UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Identifying epigenetic associations with cell type and gestational age in the neonatal immune system de Goede, Olivia Mae 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2016_may_degoede_olivia.pdf [ 6.23MB ]
Metadata
JSON: 24-1.0228788.json
JSON-LD: 24-1.0228788-ld.json
RDF/XML (Pretty): 24-1.0228788-rdf.xml
RDF/JSON: 24-1.0228788-rdf.json
Turtle: 24-1.0228788-turtle.txt
N-Triples: 24-1.0228788-rdf-ntriples.txt
Original Record: 24-1.0228788-source.json
Full Text
24-1.0228788-fulltext.txt
Citation
24-1.0228788.ris

Full Text

IDENTIFYING EPIGENETIC ASSOCIATIONS WITH CELL TYPE AND GESTATIONAL AGE IN THE NEONATAL IMMUNE SYSTEM  by Olivia Mae de Goede  B.Sc. (Hons.), University of Victoria, 2013  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  MASTER OF SCIENCE in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Medical Genetics)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  April 2016  © Olivia Mae de Goede, 2016 ii  Abstract  Neonates have a uniquely structured immune system characterized by immunotolerance, an unprimed adaptive immune system, and a heavy reliance on innate immune responses. Although this prevents excessive hyperinflammatory responses during gestation and postnatal microbial colonization of the neonate, it also confers vulnerability to infection. This risk is heightened in those born preterm (prior to 37 weeks gestation), as development of their immune system is interrupted by early birth. Throughout gestation, the predominant hematopoietic organ shifts in a defined temporal pattern. Each hematopoietic source produces different types of immune cells in different proportions, to accommodate the needs of the developing fetus. One of the greatest differences between these organs is the release of nucleated red blood cells (nRBCs) into circulation – ranging from the yolk sac, which exclusively releases primitive nRBCs, to the bone marrow, in which erythroid cells are enucleated before entering circulation. Although generally regarded as a consequence of high erythropoietic demand in the fetus, recent functional studies have indicated an immunosuppressive role for fetal nRBCs as well. DNA methylation (DNAm) is the addition of a methyl group to a cytosine base, a modification which does not change the underlying genetic sequence. DNAm mediates hematopoietic lineage commitment and can be a useful marker for cell composition and immune function in blood. Using the Illumina Infinium HumanMethylation450 BeadChip microarray, this thesis establishes DNAm profiles for major cord blood hematopoietic cells in both term and preterm births. In-depth examination of DNAm in term nRBCs revealed that epigenetic marks in this enigmatic cell population are likely highly regulated. Comparisons between cord blood iii  hematopoietic cells collected from term versus preterm births allowed for the identification of both cell-specific and systemic prematurity-associated differential methylation. These findings contribute to current understanding of the molecular mechanisms behind preterm birth and highlight candidate genes for follow-up gene expression or functional analysis of preterm hematopoietic cell populations, including CDC42EP1, CLIP2, FBXO31, the oncogene WWTR1, and tumour suppressor genes STK10 and RARRES3.    iv  Preface Ethics approval was required for Chapters 2-4, and obtained from the University of British Columbia Children’s & Women’s Research Ethics Board (certificate numbers H07-02681 and H04-70488). Written, informed parental consent to participate was obtained. Individual patient data is not reported. A version of Chapter 2 has been published. de Goede, O.M., Razzaghian, H.R., Price, E.M., Jones, M.J., Kobor, M.S., Robinson, W.P., and Lavoie, P.M. (2015) Nucleated red blood cells impact DNA methylation and expression analyses of cord blood hematopoietic cells. Clinical Epigenetics. 7(1): 95. I sorted cord blood cells for DNA methylation analyses, performed the DNA methylation data analyses, and drafted the manuscript. Dr. H.R. Razzaghian collected cord blood samples, sorted naïve CD4 T cells for transcriptomic analysis, and performed transcriptome data analysis. E.M. Price and Dr. M.J. Jones helped analyze data and reviewed the manuscript. Dr. M.S. Kobor contributed key reagents and reviewed the manuscript. Drs. W.P. Robinson and P.M. Lavoie supervised and designed the research, contributed key reagents, interpreted data, and edited the manuscript. Chapter 3 uses the same samples collected for Chapter 2. Dr. W.P. Robinson and I designed the analysis. I performed all data analysis. Additional cord blood samples for Chapter 4 were collected in the lab of Dr. P.L. Lavoie. I was responsible for cord blood cell sorting, participated in DNA methylation array processing in the lab of Dr. W.P. Robinson, and performed all data analysis. Dr. W.P. Robinson and I designed the analysis. v  Table of Contents  Abstract .......................................................................................................................................... ii Preface ........................................................................................................................................... iv Table of Contents ...........................................................................................................................v List of Tables ..................................................................................................................................x List of Figures ............................................................................................................................... xi List of Abbreviations ................................................................................................................. xiii List of Symbols .............................................................................................................................xv Acknowledgements .................................................................................................................... xvi Chapter 1: Introduction ................................................................................................................1 1.1 Overview ......................................................................................................................... 1 1.2 Immune system composition .......................................................................................... 2 1.2.1 Hematopoiesis ............................................................................................................. 2 1.2.2 Innate and adaptive immunity ..................................................................................... 4 1.2.3 Major hematopoietic immune cells ............................................................................. 5 1.3 Neonatal immunity.......................................................................................................... 7 1.3.1 Fetal hematopoiesis ..................................................................................................... 8 1.3.2 Distinguishing the neonatal and adult immune systems ........................................... 10 1.3.2.1 Differences in cellular composition .................................................................. 10 1.3.2.2 Nucleated red blood cells .................................................................................. 11 1.3.2.3 Functional differences in neonatal immunity ................................................... 15 1.3.2.4 Immunosuppression in fetal development ........................................................ 18 vi  1.3.3 Postnatal immune maturation ................................................................................... 19 1.4 Preterm birth ................................................................................................................. 20 1.4.1 Incidence and risk factors ......................................................................................... 20 1.4.2 Health impact of preterm birth .................................................................................. 21 1.4.3 The preterm immune system ..................................................................................... 22 1.5 DNA methylation .......................................................................................................... 23 1.5.1 Relationship with gene expression............................................................................ 24 1.5.2 Hematopoietic lineage commitment ......................................................................... 26 1.5.3 Epigenetics and hematopoietic cell function ............................................................ 26 1.5.4 DNA methylation and gestational age ...................................................................... 28 1.5.5 Array-based measurements of DNA methylation ..................................................... 29 1.5.5.1 Considerations for 450K array studies .............................................................. 29 1.6 Research objectives ....................................................................................................... 33 Chapter 2: Nucleated Red Blood Cell Contamination During Fluorescence-Activated Cell Sorting Impacts Epigenetic and Gene Expression Analyses of Cord Blood Cells .................34 2.1 Background ................................................................................................................... 34 2.2 Methods......................................................................................................................... 35 2.2.1 Sample collection and cell purification ..................................................................... 35 2.2.2 RNA extraction and genome-wide expression profiling .......................................... 38 2.2.3 Illumina Infinium HumanMethylation450 BeadChip ............................................... 38 2.2.4 DNA methylation data analysis ................................................................................ 39 2.3 Results and discussion .................................................................................................. 40 vii  2.3.1 Heterotopic cell interactions impact genome-wide signatures of hematopoietic cells .   ................................................................................................................................... 40 2.3.2 Revised DNA methylation profiles of hematopoietic cells obtained by a more stringent cell sorting strategy ................................................................................................ 44 2.3.3 Erythroid-specific differentially methylated sites ..................................................... 53 2.4 Conclusion .................................................................................................................... 56 Chapter 3: Characterizing the Hypomethylated DNA Methylation Profile of Nucleated Red Blood Cells from Cord Blood ......................................................................................................57 3.1 Background ................................................................................................................... 57 3.2 Methods......................................................................................................................... 59 3.2.1 Sample collection and the Illumina Infinium HumanMethlyation450 BeadChip .... 59 3.2.2 Combined blood cell and placental DNA methylation data analysis ....................... 59 3.2.3 Blood cell only DNA methylation data analysis ....................................................... 61 3.3 Results and discussion .................................................................................................. 62 3.3.1 Hypomethylation in nRBCs is distinct from placental hypomethylation ................. 62 3.3.2 nRBC DNA methylation is associated with nRBC proportion in cord blood .......... 71 3.3.3 Differentially methylated regions distinguishing nRBCs from WBCs..................... 77 3.4 Conclusion .................................................................................................................... 84 Chapter 4: Comparing Hematopoietic Cell DNA Methylation Profiles between Preterm and Term Births ...........................................................................................................................86 4.1 Background ................................................................................................................... 86 4.2 Methods......................................................................................................................... 87 4.2.1 Sample collection and cell purification ..................................................................... 87 viii  4.2.2 Illumina Infinium HumanMethylation450 BeadChip ............................................... 88 4.2.3 DNA methylation data preparation and analysis ...................................................... 88 4.3 Results and discussion .................................................................................................. 90 4.3.1 Comparing global DNA methylation of preterm and term immune cells ................ 90 4.3.2 Prematurity-associated DNA methylation at biologically relevant subsets of the genome .................................................................................................................................. 91 4.3.3 Prematurity-associated differentially methylated sites in hematopoietic cells ......... 96 4.3.4 Cell-specific DNA methylation patterns differ between term and preterm births .. 102 4.3.5 Comparing nRBC DNA methylation changes in preterm birth to changes with nRBC count ................................................................................................................................. 104 4.4 Conclusion .................................................................................................................. 105 Chapter 5: Conclusion ...............................................................................................................107 5.1 Summary of findings................................................................................................... 107 5.2 Strengths and limitations............................................................................................. 112 5.3 Future directions ......................................................................................................... 114 5.4 Conclusion .................................................................................................................. 117 References ...................................................................................................................................118 Appendices ..................................................................................................................................139 Appendix A : Supplementary material for Chapter 2 ............................................................. 139 A.1 Supplementary methods .......................................................................................... 139 A.2 Supplementary tables .............................................................................................. 142 A.3 Supplementary figures ............................................................................................ 143 Appendix B : Supplementary material for Chapter 3 ............................................................. 147 ix  B.1 Supplementary methods .......................................................................................... 147 B.2 Supplementary tables .............................................................................................. 148 B.3 Supplementary figures ............................................................................................ 153 Appendix C : Supplementary material for Chapter 4 ............................................................. 155 C.1 Supplementary methods .......................................................................................... 155 C.2 Supplementary tables .............................................................................................. 157 C.3 Supplementary figures ............................................................................................ 163  x  List of Tables  Table 1.1 Absolute blood cell count ranges for adult peripheral blood, term cord blood, and preterm (<37 weeks GA) cord blood. ............................................................................................. 7 Table 2.1 Number of cell-specific DM CpG sites (FDR <5%) following the standard and stringent FACS strategies. ............................................................................................................ 48 Table 2.2 Top eight CpG sites with nRBC-distinct DNAm from white blood cells in cord blood........................................................................................................................................................ 54 Table 3.1 Number of CpG sites in each blood cell population that show differential methylation (FDR <5%) with the proportion of nRBCs in cord blood. ........................................................... 77 Table 4.1 Subject characteristics and cell types collected from each subject ............................... 88 Table 4.2 Number of prematurity-DM sites for each cell type (FDR <5%, |Δβ| >0.10). ............. 97 Table 4.3 Overlap between cell-specific prematurity-DM sites (FDR <5%, |Δβ| >0.10) and source-DM sites. ........................................................................................................................... 98 Table 4.4 Location and genomic context of the 25 prematurity-DM sites (FDR <5%, |Δβ| >0.10) common to T cells, granulocytes, monocytes, and nRBCs......................................................... 101 Table 4.5 Number of cell type-DM sites (FDR <5%, |Δβ| >0.20) within preterm samples, within term samples, and in common between the two GA groups....................................................... 102  xi  List of Figures  Figure 1.1 Classical model of hematopoiesis ................................................................................. 3 Figure 1.2 Shifting hematopoietic sources throughout gestation.................................................... 8 Figure 1.3 Major steps of erythropoiesis. ..................................................................................... 12 Figure 1.4 Technical considerations for the Illumina 450K array. ............................................... 32 Figure 2.1 Schematic representation of cell sorting strategies. .................................................... 37 Figure 2.2 DNAm profiles of cord blood cells isolated by the standard FACS strategy. ............. 42 Figure 2.3 Genome-wide transcriptomic profiles of naïve CD4 T cells sorted by either the standard or stringent FACS strategy. ............................................................................................ 44 Figure 2.4 DNAm profiles of cord blood cells isolated using the stringent FACS strategy. ........ 46 Figure 2.5 DNAm changes with FACS strategy in nRBCs, T cells and monocytes at their top DM sites. ....................................................................................................................................... 49 Figure 2.6 Selected discordant DM sites between the standard and stringent FACS protocols. .. 51 Figure 2.7 Erythroid-WBC interactions during FACS affect DNAm based on the proportion of nRBCs. .......................................................................................................................................... 53 Figure 3.1 nRBCs and placenta are both hypomethylated cell/tissue types. ................................ 64 Figure 3.2 Cell-specific DNAm patterns differ at distinct subsets of the genome. ...................... 66 Figure 3.3 Two types of nRBC mixtures could give rise to the variable nRBC DNAm profile. . 70 Figure 3.4 nRBC DNAm is negatively correlated with nRBC proportion in cord blood. ............ 73 Figure 3.5 Summary of nRBC DMRs........................................................................................... 78 Figure 3.6 Examples of nRBC-specific DMRs............................................................................. 81 Figure 3.7 nRBC-specific DMRs with nRBC count-associated DNAm. ..................................... 83 xii  Figure 4.1 Genome-wide DNAm comparisons between major hematopoietic cells in term and preterm births. ............................................................................................................................... 91 Figure 4.2 DNAm relationships between GA groups and cell types differ at different regions of the genome. ................................................................................................................................... 93 Figure 4.3 Cell type-DM sites (FDR <5%, |Δβ| >0.20) within all preterm samples and within all term samples, grouped by CpG density and DNAm relative to other cell types. ....................... 103  xiii  List of Abbreviations  450K array Illumina HumanMethylation450 BeadChip AGM  aorta-gonad-mesonephros ANOVA analysis of variance APC  antigen-presenting cell APP  antimicrobial proteins and peptides BCG  bacille Calmette-Guérin BM  bone marrow CBMC  cord blood mononuclear cell CGI  cytosine-guanine dinucleotide island CpG  cytosine-guanine dinucleotide DAMP  danger-associated molecular patterns DM  differentially methylated DMR  differentially methylated region DNAm DNA methylation FACS  fluorescence-activated cell sorting FDR  false detection rate FL  fetal liver GA  gestational age GO  gene ontology HDAC  histone deacetylase HSC  hematopoietic stem cell xiv  Ig  immunoglobulin IL  interleukin LCR  locus control region LINE1  long interspersed element NET  neutrophil extracellular trap NK cell natural killer cell nRBCs  nucleated red blood cell PAMP  pathogen-associated molecular pattern PCA  principal component analysis PRR  pattern recognition receptor PTB  preterm birth RE  repetitive element RBC  red blood cell RefSeq Reference Sequence RNAi  RNA interference SNP  single nucleotide polymorphism SV  surrogate variable SVA  surrogate variable analysis Th  T helper cell TLR  Toll-like receptor Treg  regulatory T cell TSS  transcription start site WBC  white blood cell xv  List of Symbols  β DNA methylation beta values Δβ Change in DNA methylation beta value xvi  Acknowledgements  I would like to acknowledge my supervisor, Dr. Wendy Robinson, for welcoming me into her lab and teaching me so much. Thank you for your investment in my growth as a researcher and as an individual. Thanks to the Robinson lab for showing me the ropes and being up for any discussion: Maria, Ruby, Johanna, Irina, Kirsten, Magda, Sam, Chaini, Giulia, Tanjot, Yao, Olive, and Thamar. I would also like to acknowledge all of the helpful advice from my advisory committee, Dr. Pascal Lavoie and Dr. Carolyn Brown. Thank you to the University of British Columbia and the Canadian Institute of Health Research for their financial support. Thanks to my mom, my dad, Andrea, and Josh for all of the love and support, and thanks to the UBC soccer team for being my crazy second family. Words cannot express how much you all mean to me. 1  Chapter 1: Introduction 1.1 Overview Preterm birth (PTB), or birth prior to 37 weeks gestational age (GA), occurs in approximately 11% of live births1. PTB is the greatest direct cause of neonatal death and increases the risk of neonatal infection. PTB is associated with a variety of short- and long-term sequelae, including increased risk for respiratory illnesses and learning disabilities in childhood as well as hypertension and heart disease in adulthood2.  Premature birth interrupts fetal immune development, resulting in an immune system with distinct cell type composition and cell-specific limitations in immune responses. Epigenetic marks, such as DNA methylation (DNAm), have been found to mediate changes in both immune system structure and function3-6. The goal of my project is to characterize the DNAm profiles of key immune cells in term and preterm neonates, to improve our understanding of immune system regulation in these populations. Identifying epigenetic patterns unique to the preterm immune system may reveal methods for early identification or treatment of neonatal infection in this vulnerable group. In the introduction of this thesis, hematopoietic processes that give rise to the immune system will be discussed, focusing on how these processes develop over gestation. Ways in which the neonatal immune system deviates from the adult immune system will then be described, as well as how these differences are exaggerated in preterm infants. Finally, the role of epigenetics in cell fate decisions and fetal development will be outlined.  2  1.2 Immune system composition 1.2.1 Hematopoiesis Hematopoiesis is a popular system for studying stem cells and lineage commitment, partially due to the simplicity of its original model. The classic model of hematopoiesis has a hierarchical structure crowned by hematopoietic stem cells (HSCs), which have the potential to differentiate into any blood cell lineage and are capable of self-renewal (Figure 1.1). As a first step towards differentiation, HSCs become multipotent progenitors, which cannot self-renew but maintain the potential to commit to any blood cell lineage7. Lineage commitment then progresses into a series of bifurcating decisions, the first of which being between the common myeloid progenitor and the common lymphoid progenitor8,9. The myeloid route can lead to the granulocyte-monocyte lineage (with monocytes, macrophages, neutrophils, basophils, eosinophils and mast cells as terminally-differentiated cells) and the megakaryocyte-erythroid lineage, whereas the lymphoid route can branch into B cells, T cells, and natural killer (NK) cells.  