to Steven LeClerq who made me feel like I am part of the

NNIPS family during my visiting period. Also, I give thanks to all the other NNIPS team members who showed great hospitality to me. This NNIPS study site visit will remain unforgettable to me.

My thesis work was embedded in the Nutriproteomics project and I cannot thank each member of Nutriproteomics team enough. They actually did all the critical and outstanding work from designing the study, performing laboratory experiments, developing methods and analytic tools, and managing the whole project. I joined this team in 2012. Whenever I was lost in the complexity of this project, Lee Shu-Fune Wu was always supportive of me. She helped me not only with my work, but also gave me a big smile, comfort, and sincere encouragement. Dr. John Groopman served on my school-wide oral exam committee. I appreciate his intellectual guidance on this project

v and thoughtful feedback on my thesis. Dr. Robert Cole and his laboratory at the School of

Medicine performed all proteomics analysis and it was such a nice opportunity for me to learn the state-of-art high throughput-technologies he brought to public health nutrition. I am also thankful for his responsiveness to my requests and keen questions that made me think through my methodology and interpretation of findings. Dr. Kerry Schulze and her laboratory analyzed all micronutrients and inflammatory indicators, which were essential parts of my thesis. When I was stuck with some issues and no one seemed to solve my questions, she steered me in the right direction and gave me good alternative solutions. I sincerely appreciate all of their invaluable contributions to this project and always being supportive to my research.

There are three people who gave me invaluable guidance and provided me opportunities to learn and grow. I would like to express my great appreciation to Dr. Ingo

Ruczinski. Because of the nature of a large-scale data analysis, I had to apply different analytic skills from what I had learned from regular Biostatistics classes. Because I was new to a large-scale data analysis, I faced many statistic and analytic challenges.

Whenever I requested a meeting, Dr. Ruczinski made time for me and put effort to address my questions. Also, I am thankful that he taught me the true attitude a researcher should have; I challenged the results until I could not think of any other ways to reject the findings. It was a great lesson to me and this attitude will ground me in my future research.

I am thankful for Dr. Parul Christian who was my supervisor for CHERG work and is a committee member of my departmental, school-wide oral exam, and final defense exam. CHERG study with meta-analysis was a great research experience as a

vi student and she always gave me clear ideas and instructions to follow. As a main principal investigator of NNIPS-3 and child follow-up studies, she provided me critical feedback and fair and reliable support on my thesis research. I will never forget how she did not mind spending extra hours to help me to prepare my defense seminar.

I feel truly grateful to my advisor, Dr. Keith West for his endless generosity and patience, and financial and mental support. He guided the big picture of my study beyond what literature says or what other people conventionally believe. His pioneering spirit and enthusiasm for proteomics inspired me throughout my research. The countless meetings with him kept motivating me to have creative and innovative thinking. Due to his support, I could truly enjoy my thrilling and exciting research. Most of all, I would like to thank him for sending me to Bangladesh and Nepal to learn community-based cohort study. This experience became the foundation of my research and will help me to power through other public health nutrition research in the future.

This study would not have been possible without enormous financial support from the Bill and Melinda Gates Foundation and it actually supported me beyond. Dr. Yiwe He who is a program officer of the Foundation willingly agreed to serve on my external committee member and flew from Seattle to Baltimore to attend my defense seminar and final oral exam. He provided invaluable feedback on my research as funder’s perspective and it was unforgettable experience to have him in my committee. I am truly thankful for his support for my research as well as steady financial support for this project. Lastly, I would like to thank to Dr. Pierre Coulombe who served as a chair of my thesis committee.

I am thankful for his time and effort to review my thesis and invaluable biologic feedback on my research.

vii

Special thanks to essential people in the Human Nutrition and the department of

International Health who made my Hopkins life possible - Peggy Bremer, Rhonda

Skinner, Tom Durkin, Carol Buckley, and Cristina Salazar. Thank you for your willingness to help me.

I would like to thank my dear friends, Rachel Dlugash and Tatenda Mupfuze who were always there for me and listened my story. I will always cherish our time being together and this is one of the most precious memories among my student life in

Baltimore. I also would like to thank to the Dissertation Support Group members including Yun-Hee Kang, Megan Henry, Bess Lewis, Dorothy Chiu, Ramya

Ambikapathi, Myra Shapiro, and Jennifer Lam. We met every Friday to share our weekly goals and check up weekly achievements. They helped me to work through the most stressful time of writing my thesis. I would like to especially thank Yun-hee Kang for her sincere warm heart to public health and other people. Also, Moira Angel was a wonderful colleague working with CHERG and is also a good friend who always tries to help and cheer me. Also, I am thankful to other wonderful Hopkins friends, Max Barffour, Lenis

Chen-Edinboro, Amanda Palmer, Yong-ji Jo, and my awesome roommates Saki

Takahashi and Nuriesya Abubaker.

Also, I cannot forget my great mentors of college years, professor emeritus Yang- ja Lee and former advisor of Master program, Hye-young Kim. Professor Yang-ja Lee is the one who made me think that I can help other people improve their health with better nutrition. Without laboratory experience during my masters program, I would not be interested in proteomics study and without Dr. Hye-young Kim’s unconditional support, I would not study abroad for my Ph.D. degree.

viii

Lastly, I am deeply grateful to my family. My parents always believed in my pursuits and give me unconditional love. Although I have lived apart from them for 6 years, I feel that we are more strongly bonded and connected together than ever before. I always wanted to be a daughter that they are proud of. There is nothing I can give back to my parents compared to what they have given to me, but I hope they are happy to see me as more mature and independent daughter pursuing her life goals. Because I could meet these wonderful people at Hopkins who opened my eyes to see the broader world, and I feel cared and loved by my family and friends, I truly feel this Ph.D. journey was my most valuable gift.

Dedication:

To my parent, Kwang-woo Lee and Mee-sook Lee

for their faithful love

TABLE OF CONTENTS

ABSTRACT ...... ii ACKNOWLEDGEMENTS ...... v Table of Contents ...... xi List of Tables ...... xiii List of Figures ...... xiv List of Appendices ...... xv Acronyms and Abbreviations ...... xvi Chapter 1. Introduction and Specific Aims ...... 1

Chapter 2. Literature Review ...... 6 2.1 The Human plasma proteome ...... 6 2.2 Inflammation and the plasma proteome ...... 13 2.3 Child development and the plasma proteome ...... 22 2.4 Effect of antenatal micronutrient supplementation on child plasma proteome ...... 34

Chapter 3: Study Design and Methods ...... 69 3.1 Population, study design, and subjects ...... 69 3.2 Measurements ...... 75 3.3 Data analyses ...... 81

Chapter 4: A Plasma Proteome Associated with Inflammation in School- aged Children in Rural Nepal ...... 93 ABSTRACT ...... 93 INTRODUCTION ...... 95 SUBJECTS AND METHODS ...... 98 RESULTS ...... 101 DISCUSSION ...... 105 CONCLUSIONS ...... 111

Chapter 5: Exploring the Plasma Proteome Associated with Child General Intelligence in School-Aged Children in Rural Nepal ...... 128 ABSTRACT ...... 128 INTRODUCTION ...... 130 SUBJECTS AND METHODS ...... 132 RESULTS ...... 139 DISCUSSION ...... 143 CONCLUSIONS ...... 148 APPENDIX ...... 161

Chapter 6: Effect of Antenatal Micronutrient Supplementation on Plasma Protein Profiles in School-aged Children in Rural Nepal ...... 173 ABSTRACT ...... 173 INTRODUCTION ...... 175 SUBJECTS AND METHODS ...... 177 RESULTS ...... 184 DISCUSSION ...... 187 CONCLUSIONS ...... 193 APPENDIX ...... 201

Chapter 7: Conclusions ...... 215

Curriculum Vitae ...... 225

xii

LIST OF TABLES Table 2.1. Biochemical characteristics between AGP and CRP ...... 16 Table 2.2. Effect of antenatal micronutrient supplementation on growth, development, mortality, cardiovascular and metabolic health in children living in LMICs (≥12 months) ...... 38 Table 2.3. Effect of maternal micronutrient on gene/protein expression in offspring using untargeted methods in animal studies ...... 51 Table 3.1. Micronutrient supplement regimen ...... 70 Table 3.2. Subtest measures in the UNIT ...... 78 Table 3.3. Key results of Gene set enrichment analysis ...... 88 Table 4.1. Demographic, nutritional, and health characteristics of 6-8 year old children in rural Nepal (n = 500)1 ...... 112 Table 4.2. Plasma proteins positively associated with plasma α-1-acid glycoprotein (AGP) in 6-8 year old children in rural Nepal, ordered by P (FWER<0.01%)1 ...... 113 Table 4.3. Plasma proteins negatively associated with plasma α-1-acid glycoprotein (AGP) in 6-8 year old children in rural Nepal, ordered by P (FWER<0.01%)1 ...... 115 Table 4.4. Cellular localization and molecular/biological functions of plasma proteins associated with α-1-acid glycoprotein (AGP) in 6-8 year old children in rural Nepal1 ...... 121 Table 5.1. Demographic, anthropometric, health, and dietary characteristics of children with and without psychological assessments among school-aged children in rural Nepal1 ...... 149 Table 5.2. Plasma proteins positively associated with the Universal Non-verbal Intelligence Test (UNIT) score in school-aged children in rural Nepal (q<0.05), ordered by q ...... 152 Table 5.3. Plasma proteins negatively associated with Universal Non-verbal Intelligence Test (UNIT) score in school-aged children in rural Nepal (q<0.05), ordered by q ..... 153 Table 5.4. Multiple plasma proteins that explain Universal Non-verbal intelligence Test (UNIT) score in school-aged children in rural Nepal (n=241) ...... 155 Table 5.5. Association between plasma proteins and Universal Non-verbal intelligence Test (UNIT) score in the multivariate model (Model 1), adjusted for child iron status (Model 2), linear growth (Model 3), and household socio-economic status (SES) (Model 4) in school aged children in rural Nepal7 ...... 158 Table 5.6. Plasma proteins associated with Universal Non-verbal Intelligence Test (UNIT) score, after adjustment for child iron status, linear growth, and household socio- economic status in school aged children in rural Nepal (q<0.05), ordered by q .. 159 Table 6.1. Characteristics of children, mothers, and households at baseline and follow-up by antenatal micronutrient supplementation group1 ...... 195 Table 6.2. Expected difference in relative abundance of nesprin-1 (SYNE1) by maternal supplementation group relative to the control, stratified by sex in children 6-8 years of age in Sarlahi, Nepal1 ...... 199 Table 6.3. Enriched gene sets in maternal folic acid supplementation group relative to the control group among girls 6-8 y of age in Sarlahi, Nepal (false discovery rate cut off of 5%) ...... 200

xiii

LIST OF FIGURES Figure 2.1. Acute phase proteins and systematic response to inflammation ...... 15 Figure 2.2. Potential biological systems associated with child cognitive function ...... 27 Figure 2.3. Plausible mechanisms of maternal micronutrient status during pregnancy and fetal programming ...... 42 Figure 3.1. Flow chart of study participants ...... 73 Figure 3.2. Example of enrichment plot ...... 89 Figure 4.1A-C. Volcano plot of plasma proteins associated with plasma α-1-acid glycoprotein (AGP) in 6-8 year old children in rural Nepal...... 118 Figure 4.2A-C. Correlation matrix and bi-plots from principal components (PC) analysis using plasma proteins associated with α-1-acid glycoprotein in 6-8 year old children in rural Nepal...... 119 Figure 5.1. Flow diagram of study participant selection1 ...... 151 Figure 5.2. Hierarchical clustering of plasma proteins associated with Universal Non- verbal intelligence Test (UNIT) score in school-aged children in rural Nepal1,2 ...... 154 Figure 5.3A-D. Heatmaps of association between plasma proteins and Universal Non- verbal Intelligence Test (UNIT) (A), adjusted for child iron status (B), linear growth (C), and household socio-economic status (D), respectively, in school-aged children in rural Nepal...... 156 Figure 5.4. Risk factors associated with Universal Non-verbal intelligence Test (UNIT) score in a multivariate model in school-aged children in rural Nepal (n=247)1,2 ...... 160 Figure 6.1. Flow diagram of study design and participants ...... 194 Figure 6.2. Histograms and quantile-quantile (q-q) plots of p-values of mean differences in relative abundance of plasma proteins by maternal supplementation group relative to the control group among children 6-8 years of age in Sarlahi, Nepal1 ...... 197 Figure 6.3. Quantile-quantile plots of p-values of mean differences in relative abundance of plasma proteins by maternal supplementation group relative to the control group, stratified by sex among children 6-8 of years of age in Sarlahi, Nepal ...... 198

xiv

LIST OF APPENDICES Appendix 5.1. Associations between the Universal Non-verbal Intelligence Test (UNIT) and potential risk factors of child development in school-aged children in rural Nepal ...... 161 Appendix 5.2. Number of plasma samples per iTRAQ experiment ...... 163 Appendix 5.3. Positive association between plasma proteins and Universal Non-verbal Intelligence Test (UNIT), after adjusted for child iron status, linear growth, and household socio-economic status in school aged children in rural Nepal ...... 164 Appendix 5.4. Negative association between plasma proteins and Universal Non-verbal Intelligence Test (UNIT), after adjusted for child iron status, linear growth, and household socio-economic status in school aged children in rural Nepal ...... 165 Appendix 5.5. Plasma proteins associated with percent correct no-go in school aged children in rural Nepal (q<0.05), ordered by q ...... 166 Appendix 6.1. Comparison of characteristics of children, mothers, and household between all children followed-up and sub-samples...... 201 Appendix 6.2. Distribution of maternal micronutrient supplementation of children for proteomics analysis (n=500) across 72 iTRAQ experiments1 ...... 203 Appendix 6.3. Plasma proteins with iTRAQ channel effects (n=17)1 ...... 204 Appendix 6.4. Proteins of leading-edge subsets of enriched gene sets by maternal folic acid supplementation relative to the control group among girls 6-8 years of age in Sarlahi, Nepal1 ...... 206 Appendix 6.5. Enrichment plots of cytoskeleton (cellular component, GO:0005856) and organ development (biological process, GO:0048513) by maternal supplementation group relative to the control group, stratified by sex among children 6-8 years of age in Sarlahi, Nepal1,2,3 ...... 207

ACRONYMS AND ABBREVIATIONS 11β-HSD2 11β-hydroxysteroid dehydrogenase type 2 ACTH Adrenocorticotropic hormone ADHD Attention deficit hyperactivity disorder AGP Alpha-1-acid glycoprotein APP Acute phase protein APR Acute phase response BBB Blood-brain-barrier BLUP Best Linear Unbiased Predictor BMIZ Body mass index-for-age z-score CRP C-reactive protein DBH Dopamine β-hydroxylase DMR Differentially methylated region ES Enrichment score FDR False discovery rate FWER Family-wise error rate GH Growth hormone GO Gene Ontology GSEA Gene set enrichment analysis HAZ Height-for-age z-score HOME inventory Home Observation for Measurement of the Environment Inventory HPA-axis Hypothalamic-pituitary-adrenal-axis IGF Insulin-like growth factor iTRAQ Isobaric tag for relative and absolute quantitation IUGR Intrauterine growth restriction LME Linear mixed effect LMICs Low and Middle Income Countries LPS Lipopolysaccharide LTP Long-term potentiation MABC Movement Assessment Battery for Children MSigDB Molecular Signature Database MS Mass spectrometry MUAC Middle-upper arm circumference NMDA receptor N-methyl-D-aspartate receptor NNIPS-3 Nepal Nutrition Intervention Project, Sarlahi-3 SNP Single nucleotide polymorphism SNS Sympathetic nervous system UNIT Universal Nonverbal Intelligence Test WAZ Weight-for-age zscore

xvi

1 CHAPTER 1. INTRODUCTION AND SPECIFIC AIMS

During the last several decades, a concerted effort has been made to improve maternal

and child health in low- and middle-income countries (LIMCs). A series of

comprehensive reviews by collaborative researchers have elaborated that undernutrition

in women and children is widespread, and global disease and health burden rooted in

undernutrition is high (1, 2). A great body of research has provided evidence-based

political strategies and interventions to improve maternal and child nutritional status and

to prevent adverse health consequences (3). While we have made remarkable progress on

understanding the major global health problems, as well as potential solutions, what has

not been sufficiently explored is the way how the human body responds to nutrition-

deprived conditions or factors that interfere with normal metabolism. Conventional child

health outcomes such as morbidity and anthropometric measurements in nutritional

research are important public health indicators, but they are limited in their capacity to

reflect the complex biological processes underlying those health outcomes (4).

Modern high-throughput technologies have rapidly advanced and enabled more

comprehensive measurement of bio-specimens. The emerging –omics studies generate

large-scale data of different types of biomolecules, providing a more complete picture of

the health profiles of individuals (5). Among –omics studies, proteomics may have broad

applications in public health nutrition. It not only reflects the current physiologic and

metabolic statuses of individuals, but also indicates changes in gene expression levels,

and possibly predicts future health consequences (6). In addition, as many proteins are

co-regulated as a network, the proteome can illuminate relevant biological pathways

1 associated with nutritional exposure or health outcomes. Out of a variety of human proteomes, the plasma proteome is the most accessible and the richest among human body fluids, containing not only classical plasma proteins, but also a wide range of tissue proteomes (7). Thus, the application of plasma proteomics to public health nutrition may fill the research gap and deepen our understanding of child health.

During 1998-2009, a research team at the Johns Hopkins School of Public Health conducted a randomized controlled trial of antenatal micronutrient supplementation and a series of follow-up studies in southeastern Nepal, where undernutrition and micronutrient deficiencies are common in reproductive-age women (8, 9). In this cohort, pregnant women were randomly assigned to receive daily folic acid, iron-folic acid, iron-folic acid-zinc, or multiple micronutrients, all containing vitamin A, with a control group that received vitamin A alone, from early pregnancy to 3 months postpartum (10). Children were followed-up with at 6-8 years of age to collect plasma specimens (11, 12).

Psychological tests were administered to children a year later to assess general intelligence, executive, and motor functions (13). Previous studies highlighted that chronic inflammation is prevalent among children, and antenatal micronutrient supplementation improved different health outcomes of children, including cognitive functions (11, 12, 14, 15).

Based on the original and follow-up studies, this study performed a secondary data analysis to explore underlying biological processes associated with prenatal micronutrient supplementation, on-going inflammation, and prospective cognitive function, by profiling the plasma proteome of children. To our knowledge, this is the first study to investigate the plasma proteome of children living in LIMCs. Thus, the ultimate goal of this study is

2 to evaluate public health applications of plasma proteomics to better understand child health in undernourished populations. If such proteins that reflect prenatal micronutrient supplementation effects, childhood inflammation, or cognitive function exist, they could potentially provide valid and reliable population-based means of characterizing health status or evaluating the efficacy of nutritional interventions. Such information will be invaluable to researchers and program planners interested in developing interventions to improve the wellbeing of children in underprivileged populations.

Specific Aims

Aim 1: To identify plasma proteins that co-vary with plasma alpha-1-acid glycoprotein

(AGP) in 6-8-year-old Nepalese children.

Hypothesis 1a: Plasma proteins quantified by mass spectrometry will be

positively and negatively associated with plasma AGP measured by radial

immunodiffusion assay.

Hypothesis 1b: Plasma proteins that are co-regulated in the same biological

process will more strongly co-vary than ones in different biological process.

Aim 2: To identify plasma proteins that co-vary with psychological test performance in

7-9-year-old Nepalese children.

Hypothesis 2a: Plasma proteins quantified by mass spectrometry will be

associated with general intellectual abilities measured by the Universal Non-

verbal Intelligence Test (UNIT).

Hypothesis 2b: Observed associations between plasma proteins and the UNIT

score will be attenuated after adjustment for a priori known child development

risk factors (child iron status, long-term nutritional status, and household socio-

economic status).

Aim 3: To examine the effect of antenatal a) folic acid, b) iron-folic acid, c) iron-folic acid-zinc and d) multiple micronutrient supplementation (all containing vitamin A) compared to vitamin A alone as the control on plasma protein profiles in 6-8-year-old

Nepalese children.

Hypothesis 3a: Children of mothers who received different combinations of

micronutrient supplements will have differentially abundant proteins compared to

those whose mothers received vitamin A alone.

Hypothesis 3b: Children of mothers who received different combinations of

micronutrient supplements will have differentially enriched gene sets compared to

those whose mothers received vitamin A alone.

REFERENCES 1. Black RE, Allen LH, Bhutta ZA, et al. Maternal and child undernutrition: global and regional exposures and health consequences. Lancet 2008;371(9608):243-60. 2. Black RE, Victora CG, Walker SP, et al. Maternal and child undernutrition and overweight in low-income and middle-income countries. Lancet 2013;382(9890):427-51. 3. Bhutta ZA, Das JK, Rizvi A, et al. Evidence-based interventions for improvement of maternal and child nutrition: what can be done and at what cost? Lancet 2013;382(9890):452-77. 4. Allen LH. Micronutrient research, programs, and policy: From meta-analyses to metabolomics. Advances in nutrition 2014;5(3):344S-51S. 5. Kaput J, van Ommen B, Kremer B, et al. Consensus statement understanding health and malnutrition through a systems approach: the ENOUGH program for early life. Genes & nutrition 2014;9(1):378. 6. Dettmer K, Aronov PA, Hammock BD. Mass spectrometry-based metabolomics. Mass spectrometry reviews 2007;26(1):51-78. 7. Anderson NL, Anderson NG. The human plasma proteome: history, character, and diagnostic prospects. Molecular & cellular proteomics : MCP 2002;1(11):845-67. 8. Christian P, Shrestha J, LeClerq SC, et al. Supplementation with micronutrients in addition to iron and folic acid does not further improve the hematologic status of pregnant women in rural Nepal. The Journal of nutrition 2003;133(11):3492-8. 9. Jiang T, Christian P, Khatry SK, et al. Micronutrient deficiencies in early pregnancy are common, concurrent, and vary by season among rural Nepali pregnant women. The Journal of nutrition 2005;135(5):1106-12. 10. Christian P, Khatry SK, Katz J, et al. Effects of alternative maternal micronutrient supplements on low birth weight in rural Nepal: double blind randomised community trial. Bmj 2003;326(7389):571. 11. Schulze KJ, Christian P, Wu LS, et al. Micronutrient deficiencies are common in 6- to 8-year-old children of rural Nepal, with prevalence estimates modestly affected by inflammation. The Journal of nutrition 2014;144(6):979-87. 12. Stewart CP, Christian P, Schulze KJ, et al. Antenatal micronutrient supplementation reduces metabolic syndrome in 6- to 8-year-old children in rural Nepal. The Journal of nutrition 2009;139(8):1575-81. 13. Christian P, Murray-Kolb LE, Khatry SK, et al. Prenatal micronutrient supplementation and intellectual and motor function in early school-aged children in Nepal. JAMA : the journal of the American Medical Association 2010;304(24):2716-23. 14. Christian P, Stewart CP, LeClerq SC, et al. Antenatal and postnatal iron supplementation and childhood mortality in rural Nepal: a prospective follow-up in a randomized, controlled community trial. American journal of epidemiology 2009;170(9):1127-36. 15. Stewart CP, Christian P, LeClerq SC, et al. Antenatal supplementation with folic acid + iron + zinc improves linear growth and reduces peripheral adiposity in school-age children in rural Nepal. The American journal of clinical nutrition 2009;90(1):132-40.

2 CHAPTER 2. LITERATURE REVIEW

2.1 The Human plasma proteome

With the completion of profiling the human genome and the availability of sensitive

and comprehensive analytical instruments, proteomics has been rapidly developed and

advanced biomedical research. Proteomics analyzes an entire set of proteins present in

biological specimens at a certain point (1). Proteomics studies evaluate not a couple of

targeted proteins, but hundreds and thousands of proteins, providing opportunities to

understand constituents, variability, biological functions, and the network of human

proteins. The plasma proteome, which is one of the subsets of human proteomes, has

received attention because it is readily accessible with a simple collection procedure, as

compared with those for other human body fluids. In 2002, Anderson et al. reported 289

plasma proteins identified using 2-D gel electrophoresis-based methods, and Adkins et al.

reported 490 plasma proteins from the depleted serum using mass spectrometry (1, 2). In

2005, the Human Proteome Organization (HUPO) developed the Plasma Proteome

Database with 3,020 annotated plasma proteins, based on two or more peptides (3). In its

recent report, the database has been updated to include 10,546 plasma proteins (4).

Complementary to advances in proteomics, a variety of bioinformatics tools, annotation

databases, and analytical strategies have been developing to help the interpretation of

patterns of protein abundance from high-dimensional data in the context of relevant

biology. Together, the integration of multi-disciplines has enabled systematical analysis

of the plasma proteome.

2.1.1. Plasma components

Plasma, the liquid part of blood, represents about 55% of the entire blood volume (5).

Plasma consists of water (90% of plasma), electrolytes, metabolites, nutrients, high molecular weight compounds (peptides, polysaccharides, and polynucleotides), gases, and proteins. With a concentration of 7.0-7.5 g/dL, proteins comprise the majority of plasma solids (6). The concentration of proteins in the plasma plays a key role in regulating osmotic pressure and determining the distribution of fluid between the intravascular compartment and tissues (5).

2.1.2. Definition and categories of plasma proteome

Plasma proteins were originally defined in the 1970s as “those proteins carrying out their functions in circulation” by F.W. Putnam (7). The introduction of an untargeted protein discovery method broadened this concept to “plasma proteome,” which embraces all proteins present in the plasma for a variety of different reasons (1). In N. Leigh

Anderson’s article in 2002, plasma proteins are largely classified by several categories: classical plasma proteins (secreted primarily by the liver and intestines and act in plasma); immunoglobulins; long-distance receptor ligands (hormones, e.g. insulin); local- receptor ligands (e.g. cytokines and short-distance mediators); temporary passengers

(non-hormone proteins that pass through the plasma from the site of synthesis to the site of action, e.g. lysosomal proteins); tissue leakage products (act within cells but released into the plasma upon cell death or damage); aberrant secretion (released from tumors or diseased tissues); and foreign proteins (proteins originating from infectious pathogens)

(1).

The plasma proteome is the richest and the most complex sub-proteome of the different types of human proteomes (1). It contains not only classical proteins, but also a variety of intracellular proteins and sub-parts of other tissue proteomes (1, 8). They are transcription factors, cytoskeleton, nuclear proteins, channels, or cell membrane receptors, secreted into the blood during signaling, cell necrosis, apoptosis, and hemolysis (2). A study showed that the total amounts of cellular proteins in plasma are about two times higher than those of extracellular proteins, suggesting active cellular protein-releasing processes into the bloodstream (9).

2.1.3. Dynamic range of plasma protein concentration

The most unique characteristic of the plasma proteome is a dynamic range of protein concentration. It spans more than 10 orders of magnitude between high-abundance and low-abundance proteins (1). The classical plasma proteins are in the range of 106-1010 pg/ml, tissue leakage proteins are in the range of 103-106 pg/ml, and cytokines are in the range of a few-102 pg/ml. This dynamic range in the abundance creates great challenges for the isolation, identification, and quantification of low-abundance plasma proteins (1).

If poteins are leaked from tissues, they could be related to health or disease phenotypes of interest (10). The detection of these low-abundance proteins can be enhanced by eliminating high-abundance proteins such as albumin (55% of plasma proteins), globulin

(38%), and fibrinogen (7%), although some high-abundance proteins carry some small proteins that bind to high-abundance proteins to increase their half-lives (11).

2.1.4. Contributing factors to variation in plasma protein abundance

Age or life-stage: Ignjatovic et al. reported that up to 100 plasma proteins involved in iron transport and homeostasis, immune response, haemostasis and apoptosis showed

8 developmental differences among neonates, children, and adults (12). Using a total human proteome, Kim et al. identified numerous tissue proteins expressed only in fetuses, but not in adults, suggesting the existence of development-specific proteins (13). Scholl et al. found 61 serum proteins that were significantly differentially abundant by gestational trimesters in 23 Nepalese pregnant women (14). Buckley III et al. showed that average immunoglobulin concentration significantly declined with age among 811 healthy individuals aged from birth to 92 years (15).

Sex: Ignjatovic et al. showed that 0.87-1.8 % of total identified proteins were differentially abundant by gender (12). Corzett et al. observed a trend in gender differences in alpha-2-HS-glycoprotein and other proteins, although the differences were not significant after adjusting for multiple comparison (16). Kim et al. demonstrated that female sex was associated with an elevated abundance of pro-inflammatory proteins, apolipoproteins, insulin-resistance adipokines, and markers of calcification compared to males (17).

Nutritional status: Low concentrations of albumin, tranthyretin, and transferrin had historically been considered biomarkers of malnutrition including kwashiorkor and marasmus, although these direct biological links have been withdrawn after a better understanding of inflammation was achieved (18-20). A study demonstrated that 1.2 mg of daily folic acid supplement for 12 weeks increased the levels of proteins involved in the activation and regulation of the immune system, complement cascade, and coagulation factors in adults (21). Another double-blind randomized controlled trial showed that 6 weeks of fish oil supplementation reduced the abundance of several apolipoproteins and proteins involved in inflammation and blood coagulation in the

9 treatment group, compared with the control group (22). Garcia-Bailo et al. reported that 4 principal component analysis-driven groups of 54 high-abundance plasma proteins were associated with ethno-cultural dietary patterns in a healthy young adult population (23).

Genetic factors: Genetic contributions to protein profiles are out of the scope of this study. However, there is a consistency in the results of many studies that supports a possible genetic explanation for natural variations in protein abundance. Within- individual, a considerable proportion of quantitative variation in the selected plasma proteins was explained by genetic factors (24-27). Intra-individual variation was smaller than inter-individual variation for selected plasma proteins (28-30). On the population level, studies showed that there are considerable differences in plasma protein profiles among different ethnicities (15, 17). These results suggest potential genetic effects on plasma protein profiles.

Other sources of variability, including the season of blood draw, socio-economic status, exercise, sleep, stress, exposure to environmental toxins, and use of drugs might contribute to variation in plasma protein profiles (1, 31). As many external factors can affect abundance of plasma proteins, these factors should be carefully considered in analysis to make a valid inference between plasma proteins and exposure or outcome of interest.

2.1.5. Mass Spectrometry-based proteomics (unbiased protein discovery)

Mass spectrometry (MS) has revolutionized proteomics. The main roles of modern

MS, more specifically tandem mass spectrometry (MS/MS), are the identification and quantification of proteins. MS identifies proteins by determining their exact masses and

10 generating information on the amino acid sequences (32). MS quantifies proteins or peptides largely by labeling proteins/peptides of interest with differentiated chemical/metabolic tags or through a comparative analysis of spectral features (label- free) (33). Among the many different types of quantitative MS-based proteomics, isobaric tags for relative and absolute quantification (iTRAQ) showed sensitivity for relative (or absolute) quantification of peptides/proteins (34). This technology utilizes isobaric reagents to label the primary amines of peptides. The reagents consists of reporter group with different mass, balance group, and peptide reactive group which reacts with the primary amines of peptides (35). The purpose of balance group is to make the labeled peptides from each sample isobaric (same mass). The peptide sequence is determined from the product ions by cleaving peptide inter-residue bonds and the relative quantity of peptide is quantified by comparing the intensities of reporter ion signals in the

MS/MS spectra (35). The identified and quantified peptides are used to infer identification and relative quantification of their corresponding protein (36). The advantage of iTRAQ technology is that it facilitates the simultaneous analysis of up to 4 or 8 biological samples in the same iTRAQ experiment, producing proteomics data from a relatively large number of samples (35). In addition, this technology increases the probability of correct peptide identification, particularly for low-abundant proteins by concomitantly increasing precursor ion intensity (35).

2.1.6. Protein annotation & the Gene Ontology

Analysis and interpretation of high-throughput data is challenging. Compared to a conventional hypothesis-driven analysis, which deals with a single or a couple of proteins of interest, hypothesis-generating analysis yields hundreds or thousands of genes or

11 proteins. As understanding biological functions of a single protein requires considerable effort and time, the interpretation of the underlying biology of a list of genes/proteins associated with the phenotype of interest can be overwhelming. To facilitate this process, automated annotation is available by using bioinformatics databases that provide functional attributes of the genes/proteins. There are public repositories, such as the Gene

Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), BioCarta, and

REACTOME that provide useful information about consistent descriptions of genes or gene products (37). Among them, the GO is an integrative bioinformatics database that comprises three structured, controlled vocabularies (ontologies) that group genes or gene products in terms of their associated biological processes (BP), cellular components (CC) and molecular functions (MF) (37). The ontology term has a unique species-specific identifier (term accession number) and term name. The structure of GO is hierarchical, with child terms being more specialized than their parent terms. Although it is the most widely used public database, its limitations on the quality of annotation and resolution have been recognized (38).

2.1.7. Functional pathway analysis

Analyzing high-throughput data at the functional level is widely used in –omics studies. Identifying pathways or functions that are related to the phenotype of interest can have more explanatory power than a simple list of genes/proteins. There are a large number of analytic knowledge base-driven methods such as over-representation analysis, functional class scoring, and pathway topology-based approaches (39). Functional class scoring, in other words enrichment analysis, assumes that both individual genes with large changes and weak but coordinated changes in sets of functionally related genes can

12 have significant effects. Thus, this method can be useful when the expected effect size is marginal. This method computes gene-level statistics (e.g. t-statistic) across all measured genes associated with a phenotype and then the gene-level statistics in a pathway are aggregated into a single pathway-level statistic (e.g. Kolmogorov-Smirnov statistic) (40).

Statistical significance of the pathway-level statistics is assessed through either phenotype permutation or gene label permutation for each pathway (41). Enrichment analysis has been widely used because it does not require an arbitrary threshold for dividing data into significant and non-significant pools. It uses all genes/proteins for analysis, and it accounts for interdependencies among genes. The major limitation of this method is that it analyzes each pathway independently (39). Because of the hierarchical structure of the GO terms, multiple similar pathways are likely to be enriched due to overlapping genes across pathways.

2.2 Inflammation and the plasma proteome

Inflammation is a natural host defense mechanism to protect organisms from harmful stimuli (42). It accompanies a wide range of changes in local and distant organ systems

(43). One of the prominent characteristics of inflammation is the change in concentrations of a large number of plasma proteins (44). This reaction facilitates the elimination of the infectious organisms, the activation of the repair process, and the restoration of the disturbed physiological homeostasis. The importance of measuring inflammation has been well recognized in nutritional research because of a substantial adjustment in nutrient metabolisms during inflammation (45). In practice, only a couple of proteins such as C-reactive protein (CRP) and alpha-1-acid glycoprotein (AGP) have been dominantly measured as conventional biomarkers of inflammation. Untargeted

13 discovery approaches have been applied in animal studies and demonstrated a consortium of proteins reflecting systematic response to inflammation, but the biological scope of inflammation in healthy population has not been investigated in a holistic manner.

2.2.1. Acute phase response (APR) & acute phase proteins (APP)

A variety of stimuli including infection, tissue injury, stress, and neoplasm can activate macrophages and blood monocytes, and this local reaction initiates the APR cascade (Figure 2.1). The blood cells release inflammatory mediators called cytokines, such as interleukin-1 (IL-1) and tumor necrosis factor (TNF). These cytokines activate local fibroblasts and endothelial cells to release a second wave of cytokines, including IL-

6 (46). IL-6, coupled with IL-1 and TNF-alpha, plays a key role in the regulation of gene expression and the production of APPs (46). Due to substantial changes in the expression of a large number of genes in hepatocytes, there is a considerable alteration in the amount of proteins in plasma (43). Conventionally, it is considered that proteins with at least 25% change in concentration to be APP. Up-regulated and down-regulated proteins are called positive and negative APPs, respectively. Compared to positive APPs, negative APPs have not been well documented. In the paper by Gabay et al. in 1999, 41 positive APPs were listed, as opposed to only 8 negative APPs (43). As negative APPs are usually not involved in host defense, it is speculated that the down-regulation of proteins occurs to save amino acids to produce other positive APPs during inflammation (47). The overall goal of APPs is to restore homeostasis, but APPs are highly heterogeneous in their regulatory cytokines, functions, and duration and magnitude of concentration changes.

Figure 2.1. Acute phase proteins and systematic response to inflammation

( - ) Disturbance of Infection Tissue injury homeostasis Immunologic disorders Chronic inflammation (+) inflammation Chronic Activation of leukocytes, endothelial cells, and fibroblasts Local

reaction Release of IL-6, IL-1, TNF

Vascular Bone Liver HPA ECM endothelium

Breakdown of Systematic Adhesion Resorption Acute phase ACTH skeletal muscle molecules Haematopoiesis proteins Cortisol reaction Connective tissue

Plasma proteome

Healing

Figure has been modified from Heinrich et al. (46).

2.2.2. alpha-1-acid glycoprotein (AGP) vs. C-reactive protein (CRP)

CRP and AGP, the most well-known positive APPs, are widely measured in practice to indicate inflammation status, but they are different in many aspects (Table 2.1). While

CRP dramatically increases its concentration in a short period of time within 24-48 hours and declines sharply, AGP is a constitutively expressed protein and takes 3-5 days to reach the maximum concentration in response to inflammation and takes a long-period of time to get back to the normal concentration (45). Thus, measuring both CRP and AGP can be useful in that these inflammatory indicators reflect the initiation and convalescent phases of inflammation, respectively (45). The biological functions of AGP are largely unknown. One known function is that AGP exhibited anti-inflammatory and

15 immunomodulatory effects by inhibiting neutrophils and complement activation (48).

AGP can be highly glycosylated, and it is hypothesized that this carbohydrate moiety binds E-selectin expressed on endothelial cells, consequently inhibiting leukocyte extravasation (49). Consistent findings support that AGP has a protective effect on cellular or tissue damage that can occur as a consequence of the inflammation process

(50). CRP is widely known to have pro-inflammatory effects and to activate the complement cascade, which is a part of the innate immune system. Some studies suggest that CRP also has anti-inflammatory effects, as well (51).

Table 2.1. Biochemical characteristics between AGP and CRP

AGP CRP Normal concentration 0.5-1.0 0.001 (g/L) Concentration during Increase 2-5 fold Increase 20-1000 fold inflammation Half-life 5.2 days 2 days Time to maximum 3-5 days 24-48 hours Biological functions Anti-inflammatory Pro-inflammatory roles roles 1. Activation of 1. inhibition of neutrophil complement system activation (classical pathway) 2. modulation of 2. Recognition of foreign or lymphocyte damaged cells responsiveness 3. Production of cytokines 3. anti-complement Anti-inflammatory roles activity 1. inhibition of neutrophil 4. inhibition of platelet chemotaxis aggregation 2. inhibition of L-selectin 3. inhibition of alternative complement pathway Table contents were mainly extracted from Thurnham et al (45).

2.2.3. Systematic response to inflammation

One of the unique characteristics of inflammation is a systematic body response involving many organs to local inflammation. Because cytokines have pleiotropic actions on a wide range of different cells and tissues, APR can be accompanied by a large number of behavioral, physiologic, biochemical, and nutritional changes, interacting with many types of different organs and systems (43).

Endothelium vascular system: Pro-inflammatory cytokines induce the expression of surface receptors (e.g. selectins) on local endothelial cells. These cells recruit blood cells from the bloodstream, and those activated recruited cells increase the expression of complement receptors, parts of the cell metabolism such as oxidative burst, or the expression of chemokines, which recruit more defensive cells to inflamed tissues (47).

Hematopoietic changes: Pro-inflammatory cytokines decrease the response of erythrocyte precursor to erythropoietin, and decrease production of erythropoietin, which is the pathogenesis of anemia in chronic disease (52). IL-6 induces thrombocytosis and leukocytosis (53, 54).

Neuroendocrine changes: IL-6 and other cytokines induce fever, somnolence, and anorexia (55). An activated hypothalamic-pituitary-adrenal (HPA) axis increases corticotropin-releasing hormone, which consequently stimulates corticortropin and cortisol production from the adrenal gland (56). Increased cortisol enhances the stimulatory effects of cytokines on the production of APPs (57).

Hepatic changes: Inflammatory cytokines increase production of tissue inhibitors of metalloproteinases and other hepatic constitutive enzymes, and decrease responsiveness

17 to growth hormone by reducing growth hormone receptors on hepatocytes and subsequently lowering plasma insulin-like growth factor I (58).

Metabolic changes: Several cytokines induce the loss of muscle and negative nitrogen balance, decrease gluconeogenesis, increase osteoporosis, increase hepatic lipogenesis, increase lipolysis in adipose tissue, and decrease lipoprotein lipase activity in muscle and adipose tissue (59-61).

2.2.4. Healing

Successful tissue repair requires finely coordinated restitution of epithelial and mesenchymal cells, repair of the extracellular matrix (ECM), and vascular remodeling

(62). The ECM provides a physical scaffold for cells and also serves as a reservoir of growth factors that influence cell proliferation, differentiation, and apoptosis (63). In a late phase of inflammation, synthesis of new ECM, such as collagen and proteoglycan deposition, replaces lost or damaged tissues (64). This newly deposited ECM is remodeled over time to emulate normal tissue. Proteases, such as metalloproteinases

(MMPs), are usually up-regulated and act to fine-tune the nature of the immune response or to remodel the interstitial matrix during chronic inflammation (65). Vascular remodeling is the growth and remodeling process that modifies the existing vascular system to form a more complex branching network (66). These multi-step processes involve ECM degradation, cell proliferation, survival, and migration(67).

2.2.5. Chronic inflammation

If multiple mechanisms that ensure tissue resolution fail, acute inflammation can be prolonged. One of the straightforward causes of non-resolving inflammation is repeated

18 infection. In chronic inflammation, tissue damage and necrosis induce aberrant ECM expression and bioactive ECM fragments by MMPs. These molecules re-activate leukocyte chemotaxis and perpetuate inflammation (62, 68). Hypoxia can occur and induce accumulation of macrophages and other immune cells (69). Then, angiogenesis sustains inflammation by providing oxygen and nutrients for metabolic needs at inflammatory sites. This is an intricate interplay between inflammation, ECM remodeling, and angiogenesis (70). Chronic inflammation is not a primary cause of many chronic diseases, but contributes to the pathogenesis of many adverse health outcomes, such as cachexia, cachectic obesity, insulin resistance, dyslipidemia, altered steroid axes, anemia, osteopenia, etc (71).

2.2.6. Untargeted modern approaches & acute phase proteins

Endotoxin-challenged liver transcriptomics in mouse model (72)

Several studies examined APPs using a global discovery approach. Yoo et al. examined hepatic gene expression of the acute phase response using transcriptomics by endotoxin challenges in mouse models. The authors reported that 898 protein-encoding genes out of 8,551 markers were responsive to a single lipopolysaccharide (LPS) stimulus, and the number of up-regulated genes was similar to the number of genes down-regulated. The responsive genes were functionally annotated with the Gene

Ontology (GO) database. They observed that almost 60% of genes up-regulated are involved in defense and immunity or intracellular signaling, and 50% of genes down- regulated play roles in metabolism. In addition, another prominent change was observed in genes involved in lipid and cholesterol metabolism. Reductions in genes encoding enzymes for fatty acid synthesis, bile acid synthesis, and low- density lipoprotein receptor

19 support the conclusion that inflammation induces reduction in cholesterol uptake and conversion of cholesterol to bile acids. They also observed that LPS induced genes involved in the innate and adaptive immune system, such as serum amyloid, other lipocalins, and genes related to major histocompatibility complex class I antigen presentation machinery.

Chronic inflammation and plasma proteomics in mouse model (73)

Kelly-Spratt et al. induced sub-acute and chronic inflammation in a mouse model to examine changes in the plasma proteome profile during inflammation. This study quantified about 500 plasma proteins, and 25% of proteins were associated with inflammation. Network analysis showed that many proteins responsive to sub-acute inflammation are related to the complement system, coagulation, the fibrinolytic system, antiproteases, transport proteins, and other participants in the inflammatory response.

Proteins responsive to chronic inflammation are related to mainly the inflammatory response and ECM remodeling. This result suggests that cellular signaling networks were common for both sub-acute and chronic inflammation, but impairment in tissue repair might be a distinct mechanism in chronic inflammation.

2.2.7. Database-based profiling plasma inflammatory proteins (74)

Saha et al. dissected the human plasma proteome by systematically searching inflammatory proteins in the human proteome in available datasets. They also performed extensive GO cross-comparison analyses across datasets. They reported a set of 291 inflammatory proteins and a subset of 204 inflammatory plasma proteins by querying the

GO term “inflammatory response (GO:0006594)”. The result showed that functional

20 annotations of the human plasma proteome and a subset of the inflammatory plasma proteome were differently enriched. Compared to the human plasma proteome, plasma membrane part and extracellular space were over-represented and nucleus was under- represented for cellular components, protein binding and signal-transduction were overrepresented in molecular function, and response to stress was over-represented and transport was under-represented in the inflammatory plasma proteome. This result showed that the inflammatory plasma proteome has a profile distinct from the plasma proteome, although it is a subset of the latter one.

2.3 Child development and the plasma proteome

2.3.1. Child development at school-age

Child development is an umbrella term referring to the biological, psychological and emotional changes that occur as a child transitions from a dependent infant to an autonomous teenager (75). These changes usually include the development of language, cognitive skills (e.g. symbolic thought, memory, and logic), motor skills (e.g. gross and fine movements), and social-emotional functions (a sense of self, empathy, and how to interact with others) (76). Various parts of the brain have different developmental trajectories. The most dramatic and fundamental development begins around conception and continues up until early childhood. At the time of school age, the structural development of the brain is fairly similar to that of the adult brain. By the age of 6 years, the total size of the brain will reach approximately 90% of its adult size (77). However, marked maturational processes continue throughout childhood. The most important neuroanatomical characteristic in school-aged children is the development of frontal cortex (78). A spurt in the development of the frontal lobes occurs at approximately 7-9 years, with rapid myelination of neurons in this area. The frontal lobes act as a hub of cortical activity by connecting with all other parts of the brain through feedback and feedforward connective loops (79). Thus, children at this age are able to have a more comprehensive memory and perform higher-order cognitive activities such as planning, developing strategies, problem solving, focusing, maintaining attention, and inhibiting task-irrelevant information stimulation (80).

2.3.2. Global burden of child development in LMICs

Poor childhood development is a global health challenge that affects a substantial number of children living in the developing world. A comprehensive review estimated that approximately 200 million children under five do not reach their developmental potential in less-developed countries (81). Poor child development is known to decrease school and economic performance and affect the health and well-being of individuals, consequently leading to intergenerational poverty transmission (82).

2.3.3. Known risk factors of child development in LMICs

An influential consensus in childhood development research is that child development can be determined not by nature or nurture alone, but by a multidirectional interaction between nature and nurture (or “nature through nurture”) (83). However, a study comparing mono- and dizygotic twins indicates that the magnitude of the genetic effects

(nature) on IQ depends on SES, such that cognitive ability is almost entirely predicted by environmental factors at low-SES class levels (84). This finding might be relevant to children living in developing countries where many aspects of living conditions are not standardized, and the desirable nurture for full potential development is not assured. They are more likely to be exposed to more diverse risk factors, such as poverty, biological and psychological risk factors, and socio-cultural risk factors (gender inequity, low maternal education, or reduced access to health service) (81).

Socio-economic status

Socio-economic status (SES) is strongly associated with child development, psychological well-being, and emotional development (81, 85, 86). In the context of

23 developing countries, many studies showed that poverty as an underlying factor influences a wide range of co-occurring risk factors that exacerbate poor development of children (85). Children growing up in poverty or with a low household SES are more likely to be exposed to poor sanitation, crowded or suboptimal living conditions, lack of psychosocial stimulation, less interactive parenting styles, fewer household resources, an inadequate amount and quality of food, and great environmental hazards (81, 87).

Nutrition

There is consistent evidence that childhood undernutrition is one of the major risk factors of compromised development (88). Chronic undernutrition, reflected by linear growth retardation or stunting, is associated with delayed cognitive and psychomotor development, poor fine motor skills, altered behavior, and school achievement, adjusting for potential confounders in many different studies (81). Also, studies indicate that growth faltering at early ages is more strongly associated with the outcome than that occurring at older ages (89). In randomized trials, food supplementation to young children improved nutritional status, motor and mental development, and cognitive ability

(81). It has been recently recognized that prenatal nutritional disturbances may play a crucial role in establishing developmental trajectories and early development, but its long-term effects are less consistent (88).

Among micronutrients, deficiencies in iodine and iron have been widely recognized as nutritional risk factors (81). Iron deficiency anemia in infancy or the preschool period is associated with poorer cognitive, motor, and social-emotional development in multiple studies (81, 88). Iron treatment to iron-deficient infants did not substantially improve

24 motor development, executive functioning, and recognition memory in young adulthood, suggesting compromised cognitive abilities from iron deficiencies in early childhood are less likely to be reversible (90-92). Iodine deficiency can lead to congenital hypothyroidism and irreversible mental retardation. Meta-analysis of 18 studies showed that iodine deficiency is associated with an 13.5-point average drop in IQ score, and prenatal or postnatal iodine treatment improved IQ scores by an average of 8.7 points (93,

94).

Infectious diseases

Association between diarrhea and child development is equivocal (88). Because adjustment for stunting attenuates the association between diarrhea and cognitive outcomes, its independent association remains to be clarified (95). In many studies, malaria parasites or HIV infections were associated with poor child cognitive development, including school incompletion and language and cognitive deficits in children after covariate control (96-98). Intervention with chemoprophylaxis or highly active antiretroviral therapy showed positive effects on school performances and attendance in children (99-101). In a malaria or HIV non-endemic population, enteric infection such as helminthes was strongly associated with impaired cognitive development, after adjustment for SES, schooling, stunting, and others potential covariates in Peru (102).

Schooling

Schooling is a crucial factor contributing to the cognitive abilities of children in

LMICs where primary education is not compulsory for children (88). It is a strong

25 independent predictor of cognitive outcome, and it may lie on the same causal pathway of household socio-economic status, cultural aspects, and parents’ education in low-resource settings (103). Thus, school attendance is a contributing factor for child cognitive development, not only because it provides ample learning opportunities, but also because it is an indicator of contextual conditions of household.

Parenting style

Parenting style and household environment may not affect the cognitive development of school-aged children as much as that of young children, but they may continue to affect children’s cognitive abilities. In children under 5, caregiver and child interactions were facilitated by caregivers’ positive emotionality, sensitivity, responsiveness, and avoidance of harsh physical punishment (88). Many studies showed that intervention with improved child-parent interactions showed positive effects on child cognitive outcomes, and the effects remained later as well in some studies (88).

2.3.4. Body systems potentially associated with child cognitive development

As a regulatory process simultaneously modulates a wide range of body systems, human body systems are highly interdependent. Active interactions among the central nervous system (CNS), the endocrine system, and the immune system have been well documented (104-106). Whether plasma proteins co-vary with cognitive functions is an open question. A physical endothelial barrier, the Blood-Brain Barrier (BBB) exists to protect the brain from circulating toxins or peripheral molecules in the blood (107).

However, it has been recently recognized that the BBB has a variety of receptors and a wide range of permeability as a dynamic interface between the CNS and the rest of the

26 body systems (107). In addition, some neurotrophins regulating synaptic plasticity or proteins metabolizing neuropeptides are found in peripheral blood (108). Lastly, profiles of plasma proteins, as a snapshot of current physiological and metabolic conditions, may indirectly reflect the activities of the CNS, as they are co-regulated by common factors that can directly affect the CNS. Thus, it is possible to hypothesize that plasma proteins are associated with the CNS function (Figure 2.2.).

Figure 2.2. Potential biological systems associated with child cognitive function

Behavior Performance

CNS development & Brain function

Endocrine system

Body systems Lipid Plasma Immune metabolism proteome system

Neuropeptide

Biological & Nutrition Schooling psychological Iron status Psychological Iodine status stimulation factors

Parent eduction Context Ethnicity Income

Endocrine system

Growth hormone/Insulin-like growth factor axis (GH/IGF-axis)

There has been an increasing interest in the neurodevelopmental functions of growth hormone (GH)/insulin-like growth factors (IGFs) in children. IGFs are mainly secreted from the liver following GH stimulation and play crucial roles in both somatic growth and development (109). IGF-1 is involved in regulating neural development including neurogenesis, myelination, synaptogenesis, dendritic branching and neuroprotection after neuronal damage (110). Over-expression and deletion of IGF-1 increased and decreased brain size of mice, respectively (111, 112). Also, impairment in the IGF-1 axis affected the development of corticospinal tract, which controls discrete voluntary skilled movements, such as precise movement of the fingers (113). Gunnell et al. found that serum IGF-1 was positively associated with verbal intelligence in 547 normal children at

8-9 years age in the Avon Longitudinal Study of Parents and Children (114). Although there is not as much research into IGF-2 as IGF-1, it is reported that IGF-2 is associated with memory function. IGF-2 administration to the hippocampus significantly enhanced memory retention in rats (115). Growth hormone (GH) also showed profound effects on brain function in both animal and human studies. Nyberg et al. suggests that GH and

IGF-1 may affect cognitive functions including learning and memory by directly passing the BBB (116). GH interacts with the methyl-D-aspartate (NMDA) receptor complex located in the CNS, which leads to long-term potentiation (LTP) and affects excitatory circuits involved in synaptic plasticity [47, 48]. In clinical studies, GH treatment improved not only linear growth, but also cognition and behaviors in children born small- for-gestational age (117). Webb et al. found that children with growth hormone deficiency (IGHD) had low scores of full-scale IQ and movement skills, as compared with children with idiopathic short stature (118). It has been well established that

28 nutritional status affects the secretion of circulating IGF-1 by regulating the number of growth hormone receptors in the liver (119).

Hypothalamic–pituitary–adrenal axis (HPA)

HPA is a major part of the neuroendocrine system, which comprises three endocrine glands and complex feedback loops through corticotropin releasing hormone (CRH), adrenocorticotropic hormone (ACTH), and cortisol. It controls body reactions to stress and regulates many body processes, including digestion, immune system, mood, sexuality, and energy metabolism (120). Epidemiological studies suggested that postnatal glucocorticoids given to preterm infants to reduce the severity of chronic lung disease is associated with an impairment in brain development (121). Early postnatal glucocorticoid administration was associated with poor motor coordination and visual-motor integration, and lower IQ in school-aged children who were preterm infants (122). An in vivo study demonstrated that excess glucocorticoid exposure resulted in retardation of radial migration during cortical development in rats (123). Animal studies demonstrated that the quality of parental care can affect neurodevelopment in offspring by modifying the HPA- axis. Rat pups of mothers who exhibited frequent licking, grooming, and arched back- nursing showed increased hippocampal glucocorticoid receptor expression and reduced hypothalamic CRH and plasma cortisol responses to acute stress, as compared with offspring of dams that exhibited low licking and grooming (124). In a human study, child abuse was associated with epigenetic gene regulation of reduced hippocampal glucocorticoid receptor expression in suicide victims (125). These findings suggest that the quality of parenting can affect child cognitive development by altering the HPA stress response, which is closely related to synaptic development and cognitive function (126).

Gut hormones

A large amount of energy is utilized by brain relative to its volume and the proportion is greater in young children. With this reason, gut hormones or peptides involved in energy homeostasis are suspected to influence the cognitive function. Animal studies revealed that these hormones directly pass the BBB and affect cognitive processes, such as learning and memory, mainly by modulating synaptic plasticity (104). Insulin and glucagon-like peptide 1 (GLP1), are released into the bloodstream, reach the hypothalamus and the hippocampus, and activate signal-transduction pathways that promote synaptic activity (127). Ghrelin, endogenously produced in the stomach, can promote rapid reorganization of synaptic terminals in the hypothalamus and synapse formation in dendritic spines, subsequently promoting long-term potentiation (LTP) and spatial memory (128). Leptin, produced by adipose tissue, also activates NMDA receptor function in the hippocampus and the hypothalamus and influences learning and memory

(129). Another study showed that genetically obese rodents with leptin dysfunction showed impairments in LTP and spatial learning (130). Overall, gut hormones, which are directly regulated by diet or fat mass, ultimately bind to receptors expressed in neurons in the brain and act on the expression of genes implicated in synaptic plasticity in neurons

(104).

Immune system

Beyond its classical role in host defense, the immune system is critically involved in normal brain development. Cytokines are involved in glial cell migration, differentiation, and synaptic maturation (131). The classical complement cascade is also required for

30 developmental synapse elimination during brain development (132). Because of these crucial roles in brain development, the dysregulation of cytokines and complement components can be associated with abnormal brain development. Studies showed that bacterial or virus infection markedly increased pro-inflammatory cytokines, including IL-

6, TNF-alpha, and IL-1beta, and cortisol in the peripheral blood and also within the CNS

(133, 134). They disrupted structure in the hippocampus and cerebellum, and promoted the apoptosis of neurons through pro-apoptotic gene expression (134). In addition, increased cytokines suppressed the induction of brain-derived neurotrophic factor

(BDNF), which has neuroprotective action (135). Studies suggest that cognitive function is particularly vulnerable to infection because the hippocampus has the largest number of cytokine receptors within the CNS (136, 137). Bilbo et al. showed that perinatal E.coli infection changed the number of microglial cells in the CNS, and the microglial cells remained chronically activated. Subsequent endotoxin challenge induced the over- production of cytokines in the neonate brain (133, 138). The authors concluded that altered immune response during critical period of brain development can lead to enduring consequences for memory or other cognitive abilities (139). In addition, receptors of neurotransmitters including dopamine, serotonin, and endorphin, and neuroendocrine hormones including adrenocorticotropic hormone (ACTH), CRH, and TRH, have been found in lymphocytes, suggesting that orchestrated immune reactions are modulated by active interplay with molecules that regulate cognitive function (140, 141).

Neuropeptide

Evidence has been accumulated regarding dopamine β-hydroxylase (DBH) and its role in cognitive and motor functions. It is the only enzyme involved in converting

31 dopamine to norepinephrine (142). Norepinephrine is the predominant neurotransmitter of the sympathetic nervous system (SNS), which modulates the activity of many vital organs and facilitates adaptation to physical and metabolic challenges. DBH is an ascorbic-acid- and copper-dependent enzyme expressed in noradrenergic nerve terminals of the central and peripheral nervous systems, the prefrontal cortex, and the adrenal medulla (143). Both dopamine and norepinephrine are known for their involvement in potential biological explanations of dysfunctions in motor control, sleep cycles, learning and memory (144). Thomas et al. demonstrated that 90% of DBH knockout mice exhibited delays in rapid growth during lactation and lower performance in motor functions (143). In a human study, serum DBH level was significantly lower in attention deficit hyperactivity disorder (ADHD) children and children with acute depressive disorders than in normal children (145). Single nucleotide polymorphism (SNP) at DBH was associated with performance on the color word Stroop task in Chinese children with

ADHD, indicating that DBH plays a role in executive functions in children (146). A study showed that iron deficiency was associated with 75% increased activity of DBH by altering brain copper metabolism in rats (147), but more studies are needed to evaluate the relation between iron status and serum DBH level.

Lipid metabolism

Lipid metabolism mediated by apolipoproteins may share a common molecular pathway with neurodevelopment in children. Apolipoprotein E (ApoE) is a lipid transport protein that plays important roles in neuronal metabolism and the repair of nerve tissue

(148). ApoE transports cholesterol and fatty acids to the brain, which are essential nutrients for neurite outgrowth and synaptogenesis (149). When a peripheral nerve is

32 impaired, macrophages secrete ApoE to scavenge cholesterol and lipids from the nerve, and neurons are regenerated or repaired using the scavenged cholesterol from ApoE rather than de novo cholesterol (150). ApoE knockout mice poorly performed in the

Morris Water Maze test compared to control, implying that ApoE is involved in learning and memory functions (151, 152). In a human study, ApoE4 genotype was associated on average with a 4.4 point higher Mental Development Index score in infants at 24 months, indicating that ApoE4 isoform carriers would have advantages in neurodevelopment over

E2/E3 carriers (153). Rask-Nissila et al. reported that children with ApoE4 had lower failure rates on neuropsychiatric tests compared to control at 5 years, although this difference was not significant (7.7 vs. 10.4 %) (154). In addition, an increasing number of epidemiological studies have observed that disturbed cholesterol metabolism or altered concentrations of other apolipoproteins such as ApoB, ApoA1, and A4 are associated with psychiatric disorders such as autistic spectrum disorders (ASD) (155), although direct biological links remain unknown. Perry et al. found no association between serum cholesterol level (total cholesterol, high density lipoprotein, and non-high density lipoprotein cholesterol) and academic performance after adjustment for covariates in nationally representative school-aged children and adolescents in the U.S. (156). Overall, compared to widely accepted roles of cholesterol and lipids in brain development, empirical evidence is far from sufficient.

2.4 Effect of antenatal micronutrient supplementation on child plasma proteome 2.4.1. Micronutrient deficiencies in pregnant women living in LMICs

Micronutrient deficiency is one of the major public health challenges in LMICs (157).

Pregnant women living in LMICs are particularly at a high risk of multiple micronutrient deficiencies. A systematic review reported that folate intake is most frequently inadequate among micronutrient intakes in pregnant women in undernourished populations (158). According to the World health organization (WHO) review of nationally representative surveys from 1993 to 2005, more than 42% of pregnant women in developing countries are anemic, and 60% are iron-deficient anemic (159, 160). There is little information about zinc deficiency, but it can be roughly estimated that 82% of pregnant women worldwide likely have inadequate usual intakes of zinc using probability methods (161). Also, Torheim et al. reported that zinc intake was generally inadequate in pregnant women living in developing countries across all regions following iron and folate intakes (158). In addition, pregnant women are likely to have multiple micronutrient deficiencies, including vitamin D, vitamin E, vitamin C, vitamin Bs, vitamin K, and copper. According to Jiang et al, approximately 20% of pregnant women in rural Nepal had more than 5 micronutrient deficiencies early in their pregnancies (162).

2.4.2. Antenatal micronutrient supplementation and intrauterine growth

The WHO recommends that non-anemic women or adolescents take weekly and anemic women take daily iron and folic acid supplementation throughout pregnancy to prevent anemia in pregnancy and to improve pregnancy outcomes (163, 164). Among anemic pregnant women, standard iron-folic acid supplementation is known to reduce anemia, iron deficiency, and low birth weight, increase birth weight, and has no effect on preterm delivery and neonatal mortality (163). The rationale behind including folic acid

34 in supplementation is to prevent maternal magaloblastic anemia and neural tube defects, but the health benefits of folic acid supplementation after neural tube closure (~4 weeks of gestation) are not clear (164). The importance of other micronutrients during pregnancy has been recognized, and a considerable number of epidemiological studies have been conducted to examine additional effects of multiple micronutrient (MM) supplementation versus standard iron-folic acid supplementation. The result of meta- analysis of 12 trials showed that MM supplementation, which contains 14 micronutrients, increased birth weight by 22g and reduced low birth weight and small for gestational age by around 10%, compared to iron-folic acid supplementation (165). A recent study not included in the meta-analysis showed that daily MM (15 micronutrients) supplementation versus iron-folic acid reduced preterm birth and low birth weight by 15% and 13%, respectively, increased birth weight by 55g and head circumference by 0.21 cm, and reduced infant mortality at 6 months among female Bangladesh infants, although this effect was statistically marginal (166). In summary, there is converging epidemiologic evidence that intervention with single or multiple micronutrient supplementation during pregnancy improves intrauterine growth in undernourished populations.

2.4.3. Antenatal micronutrient supplementation and postnatal health

The effects of maternal micronutrient supplementation on health outcomes in postnatal life is largely unknown. A limited number of epidemiological studies evaluated growth, body composition, and cognitive development, and a smaller number of studies examined the mortality, and cardiovascular and metabolic health of children (Table 2.2).

Compared to the consistent positive effect on birth size, the effects on health outcomes substantially vary among studies, partially due to non-standardized maternal

35 micronutrient supplements, heterogeneous outcome measurements, and widely varying timing of follow-up studies. Studies with multiple follow-ups observed that positive effects on postnatal growth at earlier follow-ups disappeared upon later follow-ups in

Nepal (Janakpur) and Burkina Faso. This result suggests that the effect of prenatal micronutrient supplementation on child growth may be transient, although more studies are needed to confirm this result. One inherent limitation of these studies is that most studies compared outcomes between multiple micronutrient vs. iron-folic acid supplementation groups without a true placebo group. Taken together, there is insufficient evidence in epidemiological studies with robust designs to support that antenatal micronutrient supplementation has long-term effects on health.

2.4.4. Plausible biological mechanisms and roles of micronutrients in fetal programming

The crux of long-term effects of maternal micronutrient status on offspring health originates from the hypothesis of fetal programming (167, 168). Out of all stages of human life, the most intense growth and developmental processes occur during the 9 months of pregnancy. One characteristic that uniquely defines this period is developmental plasticity. Developmental plasticity is the ability of organisms to change their phenotypes in response to environmental stimuli (169). If these changes or adaptations remain persistent, they are considered “programming,” which is associated with persistent effects in structure and/or function (190). In most case, programming is beneficial to organisms in that it increases the chances for survival in a given environment (167). However, it could be harmful if a mismatch happens in which individuals adapted to a previous environment are exposed to a different environment

(191). The most famous example of fetal programming is found in studies on the Dutch

Famine. The studies showed that people who experienced prenatal undernutrition were more likely to have negative consequences for mental and physical health (192).

Compared to relatively abundant examples of protein-energy depletion, empirical evidence of fetal programming derived from maternal micronutrient status is still limited.

Table 2.2. Effect of antenatal micronutrient supplementation on growth, development, mortality, cardiovascular and metabolic health in children living in LMICs (≥12 months)

First author Setting Exposure Effect observed Timing at follow-up

Growth

Hamadani JD (170) Bangladesh Zn vs. placebo No differences in weight and height 13 months (Dhaka)

Roberfroid D (171) Burkinafaso MM vs. IFA MM reduced stunting by 12mon, but not in 30 mon up to 12 MM increased WHZ and HC at 12 mon and increased WHZ at 30 months and mon 30 months

Villamor E (172) Tanzania1 MV, MV+VA/BC, and MM improved weight, WAZ, and WHZ 24 months VA/BC vs. placebo

Huy ND(173) Vietnam MM vs. IFA Reduced stunting 2 yrs

Wang W(174) China MM vs. IFA vs. FA No differences in stunting, underweight, and wasting among any 30 months comparisons.

Islam Khan A(175) Bangladesh Food supplementation, Prenatal MM increased stunting compared to IFA. 54 months (MINIMat) MM, IFA Food and micronutrient supplementation did not affect body composition

Vaidya A(176) Nepal (Janakpur) MM vs. IFA MM increased weight, HC, CC, MUAC, and triceps. 2.5 yrs

Devakumar D(177) No differences in WAZ, HAZ, and BMIZ 8.5 yrs

Stuart CP(178) Nepal (Sarlahi) FA, IFA, IFAZn, and MM IFAZn reduced stunting 6-8 yrs vs. control (VA only)

Development

McGrath N(179) Tanzania MM vs placebo MM improved motor development but not mental development 6-18 months score

Li Q(180) China MM vs. IFA MM improved mental development but not psychomotor score 12 months

Hamadani JD(170) Bangladesh Zn vs. placebo Negative effect on cognitive and motor functions, no effect on 39 months (Dhaka) behavioral functions

Prado EL(181) Indonesia MM vs IFA MM improved motor and cognitive functions 42 months

Caulfield LE(182) Peru IFAZn vs IFA Zn improved autonomic regulation 54 months

Caulfield LE(183) Peru IFAZn vs IFA No difference in cognitive, social, or behavioral outcomes 4-5 yrs

Christian P(184) Nepal (Sarlahi) FA, IFA, IFAZn, and MM IFA improved intellectual, motor, and executive functions 7-9 yrs vs. control (VA)

Mortality

Andersen GS(185) Guinea-Bissau MM vs. IFA No effect on mortality under 2 yr

Shaheen R(186) Bangladesh Food supplementation, Food+MM lowered mortality under 5 yrs (MINIMat) MM, IFA

Christian P(187) Nepal (Sarlahi) FA, IFA, IFAZn, and MM IFA improved mortality 6-8 yrs vs. control (VA only)

Blood pressure and metabolic syndrome

Hawkesworth S(188) Bangladesh Food supplementation, MM increased diastolic blood pressure 4.5 yrs (MINIMat) MM, IFA

Vaidya A(176) Nepal (Janakpur) MM vs. IFA MM slightly lowered blood pressure 2.5 yrs

Devakumar D(177) No difference in blood pressure 8·5 yrs

Stewart CP(189) Nepal (Sarlahi) FA, IFA, IFAZn, and MM FA improved microalbuminuria and metabolic disorder 6-8 yrs vs. control (VA only)

1HIV-infected women

Abbreviations: MV, multiple vitamins (thiamine, riboflavin, vitamin B-6, niacin, vitamin B-12, vitamin C, vitamin E, and folic acid); VA/BC, vitamin A and beta-carotene; Zn, zinc; FA, folic acid; IFA, iron-folic acid; IFAZn, iron-folic acid-zinc; MM, multiple micronutrient; WHZ, weight-for-height zscore; WAZ, weight-for-age zscore; HAZ, height-for-age zscore; BMIZ, body mass index zscore-for-age zscore; HC, head circumference; CC, chest circumference; MUAC, mid-uppder-arm-circumference

Epigenetic mechanisms

There are several plausible mechanisms of factors underlying fetal programming

(Figure 2.3). Epigenetic regulation may be a key mechanism of fetal programming.

Epigenetics refers to all modifications to genes other than changes in the DNA sequence itself (193). Epigenetic modification commonly includes DNA methylation, histone modification, and chromatic structure remodeling. Among these, DNA methylation is best understood. DNA methylation is more likely to occur at CpG dinucleotides, which are common in promotor regions. In most cases, a high level of methylation at promotor regions prevents transcription factors from binding and is associated with lower levels of gene expression (194). This change could be temporary or long-term if the changes occur during the respective critical periods of specific organ or tissue development (195).

Epigenetic modification can be influenced by many factors that influence the in utero milieu, including maternal diet. A well-known study using an agouti mice model demonstrated that methyl content of maternal diet is an important determinant of permanent alteration in DNA methylation and offspring phenotypes (196). Several micronutrients, such as folate, vitamin B6, B12, and riboflavin have important implications because they are involved in methyl metabolism as methyl donors or cofactors. Evidence of early micronutrient-induced epigenetic change in human studies is still limited. Pre-conceptional and peri-conceptional folic acid use (400 μg/d) was associated with an increase in methylation of 4.5% in the differentially methylated regions (DMR) of the IGF-2 gene in children aged 17 months (197). Another study found that folic acid supplement use before and during pregnancy reduced 2.8% of methylation at H19 DMR and that this reduction was profound in the cord blood of male compared to

40 female infants (198). These findings suggest that maternal folate status during pregnancy resulted in changes in DNA methylations in IGF2/H19, which is known to mediate embryonic growth and development (199).

Glucocorticoid or Hypothalamic-Pituitary-Adrenal programming

An altered set-point of the hypothalamic-pituitary-adrenal (HPA) axis or glucocorticoid level may play an important role in fetal programming. Glucocorticoids are steroid hormones that mediate stress responses and metabolic functions by binding to glucocorticoid receptors. There is a very large difference in the concentration of glucocorticoids between the placenta and the fetal compartment (1000:1) (200). Placental

11β-hydroxysteroid dehydrogenase type 2 (11β-HSD2) plays the barrier role by converting glucocorticoids to inactive forms to prevent the passing of glucocorticoids from mother to fetus (200). Maternal exposure to a variety of stressors, including undernutrition, can elevate maternal glucocorticoid levels and reduce placenta 11β-HSD2 activity, which results in excessive exposure of the fetus to glucocorticoids (201).

Transported glucocorticoids can induce early tissue maturation, favoring differentiation over proliferation, tissue remodeling, and altered gene expression (202). The molecular mechanisms of HPA programming refer to long-lasting changes in gene expression, including the expression of the glucocorticoid receptor itself (203, 204). Differential programming of glucocorticoid receptors in different tissues reflects permanently altered responses to stress. Previous studies showed that maternal iron deficiency and anemia- induced hypoxia are associated with increased levels of norepinephrine and glucocorticoids (126). An animal experimental study showed that deficiency in copper,

41 zinc, or vitamin E in maternal diet reduced placental expression of 11β-hydroxysteroid dehydrogenase in mice (205).

Figure 2.3. Plausible mechanisms of maternal micronutrient status during pregnancy and fetal programming

Prenatal Maternal micronutrient exposure deficiency

Epigenetic regulation Potential Hypothalamic-Pituitary- mechanisms Adrenal axis programming Oxidative stress

Liver (FA, Vit B6/B12) Fetal Skeletal muscle (FA, Fe, Vit D) tissue Central nervous system (FA, Fe) programming Immune system (FA, Fe, Zn) Endocrine system (Fe, Zn, Vit C/E)

Altered Plasma phenotypes proteome in children

Oxidative stress Fetal and placental growth are sensitive to oxidative stress. Excessive reactive oxygen species (ROS) can influence gene expression, directly damage cell membranes and other molecules, and lead to endothelial and placental dysfunction (206). Inflammation, infection and undernutrition (macro and micronutrients), which are widespread among pregnant women in developing countries, are common causes of oxidative stress. Protein or energy deficiency can directly lead to a pro-oxidative state because proteins are

42 involved in antioxidant synthesis, and micronutrients such as vitamin A, C, and E are anti-oxidants themselves (207). Elevated oxidative stress has been found to be associated with unfavorable intrauterine growth, resulting in small-for-gestational age and preterm birth (208).

2.4.5. Empirical evidence of maternal micronutrient supplementation on offspring tissue/system programming

To our knowledge, no study has examined the effect of maternal micronutrient status during pregnancy on the plasma proteome of offspring in human or animal studies.

However, there is evidence that prenatal micronutrient exposure induces long-lasting changes in specific tissues or organs in experimental animal studies. Some studies examined the effects on the tissue transcriptome or proteome of offspring to identify biological processes sensitive to prenatal exposure (Table 2.3).

Folate, riboflavin, vitamin B6, and B12

Folate is a water-soluble vitamin B that cannot be synthesized in the body and is entirely dependent on dietary sources and gut micro-organisms (209). Folate/folic acid and vitamin B2, B6, and B12 are involved in one-carbon metabolism, which is fundamental for the maintenance and repair of gene expression, amino acid metabolism, and neurotransmitter synthesis (210). They also promote DNA and RNA synthesis, which are essential for cell differentiation (211). Although the long-term effects of maternal supplementation with micronutrients involved in methyl metabolism after embryogenesis is mostly unclear, epigenetic changes might occur at the local domain level, depending on the critical windows of the specific developmental milestones of the tissues (195).

Liver

Because the liver plays a central role in metabolic homeostasis, it has been frequently targeted to examine the effects of prenatal methyl consumption on metabolic changes in offspring. Animal studies showed that a maternal methyl-deficient (MD) diet lacking folic acid, choline, methionine, or riboflavin during the peri-conception period or throughout pregnancy programmed the insulin axis and glucose metabolism in the rat and sheep offspring (212, 213). Maloney et al. examined the effect of a maternal MD diet on the hepatic proteome and identified 41 differentially expressed hepatic proteins associated with energy metabolism, amino acid metabolism, and antioxidant defense in male adult offspring rats (214). Liu et al. also reported that maternal folic acid supplementation changed the expression of 11 hepatic proteins involved in immune response, energy metabolism, cellular signal transduction, proteolysis, and cell migration regulation in newborn piglets by profiling the hepatic proteome (215). Several studies reported that a maternal protein-restricted diet (PR) increased expression of the glucocorticoid receptor (GF), PPAR-alpha, and other glucose-metabolism-related proteins in the livers of the offspring throughout their adult lives (216, 217). Lillycrop et al. demonstrated that these changes can be prevented by additional folic acid supplementation (216). The authors also repeated the same experiment using a genome- wide liver transcriptome analysis (218). They found that a small subset of genes (1.3%) reflects an adaptive response to maternal protein restriction, and folic acid supplementation reduced the number of differentially expressed genes by 0.7%, exhibiting a moderate protective effect against a protein-restricted diet (218).

Skeletal muscle

Studies reported that folate deficiency during gestation caused intrauterine growth restriction (IUGR) in piglets, and skeletal dysplasia was the major manifestation of IUGR

(219, 220). Li Y et al. demonstrated that maternal folate deficiency during early-mid gestation reduced muscle fiber number and intramuscular fat deposition in newborn piglets (221). They also observed substantial changes in the muscle transcriptome, characterized by changes in the expression of about 3000~4000 genes. These genes are mostly involved in myogenesis and lipid metabolic pathways.

Immune system

Several epidemiological studies found that high doses of maternal folic acid supplementation during pregnancy are associated with risks of atopy, asthma, wheezing, and lower respiratory tract infections in childhood, suggesting adverse consequences of over doses of folic acid intake on the development of the immune systems of offspring

(222, 223). Modest restrictions in the maternal methyl deficient diet from peri-conception did not change pregnancy outcomes including birth weight, but the adult offspring showed altered immune responses to antigenic challenge in the offspring sheep (213).

Iron

Iron is a fundamental mineral for numerous biological events including oxygen transport, ATP production, DNA synthesis, mitochondrial function, and protection of cells from oxidative damage (224). In addition, iron serves as an essential cofactor for post-translational incorporation into hemo-proteins such as hemoglobin, cytochromes, and iron-sulfur clusters for redox metabolism in mitochondria (225). Prenatal iron may

45 play crucial roles in maturation and functioning of different body systems including CNS, skeletal muscle system, and hormonal regulation system.

Central nervous system

It has been well established that iron status during the prenatal period plays a crucial role in neurodevelopment, influencing myelination, brain energy metabolism, and the synthesis of neurotransmitters, including serotonin and dopamine, during fetal development (226). A study showed that moderate iron deficiency during gestation and lactation altered more than 25 gene expressions in the hippocampi of rats (227). Among the altered gene transcripts, 30%, 20%, and 11 % were involved in metabolism, signal transduction, and neuron localization, respectively. This study also found an up- regulation of genes regulating amyloid precursor protein (227). Another study investigated the impact of developmental iron deficiency on the cerebrospinal fluid (CSF) proteome in infant rhesus monkeys. After 6 months’ deprivation of iron, 12 proteins largely involved in brain-specific function, muscle function, the immune system, and thyroid hormone carriers were differentially expressed by at least two fold changes (228).

Tran PV et al. employed iTRAQ and pathway analysis to examine the effect of prenatal iron deficiency on the hippocampus proteome, called synaptosome (229). The authors identified 331 proteins differentially abundant in the hippocampi of offspring of mothers with iron-deficient diets. Selective proteins are involved in synaptic plasticity, including glutamate receptor signaling pathway, and cellular processes essential for learning and memory. Pathway analysis revealed that a maternal iron-deficient diet altered the expression of synaptic protein-mediated cellular signaling and newly discovered neuronal

46 nitric oxide synthase signaling, suggesting a critical role of prenatal iron status in synaptic efficacy.

Endocrine system

Iron deficiency can alter or “program” the HPA-axis, which regulates many developmental processes (230). (Please see 2.4.4 Glucocorticoid programming). In addition, prenatal iron status may play an important role in thyroid hormone metabolism.

Bastian et al. found that iron deficiency during gestation and lactation periods decreased serum total T3 by 43%, serum total T4 by 67%, and whole-brain T3 by 25% in rats (231).

In addition, they discovered that thyroid hormone (TH) responsive genes in the brain were altered in iron-deficient rats, suggesting that brain defects associated with neonatal iron deficiency are mediated through reduction in circulating and brain TH levels (231).

Muscle development

Prenatal iron deficiency influences muscle development by altering iron-containing enzymes. During fetal development, iron is prioritized to hemoglobin synthesis in red blood cells when available iron does not meet the demands (232). Thus, non-heme tissues, such as skeletal and heart muscle, can be iron-depleted at normal hemoglobin levels. Rats with dietary iron deficiency during fetal and lactation periods were unable to grasp a bar as long as rats without this deficiency, indicating a deficit in muscle strength and endurance (233). Because oxidative energy production is required to contract muscle and to sustain it, the result suggests thatnormal muscle development is highly dependent on many mitochondrial enzymes that contain iron (232).

Zinc

Zinc is an essential trace element, playing important roles in protein synthesis, nucleic acid metabolism, and gene transcription by forming zinc fingers in the transcription factors (234, 235). Because of its roles in fundamental cellular functions, it may influence developmental processes (236). Hanna et al. examined the effect of gestational zinc deficiency on the IGF-axis of rat fetuses (237). A maternal zinc-deficient diet reduced

IGF-1 concentration in the amniotic fluid, plasma IGF-binding protein1 concentration in the plasma, and IGF-1 expression in the livers of fetuses. Zinc is also known to be essential in the finely tuned processes of neurogenesis, neuronal migration, differentiation, and apoptosis, all of which involve the developmental shaping of the nervous system

(238). Zinc supplementation during pregnancy increased the number of proliferating stem cells of newborn mice and other stem cell markers in the developing cerebella and cortices of newborn mice (239). Zinc deficiency during the lactation period impaired the differentiation of dendrites and Purkinjie cells and was associated with elevated levels of glucocorticoids, which delayed the migration of neurons during cerebral cortex development in rat pups (123, 240, 241).

Vitamin D

Vitamin D plays a key role in tissues that are involved in the regulation of calcium and phosphate homeostasis, such as bone and muscle (242). Epidemiological studies showed that vitamin D deficiency is associated with muscle weakness and reduced muscle mass in the elderly population (243, 244). A study investigated the roles of maternal vitamin D on muscle development in newborn rats by examining skeletal muscle transcriptome (245). A maternal vitamin D-deficient diet during pregnancy from pre-conception to lactation reduced the number of muscle cells and increased intercellular

48 space in the gastrocnemius muscles of newborns (245). Muscle transcriptomes revealed that 426 genes were differentially abundant, which are involved in protein catabolism, cell differentiation and proliferation, muscle cell development, and cytoskeleton organization (245).

Multiple micronutrient

One research team evaluated the effect of maternal micronutrient supplementation on the genome-wide methylation rates of offspring in humans (246). In this study, pre- and peri-conceptional multiple micronutrient supplementation in Gambian women changed methylation rates within gene promoters particularly related to immune system development (246). The authors concluded that this finding is important because a previous study showed that the mortality of individuals born during a hunger season was associated with compromised immune systems in this population (246). They reported that epigenetic changes were substantially sex-specific, suggesting differential effects of micronutrients on gene expression related to developmental trajectories by sex (247).

This study also indicated that changes in methylation by supplementation were common, but the effect size of methylation changes were mostly modest, suggesting that subsequent health effects might be a combination of changes in numerous genes (246).

This study also added the evidence of changes in imprinted genes by maternal micronutrient supplementation. It found a significant reduction in methylation in the

DMRs of IGF2R (only in girls) and GTL2-2 (only in boys) in cord blood samples in the multimicronutrient treatment group compared to untreatment group (248). Further studies are needed to confirm that these epigenomic changes lead to changes in gene and protein expression, and ultimately health outcomes.

In summary, evidence from experimental animal studies shows that maternal micronutrient status during peri-conception or throughout pregnancy induces stable changes in the expression of genes or proteins involved in fundamental cellular and biological processes in different types of tissues in offspring.

Table 2.3. Effect of maternal micronutrient on gene/protein expression in offspring using untargeted methods in animal studies

First Maternal Timing of Methods Outcome Effect observed (differentially Timing of author micronutrient Exposure abundant/expressed) outcome exposure measurement

Maloney Methyl-deficient periconcept 2D gel- hepatic 41 proteins involved in energy metabolism male adult rat CA (214) diet (MD)1 ion electrophoresis proteome amino acid metabolism mitochondrial activity, antioxidant defense

Liu J (215) Folic acid during 2D gel- hepatic 11 proteins involved in immune response, piglet at birth supplementation pregnancy electrophoresis proteome energy metabolism, intermediary metabolism, cellular signal transduction, proteolysis, cell migration regulation

Lillycrop Protein-restricted during microarray hepatic 311 (1.3% of liver transcriptome) genes (PR) adult male rat KA (218) diet (PR) and PR pregnancy transcriptome and 191 (0.7% of liver transcriptome) genes with folic acid (PRF). PRF prevented PR-derived changes in (PRF) response to reactive oxygen species and supplementation steroid hormone response

Barua S High dose of during microarray brain cerebral 124 genes involved in neural pathways newborn mice (249) folic acid pregnancy hemisphere neurotransmitters, neuronal-ion channels, transcriptome glutamatergic-synapse, dopamine-serotonin and synaptic plasticity

Li Y (221) Folate deficiency during microarray skeletal 3000~4000 genes involved in almost all piglet diet early-mid muscle biological processes in myogenesis and gestation transcriptome intramuscular fat deposition

Carlson ES Iron deficient diet late fetal microarray hippocampal 25 genes differentially expressed involved in rat during (227) and early transcriptome cell growth, energy metabolism, dendrite postnatal day neonatal morphogenesis, synaptic connectivity, 7-65 days period mammalian target of rapamycin pathway, and network related to Alzheimer disease etiology

Tran PV Iron deficient diet during iTRAQ hippocampal 331 differentially expressed proteins adult rat (229) gestation proteome involved in glutamate signaling pathway, through cellular process for learning and memory, postnatal synaptic proteins-mediated cellular signaling, period neuronal nitric oxide synthase signaling

Swali A Iron deficient diet peri- microarray and whole tissues 7 differentially abundant proteins (cell male (250) conception 2D-gel transcriptome proliferation, protein transport and folding, embryonic rat al period electrophoresis and proteome cytoskeletal remodeling, proteasome complex) 979 up-regulated and 1545 genes down- regulated (initiation of mitosis, apoptosis, the assembly of RNA polymerase II preinitiation complexes and WNT signaling)

Max D Vitamin D prior to microarray muscle 426 differentially expressed genes involved newborn rat (245) deficient diet conception, transcriptome in myoblast development and myogenic during preg differentiation nancy, and lactation

Roy S (251) Tocotrienols supp during microarray brain 51 genes involved in cellular defense fetal rat lementation pregnancy transcriptome mechanisms and lipoprotein metabolism

1Methyl-deficient diet (MD) is diet with insufficient folic acid, choline, and methionine

REFERENCES 1. Anderson NL, Anderson NG. The human plasma proteome: history, character, and diagnostic prospects. Molecular & cellular proteomics : MCP 2002;1(11):845-67. 2. Adkins JN, Varnum SM, Auberry KJ, et al. Toward a human blood serum proteome: analysis by multidimensional separation coupled with mass spectrometry. Molecular & cellular proteomics : MCP 2002;1(12):947-55. 3. Omenn GS, States DJ, Adamski M, et al. Overview of the HUPO Plasma Proteome Project: results from the pilot phase with 35 collaborating laboratories and multiple analytical groups, generating a core dataset of 3020 proteins and a publicly-available database. Proteomics 2005;5(13):3226-45. 4. Nanjappa V, Thomas JK, Marimuthu A, et al. Plasma Proteome Database as a resource for proteomics research: 2014 update. Nucleic acids research 2014;42(Database issue):D959-65. 5. Schaller J, Gerber S, Kaempfer U. Human blood plasma proteins: structure and function: John Wiley & Sons, 2008. 6. Murray RK, Granner DK, Mayes PA, et al. Harper's Illustrated Biochemistry: McGraw-Hill Medical Publishing Division, 2006. 7. Putnam FW. The plasma proteins: structure, function, and genetic control: Academic Press, 1975. 8. Zhang H, Liu AY, Loriaux P, et al. Mass spectrometric detection of tissue proteins in plasma. Molecular & cellular proteomics : MCP 2007;6(1):64-71. 9. Schenk S, Schoenhals GJ, de Souza G, et al. A high confidence, manually validated human blood plasma protein reference set. BMC medical genomics 2008;1:41. 10. Beck M, Schmidt A, Malmstroem J, et al. The quantitative proteome of a human cell line. Molecular systems biology 2011;7:549. 11. Qian WJ, Kaleta DT, Petritis BO, et al. Enhanced detection of low abundance human plasma proteins using a tandem IgY12-SuperMix immunoaffinity separation strategy. Molecular & cellular proteomics : MCP 2008;7(10):1963-73. 12. Ignjatovic V, Lai C, Summerhayes R, et al. Age-related differences in plasma proteins: how plasma proteins change from neonates to adults. PloS one 2011;6(2):e17213. 13. Kim MS, Pinto SM, Getnet D, et al. A draft map of the human proteome. Nature 2014;509(7502):575-81. 14. Scholl PF, Cole RN, Ruczinski I, et al. Maternal serum proteome changes between the first and third trimester of pregnancy in rural southern Nepal. Placenta 2012;33(5):424-32. 15. Buckley CE, 3rd, Dorsey FC. The effect of aging on human serum immunoglobulin concentrations. Journal of immunology 1970;105(4):964-72. 16. Corzett TH, Fodor IK, Choi MW, et al. Statistical analysis of variation in the human plasma proteome. Journal of biomedicine & biotechnology 2010;2010:258494. 17. Kim CX, Bailey KR, Klee GG, et al. Sex and ethnic differences in 47 candidate proteomic markers of cardiovascular disease: the Mayo Clinic proteomic markers of arteriosclerosis study. PloS one 2010;5(2):e9065.

18. Lonberg-Holm K, Sandberg LB, Doleman MS, et al. Classification of malnutrition by statistical analysis of quantitative two-dimensional gel electrophoresis of plasma proteins. Computers and biomedical research, an international journal 1986;19(4):340-60. 19. Johnson AM. Low levels of plasma proteins: malnutrition or inflammation? Clinical chemistry and laboratory medicine : CCLM / FESCC 1999;37(2):91-6. 20. Fuhrman MP, Charney P, Mueller CM. Hepatic proteins and nutrition assessment. Journal of the American Dietetic Association 2004;104(8):1258-64. 21. Duthie SJ, Horgan G, de Roos B, et al. Blood folate status and expression of proteins involved in immune function, inflammation, and coagulation: biochemical and proteomic changes in the plasma of humans in response to long- term synthetic folic acid supplementation. Journal of proteome research 2010;9(4):1941-50. 22. de Roos B, Geelen A, Ross K, et al. Identification of potential serum biomarkers of inflammation and lipid modulation that are altered by fish oil supplementation in healthy volunteers. Proteomics 2008;8(10):1965-74. 23. Garcia-Bailo B, Brenner DR, Nielsen D, et al. Dietary patterns and ethnicity are associated with distinct plasma proteomic groups. The American journal of clinical nutrition 2012;95(2):352-61. 24. Jacobsen HP, Herskind AM, Nielsen BW, et al. IgE in unselected like-sexed monozygotic and dizygotic twins at birth and at 6 to 9 years of age: high but dissimilar genetic influence on IgE levels. The Journal of allergy and clinical immunology 2001;107(4):659-63. 25. Harrela M, Koistinen H, Kaprio J, et al. Genetic and environmental components of interindividual variation in circulating levels of IGF-I, IGF-II, IGFBP-1, and IGFBP-3. The Journal of clinical investigation 1996;98(11):2612-5. 26. de Lange M, Snieder H, Ariens RA, et al. The genetics of haemostasis: a twin study. Lancet 2001;357(9250):101-5. 27. Cesari M, Sartori MT, Patrassi GM, et al. Determinants of plasma levels of plasminogen activator inhibitor-1 : A study of normotensive twins. Arteriosclerosis, thrombosis, and vascular biology 1999;19(2):316-20. 28. Cava F, Gonzalez C, Pascual MJ, et al. Biological variation of interleukin 6 (IL-6) and soluble interleukin 2 receptor (sIL2R) in serum of healthy individuals. Cytokine 2000;12(9):1423-5. 29. Anesi A, Rondanelli M, Trotti R, et al. Biological variability of myoglobin in healthy elderly and younger subjects. Aging 2000;12(3):168-72. 30. Panteghini M, Pagani F. Pre-analytical, analytical and biological sources of variation of lipoprotein(a). European journal of clinical chemistry and clinical biochemistry : journal of the Forum of European Clinical Chemistry Societies 1993;31(1):23-8. 31. Colquhoun DR, Goldman LR, Cole RN, et al. Global screening of human cord blood proteomes for biomarkers of toxic exposure and effect. Environmental health perspectives 2009;117(5):832-8. 32. Kussmann M, Affolter M. Proteomics at the center of nutrigenomics: comprehensive molecular understanding of dietary health effects. Nutrition 2009;25(11-12):1085-93.

33. Kussmann M, Affolter M. Proteomic methods in nutrition. Current opinion in clinical nutrition and metabolic care 2006;9(5):575-83. 34. Ross PL, Huang YN, Marchese JN, et al. Multiplexed protein quantitation in Saccharomyces cerevisiae using amine-reactive isobaric tagging reagents. Molecular & cellular proteomics : MCP 2004;3(12):1154-69. 35. Yan W, Chen SS. Mass spectrometry-based quantitative proteomic profiling. Briefings in functional genomics & proteomics 2005;4(1):27-38. 36. Nesvizhskii AI, Vitek O, Aebersold R. Analysis and validation of proteomic data generated by tandem mass spectrometry. Nature methods 2007;4(10):787-97. 37. Ashburner M, Ball CA, Blake JA, et al. Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nature genetics 2000;25(1):25-9. 38. Rhee SY, Wood V, Dolinski K, et al. Use and misuse of the gene ontology annotations. Nature reviews Genetics 2008;9(7):509-15. 39. Khatri P, Sirota M, Butte AJ. Ten years of pathway analysis: current approaches and outstanding challenges. PLoS computational biology 2012;8(2):e1002375. 40. Subramanian A, Tamayo P, Mootha VK, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proceedings of the National Academy of Sciences of the United States of America 2005;102(43):15545-50. 41. Efron N, Morgan PB, Cameron ID, et al. Oxygen permeability and water content of silicone hydrogel contact lens materials. Optometry and vision science : official publication of the American Academy of Optometry 2007;84(4):328-37. 42. Ceciliani F, Giordano A, Spagnolo V. The systemic reaction during inflammation: the acute-phase proteins. Protein and peptide letters 2002;9(3):211-23. 43. Gabay C, Kushner I. Acute-phase proteins and other systemic responses to inflammation. The New England journal of medicine 1999;340(6):448-54. 44. Mackiewicz A, Kushner I, Baumann H. Acute Phase Proteins: Molecular Biology, Biochemistry, and Clinical Applications. Florida: CRC Press, 1993. 45. Thurnham DI, McCabe GP. Influence of infection and inflammation on biomarkers of nutritional status with an emphasis on vitamin A and iron. Priorities in the assessment of vitamin A and iron status in populations. Geneva: World Health Organization, 2012. 46. Heinrich PC, Castell JV, Andus T. Interleukin-6 and the acute phase response. The Biochemical journal 1990;265(3):621-36. 47. Ceciliani F, Pocacqua V. The acute phase protein alpha1-acid glycoprotein: a model for altered glycosylation during diseases. Current protein & peptide science 2007;8(1):91-108. 48. Fournier T, Medjoubi NN, Porquet D. Alpha-1-acid glycoprotein. Biochimica et biophysica acta 2000;1482(1-2):157-71. 49. Brinkman-Van der Linden EC, Varki A. New aspects of siglec binding specificities, including the significance of fucosylation and of the sialyl-Tn epitope. Sialic acid-binding immunoglobulin superfamily lectins. The Journal of biological chemistry 2000;275(12):8625-32. 50. Hochepied T, Berger FG, Baumann H, et al. Alpha(1)-acid glycoprotein: an acute phase protein with inflammatory and immunomodulating properties. Cytokine & growth factor reviews 2003;14(1):25-34.