3   Figure 1.1 Classical model of hematopoiesis  Improvements in technology and study design have revealed that hematopoiesis is more complex than the classical model indicates. Transcriptome profiling and cell surface marker analyses have revealed an increasing number of intermediate cell populations, which blur the boundaries of the traditional hematopoietic hierarchy10-13. Rather than abrupt pairwise decisions, cell lineage commitment is more likely to be a progression towards a given cell fate, which is inherently more flexible and sensitive to changes in gene expression and the microenvironment. For example, one of the first studies to assess hematopoiesis by individual cell manipulation, rather than by populations of cells, identified multilymphoid progenitors that could give rise to not only all lymphoid cells but also monocytes, macrophages, and dendritic cells14. This revealed 4  that the earliest cell fate decision is not necessarily a myeloid-lymphoid one, and is one of many studies that have shifted our understanding of hematopoiesis.  1.2.2 Innate and adaptive immunity The immune system can be divided into innate and adaptive immune responses, which work in concert to protect against microorganisms. Adaptive immunity is mediated by lymphocytes and involves an immunological “memory” for prior exposures. This memory is used to produce pathogen-specific immune responses, conferring long-term protection to the individual. Upon antigen recognition, adaptive immune cells operate to inactivate or directly kill the invading target cell as well as initiate appropriate downstream immune responses15. Innate immunity is a non-specific system that provides an immediate, short-term response to pathogens. The innate immune system relies on phagocytic cells such as granulocytes, macrophages and dendritic cells. These cells express various pattern recognition receptors (PRRs), which recognize pathogen- and danger-associated molecular patterns (PAMPs and DAMPs, respectively) and subsequently trigger an immune response cascade16,17. This culminates in the release of antimicrobial proteins and peptides (APPs), and components of the complement system. Both phagocytic cells and the soluble factors they produce work together to identify, digest, and destroy invading microorganisms. Certain phagocytes link the innate and adaptive immune systems by presenting antigenic peptides to lymphocytes, thus instructing adaptive system development17,18. The innate immune system is also capable of developing its own form of immunological memory, in a phenomenon called trained immunity19. This immune “training” involves phenotypic and gene regulatory changes to innate immune cells that result in their heightened activity in response to a secondary infection. Antigen-specific immune training 5  has been shown in invertebrates and plants, however only nonspecific responses have been observed in vertebrates to date19.   1.2.3 Major hematopoietic immune cells The myeloid lineage can be divided into two general cell types: granulocytes, and monocyte-derived cells. Granulocytes, which include neutrophils, basophils, eosinophils, and mast cells, are blood cells with multi-lobed nuclei and secretory granules in their cytoplasm. Neutrophils are the most abundant white blood cell (WBC) (Table 1.1) and are key phagocytes of the innate immune response. They are highly motile, and upon infection travel to the affected site by chemotaxis. Neutrophils then combat the invading microorganism by phagocytosis, release of antimicrobial granules, or formation of neutrophil extracellular traps (NETs)20. Basophils, eosinophils, and mast cells are much rarer than granulocytes (Table 1.1). These cells produce key cytokines, growth factors, and other signaling molecules like histamine and heparin as part of inflammatory reactions to parasitic infections, allergies, and asthma20-22.  Monocytes and macrophages are major phagocytes in the immune system that are recruited to sites of infection to attack the foreign microbe while also affecting downstream immune responses. Macrophages additionally have constitutive phagocytic functions, clearing dead cells and debris from tissues and circulation even in the absence of an immune response. Resident macrophage populations are present in tissues throughout the body, derived from a combination of yolk sac macrophages, fetal monocytes, and adult hematopoietic progenitors23. Dendritic cells can be derived from both myeloid and lymphoid lineage commitment pathways, and are an essential type of antigen-presenting cell (APC). Professional APCs – which also include macrophages and B cells – display fragments of antigens from pathogens or 6  allergens bound to major histocompatibility complex II molecules on their membrane. These APCs then inform and activate the adaptive immune system through interaction with B and T lymphocytes20,24. The lymphoid lineage is made up of B and T cells, and NK cells. NK cells have roles in both innate and adaptive immune function. They are one of the first responders to infection, releasing granules upon recruitment that perforate the foreign target’s cell membrane and induce apoptosis20. However, NK cells also regulate adaptive immunity through cytokine production, and can become “memory NK cells” that are primed to respond to previously-encountered pathogens25.  B cells have two roles: establishing humoral immunity by producing antibodies, and presenting antigens to T cells. The antibodies secreted by B cells, called immunoglobulins (Igs), fall into 5 main categories: IgM, which is involved in complement activation; IgE, which activates the allergic response; IgA, which is involved in mucosal immunity; IgD, which activates basophils; and IgG, which has broad functions in complement activation, neutralization of microbes, and opsonization20,24,26,27.  T cells are a diverse cell type, and are classified into subsets by their distinct cell-surface markers and functions. CD8 T cells, called cytotoxic T cells, release cytotoxic proteins upon antigen recognition that cause apoptosis in the target cell. CD4 T cells, termed T helper (Th) cells, coordinate immune responses through the secretion of cytokines and other factors. There are further subsets of both types of T cell, including a wide array of Th cells. Three of the most studied Th cell subtypes are Th1 cells, which promote cell-mediated immunity and respond to intracellular pathogens; Th2 cells, which promote humoral immunity and respond to extracellular parasitic pathogens; and Th17 cells, which produce interleukin (IL)-17 and respond to 7  extracellular bacterial and fungal pathogens28-30. Another key T cell population is the regulatory T cell (Treg), which is an immunomodulatory cell type that prevents immune hyperactivity20. Although not considered immune cells, erythrocytes are the most abundant cells in blood, present at around 1000 times the number of WBCs. In adults, nearly all red blood cells (RBCs) are enucleate by the time they enter circulation. In contrast, many fetal circulating RBCs retain their nucleus, with the proportion of nucleated red blood cells (nRBCs) in the fetus declining as gestation progresses31 (Table 1.1).  Table 1.1 Absolute blood cell count ranges for adult peripheral blood, term cord blood, and preterm (<37 weeks GA) cord blood. All cell counts reported as x103 cells/µL.  Adult peripheral blood Term cord blood Preterm cord blood Total 4.0 - 10.5 # 10 - 26● 5 - 19● Neutrophils 2.0 - 6.4 # 5 - 13● 2 - 9● Eosinophils 0.1 - 0.2 # 0.2 - 2.0● 0.1 - 0.7● Basophils 0 - 0.09 # 0 - 1● 0 - 1● Monocytes 0.2 - 0.9 # 0.5-3● 0.3 - 1● Lymphocytes (total) 1.0 - 3.2 # 3.5-8.5● 2.5 - 6● NK cells 0.13 - 0.25○ 0.15-1.89 ^ 0.10 - 1.62 ^ B cells 0.16 - 0.27○ 0.35-1.17 ^ 0.09 - 2.29 ^ T cells (total) 1.0 - 1.5○ 1.64-3.83 ^ 1.21 - 4.65 ^ CD4 T cells 0.60 - 0.98○ 1.07-2.56 ^ 0.91 - 3.23 ^ CD8 T cells 0.42 - 0.66○ 0.50-1.24 ^ 0.27 - 1.42 ^ Erythroblasts 0 0 –0.64* 0 – 1.67* ● = reference ranges (5-95 percentiles), Milcic32; * = reference ranges (25-75 percentiles), Perrone et al.33; ○ = reference ranges (25-75 percentiles), Hulstaert et al.34; # = reference ranges (lower-upper limit), Adeli et al.35; ^ = reference ranges (lower-upper limit), Walker et al.36  1.3 Neonatal immunity The prenatal period is a critical time in which the fetus’ immune system is preparing for life outside of the sterile womb. However, the fetus is semi-allogeneic to the mother, sharing 8  only 50% of her DNA. Nucleated cells cross the placenta in both directions, so both the maternal and fetal immune systems must be kept in check to avoid anti-fetal and anti-maternal immune responses, respectively37. Excessive inflammation during pregnancy increases risk for spontaneous abortion and preterm birth38. The fetal immune system changes markedly over gestation to adapt to these shifting demands, as does the neonatal immune system in the period immediately after birth.  1.3.1 Fetal hematopoiesis Throughout gestation, hematopoiesis transitions between multiple locations in the fetus and extraembryonic membranes in a defined temporal pattern (Figure 1.2). The blood cells produced by each hematopoietic organ vary widely and reflect the changing demands of the fetus with immune system development and increasing oxygenation.   Figure 1.2 Shifting hematopoietic sources throughout gestation. Adapted from Mikkola, 2005 and Dzierzak, 200939,40. 9  Hematopoiesis begins in the yolk sac, at day 16 of development41. Blood cells produced by the yolk sac are largely of erythroid lineage, but early myeloid cells are also produced42,43. Yolk sac erythrocytes are primitive: they are larger than definitive erythrocytes, retain their nucleus in circulation, and express embryonic hemoglobins42. By 7 weeks gestation, yolk sac hematopoietic progenitors are no longer detectable44,45. As hematopoiesis in the yolk sac declines, the aorta-gonad-mesonephros (AGM) and the placenta become dominant hematopoietic sources. The AGM, an embryonic structure consisting of the dorsal aorta and urogenital ridges, displays de novo formation of HSCs from approximately 32 days gestation, although these HSCs are relatively low in number39,46. Placental hematopoietic activity has been suggested to occur from 8-17 weeks gestation47, but HSCs are present in the placenta from 6 weeks gestation all the way to term48. Placental HSCs associate closely with stromal cells thought to support the hematopoietic process, suggesting that the placenta is not just a reservoir for HSCs but is actively contributing to fetal hematopoiesis41. Erythroid maturation is also facilitated by the placenta early in gestation (from day 24 to 7 weeks), with primitive erythroblasts associating with placental macrophages (Hofbauer cells) during the enucleation process49,50. The fetal liver is the largest contributor to hematopoietic activity from 9 to 24 weeks’ gestation44. However, the liver does not generate HSCs de novo, and requires seeding from other organs. When hematopoiesis is initiated in the fetal liver, the placental HSC pool is as much as 15 times greater than that of the AGM51. Thus, it has been suggested that the majority of liver-seeding HSCs come from the placenta, with smaller contributions from the AGM and yolk sac39. Throughout gestation, fetal liver hematopoiesis expands its repertoire from a mostly erythroid-focused process to also produce megakaryocytic cells, other myeloid cells, and B cells39. At 10  around 8-9 weeks gestation, the thymus and spleen are seeded from the liver and make moderate contributions to lymphopoiesis (the thymus)52 and to the myeloid and erythroid lineages (the spleen)24,53,54. At 24 weeks’ gestation, hematopoiesis in the liver declines to accommodate hepatocyte proliferation24.  As hematopoiesis is reduced in the fetal liver, it increases in bone marrow. Although the bone marrow does not become the predominant source of blood cells until gestational week 24, HSCs (derived from the fetal liver) have been observed in the bone marrow as early as 10 weeks gestation43,44. Early in bone marrow hematopoiesis, myeloid and erythroid cells are produced in equal number, however the myeloid:erythroid ratio becomes similar to the adult level of 3:1 by the time the bone marrow is the predominant hematopoietic organ44. The bone marrow remains the main site of hematopoiesis after birth and throughout life.  1.3.2 Distinguishing the neonatal and adult immune systems Infants are much more vulnerable to infection than adults, due to their relatively limited immune response. Neonatal immunity is characterized by an unprimed adaptive immune system and a consequently heavy reliance on innate immunity. This section will highlight some of the cellular and molecular differences between neonatal and adult immune systems, and will also describe the influence of maternal factors on fetal immune development.  1.3.2.1 Differences in cellular composition Term cord blood shows higher overall WBC counts relative to adult blood, with high numbers of neutrophils in neonates being one of the largest contributors to this difference (Table 1.1). This elevated neutrophil count is transient, with levels decreasing to a low at 72 hours after 11  birth, and stabilizing at roughly one-third of the number at birth by 120 hours of age55,56. Further distinctive features of neonatal blood include: 1) a relative lack of memory T cells compared to adult T cells24; 2) high numbers of Tregs57,58; 3) a higher ratio of CD4 Th cells to CD8 cytotoxic T cells59; 4) a higher number of unconventional B-1 (CD54+) B cells60; 5) fewer circulating dendritic cells, which may impair T cell memory development61; and 6) the frequent presence of nRBCs, whereas RBCs in the healthy adult are enucleate. The development of nRBCs and their potential impact on fetal and neonatal immunity are discussed in the next section.  1.3.2.2 Nucleated red blood cells Early in gestation, the yolk sac produces primitive nRBCs, which are distinguished from definitive RBCs in their large size, short maturation period and lifespan, and expression of embryonic hemoglobin. Primitive nRBCs also generally retain their nucleus in circulation, although there is evidence of primitive erythroblast enucleation in the placenta50. Erythropoiesis is exclusively primitive from 3-6 weeks gestation, and primitive nRBCs have been observed throughout the first trimester of gestation. At 6 weeks gestation, definitive erythropoiesis begins, with erythroblasts enucleating in their site of origin – typically the liver or bone marrow50,62. Erythropoiesis is a two-phase process: 1) erythroid progenitor proliferation, in which HSCs commit to the erythroid lineage by proliferating and differentiating into burst forming unit-erythroid and colony forming unit-erythroid; and 2) terminal erythroid differentiation, in which the proerythroblast differentiates into enucleated reticulocytes63. Terminal erythroid differentiation occurs with the proerythroblast progressing through the basophilic, polychromatic, and orthochromatic erythroblast stages in 3 cell divisions. The orthochromatic erythroblasts then enucleate to become reticulocytes, which mature in circulation to become 12  RBCs. Each intermediate cell in the erythropoietic process is markedly distinct, reflecting the large-scale changes in membrane composition, the cytoskeleton, and nuclear structure required before the erythroblast is ready to expel its nuclei (Figure 1.3). Some of the many changes occurring during terminal erythroid differentiation include decreasing cell size, increasing chromatin condensation, increasing hemoglobin, decreasing levels of adhesion molecules, and increasing levels of most major transmembrane and cytoskeletal proteins64. Once condensed, the nucleus is polarized, and the erythroblast prepares to divide into the reticulocyte, the eventual mature RBC, and the pyrenocyte, the membrane-surrounded nucleus that will likely be digested by macrophages. This causes dramatic regional differences in the erythroblast, which are mediated by protein sorting, membrane maturation, vesicle trafficking, and autophagy65. Another key change in maturing erythroblasts is the progressive global demethylation of DNA, which has been observed in both mouse and human erythropoiesis66,67.   Figure 1.3 Major steps of erythropoiesis. As the erythroid cell progresses through terminal differentiation: the nucleus condenses; hemoglobin concentration increases; the cytoskeleton polarizes the nucleus and assembles a contractile actin ring; and regional differences in membrane and cytosol composition develop, based on whether that area will be incorporated into the pyrenocyte or the reticulocyte. Adapted from Ji, 201168. 13  The proportion of nRBCs declines throughout gestation, but nRBCs are commonly seen in term cord blood: 0-10/100 WBC is considered a healthy range, with 1-2/100 WBC being the most common31,55,69. In absolute terms, 500 nRBC/µL has been suggested to be a “normal” count31. After birth, nRBC counts drop rapidly: all nRBCs are cleared from the bloodstream within 4 days of life in healthy term infants and within around 7 days in healthy preterms. There is wide inter-individual variability in nRBC proportion, as it depends on both total leukocyte count and rate of erythropoiesis and enucleation69. Levels of nRBCs above the typical neonatal range of 0-10/100 WBC are associated with a variety of acute and chronic stimuli, such as hypoxia, asphyxia, hemolysis, fetal anemia, and maternal smoking, as well as pregnancy complications including intrauterine growth restriction, premature prolonged rupture of membranes, gestational diabetes, and chorioamnionitis31,70-75. Since bone marrow architecture does not allow nRBCs to enter circulation, it is thought that adverse events increase nRBC levels either through activation of extramedullary erythropoiesis70 or through increased levels of erythropoietin, which upregulates erythropoiesis and influences bone marrow structure and blood flow31,76,77. However, a small proportion of observed elevated nRBC cases are idiopathic – including observations of nRBC counts as extreme as >100/100 WBC69 – making nRBCs an enigmatic cell population. High proportions of nRBCs are also associated with adverse postnatal outcomes, including perinatal brain damage, cerebral palsy, early-onset sepsis, intraventricular hemorrhage, and necrotizing enterocolitis78-83. The association between elevated nRBC count and newborn morbidity and mortality is particularly strong when nRBCs persist after birth71. The functional importance of nRBCs in humans, if one exists, is not well understood. In non-mammalian vertebrates, the mature erythrocyte retains its nucleus and participates in immune system processes. These mature nRBCs express PRRs, can recognize PAMPs, and 14  produce cytokines in response to PRR activation84-86. Early studies of potential immune roles for nucleated erythrocytes showed that nRBCs from mice had immunosuppressive effects in vivo, which included dampening of antibody responses and limiting B cell proliferation87-89. More recent studies have revealed other immunomodulatory interactions between RBCs and WBCs, including regulation of T cell and neutrophil survival90-92 and immunosuppression of T and B cells and dendritic cells93-96. Further evidence for a potential immune role for nRBCs came from a series of experiments performed on mouse and human CD71+ erythroid cells which, among other findings, showed that neonatal splenic CD71+ cells suppress production of innate immune cytokines via arginine depletion93. This immunomodulatory effect was found to be specific to neonatal CD71+ erythroid cells and not displayed in their adult counterparts, with neonatal cells expressing higher levels of the arginine-depleting enzyme arginase-293. The immunosuppressive effects of nRBCs have been suggested to occur by direct cell-cell interactions, nRBC-derived cytokines, and other soluble factors87,97,98. However, it should be noted that a recent study could not replicate the same specific effects of neonatal CD71+ splenocytes: for example, although these cells did reduce adult dendritic cell cytokine production ex vivo, co-culturing with either neonatal or adult bone marrow produced the same effect99. The lack of consensus about neonatal nRBC function calls for further study of this unusual cell population. The presence of nRBCs in fetal and neonatal circulation is generally considered a consequence of high erythropoietic demand, so this cell population is relatively unstudied. Based on their uniqueness to the fetus and neonate, documented association with other immune cells, and relatively uncharacterized molecular composition and function, nRBCs may provide novel insight into the differences between neonatal and adult immunity.  15  1.3.2.3 Functional differences in neonatal immunity Not only does immune cell composition differ in newborns, but the neonatal counterparts of adult immune cells often display markedly distinct responses to pathogens and inflammatory stimuli. These functional differences likely contribute to neonates’ increased susceptibility to infection, as well as the limited effectiveness of vaccines early in life. Adult naïve T cells require an initial infection and response event to establish immunological memory, but the relatively sterile environment of the womb does not provide fetal T cells with many opportunities to develop their adaptive immune system. That is not to say that the fetus is completely isolated from antigens: the placenta has its own microbiome and pathogens are present in the intrauterine environment100,101, plus maternal cells in fetal circulation provide exposure to non-inherited maternal alloantigens37,58. Indeed, CD4 T cells with an effector memory phenotype are present in cord blood at an abundance of 1-6%; although this is far less than the adult abundance of approximately 50%, it is still indicative of T cell activation in utero102. Another limitation of adaptive immune system development is the immunotolerant profile of the fetus, which is necessary to prevent maternal-fetal rejection and PTB. Maternal- and placental-mediated immunosuppression will be discussed in Section 1.3.2.4, but the fetal immune system also plays a role. One of the fetal contributions to immunotolerance is the “waves” of T and B cell production through gestation. Earlier populations of lymphocytes are more prone to innate or tolerant responses than those produced later in gestation, which behave in a more “adult-like” manner58,103,104. There are some exceptions to the limited neonatal adaptive immune response, including observations of functional, adult-like cytotoxic T cells in neonates who experienced in utero infection of human cytomegalovirus, and of Ascaris reactivity in infants whose mothers were infected by the parasite during pregnancy105,106. In 16  general, however, neonatal T and B cells are largely unprimed for immune responses immediately following birth.  Despite this lack of immune education, neonatal naïve T cells perform important immune functions by activating innate immune system responses. Recently, it has been shown that naïve CD4 T cells from both preterm and term births can produce CXCL8 (or IL-8) upon stimulation107. CXCL8 is a potent neutrophil activator that is typically associated with myeloid and epithelial cells, and innate immune responses. The ability for naïve T cells to produce cytokines without differentiation into a helper cell is unique to the neonate and likely essential for rapid immune responses before immunological memory has been established. Neonates have atypical responses following activation of Toll-like receptors (TLRs) and other PRRs that result in skewing of T helper cells towards Th2-polarized responses, and away from Th1- and Th17-polarized responses that target intracellular pathogens and extracellular mucosal pathogens, respectively. In response to certain types of TLR activation, neonatal dendritic cells and monocytes have been found to produce reduced amounts of pro-inflammatory cytokines IL-1α, IL-1β, TNF-α, IL-18 and IL-12p70 compared to adults, but equal or greater secretion of anti-inflammatory cytokines IL-6 and IL-10, as well as increased IL-4 production by basophils108-112. These differences from adult cytokine production are not due to inadequate expression of TLRs18,113, but limitations in cytokine production and secretion – for example, neonatal immune cells do not show the same capacity to produce multiple cytokines simultaneously as adult immune cells18. This unbalanced cytokine production biases the immune system towards the extracellular pathogen defence systems of Th2 responses, but limits Th1 responses16,18. Additionally, the increased proportion and activity of Treg cells in the neonate 17  promote a tolerogenic, anti-inflammatory immune profile57,58. Fetal Tregs are critical to inhibit immune responses to the maternal cells that cross the placenta into the fetal blood stream37,104. Neonatal T cell development is further impacted by altered dendritic cell function. In addition to being less abundant in infant circulation than in adults, neonatal dendritic cells are less responsive to TLR-related activation57,61. Specifically, neonatal dendritic cells produce less IL-12p70 in response to TLR stimulation, which is an important cytokine for the Th1 pathway57. This is partially caused by the increased basophilic IL-4 production, which limits IL-12p70 production in dendritic cells112. T cell polarization is further influenced by high IL-23 production in neonatal dendritic cells that, in combination with increased production of IL-1β and IL-6, promotes Th17 responses114. Infants show differences in a variety of circulating molecules relative to adults, with higher concentrations of immunosuppressive factors promoting 1) fetal tolerance of circulating maternal cells and 2) anti-inflammatory responses to postnatal microbial colonization. For example, high neonatal concentration of adenosine, a purine metabolite, is thought to inhibit the production of pro-inflammatory cytokine TNF-α while maintaining production of IL-6, an anti-inflammatory Th2-skewing cytokine113. This effect is amplified by neonatal immune cells’ high sensitivity to adenosine. Neonatal plasma also has significantly lower amounts of various APPs57,115 and decreased concentrations of various complement components relative to adults, the latter of which is accompanied by decreased activity of both the classical and alternative complement pathways44. Reduced levels of fibronectin in neonatal plasma and the extracellular matrix may reduce attachment between phagocytes and invading microbes, further limiting innate immune function44. 