51. Gabay C, Kushner I. Acute-phase proteins. Internet: http://www.els.net [doi: 10.1038/npg.els.0000497]. 52. Means RT, Jr. Pathogenesis of the anemia of chronic disease: a cytokine-mediated anemia. Stem cells 1995;13(1):32-7. 53. Zeidler C, Kanz L, Hurkuck F, et al. In vivo effects of interleukin-6 on thrombopoiesis in healthy and irradiated primates. Blood 1992;80(11):2740-5. 54. Leary AG, Ikebuchi K, Hirai Y, et al. Synergism between interleukin-6 and interleukin-3 in supporting proliferation of human hematopoietic stem cells: comparison with interleukin-1 alpha. Blood 1988;71(6):1759-63. 55. Dinarello CA. Cytokines as endogenous pyrogens. The Journal of infectious diseases 1999;179 Suppl 2:S294-304. 56. Chrousos GP. The hypothalamic-pituitary-adrenal axis and immune-mediated inflammation. The New England journal of medicine 1995;332(20):1351-62. 57. Baumann H, Richards C, Gauldie J. Interaction among hepatocyte-stimulating factors, interleukin 1, and glucocorticoids for regulation of acute phase plasma proteins in human hepatoma (HepG2) cells. Journal of immunology 1987;139(12):4122-8. 58. Wolf M, Bohm S, Brand M, et al. Proinflammatory cytokines interleukin 1 beta and tumor necrosis factor alpha inhibit growth hormone stimulation of insulin-like growth factor I synthesis and growth hormone receptor mRNA levels in cultured rat liver cells. European journal of endocrinology / European Federation of Endocrine Societies 1996;135(6):729-37. 59. Hardardottir I, Grunfeld C, Feingold KR. Effects of endotoxin and cytokines on lipid metabolism. Current opinion in lipidology 1994;5(3):207-15. 60. Moldawer LL, Copeland EM, 3rd. Proinflammatory cytokines, nutritional support, and the cachexia syndrome: interactions and therapeutic options. Cancer 1997;79(9):1828-39. 61. Tai H, Miyaura C, Pilbeam CC, et al. Transcriptional induction of cyclooxygenase-2 in osteoblasts is involved in interleukin-6-induced osteoclast formation. Endocrinology 1997;138(6):2372-9. 62. Nathan C, Ding A. Nonresolving inflammation. Cell 2010;140(6):871-82. 63. Korpos E, Wu C, Sorokin L. Multiple roles of the extracellular matrix in inflammation. Current pharmaceutical design 2009;15(12):1349-57. 64. Diegelmann RF, Evans MC. Wound healing: an overview of acute, fibrotic and delayed healing. Frontiers in bioscience : a journal and virtual library 2004;9:283- 9. 65. Midwood KS, Williams LV, Schwarzbauer JE. Tissue repair and the dynamics of the extracellular matrix. The international journal of biochemistry & cell biology 2004;36(6):1031-7. 66. Carmeliet P. Angiogenesis in life, disease and medicine. Nature 2005;438(7070):932-6. 67. Carmeliet P, Jain RK. Angiogenesis in cancer and other diseases. Nature 2000;407(6801):249-57. 68. Sorokin L. The impact of the extracellular matrix on inflammation. Nature reviews Immunology 2010;10(10):712-23.

69. Murdoch C, Muthana M, Lewis CE. Hypoxia regulates macrophage functions in inflammation. Journal of immunology 2005;175(10):6257-63. 70. Costa C, Incio J, Soares R. Angiogenesis and chronic inflammation: cause or consequence? Angiogenesis 2007;10(3):149-66. 71. Straub RH, Cutolo M, Buttgereit F, et al. Energy regulation and neuroendocrine- immune control in chronic inflammatory diseases. Journal of internal medicine 2010;267(6):543-60. 72. Yoo JY, Desiderio S. Innate and acquired immunity intersect in a global view of the acute-phase response. Proceedings of the National Academy of Sciences of the United States of America 2003;100(3):1157-62. 73. Kelly-Spratt KS, Pitteri SJ, Gurley KE, et al. Plasma proteome profiles associated with inflammation, angiogenesis, and cancer. PloS one 2011;6(5):e19721. 74. Saha S, Harrison SH, Chen JY. Dissecting the human plasma proteome and inflammatory response biomarkers. Proteomics 2009;9(2):470-84. 75. Fernald LCH, Kariger P, P. E, et al. Examining Early Child Development in Low- Income Countries. Washington, D.C.: World Bank, 2009. 76. Development CoItSoEC. From Neurons to Neighborhoods. Washington, D.C.: National Research Council and Institute of Medicine, 2000. 77. Giedd JN. Structural magnetic resonance imaging of the adolescent brain. Annals of the New York Academy of Sciences 2004;1021:77-85. 78. Hughes D, Bryan J. The assessment of cognitive performance in children: considerations for detecting nutritional influences. Nutrition reviews 2003;61(12):413-22. 79. Goldman-Rakic PS. Development of cortical circuitry and cognitive function. Child development 1987;58(3):601-22. 80. Rabbitt P. Methodology Of Frontal And Executive Function: Taylor & Francis, 1997. 81. Walker SP, Wachs TD, Gardner JM, et al. Child development: risk factors for adverse outcomes in developing countries. Lancet 2007;369(9556):145-57. 82. Engle PL, Black MM, Behrman JR, et al. Strategies to avoid the loss of developmental potential in more than 200 million children in the developing world. Lancet 2007;369(9557):229-42. 83. Development CISEC, Phillips DA, Shonkoff JP, et al. From Neurons to Neighborhoods:: The Science of Early Childhood Development: National Academies Press, 2000. 84. Turkheimer E, Haley A, Waldron M, et al. Socioeconomic status modifies heritability of IQ in young children. Psychological science 2003;14(6):623-8. 85. Bradley RH, Corwyn RF. Socioeconomic status and child development. Annual review of psychology 2002;53:371-99. 86. McLoyd VC. Socioeconomic disadvantage and child development. The American psychologist 1998;53(2):185-204. 87. Guo G, Harris KM. The mechanisms mediating the effects of poverty on children's intellectual development. Demography 2000;37(4):431-47. 88. Walker SP, Wachs TD, Grantham-McGregor S, et al. Inequality in early childhood: risk and protective factors for early child development. Lancet 2011;378(9799):1325-38.

89. Martorell R, Horta BL, Adair LS, et al. Weight gain in the first two years of life is an important predictor of schooling outcomes in pooled analyses from five birth cohorts from low- and middle-income countries. The Journal of nutrition 2010;140(2):348-54. 90. Shafir T, Angulo-Barroso R, Calatroni A, et al. Effects of iron deficiency in infancy on patterns of motor development over time. Human movement science 2006;25(6):821-38. 91. Lukowski AF, Koss M, Burden MJ, et al. Iron deficiency in infancy and neurocognitive functioning at 19 years: evidence of long-term deficits in executive function and recognition memory. Nutritional neuroscience 2010;13(2):54-70. 92. Corapci F, Calatroni A, Kaciroti N, et al. Longitudinal evaluation of externalizing and internalizing behavior problems following iron deficiency in infancy. Journal of pediatric psychology 2010;35(3):296-305. 93. Bleichrodt N. Developmental disorders associated with severe IDD. Edtion ed. In: Hetzel BS, Dunn JT, Stansbury JB, eds. The prevention and control of iodine deficiency disorders. Amsterdam: Elsevier, 1987. 94. Qian M, Wang D, Watkins WE, et al. The effects of iodine on intelligence in children: a meta-analysis of studies conducted in China. Asia Pacific journal of clinical nutrition 2005;14(1):32-42. 95. Checkley W, Buckley G, Gilman RH, et al. Multi-country analysis of the effects of diarrhoea on childhood stunting. International journal of epidemiology 2008;37(4):816-30. 96. Kihara M, Carter JA, Newton CR. The effect of Plasmodium falciparum on cognition: a systematic review. Tropical medicine & international health : TM & IH 2006;11(4):386-97. 97. Thuilliez J. Malaria and Primary Education: A cross-country analysis on primary repetition and completion rates. Documents de travail du Centre d'Economie de la Sorbonne, Université Panthéon-Sorbonne 2009;2:127-57. 98. Sherr L, Mueller J, Varrall R. A systematic review of cognitive development and child human immunodeficiency virus infection. Psychology, health & medicine 2009;14(4):387-404. 99. Fernando D, de Silva D, Carter R, et al. A randomized, double-blind, placebo- controlled, clinical trial of the impact of malaria prevention on the educational attainment of school children. The American journal of tropical medicine and hygiene 2006;74(3):386-93. 100. Clarke SE, Jukes MC, Njagi JK, et al. Effect of intermittent preventive treatment of malaria on health and education in schoolchildren: a cluster-randomised, double-blind, placebo-controlled trial. Lancet 2008;372(9633):127-38. 101. Van Rie A, Dow A, Mupuala A, et al. Neurodevelopmental trajectory of HIV- infected children accessing care in Kinshasa, Democratic Republic of Congo. Journal of acquired immune deficiency syndromes 2009;52(5):636-42. 102. Berkman DS, Lescano AG, Gilman RH, et al. Effects of stunting, diarrhoeal disease, and parasitic infection during infancy on cognition in late childhood: a follow-up study. Lancet 2002;359(9306):564-71.

103. Patel SA, Murray-Kolb LE, LeClerq SC, et al. Household wealth and neurocognitive development disparities among school-aged children in Nepal. Paediatric and perinatal epidemiology 2013;27(6):575-86. 104. Gomez-Pinilla F. Brain foods: the effects of nutrients on brain function. Nature reviews Neuroscience 2008;9(7):568-78. 105. Nyberg F, Hallberg M. Growth hormone and cognitive function. Nature reviews Endocrinology 2013;9(6):357-65. 106. Rabin BS, Cohen S, Ganguli R, et al. Bidirectional interaction between the central nervous system and the immune system. Critical reviews in immunology 1989;9(4):279-312. 107. Neuwelt EA, Bauer B, Fahlke C, et al. Engaging neuroscience to advance translational research in brain barrier biology. Nature reviews Neuroscience 2011;12(3):169-82. 108. Bus BA, Molendijk ML, Penninx BJ, et al. Determinants of serum brain-derived neurotrophic factor. Psychoneuroendocrinology 2011;36(2):228-39. 109. Ohlsson C, Mohan S, Sjogren K, et al. The Role of Liver-Derived Insulin-Like Growth Factor-I. Endocrine reviews 2009;30(5):494-535. 110. Aberg ND, Brywe KG, Isgaard J. Aspects of growth hormone and insulin-like growth factor-I related to neuroprotection, regeneration, and functional plasticity in the adult brain. TheScientificWorldJournal 2006;6:53-80. 111. Beck KD, Powell-Braxton L, Widmer HR, et al. Igf1 gene disruption results in reduced brain size, CNS hypomyelination, and loss of hippocampal granule and striatal parvalbumin-containing neurons. Neuron 1995;14(4):717-30. 112. Carson MJ, Behringer RR, Brinster RL, et al. Insulin-like growth factor I increases brain growth and central nervous system myelination in transgenic mice. Neuron 1993;10(4):729-40. 113. Lindenberg R, Renga V, Zhu LL, et al. Structural integrity of corticospinal motor fibers predicts motor impairment in chronic stroke. Neurology 2010;74(4):280-7. 114. Gunnell D, Miller LL, Rogers I, et al. Association of insulin-like growth factor I and insulin-like growth factor-binding protein-3 with intelligence quotient among 8- to 9-year-old children in the Avon Longitudinal Study of Parents and Children. Pediatrics 2005;116(5):e681-6. 115. Chen DY, Stern SA, Garcia-Osta A, et al. A critical role for IGF-II in memory consolidation and enhancement. Nature 2011;469(7331):491-7. 116. Nyberg F, Burman P. Growth hormone and its receptors in the central nervous system--location and functional significance. Hormone research 1996;45(1-2):18- 22. 117. van Pareren YK, Duivenvoorden HJ, Slijper FS, et al. Intelligence and psychosocial functioning during long-term growth hormone therapy in children born small for gestational age. The Journal of clinical endocrinology and metabolism 2004;89(11):5295-302. 118. Webb EA, O'Reilly MA, Clayden JD, et al. Effect of growth hormone deficiency on brain structure, motor function and cognition. Brain : a journal of neurology 2012;135(Pt 1):216-27. 119. Thissen JP, Ketelslegers JM, Underwood LE. Nutritional regulation of the insulin-like growth factors. Endocrine reviews 1994;15(1):80-101.

120. Chrousos GP. The role of stress and the hypothalamic-pituitary-adrenal axis in the pathogenesis of the metabolic syndrome: neuro-endocrine and target tissue-related causes. International journal of obesity and related metabolic disorders : journal of the International Association for the Study of Obesity 2000;24 Suppl 2:S50-5. 121. O'Shea TM, Doyle LW. Perinatal glucocorticoid therapy and neurodevelopmental outcome: an epidemiologic perspective. Seminars in neonatology : SN 2001;6(4):293-307. 122. Yeh TF, Lin YJ, Lin HC, et al. Outcomes at school age after postnatal dexamethasone therapy for lung disease of prematurity. The New England journal of medicine 2004;350(13):1304-13. 123. Fukumoto K, Morita T, Mayanagi T, et al. Detrimental effects of glucocorticoids on neuronal migration during brain development. Molecular psychiatry 2009;14(12):1119-31. 124. Caldji C, Tannenbaum B, Sharma S, et al. Maternal care during infancy regulates the development of neural systems mediating the expression of fearfulness in the rat. Proceedings of the National Academy of Sciences of the United States of America 1998;95(9):5335-40. 125. McGowan PO, Sasaki A, D'Alessio AC, et al. Epigenetic regulation of the glucocorticoid receptor in human brain associates with childhood abuse. Nature neuroscience 2009;12(3):342-8. 126. Hackman DA, Farah MJ, Meaney MJ. Socioeconomic status and the brain: mechanistic insights from human and animal research. Nature reviews Neuroscience 2010;11(9):651-9. 127. McNay EC. Insulin and ghrelin: peripheral hormones modulating memory and hippocampal function. Current opinion in pharmacology 2007;7(6):628-32. 128. Diano S, Farr SA, Benoit SC, et al. Ghrelin controls hippocampal spine synapse density and memory performance. Nature neuroscience 2006;9(3):381-8. 129. Harvey J. Leptin regulation of neuronal excitability and cognitive function. Current opinion in pharmacology 2007;7(6):643-7. 130. Li XL, Aou S, Oomura Y, et al. Impairment of long-term potentiation and spatial memory in leptin receptor-deficient rodents. Neuroscience 2002;113(3):607-15. 131. Nawa H, Takei N. Recent progress in animal modeling of immune inflammatory processes in schizophrenia: implication of specific cytokines. Neuroscience research 2006;56(1):2-13. 132. Stevens B, Allen NJ, Vazquez LE, et al. The classical complement cascade mediates CNS synapse elimination. Cell 2007;131(6):1164-78. 133. Bilbo SD, Biedenkapp JC, Der-Avakian A, et al. Neonatal infection-induced memory impairment after lipopolysaccharide in adulthood is prevented via caspase-1 inhibition. The Journal of neuroscience : the official journal of the Society for Neuroscience 2005;25(35):8000-9. 134. Hornig M, Weissenbock H, Horscroft N, et al. An infection-based model of neurodevelopmental damage. Proceedings of the National Academy of Sciences of the United States of America 1999;96(21):12102-7. 135. Barrientos RM, Sprunger DB, Campeau S, et al. BDNF mRNA expression in rat hippocampus following contextual learning is blocked by intrahippocampal IL- 1beta administration. Journal of neuroimmunology 2004;155(1-2):119-26.

136. Schneider H, Pitossi F, Balschun D, et al. A neuromodulatory role of interleukin- 1beta in the hippocampus. Proceedings of the National Academy of Sciences of the United States of America 1998;95(13):7778-83. 137. Cunningham ET, Jr., De Souza EB. Interleukin 1 receptors in the brain and endocrine tissues. Immunology today 1993;14(4):171-6. 138. Bilbo SD, Barrientos RM, Eads AS, et al. Early-life infection leads to altered BDNF and IL-1beta mRNA expression in rat hippocampus following learning in adulthood. Brain, behavior, and immunity 2008;22(4):451-5. 139. Bilbo SD, Schwarz JM. Early-life programming of later-life brain and behavior: a critical role for the immune system. Frontiers in behavioral neuroscience 2009;3:14. 140. Editorial. Focus on neuro-immune interactions. Nature neuroscience 2012;15:1055. 141. Franco R, Pacheco R, Lluis C, et al. The emergence of neurotransmitters as immune modulators. Trends in immunology 2007;28(9):400-7. 142. Wikipedia. Dopamine beta hydroxylase. Internet: http://en.wikipedia.org/wiki/Dopamine_beta_hydroxylase. 143. Thomas SA, Palmiter RD. Disruption of the dopamine beta-hydroxylase gene in mice suggests roles for norepinephrine in motor function, learning, and memory. Behavioral neuroscience 1997;111(3):579-89. 144. Lozoff B, Beard J, Connor J, et al. Long-lasting neural and behavioral effects of iron deficiency in infancy. Nutrition reviews 2006;64(5 Pt 2):S34-43; discussion S72-91. 145. Paclt I, Koudelova J, Pacltova D, et al. Dopamine beta hydroxylase (DBH) plasma activity in childhood mental disorders. Neuro endocrinology letters 2009;30(5):604-9. 146. Ji N, Shuai L, Chen Y, et al. Dopamine beta-hydroxylase gene associates with stroop color-word task performance in Han Chinese children with attention deficit/hyperactivity disorder. American journal of medical genetics Part B, Neuropsychiatric genetics : the official publication of the International Society of Psychiatric Genetics 2011;156B(6):730-6. 147. Bianco LE, Wiesinger J, Earley CJ, et al. Iron deficiency alters dopamine uptake and response to L-DOPA injection in Sprague-Dawley rats. Journal of neurochemistry 2008;106(1):205-15. 148. Poirier J, Sevigny P. Apolipoprotein E4, cholinergic integrity and the pharmacogenetics of Alzheimer's disease. Journal of neural transmission Supplementum 1998;53:199-207. 149. Dietschy JM, Turley SD. Cholesterol metabolism in the brain. Current opinion in lipidology 2001;12(2):105-12. 150. Ambrose JW. APOLIPOPROTEIN E AND ITS RECEPTORS: common roles in development and dementia? Impulse: The Premier Journal for Undergraduate Publications in the Neurosciences 2004;1(1):1-58. 151. Gordon I, Grauer E, Genis I, et al. Memory deficits and cholinergic impairments in apolipoprotein E-deficient mice. Neuroscience letters 1995;199(1):1-4. 152. Oitzl MS, Mulder M, Lucassen PJ, et al. Severe learning deficits in apolipoprotein E-knockout mice in a water maze task. Brain research 1997;752(1-2):189-96.

153. Wright RO, Hu H, Silverman EK, et al. Apolipoprotein E genotype predicts 24- month bayley scales infant development score. Pediatric research 2003;54(6):819- 25. 154. Rask-Nissila L, Jokinen E, Terho P, et al. Effects of diet on the neurologic development of children at 5 years of age: the STRIP project. The Journal of pediatrics 2002;140(3):328-33. 155. Woods AG, Sokolowska I, Taurines R, et al. Potential biomarkers in psychiatry: focus on the cholesterol system. Journal of cellular and molecular medicine 2012;16(6):1184-95. 156. Perry LA, Stigger CB, Ainsworth BE, et al. No association between cognitive achievements, academic performance and serum cholesterol concentrations among school-aged children. Nutritional neuroscience 2009;12(4):160-6. 157. Black RE, Allen LH, Bhutta ZA, et al. Maternal and child undernutrition: global and regional exposures and health consequences. Lancet 2008;371(9608):243-60. 158. Torheim LE, Ferguson EL, Penrose K, et al. Women in resource-poor settings are at risk of inadequate intakes of multiple micronutrients. The Journal of nutrition 2010;140(11):2051S-8S. 159. McLean E, Cogswell M, Egli I, et al. Worldwide prevalence of anaemia, WHO Vitamin and Mineral Nutrition Information System, 1993-2005. Public health nutrition 2009;12(4):444-54. 160. World Health Organization. Dept. of Nutrition for Health and Development. Iron deficiency anaemia : assessment, prevention and control : a guide for programme managers. Geneva: World Health Organization, 2001. 161. Caulfield LE, Zavaleta N, Shankar AH, et al. Potential contribution of maternal zinc supplementation during pregnancy to maternal and child survival. The American journal of clinical nutrition 1998;68(2 Suppl):499S-508S. 162. Jiang T, Christian P, Khatry SK, et al. Micronutrient deficiencies in early pregnancy are common, concurrent, and vary by season among rural Nepali pregnant women. The Journal of nutrition 2005;135(5):1106-12. 163. WHO. Daily iron and folic acid supplementation in pregnant women. . Geneva: World Health Organization, 2012. 164. WHO. Guideline: intermittent iron and folic acid supplementation in nonanemic pregnant women. . Geneva: World Health Organization, 2012. 165. Fall CH, Fisher DJ, Osmond C, et al. Multiple micronutrient supplementation during pregnancy in low-income countries: a meta-analysis of effects on birth size and length of gestation. Food and nutrition bulletin 2009;30(4 Suppl):S533-46. 166. West KP, Jr., Shamim AA, Mehra S, et al. Effect of maternal multiple micronutrient vs iron-folic acid supplementation on infant mortality and adverse birth outcomes in rural Bangladesh: the JiVitA-3 randomized trial. Jama 2014;312(24):2649-58. 167. Gluckman PD, Hanson MA. Living with the past: evolution, development, and patterns of disease. Science 2004;305(5691):1733-6. 168. Barker DJ. The origins of the developmental origins theory. Journal of internal medicine 2007;261(5):412-7. 169. Cota BM, Allen PJ. The developmental origins of health and disease hypothesis. Pediatric nursing 2010;36(3):157-67.

170. Hamadani JD, Fuchs GJ, Osendarp SJ, et al. Zinc supplementation during pregnancy and effects on mental development and behaviour of infants: a follow- up study. Lancet 2002;360(9329):290-4. 171. Roberfroid D, Huybregts L, Lanou H, et al. Impact of prenatal multiple micronutrients on survival and growth during infancy: a randomized controlled trial. The American journal of clinical nutrition 2012;95(4):916-24. 172. Villamor E, Saathoff E, Bosch RJ, et al. Vitamin supplementation of HIV- infected women improves postnatal child growth. The American journal of clinical nutrition 2005;81(4):880-8. 173. Huy ND, Le Hop T, Shrimpton R, et al. An effectiveness trial of multiple micronutrient supplementation during pregnancy in Vietnam: impact on birthweight and on stunting in children at around 2 years of age. Food and nutrition bulletin 2009;30(4 Suppl):S506-16. 174. Wang W, Yan H, Zeng L, et al. No effect of maternal micronutrient supplementation on early childhood growth in rural western China: 30 month follow-up evaluation of a double blind, cluster randomized controlled trial. European journal of clinical nutrition 2012;66(2):261-8. 175. Islam Khan A. Effects of pre- and postnatal nutrition interventions on child growth and body composition: the MINIMat trial in rural Bangladesh. Global health action 2013;6:22476. 176. Vaidya A, Saville N, Shrestha BP, et al. Effects of antenatal multiple micronutrient supplementation on children's weight and size at 2 years of age in Nepal: follow-up of a double-blind randomised controlled trial. Lancet 2008;371(9611):492-9. 177. Devakumar D, Chaube SS, Wells JC, et al. Effect of antenatal multiple micronutrient supplementation on anthropometry and blood pressure in mid- childhood in Nepal: follow-up of a double-blind randomised controlled trial. The Lancet Global health 2014;2(11):e654-63. 178. Stewart CP, Christian P, LeClerq SC, et al. Antenatal supplementation with folic acid + iron + zinc improves linear growth and reduces peripheral adiposity in school-age children in rural Nepal. The American journal of clinical nutrition 2009;90(1):132-40. 179. McGrath N, Bellinger D, Robins J, et al. Effect of maternal multivitamin supplementation on the mental and psychomotor development of children who are born to HIV-1-infected mothers in Tanzania. Pediatrics 2006;117(2):e216-25. 180. Li Q, Yan H, Zeng L, et al. Effects of maternal multimicronutrient supplementation on the mental development of infants in rural western China: follow-up evaluation of a double-blind, randomized, controlled trial. Pediatrics 2009;123(4):e685-92. 181. Prado EL, Alcock KJ, Muadz H, et al. Maternal multiple micronutrient supplements and child cognition: a randomized trial in Indonesia. Pediatrics 2012;130(3):e536-46. 182. Caulfield LE, Zavaleta N, Chen P, et al. Maternal zinc supplementation during pregnancy affects autonomic function of Peruvian children assessed at 54 months of age. The Journal of nutrition 2011;141(2):327-32.

183. Caulfield LE, Putnick DL, Zavaleta N, et al. Maternal gestational zinc supplementation does not influence multiple aspects of child development at 54 mo of age in Peru. The American journal of clinical nutrition 2010;92(1):130-6. 184. Christian P, Murray-Kolb LE, Khatry SK, et al. Prenatal micronutrient supplementation and intellectual and motor function in early school-aged children in Nepal. Jama 2010;304(24):2716-23. 185. Andersen GS, Friis H, Michaelsen KF, et al. Effects of maternal micronutrient supplementation on fetal loss and under-2-years child mortality: long-term follow-up of a randomised controlled trial from Guinea-Bissau. African journal of reproductive health 2010;14(2):17-26. 186. Shaheen R, Streatfield PK, Naved RT, et al. Equity in adherence to and effect of prenatal food and micronutrient supplementation on child mortality: results from the MINIMat randomized trial, Bangladesh. BMC public health 2014;14:5. 187. Christian P, Stewart CP, LeClerq SC, et al. Antenatal and postnatal iron supplementation and childhood mortality in rural Nepal: a prospective follow-up in a randomized, controlled community trial. American journal of epidemiology 2009;170(9):1127-36. 188. Hawkesworth S, Wagatsuma Y, Kahn AI, et al. Combined food and micronutrient supplements during pregnancy have limited impact on child blood pressure and kidney function in rural Bangladesh. The Journal of nutrition 2013;143(5):728-34. 189. Stewart CP, Christian P, Schulze KJ, et al. Antenatal micronutrient supplementation reduces metabolic syndrome in 6- to 8-year-old children in rural Nepal. The Journal of nutrition 2009;139(8):1575-81. 190. Gluckman PD, Hanson MA, Pinal C. The developmental origins of adult disease. Maternal & child nutrition 2005;1(3):130-41. 191. Bateson P, Barker D, Clutton-Brock T, et al. Developmental plasticity and human health. Nature 2004;430(6998):419-21. 192. Roseboom TJ, Painter RC, van Abeelen AF, et al. Hungry in the womb: what are the consequences? Lessons from the Dutch famine. Maturitas 2011;70(2):141-5. 193. Wolffe AP, Matzke MA. Epigenetics: regulation through repression. Science 1999;286(5439):481-6. 194. Bird A. DNA methylation patterns and epigenetic memory. Genes & development 2002;16(1):6-21. 195. Hochberg Z, Feil R, Constancia M, et al. Child health, developmental plasticity, and epigenetic programming. Endocrine reviews 2011;32(2):159-224. 196. Waterland RA, Jirtle RL. Transposable elements: targets for early nutritional effects on epigenetic gene regulation. Molecular and cellular biology 2003;23(15):5293-300. 197. Steegers-Theunissen RP, Obermann-Borst SA, Kremer D, et al. Periconceptional maternal folic acid use of 400 microg per day is related to increased methylation of the IGF2 gene in the very young child. PloS one 2009;4(11):e7845. 198. Hoyo C, Murtha AP, Schildkraut JM, et al. Methylation variation at IGF2 differentially methylated regions and maternal folic acid use before and during pregnancy. Epigenetics : official journal of the DNA Methylation Society 2011;6(7):928-36.

199. Pidsley R, Dempster E, Troakes C, et al. Epigenetic and genetic variation at the IGF2/H19 imprinting control region on 11p15.5 is associated with cerebellum weight. Epigenetics : official journal of the DNA Methylation Society 2012;7(2):155-63. 200. Edwards CR, Benediktsson R, Lindsay RS, et al. 11 beta-Hydroxysteroid dehydrogenases: key enzymes in determining tissue-specific glucocorticoid effects. Steroids 1996;61(4):263-9. 201. Lingas R, Dean F, Matthews SG. Maternal nutrient restriction (48 h) modifies brain corticosteroid receptor expression and endocrine function in the fetal guinea pig. Brain research 1999;846(2):236-42. 202. Langley-Evans SC. Nutritional programming of disease: unravelling the mechanism. Journal of anatomy 2009;215(1):36-51. 203. Weaver IC, Cervoni N, Champagne FA, et al. Epigenetic programming by maternal behavior. Nature neuroscience 2004;7(8):847-54. 204. Seckl JR, Meaney MJ. Glucocorticoid programming. Annals of the New York Academy of Sciences 2004;1032:63-84. 205. Rosario JF, Gomez MP, Anbu P. Does the maternal micronutrient deficiency (copper or zinc or vitamin E) modulate the expression of placental 11 beta hydroxysteroid dehydrogenase-2 per se predispose offspring to insulin resistance and hypertension in later life? Indian journal of physiology and pharmacology 2008;52(4):355-65. 206. Brenseke B, Prater MR, Bahamonde J, et al. Current thoughts on maternal nutrition and fetal programming of the metabolic syndrome. Journal of pregnancy 2013;2013:368461. 207. Luo ZC, Fraser WD, Julien P, et al. Tracing the origins of "fetal origins" of adult diseases: programming by oxidative stress? Medical hypotheses 2006;66(1):38-44. 208. Gupta P, Narang M, Banerjee BD, et al. Oxidative stress in term small for gestational age neonates born to undernourished mothers: a case control study. BMC pediatrics 2004;4:14. 209. Dudeja PK, Kode A, Alnounou M, et al. Mechanism of folate transport across the human colonic basolateral membrane. American journal of physiology Gastrointestinal and liver physiology 2001;281(1):G54-60. 210. Bailey LB, Gregory JF, 3rd. Folate metabolism and requirements. The Journal of nutrition 1999;129(4):779-82. 211. Fekete K, Berti C, Cetin I, et al. Perinatal folate supply: relevance in health outcome parameters. Maternal & child nutrition 2010;6 Suppl 2:23-38. 212. Maloney CA, Hay SM, Young LE, et al. A methyl-deficient diet fed to rat dams during the peri-conception period programs glucose homeostasis in adult male but not female offspring. The Journal of nutrition 2011;141(1):95-100. 213. Sinclair KD, Allegrucci C, Singh R, et al. DNA methylation, insulin resistance, and blood pressure in offspring determined by maternal periconceptional B vitamin and methionine status. Proceedings of the National Academy of Sciences of the United States of America 2007;104(49):19351-6. 214. Maloney CA, Hay SM, Reid MD, et al. A methyl-deficient diet fed to rats during the pre- and peri-conception periods of development modifies the hepatic proteome in the adult offspring. Genes & nutrition 2013;8(2):181-90.

215. Liu J, Yao Y, Yu B, et al. Effect of maternal folic acid supplementation on hepatic proteome in newborn piglets. Nutrition 2013;29(1):230-4. 216. Lillycrop KA, Phillips ES, Jackson AA, et al. Dietary protein restriction of pregnant rats induces and folic acid supplementation prevents epigenetic modification of hepatic gene expression in the offspring. The Journal of nutrition 2005;135(6):1382-6. 217. Bertram C, Trowern AR, Copin N, et al. The maternal diet during pregnancy programs altered expression of the glucocorticoid receptor and type 2 11beta- hydroxysteroid dehydrogenase: potential molecular mechanisms underlying the programming of hypertension in utero. Endocrinology 2001;142(7):2841-53. 218. Lillycrop KA, Rodford J, Garratt ES, et al. Maternal protein restriction with or without folic acid supplementation during pregnancy alters the hepatic transcriptome in adult male rats. The British journal of nutrition 2010;103(12):1711-9. 219. Wang J, Chen L, Li D, et al. Intrauterine growth restriction affects the proteomes of the small intestine, liver, and skeletal muscle in newborn pigs. The Journal of nutrition 2008;138(1):60-6. 220. Sram RJ, Binkova B, Lnenickova Z, et al. The impact of plasma folate levels of mothers and newborns on intrauterine growth retardation and birth weight. Mutation research 2005;591(1-2):302-10. 221. Li Y, Zhang X, Sun Y, et al. Folate deficiency during early-mid pregnancy affects the skeletal muscle transcriptome of piglets from a reciprocal cross. PloS one 2013;8(12):e82616. 222. Granell R, Heron J, Lewis S, et al. The association between mother and child MTHFR C677T polymorphisms, dietary folate intake and childhood atopy in a population-based, longitudinal birth cohort. Clinical and experimental allergy : journal of the British Society for Allergy and Clinical Immunology 2008;38(2):320-8. 223. Haberg SE, London SJ, Stigum H, et al. Folic acid supplements in pregnancy and early childhood respiratory health. Archives of disease in childhood 2009;94(3):180-4. 224. Andrews NC. Iron homeostasis: insights from genetics and animal models. Nature reviews Genetics 2000;1(3):208-17. 225. Wigglesworth JM, Baum H. Iron dependent enzymes in the brain. New York, 1988. 226. Rao R, Georgieff MK. Iron in fetal and neonatal nutrition. Seminars in fetal & neonatal medicine 2007;12(1):54-63. 227. Carlson ES, Stead JD, Neal CR, et al. Perinatal iron deficiency results in altered developmental expression of genes mediating energy metabolism and neuronal morphogenesis in hippocampus. Hippocampus 2007;17(8):679-91. 228. Geguchadze RN, Coe CL, Lubach GR, et al. CSF proteomic analysis reveals persistent iron deficiency-induced alterations in non-human primate infants. Journal of neurochemistry 2008;105(1):127-36. 229. Tran PV, Dakoji S, Reise KH, et al. Fetal iron deficiency alters the proteome of adult rat hippocampal synaptosomes. American journal of physiology Regulatory, integrative and comparative physiology 2013;305(11):R1297-306.

230. Allen LH. Anemia and iron deficiency: effects on pregnancy outcome. The American journal of clinical nutrition 2000;71(5 Suppl):1280S-4S. 231. Bastian TW, Prohaska JR, Georgieff MK, et al. Perinatal iron and copper deficiencies alter neonatal rat circulating and brain thyroid hormone concentrations. Endocrinology 2010;151(8):4055-65. 232. Georgieff MK. The role of iron in neurodevelopment: fetal iron deficiency and the developing hippocampus. Biochemical Society transactions 2008;36(Pt 6):1267-71. 233. Wu LL, Zhang L, Shao J, et al. Effect of perinatal iron deficiency on myelination and associated behaviors in rat pups. Behavioural brain research 2008;188(2):263- 70. 234. Keen CL, Clegg MS, Hanna LA, et al. The plausibility of micronutrient deficiencies being a significant contributing factor to the occurrence of pregnancy complications. The Journal of nutrition 2003;133(5 Suppl 2):1597S-605S. 235. Klug A. The discovery of zinc fingers and their applications in gene regulation and genome manipulation. Annual review of biochemistry 2010;79:213-31. 236. Clegg MS, Hanna LA, Niles BJ, et al. Zinc deficiency-induced cell death. IUBMB life 2005;57(10):661-9. 237. Hanna LA, Clegg MS, Ellis-Hutchings RG, et al. The influence of gestational zinc deficiency on the fetal insulin-like growth factor axis in the rat. Experimental biology and medicine 2010;235(2):206-14. 238. Adamo AM, Oteiza PI. Zinc deficiency and neurodevelopment: the case of neurons. BioFactors 2010;36(2):117-24. 239. Wang FD, Bian W, Kong LW, et al. Maternal zinc deficiency impairs brain nestin expression in prenatal and postnatal mice. Cell research 2001;11(2):135-41. 240. Dvergsten CL, Johnson LA, Sandstead HH. Alterations in the postnatal development of the cerebellar cortex due to zinc deficiency. III. Impaired dendritic differentiation of basket and stellate cells. Brain research 1984;318(1):21-6. 241. Dvergsten CL, Fosmire GJ, Ollerich DA, et al. Alterations in the postnatal development of the cerebellar cortex due to zinc deficiency. II. Impaired maturation of Purkinje cells. Brain research 1984;318(1):11-20. 242. Bischoff HA, Borchers M, Gudat F, et al. In situ detection of 1,25- dihydroxyvitamin D3 receptor in human skeletal muscle tissue. The Histochemical journal 2001;33(1):19-24. 243. Visser M, Deeg DJ, Lips P, et al. Low vitamin D and high parathyroid hormone levels as determinants of loss of muscle strength and muscle mass (sarcopenia): the Longitudinal Aging Study Amsterdam. The Journal of clinical endocrinology and metabolism 2003;88(12):5766-72. 244. Bischoff-Ferrari HA, Dietrich T, Orav EJ, et al. Higher 25-hydroxyvitamin D concentrations are associated with better lower-extremity function in both active and inactive persons aged > or =60 y. The American journal of clinical nutrition 2004;80(3):752-8. 245. Max D, Brandsch C, Schumann S, et al. Maternal vitamin D deficiency causes smaller muscle fibers and altered transcript levels of genes involved in protein

degradation, myogenesis, and cytoskeleton organization in the newborn rat. Molecular nutrition & food research 2014;58(2):343-52. 246. Khulan B, Cooper WN, Skinner BM, et al. Periconceptional maternal micronutrient supplementation is associated with widespread gender related changes in the epigenome: a study of a unique resource in the Gambia. Human molecular genetics 2012;21(9):2086-101. 247. Tobi EW, Lumey LH, Talens RP, et al. DNA methylation differences after exposure to prenatal famine are common and timing- and sex-specific. Human molecular genetics 2009;18(21):4046-53. 248. Cooper WN, Khulan B, Owens S, et al. DNA methylation profiling at imprinted loci after periconceptional micronutrient supplementation in humans: results of a pilot randomized controlled trial. FASEB journal : official publication of the Federation of American Societies for Experimental Biology 2012;26(5):1782-90. 249. Barua S, Chadman KK, Kuizon S, et al. Increasing maternal or post-weaning folic acid alters gene expression and moderately changes behavior in the offspring. PloS one 2014;9(7):e101674. 250. Swali A, McMullen S, Hayes H, et al. Processes underlying the nutritional programming of embryonic development by iron deficiency in the rat. PloS one 2012;7(10):e48133. 251. Roy S, Lado BH, Khanna S, et al. Vitamin E sensitive genes in the developing rat fetal brain: a high-density oligonucleotide microarray analysis. FEBS letters 2002;530(1-3):17-23.

3 CHAPTER 3: STUDY DESIGN AND METHODS

The specific aims of this study are to identify plasma proteins associated with

childhood inflammation, cognitive function, and prenatal micronutrient supplementation

in a malnourished rural South Asian population. This study was based on a double-blind,

controlled trial of antenatal micronutrient supplementation. Pregnant women were

randomized to receive one of 5 micronutrient supplements from mid-pregnancy to 12

weeks postpartum. The first child follow-up study was conducted when children were 6-8

years of age for health and nutritional evaluations including blood sampling. From the

blood samples, the plasma proteome was profiled by quantitative mass spectrometry and

inflammation status of children was assessed. The second child follow-up was conducted

approximately a year later for the assessment of child cognitive function by psychological

tests. The study samples for each aim is displayed in the flow chart of study participants

(Figure 3.1). Statistical analyses were performed to estimate associations between

plasma proteins and inflammation and psychological test scores for aim 1 and 2. For aim

3, statistical analyses were performed to identify differentially abundant proteins and

differentially enriched gene sets by maternal micronutrient supplementation using the

Gene Ontology (GO) database.

3.1 Population, study design, and subjects

Nepal Nutrition Intervention Project, Sarlahi-3 (NNIPS-3)

NNIPS studies are series of cohorts to reduce maternal and child morbidity and

mortality in southern plains district of rural Sarlahi, Nepal, carried out by the Center for

Human Nutrition, Johns Hopkins Bloomberg School of Public Health. NNIPS-3 was a

community-based, cluster randomized controlled trial of antenatal micronutrient

69 supplementation which was aimed to improve birth weight and reduce infant mortality in rural South Asia region during 1999-2001(1). 30 village development communities with total population of 200,000 were selected in the district based on geographic location, and further divided into 426 sectors, smaller community cluster for the randomization units.

Women who had a low risk of becoming pregnant including women who were menopausal, sterilized, widowed, currently pregnant, and breastfeeding less than 9 months babies were excluded. The remaining women were visited every 5 weeks to ask their previous months menstruation. If their pregnancy was ascertained by urine test, they were enrolled into the randomized micronutrient study.

The four supplement arms were folic acid (FA), iron-folic acid (IFA), iron-folic acid- zinc (IFAZn), and multiple micronutrient (MM) which contains all three supplements with 11 other vitamins and minerals (Table 3.1. Micronutrient supplement regimen). All five groups of mothers received vitamin A which showed approximately 44 % reduction in pregnancy related maternal mortality in this population (2). In total, 4,926 pregnant women who provided informed consent were randomized by clusters to receive one of the five types of daily micronutrient supplement from the first trimester (mean gestational age was 10.2 weeks (4.1 SD)) through 12 weeks postpartum (1). Investigators, participants, field works were blinded to the allocation code throughout the study.

Table 3.1. Micronutrient supplement regimen

Group Micronutrient regimen Control (Vitamin A only) Vitamin A (1,000 μg RE) Folate (FA) Folic acid (400 μg) Folic acid-iron (60 mg of iron in the form of ferrous Folate+iron (IFA) fumarate) Folate+iron+zinc (IFAZn) Folic acid-iron-zinc (30 mg of zinc sulfate)

Multiple micronutrient supplement containing folic acid-iron-zinc plus vitamin D (10 μg), vitamin E (10 Multiple micronutrient mg), vitamin B1 (1.6 mg), vitamin B2 (1.8 mg), niacin (MM) (20 mg), vitamin B6 (2.2 mg), vitamin B12 (2.6 μg), vitamin C (100 mg), vitamin K (65 μg), copper (2.0 mg), and magnesium (100 mg) *Micronutrient supplements are all identically shaped, sized, and colored tablets. They were provided at levels that approximated a recommended dietary allowance.

Child follow-up study 1: Nutriproteomics study

From 4,926 pregnant women who participated in the NNIPS-3, 4,130 infants were live born. In 2006-2008, 3,524 children, 6-8 years of age whose mother participated in the NNIPS-3 across the five maternal supplementation groups were followed-up for a variety of nutritional health assessment (3-5). Of the children, 2,130 children were available for 4 aliquots of plasma specimens, birth weight measured within 72 hours after birth, and other epidemiological data. The children were stratified into the five original maternal supplement allocations and ordered by calendar date of blood draw in the field during the follow-up survey. Each group, 200 specimens were randomly sampled, yielding 1,000 child plasma specimens across the five supplement arms for nutritional assessment (6). Finally, 200 samples of each group were ordered by date of filed blood collection, and every other specimen were selected, resulting in 500 specimens for nutriproteomics study (Figure 3.1). This study is aimed to identify plasma biomarkers of a variety of micronutrients in order to develop a single valid platform to assess multiple micronutrient status at the population level (7).

Child follow-up study 2: Nepal Nutrition Intervention Project, Sarlahi, Nutrition and Cognition Project (NCOG)

In 2007-2009, a subset of children who were previously enrolled in the NNIPS-3 and

NNIPS-4 trial were followed up for cognitive function assessment (8). NNIPS-4 was a

71 cluster randomised, placebo-controlled trials evaluating the effects of preschool iron-folic acid (12.5 mg of iron and 50 μg of folic acid) with or without zinc supplementation (10 mg) on child survival (9, 10). The preschool supplementation trial was conducted from

2001-2005 among 1- to 35-month-old children. From the 5 different maternal micronutrient supplement arms of original NNIPS-3 study, children born to mothers from folic acid only group were not included in the NCOG study. Children aged at 7-9 years old participated in psychological tests of intelligence, executive function, and motor functions. Among these children, 252 children participated in both nutriproteomics study and NCOG study (Figure 3.1).

Ethical statement

The original NNIPS-3 trial was approved by the Nepal Health Research Coucil,

Kathmandu and the Institutional Review Board (IRB) at the Johns Hopkins Bloomberg

School of Public Health (JHSPH), Baltimore, MD (1). Children follow-up study was approved by IRBs at the JHSPH and Institute of Medicine, Tribhuvan University,

Kathmandu, Nepal (4). Ethical approval for the NCOG follow-up study was obtained from IRBs at the JHSPH and the Pennsylvania State University, and the Institute of

Medicine, Tribhuvan University (8).

Main health outcomes related to micronutrient supplementation in NNIPS-3 and followed-up studies

NNIPS-3 is the original study of several child follow-up studies which assessed infant morbidity, child survival, growth and nutritional health (4, 8, 11, 12). Before supplementation, maternal micronutrient status was measured. Based on defined cut-off values, baseline folate, vitamin B2, B6, and B12 deficiencies were 12, 33, 40, and 28%, respectively (13). Zinc and iron deficiencies were 61 % and 40 %, respectively, and 33 % of pregnant women were anemic (Hb <110 g/L) at the first trimester of gestation (13). In the 3rd trimester, supplements containing folic acid (FA) reduced folate deficiency by 75-

86 % (14). Iron-folic acid (IFA) reduced anemia by 54 % and iron deficient anemia by

72 % (15). IFAZn reduced subclinical infection, but failed to improve zinc status (14).

Multiple micronutrient (MM) reduced folate, vitamin B2, B6, and B12, and vitamin D deficiencies (14). For birth outcomes, IFA and MM supplements increased birth weight and reduced percentage of low birth weight babies by 16 and 14 %, respectively (1). In terms of infant mortality, FA and IFA supplement reduced mortality at 3 months by 64 % and 47 %, respectively, among preterm-born infants, but not term infants (16). Child mortality rate was also significantly low in IFA group compared to control group (hazard ratio= 0.69, 95 % CI 0.49-0.99) among children at about 7 years of age (17). FA supplement alone reduced microalbuminuria and metabolic syndrome and IFAZn improved linear growth and reduced peripheral adiposity at 6-8 years of age children (4,

5). General intelligence, executive functions, and motor skills at 7-9 years of age children were significantly improved in maternal IFA supplement group compared to control group (8).

3.2 Measurements

Plasma proteomics

Blood sampling

In 2006-2008, children were visited at home and early morning venous blood samples

(10 ml in the sodium heparin-containing tubes without preservatives or antioxidants) were collected by phlebotomist (4). Biospecimens were brought to a central laboratory and centrifuged. Blood plasma was equally aliquot into 4 tubes (at least 0.5ml of plasma per tube) and immediately frozen under liquid nitrogen. They were brought to the Johns

Hopkins University Center for Human Nutrition lab and stored at -80◦C for further molecular assessment. Masterpool which comprised of equal-sized aliquots of 1,000 plasma samples from Nutritional archive was prepared for reference standard for the planned iTRAQ runs.

High abundance protein depletion

Six high abundance proteins (albumin, transferrin, IgG, IgA, anti-trypsin, and haptoglobin) which constitute 85-90 % of total proteins were immune-depleted by

Human-6 Multiple Affinity Removal System (Agilent Technologies) LC column at the

Protein Depletion Laboratory to increase the possibility of identifying low abundance proteins (18). Unbound protein fractions were concentrated by TCA/acetone precipitation and collected to secure at least 100 μg for proteomics analysis. Each depleted sample was run on the SDS-PAGE gel for quality control. iTRAQ Tandem Mass Spectrometry

Proteomics analysis was performed at the Proteomics and Mass Spectrometry Core at the Johns Hopkins School of Medicine. Immno-depleted samples (100 µg of protein) were digested overnight with trypsin (Promega, sequencing grade). Peptide samples of 7 individuals and one masterpool sample were randomly labeled with 8-plex Isobaric Tag for Relative and Absolute Quantification (iTRAQ) reagents that contained different reporter ions which can be used as measures of peptide relative abundance in the original sample. The combined sample was fractionated into 24 fractions by strong cation exchange chromatography. iTRAQ-labeled peptides were loaded on to a reverse-phase nanobore column. Eluted peptides were sprayed into an LTQ Orbitrap Velos mass spectrometer (Thermo Scientific) and interfaced with a NanoAcquity ultra-HPLC

(Waters). Full MS scans and fragmented MS/MS scans were acquired and these spectra were searched against Refseq 40 protein database using MASCOT (Matrix Science v2.3) through Proteome Discoverer software (v1.3, Thermo Scientific). Peptides were identified with a confidence threshold of <5 % false discovery rate. A total of 72 iTRAQ experiments were performed for this study.

Estimation of relative abundance of proteins

Relative abundance of protein was estimated by the Biostatistics core at the Johns

Hopkins School of Public Health. Conventionally, relative protein abundance of biological samples from iTRAQ experiments is estimated using masterpool which is a reference sample standard to combine relative abundance of proteins from multiple iTRAQ experiments. Herbrich et al. demonstrated that more precise estimates can be obtained by using summary of biological samples without masterpool (19). Based on the finding, masterpool was not used to estimate relative abundance of proteins in this study.

The relative abundance of proteins in each channel of each experiment was estimated by calculating the median of all the median-polished log2 ion intensities across all spectra belonging to each protein. Corrections for differences in amounts of material loaded in the channels and sample processing were carried out by subtracting the channel median from the relative abundance estimate, normalizing all channels to have median zero. alpha-1-acid glycoprotein

Plasma concentration of alpha-1-acid glycoprotein (AGP) of children 6-8 years old age was assessed by the laboratory at the Center for Human Nutrition, Johns Hopkins

Bloomberg School of Public Health using a radial immunodiffusion assay (Kent

Laboratories; coefficient of variation = 10.0 %) (6).