18  Recruitment and functioning of first-response innate immune cells, such as phagocytes and NK cells, are also impaired in neonates relative to adults. Neonatal neutrophils are slow to mobilize and proliferate, and have reduced phagocytic abilities once recruited to the affected site57,116,117. These cells also are less effective at forming NETs as well as releasing APPs and reactive oxygen species to kill microbes115,118-120. Monocytes display normal antibody-dependent cellular cytotoxicity in the neonate, but are limited in their chemotactic abilities and have reduced phagocyte-induced cell death abilities44,121. NK cells show poor cytotoxic function in neonates compared to adults, which manifests as reduced production of key cytokines and impaired degranulation25,57.  1.3.2.4 Immunosuppression in fetal development Diminished immune reactivity in infants was once attributed to a general immaturity of the neonatal immune system, but now is thought to be actively maintained. This immunosuppression is protective, reducing the risk of maternal-fetal rejection during pregnancy and preventing excessive inflammation in the newborn during microbial colonization. Much of this immunosuppression is mediated by the placenta. Maternal Tregs pass through the placenta into fetal circulation57, and the placenta also secretes several factors that promote Th2-polarized responses or inhibit Th1-polarized responses, such as IL-4, IL-10, prostaglandin E2, and progesterone58,122,123. In addition to their influence on T cell polarization, these placental factors promote decidual Treg development and M2 (alternative) macrophage production, although the latter has only been demonstrated in mice to date124. Additionally, both newborns and pregnant women contain high numbers of circulating myeloid-derived suppressor cells that may inhibit T cell proliferation, promote regulatory T cell development, and limit T and NK cell cytotoxic 19  activities through expression of modulatory factors like arginase-1, reactive oxygen species, and nitric oxide synthase38. Finally, arginase-expressing cells (including CD71+ erythrocytes and hepatocytes) reduce production of the antimicrobial nitric oxide, limit activation of TLR4 pro-inflammatory responses, and inhibit NK and T cell maturation93,125,126. However, it should be noted that the suggested roles of myeloid-derived suppressor cells and CD71+ erythrocytes are relatively new and contentious, and require further study.  1.3.3 Postnatal immune maturation The neonate’s immune system undergoes drastic changes immediately after birth to adjust to life beyond maternal control and protection. Two major events occur during parturition that demand these immune system changes: release of placental control, and exposure to microbes. As described in the previous section, the placenta is one of the major mediators of the distinct fetal immune profile. It curbs fetal immune reactivity through release of anti-inflammatory cytokines, prostaglandin E2, and progesterone, while also supplementing the fetal immune system with key plasma factors from the mother, most notably IgG44. Following birth, the neonatal immune system must adjust to the release of placental aid and control; this is partially achieved through breast feeding, which confers WBCs and circulating factors such as IgA, lysozyme, and lactoferrin to the newborn44,127,128. Additionally, a successful transition from the relatively sterile womb to the external environment requires that the newborn accommodates their developing skin and gut microbiome while curtailing infection and hyperinflammatory responses. These two changes, in combination with the predominance of less immunotolerant T cells arising from a later-gestation “wave” of lymphopoiesis, all contribute to the gradual 20  upregulation of Th1-polarized responses over early infancy24,104. B cell maturation is also upregulated as a result of microbial colonization129.  1.4 Preterm birth 1.4.1 Incidence and risk factors PTB, defined as live birth prior to 37 weeks GA, is a global health concern. In 2010, there were 14.9 million cases of PTB, representing 11.1% of total live births1. PTBs are overrepresented in cases of neonatal morbidity and mortality, with prematurity being a risk factor in over 50% of neonatal deaths1,17. Although rates are generally higher in low-income countries (11.8%) versus high-income countries (9.3%), there is significant variability on a per-country basis – for example, the 2011 PTB rate in the U.S. was 11.7%1,130. PTB and neonatal infections also pose a considerable economic problem, with an estimated annual cost of $25 billion in the U.S. alone131.  PTB has three main clinical etiologies: spontaneous labour with intact membranes (40-45% of PTBs), preterm pre-labour rupture of membranes (25-30% of PTBs), and medically indicated PTBs (30-35% of PTBs)132. Risk factors for PTB include male fetal sex, assisted reproductive technologies, and maternal health conditions like renal disease, hypertension, and obesity1,132,133. Although improved clinical care has increased the survival rates of PTB132, the rates of PTB are still on the rise due to factors like increased obstetrical intervention and rising maternal age1,131,132. Additionally, the heterogeneous etiology of PTBs has complicated our attempts to understand the disorder.  21  1.4.2 Health impact of preterm birth One of the major complications associated with prematurity is infection, which causes approximately 1.6 million neonatal deaths annually134. The incidence of infection is GA dependent, progressing from 46% in infants born <25 weeks gestation to 29% at 25-28 weeks, 10% at 29-32 weeks, and 2% at >32 weeks135. GA is also a significant predictor of death in infected neonates17. The major cause of increased infection susceptibility in preterm newborns is likely their underdeveloped immune systems, described in the section below. However, a variety of other physiological factors, such as underdeveloped organs, anemia, and increased oxidative stress contribute to the poor health outcomes of PTB, which include bronchopulmonary dysplasia and necrotizing enterocolitis17,136. The invasive medical treatments preterm infants typically require, as well as the extended time they are exposed to microbes in the neonatal intensive care unit, also contribute significantly to their increased risk of infection17,137. The immediate health concerns the preterm neonate faces set the stage for a disease-burdened life. In childhood, those who were born preterm are more likely to develop cerebral palsy, sensory deficits, and respiratory complications like asthma138-140. Those born preterm are more frequently hospitalized for infections during childhood141. PTB has also been associated with atypical behaviour and cognition in childhood, including learning disabilities, attention-deficit hypertension disorder, anorexia nervosa (in females), and mental illness (in males), as well as decreased cognitive performance138,142-144. Prematurely-born adults are more prone to chronic diseases such as type 2 diabetes, hypertension, coronary heart disease, and stroke2,145-147.  22  1.4.3 The preterm immune system Fetal hematopoietic development spans the entire gestational period, and PTB interrupts this process. As a result, there is a variety of deficiencies in the preterm immune system, with the greatest defects observed in immune signalling and antigen presentation rather than in pathogen killing17.  Throughout late gestation, the mother transmits Igs (mostly IgG) to the fetus via the placenta. Most of the IgG is transmitted after 32 weeks gestation, which limits the levels of IgG present in preterm infant circulation at birth148,149. Preterm neonates also produce even lower amounts of pro-inflammatory cytokines (such as IL-6 and TNF-α) relative to term infants, who already produce low amounts of these cytokines relative to adults131,150-154. As described previously, this anti-inflammatory cytokine profile biases the Th cell population against Th1- and Th17-polarized responses in term neonates, and this skewing is even more exaggerated in the preterm infant57. Impaired production of IL-12, IL-1β, and IL-23 in preterm dendritic cells likely adds to this skewing, as these cytokines are key for commitment toward the Th1 (IL-12) and Th17 (IL-1β and IL-23) cell pathways150,155. Reduced representation of the Th17 system may contribute to the high susceptibility of preterm infants to mucosal infections caused by Staphylococcus and Candida species156. Both the complement system and APPs show GA-dependent maturation; thus, preterm infants are deficient in both of these immune factors. This has implications for impaired opsonisation processes, which mark pathogens for digestion, as well as TLR-induced immune response pathways, which involve the release of APPs17,115. TLR responses display unbalanced development, with certain TLR responses weaker than others in the preterm infant: TLR7/8 responses are the most mature in preterm infants and TLR5 responses are the most immature, 23  with more heterogeneous TLR4 responses157. This asynchronous progression in TLR maturation may influence what types of microbes are most likely to infect preterm infants. Neutrophils are the last WBC population to appear in fetal circulation, so preterm infants often have very low neutrophil counts relative to term infants55. Instead, lymphocytes are the dominant WBC population in preterm blood158. Not only are neutrophils low in number, but they are even more inefficient than neutrophils of term infants in terms of neutrophil rolling and adhesion57. Ineffective opsonisation in preterm immunity, due to the limited complement factors and Igs described above, also impairs neutrophil phagocytic activity17. Finally, the preterm immune system is marked by an increased proportion of nRBCs31,55. These nRBCs may be functionally different from the nRBCs of term neonates, since a larger fraction of them are derived from the liver rather than the bone marrow159. Therefore, arginine depletion by erythroid cells may also contribute to the preterm neonate’s increased immunosuppressive profile87,125.   1.5 DNA methylation Epigenetics refers to modifications to DNA or the proteins around which it is bound that do not change the genetic sequence. These include DNAm and covalent histone modifications at specific residues, such as acetylation, methylation, phosphorylation and ubiquitination. Epigenetic marks can affect chromatin structure, thus influencing DNA-protein interactions and the potential for gene expression. DNAm is the covalent addition of a methyl group to a cytosine base, which generally occurs in the context of one of the 28 million cytosine-guanine dinucleotides (CpGs) in the human genome. Cytosines in non-CpGs can also be methylated, although these are significantly 24  rarer160. CpGs are generally depleted in the genome, due to the tendency for methylated cytosines to be deaminated to thymines. Exceptions to this are CpG islands (CGIs), high CpG density regions often associated with gene promoters that tend to be unmethylated and thus less prone to deamination161,162. DNAm is established de novo by DNA methyltransferases DNMT3A and DNMT3B, and maintained by DNMT18. In general, DNAm is mitotically heritable, with the methylation patterns in a parent cell getting passed on to the daughter cells. However, there are two discrete periods of genome-wide DNAm erasure: active demethylation in primordial germ cells, and passive loss in the pre-implantation embryo. This loss of methylation removes epigenetic marks of differentiation, producing pluripotent cells for embryonic development. Parent-of-origin-specific epigenetic marks are also removed, with the exception of imprinted differentially methylated regions (DMRs) that are established de novo in the germline or after fertilization and retained in the embryo4,161.  1.5.1 Relationship with gene expression By impacting chromatin structure and accessibility to transcription factors, DNAm can influence a gene’s potential for expression. DNAm is classically thought of as a marker of transcriptional repression at CGIs associated with gene promoters, with the methyl group of modified cytosines extending into the major groove of DNA to both alter chromatin structure and prevent transcription factor binding4. However, there is increasing evidence that the relationship between DNAm and gene expression extends beyond classical CGI promoters. For example, DNAm at the CpG shores that stretch 2 kb on either side of CGIs has been more strongly associated with hematopoietic cell-specific gene expression and myeloid-lymphoid lineage 25  decisions compared to DNAm at CGIs3. CpG shores show the greatest variation in DNAm between cell types and in cancer8. Additionally, CpG sites within CpG shelves, which extend another 2 kb beyond CpG shores, and the “open sea” are gaining more interest as potential distal regulatory regions, particularly when they are also associated with enhancer activity8,163,164. Epigenetic studies have also revealed the significance of intergenic DNAm, with a far greater proportion of cell-specific DNAm occurring in gene bodies than in CGI promoters162,165-167. This intergenic DNAm has been postulated to regulate cell-specific alternative transcripts, and limit transcriptional noise167-171. Exons are more densely methylated than introns, with the exception of first exons, which are relatively unmethylated. This has led some to suggest that DNAm closer to the transcription start site (TSS) (for example, in the first exon) blocks transcription whereas DNAm in the gene body relates to the level of gene expression172. The tentative current model of the impact of DNAm on gene expression is that, in CpG-dense promoters, methylation represses transcription. In contrast, CpG-poor regions do not have as clear of a relationship with DNAm, and genomic context needs to be considered to better understand the influence of DNAm on gene expression at these sites173. Additionally, differentially methylated regions (DMRs) often have a clearer function when considered in the context of gene regulatory sites, such as transcription factor binding sites and enhancers, rather than the genes themselves160,173. Although DNAm is often associated with gene expression, distinguishing cause and effect is difficult. That is, is it the presence of DNAm that blocks gene expression, or are genes that are rarely expressed more likely to gain DNAm? Or does the binding of transcription factors during active gene expression open the chromatin and facilitate DNAm loss? Regardless, DNAm studies are useful to uncover markers of genes that may be differentially expressed in a condition 26  of interest, and findings from these studies can highlight molecular pathways for further investigation.  1.5.2 Hematopoietic lineage commitment DNAm shows some of the greatest biological differences with cell and tissue type. Through its influence on chromatin structure and gene expression, DNAm can mediate development and maintenance of cell identity. Constitutive DNAm marks in HSCs that are required for their self-renewal abilities have been identified174. As HSCs differentiate, their unique DNAm profile is lost: genes associated with multipotency gain DNAm and lineage-specific genes are demethylated, presumably to activate genes associated with cell-specific functions4. DNAm appears to play the biggest role in lineage commitment and gene expression changes early in differentiation, particularly at intragenic CGIs169. On a global scale, it has been found that DNAm generally increases with lymphoid commitment, but decreases with myeloid commitment3,8.  1.5.3 Epigenetics and hematopoietic cell function Not only are epigenetic marks involved in hematopoietic cell lineage commitment, but they are also associated with hematopoietic cell function. For example, the regulation of lineage-specific transcription factors and effectors during Th cell differentiation is thought to be guided by epigenetic marks. The best-studied of these are located in cytokine locus control regions (LCRs), clusters of cytokine genes that are regulated in a coordinated fashion for Th-specific expression. Once activated by T cell receptor binding, permissive histone marks – such as H3 and H4 acetylation and H3K4 trimethylation – accumulate in promoters and enhancers of the 27  appropriate Th lineage-specific cytokine genes175. In cytokine genes associated with alternative Th lineages, repressive marks – such as H3K27 trimethylation – are gained while low-level permissive epigenetic marks of the naïve T cell state are lost175. In mice, decreases in DNAm have been observed in lineage-specific genes during commitment to Th2 (IL-4), Th1 (Il18r1 and Ifng), and Treg (foxp3) pathways176-179. The importance of epigenetic marks at these LCRs for Th cell function is shown by the aberrant cytokine gene expression in Th cells treated with inhibitors of DNAm and of histone deacetylation180,181. Although the majority of these studies were done in mice, the Th cytokine LCRs are conserved across mammalian genomes; thus, similar epigenetic processes likely also occur with human Th cell differentiation. Another role of epigenetics in hematopoietic cell function is in the establishment and maintenance of trained immunity. For example, after infection with cytomegalovirus, murine NK cells become resident and display rapid expansion, degranulation, and cytokine release upon reinfection182. These primed NK cells can also self-renew, maintaining their immunological training for several months after the initial infection183. Since epigenetic marks allow infection-induced gene regulation changes to be stably inherited to subsequent cell generations without the need for reinfection, they are thought to be a key component of trained immunity. This is supported by studies of histone mark changes in macrophages with trained immunity, which identified de novo enhancers that develop after infection and likely allow for faster transcriptional responses to subsequent activation184,185. Direct connections between epigenetic marks and trained immunity have been shown in monocytes. BCG vaccination in humans was found to induce a variety of expression changes in monocytes that persisted for over 3 months after vaccination, including increased surface expression of activation markers CD11b and TLR4 as well as elevated production of pro-28  inflammatory cytokines TNF and IL-1β in response to a variety of bacterial and fungal pathogens186. This monocyte priming occurred alongside significantly increased H3K4 trimethylation in the promoters of TNFα and IL6186. Treatment with a histone methyltransferase inhibitor prevented BCG-induced training, suggesting that these histone modifications are key for maintenance of innate immunological training186. Studies on the epigenetic influence of immune cell activation have focused mostly on histone modifications, since DNAm changes are not typically thought of as rapid. However, a recent study evaluating DNAm in dendritic cells found active demethylation upon M. tuberculosis infection187. Moreover, a significant proportion of these DNAm changes overlapped with infection-induced changes in chromatin state, including transitions from heterochromatin to de novo enhancers187. These findings suggest that both histone marks and DNAm reflect functional immune changes caused by infection.  1.5.4 DNA methylation and gestational age As the fetus develops, many distinct tissues and organs develop within a short time frame. DNAm changes associated with GA have been identified that reflect both the histological changes associated with tissue development as well as epigenetic changes occurring with cell maturation4. This means that DNAm measurements from a fetus or neonate are influenced by GA. Indeed, CpG sites that show GA-specific DNAm patterns have been identified in several genome-wide methylome studies of cord blood and of placental tissue188-190. PTB-associated methylation has also been identified in cord blood; some of these findings were associated with GA, and others were independent of it190-192.  29  1.5.5 Array-based measurements of DNA methylation A widely-used method of measuring genome-wide DNAm is the Illumina Infinium HumanMethylation450 BeadChip (450K array), which interrogates over 485,000 of the 28 million CpG sites in the genome. This array uses Illumina single nucleotide polymorphism (SNP) genotyping on DNA treated with sodium bisulphite, which converts unmethylated cytosines into thymines but leaves methylated cytosines unchanged. The 450K array is a popular tool for epigenetic studies because it provides a global, albeit low-coverage, perspective of the methylome while being relatively inexpensive in comparison to large-scale sequencing methods. It also offers good reproducibility, with technical replicates showing β-value correlations of 0.992 within the 450K array and 0.95 with β-values measured by whole genome bisulphite sequencing193. Although the 450K array’s coverage is genome-wide, it is not random: it was designed with particular focus on Reference Sequence (RefSeq) genes (National Center for Biotechnology Information, Bethesda, MD, USA) and CpG islands, of which it represents 99% and 96%, respectively. There is also considerable representation of CpG shores (92%), which flank CGIs, and CpG shelves (88%), which extend beyond CpG shores193.  1.5.5.1 Considerations for 450K array studies As the 450K array increased in popularity, important findings on the biology of DNAm as well as the technology itself highlighted the importance of careful study design in microarray studies. Sex, ethnicity, genetic variability, and age (including GA) have associated DNAm patterns, and thus have been identified as confounding variables in DNAm studies188-190. However, an even greater concern is cell type, and the presence of many different cell types in 30  heterogeneous tissue samples. Since each cell type has its own epigenetic signature, the proportion of different cells influences overall DNAm measured in whole tissue194. Ways to work around these confounding variables include matching samples by these variables, including them as covariates during data analysis, or, in the case of cell composition, by correcting for variability using deconvolution algorithms. The latter approach has become very popular in studies of adult blood, and reference-free methods have been developed that could, in theory, be used on any tissue type195-197. There are also technical variables to consider when implementing a study using the 450K array (Figure 1.4). The 450K array targets CpG sites with 50-base-long probes adhered to beads, which are randomly arranged on the array. The array contains two different probe types, with two different chemistries. Type I probes, which come from the 450K array’s predecessor, have paired methylated and unmethylated probes, for which the presence of a fluorescent signal (but not its colour) indicates methylation. Type II probes, which have a chemistry developed for the 450K array, have only a single probe, where methylation is inferred from the green signal and lack of methylation is inferred from the red signal193,198,199. Not only do these probes differ in signal interpretation, but they also have biological differences, with type I probes able to cover more CpG-dense regions than type II198. A variety of normalization methods have been developed to account for these probe type differences; however, there is no consensus on which normalization method to use, nor any set rules on how to evaluate data processing methods199-202. After the 450K array’s release, independent DNAm researchers identified probes with design flaws such as SNPs at the targeted CpG site, or a tendency to cross-hybridize to other regions of the genome203. At these poorly-designed probes, DNAm measurements may be 31  influenced by the subject’s genotype or by DNAm at multiple CpG sites in the genome, respectively. The conservative analytic approach is to filter these probes from the dataset. The 450K array is organized as a chip, roughly the dimensions of a microscope slide, on which twelve DNA samples can be loaded. Since the maximum number of chips that can be analyzed at one time is eight, the maximum number of samples that can be analyzed in one “batch” of the 450K array is ninety-six. This protocol introduces three technical variables: batch, chip, and position (referring to which of the twelve spots on the chip the sample is loaded onto) (Figure 1.4). Computational methods, including surrogate variable analysis (SVA) and ComBat, have been developed to adjust for this technical variability in existing data204,205. However, recent findings have shown that using batch correction methods can introduce false positives into DNAm data, resulting in overestimation of the effect of interest206,207. Ultimately, the best way to eliminate batch effects is careful consideration of the research question and data analysis before collecting measurements, and designing 450K array preparation accordingly.  32   Figure 1.4 Technical considerations for the Illumina 450K array. (A) The 450K array has two distinct probe chemistries that target different regions of the genome and have different interpretation of fluorescent signals. Adapted from Bibikova, 2011193. (B) Batch, chip, and position are technical variables in 450K array studies.   33  1.6 Research objectives The purpose of this thesis is to establish DNAm profiles of the major cord blood cell populations from both term and preterm births. An epigenetic basis for altered neonatal immune function will be evaluated by investigating how these DNAm profiles differ both between cell types and across GA. I hypothesize that changes in DNAm observed across GA in cord blood are due to a combination of (1) cell-specific DNAm and progressive changes in cell composition; and (2) epigenetic maturation within cell populations. My thesis has three main objectives: 1. To evaluate the impact of nRBCs on standard fluorescence-activated cell sorting (FACS) techniques to isolate WBCs from cord blood, and to modify the FACS protocol to account for this unique cord blood cell population. 2. Establish a term nRBC DNAm profile by identifying how nRBC DNAm differs from major WBC populations (granulocytes, monocytes, CD4 and CD8 T cells, B cells, and NK cells) isolated from term cord blood. 3. Develop a cell-specific understanding of how DNAm in major cord blood cell populations (granulocytes, monocytes, T cells and nRBCs) differs between preterm and term birth.   