Psychological Tests

In 2007-2009, six different psychological tests were administered to assess general intelligence, executive function, and motor functions of children (8). Among the tests, the

Stroop number test and Wechesler’s backward digit span were not used in this study because the results would be affected by child literacy and numeracy skills. Children were asked to visit a central site for testing and tests were administered by trained psychometrists. Their test administrations skills of the Universal Nonverbal Intelligence

Test (UNIT) and the Movement Assessment Battery for Children (MABC) were evaluated by Pennsylvania State University graduate students through randomly selected video records (8). Before tests were administered, psychometrists spent 30 minutes to build rapport with children and they were given a snack and a drink.

General intelligence - the Universal Non-verbal Intelligence Test (UNIT)

The UNIT is a comprehensive non-verbal measure of general intelligence and basic psychological processes for children (20). It is designed to provide an equitable cognitive assessment using hand and body gestures for children who would be unfairly assessed with a language-loaded ability test (20). The UNIT measures mainly two quotients: memory and reasoning (Table 3.2). It consists of 6 subtests - symbolic memory, cube design, spatial memory, analogic reasoning, object memory, and mazes. The analogic reasoning subtest was excluded because it was not culturally appropriate to Nepalese children (8). All 5 administered tests required minor motoric manipulation. Total tests took approximately 90 minutes. The original scoring is done by transforming raw scores to standard scores using manual or software. However, there were no good standard scores to compare with Nepalese children (8). Thus, exploratory factor analysis was conducted to evaluate the data structure and 2-factors model was chosen (8). Symbolic memory, object memory, cube design, and spatial memory comprised of factor 1, and maze subtest comprised of factor 2. Total scores of five subset tests were generated and raw score was converted to T-scores (mean 50 and standard deviation 10) based on child’s age (8).

Table 3.2. Subtest measures in the UNIT

Quotients Subtests Measure Memory Symbolic Memory Examinee recalls and recreates sequence of visually presented universal symbols (e.g. green boy, black women) Spatial Memory Examinee remembers and recreates the placement of black and/or green chips on 3*3 or 4*4 cell grid Object Memory Examinee is shown a visual array of common objects (e.g. shoe, telephone, etc) for five second, after which the examiee identifies the pictured objects from larger array of pictured objects Reasoning Cube Design The examinee completes three-dimensional block designs using between one and nine green and

white blocs Mazes The examinee completes mazes by tracing a path through each maze from the center starting point to an exit

Executive functions – computerized go/no-go test This test measures the capacities of inhibitory control. Computer graphic based stimulus was projected and children in a supine position viewed the stimulus. In go trials,

“go” stimulus was presented and the children were instructed to respond by promptly pushing a computer button (21). In no-go trials, “no-go” stimulus was presented and children were instructed not to respond. The task consisted of series of 210 “go” trials and 70 “no-go” trials (8). Score was expressed as the percentage of no-go stimuli on which the children’s response was correct. In this study only no-go test score was used because there was a ceiling effect in the go test results.

Motor functions-Movement Assessment Battery for Children (MABC) & Finger-tapping test MABC measures a spectrum of gross and fine motor functions of children aged 4-12 years. This consists of 8 items grouped in three sections: manual dexterity (following a line with pen, threading lace and turning pegs), ball skills (jumping and catching a ball, throwing bean bag into mat), and balance (walking on a line, hoping in squares, heal-to- toe walking). Raw performance score of each item was converted into a scaled score, provided by manual based on standardization sample, ranging from zero to fifteen or ten with the higher scores indicating a poorer performance. 0 score meant a complete success while highest score means a fail of the item test. Scaled score ranges on each of the subtests were manual dexterity, 0 to 15; ball skills, 0 to 10; and balance skills, 0 to 15.

The total scaled score had a range of 0 to 40. Total tests took approximately 20 minutes.

Finger-tapping test measures the index finger motor speed of each hand. Children were seated and finger tapper was placed on a table. Children were instructed to tap as fast as possible for 10 seconds, using the index finger of their preferred or dominant hand first. The same task was performed using the other hand. Score was expressed as the mean number of finger taps of both hands.

Other measurements

NNIPS-3

In 1998-2001, base line characteristics of pregnant women who were enrolled in the

NNIPS-3 study were collected (1). Maternal age, parity, weight, height, middle-upper arm circumference (MUAC), education, literacy and household socio-economic status

(SES) were collected. As most infants are born at home in this area, birth weight was measured by anthropometrist at home.

Nutriproteomics (children 6-8 years of age)

Child anthropometric measurements (height, weight, and MUAC) were measured by anthropometrists at home. Detailed of nutritional and health survey can be found in the previous publication (4, 5). Child education, food intake in the previous week information, past 7 days morbidity information (number of episode of high fever, watery stool, vomiting, productive coughing, rapid breathing, and ear discharge) were gathered by interviewing mothers. The Micronutrient Analysis Laboratory at the Center for

Human Nutrition carried out assessment of a variety of micronutrients including plasma ferritin (ng/ml) and transferrin receptor (ug/ml) from the nutritional bio-archive samples

(6). Plasma transferrin receptor (TfR) (Ramco Labs) was assessed using commercial

immunoassays, and plasma ferritin and thyroglobulin concentrations were measured using a benchtop clinical chemistry analyzer (Immulite 1000; Siemens Diagnostics) [26].

NCOG (children 7-9 years of age)

Demographic and environmental information of children were collected through two- home visited by interviewers (8). Child school enrollment history, past 7 day morbidity symptoms, and 7 day dietary intake information were collected. Household salt iodine use was examined with a semi-quantitative kit (MBI kits, Madras, India). Hemoglobin level was assessed using fingerstick with a B-Hemoglobin Analyzer (HemoCue, Lack forest, California). Household environmental influences was scored with the Middle

Childhood Home Observation for the Measurement of the Environment (HOME) inventory.

3.3 Data analyses

Data analyses were based on three primary aims: 1) to identify plasma proteins that co-vary with a biomarker of inflammation, 2) to identify plasma proteins that co-vary with psychological test scores, and 3) to identify differentially abundant plasma proteins or differentially enriched gene sets by maternal micronutrient supplementation. Data analyses and interpretation were based on following assumptions. Although original

NNIPS-3 study was a community based cluster-randomized trial, cluster is not considered as a source of variability assuming that correlation between protein profiles of children within the same cluster is ignorable. In addition, missing protein mechanism is assumed to be missing completely at random. Imputation of missing values was not carried out,

except for heatmaps for aim 2. All statistical analyses were carried out using the free software environment R version 3.1.0 (http://www.r-project.org/).

Exploratory data analysis

Proteomics data

Among proteins observed > 10% of total samples, the distributions of medians and standard deviations were examined to check proteins with extreme medians or dispersions. Batch effect was examined by performing hierarchical clustering analysis to check i) if protein abundance profiles of biological samples within a iTRAQ experiment are more likely to be similar than those of samples between iTRAQ experiments and ii) if protein abundance profiles of iTRAQ experiments run in a close time period are more likely to be similar than those of iTRAQ experiments run in a distant time period.

Channel effect was investigated by examining the differences in means of relative abundance of proteins between channels. Missing pattern was checked by comparing concentrations of C-reactive protein (CRP) measured by enzyme-linked immuno assay and missing in relative abundance of CRP measured by mass spectrometry. alpha-1-acid glycoprotein (AGP) and psychological test scores

Histograms of AGP and 4 different psychological test scores (T-score of UNIT, total scaled score of MABC, no-go percentage correct, and mean number of finger taps of both hands) were examined to check the ranges, extreme values, and distributions of the continuous outcomes. If a distribution of outcome was skewed, relevant transformation was employed to the outcome. Correlations among psychological indicators were examined to check if moderate correlations exist among indicators (22).

Epidemiological and other biochemical data

Previous studies highlighted that many factors including prenatal and postnatal micronutrient supplementation, household wealth through nutritional status, schooling, and home environment, iron status, and birth weight were associated with psychological test outcomes (8, 23-26). It was reasonable to hypothesize that plasma proteins might lie on the same causal pathways of some of the factors or the associations between proteins and outcomes might be confounded by shared factors. Thus, we considered a variety of child and household characteristic variables as covariates for aim 2. We derived log- transformed transferrin receptor to ferritin ratio to indicate iron status of children. The household wealth index was created based on the first principal component of the polychoric correlation of multiple indicators of household assets (materials of ground, floor, and roof of house, bicycle, radio, television, electricity, cattle, goat, and land).

Child characteristics (age, sex, prenatal and preschool childhood micronutrient supplementation, anthropometric measurements, food intake in the previous week, previous 7 days morbidity, birth weight, education, and plasma log transferrin receptor to ferritin ratio and thyroglobulin concentrations) and household characteristics (ethnicity, caste, household wealth index, and HOME inventory) were examined to check missing or unexpected values.

Statistical analysis

Aim 1

The objective of aim 1 is to identify plasma proteins that co-vary with AGP, a biomarker of subclinical inflammation. Because relative abundance of protein of samples

were compared only within an iTRAQ experiment (7 biological samples per experiment), and the mean absolute abundance of protein within a run is different by experiments, each sample was being normalized relative to a different quantity. To take into account differences in run-mean absolute abundance, linear mixed-effects (LME) models were used to analyze data from multiple iTRAQ experiments simultaneously. We fit univariate random intercept models and parameters were estimated via restricted maximum likelihood estimation (27).

Yij= (β0 + bj) + β1*Proteinij + ϵij

th Yij : plasma AGP concentration measured in an absolute scale of i biological plasma sample (i=1, …7) in jth iTRAQ experiment (j=1,…72) th th Proteinij : relative protein abundance of i biological plasma sample in j iTRAQ experiment. β0: the fixed effect for the intercept bj: random deviation from β0 in iTRAQ experiment j β1: the slope of protein-AGP association 2 2 bj~iid N(0, σb ), ϵij~iid N(0, σ ), cov(bj ,ϵij)=0

P-values were calculated by hypothesis testing of null association between plasma

AGP and protein. We controlled the family-wise error rate (FWER) using a Bonferroni correction, to only select proteins truly associated with AGP. The correlation between the

LME predictions (Best Linear Unbiased Predictors) and the outcomes was used to indicate AGP-protein correlation, and the square of these values were used for the proportion of variation in AGP explained by the LME fitted values (28).

Aim 2

The objective of aim 2 is to identify plasma proteins that co-vary with 4 different psychological test scores of children aged 7-9 years old. The same univariate random intercept model of aim 1 was applied to aim 2. Instead of adding known risk factors as

covariates to the model, we created covariates-adjusted psychological test scores using a residual method and used as outcome variables of univariate random intercept model.

In addition to univariate LME model, we constructed a multivariate model to identify an optimal subset of proteins that explain most variability in the psychological test scores.

Coefficient of determination (R2) was derived from the square of correlation coefficient between the LME predictions based on multiple proteins and the psychological test scores.

Yij= (β0 + bj) + β1*Proteinij + β2*Proteinij + …+ βp*Proteinij + ϵij

th Yij : covariates-adjusted psychological test score of i biological plasma sample (i=1, …7) in jth iTRAQ experiment (j=1,…72) β0: the fixed effect for the intercept bj: random deviation from β0 in iTRAQ experiment j β1~p: the slope of protein-psychological test score association 2 2 bj~iid N(0, σb ), ϵij~iid N(0, σ ), cov(bj ,ϵij)=0

Aim 3

The objective of aim 3 is to identify i) differentially abundant plasma proteins and ii) differentially enriched gene sets by maternal micronutrient supplementation, all compared to the control group.

Differentially abundant proteins

Although most proteins did not have random effects, LME models were employed to take into account random extreme values of protein abundance and to assess the maternal supplementation effects on relative protein abundance. We fit random intercept models with relative protein abundance as an outcome variable and dummy variables of maternal micronutrient supplementation as fixed effects and iTRAQ experiment as a random effect.

All models were adjusted for child age, sex, preschool micronutrient supplementation, and other potential confounders. P-values were derived from testing null hypothesis of the fixed effects and multiple comparison was corrected by controlling false discovery rate (29). Potential interaction between child sex and maternal micronutrient supplementation was examined by stratifying data by sex and same analysis was repeated for each sex.

Proteinij= (β0 + bj) + β1*FAij+ β2*IFAij + β3*IFAZnij + β4*MMij + …+ ϵij

th th Proteinij = relative protein abundance of i biological plasma sample (i=1,...,7) in j iTRAQ experiment (j=1,…,72) β0: the fixed effect for the intercept bj: random deviation from β0 in iTRAQ experiment j β1: mean difference in log2 relative protein abundance between maternal FA and the control group β2: mean difference in log2 relative protein abundance between maternal IFA and the control group β3: mean difference in log2 relative protein abundance between maternal IFAZn and the control group β4: mean difference in log2 relative protein abundance between maternal MM and the control group 2 2 bj~iid N(0, σb ), ϵij~iid N(0, σ ), cov(bj ,ϵij)=0, and … represents additional covariates

Differentially enriched gene sets

Gene set enrichment analysis (GSEA) is a knowledge-based computational analysis to determine whether priori defined gene sets are statistically significant (enriched) by phenotypes of interest. For the enriched gene sets, members of the sets are not randomly distributed throughout the list, but primarily found at the top or bottom. In this study, we tested whether gene sets are significantly differentially enriched by maternal micronutrient supplementation based on plasma protein abundance data. For the annotation database, the Molecular Signature Database (MSigDB, v4.0) was used.

Among its 5 major collections including BIOCARTA, KEGG, REACTOME,

TRANSFAC, and the Gene Ontology (GO), we used only the GO database which includes total 1,454 gene sets (825 for biological process, 233 for cellular component, and 396 for molecular function). GSEA was implemented in R environment using R-

GSEA script (GSEA-P-R.1.0) which is available in the GSEA website (30). GSEA was performed based on the methods described by Subramanian et al. (2005) (31, 32). Briefly, main work flow, main parameters, and results are described in Table 3.3 and Figure 3.2.

1) A whole list of t-statistics of proteins from the LME model were ranked.

2) Enrichment score (ES) was calculated by walking down the list, increasing a running- sum statistic when we encounter a member of the gene set and decreasing it when encounter a protein not in the gene set. The increment depends on the absolute magnitude of t-statistic of the protein, which indicates the effect size of maternal micronutrient supplementation. The enrichment score was the maximum deviation from zero.

3) Statistical significance of the ES was assessed by shuffling maternal supplementation groups within iTRAQ experiment to generate null distribution of ES, preserving the underlying correlation structure of protein abundance data. Then, p-value of the observed

ES was calculated using the empirically driven distribution.

4) Normalized enrichment score (NES) was calculated based on the observed ES divided by mean of ESs across all gene sets.

5) Multiple hypothesis testing was corrected controlling FDR. q-value was calculated as a ratio of two distributions: (1) the actual ES versus the ES for all gene sets against all permutations of the dataset data and (2) the actual ES versus the ES of all gene sets

against the actual dataset. Gene sets passing a FDR threshold of 5% were considered significantly enriched gene sets in this study.

Table 3.3. Key results of Gene set enrichment analysis

Enrichment score (ES) The degree to which given gene set is overrepresented at the top or bottom of the ranked list of proteins in the abundance data. Normalized enrichment The enrichment score for the gene set after it has been score (NES) normalized across analyzed gene sets Nominal p-value The statistical significance of the enrichment score. The nominal p-value is not adjusted for gene set size or multiple hypothesis testing, therefore, it is of limited use in comparing gene sets q-value The estimated probability that the normalized enrichment score represents a false positive finding. Leading-edge subset The subset of members that contribute most to the ES.

Figure 3.2. Example of enrichment plot

The top portion of enrichment plot showed running ES (green line) for the gene set.

The score at the furthest from 0 (horizontal line) is the ES for the gene set. Gene sets with peak at the beginning or at the end of the ranked list indicate positively or negatively enriched gene sets, respectively. In the middle portion of plot, bars (black vertical lines) represent gene set members. The number of bars indicates the size of given gene set. If given gene set is not enriched by maternal micronutrient supplementations, the bars would be randomly distributed across the ranked proteins. Leading-edge subsets are the set of members that appear in the ranked list prior to the peak score.

REFERENCES 1. Christian P, Khatry SK, Katz J, et al. Effects of alternative maternal micronutrient supplements on low birth weight in rural Nepal: double blind randomised community trial. Bmj 2003;326(7389):571. 2. West KP, Jr., Katz J, Khatry SK, et al. Double blind, cluster randomised trial of low dose supplementation with vitamin A or beta carotene on mortality related to pregnancy in Nepal. The NNIPS-2 Study Group. BMJ 1999;318(7183):570-5. 3. Stewart CP, Christian P, Schulze KJ, et al. Low maternal vitamin B-12 status is associated with offspring insulin resistance regardless of antenatal micronutrient supplementation in rural Nepal. The Journal of nutrition 2011;141(10):1912-7. 4. Stewart CP, Christian P, Schulze KJ, et al. Antenatal micronutrient supplementation reduces metabolic syndrome in 6- to 8-year-old children in rural Nepal. The Journal of nutrition 2009;139(8):1575-81. 5. Stewart CP, Christian P, LeClerq SC, et al. Antenatal supplementation with folic acid + iron + zinc improves linear growth and reduces peripheral adiposity in school-age children in rural Nepal. The American journal of clinical nutrition 2009;90(1):132-40. 6. Schulze KJ, Christian P, Wu LS, et al. Micronutrient deficiencies are common in 6- to 8-year-old children of rural Nepal, with prevalence estimates modestly affected by inflammation. The Journal of nutrition 2014;144(6):979-87. 7. Cole RN, Ruczinski I, Schulze K, et al. The plasma proteome identifies expected and novel proteins correlated with micronutrient status in undernourished Nepalese children. The Journal of nutrition 2013;143(10):1540-8. 8. Christian P, Murray-Kolb LE, Khatry SK, et al. Prenatal micronutrient supplementation and intellectual and motor function in early school-aged children in Nepal. Jama 2010;304(24):2716-23. 9. Tielsch JM, Khatry SK, Stoltzfus RJ, et al. Effect of daily zinc supplementation on child mortality in southern Nepal: a community-based, cluster randomised, placebo-controlled trial. Lancet 2007;370(9594):1230-9. 10. Tielsch JM, Khatry SK, Stoltzfus RJ, et al. Effect of routine prophylactic supplementation with iron and folic acid on preschool child mortality in southern Nepal: community-based, cluster-randomised, placebo-controlled trial. Lancet 2006;367(9505):144-52. 11. Christian P, Darmstadt GL, Wu L, et al. The effect of maternal micronutrient supplementation on early neonatal morbidity in rural Nepal: a randomised, controlled, community trial. Archives of disease in childhood 2008;93(8):660-4. 12. Murray-Kolb LE, Khatry SK, Katz J, et al. Preschool Micronutrient Supplementation Effects on Intellectual and Motor Function in School-aged Nepalese Children. Archives of pediatrics & adolescent medicine 2012;166(5):404-10. 13. Jiang T, Christian P, Khatry SK, et al. Micronutrient deficiencies in early pregnancy are common, concurrent, and vary by season among rural Nepali pregnant women. The Journal of nutrition 2005;135(5):1106-12. 14. Christian P, Jiang T, Khatry SK, et al. Antenatal supplementation with micronutrients and biochemical indicators of status and subclinical infection in rural Nepal. The American journal of clinical nutrition 2006;83(4):788-94.

15. Christian P, Shrestha J, LeClerq SC, et al. Supplementation with micronutrients in addition to iron and folic acid does not further improve the hematologic status of pregnant women in rural Nepal. The Journal of nutrition 2003;133(11):3492-8. 16. Christian P, West KP, Khatry SK, et al. Effects of maternal micronutrient supplementation on fetal loss and infant mortality: a cluster-randomized trial in Nepal. The American journal of clinical nutrition 2003;78(6):1194-202. 17. Christian P, Stewart CP, LeClerq SC, et al. Antenatal and postnatal iron supplementation and childhood mortality in rural Nepal: a prospective follow-up in a randomized, controlled community trial. American journal of epidemiology 2009;170(9):1127-36. 18. Scholl PF, Cole RN, Ruczinski I, et al. Maternal serum proteome changes between the first and third trimester of pregnancy in rural southern Nepal. Placenta 2012;33(5):424-32. 19. Herbrich SM, Cole RN, West KP, Jr., et al. Statistical inference from multiple iTRAQ experiments without using common reference standards. Journal of proteome research 2013;12(2):594-604. 20. Bracken BA, McCallum RS. Universal Nonverbal Intelligence Test. Itasca, IL: Riverside, 1998. 21. Konishi S, Nakajima K, Uchida I, et al. No-go dominant brain activity in human inferior prefrontal cortex revealed by functional magnetic resonance imaging. The European journal of neuroscience 1998;10(3):1209-13. 22. Deary IJ, Penke L, Johnson W. The neuroscience of human intelligence differences. Nature reviews Neuroscience 2010;11(3):201-11. 23. Christian P, Morgan ME, Murray-Kolb L, et al. Preschool iron-folic acid and zinc supplementation in children exposed to iron-folic acid in utero confers no added cognitive benefit in early school-age. The Journal of nutrition 2011;141(11):2042- 8. 24. Christian P, Murray-Kolb LE, Tielsch JM, et al. Associations between preterm birth, small-for-gestational age, and neonatal morbidity and cognitive function among school-age children in Nepal. BMC pediatrics 2014;14:58. 25. Patel SA, Murray-Kolb LE, LeClerq SC, et al. Household wealth and neurocognitive development disparities among school-aged children in Nepal. Paediatric and perinatal epidemiology 2013;27(6):575-86. 26. Murray-Kolb LE. Iron and brain functions. Current opinion in clinical nutrition and metabolic care 2013;16(6):703-7. 27. Harville DA. Maximum Likelihood Approaches to Variance Component Estimation and to Related Problems. Journal of the American Statistical Association 1977;72(358):320-38. 28. Robinson GK. That BLUP is a good thing: the estimation of random effects. Stat Sci 1991;6:15-32. 29. Storey JD. A direct approach to false discovery rates. . JRSS-B 2002;64:479-98. 30. Broad Institute. Gene Set Enrichment Analysis/MSigDB v 4.05. Internet: http://www.broadinstitute.org/gsea/downloads.jsp 2014. 31. Subramanian A, Tamayo P, Mootha VK, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles.

Proceedings of the National Academy of Sciences of the United States of America 2005;102(43):15545-50. 32. Mootha VK, Lindgren CM, Eriksson KF, et al. PGC-1alpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes. Nature genetics 2003;34(3):267-73.

4 CHAPTER 4: A PLASMA PROTEOME ASSOCIATED WITH INFLAMMATION IN SCHOOL-AGED CHILDREN IN RURAL NEPAL

ABSTRACT

Background: Inflammation is a condition stemming from complex host defense and

tissue repair mechanisms, often simply characterized by plasma levels of a single acute

reactant. Objective: We attempted to identify candidates biomarkers of inflammation

biomarkers within the plasma proteome. Design and methods: We applied quantitative

proteomics using isobaric mass tags (iTRAQ) tandem mass spectrometry to quantify

proteins in plasma of 500 Nepalese children 6-8 years of age. We evaluated those that

co-vary with inflammation, indexed by α-1-acid glycoprotein (AGP), a conventional

biomarker of chronic inflammation. Results: Among 982 proteins quantified in >10% of

samples, 99 were strongly associated with AGP at a family-wise error rate of 0.1%.

Magnitude and significance of association varied more among proteins positively (n=41)

than negatively associated (n=58) with AGP. The former included known positive acute

phase proteins including C-reactive protein, serum amyloid A, complement components,

protease inhibitors, transport proteins with anti-oxidative activity, and numerous

unexpected intracellular signaling molecules. Negatively associated proteins exhibited

distinct differences in abundance between secretory hepatic proteins involved in

transporting or binding lipids, micronutrients (vitamin A and calcium), growth factors

and sex hormones, and proteins of largely extra-hepatic origin involved in the formation

and metabolic regulation of extracellular matrix. Conclusions: Our findings have

revealed a vast plasma proteome within a free-living population of children that comprise

functional biomarkers of homeostatic and induced host defense, nutrient metabolism and

tissue repair, representing a set of plasma proteins that may be used to assess dynamics and extent of inflammation for future clinical and public health application.

INTRODUCTION

Inflammation is an evolutionarily conserved body response for protecting the host from potentially lethal stresses (1). It is often understood as a process to isolate and eliminate pathogens or other non-self or intolerant agents by immune cells (2) followed by processes that resolve inflammation, repair damaged tissues and restore homeostasis

(3). However, when the transition to tissue repair fail, inflammatory processes can persist and cause harm (4). It is well documented that chronic inflammation, often viewed as subclinical, contributes to chronic disease processes leading to obesity, diabetes, atherosclerosis, rheumatoid arthritis and cancer (4). In addition, inflammation in response to continuous exposure to infectious agents (5-7) and environmental toxins (8, 9) may contribute to childhood undernutrition and developmental deficits in impoverished areas of the world where poor sanitation and frequent infections are common.

An observable characteristic of inflammation is an increase or decrease in the release and concentration of numerous proteins in the bloodstream, a process regarded as an acute phase reaction, although it can occur in response to either acute or chronic inflammatory stress (1). At least 50 acute phase proteins (APP) are known to increase or decrease by at least 25%, a commonly used cut-off, from their baseline concentrations during inflammation (10, 11). This quantitative alteration is mainly regulated by inflammatory mediators that induce reactions within innate and adaptive immune, neuroendocrine, vascular endothelial, hematopoietic, metabolic, and other defense and repair systems (10). Due to these systemic effects, APPs are widely used as clinical diagnostic and prognostic indicators of disease processes (12). Also, APPs are often recommended for use in population studies as correction factors for assessing status with

respect to micronutrients, whose indicator concentrations may decrease (e.g., serum retinol for vitamin A) or rise (e.g, serum ferritin for iron) with inflammation (13, 14).

Although large in number, only one or a few APP are actually ever assessed in population studies to crudely represent phases of inflammation (15). This may, in part, be due to incomplete understanding of the vast number of molecular players and ability to measure and interpret their dynamics in response to infection or inflammation, limited resources to quantify multiple APPs, and a general lack of public health awareness of the dynamics of inflammation that exist in populations.

The unbiased approach of plasma proteomics offers a unique opportunity to discover, quantify and explore the utility of a wide array of proteins that respond to, initiate, maintain and resolve inflammation. In a given setting and time, all phases of response may be expected to be present in a population, reflecting potentially enormous utility of accessing and interpreting a larger inventory of biomarkers of inflammation with which to understand homeostatic mechanisms, subclinical disease processes, types of malnutrition and the range of inflammatory exposures in the environment.

In this study, we define inflammation on a continuum, as measured by the circulating concentration of α-1 acid glycoprotein (AGP), a conventional index of inflammation that reliably responds to systemic infection and numerous other pathological stresses and their resolution (16, 17). AGP or orosomucoid (ORM) is present in plasma as a mixture of

ORM1 and ORM2 each of which is encoded by two tandomly arranged genes, ORM1 and ORM2 (18). Because the concentration of AGP slowly rises and remains elevated during recovery or convalescence, AGP may be considered more sensitive in detecting chronic and subclinical inflammation in populations than C-reactive protein (CRP),

which tends to react, spike and resolve more quickly and thus be less often detected at a given time (14). We refer to the entire set of plasma proteins that definitively and quantitatively co-vary with inflammation (AGP), as a “population plasma inflammasome” that may include but extend beyond classical acute phase proteins, contain proteins of hepatic origin as well as those secreted or leaked from extra-hepatic tissues (19, 20), and include proteins involved in constitutive and induced homeostatic immune, coagulative and repair phases of inflammatory processes. We also distinguish this compound term from the “inflammasome”, used to describe an intracellular multiprotein complex that is responsible for activating the processing of pro- inflammatory cytokines (21). While limited animal experimentation has described an endotoxin-induced plasma inflammasome and compilations of inflammatory plasma proteins from databases for clinical use exist in the literature (22, 23), definition of a human population plasma inflammasome remains incomplete.

We carried out this quantitative proteomics study in a population cohort of Nepalese children 6 to 8 years of age born to mothers who, during gestation, participated in a randomized trial of antenatal micronutrient supplementation (24, 25). Twenty-five percent of the children exhibited an elevation in plasma AGP (>1g/L), but only 0.6% had an elevated C-reactive protein (CRP) (13), reflecting a population affected by chronic inflammation (13) but not acutely ill, providing an opportunity to explore markers of low- grade inflammation that may be useful for population assessment (26). In this study, we hypothesized the existence of a quantifiable population plasma inflammasome, defined as a suite of plasma proteins tracking AGP, an established, conventional biomarker of subclinical inflammation.

SUBJECTS AND METHODS

Field study and AGP analysis

In 1999-2001, a community-randomized, placebo-controlled field trial was carried out in the District of Sarlahi, Nepal to assess effects of 4 combinations of antenatal supplemental micronutrients on birth outcomes (25). The trial was registered with

ClinicalTrials.gov: NCT00115271. Of 4,130 infants live born, 3,524 were followed-up at

6-8 years of age for a nutritional, health and socio-demographic assessment by methods previously described (27, 28). During home visits, children of consenting parents were asked to fast overnight, and phlebotomists collected venous blood samples the next morning from 94% (n=3,305) of eligible children. Bio-specimens were brought to a field laboratory for plasma extraction. Plasma samples were stored in liquid nitrogen tanks and shipped to the Center for Human Nutrition, Johns Hopkins Bloomberg School of Public

Health in Baltimore, MD, USA. Among the plasma samples, 2,130 samples (64%) were selected based on having multiple plasma aliquots, complete epidemiologic data from both the original trial and follow-up assessment, and valid birth size measures (birth weight measured <72 hours after birth). Of the 2,130, 1,000 child plasma samples were randomly sampled across the five original maternal intervention groups (n=200 from each) and analyzed for multiple micronutrients and inflammation (AGP, CRP) status by conventional assays (13). Specifically, concentrations of plasma AGP were measured by a radial immunodiffusion assay (Kent Laboratories; CV = 10.0 %). From the 1,000 specimens, specimens were ordered by date of field blood collection within each original maternal supplement allocation stratum of 200 specimens, and following a chance start every other specimen was selected for inclusion into the proteomics archive, yielding a

total of 500 samples (n=100 per maternal group) (24). Socio-demographic, anthropometric and morbidity status, and dietary intakes among the 500 selected children were comparable to the 500 not in the proteomics study (13). The follow-up study protocol and biospecimen use were approved by the Institutional Review Boards at Johns

Hopkins School of Public Health, Baltimore, MD, in the United States and at the Institute of Medicine, Tribhuvan University, Kathmandu, in Nepal.

Proteomics and data analysis

Immuno-depletion and proteomics assays have been previously described (24). Briefly, a master plasma pool was prepared by combining plasma aliquots of 25 µL from each of the 1,000 samples comprising the micronutrient archive. The 500 specimens selected for proteomics analysis each comprised 40 µL of plasma. Individual specimens plus master pool aliquots were depleted of six high abundance proteins (albumin, transferrin, IgG,

IgA, anti-trypsin, and haptoglobin) using a Human-6 Multiple Affinity Removal System

LC column (Agilent Technologies). Proteomics analysis was performed at the Proteomics and Mass Spectrometry Core within the Johns Hopkins School of Medicine. Immno- depleted samples (100 µg of protein) were digested overnight with trypsin (Promega, sequencing grade). Peptide samples of 7 individuals and one masterpool sample were randomly labeled with 8-plex Isobaric Tag for Relative and Absolute Quantification

(iTRAQ) reagents that contained different reporter ions which can be used as measures of peptide relative abundance in the original sample. The combined sample was fractionated into 24 fractions by strong cation exchange chromatography. iTRAQ-labeled peptides were loaded on to a reverse-phase nanobore column. Eluted peptides were sprayed into

an LTQ Orbitrap Velos mass spectrometer (Thermo Scientific) and interfaced with a

NanoAcquity ultra-HPLC (Waters). Full MS scans and fragmented MS/MS scans were acquired and these spectra were searched against Refseq 40 protein database using

MASCOT (Matrix Science v2.3) through Proteome Discoverer software (v1.3, Thermo

Scientific). Peptides were identified with a confidence threshold of <5 % false discovery rate. A total of 72 iTRAQ experiments were performed for this study.

Details of relative abundance estimation have been described previously (29). Briefly, reporter ion intensities were log2 base-transformed and median normalized for each reporter ion intensity spectrum. The relative abundance of proteins in each channel of each experiment was estimated by calculating the median of all the median-polished log2 ion intensities across all spectra belonging to each protein. Corrections for differences in amounts of material loaded in the channels and sample processing were carried out by subtracting the channel median from the relative abundance estimate, normalizing all channels to have median zero. Plasma AGP concentration was log2 transformed due to its right-skewed distribution. Linear mixed-effects (LME) models were employed to assess the association between plasma AGP concentration and relative abundance of individual plasma proteins from multiple iTRAQ experiments. A univariate random intercept model was fit for each protein with AGP as a dependent variable, the protein as a fixed effect, and each iTRAQ experiment as a random effect. Model parameters in these mixed effects models were estimated via Restricted Maximum Likelihood (30). Estimates of absolute protein abundance were calculated as Best Linear Unbiased Predictors (31).

Observed sample sizes differed among proteins due to missing values from the mass- spectrometry (32). We report summary statistics for the association between plasma

100

protein abundance and AGP as percent change in AGP per 2-fold (100%) increase in protein relative abundance (derived from the slope of the LME model) and its statistical significance (p-value), and the correlation between estimated absolute protein abundance and plasma AGP concentration. We controlled the family-wise error rate (FWER) at the

0.1% level using a Bonferroni correction, to only select proteins truly associated with

AGP (P <1.02e-06). Since we expected that proteins associated with AGP might be co- regulated or co-vary in the same biological systems, we examined relationships between proteins by constructing a correlation matrix and performed principal component analysis

(PCA). Pairwise protein:protein correlation coefficients were calculated within each iTRAQ experiment and averaged coefficients across iTRAQ experiments to construct a correlation matrix and to perform PCA. Bi-plots were constructed to visualize the 1st, 2nd, and 3rd principal components of each protein from PCA (33).

Corresponding gene symbols of protein genInfo identifier (gi) numbers were derived from the Human Genome Organisation (HUGO) gene annotation and used in tables and figures, once linked to protein names in initial descriptive tables, to conserve space (34).

Resources of general description of proteins including cellular compartment, biological/molecular functions, and mRNA expression across tissues were extracted from the NCBI protein database, UniProt beta, Gene ontology Annotation (Uniprot-GOA) database, BioGPS, COMPARTMENTS, and in-depth review of literature (35-39).

All analyses were performed using the R Environment for Statistical Computing

(version 3.1.0; R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

Study participant characteristics

101

Demographic, nutritional, and health characteristics of study children (Table 4.1) were similar to children in the original, larger follow-up cohort (28). Children were undernourished, compared to the WHO reference population, reflected by prevalence rates of 40%, 15%, and 50% for stunting (height-for-age z-score < -2), thinness (BMI- for-age z-score <-2), and underweight (weight-for-age z-score <-2). More than half and approximately 30% of children consumed dairy food and dark green leafy vegetables equal to or greater than 3 times in the past week, respectively, but meat, fish, and eggs were less frequently consumed by children. Approximately 8% of children reported at least one episode of fever in the past week, but prevalence of other symptoms was low

(<5%). Fourteen percent of children reported any symptoms of fever, diarrhea, productive cough, or rapid breathing in the past week. Median (interquartile range) of plasma AGP concentration was 0.84 (0.70, 1.05) g/L, with 30% of children having an elevated AGP concentration (>1 g/L).

Plasma proteins and AGP A total of 3,933 proteins were identified and quantified among 72 iTRAQ experiments required to analyze the 500 child plasma samples, of which 982 proteins were quantified in >10% of all samples (n>50). Ninety-nine proteins (~10% of all proteins adequately quantified) significantly co-varied with plasma AGP passing a Bonferroni corrected significance level, of which 41 and 58 proteins were positively and negatively associated with AGP, respectively. Among proteins positively associated with AGP, TNFAIP3 interacting protein 1 (Gene symbol: TNIP1) (P=7.6x10-112) and orosomucoid 1 (ORM1)

(P=4.6x10-101) showed the strongest associations, followed by orosomucoid 2 (ORM2)

(P=3.1x10-54) (Table 4.2). Lumican (LUM) (P=9.1x10-27) and cartilage oligomeric matrix protein (COMP) (P=5.7x10-23) were most strongly associated among negative

102

correlates (Table 4.3). A volcano plot shows distinct patterns within the population plasma inflammasome (Figures. 4.1A-C). Specifically, the percent change in AGP and strength of significance of association varied more widely within the group of positively than negatively associated proteins. A 90% and 105% increase in AGP concentration was associated with a 100% (two-fold) increase in relative abundance of ORM1 and

ORM2, respectively (Figure. 4.1C). A comparable 73~105% increase in AGP was also associated with a 100% increase in complement components 2, 5, and 9 (C2/5/9) and complement factors F and I (CFB and CFI). On the other hand, a smaller 15~18% increase in AGP was associated with a 2-fold rise in other acute phase proteins such as

CRP, haptoglobin (HP), and serum amyloid A 1 and 2 (SAA1/2) (Figure. 4.1C). Overall, a narrower range in reduction in AGP, 20~40%, was associated with a 2-fold increase in the relative abundance of 46 out of the 58 negatively associated proteins (Figure. 4.1A).

A total of 206 plasma proteins passed a false discovery significance threshold of q<0.01

(~20% of all analyzed), representing a larger plasma proteome that appears to covary with plasma AGP.

Because we selected plasma proteins positively and negatively associated with AGP, we anticipated that proteins among each group would be correlated with each other. The correlation matrix, shown in Figure. 4.2A, shows that plasma proteins positively associated with AGP were more highly correlated with each other (i.e., more dark blue) than plasma proteins negatively associated with AGP (i.e., more light blue cells). There were 83 protein-pairs whose correlation coefficients (r) were greater than 0.6 within the positive population plasma inflammasome, while only 5 such pairs were observed within the negative plasma inflammasome. Within the former, high correlations were observed

103

among pairs involving ORM1-TNIP1-MAP3K14-COG3-ACTR5-NOM1 (all r >0.80),

LBP-ELL3 (r=0.92), and LRG1-EVI5 (r=0.86). Also, it appeared that there were two subgroups in the negatively associated proteins. This observation was confirmed by principal component analysis. As expected, the first principal component (PC1) divided the population plasma inflammasome into those positively (PC1<0, referred to as Group

1) and negatively associated with AGP (PC1>0). (Figure. 4.2B). The PC2 partitioned only negatively associated proteins into two groups (Group 2 vs. Group 3) and the PC3 separated two proteins (S100A8 and S100A9) from the rest of the proteins (Figure.

4.2C). To better understand the unexpected partitioning among plasma proteins negatively correlated with the AGP, we investigated cellular localization of the proteins.

Proteins in Group 2 are known to be mainly produced by the liver and secreted into the bloodstream (green) and proteins in Group 3 are largely produced by extra-hepatic tissues and localized in extracellular matrix regions (red and blue).

Localization and functions of plasma proteins associated AGP Proteins associated with AGP are summarized by their most described cellular localization and biological or molecular functions in Table 4.4. More than half of the proteins positively associated with AGP were primarily extracellular, largely secreted into circulation from the liver, and known to promote and regulate innate immune responses and scavenge oxidative stress. These included ORM1/2, CRP, SAA 1/2/4, bacterial lipopolysaccharide binding protein (LBP), LRG1 (a protein involved in granulocyte differentiation), components of the complement cascade, free hemoglobin scavengers, a copper-carrier, and several protease inhibitors. Other positively associated proteins are mainly localized in the membrane or intracellular space and involved in diverse functions including leukocyte recruitment and trafficking, cell signaling,

104

transcription, translation, DNA repair, protein methylation, modulation of cell cycle, cytokinesis and cytoskeleton, and endoplasmic reticulum-Golgi vesicle transport.

About half of the proteins negatively correlated with AGP (i.e., decline in relative abundance with inflammation) are also largely considered hepatic proteins released into the bloodstream. They are more involved in transport and metabolism of nutrients and small molecules (e.g., RBP4 and TTR for vitamin A; apolipoproteins A1/A2/H/M for lipid or cholesterol transport and metabolism; and AHSG for calcium and phosphate metabolism), sex hormone and growth factor binding (e.g., SHBG and IGFALS), or serine endopeptidase or proteinase inhibition in regulating blood coagulation and complement cascades, among other roles. Other negative correlates are physical constituents of the extracellular matrix (ECM), including collagens [e.g, collagen types

VI α1 and 3 (COL6A1, COL6A3)], glycoproteins, glucosaminoglycans and bone matrix proteins. Proteins known to facilitate interaction between cells and ECM are aminoproteinases/peptidases (e.g. PCOLCE and PEPD), protease inhibitors (e.g. TIMP2) and numerous cell-cell or cell-matrix adhesion molecules (e.g. CDH5 and ANTRX1).

DISCUSSION

We provide in this study evidence of a population plasma inflammasome, defined in relation to the continuous distribution of an established index biomarker of inflammation,

α1-acid glycoprotein. We ensured that a set of nearly 100 plasma proteins reflect true associations with AGP by using a stringent threshold to control FWER (0.1%). Study children in rural Nepal were undernourished, similar to many child populations in rural

Asia, but were active and not acutely ill, with only 14% reporting any symptom of illness in the previous week. As such, the set of proteins observed to covary with AGP can be

105

inferred to reflect interactions that occur within a homeostatic range of inflammation for the environmental conditions encountered in this typical South Asian rural setting.

Positive plasma inflammasome proteins tended to exhibit stronger associations with

AGP and greater variation in their degree of change per unit difference in AGP concentration than negatively associated proteins, possibly reflecting higher metabolic priority and functional specificity. In addition to well-established acute phase proteins, our quantitative proteomics approach identified numerous intracellular signaling, membrane-bound, and extracellular matrix molecules not widely recognized as acute phase reactants, appearing to reflect a vast systemic repertoire of proteins that response to inflammation.

Major acute phase proteins, complement components, protease inhibitors, and transport proteins with anti-oxidative activity positively covaried with AGP in plasma, in accordance to their expected roles in responding to stress (1). Many of the associated biomarkers are produced in the liver and secreted into the plasma (10). As AGP abundance in plasma is attributable to expression of highly homologous ORM1 (AGP1) and ORM2 (AGP2) genes (40), ORM1 and ORM2 were expected to be among the strongest correlates of AGP, measured by conventional radial immunodiffusion.

Confirmation of their strong association offers direct evidence of the validity of the mass spectrometric plasma proteomics approach to identify, quantify and correlate relative protein abundance. LRG1, HP, SERPINA3, CRP, SAA1, C9, and LBP were strongly and positively associated with AGP (all P <1.0x10-25), suggesting these proteins can be expected to be observed during inflammation (10). Substantial variation in the strength of association of proteins with AGP likely reflects wide differences in expression between

106

acute and chronic phase proteins (41-43). It is likely that complement components,

LRG1, and SERPINA3 co-reflect persistent low-level inflammation with AGP in this population that is frequently exposed to infectious agents such as parasitic and bacterial pathogens(44-46) and environmental toxins such as aflatoxin and arsenic (47, 48).

Positively associated proteins are known to largely involve pro-inflammatory regulation, immune activation (e.g., CRP, SAA, LBP, and complement components), control of proteolytic attack processes (SERPINA3, SERPING1, and SEPRIND1), and transport of pro-oxidative metabolites (CP and HP) (42). ORM1/2 and LRG1 are involved in immunomodulation and granulocyte differentiation, respectively, although their molecular roles have not been fully elucidated (49-51). Collectively, our results support an inference that hepatic-driven proteins that positively covary with inflammation are involved in host defense mechanisms.

Fourteen of 41 positive correlates of AGP are proteins involved in intracellular processes. In contrast to the high capture of known, circulating proteins, larger numbers of missing values in this group (Table 4.2) support the notion that these proteins may be low in abundance, and leaked or secreted from cells or tissues as part of normal cellular metabolism and tissue maintenance. The strongest (P= 7.6x10-112) and nearly 1:1 association of TNIP1 with AGP suggests TNIP1 may be co-regulated with AGP during inflammation, although no study to our knowledge has shown a direct metabolic linkage of the two. TNIP1 regulates inflammation by inhibiting cell signal transduction such as in the NF-kappa-B activation pathway (52). However, unknown extracellular functions of

TNIP1 leave its positive association with AGP in plasma unexplained. A 30~60% increase in AGP concentration was associated with a 100% increase in proteins of

107

intracellular origin, involved in signal transduction, and transcription, translation, and protein maturation and secretion. High correlations between these intracellular proteins and extracellular circulating proteins such as ORM1, LBP, and LRG1 suggest their abundance could reflect biosynthesis of acute phase or inflammatory mediators (53), or act themselves as acute phase proteins.

Among proteins negatively correlated with inflammation, about half are known to transport and regulate bioavailability of nutrients and hormones. Correlations among these proteins were negligible, suggesting independent regulation and metabolic pathways, while still being susceptible to hepatic-directed systematic reduction during inflammation (54). Some negative correlates are components of lipoprotein particles, involved in anti-inflammation (APOA1/2), anti-oxidative activities (PON1, PON3, and

PCYOX1), and reverse cholesterol transport (PLTP, CLU and LCAT) (all P <1.0x10-4).

Considering the increased abundance observed in serum amyloid As (SAAs) which are pro-inflammatory apolipoproteins, our results support considerable alteration in plasma lipoprotein composition during homeostatic inflammation (55). RBP4, TTR, AHSG, and

APOA1/2 are well-known negative acute phase proteins (10) that are consistent with known redistributions of vitamin A, calcium, phosphate and lipids during inflammation

(55-58).

In our analysis, a small decrement in AGP per two-fold increment in relative protein abundance identifies a protein that may be expected to decrease markedly during acute inflammation. The smallest decrease in AGP (20-24%) was observed with a two-fold rise in the proteins sex hormone binding globulin (SHBG), insulin-like growth factor

(IGF) acid labile subunit and IGF binding protein (BP) 3 (all P <1.0x10-5), consistent

108

with a reduction in insulin-like growth factor 1 and androgens expected during inflammation (10, 59). IGFALS and IGFBP3 form a ternary complex with most plasma

IGFs which play key roles in somatic growth and development (60, 61). SHBG binds and regulates functions of circulating androgens and estrogens (62). Population studies have revealed inverse associations between SHBG and inflammatory markers and adiposity-related early onset of puberty among girls (63, 64). Observed inverse associations between AGP and hepatic-derived proteins may reflect metabolic adaptation to altered endocrine signaling in response to inflammation.

Another group of proteins negatively associated with inflammation were components or regulators of the extracellular matrix (ECM). ECM comprises collagens, proteoglycans, elastins and glycoproteins and it is constantly remodeled through active interaction with bioactive molecules and adjacent cells (65). Our PCA results revealed that variance in abundance profiles for these proteins differed from those of hepatic origin, possibly due to differences in intravascular concentration gradients between classic plasma and extravascular proteins. Lumican (LUM), cartilage oligomeric matrix protein (COMP), tetranectin (CLEC3B), osteomodulin (OMD), and collagen α-1 and -3 type VI (COL6A1/3) are enriched in cartilage, bone matrix, skeletal muscle or adipose tissues (66-70). Beyond structural and functional components of ECM, many other proteins are involved in collagen or ECM metabolism, angiogenesis, and cell-cell and cell-matrix adhesion (71-78). These cell surface molecules or enzymatic proteins are involved in penetrating the vascular endothelial cells, regulation of pericellular proteolysis of ECM, and cell migration into inflamed tissues, critical to tissue repair and turnover (79). Our results suggest that proteins that maintain integrity of the ECM are

109

down-regulated, possibly reflecting metabolic rebalancing between host defense and healing mechanisms(80). These proteins are particularly important in chronic inflammatory conditions that are commonly accompanied by degradation of connective tissues (81-84).

Findings in this study corroborate fundamental characteristics of acute phase proteins observed in -omics studies in animals that have examined changes in gene or protein expression levels during inflammation. Yoo et al. showed in mice that 898 out of 8,551 protein-encoding genes (~7%) in the hepatic transcriptome were altered, equally up and down, by endotoxin-induced inflammation (85). We also observed that half the members of the population plasma inflammasome are predominantly hepatic in origin, approximately equally divided across positive and negative correlates of AGP and possibly reflecting a need to maintain protein equilibria in the vascular compartment (42).

Also in mice, Kelly-Spratt et al. reported that a third of ~500 plasma proteins increased or decreased by more than 1.25 fold in response to induced-inflammation (22). We observed that a large fraction (~20%) of the measurable plasma proteome covaries with inflammation, at a false discovery rate below 1%. The study also showed that induced- inflammation altered abundance of proteins involved in the complement and coagulation cascades, fibrinolysis, inhibition of protease activity, protein transport and reduced the abundance of proteins involved in ECM and collagen network remodeling (22), which were also observed in the present study.