34  Chapter 2: Nucleated Red Blood Cell Contamination during Fluorescence-Activated Cell Sorting Impacts Epigenetic and Gene Expression Analyses of Cord Blood CellsA 2.1 Background With the increased accessibility of high throughput technologies for epigenetic and gene expression studies, genome-wide approaches have gained popularity in studies of hematopoietic cell lineage relationships3,12,208,209. However, although genome-wide profiling of isolated blood cells can provide a large amount of information, data interpretation is notoriously difficult in mixed cell populations194,210,211. To address this issue, studies can be performed either on homogeneous cell populations, or on mixed cell samples with deconvolution algorithms applied to correct for differences in cell composition195,197. One concern with the former approach in blood is that RBCs have been shown to engage in functional heterotopic interactions with other hematopoietic cells90-93,96,212. If not formally excluded using lineage markers, these interactions could impact whole genome studies of hematopoietic cells sorted by FACS, particularly in cord blood which has a notable proportion of nRBCs.  The proportion of nRBCs in cord blood varies considerably between individuals. Typically these cells represent only a few percent of the total nucleated cell count; however, they can comprise up to 50% of all nucleated cells in some chronic hypoxic-ischemic-related pregnancy situations31,69. nRBCs are generally resistant to lysis protocols and tend to sediment in                                                  A A version of this chapter has been published as: de Goede, O.M., Razzaghian, H.R., Price, E.M., Jones, M.J., Kobor, M.S., Robinson, W.P., and Lavoie, P.M. (2015) Nucleated red blood cells impact DNA methylation and expression analyses of cord blood hematopoietic cells. Clinical Epigenetics. 7(1): 95. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium. 35  the mononuclear cell fraction during purification by density gradient centrifugation, further complicating the isolation of pure hematopoietic cell populations213. Depending on their proportion, the presence of nRBCs could complicate both epigenetic and gene expression studies. Under non-pathological conditions, DNAm shows great biological differences with tissue and cell type. Clustering of adult blood cells based on their DNAm profiles is consistent with the classical model of hematopoietic lineage relationships160,194,197. However, our initial analysis of genome-wide DNAm in cord blood cell populations isolated by FACS suggested significant cross-contamination between cell types. We observed low incidence WBC heterotopic interactions with nRBCs that were undetected by traditional singlet FACS gating, due to the small size of nRBCs. To obtain pure WBC populations, we developed and implemented a stringent sorting protocol that excludes erythroid-specific surface markers. The DNAm profiles of cell populations obtained by our stringent FACS method were used 1) to evaluate the impact of nRBC contamination on the DNAm profiles of T cells and monocytes; and 2) to identify nRBC-distinct DNAm markers to detect erythroid contamination in genome-wide DNAm studies.  2.2 Methods 2.2.1 Sample collection and cell purification Ethics approval for this study was obtained from the University of British Columbia Children’s and Women’s Research Ethics Board (certificate numbers H07-02681 and H04-70488). Written, informed parental consent to participate was obtained. Individual patient data is not reported. 36  Cord blood was collected from term neonates (38-40 weeks GA) delivered by elective caesarean section in absence of labor at the Children’s and Women’s Health Centre of BC (Vancouver, BC, Canada). B cells, T cells, monocytes, NK cells and nRBCs were purified by FACS as described in Appendix Section A.1. Two sorting protocols were compared, which are referred to as the “standard” and “stringent” protocols. The stringent method includes additional negative gating steps, mainly for erythroid lineage-specific cell surface protein markers (Figure 2.1, Supplementary Table A.1). For DNAm studies, cells were sorted from a total of 12 subjects as described in Appendix Section A.1. Whole (CD3+) T cells, monocytes, and nRBCs were collected from 5 individuals by the standard sorting method; B cells, CD4 T cells, CD8 T cells, granulocytes, monocytes, NK cells, and nRBCs were collected from 7 individuals by the stringent sorting method. Granulocytes were collected by hypotonic lysis as described in Appendix Section A.1. For genome-wide gene expression analysis, naïve CD4 T cells were sorted from 12 additional subjects. Naïve T cells were sorted according to the following parameters: CD3+/CD4+/CD25- /CD45RO-/CCR7+ for the standard protocol, plus CD235- for the stringent protocol.  37   Figure 2.1 Schematic representation of cell sorting strategies. (A) The standard cell sorting strategy used to purify whole (CD3+) T cells, nRBCs, and monocytes by FACS. (B) The stringent cell sorting strategy used to purify CD4 and CD8 T cells, B cells, nRBCs, monocytes, and NK cells by FACS.   38  2.2.2 RNA extraction and genome-wide expression profiling Total RNA was extracted from the samples using QIAshredder columns and RNeasy Mini Kit (both Qiagen, MD, USA). RNA samples were cleaned using RNA Clean & Concentrator kit (Cedarlane, ON, Canada). Their quantities were measured with a NanoDrop spectrophotometer (Thermo Fisher Scientific, DE, USA) and sample integrity was evaluated using Agilent RNA 6000 Nano kit and Agilent 2100 Bioanalyzer (Agilent, CA, USA). The RNA samples were then hybridized to the Illumina (CA, USA) HumanHT-12_v4_BeadChip array following manufacturer’s protocols. The resulting data were transferred to GenomeStudio (Illumina), then further processed and normalized using the lumi package214 in R software215. Any gene probes with signal intensity <100 were considered background expression and removed from analysis, for a final dataset of 20,876 probes. Average log2(expression) for each gene in T cells collected by the standard sorting strategy was compared to average log2(expression) for each gene in T cells collected by the stringent sorting strategy.  2.2.3 Illumina Infinium HumanMethylation450 BeadChip DNA was extracted from isolated cell populations using standard protocols and purified with the DNeasy Blood and Tissue kit (Qiagen). DNA was bisulphite-converted using the EZ DNA Methylation Kit (Zymo Research, CA, USA) before amplification and hybridization to the 450K array (Illumina) following manufacturer’s protocols. 450K array chips were scanned with a HiScan reader (Illumina). Raw intensity data were background normalized in GenomeStudio (Illumina). Quality control was performed using the 835 control probes included in the array. The intensity data were then exported from GenomeStudio and converted into M values using the lumi 39  package214in R software215. Sample identity and quality were evaluated and probe filtering was performed as described in Appendix Section A.1, to identify and remove one NK cell sample as an outlier and produce a final dataset of 440,315 CpG sites. Background intensity and red-green color bias were corrected for using the lumi package, and the data were normalized by subset within-array quantile normalization201.  2.2.4 DNA methylation data analysis Since the stringent FACS strategy was designed based on results from the standard FACS strategy, sample collection and 450K array runs for these two protocols were done separately. To avoid confounding by batch effects, DNAm analyses were also performed separately for the data from each FACS protocol. My analytic approach was to compare cell types sorted by the same FACS protocol to each other, and then to evaluate whether a given cell type’s epigenetic relationship with the other cell types changed between FACS methods. By making comparisons between cell populations derived from the same set of individuals, I reduced DNAm differences that can arise due to genetic effects. For the standard sorting method, DNAm data were available for nRBCs, monocytes, and T cells from 5 individuals at 440,315 sites after pre-processing. Unsupervised Euclidean clustering of the samples based on DNAm β values was performed as an initial global analysis step. Differential DNAm between each blood cell pairing was tested by linear modeling through the R package limma216. SVA using the R package sva204 was performed to account for unwanted variability in the linear modeling. Surrogate variables (SVs) were used as covariates in the model, with cell type as the main effect. Resulting p-values were adjusted for multiple comparisons by the Benjamini & Hochberg217 false discovery rate (FDR) method, and 40  statistically significant sites were limited to those that passed an FDR <5%. SV-corrected data was used for DNAm-based filtering of the statistically significant sites. At each site, a between-group difference in DNAm (Δβ) was calculated by subtracting mean DNAm for one cell type from the other. Differentially methylated (DM) sites were considered as those having both an FDR <5% and |Δβ|>0.20. For the stringent sorting method, DNAm data were available for B cells, CD4 T cells, CD8 T cells, granulocytes, monocytes, NK cells, and nRBCs from 7 individuals at 440,315 CpG sites after pre-processing. To analyze the data in a comparable way to the standard FACS protocol, only CD4 T cells, monocytes, and nRBCs were considered. The DNAm profiles of these cell populations were analyzed as described for the standard sorting protocol. To identify DNAm markers specific to nRBCs, data from the stringent sorting method for all 7 cell types were used. DM sites between nRBCs and every other cell type were detected by linear modeling with nRBCs as the reference cell type and SVs included as covariates. Significantly DM sites were defined as those with a FDR <5% and a |Δβ|>0.50. Finally, to evaluate the relationship between nRBC proportion in whole cord blood and DNAm of nRBCs, the SV-corrected M values for the 7 nRBC samples collected by stringent FACS methods were used. Linear modeling was performed with nRBC proportion (as measured by number of nRBCs/100 WBCs in whole blood) as the main effect and no covariates.  2.3 Results and discussion 2.3.1 Heterotopic cell interactions impact genome-wide signatures of hematopoietic cells Heterotopic cell interactions are a well-known occurrence that may confound cell-specific studies. To avoid this problem, cell doublets are generally excluded during FACS by 41  employing forward/side scatter-width singlet gatings. Despite applying these standard criteria, we observed small proportions of double positive events co-expressing both T cell- (CD3) and erythroid- (CD235) specific lineage markers. When analyzed by flow cytometry after sorting, these double positive events were found to be distinct cell events expressing either erythroid or T cell markers (Figure 2.2A). Examination of these sorted events under light microscopy confirmed that they were T cell-RBC doublets. To a lesser extent, we also detected events positive for expression of both erythroid (CD235) and monocyte (CD14) or B cell (CD19) markers (not shown). These findings indicate that heterotopic RBC-to-WBC doublets can be undetected by FACS using conventional singlet gating. 42   Figure 2.2 DNAm profiles of cord blood cells isolated by the standard FACS strategy. (A) A CD14-/CD19-/CD3+/CD235+ population isolated by FACS (left panel) is revealed to be T cell/RBC doublets by flow cytometry, which identifies two distinct cell types after sorting (right panel). (B) Unsupervised Euclidean clustering of genome-wide DNAm (440,315 CpG sites) between whole T cells (CD3T), nRBCs, and monocytes (Mo); numbers in the sample labels indicate different cord blood donors. (C) Number of large magnitude DM sites (FDR <5%, |Δβ| >0.20) between nRBCs and T cells, nRBCs and monocytes, and T cells and monocytes sorted using a standard approach. (D) DNAm heat map of nRBCs, T cells, and monocytes at top nRBC-DM sites identified by the standard sorting protocol (FDR <5%, |Δβ| >0.30; 457 CpG sites).  We assessed the impact of these rare heterotopic RBC-to-WBC interactions on genome-wide DNAm and gene expression analyses of hematopoietic cord blood cells. First, DNAm data 43  from T cells, monocytes, and nRBCs sorted without formal exclusion for an erythroid cell lineage marker were evaluated. Unsupervised Euclidean clustering of array-wide DNAm showed unexpected clustering of nRBCs with T cells, and monocytes as the most epigenetically distinct population (Figure 2.2B). When these cell populations were compared to identify differentially methylated (DM) sites (FDR<5%, |Δβ| >0.20), nRBCs versus T cells had fewer DM sites (3,538) than either nRBCs versus monocytes (12,852) or T cells versus monocytes (18,738) (Figure 2.2C). Even at their largest-magnitude DM sites, nRBCs sorted by the standard FACS strategy often displayed DNAm values intermediate to the DNAm of monocytes and T cells (Figure 2.2D). This is unusual, since exemplar cell-specific DM sites are typically either fully methylated (β >0.80) or unmethylated (β <0.20), with comparison cells exhibit opposing levels of DNAm194,218. Next, whole-genome expression data were compared between naïve CD4 T cells sorted by the standard FACS method and naïve CD4 T cells sorted by the stringent FACS method. We observed a high expression of hemoglobin genes in T cells sorted by the standard protocol, but not in T cells sorted by the stringent protocol (Figure 2.3). Finally, we examined gene expression datasets of hematopoietic cells publicly available in the Gene Expression Omnibus. Expression of hemoglobin genes was high in the majority of cord blood hematopoietic cell datasets, indicating contamination of these published datasets (Supplementary Figure A.1). However, increased hemoglobin gene expression was not observed in hematopoietic cells collected from adult blood. Together, these DNAm and gene expression findings suggest that heterotopic cell interactions, though a rare occurrence, can significantly impact genome-wide molecular signatures of cord blood hematopoietic cells from cord blood.  44   Figure 2.3 Genome-wide transcriptomic profiles of naïve CD4 T cells sorted by either the standard or stringent FACS strategy. Log2(expression) of 20,876 gene probes in naïve CD4 T cells sorted by either a standard (no erythroid exclusion) or stringent (exclusion of erythroid-specific surface markers) FACS strategy.  2.3.2 Revised DNA methylation profiles of hematopoietic cells obtained by a more stringent cell sorting strategy When employing a stringent sorting strategy that formally excludes RBCs, the DNAm relationships between cord blood T cells, monocytes and nRBCs were more consistent with previous hematopoietic lineage studies3,4,174,219,220. Unsupervised Euclidean clustering by array-wide DNAm showed that nRBCs were epigenetically closer to the myeloid lineage (monocytes) than to the lymphoid lineage (T cells) following this stringent sorting approach (Figure 2.4A). Additionally, each hematpoietic population was more epigenetically distinct, as reflected by both principal component analysis (PCA) (Supplementary Figure A.2) and the greater number of DM sites for each cell type comparison following stringent sorting (24,263 for nRBCs versus T cells; 12,980 for nRBCs versus monocytes; 19,278 for T cells versus monocytes; Figure 2.4B) 45  compared to standard sorting (Figure 2.2C). CD4 T cells and nRBCs sorted by the stringent protocol showed a greater number of cell-specific DM sites than whole (CD3+) T cells and nRBCs sorted by the standard protocol (Table 2.1). In contrast, monocytes sorted by the stringent protocol showed fewer DM sites, likely due to the DNAm profile of nRBCs becoming more similar to monocytes after stringent cell sorting (Figure 2.4A).   46   Figure 2.4 DNAm profiles of cord blood cells isolated using the stringent FACS strategy. (A) Unsupervised Euclidean clustering of genome-wide DNAm (440,315 CpG sites) in CD4 and CD8 T cells (CD4T and CD8T, respectively), monocytes (Mo), and nRBCs isolated by a stringent FACS protocol; letters in the sample labels indicate different cord blood donors. (B) Number of large magnitude DM sites (FDR <5%, |Δβ| >0.20) between nRBCs and CD4 T cells, nRBCs and monocytes, and CD4 T cells and monocytes sorted using a stringent approach. (C-E) Overlap of cell-specific DM sites (FDR <5%, |Δβ| >0.20) identified in the standard versus stringent sorting protocols for (C) T cells, (D) monocytes, and (E) nRBCs.  Top DM sites for each cell type (FDR <5%, |Δβ| >0.20) were then compared between the two sorting protocols. For T cells, the majority of DM sites (>98%) discovered by the standard method overlapped with the DM sites identified by the stringent protocol (Figure 2.4C). A 47  notable percentage (47%) of monocyte-DM sites found by the standard protocol were also discovered by the stringent protocol (Figure 2.4D). For nRBCs, the DM sites identified by the two protocols showed the least overlap (36%), with the stringent protocol identifying far more nRBC-DM sites than the standard protocol (8,982 versus 2,338) (Figure 2.4E). Of the 8,982 stringent nRBC-DM sites, six were located in the hemoglobin genes we found to be highly expressed in cord blood WBCs sorted by a standard protocol (and thus presumed to be contaminated with RBCs) (Figure 2.3; Supplementary Table A.2). These genes were also reported as highly expressed in publicly available datasets of cord blood WBCs, indicating widespread erythroid contamination (Supplementary Figure A.1). The DNAm differences at these loci were striking, with the mean nRBC DNAm up to 43 percentage points less than the mean DNAm for all WBCs. Several of these CpG sites are located in either the body of the associated hemoglobin gene or within 300 bases upstream of its transcriptional start site, and may be associated with erythroid-specific gene expression.    48  Table 2.1 Number of cell-specific DM CpG sites (FDR <5%) following the standard and stringent FACS strategies. Minimum |Δβ| nRBCs Monocytes T cells Standard FACS Stringent FACS Standard FACS Stringent FACS Standard FACS Stringent FACS NA 61,405 197,237 41,451 80,600 39,284 111,965 0.05 38,295 144,949 32,558 32,846 23,060 50,621 0.10 11,848 63,099 22,829 14,864 10,868 27,420 0.20 2,338 8,982 11,855 5,940 2,602 12,014 0.30 457 2,628 6,486 3,162 879 5,474 0.40 17 648 3,520 1,757 292 2,600 0.50 0 41 1,884 878 48 1,268 0.60 0 1 908 319 2 553 0.70 0 0 255 51 0 158 0.80 0 0 3 1 0 12 Cell-type comparisons were made between monocytes, whole (CD3+) T cells, and nRBCs sorted by the standard FACS strategy; and between monocytes, CD4 T cells, and nRBCs sorted by the stringent FACS strategy.  The top DM sites from the stringent protocol represent sites with the strongest cell-specific DNAm patterns (8,982 nRBC-DM sites, 12,014 CD4 T cell-DM sites, and 5,940 monocyte-DM sites). Thus, I used these sites to confirm heterotopic cell interactions in the standard protocol. The distributions of DNAm values for each cell type-by-protocol combination show a defined shift in nRBC DNAm between sorting methods (Figure 2.5). When sorted by the standard method, nRBC DNAm was more similar to the DNAm patterns of T cells. The exclusion of other hematopoietic lineages in the stringent sorting of nRBCs dramatically decreased nRBC DNAm, suggesting a cleaner population of these cells. In contrast, the impact of sorting protocol on DNAm profiles of monocytes and T cells was modest. 49   Figure 2.5 DNAm changes with FACS strategy in nRBCs, T cells and monocytes at their top DM sites. (A) 8,982 nRBC-DM sites. (B) 12,014 CD4 T cell-DM sites. (C) 5,940 monocyte-DM sites. Solid lines represent group mean, dashed lines represent individual samples.  50  To further evaluate how sorting strategy affected cell type epigenetic profiles, I looked at discordant sites: sites that were DM in one sorting protocol, but not in the other. In nRBCs, differential DNAm unique to the standard protocol was observed at 1,505 sites, while 8,149 sites were uniquely-DM in the stringent protocol (Figure 2.4E). An example nRBC-discordant site is provided in Figure 2.6A: a CpG site in BCL11B shows nRBC DNAm trending towards T cell levels in the standard FACS protocol, but exhibiting DNAm similar to other non-T cells in the stringent FACS protocol. In contrast to nRBCs, monocytes sorted by the stringent protocol had few DM sites that were not also identified in the standard protocol (Figure 2.4D). Unlike nRBC-discordant sites, there appeared to be multiple reasons for monocyte-discordant sites. At some of these sites, absolute DNAm in monocytes did not change significantly between the two sorting protocols, but the change in nRBC DNAm with stringent sorting impacted the detection of differential DNAm when compared to monocytes (Figure 2.6B). For other sites, DNAm differences were noted between protocols for all three cell types and may be attributable to technical noise or genetic differences between the different set of subjects for each sorting method. In fact, a few of discordant sites were clearly “epipolymorphisms”, in which changes in DNAm levels were associated with individuals rather than cell types221; this resulted in highly variable DNAm patterns within a cell type (Figure 2.6C). 51   Figure 2.6 Selected discordant DM sites between the standard and stringent FACS protocols. At these discordant CpG sites, a given cell type is DM only in one protocol, but not the other. (A) An example CpG site illustrating contamination of nRBCs with T cells after sorting by standard FACS methods. The nRBCs trend toward T cell DNAm in the standard method, but are hypermethylated (like all other non-T cells) after sorting by the stringent method. (B) Example of a discordant site in monocytes due to heterotopic cell interactions. (C) Example of a discordant site in monocytes due to a likely epipolymorphism. This was confirmed by comparing DNAm plotted by cell type (left boxplots) to plotting by individual (right boxplots). DNAm in each individual shows one of three distinct levels of DNAm, presumably depending on the individual’s genotype at the site influencing DNAm. 52  Comparing the cell-specific DM sites discovered by each sorting protocol further supports our hypothesis of heterotopic cell doublet contamination in the standard protocol. This contamination appears to have a much greater effect on nRBC DNAm than on T cell or monocyte DNAm. We attribute this to the relatively low proportion (~5-10%) of RBCs that was nucleated in our samples (Figure 2.7). The impact of heterotopic contamination by WBCs on the DNAm profile of sorted nRBCs is more obvious than the reverse, since all WBCs are nucleated. However, since the proportion of nRBCs can be as high as 50% of all nucleated cells in cord blood31,69, we expect that cross-contamination during FACS can have a major impact on the DNAm profile of sorted WBCs in a subset of cases.   53   Figure 2.7 Erythroid-WBC interactions during FACS affect DNAm based on the proportion of nRBCs. As illustrated, only a fraction of erythroid cells are nucleated; therefore, contaminating nRBCs contribute less DNA than contaminating WBCs. Consequently, heterotopic cell interactions during FACS have greater weight on the DNAm profiles of erythroid cells than of WBCs (in this figure, T cells), which are all nucleated. This illustrates how the relative impact of cross-contamination during FACS on DNAm can be much greater for nRBCs than for WBCs, but WBC DNAm can also be strongly affected in a subset of cases where overall nRBC proportion is high.  2.3.3 Erythroid-specific differentially methylated sites To provide a way of evaluating the impact of nRBCs on cord blood cell DNAm profiles, I defined erythroid lineage-specific DNAm markers. DNAm of B cells, CD4 and CD8 T cells, granulocytes, monocytes, and NK cells sorted using the stringent FACS strategy were compared to stringently-sorted nRBCs. For each WBC population, over 210,000 of the 440,315 CpG sites analyzed showed statistically significantly different DNAm from nRBCs (FDR <5%). Eight of these DM sites, termed erythroid DNAm markers, were selected based on an average Δβ >0.50 54  between nRBCs and every other cell type (Table 2.2; Supplementary Figure A.3). I did not consider gene function in marker selection, as the intention is to use these erythroid DNAm markers to quickly detect nRBC contamination in cell samples using targeted methylation assays, such as pyrosequencing. At all eight CpG sites, nRBCs are less methylated than WBCs; thus if a sample is significantly contaminated, DNAm should be notably lower than reference WBC levels.  Table 2.2 Top eight CpG sites with nRBC-distinct DNAm from white blood cells in cord blood. 450K array CpG identifier CpG location: chromosome, closest gene Location in gene Mean nRBC β  (min., max.) Mean non-erythroid cell β (min., max.) cg05012676 16, ZFPM1 Intron 0.421  (0.298, 0.508) 0.937  (0.896, 0.963) cg06768361 12, TESC Intron; enhancer* 0.336 (0.211, 0.426) 0.924  (0.825, 0.978) cg10018933 2, HDAC4 Intron; enhancer** 0.410  (0.346, 0.468) 0.922  (0.890, 0.953) cg15974642 10, IFIT1B TSS200 0.362  (0.234, 0.475) 0.939  (0.904, 0.957) cg18168751 4, IDUA TSS1500; enhancer* 0.435  (0.356, 0.556) 0.951  (0.924, 0.969) cg20555305 8, CPSF1 Intron 0.369  (0.277, 0.481) 0.878  (0.823, 0.917) cg25105522 17, MAP3K14 Intron; enhancer** 0.224  (0.126, 0.300) 0.872  (0.750, 0.962) cg26876834 16, SNORA64 & RPS2 SNORA64: TSS1500 RPS2: Intron 0.369  (0.227, 0.440) 0.890  (0.839, 0.938) *Based on UCSC Genome Browser: ENCODE Enhancer- and promoter-associated histone mark (H3K4Me1) in the K562 cell line **Based on Illumina 450K array annotation   55  Some of these erythroid DNAm markers are associated with genes that have known erythropoietic function, such as ZFPM1 and HDAC4. The zinc finger protein ZFPM1 acts as a cofactor for GATA-1, a key transcription factor in erythroid differentiation222,223. HDACs directly associate with GATA-1 and HDAC4 expression is specifically reduced during erythroid maturation, likely being localized to the nucleus224. HDAC4 may be involved in the enucleation process of nRBCs: histone deacetylation is essential for heterochromatin formation, and condensed chromatin is a main requirement for enucleation and terminal erythroid differentiation68. Interestingly, other erythroid DNAm markers are near genes involved in immune functions, such as MAP3K14 and IFIT1B, consistent with the idea that nRBCs have an immunoregulatory role in early life. MAP3K14 induces NF-kappa-B signaling, a major inflammatory response pathway225. IFIT1 is typically silent in most cells, but becomes highly expressed in response to interferons, viral infection, and certain molecular patterns, with IFIT proteins having antiviral effects through binding and modulation of host and viral proteins and RNA226. As these erythroid DNAm marker sites are located largely in enhancer regions, reduced DNAm in nRBCs may reflect either specific upregulation of these genes in erythroid cells or a more primitive permissive state that is actively shut off in differentiation of other cell types. Although these erythroid DNAm markers are the top nRBC-DM sites, they display notable inter-individual variability in nRBC DNAm, with β-value standard deviations ranging from 0.048 to 0.091. I hypothesize that this variability in DNAm may be related to important inter-individual differences in nRBC function or maturation, based on the negative Pearson correlation observed between array-wide median nRBC DNAm and nRBC proportion (r =-0.86, p =0.013) (Supplementary Figure A.4A). Linear modeling identified 5,935 CpG sites significantly associated (FDR <5%) with nRBC proportion, including three of the eight CpG 56  sites identified as erythroid DNAm markers (Supplementary Figure A.4B-C). These results suggest that DNAm changes in cord blood nRBCs occur dynamically with nRBC production and maturation, thereby revealing an additional level of functional complexity to consider in whole-genome DNAm analyses of nRBCs.  