Our study has strengths and limitations to consider. The diverse plasma proteins we observed to be associated with AGP in this ambient population offer an unbiased view of inflammation. With this large-scale and untargeted approach, we identified potential

110

networks of interaction and a large number of candidate biomarkers. Defining the plasma inflammasome with respect to AGP, an established index of chronic inflammation, enabled us to identify proteins that are likely relevant to the homeostatic response to inflammation. While our unit of estimated protein amount, i.e., relative abundance, restricts ability to directly translate these findings into immediate application, biomarkers of strongest association can be considered candidates for absolute quantification to expand the repertoire of applied biomarkers of inflammation. With a cross-sectional design, we could not infer metabolic proximity or causality of association. We depleted 6 highly abundant proteins, although they were not completely removed, and our current iTRAQ technology could not quantify low abundant cytokines and chemokines which mediate inflammation, revealing challenges of profiling a whole plasma inflammasome due to the dynamic range of protein abundance. However, it may be promising to investigate less abundant proteins and integrate protein profiles from different spectrums of abundance to build a more complete picture of a population plasma inflammasome. It is also possible that the plasma inflammasome could be refined using other sensitive markers of systematic inflammation markers.

CONCLUSIONS

This study provides evidence of strong association between an index biomarker of chronic inflammation and proteins of host defense, nutrient and hormonal metabolism and tissue remodeling. It is tempting to speculate that the low-grade inflammation seen in this study of young children could reflect early life processes, and thus risk, of adult chronic disease of rising prominence in undernourished and impoverished societies of

South Asia.

111

Table 4.1. Demographic, nutritional, and health characteristics of 6-8 year old children in rural Nepal (n = 500)1

Characteristics Values Age, years 7.5 (0.4) Girls, % 50.2 Ethnicity (Pahadi), % 31.8 2 Anthropometric measurements Height, cm 114.1 (5.8)

Weight, kg 18.2 (2.2) BMI, kg/m2 14.0 (1.0) Mid-upper arm circumference, cm 15.4 (1.1)

Height-for-age -1.77 (0.95) BMI-for-age -1.20 (0.89) Weight-for-age -1.99 (0.85) Stunted, %3 39.1 Thin, %3 16.4 3 Underweight, % 48.5 Dietary, ≥3 intake in the past week, %

Dairy 56.6 Meat 8.6 Fish 9.0

Eggs 2.2 Dark green leafy vegetables 32.4 Morbidity, symptoms reported in the past week, % Fever 8.2 Diarrhea 3.2 Productive cough 3.8 Rapid breathing 2.8 Any of above symptoms 14.0 Plasma Concentration of AGP 0.84 (0.70, AGP, g/L 1.05)4

5 AGP > 1.0 g/L , % 29.8 Abbreviations: AGP, α-1-acid glycoprotein; BMI, body mass index 1Values are means (SD) or percentages (95% CI). 2One outlier was excluded (n=499) 3Z-scores were calculated based on World Health Organization reference for 5-19 years. Underweight, weight-for-age Z-score< -2; stunted, height-for-age Z-score< -2; thin, BMI-for-age Z-score< -2 (86) 4Median (Interquartile range) 5Cut-off for inflammation (15)

112

Table 4.2. Plasma proteins positively associated with plasma α-1-acid glycoprotein (AGP) in 6-8 year old children in rural Nepal, ordered by P (FWER<0.01%)1

% change Gene in Protein name symbol n2 AGP3 P4 r5 Accession6 TNFAIP3 interacting protein 1 TNIP1 388 101.8 7.6E-112 0.816 116256481 Orosomucoid 1 ORM1 500 86.6 4.6E-101 0.770 167857790 Orosomucoid 2 ORM2 500 105.3 3.1E-54 0.674 4505529 Leucine-rich alpha-2- glycoprotein 1 LRG1 500 75.5 5.6E-48 0.654 16418467 Serpin peptidase inhibitor, clade A (alpha-1 antiproteinase, antitrypsin), member 3 SERPINA3 500 125.9 7.0E-41 0.627 50659080 ARP5 actin-related protein 5 homolog (yeast) ACTR5 255 60.4 9.6E-40 0.738 151301041 Component of oligomeric golgi complex 3 COG3 215 56.8 2.9E-35 0.730 13899251 Haptoglobin HP 354 16.8 2.7E-35 0.659 4826762 Mitogen-activated protein kinase kinase kinase 14 MAP3K14 307 55.6 3.2E-32 0.663 115298645 C-reactive protein, pentraxin- related CRP 438 15.6 3.2E-32 0.609 55770842 Serum amyloid A1 SAA1 493 17.6 3.1E-31 0.589 40316912 Complement component 9 C9 500 80.2 6.0E-30 0.581 4502511 Zinc finger, ZZ-type with EF- hand domain 1 ZZEF1 444 30.7 6.2E-30 0.601 73747881 Amyloid P component, serum APCS 500 57.2 3.0E-26 0.561 4502133 Lipopolysaccharide binding protein LBP 500 45.7 9.9E-26 0.559 31652249 Dynein, axonemal, assembly factor 1 DNAAF1 207 56.2 9.4E-22 0.694 157674358 NOP2/Sun domain family, member 6 NSUN6 249 53.4 6.4E-20 0.597 32698918 Ecotropic viral integration site 5 EVI5 271 49.5 1.6E-19 0.551 68299759 Complement factor I CFI 500 105.7 3.4E-19 0.521 119392081 Dynein, axonemal, heavy chain 11 DNAH11 298 37.1 8.4E-18 0.585 51479173 Complement factor B CFB 500 72.9 3.6E-17 0.507 67782358 Haptoglobin-related protein HPR 431 21.6 3.0E-16 0.509 45580723 Inter-alpha-trypsin inhibitor heavy chain 3 ITIH3 500 51.8 8.0E-15 0.489 133925809 MAP/microtubule affinity- regulating kinase 3 MARK3 231 50.3 1.1E-13 0.560 193083131

113

S100 calcium binding protein A9 S100A9 500 21.3 8.9E-13 0.477 4506773 Nucleolar protein with MIF4G domain 1 NOM1 119 44.8 1.1E-12 0.625 61097912 Elongation factor RNA polymerase II-like 3 ELL3 214 34.5 1.4E-12 0.537 13376768 S100 calcium binding protein A8 S100A8 500 20.0 2.2E-12 0.474 21614544 Asunder, spermatogenesis regulator ASUN 235 36.2 1.9E-11 0.524 155030185 Serum amyloid A2 SAA2 189 18.0 1.0E-10 0.546 188497671 Complement component 5 C5 500 73.8 8.7E-10 0.451 38016947 Complement component 2 C2 423 94.8 4.5E-09 0.445 14550407 Serum amyloid A4, constitutive SAA4 493 31.8 4.6E-09 0.442 10835095 Ceruloplasmin (ferroxidase) CP 500 63.8 1.3E-08 0.443 4557485 Integrin, alpha L (antigen CD11A (p180), lymphocyte function-associated antigen 1; alpha polypeptide) ITGAL 82 80.5 2.3E-08 0.645 167466217 Serpin peptidase inhibitor, clade G (C1 inhibitor), member 1 SERPING1 500 64.7 2.9E-08 0.439 73858570 Inter-alpha-trypsin inhibitor heavy chain family, member 4 ITIH4 500 76.9 5.6E-08 0.437 31542984 T-box 22 TBX22 103 32.5 2.0E-07 0.459 18375603 Serpin peptidase inhibitor, clade D (heparin cofactor), member 1 SERPIND1 500 40.9 3.6E-07 0.431 73858566 Complement component 4 binding protein, alpha C4BPA 500 24.4 4.6E-07 0.431 4502503 Mediator complex subunit 23 MED23 196 41.7 8.6E-07 0.483 28558969 Abbreviations: FWER, family-wise error rate 1Plasma proteins that achieved a Bonferroni corrected significance level (P <0.001/982=1.02e-06) 2The number of child plasma samples of each listed protein (50

114

Table 4.3. Plasma proteins negatively associated with plasma α-1-acid glycoprotein (AGP) in 6-8 year old children in rural Nepal, ordered by P (FWER<0.01%)1

% change Gene in Protein name symbol n2 AGP3 P4 r5 Accession6 Lumican LUM 500 -48.9 9.1E-27 -0.564 4505047 Cartilage oligomeric matrix protein COMP 500 -36.0 5.7E-23 -0.542 40217843 C-type lectin domain family 3, member B CLEC3B 500 -53.0 1.0E-20 -0.529 156627579 Transthyretin TTR 500 -48.8 4.7E-20 -0.524 4507725 Peptidoglycan recognition protein 2 PGLYRP2 500 -48.6 1.2E-18 -0.515 156616294 Apolipoprotein A-I APOA1 500 -43.2 1.9E-18 -0.516 4557321 Serpin peptidase inhibitor, clade A, member 4 SERPINA4 500 -46.5 4.9E-18 -0.512 21361302 Inter-alpha-trypsin inhibitor heavy chain 2 ITIH2 500 -53.4 1.1E-16 -0.505 70778918 Alpha-2-macroglobulin A2M 500 -58.5 3.1E-16 -0.500 66932947 Afamin AFM 500 -38.0 4.1E-16 -0.499 4501987 Attractin ATRN 500 -49.2 4.3E-16 -0.500 21450863 Fibroblast activation protein, alpha FAP 423 -31.7 1.5E-14 -0.514 16933540 Interferon-related developmental regulator 2 IFRD2 486 -29.9 2.2E-14 -0.493 197333755 Apolipoprotein A-II APOA2 500 -36.3 2.7E-14 -0.486 4502149 Collagen, type VI, alpha 3 COL6A3 472 -38.9 4.8E-14 -0.495 55743106 Alpha-2-HS-glycoprotein AHSG 500 -39.3 2.4E-13 -0.477 156523970 Serpin peptidase inhibitor, clade C (antithrombin), member 1 SERPINC1 500 -61.6 1.5E-12 -0.474 4502261 Kelch-like family member 34 KLHL34 403 -36.7 1.7E-12 -0.493 23397572 Peptidase inhibitor 16 PI16 500 -31.0 2.6E-12 -0.470 70780384 Thrombospondin 4 THBS4 451 -25.5 2.7E-12 -0.470 31543806 Gelsolin GSN 493 -36.9 3.0E-12 -0.481 4504165 Anthrax toxin receptor 2 ANTXR2 367 -33.4 3.3E-12 -0.526 50513243 Retinol binding protein 4, plasma RBP4 500 -31.6 3.7E-12 -0.469 55743122 CD109 molecule CD109 300 -34.4 3.9E-12 -0.558 227430301 TIMP metallopeptidase inhibitor 2 TIMP2 368 -33.5 4.7E-12 -0.491 4507511 SPARC-like 1 (hevin) SPARCL1 493 -39.5 4.0E-11 -0.464 190341024 Cadherin 5, type 2 (vascular endothelium) CDH5 500 -34.1 1.1E-10 -0.458 166362713

115

Olfactomedin 1 OLFM1 465 -36.8 1.5E-10 -0.466 17136143 Apolipoprotein M APOM 500 -33.8 1.9E-10 -0.457 22091452 Anthrax toxin receptor 1 ANTXR1 339 -28.1 2.4E-10 -0.489 14149904 Serpin peptidase inhibitor, clade A (alpha-1 antiproteinase, antitrypsin), member 5 SERPINA5 500 -31.9 5.2E-10 -0.454 194018472 Cadherin 13, H-cadherin (heart) CDH13 465 -34.4 7.6E-10 -0.465 4502719 Tenascin XB TNXB 500 -29.5 1.2E-09 -0.450 188528648 Dipeptidyl-peptidase 4 DPP4 416 -35.3 2.0E-09 -0.431 18765694 CD93 molecule CD93 416 -28.2 4.9E-09 -0.468 88758613 Peptidase D PEPD 466 -31.2 6.5E-09 -0.454 149589008 Microtubule-actin crosslinking factor 1 MACF1 430 -28.2 6.6E-09 -0.465 33188445 Procollagen C- endopeptidase enhancer PCOLCE 500 -35.3 6.7E-09 -0.443 157653329 Prenylcysteine oxidase 1 PCYOX1 493 -27.1 7.7E-09 -0.441 166795301 Quiescin Q6 sulfhydryl oxidase 1 QSOX1 500 -49.3 1.0E-08 -0.443 13325075 Apolipoprotein H (beta-2- glycoprotein I) APOH 500 -29.6 1.5E-08 -0.440 153266841 Integrin, alpha 2 (CD49B, alpha 2 subunit of VLA-2 receptor) ITGA2 187 -33.7 1.6E-08 -0.494 116295258 Inter-alpha-trypsin inhibitor heavy chain 1 ITIH1 500 -41.2 4.0E-08 -0.439 156119625 Alanyl (membrane) aminopeptidase ANPEP 500 -35.2 4.0E-08 -0.440 157266300 Biotinidase BTD 500 -35.5 4.2E-08 -0.438 4557373 Insulin-like growth factor binding protein, acid labile subunit IGFALS 500 -23.7 4.3E-08 -0.440 4826772 Glycosylphosphatidylinosito l specific phospholipase D1 GPLD1 500 -30.1 5.0E-08 -0.437 29171717 Heparan sulfate proteoglycan 2 HSPG2 493 -40.2 5.4E-08 -0.431 126012571 Lymphatic vessel endothelial hyaluronan receptor 1 LYVE1 479 -22.9 6.8E-08 -0.448 40549451 Butyrylcholinesterase BCHE 500 -31.4 8.9E-08 -0.437 4557351 Osteomodulin OMD 465 -24.9 9.0E-08 -0.466 4826876 Kallikrein B, plasma (Fletcher factor) 1 KLKB1 500 -36.6 9.4E-08 -0.436 78191798 Sex hormone-binding globulin SHBG 486 -19.6 2.3E-07 -0.429 7382460 Multimerin 2 MMRN2 444 -26.3 2.7E-07 -0.453 221316695 Mannan-binding lectin MASP1 486 -26.6 2.9E-07 -0.434 21264359

116

serine peptidase 1 Paraoxonase 1 PON1 500 -21.1 3.7E-07 -0.434 19923106 Neural cell adhesion molecule 1 NCAM1 453 -31.9 5.4E-07 -0.446 115529478 Collagen, type VI, alpha 1 COL6A1 472 -24.3 7.9E-07 -0.452 87196339 Abbreviations: FWER, family-wise error rate 1Plasma proteins that achieved a Bonferroni corrected significance level (P <0.001/982=1.02e-06) 2The number of child plasma samples of each listed protein (50

117

Figure 4.1A-C. Volcano plot of plasma proteins associated with plasma α-1-acid glycoprotein (AGP) in 6-8 year old children in rural Nepal.

Plot (A) and (C) are enlarged rectangles in plot (B). (A) Plasma proteins negatively associated with AGP, presented by gene symbol (n=58); (B) Plasma proteins positively and negatively associated with AGP were colored in blue and red, respectively (n=99); (C) Plasma proteins positively associated with AGP, presented by gene symbol (n=41). x- and y-axes are logarithmic.

118

Figure 4.2A-C. Correlation matrix and bi-plots from principal components (PC) analysis using plasma proteins associated with α-1-acid glycoprotein in 6-8 year old children in rural Nepal.

(A) Bottom-triangle is the correlation matrix of plasma proteins positively associated with AGP (Group1). Upper-triangle is the correlation matrix of plasma protein negatively associated with AGP (Group 2 & 3). (B and C) Bi-plot was constructed by the first three principle components. Color depicts representative tissue origins or subcellular

119

localization of proteins: black-intracellular space; green-hepatic origin and secreted into plasma; red-extracellular matrix; blue-extracellular matrix membrane binding). Proteins with PC1 less than 0 were assigned into group 1, proteins with PC1 and PC2 greater than 0 were assigned into group 2, and proteins with PC1 greater than 0 and PC2 less than 0 were assigned into group 3 (4 proteins were not included due to missings and lack of information about subcellular localization).

120

Table 4.4. Cellular localization and molecular/biological functions of plasma proteins associated with α-1-acid glycoprotein (AGP) in 6-8 year old children in rural Nepal1

Association Cellular localization Molecular/biological function Protein Immune system ORM1/2, APCS, LBP, LRG1 Extracellular Lipoproteins SAA1/2/4 (plasma) Complement system C2/5/9, CFB, CFI, C4BPA, CRP (n=23) Transport/scavenger protein HP/HPR, CP Serine proteases inhibitor SERPINA3, SERPING1, SERPIND1, ITIH3/4 Extracellular (Plasma membrane) Leukocyte-endothelial interaction ITGAL Positively (n=1) associated Extra- & intra- with AGP Leukocyte trafficking S100A8/9 cellular (n=2) (n=40) Regulation of cell signaling MAP3K14, TNIP1 Transcription and translation ELL3, NOM1, TBX22, MED23, NSUN6, regulation, DNA/RNA binding ACTR5, MARK3 Intracellular Cell cycle, Cell division, Mitosis EVI5, ASUN (n=14) Cytoskeleton DNAAF1, DNAH11 ER-Golgi vesicle-mediated transport COG3 Transport RBP4, TTR, AFM, AHSG Lipoproteins APOA1/2, APOH, APOM SERPINA5, SERPINC1, A2M, SERPINA4, Extracellular Serine proteinase inhibitor ITIH1/ ITIH2 Negatively (plasma) Serine type endopeptidase KLKB1, MASP1 associated (n=25) Other enzymes BCHE, PON1, BTD, PCYOX1, GPLD1 with AGP Inflammatory response ATRN (n=56) Growth factor/Hormone binding IGFALS, SHBG Peptidoglycan recognition PGLYRP2 Collagen COL6A1/3 Extracellular matrix non-collagenous glycoprotein COMP, THBS4, TNXB, SPARCL1, MMRN2 (n=11) Proteoglycan HSPG2, LUM

121

Bone matrix, mineralization CLEC3B, OMD Aminoprotease or peptidase PCOLCE, PEPD, DPP4, FAP, ANPEP CDH13, CDH5, NCAM1, OLFM1, CD93, Extracellular Cell-cell/cell-ECM adhesion ANTRX1, ANTRX2, QSOX1, ITGA2 (Plasma membrane) Cytoskeleton modulation GSN, MACF1 (n=20) Peptidase inhibitor TIMP2, PI16, CD109 Hyaluronan receptor LYVE1 1Proteins of unknown function were not presented (KLHL34, IFRD2, and ZZEF1)

122

REFERENCES

1. Kushner I. The phenomenon of the acute phase response. Annals of the New York Academy of Sciences 1982;389:39-48. 2. Ceciliani F, Giordano A, Spagnolo V. The systemic reaction during inflammation: the acute-phase proteins. Protein and peptide letters 2002;9(3):211-23. 3. Kumar V, Abbas AK, Aster JC. Robbins Basic Pathology. Edtion ed. Philadelphia, U.S.: ELSEVIER Saunders, 2012:29-73. 4. Nathan C, Ding A. Nonresolving inflammation. Cell 2010;140(6):871-82. 5. Investigators M-EN. The MAL-ED study: a multinational and multidisciplinary approach to understand the relationship between enteric pathogens, malnutrition, gut physiology, physical growth, cognitive development, and immune responses in infants and children up to 2 years of age in resource-poor environments. Clinical infectious diseases : an official publication of the Infectious Diseases Society of America 2014;59 Suppl 4:S193-206. 6. Prendergast AJ, Rukobo S, Chasekwa B, et al. Stunting is characterized by chronic inflammation in Zimbabwean infants. PloS one 2014;9(2):e86928. 7. Rodriguez L, Cervantes E, Ortiz R. Malnutrition and gastrointestinal and respiratory infections in children: a public health problem. International journal of environmental research and public health 2011;8(4):1174-205. 8. Ahmed S, Moore SE, Kippler M, et al. Arsenic exposure and cell-mediated immunity in pre-school children in rural bangladesh. Toxicological sciences : an official journal of the Society of Toxicology 2014;141(1):166-75. 9. Turner PC. The molecular epidemiology of chronic aflatoxin driven impaired child growth. Scientifica 2013;2013:152879. 10. Gabay C, Kushner I. Acute-phase proteins and other systemic responses to inflammation. The New England journal of medicine 1999;340(6):448-54. 11. AmiGO2 (v.2.1.4). acute-phase response (GO:0006953). 2015-01-29. Internet: http://amigo.geneontology.org/amigo/term/GO:0006953 (accessed Nov 2014. 12. Whicher T, Bienvenu J, Price CP. Molecular biology, measurement and clinical utility of the acute phase proteins. Pure and Applied Chemistry 1991;63(8):1111- 6. 13. Schulze KJ, Christian P, Wu LS, et al. Micronutrient deficiencies are common in 6- to 8-year-old children of rural Nepal, with prevalence estimates modestly affected by inflammation. The Journal of nutrition 2014;144(6):979-87. 14. Thurnham DI, McCabe GP. Influence of infection and inflammation on biomarkers of nutritional status with an emphasis on vitamin A and iron In: Organization WH, ed. Priorities in the Assessment of Vitamin A and Iron Status in Populations. Geneva, Switzerland: World Health Organization, 2012:63-80. 15. Thurnham DI, Mburu AS, Mwaniki DL, et al. Micronutrients in childhood and the influence of subclinical inflammation. The Proceedings of the Nutrition Society 2005;64(4):502-9. 16. Ceciliani F, Pocacqua V. The acute phase protein alpha1-acid glycoprotein: a model for altered glycosylation during diseases. Current protein & peptide science 2007;8(1):91-108.

123

17. Fournier T, Medjoubi NN, Porquet D. Alpha-1-acid glycoprotein. Biochimica et biophysica acta 2000;1482(1-2):157-71. 18. Yuasa I, Umetsu K, Suenaga K, et al. Orosomucoid (ORM) typing by isoelectric focusing: evidence of two structural loci ORM1 and ORM2. Human genetics 1986;74(2):160-1. 19. Anderson NL, Anderson NG. The human plasma proteome: history, character, and diagnostic prospects. Molecular & cellular proteomics : MCP 2002;1(11):845-67. 20. Kalmovarin N, Friedrichs WE, O'Brien HV, et al. Extrahepatic expression of plasma protein genes during inflammation. Inflammation 1991;15(5):369-79. 21. Petrilli V, Papin S, Tschopp J. The inflammasome. Current biology : CB 2005;15(15):R581. 22. Kelly-Spratt KS, Pitteri SJ, Gurley KE, et al. Plasma proteome profiles associated with inflammation, angiogenesis, and cancer. PloS one 2011;6(5):e19721. 23. Saha S, Harrison SH, Chen JY. Dissecting the human plasma proteome and inflammatory response biomarkers. Proteomics 2009;9(2):470-84. 24. Cole RN, Ruczinski I, Schulze K, et al. The plasma proteome identifies expected and novel proteins correlated with micronutrient status in undernourished Nepalese children. The Journal of nutrition 2013;143(10):1540-8. 25. Christian P, Khatry SK, Katz J, et al. Effects of alternative maternal micronutrient supplements on low birth weight in rural Nepal: double blind randomised community trial. Bmj 2003;326(7389):571. 26. Calder PC, Ahluwalia N, Albers R, et al. A consideration of biomarkers to be used for evaluation of inflammation in human nutritional studies. The British journal of nutrition 2013;109 Suppl 1:S1-34. 27. Stewart CP, Christian P, Schulze KJ, et al. Antenatal micronutrient supplementation reduces metabolic syndrome in 6- to 8-year-old children in rural Nepal. The Journal of nutrition 2009;139(8):1575-81. 28. Stewart CP, Christian P, LeClerq SC, et al. Antenatal supplementation with folic acid + iron + zinc improves linear growth and reduces peripheral adiposity in school-age children in rural Nepal. The American journal of clinical nutrition 2009;90(1):132-40. 29. Herbrich SM, Cole RN, West KP, Jr., et al. Statistical inference from multiple iTRAQ experiments without using common reference standards. Journal of proteome research 2013;12(2):594-604. 30. Harville DA. Maximum Likelihood Approaches to Variance Component Estimation and to Related Problems. Journal of the American Statistical Association 1977;72(358):320-38. 31. Robinson GK. That BLUP is a good thing: the estimation of random effects. Stat Sci 1991;6:15-32. 32. Luo R, Zhao H. Protein quantitation using iTRAQ: Review on the sources of variations and analysis of nonrandom missingness. Statistics and its interface 2012;5(1):99-107. 33. Gower JC, Lubbe SG, Roux NJL. Understanding Biplots: Wiley, 2011. 34. Gray KA, Yates B, Seal RL, et al. Genenames.org: the HGNC resources in 2015. Nucleic acids research 2015;43(Database issue):D1079-85.

124

35. Binder JX, Pletscher-Frankild S, Tsafou K, et al. COMPARTMENTS: unification and visualization of protein subcellular localization evidence. Database : the journal of biological databases and curation 2014;2014:bau012. 36. Geer LY, Marchler-Bauer A, Geer RC, et al. The NCBI BioSystems database. Nucleic acids research 2010;38(Database issue):D492-6. 37. Huntley RP, Sawford T, Mutowo-Meullenet P, et al. The GOA database: Gene Ontology annotation updates for 2015. Nucleic acids research 2015;43(Database issue):D1057-63. 38. UniProt C. Activities at the Universal Protein Resource (UniProt). Nucleic acids research 2014;42(Database issue):D191-8. 39. Wu C, Orozco C, Boyer J, et al. BioGPS: an extensible and customizable portal for querying and organizing gene annotation resources. Genome biology 2009;10(11):R130. 40. Nakamura H, Yuasa I, Umetsu K, et al. The rearrangement of the human alpha(1)-acid glycoprotein/orosomucoid gene: evidence for tandemly triplicated genes consisting of two AGP1 and one AGP2. Biochemical and biophysical research communications 2000;276(2):779-84. 41. Kushner I, Mackiewicz A. Acute phase proteins as disease markers. Disease markers 1987;5(1):1-11. 42. Mackiewicz A, Kushner I, Baumann H. Acute Phase Proteins: Molecular Biology, Biochemistry, and Clinical Applications. Florida: CRC Press, 1993. 43. Samols D, Agrawal A, Kushner I. Acute Phase Proteins. In: Oppenheim JJ, Feldman M, eds. Cytokine reference online. London, U.K.: Academic Press, 2002. 44. Navitsky RC, Dreyfuss ML, Shrestha J, et al. Ancylostoma duodenale is responsible for hookworm infections among pregnant women in the rural plains of Nepal. The Journal of parasitology 1998;84(3):647-51. 45. Chowdhury R, Huda MM, Kumar V, et al. The Indian and Nepalese programmes of indoor residual spraying for the elimination of visceral leishmaniasis: performance and effectiveness. Annals of tropical medicine and parasitology 2011;105(1):31-5. 46. Coles CL, Sherchand JB, Khatry SK, et al. Nasopharyngeal carriage of S. pneumoniae among young children in rural Nepal. Tropical medicine & international health : TM & IH 2009;14(9):1025-33. 47. Groopman JD, Egner PA, Schulze KJ, et al. Aflatoxin exposure during the first 1000 days of life in rural South Asia assessed by aflatoxin B(1)-lysine albumin biomarkers. Food and chemical toxicology : an international journal published for the British Industrial Biological Research Association 2014;74:184-9. 48. Thakur JK, Thakur RK, Ramanathan A, et al. Arsenic Contamination of Groundwater in Nepal—An Overview. Water 2010;3(1):1-20. 49. Hochepied T, Berger FG, Baumann H, et al. Alpha(1)-acid glycoprotein: an acute phase protein with inflammatory and immunomodulating properties. Cytokine & growth factor reviews 2003;14(1):25-34. 50. Wang X, Abraham S, McKenzie JA, et al. LRG1 promotes angiogenesis by modulating endothelial TGF-beta signalling. Nature 2013;499(7458):306-11.

125

51. O'Donnell LC, Druhan LJ, Avalos BR. Molecular characterization and expression analysis of leucine-rich alpha2-glycoprotein, a novel marker of granulocytic differentiation. Journal of leukocyte biology 2002;72(3):478-85. 52. Ramirez VP, Gurevich I, Aneskievich BJ. Emerging roles for TNIP1 in regulating post-receptor signaling. Cytokine & growth factor reviews 2012;23(3):109-18. 53. Fey GH, Gauldie J. The acute phase response of the liver in inflammation. Progress in liver diseases 1990;9:89-116. 54. Aldred AR, Schreiber G. The Negative Acute Phase Proteins. Edtion ed. In: Mackiewicz A, Kushner I, Baumann H, eds. Acute Phase Proteins Molecular Biology, Biochemistry, and Clinical Applications: Taylor & Francis, 1993:21-38. 55. Khovidhunkit W, Kim MS, Memon RA, et al. Effects of infection and inflammation on lipid and lipoprotein metabolism: mechanisms and consequences to the host. Journal of lipid research 2004;45(7):1169-96. 56. Moody BJ. Changes in the serum concentrations of thyroxine-binding prealbumin and retinol-binding protein following burn injury. Clinica chimica acta; international journal of clinical chemistry 1982;118(1):87-92. 57. Rosales FJ, Ross AC. Acute inflammation induces hyporetinemia and modifies the plasma and tissue response to vitamin A supplementation in marginally vitamin A-deficient rats. The Journal of nutrition 1998;128(6):960-6. 58. Jahnen-Dechent W, Heiss A, Schafer C, et al. Fetuin-A regulation of calcified matrix metabolism. Circulation research 2011;108(12):1494-509. 59. Tsilidis KK, Rohrmann S, McGlynn KA, et al. Association between endogenous sex steroid hormones and inflammatory biomarkers in US men. Andrology 2013;1(6):919-28. 60. Le Roith D, Bondy C, Yakar S, et al. The somatomedin hypothesis: 2001. Endocrine reviews 2001;22(1):53-74. 61. Baxter RC. Insulin-like growth factor binding proteins in the human circulation: a review. Hormone research 1994;42(4-5):140-4. 62. Anderson DC. Sex-hormone-binding globulin. Clinical endocrinology 1974;3(1):69-96. 63. Liao CH, Li HY, Yu HJ, et al. Low serum sex hormone-binding globulin: marker of inflammation? Clinica chimica acta; international journal of clinical chemistry 2012;413(7-8):803-7. 64. Pinkney J, Streeter A, Hosking J, et al. Adiposity, chronic inflammation, and the prepubertal decline of sex hormone binding globulin in children: evidence for associations with the timing of puberty (earlybird 58). The Journal of clinical endocrinology and metabolism 2014;99(9):3224-32. 65. Hynes RO. The extracellular matrix: not just pretty fibrils. Science 2009;326(5957):1216-9. 66. Byron A, Humphries JD, Humphries MJ. Defining the extracellular matrix using proteomics. International journal of experimental pathology 2013;94(2):75-92. 67. Wilson R. The extracellular matrix: an underexplored but important proteome. Expert review of proteomics 2010;7(6):803-6. 68. Wewer UM, Ibaraki K, Schjorring P, et al. A potential role for tetranectin in mineralization during osteogenesis. The Journal of cell biology 1994;127(6 Pt 1):1767-75.

126

69. Khan T, Muise ES, Iyengar P, et al. Metabolic dysregulation and adipose tissue fibrosis: role of collagen VI. Molecular and cellular biology 2009;29(6):1575-91. 70. Urciuolo A, Quarta M, Morbidoni V, et al. Collagen VI regulates satellite cell self-renewal and muscle regeneration. Nature communications 2013;4:1964. 71. Yucel G, Oro AE. Cell migration: GSK3beta steers the cytoskeleton's tip. Cell 2011;144(3):319-21. 72. O'Brien P, O'Connor BF. Seprase: an overview of an important matrix serine protease. Biochimica et biophysica acta 2008;1784(9):1130-45. 73. Man XY, Finnson KW, Baron M, et al. CD109, a TGF-beta co-receptor, attenuates extracellular matrix production in scleroderma skin fibroblasts. Arthritis research & therapy 2012;14(3):R144. 74. Kim SH, Turnbull J, Guimond S. Extracellular matrix and cell signalling: the dynamic cooperation of integrin, proteoglycan and growth factor receptor. The Journal of endocrinology 2011;209(2):139-51. 75. Hotchkiss KA, Basile CM, Spring SC, et al. TEM8 expression stimulates endothelial cell adhesion and migration by regulating cell-matrix interactions on collagen. Experimental cell research 2005;305(1):133-44. 76. Fukasawa K, Fujii H, Saitoh Y, et al. Aminopeptidase N (APN/CD13) is selectively expressed in vascular endothelial cells and plays multiple roles in angiogenesis. Cancer letters 2006;243(1):135-43. 77. Bell SE, Mavila A, Salazar R, et al. Differential gene expression during capillary morphogenesis in 3D collagen matrices: regulated expression of genes involved in basement membrane matrix assembly, cell cycle progression, cellular differentiation and G-protein signaling. Journal of cell science 2001;114(Pt 15):2755-73. 78. Arora PD, Glogauer M, Kapus A, et al. Gelsolin mediates collagen phagocytosis through a rac-dependent step. Molecular biology of the cell 2004;15(2):588-99. 79. Korpos E, Wu C, Sorokin L. Multiple roles of the extracellular matrix in inflammation. Current pharmaceutical design 2009;15(12):1349-57. 80. Schreiber G. Synthesis and secretion of APP London: Academic Press, 1983. 81. Redlich K, Smolen JS. Inflammatory bone loss: pathogenesis and therapeutic intervention. Nature reviews Drug discovery 2012;11(3):234-50. 82. Ademowo OS, Staunton L, FitzGerald O, et al. Biomarkers of inflammatory arthritis and proteomics. Edtion ed. In: Stanilova SA, ed. Genes and Autoimmunity - Intracellular Signaling and Microbiome Contribution: InTech, 2013:237-67. 83. Delano MJ, Moldawer LL. The origins of cachexia in acute and chronic inflammatory diseases. Nutrition in clinical practice : official publication of the American Society for Parenteral and Enteral Nutrition 2006;21(1):68-81. 84. Tseng S, Reddi AH, Di Cesare PE. Cartilage Oligomeric Matrix Protein (COMP): A Biomarker of Arthritis. Biomarker insights 2009;4:33-44. 85. Yoo JY, Desiderio S. Innate and acquired immunity intersect in a global view of the acute-phase response. Proceedings of the National Academy of Sciences of the United States of America 2003;100(3):1157-62. 86. WHO. Growth reference 5-19 years. Internet: www.who.int/growthref/en/.

127

5 CHAPTER 5: EXPLORING THE PLASMA PROTEOME ASSOCIATED WITH CHILD GENERAL INTELLIGENCE IN SCHOOL-AGED CHILDREN IN RURAL NEPAL

ABSTRACT

Background: The global burden of poor child development is high. Long-term

undernutrition, iron deficiency, infectious diseases, and psychological deficits stemming

from poverty have been identified as risk factors for undesirable developmental loss.

However, our understanding of underlying biological mechanisms remains unexplored.

Objective: This study aimed to identify plasma proteins that co-vary with intellectual

function of school-aged children. Design and methods: We applied quantitative

proteomics to profile plasma proteins of 249 children aged 6-8 years in rural Nepal.

Approximately one year later, we assessed general intelligence using the Universal

Nonverbal Intelligence Test (UNIT) in children at ages 7-9. Results: Among 751 proteins

quantified in >50 samples, 9 and 13 proteins were positively and negatively associated

with UNIT score, respectively, passing a false discovery rate threshold of 5.0%.

Positively associated proteins were largely binding proteins of insulin-like growth factors,

apolipoproteins, and transthyretin. Negatively associated proteins included complement

components, pyruvate kinase, and mostly proteins involved in subclinical inflammation.

After adjustment for known risk factors of cognitive abilities including child iron status,

long-term nutritional status, and household socio-economic status, only the associations

with proteins involved in chronic inflammation remained significant, explaining 5~9% of

variance in UNIT score. Conclusions: Using unbiased quantitative proteomics, this study

provides preliminary evidence of negative associations between child intellectual

128 function and plasma proteins involved in chronic low-degree inflammation. Further studies are needed to evaluate the roles and impact of chronic inflammation in child development in undernourished populations.

129

INTRODUCTION

Child development is shaped by dynamic interaction between individuals and their surrounding environment (1). More than 200 million children under five do not reach their developmental potential in less-developed countries (2). This conservative global estimate underscores not only the burden of deficits in child development, but also the vulnerability of children to the adverse environments with which they are confronted in daily life. Constant exposure to poverty, inadequate diets, infectious agents, stress, poor sanitation, and lack of cognitive stimulation and learning opportunities throughout early life can profoundly disturb fundamental biological processes of brain development (3).

This sub-optimal development in early life increases the risk of poor school performance, low income, high fertility, and poor child care in later life, consequently deepening global health inequalities (4).

Although the magnitude of the problem is indisputable and our knowledge about contributing factors is converging, our understanding of how those factors affect developmental processes has been hampered by inherent limitations in direct and indirect assessment of cognitive function. Because most psychological tests were designed for populations in developed countries, they have to be modified based on the cultural contexts of target settings, leaving the assessment of child development in low-resource settings limited (5). During the past decade, integrative sciences have begun to unlock the black box of neurological bases of high cognitive activities or pathologies associated with neurological disorders by applying neuro- or omics technologies (6-10). Thus, there is an increasing need to enhance our abilities to evaluate child development in low-resource

130 settings and to advance our understanding of the complexities of the biological processes underlining variation in child development using new technologies.

Although there is a huge amount of uncertainty in the interplay between peripheral proteins and CNS development, plasma proteome may provide a unique opportunity to explore biological processes or markers associated with child cognitive development.

Some proteins may directly act on the CNS by traversing the Blood-Brain-Barrier (BBB) which is a capillary endothelial cell layer that has a variety of protein receptors and selective permeability (11). Proteins involved in metabolism of neuropeptides or neurotrophins have been detected in plasma or serum (12-14). Plasma proteome may reflect crosstalk between other biological systems and the CNS that are co-regulated by small bioactive molecules such as neuroendocrine hormones (15-17). Clinical studies show that neurodevelopmental disorders are associated with dysregulation in the immune and endocrine systems and lipid metabolism, abnormal cellular structure, and disequilibrium in redox systems, all of which were reflected by serum peptides or proteins (10, 18-21).

We used previously collected plasma samples from school-aged children living in rural plain areas in southern Nepal and assessed intellectual, motor, and executive functions approximately a year later to identify plasma proteins that co-vary with general intellectual function and to examine the relationship between these proteins and contextual and biological factors that are known to influence cognitive development (22,

23). The results of this explorative study will provide useful preliminary information about biological pathways that are important for child cognitive development in healthy and relatively malnourished populations.

131

SUBJECTS AND METHODS

Study design and population

In 1999-2009, a research team at the Johns Hopkins Bloomberg School of Public

Health carried out community-based, cluster randomized controlled trials of antenatal and preschool micronutrient supplementation, as well as a series of follow-up studies, in the rural southern plains district of Sarlahi, Nepal. The maternal micronutrient intervention study was conducted from 1999 to 2001 and involved 4,926 pregnant women who received one of five types of micronutrient supplement daily from the first trimester through 3 months postpartum to reduce low birth weight and infant mortality (24). The child micronutrient supplementation trial was conducted from 2001 to 2005 and involved

3,675 children at 12-35 months of age who received one of four types of daily micronutrient supplement to evaluate the effects of iron-folic acid and zinc supplementation on child survival (25). In 2006-2008, we followed-up 3,524 children born to mothers who participated in the antenatal micronutrient supplementation trial for nutritional and health assessments and blood sample collection (23, 26). Children who had blood samples available for 4 plasma aliquots, completed epidemiologic data, and birth weight measured within 72 hours after birth were stratified by maternal supplementation groups; from these, 1,000 were randomly selected for nutritional assessment, and 500 of these were further randomly selected for proteomics analysis (27).

In 2007-2009, a subset of children from treatment groups in the prenatal and preschool micronutrient supplementation trials was selected for a follow-up neurocognitive assessment study (28). The study participants of the present study included children

132 whose plasma specimens were analyzed for proteomics and cognitive functions were assessed at 7-9 years of age (22, 27).

Preparation and analysis of plasma protein

During home visit of the first follow-up study in 2006-2008, phlebotomists collected early morning fasting venous blood (10 ml in the sodium heparin-containing tubes without preservatives or antioxidants) of children at 6-8 years of age. The bio-specimens were brought to a central laboratory and centrifuged for plasma extraction. Plasma samples were equally aliquoted into 4 tubes (at least 0.5ml of plasma per tube) and immediately frozen under liquid nitrogen. They were shipped to the Center for Human

Nutrition at the Johns Hopkins University, and stored at -80 ◦C for further assessment.

The procedures of proteomics have been described in detail in a previous publication (27).

Briefly, plasma samples of the 500 children and master pool sample which consisted of equal amount of plasma samples of 1,000 children were immune-depleted to remove 6 high abundant plasma proteins (albumin, immunoglobulin A, immunoglobulin G, transferrin, haptoglobin, and anti-trypsin) using a Human-6 Multiple Affinity Removal

200 System (MARS) LC column (Agilent Technologies, http://www.chem.agilent.com).

Immune-depleted samples were digested overnight by trypsin. 7 randomly selected samples and one masterpool sample were randomly labeled with 8 isobaric Tags for relative and absolute quantitation (iTRAQ) reagent (AB Sciex, http://www.absciex.com) according to manufacturer’s instructions. All samples were mixed together and fractionated into 24 fractions by strong cation exchange (SCX) chromatography. iTRAQ labeled peptides in each SCX fraction were loaded to reverse phase nanobore column and

133 eluted peptides were sprayed through a 10 μm emitter tip into an LTQ Orbitrap Velos mass spectrometer (Thermo Scientific, www.thermo.com/orbitrap) interfaced with a

NanoAcquity UPLC (Waters Corp). From each survey scan up to 10 peptide masses were individually isolated, fragmented and analyzed. Isotopically resolved masses in mass spectrometric (MS) and MS/MS spectra were extracted with and without deconvolution using Thermo Scientific Xtract software and searched against the RefSeq 40 protein database using Mascot (Matrix Science, www.matrixscience.com) through Proteome

Discoverer software (v1.3, Thermo Scientific). Peptides with high confidence (> 95%) from Mascot searches were filtered within the Proteome Discoverer with false Discovery

Rate (FDR) less than 5%.

Assessment of general intelligence

Details about the administration of the UNIT were previously reported (22). Briefly, children were asked to visit a central site for testing and tests were administered by trained psychological evaluators. The UNIT was designed to provide an equitable assessment using hand and body gestures for children who would be unfairly assessed with a language-loaded ability test (29). It measures mainly two quotients using six subtests: symbolic, spatial, and object memory, analogic reasoning, cube design, and mazes. The analogic reasoning subtest was removed because it was not culturally appropriate to Nepalese children. Total scores of the five subtests were generated and raw scores were converted to T-scores (mean 50 and standard deviation 10) based on child’s age.

Child nutritional and health assessments and household socio-economic interview

134

Other nutritional and health measurements of children and demographic and socio- economic status of households during home visits were described in details in elsewhere

(22, 23, 26). Briefly, specialized field assessment teams visited the children in their homes for household socioeconomic interviews and nutritional and health survey (26).

The heads of household were asked about children’s educational attainments, as well as household SES information (caste, religion, household asset ownership, parental education, and parental literacy). Trained anthropometrists measured children’s height, weight, and mid-upper arm circumference (MUAC) using standard instruments (26). In addition, children’s past 7-day morbidity symptoms and dietary intake information were also collected (22). Child iron status was assessed using the same plasma specimens collected for proteomics analysis (27). Plasma transferrin receptor (TfR) (Ramco Labs) was assessed using commercial immunoassays, ferritin and thyroglobulin concentrations were measured using a benchtop clinical chemistry analyzer (Immulite 1000; Siemens

Diagnostics), and alpha-1-acid glycoprotein (AGP) was measured using radial immunodiffusion assay (Kent Laboratories) (30). Weight and height data were transformed to z-scores based on the WHO Reference growth charts for children aged 5 to 19 years using AnthroPlus software (31).

Statistical analysis

Psychological test outcome and selection of potential confounders/mediators

The median (interquartile range) score of UNIT was 50.0 (43.0, 57.1), ranging

from 29.3 to 78.5. The distribution of UNIT was normally distributed without

noticeable outliers. Because child sex, age, and prenatal iron-folic acid supplementation

135

are known to be associated with intelligence score and childhood iron-folic acid

supplementation can be potentially associated with the outcome in this population (22,

32), we considered the child characteristics fixed covariates and adjusted for these

variables in all models. Because it is possible that observed associations between

plasma proteins and intelligence scores are derived by common factors or some plasma

proteins can potentially mediate the effect of factors on the intelligence outcome, we

considered a list of risk factors that are potentially associated with plasma proteins.

Selection of the covariates was based on known risk factors in the literature and the

results of preliminary data analysis (3, 22, 28, 33, 34). The following variables were

evaluated for associations with UNIT score as potential risk factors: child iron status

(log transformed transferrin receptor to ferritin ratio), iodine status (plasma

thyroglobulin concentration), long-term nutritional status reflected by height-for-age z-

score (HAZ), body mass index z-score, school attendance, food intake in the previous

week, morbidity history in the past week, birth weight, household ethnicity (Pahadi or

Madheshi), caste (Brahmin or Chhetri, Vaishya, and Shudra or non-hindu), HOME

inventory score, wealth index, and maternal literacy and education level (Appendix

5.1). We fit multivariate linear regression models with the fixed variables and selective

additional variables based on their independent associations with outcome (Appendix

5.1). The household wealth index was created based on the first principal component of

the polychoric correlation of multiple indicators of household assets (materials of

ground, floor, and roof of house, bicycle, radio, television, electricity, cattle, goat, and

land).

Linear mixed effects models

136

Estimation of relative abundances of protein measured by multiple iTRAQ experiments has been documented in detail elsewhere (35). Briefly, reporter ion intensities derived from Proteome Discoverer v1.3 were log2 base transformed and median normalized for each reporter ion intensity spectrum. The relative abundance of proteins in each channel of each experiment was estimated by calculating the median of all the median-polished log2 ion intensities across all spectra belonging to each protein. Corrections for differences in amounts of material loaded in the channels and sample processing were carried out by subtracting the channel median from the relative abundance estimate, normalizing all channels to have median zero.

We used linear mixed effects models (LME) to combine the proteomic data from different experiments and to estimate the association of protein relative abundances with outcome. To examine whether abundance of protein independently explains variance in

UNIT score after adjustment for other known covariates or risk factors, the UNIT test outcome was adjusted for covariates or risk factors by fitting a multivariate linear regression model and residuals were obtained. We fit a univariate random intercept model with the adjusted UNIT test score as a dependent variable, each protein as a fixed effect and iTRAQ experiment as a random effect. Parameters were estimated using restricted maximum likelihood estimation (36). Estimates of absolute protein abundance were calculated as Best Linear Unbiased Predictors (37). We transformed beta- coefficients (the fixed effect of the slope of the adjusted UNIT score and protein association) to be interpreted as estimated change in the adjusted UNIT score associated with 50% increase (1.5 fold-change) in relative abundance of protein. P-value was calculated by testing the hypothesis of null association between relative protein

137 abundance and the test outcome. Multiple hypothesis testing was corrected by controlling false discovery rate (FDR) (38). Proteins passing a FDR threshold of 5% were considered significantly associated with outcome. Coefficient of determination (R2) was estimated by calculating correlation between psychological test score and estimated absolute protein abundance. We performed sensitivity analyses by restricting iTRAQ experiments with samples at least 3 to check whether sample size per iTRAQ experiment affects estimation. We fit a multivariate model to estimate variability in the adjusted UNIT score explained by plasma proteins. Candidate proteins for the model were selected based on their marginal association with outcome. The model started with a protein which was the most strongly associated with outcome. Additional proteins were allowed to the model as covariates based on forward stepwise procedure with criterion that all included proteins in the model were significantly associated with outcomes with a significance level 0.05.

Correlation coefficients were calculated by averaging within-iTRAQ correlation coefficients between relative protein abundance across all iTRAQ experiments.

Hierarchical clustering was constructed using dissimilarity between proteins (1-Pearson correlation coefficient).

In addition, we examined changes in associations between proteins and outcome by adjusting for different risk factors to assess which plasma proteins associated with UNIT score are possibly associated with risk factors. To illustrate changes in the magnitude of associations, heatmaps were constructed with plasma proteins associated with the UNIT among the completed number of observation at least 70% of all samples. Color represents predicted values of relative abundance of protein from the LME models. Missing values of some proteins were imputed by multiple imputation (number of imputation=10) (39).

138

Imputed values were estimated by random sampling from 95% prediction interval of outcome-protein linear regression. LME models were fit with each imputed datasets and averaged predicted values of protein abundance were used for missing values.