2.4 Conclusion While nRBCs are generally absent or rare in adult blood, they are commonly present in low proportion in cord blood, with a higher nRBC count associated with a variety of maternal and fetal health factors. Our data show that nRBCs have a distinct DNAm profile, with an association between nRBC DNAm and overall nRBC proportion in cord blood. The complex DNAm profile of nRBCs has implications for epigenetic studies of whole cord blood and mononuclear cells, in which nRBCs have a demonstrable effect on DNAm (M.J.J. et al., manuscript in preparation). Despite the variability in nRBC DNAm at our identified erythroid DNAm markers, we believe that these sites will be sufficient to detect erythroid cells due to the low variation within all WBCs at these sites, as well as the large magnitude of DNAm difference between nRBCs and other cell types. Heterotopic interactions between erythroid cells and WBCs are likely biologically meaningful events, since RBCs have immune functions that require cell-to-cell contact with WBCs87,90-93,96. Additionally, RBCs adhere to macrophages to form erythroblastic islands during both fetal and adult RBC maturation227. Our data highlight the importance of formally excluding these interactions in lineage studies of cord blood hematopoietic cells using flow cytometry. This has major ramifications for the design of epigenetic, transcriptomic, and functional studies of cord blood cells. 57  Chapter 3: Characterizing the Hypomethylated DNA Methylation Profile of Nucleated Red Blood Cells from Cord Blood 3.1 Background In adults, nearly all RBCs are enucleate by the time they enter circulation; in contrast, many fetal circulating RBCs retain their nucleus, with the proportion of nRBCs in the fetus declining as gestation progresses31. nRBCs are often still present in low proportion in the neonate, with 0-10/100 WBCs considered the typical range31. However, increased nRBC levels have been associated with various pregnancy complications including infection, preeclampsia, maternal smoking, maternal obesity and diabetes31,69,72-74,228,229. The presence of nRBCs in fetal and neonatal circulation is often ascribed to high erythropoietic demand, but these cells may also impact immune function. Experimental data on human and mouse nRBCs shows that they are actively immunosuppressive93, and increased nRBC count is linked to increased risk of neonatal sepsis and perinatal brain damage230,231. Hematopoietic stem cells migrate to multiple anatomic locations during development. The bone marrow becomes the predominant source during the end of the third trimester, although other sites, including the placenta and liver, continue to serve as hematopoietic niches throughout development24,40. Since nRBCs are typically too large and non-deformable to pass through bone marrow fenestrations, their presence in fetal or adult circulation can be due to either extramedullary hematopoiesis or erythropoietin-induced changes in bone marrow structure and blood flow; both of these changes occur in response to hypoxia, infection, and other forms of stress31,70,76,77. 58  Under normal conditions, nRBCs disappear from circulation within 4 days in term infants and 7 days in preterm infants55. The persistence of nRBCs after birth is highly predictive of newborn morbidity and mortality71. Understanding the origin of nRBCs will be relevant to evaluating the likelihood that they will persist in the neonate. For example, I would expect that placental-derived nRBCs disappear more rapidly than those produced by a fetal organ, since connection to the placenta is lost after birth. DNAm has been well-studied in hematopoietic cell lineage commitment3,4,232. Generally, as cells commit to a specific lineage, DNAm decreases in appropriate lineage-specific genes while DNAm increases in genes associated with other blood cell lineages as well as the “stem-like” properties of self-renewal and multipotency4. nRBCs are an exception to this trend, displaying global demethylation as erythropoiesis progresses in both human and mouse66,67. This unusual epigenetic pattern may be because the enucleate RBC is the mature erythroid cell; thus, DNAm is not relevant to the fully mature state and may not be as tightly regulated as in other hematopoietic cells. However, DNAm may still be important for erythroid maturation. After observing that nRBCs are a distinctly hypomethylated cell population in cord blood (Section 2.3.2), I sought to establish a term nRBC methylome using the 450K array. DNAm in nRBCs was compared to CD4 T cells, CD8 T cells, NK cells, B cells, monocytes, and granulocytes. Low global DNAm has only been reported in gametes, early human embryos, placenta, and cancer233-236, but not in normal somatic tissues. I therefore hypothesized that hypomethylation of nRBCs could be due to: 1) retention of an early developmental signature, in which case nRBCs may be epigenetically similar to the placenta; or 2) a passive loss of DNAm during erythropoiesis. I compared patterns of DNAm between nRBCs and placenta to identify the origin of nRBC hypomethylation. These comparisons were performed on a range of genomic 59  regions including repetitive elements (REs), imprinted DMRs, and regions of high, intermediate, and low CpG density. I also show that DNAm in term nRBCs is associated with their proportion in cord blood, and may reflect shifting maturity or anatomic source of this cell population.  3.2 Methods 3.2.1 Sample collection and the Illumina Infinium HumanMethlyation450 BeadChip The blood cell samples used for this Chapter are the same ones collected for Chapter 2. Their collection is described as the “stringent sorting protocol” in Section 2.2.1 and Appendix Section A.1.  Whole chorionic villi were sampled from placentas of 5 term births (3 female, 2 male) delivered at the BC Women’s Hospital (Vancouver, Canada), for use in a previous study237. Villi were sampled from four sites on the fetal side of the placenta and rinsed of maternal blood, with the amniotic membrane and chorion removed.  DNA was extracted from all samples using standard protocols and purified with the DNeasy Blood and Tissue kit (Qiagen). For the villi samples, DNA from the multiple sampling sites of the same placenta were pooled in equal quantities to better represent the whole placenta. DNA was bisulphite-converted using the EZ DNA Methylation Kit (Zymo Research) before amplification and hybridization to the 450K array following manufacturer’s protocols (Illumina). 450K array chips were scanned with a HiScan reader (Illumina).  3.2.2 Combined blood cell and placental DNA methylation data analysis Raw intensity data for all blood cells, blood cell mixtures and placental whole villi were background normalized in GenomeStudio (Illumina). Quality control was performed using the 60  835 control probes included in the array. The intensity data were then exported from GenomeStudio and converted into M values using the lumi package214in R software215. Sample identity and quality were evaluated as described in Appendix Section B.1, and one NK cell sample was removed as an outlier. The 450K array targets 485,577 DNAm sites, but probe filtering was performed as described in Appendix Section B.1 to produce a final dataset of 431,767 sites. Background intensity and red-green color bias were corrected for using the lumi package214.  Unsupervised Euclidean clustering of the samples based on DNAm β values and PCA based on DNAm M values were performed as exploratory global analysis steps. DNAm was then evaluated at specific subsets of the 450K array, based on 1) CpG density; 2) location in REs; 3) previous association with age-related DNAm changes238; or 4) overlap with hematopoietic “source-DM sites” from Lessard et al.239, which display differential methylation between erythroblasts derived from fetal liver (FL) and erythroblasts derived from adult bone marrow (BM). The groups were as follows: 1) CpG sites in regions of high (142,380), intermediate-shore (30,549), intermediate (101,579) or low (157,259) CpG density; 2) CpG sites overlapping with long interspersed element 1 (LINE1) (7,591) or Alu (9,044) repetitive regions; 3) CpG sites linked to age-related DNAm changes238, separated by whether they increased (1,036) or decreased (1,803) with age; and 4) the top 100 CpG sites hypomethylated in adult BM erythroblasts (“BM-hypomethylated sites”) or hypomethylated in FL erythroblasts (“FL-hypomethylated sites”), based on Lessard et al.’s239 Δβ values. CpG sites located in known somatic and placenta-specific imprinted DMRs were also assessed240. DNAm at these CpG site groups were compared (as β-values) between all cell and tissue types using analysis of variance 61  (ANOVA), followed by Tukey’s honest significant difference test, using a multiple comparison-adjusted p-value significance threshold of 0.0045.  3.2.3 Blood cell only DNA methylation data analysis Raw intensity data for only the blood cell samples were prepared in GenomeStudio and converted into M values in R software as described in Section 3.2.2. The identity and quality of these samples had already been evaluated; the previously-identified outlier NK cell sample was removed in this dataset, and the same 431,767 CpG sites selected in the blood and placenta dataset were assessed. Background intensity and red-green color bias were corrected for using the lumi package214, the data were normalized by subset within-array quantile normalization201, and the technical variables of 450K array chip and row were adjusted for with ComBat205. To assess if the relationship between nRBC proportion in whole cord blood (as measured by number of nRBCs/100 WBCs) and DNAm was exclusive to nRBCs, I performed linear modelling using the R package limma216. Cell type and the interaction term of cell type and nRBC count were variables of interest, with sex included in the model as a covariate. Resulting p-values were adjusted for multiple comparisons by the Benjamini & Hochberg217 FDR method, and statistically significant sites were limited to those that passed an FDR <5%. Then, nRBC DNAm β-values were evaluated as a function of nRBC proportion in whole cord blood at the subsets of the genome described in Section 3.2.2. I next looked for nRBC DMRs associated with nRBC count and with cell type using the R package DMRcate241. The package reduced the dataset to 426,732 probes, filtering out any probes with a SNP (minor allele frequency ≥0.05) within 2 nucleotides of the CpG site. The model included cell type and the interaction term of cell type and nRBC count as variables of 62  interest, with sex as a covariate. Since each cell type was collected from the same set of individuals, DNAm measurements may be influenced by inter-individual differences. To adjust for this, the model included a within-individual consensus correlation estimated using the duplicateCorrelation() function in limma216. With this model, two types of DMRs were identified: (1) nRBC-count DMRs, in which nRBC DNAm was significantly associated with nRBC count; and (2) nRBC-WBC DMRs, or pairwise DMRs between nRBCs and each WBC type. Resulting p-values were adjusted for multiple comparisons by the Benjamini & Hochberg FDR method217, and statistically significant regions were limited to those that passed an FDR <5%. The nRBC-WBC DMRs were further filtered such that the maximum β-fold change within the DMR was ≥0.20.  All nRBC-WBC DMRs were inclusively overlapped to produce a set of “nRBC-specific DMRs”, at which nRBCs were significantly DM from all WBCs. These DMRs were then subdivided based on whether nRBC DNAm in these regions was also associated with nRBC count. If an nRBC-specific DMR overlapped with an nRBC-count DMR, it was classified as a “count-associated nRBC DMR”; if it did not, it was classified as a “cell-only nRBC DMR”. ErmineJ was used to evaluate enrichment of gene ontology (GO) terms in genes associated with the count-associated nRBC DMRs and cell-only nRBC DMRs242.  3.3 Results and discussion 3.3.1 Hypomethylation in nRBCs is distinct from placental hypomethylation Analysis by array-wide Euclidean clustering and PCA reveal that nRBCs have the most distinct genomic DNAm profile of the cord blood cells studied (Figure 3.1A; Supplementary Figure B.1). The uniqueness of nRBC DNAm is reflected in its density distribution (Figure 63  3.1B): although nRBCs and WBCs have similar distributions at unmethylated CpG sites, nRBCs have notably decreased DNAm relative to WBCs at CpG sites that are typically highly methylated. This results in an overall intermediate nRBC DNAm profile. nRBCs also have high inter-individual variability, with higher array-wide median absolute deviation than WBCs and less compact clustering when analyzed by PCA (Supplementary Figure B.1).64   Figure 3.1 nRBCs and placenta are both hypomethylated cell/tissue types. (A) Array-wide β value density distribution. Solid lines represent cell/tissue type mean, dashed lines represent individual samples. (B) Array-wide Euclidean clustering. Numbers refer to different placental donors, letters refer to different cord blood donors. WB = whole cord blood.65  Global hypomethylation is also observed in the placenta, as well as the early embryo and some cancers233-236. I sought to compare the DNAm profile of nRBCs to that of placental whole chorionic villi, which are largely composed of trophoblast cells as well as a range of villus core cells including endothelial cells, fibroblasts, macrophages, and pericytes. On a global scale, placenta appears to be less methylated than nRBCs, but displays more of a trimodal DNAm distribution whereas nRBCs exhibit a bimodal distribution with a left-shifted “methylated” peak (Figure 3.1A). In other words, despite overall lower DNAm, the placenta maintains high levels of DNAm at more sites than nRBCs. Although nRBCs and the placenta are both hypomethylated relative to WBCs, nRBCs still cluster more closely with other blood cells than with villi (Figure 3.1B). This may be attributable to different gene regulation requirements between the two tissues, or it may reflect nRBCs’ origin from a precursor hematopoietic cell. To better understand the epigenetic similarities and differences, I further evaluated tissue- and cell-type DNAm at specific subsets of the array selected based on genomic context. This included surrounding CpG density, the presence of REs and prior association with biological variables of interest, specifically age and hematopoietic source. Detailed descriptions of the CpG site groups are provided in Section 3.2.2.  On an array-wide level, placenta and nRBCs were similarly significantly hypomethylated compared to WBCs and blood cell mixtures (p <8.5e-05) (Figure 3.2A). LINE1 and Alu REs showed similar hypomethylation in the two tissues (Figure 3.2B). In placenta, hypomethylation of LINE1 and Alu elements was suggested to reflect their contribution to the organ’s distinct invasive and proliferative properties243. In nRBCs, hypomethylation at LINE1 and Alu regions may instead be due to loss of DNA methyltransferase (DNMT) activity as erythropoiesis progresses. REs need cooperative action of both maintenance and de novo DNMTs to maintain 66  DNAm244. DNMT3A and DNMT3B, which establish DNAm, are markedly downregulated during erythropoiesis, likely contributing to RE hypomethylation66.   Figure 3.2 Cell-specific DNAm patterns differ at distinct subsets of the genome. (A) Array-wide. (B) Repetitive elements: (i) Alu and (ii) LINE1. (C) Regions of (i) low, (ii) intermediate, (iii) intermediate shore-associated, and (iv) high CpG density. (D) CpG sites found to (i) decrease with age or (ii) increase with age, and CpG sites hypomethylated in nRBCs derived from (iii) adult bone marrow or (iv) fetal liver. Brackets indicate significantly different pairs and are coloured relative to the comparison’s reference cell type (p <0.0045). WB = whole cord blood.  67  Both placenta and nRBCs displayed significant hypomethylation compared to WBCs in regions of low and intermediate CpG density, however at the intermediate CpG density regions placental DNAm was also significantly lower than nRBCs (Figure 3.2C). The genome is generally CpG-poor and methylated, with the exception of CGIs, which are CpG-rich and typically unmethylated. Regions of low and intermediate CpG density are widespread across the genome and encompass biologically relevant regions such as distal gene regulatory elements and REs161,163,168. Thus, hypomethylation in these regions in nRBCs could reflect a combination of global demethylation and erythroid lineage-specific gene regulation. In contrast, nRBCs were the only hypomethylated population at intermediate shore-associated CpG density regions (Figure 3.2C). This nRBC-unique pattern could be due to the role of DNAm at CpG island shores in hematopoietic cell differentiation and associated gene expression3,8, with nRBC hypomethylation in these regions reflecting processes that distinguish erythroid-specific gene regulation from the myeloid and lymphoid lineages. At high CpG density regions, placenta was significantly more methylated, whereas nRBCs were unmethylated like all other blood cell populations (Figure 3.2C). In villi, there is active protection of DNAm at promoters of specific genes, including tumour suppressors, which are generally regions of high CpG density245. The lack of increased DNAm at CpG-dense regions in nRBCs is consistent with a passive demethylation process. Over time, DNAm changes across the genome in a manner that is thought to be partially stochastic and partially influenced by environmental exposures. These age-associated DNAm changes result in epigenetic drift, in which inter-individual variability in DNAm increases with age246-248. Looking at nRBC DNAm in sites previously associated with aging and epigenetic drift may shed light on the specificity of nRBC demethylation. I selected previously-identified CpG 68  sites where DNAm was significantly associated with age238 and grouped them by direction of DNAm change with age (“age loss sites”, at which DNAm decreases with age, and “age gain sites”, at which DNAm increases with age). Placental DNAm conforms to the age-associated DNAm patterns (Figure 3.2D). In contrast, although nRBCs are significantly hypomethylated at age loss sites, it is not as extreme as placental demethylation, and nRBCs have the same DNAm levels as myeloid blood cells at age gain sites. This trend is likely attributable to overlap between age-associated sites and CpG-rich regions, as 11% of age gain sites are in regions of high CpG density compared to only 1% of age loss sites. In placenta, strong CpG islands and certain promoter-associated CpG islands are specifically methylated243,245 (Figure 3.2C), which is perhaps why placenta has elevated DNAm at the age gain sites. In contrast, nRBCs do not appear to gain DNAm at any genomic regions, suggesting a passive demethylation process that is not as tightly controlled as that of the placenta. It is possible that the notably variable DNAm in nRBCs may reflect inter-individual differences in contribution from different hematopoietic sources. In term infants, blood cells generally come from the BM; however, extramedullary sources can be activated in response to stressors like hypoxia or anemia76,77. A recent study comparing nRBCs derived ex vivo from either adult BM HSCs or FL HSCs identified widespread differences in DNAm between the two populations239. From this study’s findings, I created two groups of CpG sites: sites that were hypomethylated in adult BM-derived nRBCs relative to FL-derived nRBCs (“BM-hypomethylated sites”), and sites with the opposite pattern (“FL-hypomethylated sites”). By looking at these sites in our data, I sought to evaluate whether DNAm in nRBCs from term cord blood was more similar to cultured nRBCs of BM or FL origin. nRBCs are the most hypomethylated tissue in BM-hypomethylated sites, but they do not differ significantly from all 69  WBCs at the FL-hypomethylated sites (Figure 3.2D). Thus, it appears that term nRBC DNAm conforms more to BM- rather than FL-specific DNAm patterns.  Intriguingly, nRBC DNAm is also more variable at these adult BM-hypomethylated sites, which suggests that: 1) inter-individual variability at these sites reflects differences in maturity of nRBCs, with all nRBCs derived from the BM; or 2) nRBCs are produced from other hematopoietic sources, and inter-individual variability at BM-hypomethylated sites reflects varying contribution of BM to the nRBC population (Figure 3.3). If the latter is the case, the lack of variability in DNAm at FL-hypomethylated sites suggests that the alternative source of nRBCs is not the liver, but perhaps the placenta or spleen70,76,77. Placenta was one of the most highly methylated tissues at these source-DM sites. This may reflect their role as an alternative hematopoietic source during fetal development47,48, suggesting that placental-derived hematopoietic cells have their own set of hypomethylated sites and have increased DNAm at signature CpG sites for FL- and BM-derived cells. However, it is important to acknowledge that the placenta is a mixed-cell population consisting largely of trophoblast, with a very low proportion of immune cells. Thus, these source-DM sites might not be as relevant to placental biology as they are to nRBCs. 70   Figure 3.3 Two types of nRBC mixtures could give rise to the variable nRBC DNAm profile. (A) A given nRBC population’s DNAm levels are influenced by the overall maturity of the population: nRBCs are on a spectrum of maturity and readiness for enucleation, reflected in their genome-wide DNAm. (B) A given nRBC population’s DNAm levels are influenced by the proportion of nRBC source subtypes: nRBCs come from more than one hematopoietic source, and each subtype has its own DNAm profile.  The distinctive trends in nRBC DNAm relative to placenta suggest that nRBC hypomethylation, although global and partially attributable to stochastic loss, also has some selectivity. This is further supported by the observation that nRBCs, like WBCs and villi, maintain ~50% methylation in imprinted DMRs, but are unmethylated like WBCs at placental-specific imprinted DMRs (Supplementary Figure B.2A-D), although this contrasts with a previous report of demethylation in some germline imprinted DMRs during erythropoiesis66. The finding that some of the most variable and hypomethylated CpG sites in nRBCs have been 71  previously linked to age and hematopoietic source suggest that DNAm in nRBCs could reflect the overall maturity of the population, or their site of production.  An interesting follow up to these findings would be to compare nRBC and placenta DNA hydroxymethylation. Hydroxymethylation is the oxidation of 5-methylcytosine – the methylated cytosine in DNAm – to 5-hydroxymethylcytosine (5hmC). This conversion was originally thought to be an intermediate step in demethylation, however it has also been suggested that 5hmC has its own role in gene regulation249. Similar to the trends in DNAm, genome-wide 5hmC levels are notably lower in placenta250, cancer cells250, and stem cells251 than in somatic cells. 5hmC typically increases with lineage commitment251; however, in erythropoiesis an initial spike in 5hmC occurs with early erythroid lineage commitment but is followed by a steady decline in 5hmC during terminal erythroid differentiation252. Thus, a direct comparison of the low-hydroxymethylation profiles of nRBCs and placenta may provide insight on whether loss of DNAm in these tissues serves to increase functional hydroxymethylation, or to completely erase epigenetic marks.  3.3.2 nRBC DNA methylation is associated with nRBC proportion in cord blood The observed variability in nRBC DNAm may be explained by 1) stochastic loss, or 2) different proportions of nRBCs by either maturation or anatomical origin. Varying proportions of nRBCs in each cord blood sample may also be related to the observed inter-individual differences in nRBC DNAm (Supplementary Table B.1). To test this, I evaluated the relationship between nRBC DNAm and nRBC count. Genome-wide median nRBC DNAm was significantly negatively correlated with nRBC proportion (Pearson r =-0.928, p =0.003), with a rate of change of -0.0080 Δβ per unit increase in nRBCs/100 WBCs (Figure 3.4A). This relationship was 72  strongest in intermediate and low CpG density regions (-0.0078 Δβ/nRBC and -0.0110 Δβ/nRBC, respectively), and non-significant in high and intermediate shore CpG density regions (Figure 3.4B). There was no significant correlation between nRBC DNAm and nRBC count at CpG sites located in known imprinted DMRs, including placental-specific ones (Supplementary Figure B.2E).73   Figure 3.4 nRBC DNAm is negatively correlated with nRBC proportion in cord blood. Asterisks in legend indicate significant Pearson correlation (p <0.05) for that group of CpG sites. (A) Array-wide median DNAm. (B) Median DNAm in regions of low, intermediate (Int.), intermediate shore-associated (Int. shore), and high CpG density. (C) Median DNAm at CpG sites in Alu and LINE1 repetitive elements. (D) Median DNAm of CpG sites where DNAm has been previously associated with age, separated by direction of DNAm change with age. (E) Median DNAm of CpG sites previously found to vary in DNAm in nRBCs based on hematopoietic source, either adult bone marrow or fetal liver. 74  Significant negative correlations between median nRBC DNAm and nRBC count were also observed in CpG sites overlapping with LINE1 and Alu REs (Pearson r =-0.929, p =0.003 and r =-0.889, p =0.007, respectively) (Figure 3.4C). At the age-associated sites, only the age loss sites showed a significant correlation between median nRBC DNAm and nRBC count (Pearson r =-0.948, p =0.001; Figure 3.4D). In contrast, median nRBC DNAm was the same across all individuals at sites where DNAm increases with age. However, the lack of variability at these age-gain sites is likely because of their high representation of CpG-dense regions, which are generally unmethylated and not prone to changes in DNAm (Figure 3.2D). Intriguing patterns emerged when the nRBC DNAm-count relationship was evaluated in the source-DM sites (Figure 3.4D). My initial hypothesis was that increasing nRBC count may reflect the activation and relative contribution of extramedullary hematopoietic sources, with a greater proportion of circulating “alternative source” nRBCs as overall nRBC count increases70,76,77. Were that the case, I anticipated seeing a high-magnitude decrease in nRBC DNAm at the FL-hypomethylated sites as nRBC count increases, as that would reflect a greater proportion of FL-derived nRBCs. Instead, the opposite trend emerged: as nRBC count increased, nRBC DNAm became more similar to the adult BM-derived epigenetic signature, with no significant change in the FL-hypomethylated sites. There was a significant negative correlation between nRBC DNAm at BM-hypomethylated sites and nRBC count (Pearson r =-0.854, p =0.015), with these sites showing the greatest magnitude of decline of all genomic regions evaluated (-0.0144 Δβ/nRBC). In contrast, there was no significant correlation between nRBC DNAm and nRBC count at the FL-hypomethylated sites (Pearson p =0.482). This suggests that, although nRBC DNAm may reflect the contribution of BM to erythropoiesis, the liver is not an alternative hematopoietic source that varies significantly in its contribution between term 75  neonates. It may be that hematopoietic organs other than the fetal liver are being activated, such as the placenta or the spleen.  It is important to consider not only the differing hematopoietic source, but also the difference in age between the Lessard et al. samples from which the source-DM sites were defined239. The BM-hypomethylated sites were originally identified in erythroblasts derived from adult BM, whereas the FL-hypomethylated sites were identified in erythroblasts from the fetal liver. Thus in my data, the demethylation in nRBCs at BM-hypomethylated sites with increasing nRBC count in our data could reflect an increasingly “adult-like” quality in this cell population. I suggest that, if this is the case, that fetal and adult erythroblasts may differ in the proportion of different erythroblast stages, as depicted in Figure 3.3A. The original paper showed that their fetal and adult erythroblasts had similar differentiation kinetics, based on expression of stem cell marker CD34 and erythroid markers CD71 and CD235a239. However, recent work has shown that further categorizing of erythroblasts with cell surface markers band-3 and α4-integrin isolates more distinct erythroblast stages253. This means that the adult and fetal erythroblasts studied by Lessard et al.239 may have been age-specific mixtures of erythroblast sub-populations. Since global demethylation occurs during erythropoiesis66,67, the decrease in nRBC DNAm I observed with increasing nRBC count could reflect higher proportions of developmentally mature nRBCs in high-nRBC count populations (Figure 3.3A). If this is the case, a possible reason for this trend could be low numbers of macrophages, which are key facilitators of enucleation227.  This would likely result in an increased number of nRBCs that are ready for enucleation, but “stuck”. Another possible reason for more mature nRBCs in high-nRBC count populations could be erythropoietin-mediated changes in bone marrow structure, which occurs in response to hypoxic or inflammatory stressors. Increasing the size of bone marrow fenestrations 76  would allow for the premature release of some nRBCs, but likely only the smaller, more mature sub-populations. The physiological changes reflected by nRBC count may also be associated with DNAm changes in WBCs. Linear modelling performed between DNAm in each blood cell type and nRBC count showed that all blood cells had some CpG sites significantly associated with nRBC count (Table 3.1). As expected, the effect was most apparent in nRBCs themselves: 106,810 CpG sites showed nRBC DNAm significantly associated with count, with nearly all of these sites (97%) decreasing in DNAm as nRBC count increases. In WBCs, a far smaller number of sites showed changes in DNAm with nRBC count, ranging from 217 to 972. With the exception of CD4 T cells, which had 79% of its count-DM sites displaying increasing DNAm with count, none of the WBC count-DM sites showed an obvious trend towards increasing or decreasing DNAm with count. This suggests that whatever nRBC count reflects also influences WBC DNAm, albeit to a lesser extent than the relationship with nRBC DNAm. If it is a change in nRBC maturity (Figure 3.3A), this may impact WBC populations by changing the immune environment. Immature erythrocytes in fetal and neonatal circulation are thought to promote immunosuppression by arginine depletion93. Maturation of the nRBC population towards readiness for enucleation could make them less transcriptionally active, relieving immunomodulatory pressures on fetal WBCs and potentially affecting DNAm. Alternatively, if nRBC count reflects varying contributions from hematopoietic sources, the epigenetic change likely reflects introduction of an nRBC subgroup with distinct DNAm patterns (Figure 3.3B). The same process would likely also occur in WBCs, with the alternative hematopoietic organ producing WBCs with a DNAm profile distinct from those produced in the BM.   77  Table 3.1 Number of CpG sites in each blood cell population that show differential methylation (FDR <5%) with the proportion of nRBCs in cord blood.  nRBCs B cells CD4  T cells CD8  T cells Gran. Mono. NK cells All WBCs Total DM sites 106,810 478 972 217 711 397 435 57 Decreasing DNAm with nRBC count 103,005 249 203 101 406 178 158 20 Increasing DNAm with nRBC count 3,805 238 769 116 305 219 277 35 |Δβ > 0.02 /nRBC| 8,459 45 50 34 59 37 24 1 Gran. = granulocytes; Mono. = monocytes.  3.3.3 Differentially methylated regions distinguishing nRBCs from WBCs After evaluating the nRBC DNAm profile by genomic region, I performed DMR analysis to identify DM CpG sites most strongly associated with the erythroid lineage. DMRs were identified between nRBCs and each WBC type (FDR <5%, |Δβ| >0.20) then overlapped to produce a total of 2,965 nRBC-specific DMRs, at which nRBCs have distinct DNAm levels from all WBC types. The majority of nRBC-specific DMRs (2,061, 69.5%) were hypomethylated relative to WBCs at the CpG site with the largest DNAm difference. nRBCs were an intermediately methylated cell population at 852 (28.7%) of nRBC-specific DMRs, and were hypermethylated relative to all WBCs at only 54 (1.8%) of nRBC-specific DMRs. Section 3.3.2 describes the association between nRBC DNAm and the proportion of nRBCs in cord blood. To follow up on these findings, I assessed how many of the nRBC-specific DMRs also show an epigenetic relationship with nRBC count. Of the 2,965 nRBC-specific 78  DMRs, 2,515 overlapped with regions in which nRBC DNAm is significantly associated (FDR <5%) with nRBC count (termed “count-associated nRBC DMRs”), whereas 450 did not (termed “cell-only nRBC DMRs”) (Figure 3.5A). 2,350 (93.4%) of the count-associated nRBC DMRs showed decreasing DNAm in nRBCs as their proportion in cord blood increased. Next, I wanted to ask if count-associated nRBC DMRs over-represent certain gene regions compared to cell-only nRBC DMRs. A significantly larger proportion of cell-only DMRs are found in the gene body (χ2=6.69, p=0.009) (Figure 3.5B).   Figure 3.5 Summary of nRBC DMRs. (A) Flowchart of nRBC-specific DMR identification and sub-groupings based on association between nRBC DNAm and nRBC count. (B) Proportion of nRBC-specific DMRs covering each type of gene region, *: p <0.05. 79  GO analysis was then performed using ErmineJ on the nRBC-specific DMRs. The 450 cell-only nRBC DMRs were first evaluated. No GO terms were significantly enriched (FDR <10%) for these DMRs, however many of the top-ranked terms were related to immune cell processes such as growth of symbionts and bacteria, chemokine processes, and WBC migration (Supplementary Table B.2). Since these DMRs are associated with nRBC-specific DNAm regardless of nRBC count, their corresponding GO terms shed light on consistent epigenetic differences between erythroid cells and other blood cells. The presence of immune functions in the higher-ranked GO terms likely reflects that all of the cell populations being compared to nRBCs are WBCs, which have epigenetic profiles suited for immune function. The 2,515 count-associated nRBC DMRs were significantly enriched for 6 GO terms (FDR <10%). These terms were related to activity of the Ras and Rho guanyl nucleotide exchange factor, RNAi, the cytoskeleton, and the phosphatidylinositol 3-kinase cascade (Supplementary Table B.3). These enriched terms may reflect the unique structural and biochemical changes occurring in erythroblasts as they prepare for enucleation such as alterations in plasma membrane composition, reorganization of the cytoskeleton, chromatin condensation, and polarization of the nucleus65,68. For example, RNAi has been reported to be a mediator of chromatin condensation during enucleation68.  Closer inspection of nRBC-specific DMRs illustrates that nRBCs do not follow the traditional pattern of cell-specific DNAm, in which DM sites are either fully methylated (β>0.80) or unmethylated (β<0.20), with comparison cells exhibiting opposing levels of DNAm. Instead, the typical pattern I observed is intermediate absolute DNAm in nRBCs, which was often sandwiched between opposite extremes in DNAm displayed by lymphoid and myeloid cell types (Figure 3.6A-C). This may seem incongruous with the earlier statement that the majority of 80  nRBC-specific DMRs were hypomethylated relative to WBCs; this is due to the nRBC-specific DMRs being derived from multiple nRBC-WBC pairwise comparisons. A given DMR had to meet the significance thresholds (FDR <5%, |Δβ| >0.20) for each nRBC-WBC pairwise comparison to be classified as an nRBC-specific DMR. However, within the DMR, it may be a different CpG site with the greatest DNAm difference between nRBCs and each WBC type. Thus, the majority of nRBC-specific DMRs have a maximum Δβ that is negative (i.e. nRBCs are hypomethylated), but this nRBC hypomethylation can occur at different CpG sites within the DMR for different WBC populations. Some nRBC-specific DMRs showed “true” nRBC hypomethylation, with all WBCs fully methylated; however, this hypomethylation was only moderate and nRBC DNAm ranged from 0.20 to 0.50 (Figure 3.6D). Since our data and prior studies have shown progressive demethylation in nRBCs with maturation, the lack of fully methylated regions in nRBCs is not surprising.81   Figure 3.6 Examples of nRBC-specific DMRs. Top plot is all samples, coloured by cell type (legend at top right); bottom plot is individual nRBC samples, coloured by nRBC count (legend at bottom right). (A) AZU1. (B) MAFG. (C) EPHB3. (D) PRKACA.  82  Although the I did not put a |Δβ| threshold on the association between nRBC DNAm and nRBC count, many count-associated nRBC-specific DMRs displayed striking DNAm changes of >0.02 per unit increase in % WBC (Figure 3.7A). This is the equivalent of a ~0.20 β-value range in nRBC DNAm from individuals with a very low nRBC count (e.g. 0% WBC) to individuals with a higher (but not abnormal) nRBC count of 10% WBC. Several of the high-magnitude count-associated nRBC DMRs were located within 200bp upstream of the TSS or in the 5’ UTR, and thus may be impacting transcription of these genes (Figure 3.7B-D). 83   Figure 3.7 nRBC-specific DMRs with nRBC count-associated DNAm. (A) Distribution of count-associated nRBC-specific DMRs by DNAm change with nRBC count (Δβ per unit increase in % WBCs). (B-D) Examples of count-associated nRBC-specific DMRs; top plot is all samples coloured by cell type (legend at top right), bottom plot is individual nRBC samples coloured by nRBC count (legend at bottom right). (B) TPM4. (C) EPN1. (D) RFFL. 84  3.4 Conclusion nRBCs are a cell population that is given little thought in studies of neonatal immunity. Although erythroid cells have been found to impact proliferation and immune function of T cells, B cells and dendritic cells91-96, and fetal nRBCs display immunosuppressive effects not shown by their adult counterparts93, an immune role for nRBCs in neonatal circulation is not yet widely accepted. It has actually been further called into question by a recent study that was unable to replicate these fetal erythroid-specific immune effects99. Following this limited understanding of nRBCs, the implications of the global, passive demethylation during erythropoiesis may not be meaningful in terms of cell function. Our data suggests that the nRBC DNAm profile partially supports passive loss of DNAm in this cell population, with tendency for demethylation at high-methylated regions that lack obvious associations with gene expression, such as REs (Figure 3.2B). However, analysis of nRBC DNAm at sites previously associated with age and hematopoietic source reveals considerable heterogeneity in this population (Figure 3.2D, Figure 3.4D-E), and DMR analysis suggests a link between nRBC-specific DNAm and erythropoietic processes (Supplementary Tables B.2-B.3). As their proportion in cord blood increases, nRBCs conform more to the adult BM DNAm profile. I suggest that this could reflect two possible changes in the nRBC population. First, that individuals with high overall nRBC count may have a higher proportion of mature nRBCs, which may be due to low abundance of macrophages to mediate that final enucleation step. Second, these DNAm changes could reflect an increasing proportion of nRBCs coming from the BM with increasing representation in cord blood. Since it’s unusual for BM to allow erythroid cells to enter circulation prior to enucleation, this increasing leniency could reflect some erythropoiesis-inducing stressor. It is also interesting to consider what the 85  alternative hematopoietic source (which would decline in nRBC production with increasing nRBC count) may be, since our data suggests that it is not the fetal liver (Figure 3.2D, Figure 3.4E). If this alternative source is the placenta, the relative contribution of these two organs to the nRBC population in term births could reflect how quickly nRBC counts drop after birth: nRBCs coming mostly from the placenta would be rapidly cleared after birth due to the lost connection to this organ. In contrast, production of nRBCs from the BM could continue for some time after birth. In that case, these DNAm changes could be useful to identify neonates likely to have persistently elevated nRBC counts after birth, which is associated with an increased risk of morbidity and mortality71. Ultimately, neonatal nRBCs are a complex population with unclear function; further study is required to understand whether their distinctive epigenetic profile reflects transience or functional processes.    86  Chapter 4: Comparing Hematopoietic Cell DNA Methylation Profiles between Preterm and Term Births 4.1 Background PTB is a complex disorder involved in over 50% of neonatal deaths, either as a direct cause or through prematurity-associated complications1. If a premature infant survives the immediate postnatal period, they face increased risk of developing a variety of short- and long-term health conditions, including childhood infections, asthma, and cognitive disabilities, and adult hypertension and heart disease2,139-141,145,147. This disease-burdened life is attributable to systemic immaturity, as well as increased oxidative stress and altered metabolism as a result of invasive medical procedures and parenteral nutrition in early life17,136,137. The immune system is not spared from the effects of PTB. Blood cell composition and functional capacity change throughout gestation; the interruption of this dynamic process results in a preterm immune system that is unprepared for the microbe-ridden external environment. A variety of general and cell-specific deficits in immune function have been identified in preterm infants17,57,131, which increase the preterm neonate’s vulnerability to infection. The association between DNAm and hematopoietic cell lineage commitment has been well established3,4, and multiple studies have found differential methylation between preterm and term infants190,191,254. However, these studies used whole cord blood for comparison, which is a mixed-cell sample in which overall DNAm levels are influenced by cell composition194,210. As a result, these studies cannot distinguish prematurity-associated DNAm patterns due to differences in cell composition from DNAm patterns due to differences in immune function. 87  The objective of this chapter is to identify DNAm changes with PTB in the major immune cells of cord blood. I hypothesize that prematurity-associated DNAm patterns will reflect both changes in hematopoietic source and cell-specific maturation processes. Using the 450K array, I compared the genome-wide DNAm profiles between term infants and very preterm infants (<30 weeks GA) in granulocytes, monocytes, T cells, and nRBCs sorted from cord blood by FACS.  4.2 Methods 4.2.1 Sample collection and cell purification Ethics approval for this study was obtained from the University of British Columbia Children’s and Women’s Research Ethics Board (certificate numbers H07-02681 and H04-70488). Written, informed parental consent to participate was obtained. Individual patient data is not reported. Cord blood was collected from neonates delivered by elective caesarean section in absence of labor at the Children’s and Women’s Health Centre of BC (Vancouver, Canada). A total of 10 infants were involved in the study: 5 preterm (GA range 26-31 weeks) and 5 term (GA range 38-40 weeks) (Table 4.1). Hematopoietic cell purification and sorting is described in Appendix Section C.1. All cell populations (T cells, granulocytes, monocytes, and nRBCs) were collected from all 5 term subjects; however, due to small sample volumes and variability in blood cell counts, not all cell populations were collected from all preterm subjects (Table 4.1).    88  Table 4.1 Subject characteristics and cell types collected from each subject  Sex GA, weeks T cells Granulocytes Monocytes nRBCs term_1 M 38-40* Y Y Y Y term_2 M 38-40* Y Y Y Y term_3 F 38-40* Y Y Y Y term_4 M 38-40* Y Y Y Y term_5 M 38-40* Y Y Y Y preterm_A M 26 Y N N Y preterm_B F 29 Y Y Y N preterm_C M 30 Y Y Y Y preterm_D F 30 Y N Y Y preterm_E** M 30 Y Y Y Y * = exact GA of term infants not known ** = one of dichorionic/diamniotic twins  4.2.2 Illumina Infinium HumanMethylation450 BeadChip DNA was extracted from all samples using standard protocols and purified with the DNeasy Blood and Tissue kit (Qiagen). DNA was bisulphite-converted using the EZ DNA Methylation Kit (Zymo Research) before amplification and hybridization to the 450K array following manufacturer’s protocols (Illumina). 450K array chips were scanned with a HiScan reader (Illumina).  4.2.3 DNA methylation data preparation and analysis Raw intensity data for all blood cells and cell mixtures were background normalized in GenomeStudio (Illumina). Quality control was performed using the 835 control probes included in the array. The intensity data were then exported from GenomeStudio and converted into M values using the lumi package214 in R software215. Sample identity and quality were evaluated as described in Appendix Section C.1. The 450K array targets 485,577 DNAm sites, but probe 89  filtering was performed as described in Appendix Section C.1 to produce a final dataset of 429,765 sites. Background intensity and red-green color bias were corrected for using the lumi package, and the data were normalized by subset within-array quantile normalization201. Unsupervised Euclidean clustering of the samples based on DNAm β values and PCA based on DNAm M values were performed as exploratory global analysis steps. I then looked at DNAm at specific subsets of the 450K array, based on 1) CpG density; 2) previous association with age-related DNAm changes238; 3) previous association of cord blood DNAm with PTB191; or 4) overlap with hematopoietic “source-DM sites”239, which display differential methylation between erythroblasts derived from FL and erythroblasts derived from adult BM. The groups were as follows: 1) CpG sites in regions of high (141,866), intermediate-shore (30,369), intermediate (101,122) or low (156,408) CpG density; 2) CpG sites linked to age-related DNAm changes238, separated by whether they increased (1,030) or decreased (1,797) with age; 3) CpG sites hypomethylated in PTB (961) and hypermethylated in PTB (505)191; and 4) the top 100 CpG sites hypomethylated in adult BM erythroblasts (“BM-hypomethylated sites”) or hypomethylated in FL erythroblasts (“FL-hypomethylated sites”), based on Lessard et al.’s239 Δβ values. DNAm at these CpG site groups were compared (as β-values) between all cell and tissue types using ANOVA followed by Tukey’s honest significant difference test, using a multiple comparison-adjusted p-value threshold of 0.0045. To assess GA-associated DNAm changes in each blood cell population, linear modelling using the R package limma216 was performed, with the combination of cell type and birth group (preterm or term) as a variable of interest and sex included in the model as a covariate. Since each cell type was collected from the same set of individuals, DNAm may have been influenced by inter-individual differences. To adjust for this, the model included a within-individual 90  consensus correlation estimated using the duplicateCorrelation() function in limma216. Resulting p-values were adjusted for multiple comparisons by the Benjamini & Hochberg217 FDR method, and statistically significant sites (“prematurity-DM sites”) were limited to those with an FDR <5% and a |Δβ| >0.10. To identify cell-type specific DNAm in the preterm and term immune system, I performed linear modelling using the R package limma216 with the same model described above. Resulting p-values were adjusted for multiple comparisons by the Benjamini & Hochberg217 FDR method, and statistically significant sites (“cell-type DM sites”) were limited to those with an FDR <5% and a |Δβ| >0.20. ErmineJ was used to evaluate enrichment of GO terms in genes associated with the cell-type and prematurity-DM sites242.  4.3 Results and discussion 4.3.1 Comparing global DNA methylation of preterm and term immune cells Cell type is the dominant influence when global DNAm profiles of term and preterm cell populations are compared by array-wide Euclidean clustering (Figure 4.1A). Prematurity also has an observable impact on epigenetic relationships between the samples, with some preterm samples clustering on their own within each cell type. However, these GA subgroups are not perfect, with some preterm samples clustering more closely with their term counterparts (Figure 4.1A). Evaluating genome-wide DNAm by β-value density distributions suggests that the effect of prematurity may be largest in nRBCs: all of the WBCs show similar distributions between term and preterm samples, whereas term nRBCs appear strongly hypomethylated relative to preterm nRBCs (Figure 4.1B). Since global demethylation occurs with successive cell divisions 91  during erythropoiesis66,67, this difference may reflect a greater proportion of more mature nRBCs in the term nRBC populations.   Figure 4.1 Genome-wide DNAm comparisons between major hematopoietic cells in term and preterm births. (A) 450K-array wide Euclidean clustering. Letters and numbers refer to different preterm and term cord blood donors, respectively. (B) DNAm (β value) density distribution; dashed lines represent individual samples, solid lines represent the mean of that GA/cell type group.  4.3.2 Prematurity-associated DNA methylation at biologically relevant subsets of the genome To identify genomic regions where the association between DNAm and prematurity are strongest, specific subsets of the 450K array were evaluated. These groups were based on CpG density, as well as prior associations between DNAm and hematopoietic source, age, or PTB, as described in Section 4.2.3191,238,239.  None of the WBC populations differed in median genome-wide DNAm between preterm and term infants, and although term nRBCs were notably hypomethylated compared to preterm 92  nRBCs, this difference was not significant (Figure 4.2A). When the genome is divided by CpG density, no significant differences were observed between preterm and term samples at regions of high or shore-associated intermediate CpG density (Figure 4.2B). WBC populations also displayed similar DNAm levels at intermediate and low CpG density regions, regardless of prematurity (Figure 4.2B). This indicates that there are too few prematurity-associated DNAm changes in WBCs to produce observable differential methylation on a large scale. In contrast, term nRBCs displayed hypomethylation relative to their preterm counterparts in these regions; however these differences did not pass the multiple-test corrected significance threshold (p-values <0.05, but >0.0045). This trend is likely due to higher DNAm levels in intermediate and low CpG density regions, which are also generally more dynamic than high and shore-associated intermediate CpG density regions161,255. 