General information of proteins was derived from National Center for Biotechnology

Information (NCBI) Protein Database and UniProt Knowledgebase (40, 41).

All analyses were carried out using open-source software implemented in the statistical environment R (42).

RESULTS

Among 500 children whose plasma specimens were analyzed by mass spectrometry,

252 children completed one of four cognitive tests. Characteristics of children who participated in the tests (n=252) and those of children who did not (n=248) were similar, except that children enrolled in the cognitive assessment were heavier (P=0.022) than children who did not (Table 5.1). Among children with cognitive test outcomes, the average (SD) ages at blood collection and at cognitive assessment were 7.5 (0.4) and 8.4

(0.7) years, respectively. Child sex was well balanced. About 30% of them were Pahadi

(ethnic group from hill region) and the rest of them were Madheshi (ethnic group from southern plains region). Mother was a primary care taker for almost every child (94.4%).

About 68% of children had been ever sent to school at the time when blood samples were collected. Children were relatively malnourished, characterized by 34.9% of stunting,

44.0% of underweight and 17.1% of low body mass index (BMI). The prevalence of lower respiratory tract infection and diarrhea episodes in the past week were relatively low (2.0 and 4.4%, respectively). About 27.5%, 81.7%, 48%, and 25% of children did not

139 consume milk, egg, meat/fish/chicken, and dark green leafy vegetable in the past week, respectively. The average (SD) age of mothers was 31.7 (5.3) years. 75% of them were illiterate and almost 80% of them had never attended to school. About 60% of households was Vaishya (middle class), and about 20% of households were Brahmin/Chhetri (high class) and Shudra/non-hindu (low class), each. More than half of households had no wall or wall made of thatch, grass, sticks, or branches.

One child with a missing value for the UNIT score and two children of a single observation within iTRAQ experiment were excluded, yielding 249 children included in analysis (Figure 5.1). The average (standard deviation, SD) number of plasma specimens in a single iTRAQ experiment was 3.5 (1.2), ranging from 2 to 6 (Appendix 5.2). Among

3,933 identified and quantified proteins in the plasma of 500, 751 plasma proteins were observed in greater than 50 children. Among the 751 proteins, 22 proteins (about 3% of total number of proteins analyzed) were associated with UNIT score, passing a FDR threshold of 5.0%. Among 9 proteins positively associated with UNIT score, the expected

UNIT score (95% confidence interval) increased 7.1 (4.3, 9.9) per 50% increase in relative abundance of insulin-like growth factor (IGF) binding protein, acid labile subunit

(IGFALS) (q=0.0003) (Table 5.2). It accounted for 24.8% of variance in UNIT score adjusted for child age, sex, and previous micronutrient supplementation. Following

IGFALS, 50% increases in relative abundance of transthyretin (TTR), apolipoproteins

(APOA1, A2, C3, C1, M, and D), and IGF-binding protein 3 (IGFBP3) were associated with 2.3~9.2 unit increases, explaining 18~21% of variance in UNIT score (all q<0.05).

Among 13 proteins negatively associated with UNIT score, orosomucoid 1 (ORM1) was most strongly associated with UNIT score. It decreased by 5.3 (3.1, 7.5) per 50%

140 increase in relative abundance of ORM1 (q=0.0007) (Table 5.3). It explained 23.5% of variance in the adjusted UNIT score. Components of complement system (C2, C5, C9, and complement factor I), plasma proteins involved in inflammatory response

(SERPINA3, LRG1, and LBP), intracellular protein (RCN1), protein timeless homolog

(TIMELESS), and pyruvate kinase isozymes M1/M2 (PKM) were negatively associated with UNIT score (all q<0.05). Hierarchical clustering analysis showed that proteins positively associated with the outcome were mainly divided into insulin-like growth factors or thyroid hormone binding proteins and apolipoproteins, while proteins negatively associated with UNIT score were clearly clustered among proteins participating in inflammation except PKM (Figure 5.2). Sensitivity analysis of restricting iTRAQ experiments with at least 3 observations showed similar estimates and slightly stronger associations of proteins with UNIT score. In a multivariate model, IGFALS,

ORM1, APOC1, and PKM jointly explained 37.8% of variance in the adjusted UNIT score (Table 5.4).

Among proteins associated UNIT score, we examined if the associations remain significant after additional adjustment for known risk factors including child iron status, long-term nutritional status (reflected by linear growth indicator), and household socio- economic status (SES) (Figure 5.3A-D. and Appendix 5.3A-B). Heatmaps showed that controlling for child iron status did not substantially change the associations between proteins and outcome (Figure 5.3A vs. B), but controlling for child HAZ moderately

(Figure 3A vs. C) and household SES considerably (Figure 5.3A vs. D) attenuated the observed associations and this was more prominent among proteins positively associated with outcome. We applied the same approach to the multivariate model with the 4

141 identified proteins in the previous multivariate model (Table 5.5). The associations remained significant after adjustment for child iron status, except APOC1 (Model 2). The associations with IGFALS and APOC1 became non-significant after controlling for child

HAZ and household SES, respectively, while the associations with ORM1 and PKM persisted significant after adjustment for iron status, linear growth, and household SES.

In a fully adjusted model with all risk factors, 7 proteins out of 751 proteins were significantly associated with UNIT score (Table 5.6). They are components of complement system (C9, CFI, and CFHR5), proteins involved in inflammation (ORM1,

LRG1, and SERPINA3), and an intracellular protein (RCN1). These proteins were all negatively associated with UNIT score by 4~8 unit decreases per 50% increase in relative abundance of proteins, explaining 5~9% of variance in the adjusted UNIT score for all risk factors, except RCN1.

To evaluate change in UNIT score associated with absolute change in abundance of protein involved in inflammation, we fit a model with orosomucoid [alpha-1-acid glycoprotein (AGP)], which was measured by a conventional assay in absolute scale.

UNIT score decreased by 1.7 (0.7, 2.8) units per increase in one standard deviation of plasma AGP concentration, adjusting for age, sex, previous micronutrient supplementation, ethnicity, school attendance, linear growth, and iron status (P=0.0015)

(Figure 5.4). The magnitude of this association was almost similar with the magnitude of association with one SD increase in HAZ 2.0 (0.9, 3.1).

Four proteins were associated with percent correct No-go score with q-value less than

0.05 (Appendix 5.5). Neural cell adhesion molecule 1 (NCAM) and membrane alanine

142 aminopeptidase (ANPEP) were positively associated with the outcome, while complement factor properdin (CFP) and Charcot-Leyden crystal protein (CLC) was negatively associated with the outcome (q<0.05). They accounted for about 20~26% of variance in the inhibitory control score. No protein was associated with scores of motor tests such as MABC and finger-tapping.

DISCUSSION

Using untargeted quantitative proteomics, this study found that plasma proteins involved in a variety of biological pathways measured in school-aged children covaried with child general intelligence test performance, assessed approximately one year later.

After controlling for known risk factors, plasma proteins involved in inflammatory response remained significantly associated with the outcome, explaining variance in the outcome in addition to the other risk factors. This result suggests that childhood inflammation may affect cognitive development through a distinct biological pathway in this population.

Major binding proteins to a somatic insulin-like growth factors were strongly associated with child intelligence score in the least adjusted model. In the circulation, IGFALS and

IGFBP3 form a 150-kD complex with 80~85% of circulating IGF1, increasing its half- life in the intravascular compartment (43, 44). There are a lack of studies examining associations between plasma IGF1 and IGFALS/BP3 and cognitive abilities. In a population-level study, IGF1 but not IGFBP3 concentration was associated with intelligence quotient test outcomes in school-aged children in the study of the Avon

Longitudinal Study of Parents and Children (45). The substantial attenuation in the observed associations after adjustment for HAZ and household SES suggests that plasma

143 abundance of these proteins might be associated with long-term nutritional status or contextual household SES conditions (46). It is possible that long-term nutritional status up-regulates growth hormone that stimulates hepatic production of IGFALS/BP3 as well as affects child cognitive development (47-49). There is accumulating evidence of the central role of the IGF1 in central nervous system (CNS) development (50-52). However, in the present study, it was not strongly associated (P=0.0509) with the outcome, although this interpretation requires extra caution because it was observed in a small number of samples (n=90). Because locally produced IGF1 has autocrine and paracrine actions within the CNS (53), more studies are required to clarify the roles of circulating

IGF1 and its binding proteins in the neurocognitive function in children.

Various types of plasma apolipoproteins were positively associated with intellectual functioning score. Apolipoproteins are structural components of lipoproteins such as chylomicron (APOA1/C1/C3), high-density lipoprotein (HDL) (APOA1/A2/M/D), and

(very) low-density lipoprotein (LDL) (APOC1/C3/M) (54). Based on the results of our independent analysis, these 6 apolipoproteins were all strongly positively associated with serum HDL cholesterol (all P<0.0001), but not LDL concentration in these children, suggesting that HDL cholesterol metabolism might be associated with the outcome. The positive associations were nullified by adjustment for household SES. This result is consistent with the study by Perry et. al, showing that a positive association between serum HDL cholesterol and cognitive achievement and academic performance became non-significant after adjusting for socio-economic/demographic and other risk factors in a nationally representative sample of school-aged children (55). Although cholesterol is vital in brain function and HDL cholesterol and APOA1 are known to exhibit anti-

144 inflammation properties in the CNS (56), there is lack of evidence that plasma cholesterol or peripheral apolipoproteins are associated with cognitive function at this age of children

(57, 58). The observed associations might be explained by household SES that may affect plasma apolipoprotein abundances in children through supply of different quantities and qualities of dietary lipids.

On the other hand, the positive association with TTR was attenuated but remained marginally significant (q=0.0613) after controlling risk factors. TTR transports thyroid hormone (T4) and retinol in plasma and cerebrospinal fluid (59, 60). Some studies suggest neuroprotective effect of TTR against neurodegenerative diseases (61, 62), but little is known about its role in neurodevelopment in children.

A group of plasma proteins involved in inflammation and the complement cascade negatively covaried with intellectual functioning score. Based on the results of our previous study [Chapter 4], ORM1, complement components (C9/2/5, and CFI), LRG1, and SERPINA3 showed similar magnitude of association with plasma AGP concentration, suggesting their constitutive abundance in circulation and roles in low-degree inflammation. Interestingly, C-reactive protein (CRP) and serum amyloid A1 (SAA1), which are known to immediately respond to inflammation, were not associated with

UNIT score (all q>0.3).

The heat maps and multivariate models showed that these proteins might be associated with intelligence score independently from contextual conditions. This result indicates that long-term low-level of inflammation might be a distinct pathway of developmental loss in children in this ambient conditions, although residual confounding cannot be ruled out. Similar results were observed in other studies. Jiang et al. reported that fever and

145 pro-inflammatory cytokines (IL-1beta and IL-6) as clinical and biochemical markers of inflammation in the first year of life were associated with decreased language and motor scores measured by a culturally adapted version of the Bayley Scales of Infant and

Toddler Development at 24 months in Bangladesh infants, independently from stunting, household SES, and maternal education (63). In addition, studies reported an inverse association between inflammation biomarkers and cognitive abilities in general adult populations, controlling for SES (64-66). There is a biologically plausible explanation of the negative effect of chronic inflammation on CNS development. The production of proteins participating in the inflammation is induced by cytokines and stress hormones

(67). These small molecules are known to cross or disrupt the BBB and affect the hippocampus which is responsible for memory consolidation and synaptic plasticity (68-

71). Thus, it is possible that proteins participating in peripheral chronic inflammation may indirectly reflect risk of sub-optimal cognitive development of children.

Pyruvate kinase isoenzyme (PKM or PKM2) was negatively associated with child intellectual score. In the fully adjusted model, it was marginally associated with the outcome (q=0.0857). It is a rate-limiting enzyme in glycolysis and regulated by thyroid hormones (72), and mediates cellular metabolic processes, generating ATP and pyruvate

(73, 74). It is not clear how a cytosolic glycolytic enzyme is negatively associated with child intellectual score. One possibility is that PKM in plasma may be associated with tissue damage due to oxidative stress or infection (75). Some infectious pathogens such as Neisseria. bacteria are known to directly interact with PKM to obtain pyruvate for their growth and survival (74, 76). Thus, we can hypothesize that a disturbance in

146 fundamental cellular energy metabolism or tissue damage by toxic agents may be negatively associated with child intellectual function.

In addition, we found that protein timeless homolog (TIMELESS) was negatively associated with child intelligence score. In the fully adjusted model, the association was attenuated, but remained marginally significant with outcome (q= 0.0676). TIMELESS is encoded from one of the CLOCK genes, whose expression oscillates rhythmically with time (77). It plays roles in the maintenance of genome stability throughout DNA replication, cell survival after DNA damage, and the regulation of the circadian clock (78,

79). In Drosophila, its heterodimer, PERIOD, was found to be essential for long-term memory formation (80). A considerable amount of missing values (only observed in n=62) of this protein suggest that it is probably leaked from a variety of cells or tissues.

No studies exist supporting direct or indirect biological links between this protein and child cognitive development.

This is the first study that explored the associations between plasma proteins and child cognitive functions applying an untargeted approach of proteomics in healthy, free-living children. Approximately one year time lag between blood sample collection and psychological test administration enabled us to rule out tentative associations raised from ongoing transient physiological conditions of children that can affect psychological test performance. Although this study included a small subset of children from a large-scale community-based cohort study, we believe children in this study are representative of

Terai regions in southern Nepal. Children were fairly homogeneous in terms of their ages and socio-cultural economic circumstances, minimizing unwanted bias. Also, rich epidemiologic and nutritional data allowed us to expand our analysis by including other

147 known risk factors of child development. We were able to measure comprehensive intellectual function of school-aged children using a valid and reliable test comprised of multiple subtests (81).

On the other hand, we could not detect and identify proteins that are known to directly function in neurological processes. Small number of samples affected precision of estimation of relative abundance of proteins and reduced power. In addition, observed associations in this prospective study design do not imply any causal relationship between proteins and outcomes. Also, substantial proportion of the observed associations could be due to residual confounding (46). Thus, further validation by longitudinal, sibling, or intervention studies should be followed to confirm the roles of the identified proteins in child development.

CONCLUSIONS

This exploratory study provides preliminary evidence that child intellectual function is negatively associated with low-degree chronic inflammatory status, reflected by plasma proteins involved in homeostatic control of inflammation. More research is requested to understand underlying factors and developmental consequences of prolonged low-grade inflammation in children living in areas where undernutrition and frequent infections are common. This will help us to identify more vulnerable children and to develop more specific and integrative interventions for preventing developmental loss in children.

148

Table 5.1. Demographic, anthropometric, health, and dietary characteristics of children with and without psychological assessments among school-aged children in rural Nepal1

Children Children with without psychological psychological P2 assessment assessment n (n=252) (n=248) Age, years (at blood specimen collection) 252 7.5 (0.4) 7.4 (0.5) 0.626 Age, years (at psychological test administered)3 252 8.4 (0.7) - - Male, n (%) 252 123 (48.8) 126 (50.4) 0.721 Ethnicity (Pahadi), n (%) 252 81 (32.7) 78 (31.5) 0.944 Primary caretaker mother3, n (%) 252 238 (94.4) - - Ever sent to school, n (%) 252 170 (67.5) 164 (66.1) 0.825 Nutritional status at baseline Weight, kg 252 18.6 (2.9) 18.0 (2.2) 0.022 Height, cm 252 114.5 (5.9) 113.7 (6.1) 0.116 BMI, kg/m2 252 14.1 (1.2) 13.9 (1.1) 0.123 Child MUAC, cm 252 15.5 (1.3) 15.4 (1.1) 0.167 WAZ 252 -1.88 (0.98) -2.06 (0.81) 0.023 HAZ 252 -1.69 (0.97) -1.83 (1.00) 0.117 BMIZ 252 -1.14 (0.88) -1.25 (0.93) 0.154 Stunting (HAZ<-2), n (%) 252 88 (34.9) 107 (43.1) 0.073 Underweight (WAZ<-2), n (%) 252 111 (44.0) 131 (52.8) 0.061 Low BMI (bmiz<-2), n (%) 252 43 (17.1) 39 (15.7) 0.777 Morbidity in the past 7 day, n (%) LRTI4 252 5 (2.0) 2 (0.8) 0.459 Diarrhea5 252 11 (4.4) 9 (3.6) 0.848 Diet in the past 7 days (any intake), n (%) Milk6 251 182 (72.5) 160 (64.5) 0.068

149

Egg 252 46 (18.3) 43 (17.3) 0.880 Any meat 252 131 (52.0) 132 (53.2) 0.851 Dark green leafy vegetables 252 189 (75.0) 187 (75.4) 0.999 Maternal characteristics Literacy, n (%) 252 63 (25.0) 48 (19.4) 0.158 No school education, n (%) 252 200 (79.4) 203 (81.9) 0.555 Household characteristics Caste, n (%) 252 0.079 Brahmin or Chhetri 44 44 (17.5) 26 (10.5) Vaiysha 157 157 (63.3) 169 (67.1) Shudra or non-hindu 51 51 (20.2) 53 (21.4) First floor material 252 0.499 No wall, thatch, grass, sticks, branches 130 130 (51.6) 135 (54.4) Katcha 88 88 (34.9) 85 (34.3) Wood planks 19 19 (7.7) 11 (4.4) Pakka 15 15 (6.0) 17 (6.9) Abbreviations: BMI, body mass index; MUAC, middle-upper arm circumference; WAZ, weight-for-age zscore; HAZ, height-for-age zscore; BMIZ, body mass index zscore; LRTI, Lower respiratory tract infection; HOME, Home Observation for the Measurement of the Environment 1Data are given as mean and standard deviation unless otherwise indicated. 2Using analysis of variance test for continuous variables and the chi-square test for categorical variables 3Measured when psychological tests were administered 4Lower respiratory tract infection is defined by productive cough or rapid breathing and fever 5Diarrhea is defined by watery stools or dysentery 6Data are missing for milk intake (n=1)

150

Figure 5.1. Flow diagram of study participant selection1

Live born infants n=4,130

Death, n=359 Lost to follow-up, n=247 Children followed-up at 6-8 yr of age n=3,524 Refusal blood-draw Absence during home visit n=219 Blood samples n=3,305 Not available for 4 plasma alliquots Not available for completed epidemiological data Birth weight measured ≥ 72h Eligible for n=1,175 nutritional bioarchive n=2,130 Random sampling stratified by maternal micronutrient supplementation Nutritional bioarchive n=1,000 Random sampling stratified by maternal micronutrient supplementation Proteomics bioarchive n=500 Children without cognitive test assessmentt n=248 Children with proteomics data and cognitive test outcomes n=252 No UNIT test score, n=1 Single plasma sample per iTRAQ experiment, n=2 Included for data analysis n=249

Abbreviations: UNIT, the Universal Nonverbal Intelligence Test, iTRAQ, isobaric tag for relative and absolute quantitation 1The original study was a community-clustered randomized controlled trial of antenatal micronutrient supplementation

151

Table 5.2. Plasma proteins positively associated with the Universal Non-verbal Intelligence Test (UNIT) score in school-aged children in rural Nepal (q<0.05), ordered by q

Gene Estimated No. Accession Protein n1 P 3 q4 R2,5 symbol change2 1 gi4826772 IGF-binding protein, acid labile subunit IGFALS 249 7.1 (4.3, 9.9) 6.0E-07 3E-04 24.8 2 gi4507725 Transthyretin TTR 249 9.2 (4.8, 13.7) 4.8E-05 0.004 20.8 3 gi4557321 apolipoprotein A-I APOA1 249 7.4 (3.7, 11.1) 8.6E-05 0.006 20.3 4 gi4502149 apolipoprotein A-II APOA2 249 6.6 (3.1, 10) 0.0002 0.01 19.6 5 gi62243068 IGF-binding protein 3 IGFBP3 249 4.4 (1.9, 6.9) 0.0006 0.022 18.6 6 gi4557323 apolipoprotein C-III APOC3 249 3.6 (1.5, 5.7) 0.0007 0.028 18.3 7 gi4502157 apolipoprotein C-I APOC1 249 2.3 (0.9, 3.7) 0.0014 0.043 17.9 8 gi22091452 apolipoprotein M APOM 249 5.8 (2.1, 9.5) 0.0020 0.048 17.6 9 gi4502163 apolipoprotein D APOD 249 5.2 (1.9, 8.6) 0.0020 0.048 17.6 Abbreviations: IGF, insulin-like growth factor 1Number of observation (maximum number=249) 2Expected difference in UNIT score per 50% increase in relative abundance of plasma protein, adjusted for child sex, age, and prenatal/childhood micronutrient supplementation 3P-value of hypothesis test of null association between protein and adjusted UNIT score 4Adjusted P-value correcting multiple hypothesis testing (false discovery rate) 5Percentage of variance in UNIT score (adjusted for child age, sex, and prenatal/childhood micronutrient supplementation) explained by protein

152

Table 5.3. Plasma proteins negatively associated with Universal Non-verbal Intelligence Test (UNIT) score in school-aged children in rural Nepal (q<0.05), ordered by q

Gene Estimated No. Accession Protein n1 P3 q4 R2,5 symbol change2 1 gi167857790 orosomucoid 1 ORM1 249 -5.3 (-7.5, -3.1) 2.5E-06 7.0E-04 23.5 2 gi4502511 complement component 9 C9 249 -7.6 (-11, -4.3) 8.8E-06 0.001 22.4 3 gi50659080 serpin peptidase inhibitor, member 3 SERPINA3 249 -9.8 (-14.2, -5.4) 1.1E-05 0.001 22.2 4 gi119392081 complement factor I CFI 249 -9.9 (-14.3, -5.4) 1.3E-05 0.001 22.0 5 gi14550407 complement component 2 C2 211 -15.3 (-22.7, -7.9) 5.1E-05 0.004 23.0 6 gi16418467 leucine-rich alpha-2-glycoprotein 1 LRG1 249 -5.0 (-7.6, -2.5) 0.0001 0.007 19.9 7 gi4506455 reticulocalbin 1 RCN1 59 -6.2 (-9.6, -2.9) 0.0003 0.012 43.2 8 gi38016947 complement component 5 C5 249 -8.4 (-13.2, -3.7) 0.0005 0.022 18.7 9 gi222136585 protein timeless homolog TIMELESS 62 -13.4 (-21.1, -5.8) 0.0006 0.022 31.6 10 gi31652249 lipopolysaccharide-binding protein LBP 249 -4.0 (-6.3, -1.7) 0.0007 0.024 18.5 11 gi33286418 pyruvate kinase isozymes M1/M2 PKM 239 -6.1 (-9.8, -2.3) 0.0015 0.043 20.7 leucine-rich repeat-containing 12 gi157674358 protein 50 DNAAF1 107 -4.4 (-7.2, -1.7) 0.0016 0.043 45.0 ecotropic viral integration site 5 13 gi68299759 protein EVI5 134 -4.5 (-7.2, -1.7) 0.0017 0.046 27.8 1Number of observation (maximum number=249) 2Expected difference in UNIT score per 50% increase in relative abundance of plasma protein, adjusted for child sex, age, and prenatal/childhood micronutrient supplementation 3P-value of hypothesis test of null association between protein and adjusted UNIT score 4Adjusted P-value correcting multiple hypothesis testing (false discovery rate) 5Percentage of variance in UNIT score (adjusted for child age, sex, and previous prenatal/childhood micronutrient supplementation) explained by protein

153

Figure 5.2. Hierarchical clustering of plasma proteins associated with Universal Non-verbal intelligence Test (UNIT) score in school-aged children in rural Nepal1,2

1Dissimilarity was calculated by 1-Pearson correlation coefficients 2Two proteins (TIMESS and RCN1) were excluded due to small number of observation

154

Table 5.4. Multiple plasma proteins that explain Universal Non-verbal intelligence Test (UNIT) score in school-aged children in rural Nepal (n=241)

Univariate model Multivariate model1 Gene Protein symbol Estimates2 P3 R2,4 Estimates2 P3 R2,5 IGF-binding protein, acid labile subunit IGFALS 7.2 (4.3, 10.0) 6.3E-07 27.2 4.9 (2.1, 7.7) 0.0006 orosomucoid 1 ORM1 -5.2 (-7.4, -3.0) 4.8E-06 25.4 -3.5 (-5.7, -1.3) 0.0020 37.8 apolipoprotein C-I APOC1 2.5 (1.0, 3.9) 7.7E-04 21.2 1.4 (0.0, 2.8) 0.0455 pyruvate kinase isozymes M1/M2 PKM -6.1 (-9.8, -2.3) 0.0015 20.7 -4.6 (-8.1, -1.1) 0.0092 1Candidate proteins (n=22) for the multivariate model were selected based on marginal association (false discovery rate threshold of 5.0%). The multivariate model was derived using forward stepwise procedure, including additional proteins that achieved significance level 0.05. 2Expected difference in UNIT score per 50% increase in protein abundance 3P-value of hypothesis test of null association between relative protein abundance and UNIT score 4Percentage of explained variability in UNIT score (adjusted for child age, sex, prenatal/childhood micronutrient supplementation) by protein 5Percentage of explained variability in UNIT score (adjusted for child age, sex, prenatal/childhood micronutrient supplementation) by multiple proteins

155

Each column and row represents child and protein, respectively. Children were ordered by adjusted UNIT score. Proteins (n=18) were selected based on marginal association (false discovery rate <0.05) and number of observation (n>70% of total sample size). Color represents best linear unbiased predictions of adjusted UNIT score from the proteomics data. A (Model 1): basic model (UNIT score was adjusted for child sex, age, and prenatal/postnatal micronutrient supplementation) B (Model 2): additional adjustment for child iron status (log transformed transferrin receptor to ferritin ratio) to model 1 C (Model 3): additional adjustment for child linear growth (height-for-age z-score) to model 1 D (Model 4): additional adjustment for household socio-economic status (wealth index, maternal education, ethnicity, and schooling) to model 1

157

Table 5.5. Association between plasma proteins and Universal Non-verbal intelligence Test (UNIT) score in the multivariate model (Model 1), adjusted for child iron status (Model 2), linear growth (Model 3), and household socio-economic status (SES) (Model 4) in school aged children in rural Nepal7

1 2 3 4 Gene Model1 Model 2 Model 3 Model 4 symbol Estimates5 P6 Estimates5 P6 Estimates5 P6 Estimates5 P6 IGFALS 4.9 (2, 7.7) 0.0008 4.8 (2.0, 7.5) 0.0008 2.4 (-0.4, 5.2) 0.0965 1.5 (-1.1, 4.1) 0.2638 ORM1 -3.5 (-5.8, -1.3) 0.0021 -4.0 (-6.2, -1.9) 0.0003 -3.4 (-5.6, -1.2) 0.0021 -3.4 (-5.4, -1.3) 0.0011 APOC1 1.4 (0.0, 2.8) 0.0478 1.2 (-0.2, 2.5) 0.0867 1.2 (-0.1, 2.6) 0.0726 0.6 (-0.6, 1.9) 0.3203 PKM -4.6 (-8.1, -1.1) 0.0094 -4.8 (-8.2, -1.4) 0.0057 -5.0 (-8.4, -1.6) 0.0042 -3.9 (-7.0, -0.7) 0.0168 1Model 1: basic model (UNIT score was adjusted for child age, sex and prenatal and child micronutrient supplementation) 2Model 2: additionally adjusted for iron status (log transformed transferrin receptor to ferritin ratio) to model1 3Model 3: additionally adjusted for child height-for-age z score to model 1 4Model 4: additionally adjusted for socio-economic status (wealth index, maternal education, ethnicity, and schooling) to model 1 5Expected difference in adjusted UNIT score per 50% increase in relative abundance of plasma protein (95% confidence interval) 6p-value of hypothesis test of null association between relative protein abundance and adjusted UNIT score 7Analysis was restricted to children with completed data for outcome and all covariates

158

Table 5.6. Plasma proteins associated with Universal Non-verbal Intelligence Test (UNIT) score, after adjustment for child iron status, linear growth, and household socio-economic status in school aged children in rural Nepal (q<0.05), ordered by q

Gene Estimated No. Accession Protein symbol n1 change2 P3 q4 R2,5 1 gi119392081 complement factor I CFI 247 -8.4 (-12.1, -4.7) 7.9E-06 0.0021 8.59 2 gi4502511 complement component 9 C9 247 -6.3 (-9, -3.5) 8.2E-06 0.0021 8.43 3 gi167857790 orosomucoid 1 ORM1 247 -4.0 (-5.9, -2.2) 1.3E-05 0.0023 7.89 4 gi50659080 serpin peptidase inhibitor, clade A, SERPINA3 247 -7.5 (-11.1, -3.8) 5.6E-05 0.0072 6.45 5 gi13540563 complement factor H-related 5 CFHR5 239 -4.6 (-7.0, -2.2) 0.0001 0.0140 9.19 6 gi16418467 leucine-rich alpha-2-glycoprotein 1 LRG1 247 -3.9 (-6.0, -1.8) 0.0003 0.0243 5.06 7 gi4506455 reticulocalbin 1 RCN1 59 -4.8 (-7.4, -2.2) 0.0003 0.0243 42.2 1Number of observation (maximum number=247). Data missing for plasma transferrin receptor to ferritin ratio (n=2) 2Expected difference in the fully adjusted UNIT score per 50% increase in the relative protein abundance (adjustment for child sex, prenatal/childhood iron-folic acid supplementation, wealth index, maternal education, ethnicity, and schooling, height-for-age z score, and log transformed transferrin receptor to ferritin ratio) 3P-values of hypothesis test of null association between protein abundance and adjusted UNIT score 4Adjusted P-value correcting multiple hypothesis testing (false discovery rate) 5Percentage of variance in the fully adjusted UNIT score explained by protein

159

Figure 5.4. Risk factors associated with Universal Non-verbal intelligence Test (UNIT) score in a multivariate model in school-aged children in rural Nepal (n=247)1,2

Abbreviation: AGP, alpha-1-acid glycoprotein, TfR:ferr index, log-transformed transferrin receptor to ferritin index (g/mg), HAZ, height-for-age z score; IFA, iron-folic acid supplementation. AGP* indicates standardized plasma AGP, measured using radial immunodiffusion assay. 1x-axis denotes expected difference in UNIT score of each variable adjusting for rest of other covariates in the model. 2Maternal education and household asset variables were dropped from the multivariate model (P>0.05)

160

APPENDIX

Appendix 5.1. Associations between the Universal Non-verbal Intelligence Test (UNIT) and potential risk factors of child development in school-aged children in rural Nepal

Multivariate9 Univariate (n=247) Variables n beta (95% CI) beta (95% CI) Age, year 249 1.1 (-0.9, 3.1) 1.8 (0.3, 3.4) Girl (ref: Boy) 249 -5.5 (-8.1, -3.0) -4.7 (-6.8, -2.6) Prenatal IFA supplementation (ref: placebo)1 249 2.2 (-0.5, 4.9) 1.7 (-0.5, 3.8) Child IFA supplementation (ref: placebo and zinc) 2 249 -2.3 (-4.9, 0.3) -1.3 (-3.4, 0.7) Ever sent to school 249 11.0 (8.5, 13.4) 6.6 (4.1, 9.0) HAZ 249 3.3 (2.1, 4.6) 2.1 (1.0, 3.2) BAZ 249 1.7 (0.2, 3.2) - Iron status3 247 -2.4 (-4.3, -0.5) -1.5 (-3.0, 0.0) Iodine status4 247 -1.2 (-2.9, 0.51) - WAZ at birth 249 2.4 (1.1, 3.7) - LRTI5 (7-9 yr) 247 4.9 (-3.7, 13.5) - Diarrhea6 (7-9 yr) 249 -1.5 (-10.9, 7.9) - Fruit intake (7-9 yr) 248 1.6 (-1.2, 4.4) - DGLV intake (7-9 yr) 249 2.5 (-0.5, 5.6) - Meat, chicken, fish intake (7-9 yr) 249 3.4 (0.8, 6.0) - Milk (7-9 yr) 249 1.0 (-1.7, 3.7) - Ethnicity, Pahadi (ref: Madheshi) 249 8.8 (6.2, 11.4) 3.3 (0.7, 6.0) Caste (ref: Non-Hindu) 249 - Vaisha 6.3 (3.2, 9.5) - Brahmin or Chhetri 12.2 (8.2, 16.2) - Wealth7 249 -2.2 (-2.9, -1.5) -0.7 (-1.4, 0.0) HOME inventory score8 248 0.5 (0.3, 0.7) -

161

Maternal education 249 1.3 (0.9, 1.7) 0.4 (0.0, 0.8) Maternal literacy 249 8.4 (5.5, 11.2) - Total explained variance in UNIT score, % - 43.4% Abbreviations: IFA, iron-folic acid supplementation; HAZ, height-for-age zscore; BAZ, body mass index-for-age zscore; WAZ, weight-for-age zscore; LRTI, lower respiratory track infection; DGLV, dark green leafy vegetable. 1Prenatal IFA indicates antenatal micronutrient supplements containing iron and folic acid 2Child IFA indicates childhood micronutrient supplements containing iron and folic acid 3Iron status was defined by log-transformed transferrin receptor (mg/L) to ferritin (ug/L) ratio. High value represents low iron status 4Iodine status was defined by log-transformed thyroglobulin (ug/L). 5Lower respiratory tract infection was defined by episode of productive cough or rapid breathing and fever in the past week 6Diarrhea was defined by episode of watery stools or dysentery in the past week 7High score represents low assets. 8Environmental influences on child outcomes were indexed with the Middle Childhood Home Observation for the Measurement of the Environment (HOME) Inventory 9Child age, sex, and prenatal/childhood micronutrient supplementation variables were fixed in the multivariate linear regression model. Other variables were selected based on prior knowledge or independent associations with UNIT score.

162

Appendix 5.2. Number of plasma samples per iTRAQ experiment

No. of plasma samples 2 3 4 5 6 7 per iTRAQ experiment1 = Total, 70 iTRAQ 16 18 20 13 3 0 No. of iTRAQ experiments experiments 1Multiplex-iTRAQ has 8 channels. One master pool and seven biological samples randomly selected from the proteomics bio-archive were loaded to each channel.

163

Appendix 5.3. Positive association between plasma proteins and Universal Non-verbal Intelligence Test (UNIT), after adjusted for child iron status, linear growth, and household socio-economic status in school aged children in rural Nepal

1 Gene UNIT:protein association symbol Model12 Model23 Model34 Model45 Model56 IGFALS 7.1 (4.3, 9.9) 7.1 (4.3, 10) 4.6 (1.8, 7.3) 3.3 (0.8, 5.8) 2.3 (-0.2, 4.7) TTR 9.2 (4.8, 13.7) 8.7 (4.2, 13.1) 7.7 (3.4, 11.9) 6.9 (3.1, 10.8) 6.0 (2.3, 9.7) APOA1 7.4 (3.7, 11.1) 6.8 (3.1, 10.5) 6.2 (2.7, 9.7) 3.6 (0.4, 6.8) 3.0 (-0.1, 6.1) APOA2 6.6 (3.1, 10) 6.4 (3, 9.8) 5.2 (1.9, 8.5) 3.5 (0.5, 6.5) 3.0 (0.1, 5.9) IGFBP3 4.4 (1.9, 6.9) 4.4 (1.9, 7) 2.7 (0.3, 5.1) 1.8 (-0.3, 4) 1.2 (-0.9, 3.3) APOC3 3.6 (1.5, 5.7) 3.1 (0.9, 5.2) 3.3 (1.3, 5.4) 2.0 (0.2, 3.9) 1.7 (-0.1, 3.5) APOC1 2.3 (0.9, 3.7) 2.2 (0.8, 3.6) 2.0 (0.6, 3.3) 1.2 (-0.1, 2.4) 1.0 (-0.1, 2.2) APOM 5.8 (2.1, 9.5) 5.6 (2, 9.3) 5.6 (2.2, 9.1) 2.5 (-0.7, 5.7) 2.7 (-0.4, 5.7) APOD 5.2 (1.9, 8.6) 4.3 (1, 7.6) 4.5 (1.4, 7.7) 2.5 (-0.3, 5.4) 1.6 (-1.1, 4.4) 1Expected difference in adjusted UNIT score per 50% increase in protein abundance 2Model 1: basic model (UNIT score adjusted for child age, sex and prenatal and child micronutrient supplementation) 3Model 2: additionally adjusted for child iron status (log transformed transferrin receptor to ferritin ratio to model 1 4Model 3: additionally adjusted for child height-for-age z score to model 1 5Model 4: additionally adjusted for household socio-economic status (wealth index, maternal education, ethnicity, and schooling) to model 1 6Model 5: full model (adjusted for all covariates listed above)

164

Appendix 5.4. Negative association between plasma proteins and Universal Non-verbal Intelligence Test (UNIT), after adjusted for child iron status, linear growth, and household socio-economic status in school aged children in rural Nepal

UNIT:protein association1 Gene symbol Model12 Model23 Model34 Model45 Model56 ORM1 -5.3 (-7.5, -3.1) -5.7 (-7.8, -3.5) -4.5 (-6.6, -2.5) -4.1 (-6.0, -2.1) -4 (-5.9, -2.2) C9 -7.6 (-11, -4.3) -8.1 (-11.4, -4.8) -7.3 (-10.5, -4.1) -5.9 (-8.8, -3.0) -6.3 (-9.0, -3.5) SERPINA3 -9.8 (-14.2, -5.4) -10.6 (-14.9, -6.4) -8.7 (-12.8, -4.5) -7.4 (-11.2, -3.6) -7.5 (-11.1, -3.8) CFI -9.9 (-14.3, -5.4) -10.9 (-15.3, -6.4) -10.0 (-14.1, -5.8) -7.1 (-10.9, -3.2) -8.4 (-12.1, -4.7) C2 -15.3 (-22.7, -7.9) -15.3 (-22.6, -8.1) -13.9 (-20.8, -7.0) -10.5 (-16.9, -4.1) -10.1 (-16.1, -4.0) LRG1 -5 (-7.6, -2.5) -5.4 (-7.9, -2.9) -4.5 (-6.9, -2.0) -3.9 (-6.1, -1.7) -3.9 (-6.0, -1.8) RCN1 -6.2 (-9.6, -2.9) -5.9 (-9.1, -2.7) -6.2 (-9.1, -3.2) -4.9 (-7.8, -2.0) -4.8 (-7.4, -2.2) C5 -8.4 (-13.2, -3.7) -9.5 (-14.2, -4.8) -7.2 (-11.8, -2.7) -5.5 (-9.6, -1.4) -6 (-9.9, -2.0) TIMELESS -13.4 (-21.1, -5.8) -12.5 (-20.1, -4.9) -12.5 (-19.6, -5.4) -9.9 (-16.1, -3.7) -9.2 (-15.0, -3.3) LBP -4.0 (-6.3, -1.7) -4.2 (-6.5, -1.9) -3.5 (-5.6, -1.3) -3.3 (-5.3, -1.3) -3.3 (-5.2, -1.4) PKM -6.1 (-9.8, -2.3) -6.1 (-9.8, -2.4) -5.9 (-9.4, -2.3) -4.4 (-7.6, -1.1) -4.5 (-7.6, -1.4) DNAAF1 -4.4 (-7.2, -1.7) -4.8 (-7.5, -2.0) -3.5 (-6.2, -0.8) -2 (-4.6, 0.7) -1.9 (-4.5, 0.6) EVI5 -4.5 (-7.2, -1.7) -4.4 (-7.2, -1.7) -3.9 (-6.6, -1.2) -2.7 (-5.1, -0.3) -2.7 (-5.0, -0.3) 1Expected difference in adjusted UNIT score per 50% increase in protein abundance 2Model 1: basic model (UNIT score adjusted for child age, sex and prenatal and child micronutrient supplementation) 3Model 2: additionally adjusted for child iron status (log transformed transferrin receptor to ferritin ratio) to model 1 4Model 3: additionally adjusted for child height-for-age z score to model 1 5Model 4: additionally adjusted for household socio-economic status (wealth index, maternal education, ethnicity, and schooling) to model 1 6Model 5: full model (adjusted for all covariates listed above)

165

Appendix 5.5. Plasma proteins associated with percent correct no-go in school aged children in rural Nepal (q<0.05), ordered by q

Gene No. Accession Protein symbol n1 Estimated change2 P3 q4 R2,5 1 gi115529478 neural cell adhesion molecule 1 NCAM1 219 17.1 (9.2, 25.0) 2.4E-05 0.0126 24.0 2 gi157266300 membrane alanine aminopeptidase ANPEP 248 18.5 (9.1, 28.0) 0.0001 0.0328 26.1 3 gi4505737 complement factor properdin CFP 248 -12.2 (-18.7, -5.8) 0.0002 0.0345 26.1 4 gi20357559 Charcot-Leyden crystal protein CLC 207 -3.8 (-5.9, -1.7) 0.0003 0.0424 20.6 1Number of observation (maximum number=248) 2Expected difference in no-go score per 50% increase in relative abundance of plasma protein, adjusted for child sex, age, prenatal/childhood micronutrient supplementation 3P-value of hypothesis test of null association between protein and adjusted No-go score 4Adjusted P-value correcting multiple hypothesis testing (false discovery rate) 5Percentage of variance in adjusted correct No-go score explained by protein

166

REFERENCES

1. Committee on Integrating the Science of Early Childhood Development. From Neurons to Neighborhoods : The Science of Early Childhood Development. Washington, D.C.: National Academy Press, 2000. 2. Grantham-McGregor S, Cheung YB, Cueto S, et al. Developmental potential in the first 5 years for children in developing countries. Lancet 2007;369(9555):60- 70. 3. Walker SP, Wachs TD, Grantham-McGregor S, et al. Inequality in early childhood: risk and protective factors for early child development. Lancet 2011;378(9799):1325-38. 4. Engle PL, Black MM, Behrman JR, et al. Strategies to avoid the loss of developmental potential in more than 200 million children in the developing world. Lancet 2007;369(9557):229-42. 5. Prado EL, Hartini S, Rahmawati A, et al. Test selection, adaptation, and evaluation: a systematic approach to assess nutritional influences on child development in developing countries. The British journal of educational psychology 2010;80(Pt 1):31-53. 6. O'Shaughnessy ES, Berl MM, Moore EN, et al. Pediatric functional magnetic resonance imaging (fMRI): issues and applications. Journal of child neurology 2008;23(7):791-801. 7. Webb EA, O'Reilly MA, Clayden JD, et al. Effect of growth hormone deficiency on brain structure, motor function and cognition. Brain : a journal of neurology 2012;135(Pt 1):216-27. 8. Shi T, Li X, Song J, et al. EEG characteristics and visual cognitive function of children with attention deficit hyperactivity disorder (ADHD). Brain & development 2012;34(10):806-11. 9. Chabris CF. Cognitive and Neurobiological Mechanisms of the Law of General Intelligence. Edtion ed. In: Roberts MJ, ed. Integrating the Mind: Domain General Versus Domain Specific Processes in Higher Cognition. Hove, U.K.: Psychology Press, 2008. 10. Ritsner MS, Gottesman II. Where do we stand in the quest for neuropsychiatric biomarkers and endophenotypes and what next? Edtion ed. In: Ritsner MS, ed. The Handbook of Neuropsychiatric Biomarkers, Endophenotypes and Genes: Volume II: Neuroanatomical and Neuroimaging Endophenotypes and Biomarkers: Springer, 2009:3-21. 11. Neuwelt EA, Bauer B, Fahlke C, et al. Engaging neuroscience to advance translational research in brain barrier biology. Nature reviews Neuroscience 2011;12(3):169-82. 12. Bus BA, Molendijk ML, Penninx BJ, et al. Determinants of serum brain-derived neurotrophic factor. Psychoneuroendocrinology 2011;36(2):228-39. 13. Paclt I, Koudelova J, Pacltova D, et al. Dopamine beta hydroxylase (DBH) plasma activity in childhood mental disorders. Neuro endocrinology letters 2009;30(5):604-9.

167

14. Shim SH, Hwangbo Y, Kwon YJ, et al. Increased levels of plasma brain-derived neurotrophic factor (BDNF) in children with attention deficit-hyperactivity disorder (ADHD). Progress in neuro-psychopharmacology & biological psychiatry 2008;32(8):1824-8. 15. Anderson NL, Anderson NG. The human plasma proteome: history, character, and diagnostic prospects. Molecular & cellular proteomics : MCP 2002;1(11):845-67. 16. Gomez-Pinilla F. Brain foods: the effects of nutrients on brain function. Nature reviews Neuroscience 2008;9(7):568-78. 17. Ziemssen T, Kern S. Psychoneuroimmunology--cross-talk between the immune and nervous systems. Journal of neurology 2007;254 Suppl 2:II8-11. 18. Schwarz E, Guest PC, Rahmoune H, et al. Sex-specific serum biomarker patterns in adults with Asperger's syndrome. Molecular psychiatry 2011;16(12):1213-20. 19. Nelson PG, Kuddo T, Song EY, et al. Selected neurotrophins, neuropeptides, and cytokines: developmental trajectory and concentrations in neonatal blood of children with autism or Down syndrome. International journal of developmental neuroscience : the official journal of the International Society for Developmental Neuroscience 2006;24(1):73-80. 20. Nelson KB, Grether JK, Croen LA, et al. Neuropeptides and neurotrophins in neonatal blood of children with autism or mental retardation. Annals of neurology 2001;49(5):597-606. 21. Miyazaki K, Narita N, Sakuta R, et al. Serum neurotrophin concentrations in autism and mental retardation: a pilot study. Brain & development 2004;26(5):292-5. 22. Christian P, Murray-Kolb LE, Khatry SK, et al. Prenatal micronutrient supplementation and intellectual and motor function in early school-aged children in Nepal. Jama 2010;304(24):2716-23. 23. Stewart CP, Christian P, Schulze KJ, et al. Antenatal micronutrient supplementation reduces metabolic syndrome in 6- to 8-year-old children in rural Nepal. The Journal of nutrition 2009;139(8):1575-81. 24. Christian P, Khatry SK, Katz J, et al. Effects of alternative maternal micronutrient supplements on low birth weight in rural Nepal: double blind randomised community trial. Bmj 2003;326(7389):571. 25. Tielsch JM, Khatry SK, Stoltzfus RJ, et al. Effect of routine prophylactic supplementation with iron and folic acid on preschool child mortality in southern Nepal: community-based, cluster-randomised, placebo-controlled trial. Lancet 2006;367(9505):144-52. 26. Stewart CP, Christian P, LeClerq SC, et al. Antenatal supplementation with folic acid + iron + zinc improves linear growth and reduces peripheral adiposity in school-age children in rural Nepal. The American journal of clinical nutrition 2009;90(1):132-40. 27. Cole RN, Ruczinski I, Schulze K, et al. The plasma proteome identifies expected and novel proteins correlated with micronutrient status in undernourished Nepalese children. The Journal of nutrition 2013;143(10):1540-8.

168

28. Patel SA, Murray-Kolb LE, LeClerq SC, et al. Household wealth and neurocognitive development disparities among school-aged children in Nepal. Paediatric and perinatal epidemiology 2013;27(6):575-86. 29. Bracken BA, McCallum RS. Universal Nonverbal Intelligence Test. Itasca, IL: Riverside, 1998. 30. Schulze KJ, Christian P, Wu LS, et al. Micronutrient deficiencies are common in 6- to 8-year-old children of rural Nepal, with prevalence estimates modestly affected by inflammation. The Journal of nutrition 2014;144(6):979-87. 31. de Onis M, Onyango AW, Borghi E, et al. Development of a WHO growth reference for school-aged children and adolescents. Bulletin of the World Health Organization 2007;85(9):660-7. 32. Murray-Kolb LE, Khatry SK, Katz J, et al. Preschool micronutrient supplementation effects on intellectual and motor function in school-aged Nepalese children. Archives of pediatrics & adolescent medicine 2012;166(5):404-10. 33. Christian P, Morgan ME, Murray-Kolb L, et al. Preschool iron-folic acid and zinc supplementation in children exposed to iron-folic acid in utero confers no added cognitive benefit in early school-age. The Journal of nutrition 2011;141(11):2042- 8. 34. Christian P, Murray-Kolb LE, Tielsch JM, et al. Associations between preterm birth, small-for-gestational age, and neonatal morbidity and cognitive function among school-age children in Nepal. BMC pediatrics 2014;14:58. 35. Herbrich SM, Cole RN, West KP, Jr., et al. Statistical inference from multiple iTRAQ experiments without using common reference standards. Journal of proteome research 2013;12(2):594-604. 36. Harville DA. Maximum Likelihood Approaches to Variance Component Estimation and to Related Problems. Journal of the American Statistical Association 1977;72(358):320-38. 37. Robinson GK. That BLUP is a good thing: the estimation of random effects. Stat Sci 1991;6:15-32. 38. Storey JD. A direct approach to false discovery rates. Journal of the Royal Statistical Society: Series B 2002;64:479-98. 39. Rubin DB. Multiple Imputation for Nonresponse in Surveys. New York: Wiley & Sons, 1987. 40. Geer LY, Marchler-Bauer A, Geer RC, et al. The NCBI BioSystems database. Nucleic acids research 2010;38(Database issue):D492-6. 41. UniProt C. Activities at the Universal Protein Resource (UniProt). Nucleic acids research 2014;42(Database issue):D191-8. 42. R Development Core Team. R: a language and environment for statistical computing. Vienna, Austria, 2004. 43. Baxter RC. Circulating levels and molecular distribution of the acid-labile (alpha) subunit of the high molecular weight insulin-like growth factor-binding protein complex. The Journal of clinical endocrinology and metabolism 1990;70(5):1347- 53. 44. Cohen KL, Nissley SP. The serum half-life of somatomedin activity: evidence for growth hormone dependence. Acta endocrinologica 1976;83(2):243-58.