93   Figure 4.2 DNAm relationships between GA groups and cell types differ at different regions of the genome. (A) Genome-wide. (B) CpG sites grouped by CpG density. (C) CpG sites hypomethylated in nRBCs derived from adult bone marrow (top) or fetal liver (bottom). (D) CpG sites found to increase with age (top) or decrease with age (bottom). (E) CpG sites found to be differentially methylated in cord blood between preterm and term births, either hypomethylated in PTB (top) or hypomethylated at term (bottom). *: p<0.05; **: p <0.0045 94  During fetal development, the shift of the predominant hematopoietic organ from FL to BM occurs around 24 weeks gestation24. Thus, it is likely that the preterm samples, which range from 26-30 weeks GA, have a greater proportion of FL-derived cells than the term samples. To see if hematopoietic source is related to DNAm differences between preterm and term samples for specific cell types, I analyzed candidate CpG sites identified in a previous study239. The original study identified DNAm differences between nRBCs derived ex vivo from adult BM HSCs and those derived from FL HSCs. From that work, I created two groups of CpG sites: sites that were hypomethylated in adult BM-derived nRBCs relative to FL-derived nRBCs (“BM-hypomethylated sites”), and sites with the opposite pattern (“FL-hypomethylated sites”). All cell types showed the same patterns at these source-DM sites, with preterm samples hypomethylated in FL-hypomethylated sites, and term samples hypomethylated in BM-hypomethylated sites (Figure 4.2C). It is interesting that both WBCs and nRBCs show the same DNAm patterns, despite these candidate sites being identified only using nRBCs. The consistent DNAm differences at these source-DM sites suggest that hematopoietic sources have epigenetic signatures that are conferred to all cell types derived from that source, regardless of lineage. Another intriguing trend in this data is the finding that there are CpG sites where nRBCs gain DNAm with GA, as the overwhelming trend is for nRBCs to become demethylated as the fetus approaches term. This indicates that nRBC demethylation, although global and passive66,67, is selective, with some CpG sites being protected from DNAm loss. These results are also an in vivo verification of the original findings, which were performed on nRBCs ex vivo239. It is important to note that the source-DM sites are comparisons of not only hematopoietic source (liver versus BM), but also age (fetal versus adult). To evaluate whether the term-preterm differences observed at these sites may instead be more reflective of age than 95  hematopoietic source, these samples were next compared at a set of CpG sites previously identified as DM between newborns and nonagenarians238. At sites where DNAm increases with age, none of the cell types show significant DNAm differences between preterm and term samples, and only nRBC DNAm decreases with GA at sites where DNAm decreases with age (Figure 4.2D). Since these adult-neonate DM sites do not show the same trends observed at the source-DM sites, this suggests that the source-DM sites are more a reflection of hematopoietic source than of age-related differences. The hypomethylation of term nRBCs at sites where DNAm decreases with age may be more reflective of the high DNAm levels at these sites, as well as their increased representation of intermediate and low CpG dense regions (35% and 61% relative to 24% and 36% in the 450K array, respectively), as these types of CpG sites are generally more variably methylated and thus more likely to lose DNAm. Finally, the samples were compared at CpG sites previously found to be DM in umbilical cord blood between preterm (<31 weeks GA) and term births191. All cell types conformed to the expected trend at sites originally hypomethylated in term cord blood, with term cell populations significantly hypomethylated compared to their preterm counterparts (p <0.0045 for granulocytes, monocytes, nRBCs; p <0.05 for T cells; Figure 4.2E). At sites originally hypomethylated in preterm cord blood, only WBCs showed significant hypomethylation in the preterm samples relative to term samples (all p <0.0045); nRBCs did not differ significantly between preterm and term (Figure 4.2E). This is likely because these sites were selected based on DNAm differences in whole cord blood, and are not specific to nRBCs. Unlike with the source-DM sites, which were identified in pure erythroid populations, nRBC demethylation with gestation may have been masked in these PTB-DM sites by DNAm changes in WBC populations. 96  Overall, these findings suggest that, rather than reflecting a gradual change in DNAm, prematurity-associated DNAm likely reflects shifting proportions of distinct sub-populations of each hematopoietic cell type, with each one derived from a different hematopoietic organ. The exception to this trend is nRBCs, which are gradually demethylated with successive cell divisions; however, even this unusual cell population still retains an epigenetic signature of hematopoietic source (Figure 4.2C).  4.3.3 Prematurity-associated differentially methylated sites in hematopoietic cells To follow up the large-scale evaluation of how DNAm changes in hematopoietic cells with prematurity, I performed linear modelling within each cell type to identify cell-specific prematurity-associated DM sites (FDR <5%, |Δβ| >0.10). nRBCs showed the greatest difference between preterm and term samples, with 9,258 DM sites; granulocytes, monocytes, and T cells had a total of 987, 692, and 273 prematurity-DM sites, respectively (Table 4.2). The direction of DNAm change in these prematurity-DM sites followed known patterns of DNAm in terminally differentiated blood cells. Lymphoid cells display increased DNAm with terminal differentiation 3,8, which is mirrored by the higher proportion of T cell prematurity-DM sites (72%) with increased DNAm in term samples. The majority of granulocyte and monocyte prematurity-DM sites were less methylated in term samples (69% and 61%, respectively), in keeping with the documented loss of DNAm in myeloid lineage commitment3,8. The vast majority of nRBC prematurity-DM sites (94%) showed reduced DNAm in term samples. Considering the global demethylation that occurs during erythropoiesis66,67, this suggests that the term nRBC population is a more mature one, with a greater proportion of nRBCs further along in the erythropoietic process. This could be a reflection of shifting hematopoietic sources in fetal development: since 97  BM, the predominant hematopoietic organ at term, typically does not release nRBCs into circulation, perhaps the few that are released are closer to enucleation than those released by the liver or other hematopoietic organs, which are more lenient about releasing nRBCs into circulation.  Table 4.2 Number of prematurity-DM sites for each cell type (FDR <5%, |Δβ| >0.10).  T cells Granulocytes Monocytes nRBCs Total 273 987 692 9,258 DNAm decreases with GA 76 (28%) 679 (69%) 425 (61%) 8,731 (94%) DNAm increases with GA 197 (72%) 308 (31%) 267 (39%) 527 (6%)  The idea that prematurity-associated DNAm reflects differences in hematopoietic source is supported by the overlap between prematurity-DM sites and Lessard et al.’s source-DM sites239 (Table 4.3). CpG sites that decline in DNAm over gestation overlap almost exclusively with BM-hypomethylated sites, whereas CpG sites increasing in DNAm over gestation overlap almost exclusively with FL-hypomethylated sites. Thus, DNAm changes in the prematurity-DM sites are consistent with a shift from liver to BM as the predominant hematopoietic source in late gestation. The FL-hypomethylated CpG sites increase in DNAm over gestation in these data, as the liver reduces hematopoietic cell production. In contrast, DNAm declines with GA in CpG sites associated with the hypomethylated epigenetic signature of the BM, corresponding with this organ becoming the primary source of blood cells.   98  Table 4.3 Overlap between cell-specific prematurity-DM sites (FDR <5%, |Δβ| >0.10) and source-DM sites from Lessard et al.239.  T cells Granulocytes Monocytes nRBCs DNAm decreases with GA Total 76 679 425 8,731 Overlap with BM-hypo. sites 25 (32.9%) 197 (29.0%) 213 (50.1%) 895 (10.3%) Overlap with FL-hypo. sites 0 (0.0%) 1 (0.1%) 0 (0.0%) 2 (0.0%) DNAm increases with GA Total 197 308 267 527 Overlap with BM-hypo. sites 1 (0.5%) 0 (0.0%) 1 (0.4%) 1 (0.2%) Overlap with FL-hypo. sites 22 (11.2%) 74 (24.0%) 70 (26.2%) 89 (16.9%)  The number of prematurity-DM sites may also relate to the magnitude of phenotypic differences between preterm and term cell populations. Deficits in innate immune function are well-documented in preterm infants, many of which stem from reduced pro-inflammatory and antiviral cytokine production by monocytes, dendritic cells, and macrophages131,150-154. In contrast, fewer functional differences between preterm and term T cells have been observed154,256. The greater number of prematurity-DM sites in monocytes compared to T cells (692 versus 273) may reflect the more severe impact of PTB on monocyte development. In granulocytes, differences with prematurity have not been as extensively studied, although some reductions in neutrophil motility and function have been observed57,120,151,257. The high number of granulocyte prematurity-DM sites (987) could mean that they show more dynamic maturation across late gestation than both monocytes and T cells. Alternatively, these granulocyte DNAm changes could be a reflection of cell composition. Although overwhelmingly represented by 99  neutrophils in term cord blood (>95%; data not shown), granulocytes are a cell mixture that also includes eosinophils, basophils, and mast cells. Since neutrophils are one of the last cell types to be produced during fetal hematopoietic development55, they may be sufficiently underrepresented in preterm granulocyte populations to impact the overall DNAm profile. Regardless, these findings suggest that functional studies comparing preterm and term granulocytes may provide new insight into limitations of the preterm immune system. Finally, the drastic number of DNAm changes in nRBCs is likely associated with widespread structural changes occurring in both the cell and its DNA as erythroid differentiation progresses66-68.  GO term enrichment analysis of the prematurity-DM sites revealed distinct sets of significantly enriched terms for each cell type (Supplementary Table C.1). The two significant GO terms in granulocytes (FDR <10%) related to negative regulation of the ERK1/2 cascades, and Ras guanyl-nucleotide exchange factor activity. Both of these terms refer to the same cascade, the Ras-Raf-MEK-ERK pathway. Defects in this pathway have been associated with impaired neutrophil function in both oxidative responses and NET formation258,259; thus, epigenetic differences in genes involved in this pathway may be related to preterm neutrophils’ observed deficiencies in respiratory burst and NET formation, as well as over-production of superoxide120,257. The ERK1/2 pathways are also associated with immunotolerance, with monocytes and macrophages displaying inhibited ERK1/2 activation in the presence of immunosuppressive factors adenosine and prostaglandin E2260,261. This suggests that DNAm changes in granulocytes could relate to changes in fetal immunotolerance as gestation progresses. In T cells, the only significant GO term (FDR <5%) was embryonic placenta development. This may reflect the link between T cells and placenta during fetal development, which has been well documented both in terms of T cell function and the increased ratio of 100  regulatory T/effector T cell ratio earlier in gestation24,58,104,124. Monocyte prematurity-DM sites were associated with 8 significant GO terms (FDR <10%), all of which were related to epidermal and hair growth and development. These unexpected processes may relate to the key role of monocytes and monocyte-derived cells, such as dendritic cells and macrophages, in the cutaneous innate immune system262. Evaluating the nRBC prematurity-DM sites revealed 152 significantly-enriched GO terms (FDR <10%); recurring themes in this list included Ras- and Rho-related activity, the cytoskeleton, and terms related to renal, muscle, and neuronal processes. The enrichment of these terms may be a reflection of the extensive structural changes occurring in nRBCs throughout terminal erythroid differentiation, such as nuclear polarization, protein sorting, regional changes in cell membrane composition, and development of an actin contractile ring for enucleation65,68.  The prematurity-DM sites for each cell type have 25 CpG sites in common, 17 of which increase in DNAm and 8 of which decrease in DNAm with GA (Table 4.4; Supplementary Figure C.1). The genomic context of some of these common prematurity-DM sites, in combination with the direction of DNAm change, suggests inhibition of growth and proliferative processes in term immune cells. For example, two tumour suppressor genes have CpG sites in this list: a site under 1500 bp upstream of the TSS of STK10 that increases in DNAm with GA, and a site in an enhancer region within the RARRES3 gene body that decreases in DNAm with GA. Additionally, the oncogene WWTR1 contains 2 common prematurity-DM sites, both of which are located in an intergenic CpG island and increase in DNAm with GA.  101  Table 4.4 Location and genomic context of the 25 prematurity-DM sites (FDR <5%, |Δβ| >0.10) common to T cells, granulocytes, monocytes, and nRBCs. Increasing DNAm with GA 450K array CpG identifier Chromosome, location Relation to CpG island UCSC-associated gene, location In an enhancer region? cg20593826 1, 59280370 Island n/a Yes cg08731696 1, 59280489 Island n/a Yes cg11254532 1, 61520399 South shore n/a No cg17476910 1, 147806172 North shore n/a No cg13051013 1, 149684418 n/a n/a No cg15043384 3, 149374761 Island WWTR1, body No cg14557185 3, 149374763 Island WWTR1, body No cg08991643 3, 194014928 Island n/a Yes cg12155036 3, 194015171 South shore n/a Yes cg18043514 5, 158531132 North shore n/a No cg19873297 5, 171616388 South shore STK10, TSS1500 No cg15630458 8, 142986649 South shore n/a Yes cg26542567 8, 142986670 South shelf n/a Yes cg14616234 9, 84305683 South shore n/a No cg05135499 10, 21798698 Island n/a No cg22190721 16, 68269295 Island ESRP2, body No cg18284022 20, 2676474 South shore EBF4, body No Decreasing DNAm with GA 450K array CpG identifier Chromosome, location Relation to CpG island UCSC-associated gene, location In an enhancer region? cg23062810 7, 73720808 Island CLIP2, 5’ UTR Yes cg16356456 7, 73720870 Island CLIP2, 5’ UTR Yes cg13687834 10, 3514783 n/a n/a Yes cg22980293 10, 126158010 n/a LHPP, body No cg04999352 11, 63304614 n/a RARRES3, body Yes cg01787084 16, 87371097 South shelf FBXO31, body No cg03008387 20, 62089152 North shelf KCNQ2, body No cg15035133 22, 37962637 Island CDC42EP1, body No  102  4.3.4 Cell-specific DNA methylation patterns differ between term and preterm births After establishing that there are cell-specific DNAm differences between preterm and term births, I next investigated whether prematurity affects cell-type differences in DNAm. Linear modelling revealed that nRBCs were the most distinct cell type in term cell populations, whereas T cells were the most epigenetically distinct of the preterm cell populations (Table 4.5). The relatively low number of monocyte- and granulocyte-DM sites in both GA groups was likely because these cell types are both of the myeloid lineage and thus epigenetically similar, in contrast to T cells and nRBCs, which are the only representatives of their respective hematopoietic lineages. In the WBC populations, the number of cell-specific DM sites does not change drastically between preterm and term samples. Additionally, the majority of cell type-DM sites for WBCs are identified in both preterm and term samples (ranging from 63-87% of the DM sites) (Table 4.5). In contrast, the number of nRBC-DM sites nearly tripled between the preterm and term samples, increasing from 9,056 preterm nRBC-DM sites to 26,176 term nRBC-DM sites. This large change coincides with earlier observations of increased hypomethylation in term nRBCs relative to their preterm counterparts (Figure 4.1B), which made term nRBCs more distinct from WBCs.  Table 4.5 Number of cell type-DM sites (FDR <5%, |Δβ| >0.20) within preterm samples, within term samples, and in common between the two GA groups. Common percentages reported relative to number of preterm hits first, then number of term hits.  T cells Granulocytes Monocytes nRBCs Preterm 12,974 1,410 1,665 9,056 Term 12,662 1,900 1,508 26,176 Common 10,991 (85%, 87%) 1,201 (85%, 63%) 1,221 (73%, 81%) 7,645 (84%, 29%) 103  When these cell-type DM sites are compared between preterm and term samples based on CpG density and the cell type of interest’s relative DNAm – that is, whether the unique cell population has DNAm that is higher, lower, or in between the DNAm levels of the other cell types – there appears to be very little difference in genomic representation or direction of DNAm change, particularly within WBCs (Figure 4.3). In nRBCs, the dramatic increase in the number of CpG sites hypomethylated relative to WBCs occurs largely in regions of low and intermediate CpG density.   Figure 4.3 Cell type-DM sites (FDR <5%, |Δβ| >0.20) within all preterm samples (top row) and within all term samples (bottom row), grouped by CpG density and DNAm relative to other cell types. Int. shore = shore-associated regions of intermediate CpG density. 104  4.3.5 Comparing nRBC DNA methylation changes in preterm birth to changes with nRBC count Analysis of the term nRBC methylome (Section 3.3) revealed an intriguing relationship between nRBC DNAm and the proportion of nRBCs in cord blood, with DNAm decreasing as proportion increased. I suggested that this could be due to either: 1) changes in nRBC maturity, with a greater proportion of mature nRBCs with increasing overall nRBC count; or 2) an increasing proportion of BM-derived nRBCs, reflecting some erythropoiesis-inducing stressor like hypoxia or infection. Since preterm and term nRBCs also showed DNAm differences at CpG sites associated with age and hematopoietic source, I compared the count-related DNAm differences in term nRBCs to prematurity-DM sites to evaluate the epigenetic similarity between these two biological processes. As gestation progresses, the proportion of circulating nRBCs declines31. This is paralleled by widespread reduction in nRBC DNAm between preterm and term samples (Figure 4.1; Table 4.2). However, in term nRBCs, an increasing proportion of nRBCs is associated with reduced DNAm (Figure 3.4A), which contradicts how nRBC proportion changes through gestation relate to DNAm between preterm and term births. This suggests that the nRBC DNAm-nRBC count relationship in term samples is not the same as the epigenetic changes associated with prematurity.  This is supported by the DNAm trends at hematopoietic source-DM sites. At FL-hypomethylated sites there was no significant change in term nRBC DNAm with count, but preterm nRBCs are significantly hypomethylated compared to term nRBCs at these sites (Figure 3.4D; Figure 4.2C). At BM-hypomethylated sites, DNAm decreases with gestation, suggesting increased representation of BM-derived nRBCs relative to FL nRBCs as the fetus approaches 105  term. This same trend in term nRBC DNAm is observed with nRBC count. Thus, it may be that the proportion of BM-derived nRBCs in term infants increases with increasing nRBC count relative to a hematopoietic source other than the FL, for example the placenta. This could have implications for the persistence of nRBCs after birth. If the majority of nRBCs are produced by a fetal organ, like the BM, they are more likely to keep being produced after birth and remain in circulation. In contrast, if nRBCs are mostly produced by the placenta they would not be expected to persist after birth, since the connection to that hematopoietic source is lost. Understanding the source of elevated nRBC counts after birth could shed light on biological mechanisms behind poor neonatal health outcomes, as persistently high nRBC counts are associated with increased risk of neonatal morbidity and mortality71. Comparing the GO analysis of nRBC prematurity-DM sites and nRBC count DMRs indicates that they are related to similar processes, as both were significantly enriched for terms related to the cytoskeleton, Ras and Rho signalling, and muscle and neuron function. It may be that these two characteristics, GA and nRBC count at term, reflect two different types of erythropoietic maturity and thus show different patterns of DNAm change in genes related to similar processes. Specifically, increasing GA appears to be a measure of increased contribution from a more mature hematopoietic source, whereas increasing term nRBC count may reflect a more mature, inactive nRBC functional phenotype.  4.4 Conclusion DNAm differences have previously been observed between preterm and term births in cord blood190,191,254, but this is one of the first studies on the cell-specific epigenetic impact of PTB. Trends in DNAm change for each hematopoietic cell type followed prior studies of 106  terminally-differentiated hematopoietic cells, reflecting the progressive maturation of these cell populations throughout gestation (Table 4.2). The relative number of prematurity-DM sites for each cell population may reflect changes in that cell type’s functional capacity over gestation, with the highly dynamic monocyte population displaying over twice the number of prematurity-DM sites as the more stable T cells. nRBCs showed the most drastic changes in DNAm between preterm and term samples, displaying widespread hypomethylation during gestation. Given that terminal erythroid differentiation is associated with global demethylation66,67, this change likely reflects an increasing proportion of mature erythroblasts in the nRBC population at term. DNAm changes between preterm and term hematopoietic cells appear to reflect the shift from FL to BM as the predominant hematopoietic source (Table 4.3, Figure 4.2C), and may also relate to changes in epigenetic regulation of cell-specific functions (Supplementary Table C.1). 25 CpG sites showed consistent shifts in DNAm between preterm and term birth in all four cell types, with many of these sites located in genes involved in growth and proliferation, hematopoietic lineage commitment, and the cytoskeleton (Table 4.4). These results may have important clinical implications, as they highlight gene regulatory mechanisms on both cell-specific and systemic levels that are involved in neonatal immune system maturity. From these findings there are a variety of candidates to follow up with gene expression and functional studies in preterm hematopoietic cells, to further clarify molecular mechanisms contributing to the altered preterm immune system.    107  Chapter 5: Conclusion In this thesis, I established DNAm profiles of the main neonatal hematopoietic cells, and identified epigenetic differences in these cell populations between infants born preterm versus those born at term. I also documented the cord blood-specific phenomenon of nRBC-WBC contamination during FACS, and demonstrated that this contamination can impact epigenetic and gene expression studies if not eliminated by stringent cell-sorting strategies. The conclusion of this thesis will summarize my findings and their significance, address the strengths and limitations of the studies performed, and outline future directions to expand this research.  5.1 Summary of findings In Chapter 2 of this thesis, I identify a previously unaddressed technical complication in studying hematopoietic cell populations isolated from cord blood, namely the tendency for erythroid-WBC interactions to produce contaminated cell samples. Cell composition is a well-known confounding variable in epigenetic and gene expression studies of mixed cell samples194,210,211. To work around this obstacle, one can either use a deconvolution algorithm to correct for differences in cell composition195,197, or isolate and study one specific cell population. FACS is the gold standard for cell isolation; however, the purity one can achieve by FACS is only as good as the sorting strategy. In cord blood, heterotopic interactions between erythroid cells and WBCs allow erythroid cells to “hitch-hike” into WBC populations during FACS if they are not specifically excluded by negative gating for erythroid-specific markers. This finding is of little concern for studies of adult blood, since adult circulating nRBCs are enucleate and contain relatively low amounts of RNA263,264. However, I show that this cross-contamination can significantly impact both DNAm and gene expression measurements in 108  hematopoietic cell populations isolated from cord blood. Although not evaluated in Chapter 2, contamination of WBC samples with erythroid cells could also impact functional studies, as RBCs have been shown to have a variety of effects on proliferation and function in T cells, B cells, and dendritic cells93-96. These findings are of significance for any future studies of specific cord blood cell populations, which should consider implementing an erythroid exclusion step in sorting strategies. I also provide erythroid-specific DNAm markers (FDR <5%, |Δβ| from all WBCs >0.50) as candidates for quick, targeted evaluation of nRBC contamination in WBC populations by pyrosequencing. In Chapter 3, I investigated the DNAm profile of nRBCs collected from term cord blood, and found that this cell type has a distinct methylome from WBCs characterized by intermediate DNAm in regions where other hematopoietic cells are generally methylated. Previous studies of erythropoiesis in vitro from HSCs derived from murine fetal liver and adult bone marrow observed global demethylation. Shearstone et al.66 asserted that the process was passive, and were unable to associate changes in murine nRBC DNAm with gene expression. In contrast, Yu et al.67 proposed that demethylation during adult human erythropoiesis is functional, and correlated DNAm changes with gene expression changes in corresponding genes. Based on my findings, I suggest nRBC demethylation is an intermediate of these two extremes, with global loss of DNAm occurring passively as well as accelerated demethylation at biologically relevant regions. By comparing patterns of nRBC hypomethylation to placental hypomethylation, nRBC demethylation appears to be partially stochastic, occurring in areas that are less regulated and more dynamic areas, such as REs and regions of low CpG density. However, the strongest nRBC-specific DNAm loss was observed at CpG sites previously associated with hematopoietic source and age, and DNAm was maintained at known imprinted DMRs. Coupled with the 109  intriguing correlation between nRBC DNAm and nRBC proportion in cord blood, these findings suggest that there is also specificity to nRBC demethylation. Analysis of the relationship between nRBC DNAm and nRBC count revealed demethylation with increasing nRBC count. Specifically, term neonatal nRBCs conformed more to the DNAm profile of nRBCs derived from adult BM239 as their proportion in cord blood increased. This indicates that there is heterogeneity within the nRBC population, which I postulate could reflect two kinds of mixes: 1) by maturity, or 2) by hematopoietic source. If it is the former, higher nRBC counts appear to be associated with a greater proportion of mature nRBCs, based on both my analyses and prior findings of global demethylation with erythropoiesis66,67. Alternatively, this trend in nRBC DNAm could reflect variation in hematopoietic source, with a greater number of nRBCs derived from the BM as count increases. Since BM fenestrations are typically too small to allow nRBCs to enter circulation, increased contribution of nRBCs from this organ would likely be due to some erythropoiesis-inducing stressor, like hypoxia or infection.  