169

45. Gunnell D, Miller LL, Rogers I, et al. Association of insulin-like growth factor I and insulin-like growth factor-binding protein-3 with intelligence quotient among 8- to 9-year-old children in the Avon Longitudinal Study of Parents and Children. Pediatrics 2005;116(5):e681-6. 46. Cheung YB, Lam KF. Three estimates of the association between linear growth failure and cognitive ability. Tropical medicine & international health : TM & IH 2009;14(9):1020-4. 47. Donovan SM, Atilano LC, Hintz RL, et al. Differential regulation of the insulin- like growth factors (IGF-I and -II) and IGF binding proteins during malnutrition in the neonatal rat. Endocrinology 1991;129(1):149-57. 48. Ooi GT, Cohen FJ, Tseng LY, et al. Growth hormone stimulates transcription of the gene encoding the acid-labile subunit (ALS) of the circulating insulin-like growth factor-binding protein complex and ALS promoter activity in rat liver. Molecular endocrinology 1997;11(7):997-1007. 49. Thissen JP, Ketelslegers JM, Underwood LE. Nutritional regulation of the insulin-like growth factors. Endocrine reviews 1994;15(1):80-101. 50. Popken GJ, Dechert-Zeger M, Ye P, et al. Brain Development. Edtion ed. In: Varela-Nieto I, Chowen JA, eds. The Growth Hormone/Insulin-Like Growth Factor Axis during Development: Advances in Experimental Medicine and Biology: Springer, 2005. 51. Nyberg F, Hallberg M. Growth hormone and cognitive function. Nature reviews Endocrinology 2013;9(6):357-65. 52. Yan H, Mitschelen M, Bixler GV, et al. Circulating IGF1 regulates hippocampal IGF1 levels and brain gene expression during adolescence. The Journal of endocrinology 2011;211(1):27-37. 53. Popken GJ, M. D-Z, P. Y, et al. Brain development. Edtion ed. In: Varela-Nieto- Chowen, Chowen JA, eds. The Growth Hormone/Insulin-Like Growth Factor Axis During Development: Springer, 2005:187-221. 54. Irshad M, Dubey R. Apolipoproteins and their role in different clinical conditions: an overview. Indian journal of biochemistry & biophysics 2005;42(2):73-80. 55. Perry LA, Stigger CB, Ainsworth BE, et al. No association between cognitive achievements, academic performance and serum cholesterol concentrations among school-aged children. Nutritional neuroscience 2009;12(4):160-6. 56. Sena A, Capela C, Nó rega C, et al. Plasma Lipoproteins in Brain Inflammatory and Neurodegenerative Diseases. Edtion ed. In: Frank S, Kostner G, eds. Lipoproteins - Role in Health and Diseases: InTech, 2012. 57. Woods AG, Sokolowska I, Taurines R, et al. Potential biomarkers in psychiatry: focus on the cholesterol system. Journal of cellular and molecular medicine 2012;16(6):1184-95. 58. Rask-Nissila L, Jokinen E, Terho P, et al. Neurological development of 5-year-old children receiving a low-saturated fat, low-cholesterol diet since infancy: A randomized controlled trial. Jama 2000;284(8):993-1000. 59. Kanai M, Raz A, Goodman DS. Retinol-binding protein: the transport protein for vitamin A in human plasma. The Journal of clinical investigation 1968;47(9):2025-44.

170

60. Woeber KA, Ingbar SH. The contribution of thyroxine-binding prealbumin to the binding of thyroxine in human serum, as assessed by immunoadsorption. The Journal of clinical investigation 1968;47(7):1710-21. 61. Stein TD, Anders NJ, DeCarli C, et al. Neutralization of transthyretin reverses the neuroprotective effects of secreted amyloid precursor protein (APP) in APPSW mice resulting in tau phosphorylation and loss of hippocampal neurons: support for the amyloid hypothesis. The Journal of neuroscience : the official journal of the Society for Neuroscience 2004;24(35):7707-17. 62. Velayudhan L, Killick R, Hye A, et al. Plasma transthyretin as a candidate marker for Alzheimer's disease. Journal of Alzheimer's disease : JAD 2012;28(2):369-75. 63. Jiang NM, Tofail F, Moonah SN, et al. Febrile illness and pro-inflammatory cytokines are associated with lower neurodevelopmental scores in Bangladeshi infants living in poverty. BMC pediatrics 2014;14:50. 64. Gimeno D, Marmot MG, Singh-Manoux A. Inflammatory markers and cognitive function in middle-aged adults: the Whitehall II study. Psychoneuroendocrinology 2008;33(10):1322-34. 65. Kuo HK, Yen CJ, Chang CH, et al. Relation of C-reactive protein to stroke, cognitive disorders, and depression in the general population: systematic review and meta-analysis. The Lancet Neurology 2005;4(6):371-80. 66. Karlsson H, Ahlborg B, Dalman C, et al. Association between erythrocyte sedimentation rate and IQ in Swedish males aged 18-20. Brain, behavior, and immunity 2010;24(6):868-73. 67. Gabay C, Kushner I. Acute-phase proteins and other systemic responses to inflammation. The New England journal of medicine 1999;340(6):448-54. 68. Hagberg H, Mallard C. Effect of inflammation on central nervous system development and vulnerability. Current opinion in neurology 2005;18(2):117-23. 69. Chen J, Buchanan JB, Sparkman NL, et al. Neuroinflammation and disruption in working memory in aged mice after acute stimulation of the peripheral innate immune system. Brain, behavior, and immunity 2008;22(3):301-11. 70. Krabbe KS, Reichenberg A, Yirmiya R, et al. Low-dose endotoxemia and human neuropsychological functions. Brain, behavior, and immunity 2005;19(5):453-60. 71. Reichenberg A, Yirmiya R, Schuld A, et al. Cytokine-associated emotional and cognitive disturbances in humans. Archives of general psychiatry 2001;58(5):445- 52. 72. Kato H, Fukuda T, Parkison C, et al. Cytosolic thyroid hormone-binding protein is a monomer of pyruvate kinase. Proceedings of the National Academy of Sciences of the United States of America 1989;86(20):7861-5. 73. Ashizawa K, Fukuda T, Cheng SY. Transcriptional stimulation by thyroid hormone of a cytosolic thyroid hormone binding protein which is homologous to a subunit of pyruvate kinase M1. Biochemistry 1992;31(10):2774-8. 74. Ashizawa K, McPhie P, Lin KH, et al. An in vitro novel mechanism of regulating the activity of pyruvate kinase M2 by thyroid hormone and fructose 1, 6- bisphosphate. Biochemistry 1991;30(29):7105-11. 75. Chow CK. Increased activity of pyruvate kinase in plasma of vitamin E-deficient rats. The Journal of nutrition 1975;105(9):1221-24.

171

76. Dehio C, Gray-Owen SD, Meyer TF. The role of neisserial Opa proteins in interactions with host cells. Trends in microbiology 1998;6(12):489-95. 77. Unsal-Kacmaz K, Mullen TE, Kaufmann WK, et al. Coupling of human circadian and cell cycles by the timeless protein. Molecular and cellular biology 2005;25(8):3109-16. 78. Allada R, White NE, So WV, et al. A mutant Drosophila homolog of mammalian Clock disrupts circadian rhythms and transcription of period and timeless. Cell 1998;93(5):791-804. 79. Engelen E, Janssens RC, Yagita K, et al. Mammalian TIMELESS is involved in period determination and DNA damage-dependent phase advancing of the circadian clock. PloS one 2013;8(2):e56623. 80. Sakai T, Tamura T, Kitamoto T, et al. A clock gene, period, plays a key role in long-term memory formation in Drosophila. Proceedings of the National Academy of Sciences of the United States of America 2004;101(45):16058-63. 81. Hooper VS, Bell SM. Concurrent validity of the Universal Nonverbal Intelligence Test and the Leiter International Performance Scale–Revised. Psychol Schs 2006;43:143–8.

172

6 CHAPTER 6: EFFECT OF ANTENATAL MICRONUTRIENT SUPPLEMENTATION ON PLASMA PROTEIN PROFILES IN SCHOOL-AGED CHILDREN IN RURAL NEPAL

ABSTRACT

Background: Antenatal micronutrient (MN) supplementation in malnourished

populations may have long-term effects on health outcomes in children. However,

phenotypical changes are poorly understood at the molecular level. Proteins that are

proximate mediators of health attainment and consequence may offer a unique

opportunity to explore biological processes sensitive to early micronutrient exposure.

Objective: To investigate the effect of micronutrient supplementation during pregnancy

on plasma protein profiles of school-aged children in rural Nepal. Design and methods:

Pregnant women were cluster-randomized early in pregnancy to take daily one of five

dietary supplements daily through 3 months postpartum containing recommended

amounts of folic acid (FA), iron-folic acid (IFA), iron-folic acid-zinc (IFAZn), multiple

micronutrients (i.e., IFAZn + 11 others, MM) or a placebo control (all tablets contained

vitamin A). We followed at 6-8 years of age a subset of 500 children born to mothers in

the trial (100 per maternal group) and quantified protein abundance in plasma using

quantitative mass spectrometry. We estimated mean differences in relative abundance of

proteins to identify differentially abundant proteins and performed Gene Set Enrichment

Analysis to identify differentially enriched pathways by maternal MN intervention (each

compared to the control group). Results: Among 965 child plasma proteins analyzed, no

proteins or gene sets were significantly differentially abundant or enriched in any

maternal intervention groups, compared to the maternal control group, passing a false

173 discovery rate (FDR) threshold of 5%. In sex-stratified analyses, maternal folic acid supplementation increased relative abundance of nesprin-1 (SYNE1) by 50.9 (24.7,

82.8)% among boys (q=0.0144) and positively enriched 4 Gene Ontology gene sets related to cytoskeleton and development of organs among girls (all passing FDR threshold of 5%). No effects were found in other maternal supplementation groups.

Conclusion: This exploratory study showed that there are no prominent differences in the abundance of plasma proteins in children by early school age by antenatal micronutrient supplementation. Further studies are warranted to evaluate subtle but coordinated long- term effects of prenatal folic acid supplementation on the integrity of cellular structure and developmental processes in undernourished children.

174

INTRODUCTION

Although great effort and substantial progress have been made to improve micronutrient status in pregnant women and intrauterine growth through antenatal micronutrient supplementation in undernourished populations, long-term effects on child health remain to be fully elucidated (1, 2). A limited number of follow-up studies, mostly comparing multiple micronutrient to iron-folic acid, suggest that effects may persist through infancy or early childhood, conferring advantage to infant development (3-5), growth (6), survival (7, 8) and cognitive function (9, 10), as well as reduced risks of morbidity and metabolic syndrome (11, 12). On the other hand, some studies have shown that early gains in growth or other health outcomes may not persist through infancy or disappear entirely over a longer period time (13-18), leaving the long-term effects of antenatal micronutrient supplementation on child health inconclusive and a priority for public health research.

The main challenge of evaluating child health in response to maternal micronutrient supplementation is that the effects might be subclinical or metabolic in function, which would render them not easily captured by conventional assessments in population-based nutritional research (19, 20). It is also possible that challenging postnatal environmental exposures may overshadow phenotypic responses to micronutrient supplements that occur in prenatal life. Although there is a lack of consensus on the persistence of micronutrient effects, accumulating studies suggest that peri-conceptional or early exposure to folic acid or methyl donor micronutrients are associated with metabolic, immune, and neurodevelopmental health outcomes in children (21-26). More consistent evidence has been obtained through carefully controlled animal studies, providing

175 empirical evidence of nutritional programming through epigenetic regulation or changes in cellular functions or structure. These can cause quantitative or qualitative alterations in specific or global levels of tissues or organ (27-32). A limited number of studies exist supporting programming driven by nutrients such iron, zinc, and other vitamins (33-37).

We hypothesize that the plasma proteome, the entire set of proteins in the blood stream, might be a good tissue in which to explore differences in function and health associated with variation in micronutrient nutriture (38). Proteins are intermediate biomolecules that can reflect the pattern of gene expression and mediate measurable phenotypical changes. As active participants in every physiological and metabolic process, plasma proteins are involved in blood homeostasis, the immune system, hormonal regulation, and nutrient metabolism. But, proteins can be leaked or secreted from cells and tissues throughout the human body as a result of regular cellular metabolic processes. (39). Thus, profiling plasma proteins with an untargeted approach may expand our ability to identify and quantify subtle biological responses to early-life micronutrient exposure.

A double-blind randomized community-based antenatal micronutrient intervention trial in southeastern Nepal was previously conducted in where multiple micronutrient deficiencies were frequently detected during pregnancy (40, 41). In this trial, community clusters were randomly assigned for women, once pregnant, to receive a placebo control, folic acid alone or folic acid with other micronutrients from early pregnancy to 3 months postpartum, and their children were followed up at 6 to 8 years of age (11, 42). Previous analyses have revealed antenatal micronutrient supplementation improved birth size

(iron-folic acid and multiple micronutrient) and benefited a variety of child health

176 outcomes, including child survival, cognitive function (iron and folic acid) and linear growth (iron-folic acid-zinc), and reduced risk of metabolic syndrome (folic acid) (9, 11,

41-43). In the present study, we aim to explore the plasma proteome of the children to identify differentially abundant proteins or differentially enriched biological pathways related to maternal micronutrient supplementation. This study provides a unique opportunity to evaluate the long-term effects of micronutrient supplementation during pregnancy on untargeted phenotypical changes in offspring in undernourished populations.

SUBJECTS AND METHODS

Antenatal micronutrient intervention study In 1999-2001, the Nepal Nutrition Intervention Project-Sarlahi (NNIPS) of the

Center for Human Nutrition at the Johns Hopkins Bloomberg School of Public Health carried out its third major trial in the southern plains District of Sarlahi, a community- based, cluster randomized controlled trial of antenatal micronutrient supplementation.

This study aimed to improve birth weight and reduce infant mortality. The study procedures and outcomes have been described in detail in previous publications (41, 43).

Briefly, 30 contiguous village development communities (sub-districts), comprising one- third of the district, with a total population of 200,000 were selected for the study.

Criteria for area selection was that communities not be located in the northern hills or too close to the border with Bihar, India (>5 km) to favor population stability. They were further divided into 426 sectors, which served as units (clusters) of randomization.

Married women of reproductive age were enumerated and visited every 5 weeks to be asked if they menstruated in the previous month. Excluded from surveillance were

177 menopausal, sterilized or widowed women, as well as, initially, women who were currently pregnant or breastfeeding an infant less than 9 months of age. Once pregnancy was ascertained by urine test, women were enrolled into the study. In total, 4,926 consenting pregnant women participated in the trial, receiving daily from first trimester to

3 months postpartum consumption one of the five micronutrient supplements: 1) a placebo control; 2) folic acid (400 μg); 3) folic acid with iron (60 mg in the form of ferrous fumarate); 4) folic acid with iron and zinc (30 mg); or 5) a multiple micronutrient containing folic acid, iron, zinc, and an additional 11 vitamins and minerals (10 mg vitamin D as cholecalciferol, 10 mg vitamin E as d-a tocopherol, 1.6 mg thiamine, 1.8 mg riboflavin, 20 mg niacin, 2.2 mg vitamin B-6, 2.6 mg vitamin B-12, 100 mg vitamin C,

65 mg vitamin K as phylloquinone, 2.0 mg Cu, and 100 mg Mg). All supplements contained vitamin A (1,000 μg retinol equivalents of vitamin A in the form of retinyl acetate) based on earlier findings that vitamin A supplementation could reduce pregnancy-related maternal mortality in this population (44). With the exception of iron, administered doses of micronutrients in the supplements approximated a recommended dietary allowance. Supplements comprised tablets of identical shape, size and color.

Investigators, participants, and field workers were blinded to the allocation code throughout the study. Compliance of daily micronutrient supplement intake was 88%

(median) and did not vary by randomized group (41).

Child follow-up assessments and sample selection Among 4,130 live-born infants during the trial, 3,524 children were followed in

2006-2008 for growth, diet, nutritional status and health assessments, which included blood collection at 6 to 8 years of age. All follow-up procedures and outcomes have been

178 reported previously (11, 42). Field assessment teams interviewed and obtained information from heads of household about socioeconomic status and child education.

Anthropometrists measured childrens’ weight, height, triceps and subscapular skinfold thickness, mid-upper arm circumference (MUAC), and waist circumference.

Phlebotomists collected early-morning fasting venous blood (10 ml in the sodium heparin-containing tubes without preservatives or antioxidants). The bio-specimens were brought to a central laboratory and centrifuged for plasma extraction. Plasma samples were equally aliquoted into 4 tubes (at least 0.5 ml of plasma per aliquot), frozen in liquid nitrogen, periodically shipped to the Johns Hopkins University Center for Human

Nutrition, and stored at -80 ◦C for further assessment.

Among the 3,305 children whose follow-up blood samples were available, specimens for 2,726 children had all 4 full plasma aliquots. Of these children, 2,130 specimens were eligible for proteomics study consideration because they met additional criteria of having no missing values for any epidemiologic data from either the original trial or follow up study, and their birth weight was measured within 72 hours after birth.

We stratified eligible children by maternal supplementation group and ordered them by calendar date of blood draw in the field during the follow-up study. Two hundred children were systematically selected following a random start within each maternal supplementation group, employing a selection interval to yield n=200 per group or

N=1000. Multiple micronutrient and inflammation biomarker assays were performed on this set of child plasma samples forming a designated nutrition bio-archive. Micronutrient status of 1000 children have been reported previously (45). In the final step for the present study, each set of 200 specimens per original maternal supplement stratum were

179 ordered by date of field blood collection, and following a flip of coin every other specimen was designated for proteomics analysis, yielding a set of specimens from n=500 children, balanced by calendar time across the five maternal intervention arms.

This final sample size (100 per each maternal supplementation group) was assessed to be sufficient to detect an approximate 30% difference in mean relative protein abundance between each maternal supplement versus the control group, given a standard deviation of delta 1 with 80% power (1-β error) and a significance level (α error) of 0.05.

Proteomics analysis

The procedures of proteomics have been previously described in detail (38).

Briefly, plasma samples of the 500 children and a masterpool sample, which consisted of an equal amount of plasma drawn from the nutrition bioarchive sample of 1000 children, were immune-depleted to remove 6 high abundant plasma proteins (albumin, immunoglobulin A, immunoglobulin G, transferrin, haptoglobin, and anti-trypsin) using a

Human-6 Multiple Affinity Removal 200 System (MARS) LC column (Agilent

Technologies). Proteomics analysis was performed at the Proteomics and Mass

Spectrometry Core within the Johns Hopkins School of Medicine. Immune-depleted samples were digested overnight by trypsin (Promega, sequencing grade). Seven randomly selected samples and one masterpool sample were randomly labeled with 8 isobaric tags for relative and absolute quantitation (iTRAQ) reagents (AB Sciex, http://www.absciex.com) that contain different reporter ions which can be used as measure of relative amount of peptide according to manufacturer’s instructions. All samples were mixed together and fractionated into 24 fractions by strong cation exchange

(SCX) chromatography. iTRAQ labeled peptides in each SCX fraction were loaded to a

180 reverse phase nanobore column, and eluted peptides were sprayed through a 10 μm emitter tip into an LTQ Orbitrap Velos mass spectrometer (Thermo Scientific, www.thermo.com/orbitrap) interfaced with a NanoAcquity UPLC (Waters Corp). From each survey scan, up to 10 peptide masses were individually isolated, fragmented and analyzed. Isotopically resolved masses in mass spectrometric (MS) and MS/MS spectra were extracted with and without deconvolution using Thermo Scientific Xtract software and searched against the RefSeq 40 protein database using Mascot (Matrix Science, www.matrixscience.com) through Proteome Discoverer software (v1.3, Thermo

Scientific). Peptides with high (> 95%) confidence from Mascot searches were filtered within the Proteome Discoverer with a false discovery rate (FDR) less than 5%.

Statistical analyses Details of estimation of relative abundance of proteins is reported elsewhere (46).

Briefly, reporter ion intensities derived from Proteome Discoverer v1.3 were log2 transformed and median normalized for each reporter ion intensity spectrum. Relative protein abundance was estimated by calculating the median of all the de-median log2 ion intensities across all spectra belonging to each protein. Corrections for differences in amounts of material loaded in the channels and sample processing were carried out by subtracting the channel median from the relative abundance estimate, normalizing all channels to have median zero. We employed linear mixed effects models (LME) with antenatal micronutrient supplementation groups as fixed effects and iTRAQ experiment as a random effect to take into account any random effects of iTRAQ experiments which can be derived by extreme values of relative abundance of proteins. We estimated mean differences in protein relative abundance between each of the 4 maternal micronutrient

181 supplementation groups and the placebo control group. Statistical significance was calculated by testing the null hypothesis of no difference in relative protein abundance in each 2-group comparison. We estimated a q-value to control false discovery and considered proteins passing an FDR threshold <5.0% as being significantly differentially abundant (47). Maternal characteristics at baseline and child characteristics at follow-up were included in the model as covariates to account for imbalance across groups. Among the 500 children, 259 children participated in an early childhood micronutrient supplementation trial during which they were administered daily zinc alone, iron with folic acid, both combined, or neither from 12 to 36 months of age (48). Thus, dummy variables were added to the model to adjust for potential confounding effects resulting from participation in this trial. Interaction between antenatal and child intervention terms could not be tested because of inadequate cell sample sizes.

Each protein was functionally annotated based on its gene identification using the

Gene Ontology (GO) database. GenInfo identifiers (GI) of proteins were converted to corresponding official gene symbols based on the Human Genome Organization (49).

Because all gene symbols should be unique to run Gene Set Enrichment Analysis

(GSEA), redundant gene symbols due to different protein isoforms were searched and an isoform with a larger sample size was selected for analysis. The GO database for gene sets was downloaded from the Molecular Signatures Database (MSigDB, v4.0) which is a collection of annotated gene sets for GSEA analysis (50, 51). Gene sets whose members matched at least 5 proteins with our protein list were included in analysis. All analytical processes of GSEA have been described in detail elsewhere (51). Briefly, proteins were ranked based on observed t-statistics derived from the LME models, which were used to

182 identify differentially abundant proteins by maternal supplementation group. By walking down the ordered list of proteins, a running-sum statistic of each set was estimated by increasing enrichment score (ES) when a protein was a member of the set, and decreasing

ES when it was not. The ES was defined as a maximum or minimum value from zero. A positive/negative ES indicates positive/negative enrichment of a given set in the protein abundance profile of children in each maternal supplementation group, compared to the control group. To calculate the significance of the ES, we shuffled maternal micronutrient supplementation indicators within iTRAQ experiments to preserve the correlations between proteins in the dataset. Significance of the ES was defined as the portion of the null distribution of the ES against 1,000 permuted ES corresponding to a value equal or greater than the observed ES. A Normalized Enrichment Score (NES) of each set was calculated by an observed ES dividing by the mean of permuted ESs across all gene sets to takes into account differences in the size of gene sets. Multiple hypothesis testing was corrected by calculating a ratio of the two distributions: observed ES versus ES from the permuted dataset and observed ES versus ES of all gene sets in the actual dataset.

Leading-edge members of a gene set, which are defined as actively participating members in gene sets, were identified by a subset of members of a given set which appears in the ranked protein list at or before the point where the running sum reaches its maximum deviation from zero. We considered gene sets passing a FDR threshold of 5% significantly enriched.

Based on a literature suggesting potential sex-specific effects of antenatal micronutrient supplementation (52, 53), we stratified data by child sex and repeated

183 analyses to identify differentially abundant proteins and enriched gene sets among girls and boys, separately.

All statistical analyses were carried out using the free software environment R version 3.1.0 (54). The original antenatal micronutrient supplementation trial received ethical approval from the Nepal Health Research Coucil, Kathmandu and the Institutional

Review Board (IRB) at the Johns Hopkins Bloomberg School of Public Health (JHSPH),

Baltimore, MD, the USA. Children follow-up study was approved by IRBs at the JHSPH and Institute of Medicine and Tribhuvan University, Kathmandu, Nepal.

RESULTS

A flow diagram of enrollees into the maternal micronutrient supplementation, child follow-up and proteomics studies is provided in Figure 6.1. The characteristics of children at the time of follow-up assessment, and maternal characteristics in the 1st trimester at the outset of the trial are compared by original supplementation groups in

Table 6.1. The mean (SD) age of children at follow-up was 7.5 (0.4) years. Half (50.5%) of the children were male, one-third of Pahadi ethnicity, reflecting a hills origin areas in

Nepal, with the remaining being Madheshi ethnicity, reflecting a north Indian origin.

Two-thirds (66.8%) of the children had ever attended school. These characteristics did not differ by randomized maternal supplementation group. Weight, height, left mid-upper arm circumference (MUAC) and body mass index (BMI) were also similar across groups.

Children were generally undernourished, reflected by a high prevalence of stunting, underweight, and low body mass index (BMI), with some variation evident with respect to the prevalence of underweight (P=0.045). Diet in the previous week was generally similar across groups, except for frequencies of egg and dark green leafy vegetable

184 intakes (P=0.020 and 0.027, respectively). There were no differences in morbidity history in the previous week by maternal group. The prevalence of acute inflammation, indicated by elevated C-reactive protein concentration (CRP >5 mg/L), varied (P=0.048), but the prevalence of elevated α-1-acid glycoprotein concentration (AGP>100 mg/dL) was comparable across groups. Characteristics of this subset of 500 children were similar to those of children from the larger, original child follow-up cohort (Appendix 6.1).

Maternal height was similar across groups, while age, parity, weight, BMI, MUAC, and literacy at as assessed at the outset of the original trial significantly differed by maternal allocation (P<0.02 for all) and these measures were considered as covariates in final models to adjust findings for these imbalances.

Plasma samples from 500 children born to mothers allocated to the 5 arms of original micronutrient supplementation trial were well balanced across 72 iTRAQ experiments

(Appendix 6.2). Mass spectrometry identified a total of 4,705 plasma proteins, among which 988 were quantified in >10% of all plasma samples (n>50). Six hypothetical proteins and 17 proteins of significant mean differences by iTRAQ channels were excluded, yielding a total of 965 proteins for analysis (Appendix 6.3).

The distributions of P values testing mean differences in relative protein abundance between each maternal supplement and the control group were almost uniform in histograms and not different from expected distributions in the quantile-quantile plots

(qq-plots) (Figure 6.2 A-B). Among 965 plasma proteins analyzed, no proteins were differentially abundant in any comparison to the control group. Sex-stratified analyses were restricted to 587 and 589 proteins with at least 75 samples (approximately 15 samples per each maternal supplementation group) for boys and girls, respectively. The

185 q-q plots showed that observed distributions of P of mean differences in relative protein abundance were not different from expected distributions of P values in both boys and girls (Figure 6.3 A-B). Compared to the maternal control group, one protein, nesprin-1

(gene symbol: SYNE1), was significantly more abundant in children of maternal folic acid supplementation group among boys (q= 0.0144). Maternal folic acid supplementation increased relative abundance of nesprin-1 by 50.9% (95% CL: 24.7,

82.8%) among boys. The 95% confidence intervals of mean differences in nesprin-1 between folic acid and the control group of boys and girls slightly overlapped, suggesting that the folic acid effect was greater in boys than girls and this sex-specific effect was marginally significant (Table 6.2).

Among 1,454 gene sets available in the GO database, we excluded gene sets smaller than 5. This resulted in 450 and 312 gene sets available for sex-combined and sex- stratified GSEA, respectively. In the sex-combined analysis, no gene sets were significantly enriched by antenatal micronutrient supplementation. In sex-stratified analyses, 4 gene sets labeled intracellular organelle part (GO: 44446), cytoskeleton (GO:

5856), intracellular non membrane bound organelle (GO: 43232) and organ development

(GO:48731) were positively enriched by maternal folic acid supplementation among girls

(Table 6.3). The leading-edge members of intracellular organelle part, cytoskeleton, and intracellular non membrane bound organelle gene sets mostly overlapped, showing that the positive enrichments were attributable to the same proteins (Appendix 6.4). They are alpha and beta-actins (ACTA and ACTB), myosins (MYH10, MYH9, and MYL6), keratin1/2, tropomyosin (TPM4), and moesin (MSN) among others. Leading-edge proteins of organ development gene set included extracellular proteins such as growth

186 promoting proteins (IGF1/2/BP3, and plasminogen), bone development (SPARC), and glycosaminoglycan binding proteins (COMP, APOA4, ENG, and PF4) and keratins

(KRT 5/9/10/14/16) among others. The enrichment plots of cytoskeleton and organ development between boys and girls showed substantially different patterns (Appendix

6.5 A-D). No gene sets were enriched by antenatal maternal micronutrient supplementation among boys. All findings of this study remained essentially unchanged when the FDR threshold was relaxed from 5% to 10%.

DISCUSSION

Using untargeted quantitative proteomics, we examined the effects of daily antenatal supplementation with different combinations of micronutrients, from early pregnancy to childbirth, on the plasma protein profiles of school-aged children in rural Nepal.

Inference favors a cause-effect interpretation because maternal micronutrient supplement exposure was randomized, as occurred through participation in a community-based, placebo-controlled trial. Among 965 plasma proteins analyzed, no plasma proteins or gene sets were differentially abundant or enriched by different combinations of maternal micronutrient supplementation. Effects of folic acid supplementation appeared in sex- stratified analyses with one positively abundant protein among boys and a couple of positively enriched gene sets among girls. Overall, these results suggest that no prominent but only subtle sex-specific changes in abundance of plasma proteins were detected in children 6-8 years after being born to mothers who were regularly supplemented various combinations of prenatal micronutrient supplements.

We can postulate several explanations about the lack of overall effect of maternal micronutrient supplement exposure on the early school-aged childhood plasma proteome.

187

It is possible that there were truly no such quantitative changes in plasma proteins. It may be that changes in protein abundance were initially present in postnatal period but were not stable and may have been vitiated or normalized by cumulative, post-natal environmental exposures by the early school aged years. There may have been epigenetic changes during fetal life, but biochemical alterations were not translated into differences in protein expression. It is also possible that biological responses were not detectable at the sensitivity level of our instruments. It is likely that proteins that may be responsive to prenatal nutrient exposure are tissue proteins. Due to the substantial dilution of blood and the dynamic range of protein concentrations in plasma, these low-abundant proteins may not have been detected by mass spectrometry, unless changes were uni-directional and systemic. In addition, we cannot rule out weak statistical power caused by multiple hypothesis testing of a large number of proteins in a modest number of subjects per group.

There are no comparable studies to the present study, leaving these explanations in need of future testing.

A single, differentially abundant protein and several enriched gene sets were revealed when between-group analysis was stratified by child sex. While valid in a cluster randomized design, as conducted (41), we are cautious to not over-interpret these findings because of increased random chances with consequent smaller between group sample sizes after sex stratification. However, sex-specific effects of maternal nutritional status or interventions have been frequently reported to affect the postnatal phenotype.

For example, maternal multiple micronutrient (vs. standard-of-care iron-folic acid) supplementation increased concentrations of IGF-1 and leptin in the cord blood of

Burkina Faso infants, but only among boys (53). A global epigenome study showed that

188 there were few overlaps in the differentially methylated loci by periconceptional multiple micronutrient supplementation between male and female Gambian infants (52). Maternal folic acid supplementation before and during pregnancy reduced methylation at the differentially methylated region (DMR) of H19 more significantly in cord blood leucocytes of male infants (55). In animal studies, maternal folate or methyl-deficient diets (MD) during the periconceptional period has shown sex-specific effects on DNA methylation of specific genes in the gut and glucose homeostasis in murine offspring (56,

57). Although the underpinning mechanisms are poorly understood and we could not formally test interactions, the results of this study support sexual dimorphism in response to prenatal folic acid supplementation.

Our study showed that among 589 proteins analyzed among boys, one, nesprin-1

(SYNE1) was differentially abundant attributed to prenatal folic acid supplementation.

Although this protein was identified and quantified in a small number of children (~30 boys per group), the strong significance of the difference (P=0.00002) suggests that this change may reflect a true effect of maternal folic acid supplementation. Interestingly, the effects of other supplements that also contained folic acid were not as large or significant as folic acid alone, suggesting possible antagonism from other micronutrients or effects of chance. Nesprin-1, or spectrin repeat containing, nuclear envelope 1, is expressed in skeletal and smooth muscle and is also abundant in the cerebellum (58, 59). Nesprin-1 mediates actin microfilament interactions with the nuclear envelope and defects in

SYNE1 is associated with nuclear mislocalization and autosomal recessive cerebellar ataxia type 1 (60, 61). A study showed that its expression increased during embryonic stem cell differentiation (62). There is no previous report that expression of this protein

189 is modified by maternal nutrition. An animal study reported that prenatal tobacco smoke exposure up-regulated expression of this gene in the mouse embryo hippocampus, suggesting it may be susceptible to environmental factors during the early developmental period (63). Because folic acid is a critical micronutrient for normal embryonic developmental processes (64), more research is required to investigate a biological link between maternal folate status and expression of nesprin-1 in offspring.

Positive enrichments in cytoskeleton-related and organ development gene sets by maternal folic acid supplementation among girls suggest that prenatal folic acid exposure may induce moderate but coordinated changes in the abundance of proteins involved in maintaining the integrity of cellular structure and in promoting development of tissues or organs. This inference is supported by a study in rats that reported both high and low maternal dietary folate intake to change expression of ~50 transcripts related to cytoskeleton in the fetal cardiac tissue transcriptome (65). A maternal methyl-deficient diet has been noted to change abundance of cytoskeletal proteins, among others, in the rat liver proteome (66). Human studies have found that periconceptional or pregnancy maternal folic acid supplementation can affect the DNA methylation rate on the DMR that regulates expression of IGF-2 in the cord blood of infants, although these studies did not report gene expression (55, 67, 68). One study reported that a folic-acid-restricted maternal diet reduced expression of IGF1 in skull bone and serum IGF1 in rat embryos

(69). In addition, animal studies have shown that maternal folate status or folic acid supplementation changes expression of genes or proteins involved in a wide range of cellular and metabolic functions of the liver, skeletal muscle, and brain cerebral hemispheres (29-32).

190

Although there is little empirical evidence, biological plausibility of long-lasting effects of prenatal folic acid supplementation exists. Folate and folic acid are involved in

DNA methylation which is a core process in the regulation of gene expression (70).

Folate also regulates DNA and RNA synthesis which are essential for cell proliferation and differentiation (71). Due to these roles, folate status of women may affect expression of proteins involved in fundamental processes that maintain cellular structure and function, and therefore tissue and organ function. The study by Swali et al. supports that different phenotypical changes in tissues and organs caused by nutritional programming are derived from common gatekeeper processes during early development such as cytoskeleton remodeling and cell cycle regulation, which are important for cell proliferation and DNA integrity (72). Previous studies have shown that maternal folic acid supplementation reduces folate deficiency in pregnant women and reduces risks of microalbuminuria and metabolic syndrome in children, suggesting supplementation with this nutrient may induce long-lasting changes in kidney function (11, 73). However, it is unclear whether the positive enrichments we observed are associated with specific tissues or systemic change, or if there are child health implications.

Keratins were among the observed enriched gene sets. These proteins are frequently detected in the plasma proteome (74) and commonly considered contaminants related to sample processing (75). However, some keratins in serum or plasma have been increasingly viewed as clinical markers of abnormal cellular proliferation (76-78), suggesting their abundance in plasma may have cellular physiological or metabolic relevance (79). We randomly assigned plasma specimens to undergo sample preparation and iTRAQ experimentation. Thus, while it is unlikely that contamination selectively

191 occurred by maternal supplementation group, the biological meaning of variation in plasma keratin abundance remains yet unclear.

This study has unique strengths as well as known methodological limitations. To our knowledge, this is the first human study that has explored the impact of prenatal micronutrient supplementation on the plasma protein profile of offspring. In addition, this study is one of a few randomized control trials during pregnancy where the referent group did not receive iron-folic acid. This allowed us a rare opportunity to explore specific effects of folic acid supplementation within an experimental design. Although we selected a sub-sample from a larger child follow-up cohort, the random selection process appeared to protect the representativeness of study children with respect to those in the larger trial that was carried out in a typical, rural South Asian context. Gene set enrichment analysis, which uses weighted sums of associations between proteins and exposure, provides power to detect modest changes in the abundance of proteins.

Among its limitations, this study had a considerable amount of missing data, common to proteomics studies, which substantially restricted the number of proteins, power to detect differences between groups and the scope of gene sets available for analyses. While our untargeted investigation may be considered unbiased, targeting a specific range of plasma proteins guided by hypotheses to test may be preferred since low-abundant tissue proteins may be a particularly responsive set of biomolecules to changes in prenatal micronutrient exposure. Lastly, a small sample size restricted our ability to test interactions between prenatal and childhood micronutrient supplementation, as occurred during a postnatal trial (48), which may have limited our ability to isolate effects attributable solely to prenatal supplementation.

192

CONCLUSIONS

To date, little evidence is available on the postnatal proteome responses among offspring to prenatal micronutrient interventions or maternal status. This exploratory study suggests that prenatal micronutrient supplementation may not substantially change relative abundance of circulating proteins in children by early school age. Given folic acid supplementation is a widely accepted standard of prenatal care, its use or status has been shown to be associated with epigenetic changes, and this study provides suggestive evidence of subgroup proteomic differences, additional studies that evaluate enduring effects of folic acid supplementation on gene-specific, functional and health outcomes in offspring appear warranted.

193

Table 6.1. Characteristics of children, mothers, and households at baseline and follow-up by antenatal micronutrient supplementation group1

Control FA IFA IFAZn MM P2 Child characteristics at follow-up (n=100) (n=100) (n=100) (n=100) (n=100) Male, % 51 52 47 51 48 0.945 Age, years 7.5 (0.4) 7.5 (0.4) 7.4 (0.5) 7.5 (0.5) 7.5 (0.5) 0.373 Pahadi ethnicity, % 34 34 34 28 29 0.791 Ever sent to school, % 63 64 69 69 69 0.798 Literacy ,% 24 13 16 17 17 0.338 Anthropometric measurements Weight, kg 18.4 (1.8) 18.2 (2.3) 18.4 (3.7) 18.3 (2.4) 18.0 (2.2) 0.738 Height, cm 114.5 (4.8) 113.8 (5.7) 114.2 (6.4) 114.6 (7.1) 113.6 (5.6) 0.695 MUAC, cm 15.4 (1.0) 15.5 (1.2) 15.5 (1.5) 15.6 (1.2) 15.4 (1.0) 0.768 BMI, kg/m2 14.1 (1.0) 14.0 (1.1) 14.0 (1.3) 14.0 (1.3) 13.9 (0.9) 0.917 Undernutrition3, % Stunting (HAZ<-2) 34 43 36 37 45 0.436 Underweight (WAZ<-2) 39 52 49 43 59 0.045 Low BMI (BMIZ<-2) 17 17 15 19 14 0.893 Diet, any intake in the past week, % Milk 73 64 68 71 67 0.684 Chicken 12 20 21 26 20 0.175 Fish 27 24 18 20 19 0.497 Other meat 30 34 35 30 32 0.917 Eggs 15 16 17 29 12 0.020 DGLV 67 63 79 80 73 0.027 Morbidity, any symptom reported in previous week, % Poor appetite 14 11 17 9 10 0.407 Fever 7 9 5 10 10 0.645 Diarrhea 4 3 2 3 4 0.924 Productive cough 1 3 5 5 5 0.478 Rapid breathing 3 3 2 4 2 0.905

195

Any symptom 24 23 24 25 25 0.997 Inflammation, % CRP > 5 mg/L 1 9 10 6 4 0.048 AGP > 100 mg/dL 28 30 38 26 27 0.350 CRP > 5 mg/L and AGP > 100 mg/dL 1 7 9 6 3 0.082 Any inflammation 28 32 39 26 28 0.282 Household characteristic, % Caste/Religion Brahmin and Chhetri 17 7 16 17 13 0.047 Vaiysha 55 69 71 68 63 Shudra, Muslim, and others 28 24 13 15 24 First floor No walls, grass sticks, or branches 61 51 51 48 54 0.206 Katcha 30 39 31 35 38 Wood planks or Pakka 9 10 18 17 8 Electricity in home 58 41 53 56 47 0.101 Land ownership 72 77 81 73 82 0.327

Control FA IFA IFAZn MM P2 Maternal characteristics at baseline (n=100) (n=100) (n=100) (n=100) (n=100) Age, years 24.1 (5.4) 23.8 (6.3) 23.8 (5.0) 23.8 (5.9) 21.7 (5.5) 0.019 Parity 2.6 (2.4) 2.5 (2.3) 2.6 (2.1) 2.2 (1.9) 1.7 (2.0) 0.017 Weight, kg 42.5 (5.3) 42.6 (5.7) 45.0 (6.2) 43.7 (5.4) 43.6 (5.3) 0.011 Height, cm 150.4 (5.6) 150.8 (5.4) 150.8 (5.4) 150.0 (5.1) 150.6 (4.8) 0.807 BMI, kg/m2 18.8 (2.0) 18.7 (2.0) 19.7 (2.1) 19.4 (1.9) 19.2 (1.8) 0.001 MUAC, cm 21.4 (1.8) 21.4 (1.7) 22.2 (2.0) 22.0 (1.8) 21.9 (1.9) 0.008 Literacy, % 19 14 29 24 25 0.099 Abbreviations: FA, folic acid; IFA, Iron-folic acid; IFAZn, Iron-folic acid-zinc; MM, Multiple micronutrient; BMI, body mass index; MUAC, middle- upper arm circumference; HAZ, height-for-age z-score; WAZ, weight-for-age z-score; BMIZ, BMI-for-age z-score; CRP, c-reactive protein; AGP, alpha-1-acid glycoprotein 1Values are means (SD) or percentages 2Using analysis of variance test for continuous variables and the chi-square test for categorical variables. 3Calculated by using the World Health Organization growth reference for school-aged children.

196

Figure 6.2. Histograms and quantile-quantile (q-q) plots of p-values of mean differences in relative abundance of plasma proteins by maternal supplementation group relative to the control group among children 6-8 years of age in Sarlahi, Nepal1 A. Histogram B. q-q plots

1Mean differences in relative abundance of 965 plasma proteins between each maternal supplementation groups and the control group were estimated using linear mixed models adjusting for childhood micronutrient intervention, child age, egg and dark green leafy vegetable intakes in the past week, plasma C-reactive protein concentration at follow-up, caste, and maternal age, literacy, and body mass index at baseline.

197

Figure 6.3. Quantile-quantile plots of p-values of mean differences in relative abundance of plasma proteins by maternal supplementation group relative to the control group, stratified by sex among children 6-8 of years of age in Sarlahi, Nepal A. Girls B. Boys

1Mean differences in relative abundance of 589 (for girls) and 587 (for boys) plasma proteins between each maternal supplementation groups and the control group were estimated using linear mixed models adjusting for childhood micronutrient intervention, child age, egg and dark green leafy vegetable intakes in the past week, plasma C-reactive protein concentration at follow-up, caste, and maternal age, literacy, and body mass index at baseline.

198

Table 6.2. Expected difference in relative abundance of nesprin-1 (SYNE1) by maternal supplementation group relative to the control, stratified by sex in children 6-8 years of age in Sarlahi, Nepal1

Maternal supplement group Multiple Protein Sex n Folic acid2 Iron-folic acid2 Iron-folic acid-zinc2 micronutrient2 Boys 132 50.9 (24.7, 82.8) 15.0 (-1.3, 33.8) 25.6 (6.5, 48.3) 29.7 (8.4, 55.2) Nesprin-1 (gi23097308) Girls 124 2.1 (-16.6, 24.9) -3.4 (-17.2, 12.8) 7.6 (-8.2, 26.0) -3.1 (-18.3, 14.9) 1Among 587 plasma proteins, SYNE1 was differentially abundant in maternal folic acid supplementation group compared to the control group among boys, passing a false discovery rate threshold of 5.0%. 2Expected difference in relative abundance of protein between maternal intervention and the control group was estimated adjusted for childhood micronutrient intervention, child age, egg and dark green leafy vegetable intakes in the past week, plasma C-reactive protein concentration, caste, and maternal age, literacy and body mass index at baseline.

199

Table 6.3. Enriched gene sets in maternal folic acid supplementation group relative to the control group among girls 6-8 y of age in Sarlahi, Nepal (false discovery rate cut off of 5%)

Gene set (Gene Ontology category, number) 1 Size2 ES3 NES4 P5 q6 INTRACELLULAR ORGANELLE PART (CC, GO:0044446) 30 0.60 1.89 <0.001 0.0377 CYTOSKELETON (CC, GO:0005856) 27 0.67 1.93 0.0017 0.0432 INTRACELLULAR NON MEMBRANE BOUND ORGANELLE (GO:0043232) 29 0.66 1.95 0.0016 0.0468 ORGAN_DEVELOPMENT (BP, GO:0048513) 42 0.55 1.87 <0.001 0.0470 Abbreviations: GO, Gene Ontology; CC, cellular component, BP, biological process 1Gene set name in the Gene Ontology database. In total, 312 gene sets were analyzed. 2Number of genes in the gene set 3Enrichment score of gene set. It indicates the degree to which given gene set is overrepresented at the top or bottom of the ranked list of proteins 4Normalized enrichment score. It indicates the enrichment score of given gene set after normalization across analyzed gene sets to take into account different gene set size 5Statistical significance of the enrichment score. 6Multiple hypothesis testing corrected P value for each NES. It indicates estimated probability that the NES score represents a false positive finding.

200

APPENDIX

Appendix 6.1. Comparison of characteristics of children, mothers, and household between all children followed-up and sub- samples

Nutritional- Bio-archive5 Followed- All archive4 (not Proteomics- up6 (not Characteristics N followed- n n (not n included in n archive3 included in up2 included in nutritional- bio-archive) proteomics) archive) Age, year 3172 7.5 (0.4) 500 7.5 (0.4) 500 7.5 (0.4) 1130 7.5 (0.4) 1042 7.5 (0.5) Male, % 3524 50.4 500 49.8 500 52.4 1130 52.7 1394 48.1 Pahadi ethnicity, % 3511 28.6 500 31.8 500 34.2 1130 31.2 1381 23.3 Ever sent to school, % 3494 66.1 500 66.8 500 72.6 1130 67.8 1364 62.1 Literacy, % 3495 16.5 500 17.4 500 16.4 1130 17.8 1365 15.2 WAZ7 3351 -2.08 (0.89) 499 -1.98 (0.85) 500 -2.01 (0.85) 1130 -2.07 (0.91) 1222 -2.16 (0.89) HAZ7 3353 -1.89 (0.9) 499 -1.77 (0.95) 500 -1.86 (0.85) 1130 -1.86 (0.89) 1222 -1.98 (0.88) BMIZ7 3351 -1.23 (0.87) 499 -1.20 (0.89) 500 -1.15 (0.85) 1130 -1.24 (0.86) 1222 -1.25 (0.87) MUAC 3354 15.4 (1.12) 500 15.5 (1.2) 500 15.5 (1.1) 1130 15.4 (1.1) 1224 15.3 (1.1) Diet intake in the past week, % Dairy >=2 3359 67.2 500 66.4 500 69.4 1130 65.8 1229 67.9 Meat >=2 3524 18.1 500 18.4 500 20.2 1130 19.0 1231 16.2 Fish >=2 3360 17.4 500 18.2 500 18.6 1130 17.1 1230 16.8 Eggs >=2 3358 5.9 500 5.6 500 6.6 1130 5.7 1228 6.0 DGLV>=2 3359 60.1 500 60.0 500 62.8 1130 57.8 1229 61.1 Morbidity, any symptom reported in previous week, % Fever 3359 9.4 500 8.2 500 8.4 1130 10.0 1229 9.9 Diarrhea 3357 2.2 500 3.2 500 2..2 1130 2.0 1227 2.0 Coughing 3353 4.0 500 3.8 500 4.2 1130 3.8 1223 4.3 Breathing 3357 3.1 500 2.8 500 3.6 1130 3.4 1227 2.8 Caste 3517 500 500 1130 1387 Brahmin and Chhetri 13.6 14.0 15.2 15.4 11.4

201

Vaiysha 65.1 65.2 67.6 63.3 65.6 Shudra, Muslim, and others 21.3 20.8 17.2 21.3 23.2 Cattle 3517 69.7 500 69.6 500 71.8 1130 70.7 1387 68.1 Electricity 3515 50.0 500 51.0 500 49.4 1130 50.0 1385 50.0 Maternal characteristics8 Age, year 3510 23.1 (5.6) 500 23.4 (5.7) 500 23.3 (5.8) 1130 23.4 (5.7) 1380 22.6 (5.5) Parity 3501 2.1 (2.1) 500 2.3 (2.2) 500 2.3 (2.2) 1130 2.3 (2.1) 1371 1.9 (1.9) Weight 3328 43.5 (5.5) 500 43.5 (5.6) 500 43.9 (5.5) 1130 43.5 (5.4) 1198 43.4 (5.6) Height 3325 150.3 (5.5) 500 150.5 (5.3) 500 150.4 (5.4) 1130 150.5 (5.5) 1195 150.0 (5.6) MUAC 3318 21.9 (1.8) 500 21.8 (1.9) 500 21.9 (1.8) 1130 21.8 (1.8) 1188 22.0 (1.9) Literacy 3500 20.4 500 22.2 500 21.4 1130 18.7 1370 20.8 Education 3496 500 500 1130 1366 No schooling 81.6 80.6 80.6 83.1 81.0 From class 1-9 13.0 14.4 12.6 12.3 13.2 High education9 5.5 5.0 6.8 4.6 5.9 Abbreviations: BMI, body mass index; MUAC, middle-upper arm circumference; HAZ, height-for-age z-score; WAZ, weight- for-age z-score; BMIZ, BMI-for-age z-score; CRP, c-reactive protein; AGP, alpha-1-acid glycoprotein 1Values are means (SD) or percentages 2Children followed-up at 6-8 years of age (maximum N=3,524) 3Children in the proteomics archive (n=500) 4Children in the nutritional archive but not in the proteomics archive (n=500) 5Children in the bio-archive but not in the nutritional archive (n=1,130) 6Children followed-up but not in the bio-archive (maximum n=1,394) 7Calculated by using the World Health Organization growth reference for school-aged children. 8Maternal characteristics at the enrollment in the antenatal micronutrient intervention trial 9High education indicates SLC test taken, intermediate or certificate completed, Bachelor’s or higher degree

202

Appendix 6.2. Distribution of maternal micronutrient supplementation of children for proteomics analysis (n=500) across 72 iTRAQ experiments1

Abbreviations: iTRAQ, isobaric tag for relative and absolute quantitation; Cont, Control; FA, folic acid; IFA, Iron-folic acid; IFAZn, Iron-folic acid- zinc; MM, multiple micronutrient 1x-axis indicates in total 72 iTRAQ experiments run for 500 plasma specimens of children. Randomly selected seven biological plasma samples were analyzed per each iTRAQ experiment (three biological samples were analyzed in the last experiment). Different colors represent different maternal micronutrient supplementation of children within each experiment.

203

Appendix 6.3. Plasma proteins with iTRAQ channel effects (n=17)1

204

1 x-axis represents 8 channels of iTRAQ experiment (8-plex iTRAQ) and y-axis represents log2 based relative abundance of protein. Among 982 proteins, mean differences in relative abundance of proteins by iTRAQ channels were examined by analysis of variance test. Proteins with substantial mean differences or implausible patterns by channels were excluded from analysis.

205

Appendix 6.4. Proteins of leading-edge subsets of enriched gene sets by maternal folic acid supplementation relative to the control group among girls 6-8 years of age in Sarlahi, Nepal1

NAME INTRACELLULARCYTOSKELETONINTRACELLULAR NONORGAN MEMBRANE DEVELOPMENT ORGANELLE BOUND PARTORGANELLE ACTB AKAP9 MYH10 TPM4 ANXA1 ARHGDIB BASP1 CDH1 FLNA MARCKS MCF2 VASP MSN SMC4 MED23 ACTA1 KRT1 KRT2 MYH9 MYL6 DSP ERAP1 AHSG APOA5 BTD COMP ENG IGF1 IGF2 IGFBP3 KRT10 KRT14 KRT16 KRT5 KRT9 PF4 PLG SGCD SOD1 SPARC SRGN 1Leading-edge proteins (row, denoted by gene symbols) of enriched gene sets (column) are colored in red. This figure visualizes the overlap of leading-edge proteins between different enriched gene sets (4 gene sets were differentially enriched by maternal folic acid supplementation among girls, passing a false discoveray rate threshold of 5%). Leading-edge proteins are ones that contribute to enrichment score.

206

Appendix 6.5. Enrichment plots of cytoskeleton (cellular component, GO:0005856) and organ development (biological process, GO:0048513) by maternal supplementation group relative to the control group, stratified by sex among children 6-8 years of age in Sarlahi, Nepal1,2,3 A. Cytoskeleton - Girls B. Cytoskeleton - Boys

207

C. Organ development - Girls D. Organ development - Boys

1The top portion of the plot shows the running enrichment score for given gene set as the analysis walks down the ranked list of proteins by t-statistics. The score at the peak (furthest from horizontal line) is the enrichment score. The bottom portion of the plot shows where the members of the gene set appear in the ranked list of proteins. In total, 573 (for boys) and 572 (for girls) plasma proteins were analyzed for GSEA using 312 gene sets from the Gene Ontology database. 2Cytoskeleton (GO:0005856) is defined as any of the various filamentous elements that form the internal framework of cells. The term embraces intermediate filaments, microfilaments, microtubules, the microtrabecular lattice, and other structures characterized by a polymeric filamentous nature and long-range order within the cell. The various elements of the cytoskeleton not only serve in the maintenance of cellular shape but also have roles in other cellular functions, including cellular movement, cell division, endocytosis, and movement of organelles (80). 3Organ development (GO:0048513) is defined as development of a tissue or tissues that work together to perform a specific function or functions. Development pertains to the process whose specific outcome is the progression of a structure over time, from its formation to the mature structure. Organs are commonly observed as visibly distinct structures, but may also exist as loosely associated clusters of cells that work together to perform a specific function or functions (80).

208

REFERENCES 1. Ramakrishnan U, Grant FK, Imdad A, et al. Effect of multiple micronutrient versus iron-folate supplementation during pregnancy on intrauterine growth. Nestle Nutrition Institute workshop series 2013;74:53-62. 2. WHO. Guideline: Daily Iron and Folic Acid Supplementation in Pregnant Women. Edtion ed. Geneva, 2012. 3. Li Q, Yan H, Zeng L, et al. Effects of maternal multimicronutrient supplementation on the mental development of infants in rural western China: follow-up evaluation of a double-blind, randomized, controlled trial. Pediatrics 2009;123(4):e685-92. 4. McGrath N, Bellinger D, Robins J, et al. Effect of maternal multivitamin supplementation on the mental and psychomotor development of children who are born to HIV-1-infected mothers in Tanzania. Pediatrics 2006;117(2):e216-25. 5. Tofail F, Persson LA, El Arifeen S, et al. Effects of prenatal food and micronutrient supplementation on infant development: a randomized trial from the Maternal and Infant Nutrition Interventions, Matlab (MINIMat) study. The American journal of clinical nutrition 2008;87(3):704-11. 6. Huy ND, Le Hop T, Shrimpton R, et al. An effectiveness trial of multiple micronutrient supplementation during pregnancy in Vietnam: impact on birthweight and on stunting in children at around 2 years of age. Food and nutrition bulletin 2009;30(4 Suppl):S506-16. 7. Christian P, Stewart CP, LeClerq SC, et al. Antenatal and postnatal iron supplementation and childhood mortality in rural Nepal: a prospective follow-up in a randomized, controlled community trial. American journal of epidemiology 2009;170(9):1127-36. 8. Shankar AH, Jahari AB, Sebayang SK, et al. Effect of maternal multiple micronutrient supplementation on fetal loss and infant death in Indonesia: a double-blind cluster-randomised trial. Lancet 2008;371(9608):215-27. 9. Christian P, Murray-Kolb LE, Khatry SK, et al. Prenatal micronutrient supplementation and intellectual and motor function in early school-aged children in Nepal. Jama 2010;304(24):2716-23. 10. Prado EL, Alcock KJ, Muadz H, et al. Maternal multiple micronutrient supplements and child cognition: a randomized trial in Indonesia. Pediatrics 2012;130(3):e536-46. 11. Stewart CP, Christian P, Schulze KJ, et al. Antenatal micronutrient supplementation reduces metabolic syndrome in 6- to 8-year-old children in rural Nepal. The Journal of nutrition 2009;139(8):1575-81. 12. Iannotti LL, Zavaleta N, Leon Z, et al. Maternal zinc supplementation reduces diarrheal morbidity in peruvian infants. The Journal of pediatrics 2010;156(6):960-4, 4 e1-2. 13. Andersen GS, Friis H, Michaelsen KF, et al. Effects of maternal micronutrient supplementation on fetal loss and under-2-years child mortality: long-term follow-up of a randomised controlled trial from Guinea-Bissau. African journal of reproductive health 2010;14(2):17-26.

209

14. Khan AI, Kabir I, Ekstrom EC, et al. Effects of prenatal food and micronutrient supplementation on child growth from birth to 54 months of age: a randomized trial in Bangladesh. Nutrition journal 2011;10:134. 15. Khan AI, Kabir I, Hawkesworth S, et al. Early invitation to food and/or multiple micronutrient supplementation in pregnancy does not affect body composition in offspring at 54 months: follow-up of the MINIMat randomised trial, Bangladesh. Maternal & child nutrition 2012. 16. Wang W, Yan H, Zeng L, et al. No effect of maternal micronutrient supplementation on early childhood growth in rural western China: 30 month follow-up evaluation of a double blind, cluster randomized controlled trial. European journal of clinical nutrition 2012;66(2):261-8. 17. Devakumar D, Chaube SS, Wells JC, et al. Effect of antenatal multiple micronutrient supplementation on anthropometry and blood pressure in mid- childhood in Nepal: follow-up of a double-blind randomised controlled trial. The Lancet Global health 2014;2(11):e654-63. 18. Vaidya A, Saville N, Shrestha BP, et al. Effects of antenatal multiple micronutrient supplementation on children's weight and size at 2 years of age in Nepal: follow-up of a double-blind randomised controlled trial. Lancet 2008;371(9611):492-9. 19. Allen LH. Micronutrient research, programs, and policy: From meta-analyses to metabolomics. Advances in nutrition 2014;5(3):344S-51S. 20. Owens S, Fall CH. Consequences of poor maternal micronutrition before and during early pregnancy. Transactions of the Royal Society of Tropical Medicine and Hygiene 2008;102(2):103-4. 21. Yajnik CS, Deshpande SS, Jackson AA, et al. Vitamin B12 and folate concentrations during pregnancy and insulin resistance in the offspring: the Pune Maternal Nutrition Study. Diabetologia 2008;51(1):29-38. 22. Schmidt RJ, Tancredi DJ, Ozonoff S, et al. Maternal periconceptional folic acid intake and risk of autism spectrum disorders and developmental delay in the CHARGE (CHildhood Autism Risks from Genetics and Environment) case- control study. The American journal of clinical nutrition 2012;96(1):80-9. 23. Steenweg-de Graaff J, Roza SJ, Steegers EA, et al. Maternal folate status in early pregnancy and child emotional and behavioral problems: the Generation R Study. The American journal of clinical nutrition 2012;95(6):1413-21. 24. Whitrow MJ, Moore VM, Rumbold AR, et al. Effect of supplemental folic acid in pregnancy on childhood asthma: a prospective birth cohort study. American journal of epidemiology 2009;170(12):1486-93. 25. Haberg SE, London SJ, Stigum H, et al. Folic acid supplements in pregnancy and early childhood respiratory health. Archives of disease in childhood 2009;94(3):180-4. 26. Stewart CP, Christian P, Schulze KJ, et al. Low maternal vitamin B-12 status is associated with offspring insulin resistance regardless of antenatal micronutrient supplementation in rural Nepal. The Journal of nutrition 2011;141(10):1912-7. 27. Burdge GC, Lillycrop KA. Folic acid supplementation in pregnancy: Are there devils in the detail? The British journal of nutrition 2012;108(11):1924-30.

210

28. O'Neill RJ, Vrana PB, Rosenfeld CS. Maternal methyl supplemented diets and effects on offspring health. Frontiers in genetics 2014;5:289. 29. Li Y, Zhang X, Sun Y, et al. Folate deficiency during early-mid pregnancy affects the skeletal muscle transcriptome of piglets from a reciprocal cross. PloS one 2013;8(12):e82616. 30. Lillycrop KA, Rodford J, Garratt ES, et al. Maternal protein restriction with or without folic acid supplementation during pregnancy alters the hepatic transcriptome in adult male rats. The British journal of nutrition 2010;103(12):1711-9. 31. Maloney CA, Hay SM, Reid MD, et al. A methyl-deficient diet fed to rats during the pre- and peri-conception periods of development modifies the hepatic proteome in the adult offspring. Genes & nutrition 2013;8(2):181-90. 32. Barua S, Chadman KK, Kuizon S, et al. Increasing maternal or post-weaning folic acid alters gene expression and moderately changes behavior in the offspring. PloS one 2014;9(7):e101674. 33. Swali A, McMullen S, Hayes H, et al. Processes underlying the nutritional programming of embryonic development by iron deficiency in the rat. PloS one 2012;7(10):e48133. 34. Christian P, Stewart CP. Maternal micronutrient deficiency, fetal development, and the risk of chronic disease. The Journal of nutrition 2010;140(3):437-45. 35. Max D, Brandsch C, Schumann S, et al. Maternal vitamin D deficiency causes smaller muscle fibers and altered transcript levels of genes involved in protein degradation, myogenesis, and cytoskeleton organization in the newborn rat. Molecular nutrition & food research 2014;58(2):343-52. 36. Roy S, Lado BH, Khanna S, et al. Vitamin E sensitive genes in the developing rat fetal brain: a high-density oligonucleotide microarray analysis. FEBS letters 2002;530(1-3):17-23. 37. Tran PV, Dakoji S, Reise KH, et al. Fetal iron deficiency alters the proteome of adult rat hippocampal synaptosomes. American journal of physiology Regulatory, integrative and comparative physiology 2013;305(11):R1297-306. 38. Cole RN, Ruczinski I, Schulze K, et al. The plasma proteome identifies expected and novel proteins correlated with micronutrient status in undernourished Nepalese children. The Journal of nutrition 2013;143(10):1540-8. 39. Anderson NL, Anderson NG. The human plasma proteome: history, character, and diagnostic prospects. Molecular & cellular proteomics : MCP 2002;1(11):845-67. 40. Jiang T, Christian P, Khatry SK, et al. Micronutrient deficiencies in early pregnancy are common, concurrent, and vary by season among rural Nepali pregnant women. The Journal of nutrition 2005;135(5):1106-12. 41. Christian P, Khatry SK, Katz J, et al. Effects of alternative maternal micronutrient supplements on low birth weight in rural Nepal: double blind randomised community trial. BMJ 2003;326(7389):571. 42. Stewart CP, Christian P, LeClerq SC, et al. Antenatal supplementation with folic acid + iron + zinc improves linear growth and reduces peripheral adiposity in school-age children in rural Nepal. The American journal of clinical nutrition 2009;90(1):132-40.

211

43. Christian P, West KP, Khatry SK, et al. Effects of maternal micronutrient supplementation on fetal loss and infant mortality: a cluster-randomized trial in Nepal. The American journal of clinical nutrition 2003;78(6):1194-202. 44. West KP, Jr., Katz J, Khatry SK, et al. Double blind, cluster randomised trial of low dose supplementation with vitamin A or beta carotene on mortality related to pregnancy in Nepal. The NNIPS-2 Study Group. BMJ 1999;318(7183):570-5. 45. Schulze KJ, Christian P, Wu LS, et al. Micronutrient deficiencies are common in 6- to 8-year-old children of rural Nepal, with prevalence estimates modestly affected by inflammation. The Journal of nutrition 2014;144(6):979-87. 46. Herbrich SM, Cole RN, West KP, Jr., et al. Statistical inference from multiple iTRAQ experiments without using common reference standards. Journal of proteome research 2013;12(2):594-604. 47. Storey JD. The positive false discovery rate: a bayesian interpretation and the q- value. The Annals of Statistics 2003;31(6):2013–35. 48. Tielsch JM, Khatry SK, Stoltzfus RJ, et al. Effect of routine prophylactic supplementation with iron and folic acid on preschool child mortality in southern Nepal: community-based, cluster-randomised, placebo-controlled trial. Lancet 2006;367(9505):144-52. 49. Gray KA, Yates B, Seal RL, et al. Genenames.org: the HGNC resources in 2015. Nucleic acids research 2014. 50. Ashburner M, Ball CA, Blake JA, et al. Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nature genetics 2000;25(1):25-9. 51. Subramanian A, Tamayo P, Mootha VK, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proceedings of the National Academy of Sciences of the United States of America 2005;102(43):15545-50. 52. Khulan B, Cooper WN, Skinner BM, et al. Periconceptional maternal micronutrient supplementation is associated with widespread gender related changes in the epigenome: a study of a unique resource in the Gambia. Human molecular genetics 2012;21(9):2086-101. 53. Roberfroid D, Huybregts L, Lanou H, et al. Effect of maternal multiple micronutrient supplements on cord blood hormones: a randomized controlled trial. The American journal of clinical nutrition 2010;91(6):1649-58. 54. The Comprehensive R Archive Network. Internet: http://cran.r-project.org/. 55. Hoyo C, Murtha AP, Schildkraut JM, et al. Methylation variation at IGF2 differentially methylated regions and maternal folic acid use before and during pregnancy. Epigenetics : official journal of the DNA Methylation Society 2011;6(7):928-36. 56. Maloney CA, Hay SM, Young LE, et al. A methyl-deficient diet fed to rat dams during the peri-conception period programs glucose homeostasis in adult male but not female offspring. The Journal of nutrition 2011;141(1):95-100. 57. McKay JA, Wong YK, Relton CL, et al. Maternal folate supply and sex influence gene-specific DNA methylation in the fetal gut. Molecular nutrition & food research 2011;55(11):1717-23. 58. Noreau A, Bourassa CV, Szuto A, et al. SYNE1 mutations in autosomal recessive cerebellar ataxia. JAMA neurology 2013;70(10):1296-31.

212

59. Zhang Q, Skepper JN, Yang F, et al. Nesprins: a novel family of spectrin-repeat- containing proteins that localize to the nuclear membrane in multiple tissues. Journal of cell science 2001;114(Pt 24):4485-98. 60. Gros-Louis F, Dupre N, Dion P, et al. Mutations in SYNE1 lead to a newly discovered form of autosomal recessive cerebellar ataxia. Nature genetics 2007;39(1):80-5. 61. Zhang J, Felder A, Liu Y, et al. Nesprin 1 is critical for nuclear positioning and anchorage. Human molecular genetics 2010;19(2):329-41. 62. Smith ER, Zhang XY, Capo-Chichi CD, et al. Increased expression of Syne1/nesprin-1 facilitates nuclear envelope structure changes in embryonic stem cell differentiation. Developmental dynamics : an official publication of the American Association of Anatomists 2011;240(10):2245-55. 63. Mukhopadhyay P, Horn KH, Greene RM, et al. Prenatal exposure to environmental tobacco smoke alters gene expression in the developing murine hippocampus. Reproductive toxicology 2010;29(2):164-75. 64. Taruscio D, Carbone P, Granata O, et al. Folic acid and primary prevention of birth defects. BioFactors 2011;37(4):280-4. 65. Caldwell PT, Manziello A, Howard J, et al. Gene expression profiling in the fetal cardiac tissue after folate and low-dose trichloroethylene exposure. Birth defects research Part A, Clinical and molecular teratology 2010;88(2):111-27. 66. McNeil CJ, Hay SM, Rucklidge GJ, et al. Maternal diets deficient in folic acid and related methyl donors modify mechanisms associated with lipid metabolism in the fetal liver of the rat. The British journal of nutrition 2009;102(10):1445-52. 67. Haggarty P, Hoad G, Campbell DM, et al. Folate in pregnancy and imprinted gene and repeat element methylation in the offspring. The American journal of clinical nutrition 2013;97(1):94-9. 68. Steegers-Theunissen RP, Obermann-Borst SA, Kremer D, et al. Periconceptional maternal folic acid use of 400 microg per day is related to increased methylation of the IGF2 gene in the very young child. PloS one 2009;4(11):e7845. 69. Santoso MI, Rohman MS. Decreased TGF-beta1 and IGF-1 protein expression in rat embryo skull bone in folic acid-restricted diet. The Journal of nutritional biochemistry 2006;17(1):51-6. 70. Bailey LB, Gregory JF, 3rd. Folate metabolism and requirements. The Journal of nutrition 1999;129(4):779-82. 71. Fekete K, Berti C, Cetin I, et al. Perinatal folate supply: relevance in health outcome parameters. Maternal & child nutrition 2010;6 Suppl 2:23-38. 72. Swali A, McMullen S, Hayes H, et al. Cell cycle regulation and cytoskeletal remodelling are critical processes in the nutritional programming of embryonic development. PloS one 2011;6(8):e23189. 73. Christian P, Jiang T, Khatry SK, et al. Antenatal supplementation with micronutrients and biochemical indicators of status and subclinical infection in rural Nepal. The American journal of clinical nutrition 2006;83(4):788-94. 74. Schenk S, Schoenhals GJ, de Souza G, et al. A high confidence, manually validated human blood plasma protein reference set. BMC medical genomics 2008;1:41.

213

75. Farrah T, Deutsch EW, Omenn GS, et al. A high-confidence human plasma proteome reference set with estimated concentrations in PeptideAtlas. Molecular & cellular proteomics : MCP 2011;10(9):M110 006353. 76. Holdenrieder S, Stieber P, Liska V, et al. Cytokeratin serum biomarkers in patients with colorectal cancer. Anticancer research 2012;32(5):1971-6. 77. Ostergaard M, Rasmussen HH, Nielsen HV, et al. Proteome profiling of bladder squamous cell carcinomas: identification of markers that define their degree of differentiation. Cancer research 1997;57(18):4111-7. 78. Barak V, Goike H, Panaretakis KW, et al. Clinical utility of cytokeratins as tumor markers. Clinical biochemistry 2004;37(7):529-40. 79. Gonzalez-Quintela A, Tome S, Fernandez-Merino C, et al. Synergistic effect of alcohol consumption and body mass on serum concentrations of cytokeratin-18. Alcoholism, clinical and experimental research 2011;35(12):2202-8. 80. Carbon S, Ireland A, Mungall CJ, et al. AmiGO: online access to ontology and annotation data. Bioinformatics 2009;25(2):288-9.

214

CHAPTER 7: CONCLUSIONS

Although we have made successful progress in reducing child mortality, undernutrition, and other health conditions (1, 2), our understanding of the underlying biological mechanisms of nutrition and health remains largely unexplored (3). In this thesis, we had unique opportunities to explore the childhood plasma proteome and sought potential public health applications of using high-throughput technology to deepen our understanding and to identify potential plasma biomarkers that can be used in nutritional and health research in undernourished populations.

We evaluated retrospective, cross-sectional, and prospective associations between childhood plasma proteins and different, but potentially interdependent nutritional exposure, a health indicator and outcome, respectively. In chapter 4, we examined plasma proteins that co-vary with plasma alpha-1-acid glycoprotein (AGP), a conventional biomarker of inflammation. In chapter 5, we investigated associations between plasma proteins and child intellectual function assessed about a year after the collection of plasma specimens. In chapter 6, we evaluated the effect of prenatal micronutrient supplementation on plasma protein profiles in childhood. The results of the chapters showed the gradients in the strength of the associations, reflecting direct or indirect and temporary or long-term biological relationships between proteins and the exposure and outcomes.

7.1. Summary of findings & public health interpretation/implication Chapter 4

Summary of findings: In chapter 4, we identified 99 plasma proteins that strongly co- vary with AGP at a family-wise error rate of 0.1% and 206 plasma proteins passing a

215 false discovery rate threshold of 1%. These are ~10% and ~20% of the total number of proteins we analyzed, demonstrating substantial changes in abundance of a large number of proteins in plasma in response to inflammation. We refer to this consortium of proteins as a “population plasma inflammasome” to embrace the global manner of the biological response to inflammation. In addition to widely known positive acute phase proteins, we identified numerous intracellular proteins in the positive plasma inflammasome that were strongly correlated with the known acute phase reactants. This study also expanded the scope of negative acute phase proteins, which haven’t been studied as much as positive acute phase proteins in a holistic manner. Our findings reveal that hepatic vs. extra- hepatic origins of the negative plasma inflammasome showed distinct abundance profiles in circulation, suggesting a different gradient in the abundance of proteins in the intravascular compartment depending on tissue origin. Overall, the findings of this study demonstrate homeostatic control of inflammation with systematic changes in abundance of plasma proteins from a wide spectrum of body systems in a healthy pediatric population.

Public health interpretation/applications: The results of this study suggest that chronic inflammation may affect various aspects of child health. It may compromise bone health by down-regulation of proteins involved in linear growth (IGFALS) (4, 5), bone density

[tetranectin (CLEC3B) and osteomodulin (OMD)] (6, 7), and bone mineralization [fetuin-

A (AHSG)] (8-10). In addition, it may induce lean body mass loss by suppressing the expression of extracellular matrix (ECM) proteins in connective or soft tissues such as skeletal muscle or cartilage [lumican (LUM), cartilage oligomeric matrix protein

(COMP), and collagen (COL6A1/3)] (11-13) and numerous proteins regulating ECM

216 metabolism (14). In addition, it may be associated with fat deposition by lowering concentration of circulating SHBG (15). If these results are validated, it implies that chronic inflammation should be carefully considered in the interpretation of anthropometric measurements of populations with a high risk of subclinical inflammation.

We propose potential biomarkers for low-grade inflammation, including LRG1, complement component 9, and SERPINA3, as opposed to acutely reacting proteins, including SAA1/2, HP/HPR, and S100A8/9. These proteins should be validated by further studies for their future uses of characterizing inflammation status at the population level, in addition to AGP and CRP.

Chapter 5

Summary of findings: The main outcome of this study was that IGF-binding proteins

(IGFALS/BP3), transthyretin, and apolipoproteins (APOA1/2/C1/3/M/D) were positively and proteins involved in inflammation (complement components 2/5/9, ORM1,

SERPINA3, LRG1, and others) and PKM were negatively associated with general intelligence scores of children. After adjusting for known risk factors or potential confounding factors, only 7 proteins remained significantly associated with the outcome, additionally explaining 5~9% of variance in the intelligence scores. These proteins are the ones that we proposed as potential biomarkers of subclinical inflammation based on the results of chapter 4. Interestingly, CRP and other acutely reacting proteins were not associated with the outcome. Also, our result showed that the contribution of chronic inflammation to developmental loss might be similar to that of long-term undernutrition.

The findings of this chapter should be carefully interpreted, because the proteins we

217 identified do not directly interact with the CNS and we cannot rule out the impact of unmeasured or residual confounding.

Public health interpretation/applications: In addition to widely recognized biological risk factors including intrauterine growth restriction, childhood undernutrition, micronutrient deficiencies (iron and iodine), and infectious diseases, such as diarrhea,

HIV, and malaria (16, 17), non-specific chronic inflammation should be considered as one of the important risk factors of child development in undernourished populations. In this non-malaria and non-HIV area, morbidity rate was low, but one third of children showed elevated AGP (1>g/L), suggesting asymptomatic subclinical inflammation is common in children. If we find more evidence in the role and burden of chronic inflammation in child development, it can be integrated into child development interventions in high-risk chronic inflammation areas. As we suggested in chapter 4, AGP, complement component 9, LRG1, and SERPINA3 can be tested as potential markers to evaluate the efficacy of interventions. In addition, these proteins can be used to identify major determinants of chronic inflammation in children, such as enteric infection, the gut microbiome or environmental toxins, all of which are known to contribute to poor child development.

Chapter 6

Summary of findings: There were no overall effects of antenatal micronutrient supplementation on child plasma proteins. Among 587 plasma proteins, maternal folic acid supplementation increased the relative abundance of only one protein, nesprin-1, by

50 (24.7, 82.8)% among boys. Nesprin-1 is an outer nuclear membrane protein anchoring the actin cytoskeleton, and its mutation is associated with muscular dystrophy and

218 autosomal recessive cerebella ataxia (18-20). Out of 312 gene sets from the Gene

Ontology database, maternal folic acid supplementation positively enriched 4 gene sets related to cytoskeleton and organ development only among girls. Other maternal micronutrient supplementation did not have effects on the abundance of plasma proteins in children.

Public health implications: Because folic acid supplementation is widely recommended as a standard of prenatal care for pregnant women, further research is required to confirm the effect of antenatal folic acid supplementation on long-lasting gene-specific or functional changes in tissues or organs and to examine their potential health consequences.

Study strengths and limitations

Study strengths

This is the first study that explored the childhood plasma proteome in non-clinical settings. Studies have applied plasma proteomics to discover biomarkers of developmental disorders or pediatric diseases (21-23). To the best of our knowledge, no studies have investigated the plasma proteomes of free-living healthy children at the population level.

We applied multiplex quantitative mass spectrometry (MS)-based proteomics that produced high-throughput protein abundance data on a relatively large number of population-based samples (n=500). With immuno-depletion of the 6 most abundant proteins, which comprise 85% of total proteins in plasma, we identified almost 4,700 non-redundant proteins with high mass accuracy, of which almost 1000 plasma proteins

219 could be included in analysis. This untargeted approach allowed us to confirm expected biological validity of data, and to identify potential new biomarkers or to generate new hypotheses.

This study benefited from a community-based antenatal supplementation trial and a series of child follow-up studies. Causal inference was possible based on the double-blind randomized-controlled study design. Numerous plasma biomarkers of micronutrients enriched the secondary analyses. The temporal lag between the first and the second child follow-up studies minimized putative associations that can randomly occur due to transient physiological conditions of children at the time of blood collection. Study samples were fairly homogeneous, and a random sample selection procedure from the larger child follow-up cohort ensured the generalizability of findings to similar settings in

South Asia and elsewhere where maternal and child nutritional health are public health problems.

This study applied a newly developed statistical approach that estimates the relative abundance of proteins using a summary of biological samples within an iTRAQ experiment rather than using a masterpool, generating more precise estimates (24). We performed gene set enrichment analysis to increase power to detect sums of small effect sizes of individual proteins grouped based on their ontology (25). We annotated biological, molecular, and cellular information of proteins by in-depth literature review and using multiple annotation databases to incorporate existing knowledge.

220

Study limitations

This study has significant limitations to consider. The low sensitivity of mass spectrometry toward low-abundant proteins produced a large number of missing values in low-abundant proteins. This restricted the number of proteins and gene sets limiting full exploration of plasma proteome. This might particularly affect the results of chapter 6, in which we expected to examine the effect of maternal micronutrient supplementation on low-abundant tissue proteins. Because protein abundance was on a relative scale within the experiment, estimated parameters do not directly reflect physiologically relevant values. Lastly, approximately less than half of the children in this study were exposed to zinc, iron-folic acid, or iron-folic acid-zinc during childhood (26). Although we took into account the unbalanced distribution of childhood supplementation across prenatal supplementation groups, this study could not consider potential interactions between prenatal and childhood micronutrient supplementation due to a small sample size.

7.3. Conclusion

As child health is profoundly affected by cumulative environmental exposures even before birth, this study demonstrates the complex nature of biological responses and processes associated with prenatal micronutrient supplementation, childhood inflammation, and cognitive function. As proteomics studies highly depend on the advancement of other disciplines, this study faced multidisciplinary challenges in sensitive and reliable identification and quantification of proteins by mass spectrometry, finding appropriate statistical approaches that consider protein-protein dependencies, and using valid and complete bioinformatics databases to systematically annotate the functions of a large number of proteins. We learned that the dynamics of the plasma

221 proteome is complex and extracting biologically meaningful information from it is challenging. In addition, we face a new challenge on translating the findings into public health applications, which requires thorough validation and consensus from the international nutrition research community. Notwithstanding these challenges, it is promising that this large-scale comprehensive analysis on biological samples profoundly deepens and expands our previous knowledge of child health and can contribute to the development of sound interventions that will reduce harm and optimize health benefits.

From this point of view, this exploratory study provides invaluable information for future research and programs to improve child health in less-privileged areas.

222

REFERENCES 1. Black RE, Victora CG, Walker SP, et al. Maternal and child undernutrition and overweight in low-income and middle-income countries. Lancet 2013;382(9890):427-51. 2. the United Nations Children’s Fund. Levels and trends in child mortality : estimates developed by the UN Inter-agency Group for child Mortality Estimation (UN IGME). 2014. 3. Prentice AM, Gershwin ME, Schaible UE, et al. New challenges in studying nutrition-disease interactions in the developing world. The Journal of clinical investigation 2008;118(4):1322-9. 4. Wit JM, Camacho-Hubner C. Endocrine regulation of longitudinal bone growth. Endocrine development 2011;21:30-41. 5. Wang J, Zhou J, Bondy CA. Igf1 promotes longitudinal bone growth by insulin- like actions augmenting chondrocyte hypertrophy. FASEB journal : official publication of the Federation of American Societies for Experimental Biology 1999;13(14):1985-90. 6. Wewer UM, Ibaraki K, Schjorring P, et al. A potential role for tetranectin in mineralization during osteogenesis. The Journal of cell biology 1994;127(6 Pt 1):1767-75. 7. Wendel M, Sommarin Y, Heinegard D. Bone matrix proteins: isolation and characterization of a novel cell-binding keratan sulfate proteoglycan (osteoadherin) from bovine bone. The Journal of cell biology 1998;141(3):839-47. 8. Colclasure GC, Lloyd WS, Lamkin M, et al. Human serum alpha 2HS- glycoprotein modulates in vitro bone resorption. The Journal of clinical endocrinology and metabolism 1988;66(1):187-92. 9. Jahnen-Dechent W, Heiss A, Schafer C, et al. Fetuin-A regulation of calcified matrix metabolism. Circulation research 2011;108(12):1494-509. 10. Lebreton JP, Joisel F, Raoult JP, et al. Serum concentration of human alpha 2 HS glycoprotein during the inflammatory process: evidence that alpha 2 HS glycoprotein is a negative acute-phase reactant. The Journal of clinical investigation 1979;64(4):1118-29. 11. Wilson R. The extracellular matrix: an underexplored but important proteome. Expert review of proteomics 2010;7(6):803-6. 12. Byron A, Humphries JD, Humphries MJ. Defining the extracellular matrix using proteomics. International journal of experimental pathology 2013;94(2):75-92. 13. Urciuolo A, Quarta M, Morbidoni V, et al. Collagen VI regulates satellite cell self-renewal and muscle regeneration. Nature communications 2013;4:1964. 14. Straub RH, Cutolo M, Buttgereit F, et al. Energy regulation and neuroendocrine- immune control in chronic inflammatory diseases. Journal of internal medicine 2010;267(6):543-60. 15. Pinkney J, Streeter A, Hosking J, et al. Adiposity, chronic inflammation, and the prepubertal decline of sex hormone binding globulin in children: evidence for associations with the timing of puberty (earlybird 58). The Journal of clinical endocrinology and metabolism 2014;99(9):3224-32. 16. Walker SP, Wachs TD, Gardner JM, et al. Child development: risk factors for adverse outcomes in developing countries. Lancet 2007;369(9556):145-57.

223

17. Walker SP, Wachs TD, Grantham-McGregor S, et al. Inequality in early childhood: risk and protective factors for early child development. Lancet 2011;378(9799):1325-38. 18. Zhang J, Felder A, Liu Y, et al. Nesprin 1 is critical for nuclear positioning and anchorage. Human molecular genetics 2010;19(2):329-41. 19. Gros-Louis F, Dupre N, Dion P, et al. Mutations in SYNE1 lead to a newly discovered form of autosomal recessive cerebellar ataxia. Nature genetics 2007;39(1):80-5. 20. Zhang Q, Bethmann C, Worth NF, et al. Nesprin-1 and -2 are involved in the pathogenesis of Emery Dreifuss muscular dystrophy and are critical for nuclear envelope integrity. Human molecular genetics 2007;16(23):2816-33. 21. Yu HR, Kuo HC, Sheen JM, et al. A unique plasma proteomic profiling with imbalanced fibrinogen cascade in patients with Kawasaki disease. Pediatric allergy and immunology : official publication of the European Society of Pediatric Allergy and Immunology 2009;20(7):699-707. 22. Li Y, Dang TA, Shen J, et al. Identification of a plasma proteomic signature to distinguish pediatric osteosarcoma from benign osteochondroma. Proteomics 2006;6(11):3426-35. 23. Corbett BA, Kantor AB, Schulman H, et al. A proteomic study of serum from children with autism showing differential expression of apolipoproteins and complement proteins. Molecular psychiatry 2007;12(3):292-306. 24. Herbrich SM, Cole RN, West KP, Jr., et al. Statistical inference from multiple iTRAQ experiments without using common reference standards. Journal of proteome research 2013;12(2):594-604. 25. Subramanian A, Tamayo P, Mootha VK, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proceedings of the National Academy of Sciences of the United States of America 2005;102(43):15545-50. 26. Tielsch JM, Khatry SK, Stoltzfus RJ, et al. Effect of routine prophylactic supplementation with iron and folic acid on preschool child mortality in southern Nepal: community-based, cluster-randomised, placebo-controlled trial. Lancet 2006;367(9505):144-52.

224

7 CURRICULUM VITAE

SUN-EUN LEE, MS 615 N. Wolfe Street, W2501, Baltimore, MD 21202 Telephone: 410-917-0922 E-mail: [email protected]

EDUCATION

2009-present Doctor of Philosophy Candidate in International Health Center for Human Nutrition, Department of International Health, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD Dissertation: The childhood plasma proteome: discovering its applications in public health Advisor: Dr. Keith P. West, Jr.

2008 Master of Science in Food and Nutrition Yonsei University Graduate School, Seoul, Korea Thesis: Anti-inflammatory mechanism of polyunsaturated fatty acids in Helicobacter pylori-infected gastric epithelial cells Advisor: Dr. Hye-young Kim

2006 Bachelor of Science in Food and Nutrition Yonsei University, Seoul, Korea

2004-2005 Study Abroad Program University of Northern Colorado, CO, USA

PROFESSIONAL EXPERIENCE

2011-2013 Research Assistant, Child Health Epidemiology Reference Group (CHERG), Johns Hopkins Bloomberg School of Public Health, Baltimore, MD Project: Prenatal Origins of Childhood Undernutrition Managed datasets, analyzed the contribution of preterm and fetal growth restriction to childhood undernutrition, prepared manuscript (Supervisor: Dr. Parul Christian)

2010-2012 Research Assistant, Center for Human Nutrition, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD Project: Dietary pattern of pregnant women in developing countries

225

Collected, integrated, and analyzed dietary data of pregnant women in developing countries, prepared manuscript (Supervisor: Dr. Laura Caulfield)

2009 Intern, Organo Transition Metal Catalysis-Hybrid Materials Laboratory, Department of Chemistry, College of Science, Yonsei University Institute of Life, Seoul, Korea Developed an enzyme-linked immunosorbent assay of Interleukin-8 using gold-nano with silver staining method (Supervisor: Chul-Ho Jun)

2006-2008 Researcher, Brain Korea 21 Project for Functional Foods and Nutrigenomics, Yonsei University, Seoul, Korea Conducted research on anti-inflammatory effects of poly unsaturated fatty acids in gastric epithelial cells (PI: Hyeyoung Kim)

2005-2006 Student Researcher, Department of Pharmacology, Yonsei UniversityCollege of Medicine, Seoul, Korea Conducted research on the effects of glutamine on Interleukin-8 expression in Mycoplasmapneumoniae infected lung epithelial cells (PI: Hyeyoung Kim)

HONORS AND SCHOLARSHIPS

Honors

2015 American Society of Nutrition’s Emerging Leaders in Nutrition Science Poster Competition, Finalist, American Society for Nutrition

2006 Honors Student, Yonsei University, Seoul, Korea

2003 High Honors Student, Yonsei University, Seoul, Korea

Scholarships

2015 Richard Hall Fund Johns Hopkins Bloomberg School of Public Health, Baltimore, MD

2014 George G. Graham Professorship Endowment Johns Hopkins Bloomberg School of Public Health, Baltimore, MD

2013 George G. Graham Professor ship Endowment Johns Hopkins Bloomberg School of Public Health, Baltimore, MD

2012 Harry J. Prebluda Fellowship in Nutritional Biochemistry

226

Johns Hopkins Bloomberg School of Public Health, Baltimore, MD

Harry D. Kruse Publication Award in Human Nutrition Johns Hopkins Bloomberg School of Public Health, Baltimore, MD

2010-2011 International fellowship American Association of University Women (AAUW), Washington, DC, USA

2009-2010 Bacon Field Chow Memorial Fellowship Center for Human Nutrition, Johns Hopkins School of Public Health, MD, USA

2006-2008 Full Scholarship Brain Korea 21 Project for Functional Foods and Nutrigenomics, Yonsei University, Seoul, Korea

2008 Internal Scholarship Yonsei University, Seoul, Korea

2006 External Scholarship Halim Scholarship Foundation, Seoul, Korea

2004 Fund Scholarship Yonsei University, Seoul, Korea

2003 University Designated Scholarship Yonsei University, Seoul, Korea

2002 University Designated Scholarship Yonsei University, Seoul, Korea

PUBLICATIONS

Lee SE, Lim JW, Kim JM, Kim H. Anti-inflammatory mechanism of polyunsaturated fatty acids in Helicobacter pylori-infected gastric epithelial cells. Mediators of Inflammation. 2014 June; 2014:128919.

P Christian, SE Lee, M Donahue Angel, LS Adair, SE Arifeen, P Ashorn, FC Barros, CHD Fall, WW Fawzi, W Hao, G Hu, JH Humphrey, L Huybregts, CV Joglekar, SK Kariuki, PK Kolsteren, GV Krishnaveni, E Liu, R Martorell, D Osrin, L Persson, U Ramakrishnan, L Richter, D Roberfroid, A Sania, FO Ter Kuile, J Tielsch, CG Victora, CS Yajnik, H Yan, L Zeng, RE Black. Risk of childhood undernutrition related to small- for-gestational age and preterm birth in low and middle income countries. International Journal of Epidemiology. 2013 Oct;42(5):1340-55.

227

SE Lee, SA Talegawkar, M Merialdi, LE Caulfield. Dietary intakes of women during pregnancy in low and middle-income countries. Public Health Nutrition. 2013 Aug;16(8):1340-53.

SE Lee, JW Lim, H Kim. AP-1 mediates DHA-induced apoptosis in human gastric cancer cells. Ann NY Acad Sci. 2009 Aug;1171:163-9.   PRESENTATIONS

Oral presentations

SE Lee, KP West, Jr, RN Cole, KJ Schulze, JD Yager, J Groopman, P Christian. “A Plasma Proteome Associated with Inflammation in Nepalese School-aged Children.” Experimental Biology 2015, Boston, U.S.A. March 2015. Abstract #4598.

SE Lee, JW Lim, H Kim. “Mechanisms of docosahexaenoic acid-induced apoptosis via stimulation of activation protein-1 pathway in gastric cancer AGS cells.” Symposium on Recent Advances in Gastrointestinal Pharmacology and Nutrigeomics, Seoul, Korea. Feb. 2008

SE Lee, JW Lim, H Kim. “Inhibitory effect of polyunsaturated fatty acids on Helicobacter pylori-induced Interleukin-8 expression in gastric epithelial cells.” The 1st Collaboration Meeting on GI Pharmacology Co-sponsored by Yonsei University and Kyoto Pharmaceutical University, Kyoto, Japan. Jan. 2007

Poster presentations

SE Lee, KP West, Jr, RN Cole, KJ Schulze, JD Yager, J Groopman, P Christian. A Plasma Proteome Associated with Inflammation in Nepalese School-aged Children. Experimental Biology 2015, Boston, U.S.A. March 2015. Abstract #4598. Selected as finalist for ASN's Emerging Leaders in Nutrition Science Poster Competition.

SE Lee, K West, Jr, RN Cole, I Ruczinski, K Schulze, JD Yager, J Groopman, P Christian. Effect of antenatal micronutrient supplementation on the plasma proteome in school-aged children in Nepal. Micronutrient Forum Global Conference. 2014. Addis Ababa, Ethiopia. June. 2-6. Abstract #0436

SE Lee, Keith P. West, Jr, Robert N. Cole, Kerry Schulze1, Ingo Ruczinski, Joshua Betz, James D. Yager, John Groopman, Sudeep Shrestha, Parul Christian. Plasma proteins and vitamin K (PIVKA-II) status in school aged children in Nepal. Experimental Biology 2014, San Diego, USA. Apr. 26-30. Abstract #3392

SE Lee, KP West, Jr., RN Cole, I Ruczinski, K Schulze, JD Yager, J Groopman, P Christian. Effects of Antenatal Micronutrient Supplementation on Plasma Protein Profiles

228 in Nepalese Children. Experimental Biology 2013, Boston, USA. Apr. 20-24. Abstract #9268

SE Lee, S Talegawkar, M Merialdi, LE Caulfield. What Are African Women Eating During Pregnancy? Experimental Biology 2011, Washington, DC, USA. Apr. 9-13. Abstract #6129

SE Lee, JW Lim, H Kim. Mechanisms of docosahexaenoic acid-induced apoptosis of gastric cancer AGS cells. Apoptosis World 2008, Luxembourg. Jan. 2008

SE Lee, JW Lim, H Kim. Effect of polyunsaturated fatty acids on Helicobacter pylori- induced Interleukin-8 expression in gastric epithelial AGS Cells. Experimental Biology 2007, Washington, DC, USA. Apr. 2007

SE Lee, JY Seo, H Kim. Effect of glutathione and polyunsaturated fatty acids on Helicobacter pylori- induced IL-8 expression in gastric epithelial AGS Cells. The 6th Western Pacific Helicobacter Congress, Bangkok, Thailand. Nov. 2006

TEACHING EXPERIENCE

Graduate

2011-2013 Teaching Assistant, Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD Leaded course office hours, held STATA office hours, and graded problem sets, quizzes and exams for Statistical Methods in Public Health I-III (140.621-623) with Dr. Marie Diener-West and Dr. John McGready

2011 Teaching Assistant, Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD Held STATA office hours, and helped students analyze their analysis term projects for Statistical Methods in Public Health IV (140.624) with Dr. James Tonascia

2011, 2013 Teaching Assistant, Center for Human Nutrition, Department of Department of International Health, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD Graded assignments, answered questions regarding course materials, organized computer lab for dietary analysis program for Assessment of Nutritional Status (222.642) with Dr. Kerry Schulze

Undergraduate

2008 Teaching Assistant, Department of Human Ecology, Yonsei

229

University, Seoul, Korea Lectured weekly and assisted grading problem sets and exams for Advanced Nutrition Metabolism with Dr. Yang Ja Lee

2007 Teaching Assistant, Department of Human Ecology, Yonsei University, Seoul, Korea Lectured weekly and assisted grading for Nutrient Metabolism and Experiment with Dr. Yang Ja Lee

2007 Teaching Assistant, Department of Human Ecology Yonsei University, Seoul, Korea Leaded weekly lab and assisted grading problem sets and exams for Biochemistry with Dr. Hyeyoung Kim

PROFESSIONAL ACTIVITIES

2010 Attendee, Dietary Supplement Research Practicum, National Institutes of Health (NIH), Washington, D.C., USA  2007 Dietitian License  Korea Dietetic Association, Seoul, Korea

2006 Health Delegate (certificated), Harvard Project for Asian and International Relations (HPAIR), Singapore

2006 Biotechnology Delegate (certificated), North East Asian Network (NEAN) 2006, Seoul, Korea

2006 Participant (certificated), Functional Food and Consultant Program, Yonsei University, Seoul, Korea

Society Memberships

Member of American Society of Nutrition (ASN) Member of student interest group (SIG) of American Society for Nutrition

ADDITIONAL INFORMATION

Language

Korean (Native) English (Competent)

230