nRBC DNAm as a marker of hematopoietic source could also reflect the persistence of nRBCs after birth: if the majority of nRBCs are produced by a fetal organ like the BM, they are more likely to stay in circulation than nRBCs produced by the placenta, as the connection to that hematopoietic organ is lost after parturition. Persistence of nRBCs after birth is linked to increased risk of neonatal morbidity and mortality71; these results indicate that there may be epigenetic markers of high-risk infants, which could be useful for early health care monitoring or intervention. DMR analysis of nRBCs showed that nRBC-specific DNAm associated with nRBC count was significantly enriched for genes related to erythroid maturation and enucleation 110  processes, such as cytoskeletal structure and RNAi. Whether these findings relate to differences in maturity in nRBCs derived from the same source, or variability in hematopoietic sources with different thresholds of nRBC maturity before release into circulation, remains to be seen. However, these findings reveal that epigenetic marks in nRBCs are not random, and hint that more sophisticated gene regulatory processes may be occurring in these cells. In Chapter 4, I evaluated DNAm differences between term and preterm infants in the major hematopoietic cells of the neonatal immune system: T cells, granulocytes, monocytes, and nRBCs. The preterm immune system differs from that of the term neonate in both cell composition and cell function, resulting in heightened vulnerability to infection in preterm infants. Some of these deviations in immune cell function are mediated by epigenetic mechanisms, such as the limited dendritic cell production of the p35 IL-12 subunit and p40 IL-23 subunit5,155 and reduced expression of TLR36. Thus I wanted to compare DNAm profiles of term hematopoietic cells and their preterm counterparts, to identify other pathways displaying GA-dependent epigenetic regulation and therefore likely deficient in preterm infants. Previous DNAm studies using cord blood have identified significant differences between preterm and term infants190,191,254; however, interpretation of these studies is limited by the confounding factor of cord blood cell composition. This is the first evaluation of multiple hematopoietic cell types isolated from the same individuals, which allowed for the identification of cell-specific epigenetic changes with prematurity. Each cell type differed in the number of prematurity-associated DNAm differences (Table 4.2), which I suggest could reflect how premature birth impacts that cell population’s development. For example, preterm monocytes display reduced pro-inflammatory and antiviral cytokine production131,150-154, whereas preterm T cells do not show obvious functional differences from their term counterparts154,256; this is 111  paralleled by a greater number of prematurity-DM sites in monocytes (692) than in T cells (273). Since granulocytes were found to have 987 prematurity-DM sites, they are a strong candidate for further functional study. Their relatively high degree of differential methylation suggests that they undergo drastic phenotypic changes across late gestation. Alternatively, this differential methylation could reflect a change in granulocyte cell composition, as neutrophils are one of the last cell types produced in fetal hematopoietic development55. My findings highlight pathways with potentially altered gene regulation that are unique to each cell type. Some of these findings were enigmatic, such as the enrichment for DM sites in genes associated with embryonic placental development in T cells and with hair and dermal development in monocytes. However, in granulocytes, prematurity-DM sites were significantly enriched for genes associated with the Ras-Raf-MEK-ERK cascade; changes in gene regulation in this cascade has been associated with impaired NET formation and respiratory burst in neutrophils258,259, both of which are also deficient in preterm birth. The prevalence of DM sites in genes associated with these functions could reflect either reduced functional ability in preterm neutrophils, or a low proportion of neutrophils within the preterm granulocyte population. In nRBCs, prematurity-associated DNAm changes were widespread and similar in biological association to those identified in the term nRBC DNAm-count relationship, including significantly enriched terms related to the cytoskeleton, membrane composition and cell-cell junctions, and motility. As with the term nRBCs, this may reflect a change in the predominant hematopoietic source – although in this case, due to the natural shift from FL to BM that occurs in late gestation – or it may reflect a change in the overall maturity of the nRBC population. Additionally, prematurity-associated DNAm changes common to all cell types were identified. With many of these changes occurring in genes related to growth and proliferation, lineage 112  commitment, and the cytoskeleton, these common prematurity-DM sites may reflect root causes of aberrant immune responses in preterm infants. These findings eliminate the question of cell composition differences between cord blood samples from preterm and term births and reveal that there are prematurity-associated DNAm changes in each hematopoietic cell type. Some of these DNAm differences are shared and some are unique to that cell population, suggesting that DNAm marks both systemic and cell-specific differences in maturity, and thus may also relate to neonatal immune function.  5.2 Strengths and limitations The main strength of this thesis is that DNAm analysis was performed on isolated hematopoietic cell populations, rather than whole cord blood. DNAm is closely associated with hematopoietic lineage commitment, and because of that cell composition is one of the greatest confounding variables in epigenetic studies of blood. Although there are deconvolution algorithms available to correct for differences in cell composition in adult blood195,197, these algorithms do not perform as well in cord blood, which is likely largely attributable to the impact of nRBCs on whole cord blood DNAm levels. Our cell-specific approach avoided this confounding factor, plus allowed for the analysis of individual cell types. Not only were isolated cell populations used, but they were obtained from the same individuals for all term samples. Collecting “full sets” of the four cell populations was not always possible in preterm cord blood samples, due to complications such as low blood volume and variable cell composition. The main advantage of matched samples is that it limited the impact of inter-individual variability. It also allowed for identification and removal of 450K 113  array probes targeting epipolymorphisms, in which DNAm appears to be influenced by genetic differences rather than the variable of interest. Another strength of this study is the use of the 450K array for DNAm measurements. This method provides a genome-wide look at DNAm, with high coverage of RefSeq genes and CGIs (99% and 96%, respectively)193. The 450K array also covers enhancer regions and CpG shores and shelves, all of which are gaining attention as more dynamically methylated regions with greater epigenetic associations with lineage commitment3,8. Additionally, extended annotation of the array allows for more informed probe filtering, including probes that cross-hybridize with other regions or probes that contain single nucleotide polymorphisms at the target CpG, as well as more extensive analysis of the genomic context of CpG sites, such as relationship with the nearest gene203. This thesis also was subject to several limitations, the greatest of which was small sample size. Chapter 2 compared DNAm profiles in hematopoietic cells collected from 5 individuals by a “standard” approach to those collected from 7 additional subjects by a “stringent” approach; Chapter 3 used the data from the same 7 “stringent” subjects of Chapter 2; and Chapter 4 compared DNAm profiles in hematopoietic cells collected from 5 preterm subjects to those isolated from 5 term subjects. This small sample size limits a given study’s power for detecting differential methylation. In the case of cell-type differences (Chapter 3), epigenetic differences are typically so great that this is not a major concern. However, differential methylation associated with prematurity is on a smaller scale, and may have led to an underestimation of prematurity-DM sites. Unfortunately, we did not anticipate the relationship between nRBC DNAm and nRBC count described in Chapter 3 and did not obtain complete blood counts for the samples collected 114  in the preterm analysis of Chapter 4. Having blood count data for the preterm samples, as well as the additional term samples, would have provided a way of directly comparing the prematurity-associated DNAm differences in nRBCs with the findings of the nRBC count analysis from Chapter 3, rather than making speculative parallels. It is also important to acknowledge that, although the 450K array does provide a genome-wide perspective on DNAm, this is only one aspect of epigenetic regulation. Other changes, such as histone marks, may yield more information about differences between hematopoietic cells and changes with maturity. This may be especially important in nRBCs, as HDAC recruitment is key for murine erythroid differentiation, and significant increases in histone H3(K9) demethylation and decreases in histone H4(K12) acetylation have been observed in murine erythropoiesis68,265. Finally, there is the complication of heterogeneity in PTB. PTB has three main etiologies: spontaneous labour with intact membranes, preterm pre-labour rupture of membranes, and medically indicated PTBs. We did not have this information for our preterm subjects, so we could not use a specific etiology as inclusion criteria nor could I factor for it during analysis. This potential variable in the preterm cases could impact DNAm, and thus may have limited our ability to uncover prematurity-associated DNAm in Chapter 4.  5.3 Future directions This thesis opens up multiple avenues for further study. I believe that two major routes of interest are 1) clarifying how nRBC DNAm varies between fetal hematopoietic sources, and applying those findings to both the putative immune role of fetal nRBCs and the likelihood that nRBCs will persist after birth; and 2) functional follow-up studies of the differential methylation observed in premature hematopoietic cell populations. 115  To address the first point, identifying DNAm markers associated with distinct fetal hematopoietic sources would reveal whether variability in term nRBC DNAm is influenced by heterogeneity in hematopoietic source. This study could be done in a similar manner to that of Lessard et al.239, in which CD34+ HSCs are collected from the organs of interest. The difference would be that only fetal samples would be used, to avoid the confounding influence of age introduced by collecting BM HSCs from adult. HSCs could be collected from the placenta, spleen, liver, and BM. Identifying DNAm patterns specific to each of these sources may reveal epigenetic markers for infants at high risk of persistently high nRBC count, which is itself associated with increased risk of morbidity and mortality71. Since the nRBC preterm and term DNAm trends observed at the source-DM sites were also exhibited by WBCs, CpG sites identified in this study may also be applicable to WBCs. In addition to epigenetic studies of fetal nRBCs isolated from different hematopoietic sources, functional studies would also be useful. For example, one could follow the lead of Elahi et al.93 and co-culture adult immune cells with these distinct fetal nRBC populations, then assess adult immune cell function. This may reveal that nRBCs derived from certain hematopoietic organs are more immunosuppressive than others. Since the placenta plays a variety of immunomodulatory roles throughout gestation, I predict that placental-derived nRBCs are the most likely to have immunosuppressive function. As BM is the most “mature” hematopoietic organ and continues to be the predominant source of hematopoietic cells throughout adult life, this organ is the most likely to produce relatively inactive nRBCs. It would also be informative to compare results from the ex vivo study described above to DNAm analyses of nRBCs from infants with persistently high circulating nRBCs, versus control infants that rapidly clear nRBCs from circulation. Collection of blood samples and complete blood counts would be acquired both at birth (cord blood) and ~4 days after birth (peripheral 116  blood), at which point nRBCs are typically no longer present in healthy term infants31. Samples would be classified based on whether or not nRBC count declined in the 4-day period, and DNAm would be measured in nRBCs collected from the cord blood. Not only could differential methylation studies be performed between these two groups, but DNAm could also be compared at the hematopoietic source signature sites defined by the previously-described study. This may reveal whether “persistent nRBCs” can be linked back to a particular fetal hematopoietic organ, furthering our understanding of nRBC development and their role in fetal and neonatal circulation. Since the persistence of nRBCs in neonates is rare, this study would need to be performed on either a large cohort or a high-risk cohort, such as infants born preterm.  To address the second point, gene expression and functional studies of preterm hematopoietic cells should be performed as a follow up on the prematurity-DM sites identified in this thesis. Primary focus should be on those most likely to relate to cell function; in my opinion, these are the GO pathways associated with granulocyte prematurity-DM sites, and the genes containing prematurity-DM sites common to all four cell populations. Although the nRBC-specific prematurity-DM sites also yielded interesting results, the hematopoietic source studies described above may also explain some of the prematurity-associated DNAm patterns in nRBCs. Specifically, granulocyte candidate genes related to the Ras-Raf-MAK-ERK pathway that contained prematurity-DM sites, such as CSK, DAB2IP, and LYN, should be evaluated for differential expression between preterm and term neutrophils. This should be coupled with targeted DNAm measurements in these genes to verify the findings of this thesis. Preterm and term neutrophils could additionally be compared by functional capacities related to this cascade, including NET formation and respiratory burst258,259. Of the genes containing common prematurity-DM sites, some of the most promising candidates for follow-up include tumour 117  suppressors STK10, RARRES3, and FBXO31; the oncogene WWTR1; CDC42EP1, which encodes a Rho GTPase; and CLIP2, which encodes a cytoplasmic linker protein. These genes appear likely to impact hematopoietic cell growth, thus studies relating DNAm at these candidate sites to the proliferative capacity of hematopoietic cells may be revealing. These analyses have the potential to identify prematurity-associated functional differences affecting multiple hematopoietic cell populations in neonatal immune system.  5.4 Conclusion Neonates have a uniquely-structured immune system that increases their risk of infection. These differences in immunity are exaggerated in infants born preterm, as is their vulnerability to infection. As epigenetic marks are closely linked to hematopoietic lineage commitment and also have been associated with gestational age, preterm birth, and fetal immune cell function, DNAm patterns in key hematopoietic cells in neonates from term and preterm births may reveal molecular mechanisms regulating their altered immune structure. This thesis contributes to current understanding of the neonatal immune system by 1) revealing intriguing epigenetic complexity in the nRBC population, suggestive of a regulated rather than stochastic DNAm profile; and 2) identifying differential methylation associated with PTB in major hematopoietic cell populations. Additionally, adjustments to cell sorting techniques are described to prevent contamination of cell populations isolated from cord blood, to improve the quality of future studies in the field. 118  References 1. Blencowe H, Cousens S, Chou D, et al. Born too soon: The global epidemiology of 15 million preterm births. Reprod Health. 2013;10 Suppl 1:S2-4755-10-S1-S2. Epub 2013 Nov 15. 2. Menon R, Conneely KN, Smith AK. DNA methylation: An epigenetic risk factor in preterm birth. Reprod Sci. 2012;19(1):6-13. 3. Ji H, Ehrlich LI, Seita J, et al. Comprehensive methylome map of lineage commitment from haematopoietic progenitors. Nature. 2010;467(7313):338-342. 4. Cedar H, Bergman Y. Epigenetics of haematopoietic cell development. Nat Rev Immunol. 2011;11(7):478-488. 5. Goriely S, Van Lint C, Dadkhah R, et al. A defect in nucleosome remodeling prevents IL-12(p35) gene transcription in neonatal dendritic cells. J Exp Med. 2004;199(7):1011-1016. 6. Porras A, Kozar S, Russanova V, et al. Developmental and epigenetic regulation of the human TLR3 gene. Mol Immunol. 2008;46(1):27-36. 7. Morrison SJ, Weissman IL. The long-term repopulating subset of hematopoietic stem cells is deterministic and isolatable by phenotype. Immunity. 1994;1(8):661-673. 8. Alvarez-Errico D, Vento-Tormo R, Sieweke M, Ballestar E. Epigenetic control of myeloid cell differentiation, identity and function. Nat Rev Immunol. 2015;15(1):7-17. 9. Dorshkind K. Not a split decision for human hematopoiesis. Nat Immunol. 2010;11(7):569-570. 10. Adolfsson J, Mansson R, Buza-Vidas N, et al. Identification of Flt3+ lympho-myeloid stem cells lacking erythro-megakaryocytic potential a revised road map for adult blood lineage commitment. Cell. 2005;121(2):295-306. 11. Challen GA, Boles NC, Chambers SM, Goodell MA. Distinct hematopoietic stem cell subtypes are differentially regulated by TGF-beta1. Cell Stem Cell. 2010;6(3):265-278. 12. Laslo P, Pongubala JM, Lancki DW, Singh H. Gene regulatory networks directing myeloid and lymphoid cell fates within the immune system. Semin Immunol. 2008;20(4):228-235. 13. Novershtern N, Subramanian A, Lawton LN, et al. Densely interconnected transcriptional circuits control cell states in human hematopoiesis. Cell. 2011;144(2):296-309. 119  14. Doulatov S, Notta F, Eppert K, Nguyen LT, Ohashi PS, Dick JE. Revised map of the human progenitor hierarchy shows the origin of macrophages and dendritic cells in early lymphoid development. Nat Immunol. 2010;11(7):585-593. 15. Alberts B, Johnson A, Lewis J, Raff M, Roberts K, Walter P. Chapter 24, The adaptive immune system . In: Molecular Biology of the Cell. 4th edition ed. New York: Garland Science; 2002. 16. Corbett NP, Blimkie D, Ho KC, et al. Ontogeny of toll-like receptor mediated cytokine responses of human blood mononuclear cells. PLoS One. 2010;5(11):e15041. 17. Strunk T, Currie A, Richmond P, Simmer K, Burgner D. Innate immunity in human newborn infants: Prematurity means more than immaturity. J Matern Fetal Neonatal Med. 2011;24(1):25-31. 18. Kollmann TR, Crabtree J, Rein-Weston A, et al. Neonatal innate TLR-mediated responses are distinct from those of adults. J Immunol. 2009;183(11):7150-7160. 19. Netea MG, Quintin J, van der Meer JW. Trained immunity: A memory for innate host defense. Cell Host Microbe. 2011;9(5):355-361. 20. National Institute of Allergy and Infectious Diseases. Immune cells. https://www.niaid.nih.gov/topics/immunesystem/Pages/immuneCells.aspx. Accessed December 15, 2015. 21. Park YM, Bochner BS. Eosinophil survival and apoptosis in health and disease. Allergy Asthma Immunol Res. 2010;2(2):87-101. 22. Min B, Brown MA, Legros G. Understanding the roles of basophils: Breaking dawn. Immunology. 2012;135(3):192-197. 23. Epelman S, Lavine KJ, Randolph GJ. Origin and functions of tissue macrophages. Immunity. 2014;41(1):21-35. 24. Holt PG, Jones CA. The development of the immune system during pregnancy and early life. Allergy. 2000;55(8):688-697. 25. Jacobson A, Bell F, Lejarcegui N, Mitchell C, Frenkel L, Horton H. Healthy neonates possess a CD56-negative NK cell population with reduced anti-viral activity. PLoS One. 2013;8(6):e67700. 26. Miller DL, Hiravonen T, Gitlin D. Synthesis of IgE by the human conceptus. J Allergy Clin Immunol. 1973;52(3):182-188. 120  27. de Alarcon PA, Werner EJ. Neonatal hematology. Cambridge: Cambridge University Press; 2005. 28. Adkins B, Leclerc C, Marshall-Clarke S. Neonatal adaptive immunity comes of age. Nat Rev Immunol. 2004;4(7):553-564. 29. Romagnani S. Th1/Th2 cells. Inflamm Bowel Dis. 1999;5(4):285-294. 30. Curtis MM, Way SS. Interleukin-17 in host defence against bacterial, mycobacterial and fungal pathogens. Immunology. 2009;126(2):177-185. 31. Hermansen MC. Nucleated red blood cells in the fetus and newborn. Arch Dis Child Fetal Neonatal Ed. 2001;84(3):F211-5. 32. Milcic TL. The complete blood count. Neonatal Netw. 2010;29(2):109-115. 33. Perrone S, Vezzosi P, Longini M, et al. Nucleated red blood cell count in term and preterm newborns: Reference values at birth. Arch Dis Child Fetal Neonatal Ed. 2005;90(2):F174-5. 34. Hulstaert F, Hannet I, Deneys V, et al. Age-related changes in human blood lymphocyte subpopulations. II. Varying kinetics of percentage and absolute count measurements. Clin Immunol Immunopathol. 1994;70(2):152-158. 35. Adeli K, Raizman JE, Chen Y, et al. Complex biological profile of hematologic markers across pediatric, adult, and geriatric ages: Establishment of robust pediatric and adult reference intervals on the basis of the Canadian health measures survey. Clin Chem. 2015;61(8):1075-1086. 36. Walker JC, Smolders MA, Gemen EF, Antonius TA, Leuvenink J, de Vries E. Development of lymphocyte subpopulations in preterm infants. Scand J Immunol. 2011;73(1):53-58. 37. Lo YM, Lo ES, Watson N, et al. Two-way cell traffic between mother and fetus: Biologic and clinical implications. Blood. 1996;88(11):4390-4395. 38. Gantt S, Gervassi A, Jaspan H, Horton H. The role of myeloid-derived suppressor cells in immune ontogeny. Front Immunol. 2014;5:387. 39. Mikkola HK, Gekas C, Orkin SH, Dieterlen-Lievre F. Placenta as a site for hematopoietic stem cell development. Exp Hematol. 2005;33(9):1048-1054. 40. Dzierzak E, Speck NA. Of lineage and legacy: The development of mammalian hematopoietic stem cells. Nat Immunol. 2008;9(2):129-136. 121  41. Robin C, Bollerot K, Mendes S, et al. Human placenta is a potent hematopoietic niche containing hematopoietic stem and progenitor cells throughout development. Cell Stem Cell. 2009;5(4):385-395. 42. Medvinsky A, Rybtsov S, Taoudi S. Embryonic origin of the adult hematopoietic system: Advances and questions. Development. 2011;138(6):1017-1031. 43. Baron MH. Embryonic origins of mammalian hematopoiesis. Exp Hematol. 2003;31(12):1160-1169. 44. Segel GB, Palis J. Hematology of the fetus and newborn. In: Kaushansky K, Lichtman M, Beutler E, Kipps T, Prchal J, Seligsohn U, eds. Williams Hematology. 8th ed. New York: McGraw-Hill; 2010. 45. Dommergues M, Aubeny E, Dumez Y, Durandy A, Coulombel L. Hematopoiesis in the human yolk sac: Quantitation of erythroid and granulopoietic progenitors between 3.5 and 8 weeks of development. Bone Marrow Transplant. 1992;9 Suppl 1:23-27. 46. Ivanovs A, Rybtsov S, Welch L, Anderson RA, Turner ML, Medvinsky A. Highly potent human hematopoietic stem cells first emege in the intraembryonic aorta-gonad-mesonephros region. J Exp Med. 2011;208:2417-2427. 47. Barcena A, Muench MO, Kapidzic M, Fisher SJ. A new role for the human placenta as a hematopoietic site throughout gestation. Reprod Sci. 2009;16(2):178-187. 48. Dzierzak E, Robin C. Placenta as a source of hematopoietic stem cells. Trends Mol Med. 2010;16(8):361-367. 49. Challier JC, Galtier M, Cortez A, Bintein T, Rabreau M, Uzan S. Immunocytological evidence for hematopoiesis in the early human placenta. Placenta. 2005;26(4):282-288. 50. Van Handel B, Prashad SL, Hassanzadeh-Kiabi N, et al. The first trimester human placenta is a site for terminal maturation of primitive erythroid cells. Blood. 2010;116(17):3321-3330. 51. Gekas C, Dieterlen-Lievre F, Orkin SH, Mikkola HK. The placenta is a niche for hematopoietic stem cells. Dev Cell. 2005;8(3):365-375. 52. Haynes BF, Martin ME, Kay HH, Kurtzberg J. Early events in human T cell ontogeny. phenotypic characterization and immunohistologic localization of T cell precursors in early human fetal tissues. J Exp Med. 1988;168(3):1061-1080. 53. Calhoun DA, Li Y, Braylan RC, Christensen RD. Assessment of the contribution of the spleen to granulocytopoiesis and erythropoiesis of the mid-gestation human fetus. Early Hum Dev. 1996;46(3):217-227. 122  54. Tavassoli M. Embryonic and fetal hemopoiesis: An overview. Blood Cells. 1991;17(2):269-81; discussion 282-6. 55. Proytcheva MA. Issues in neonatal cellular analysis. Am J Clin Pathol. 2009;131(4):560-573. 56. Manroe BL, Weinberg AG, Rosenfeld CR, Browne R. The neonatal blood count in health and disease. I. Reference values for neutrophilic cells. J Pediatr. 1979;95(1):89-98. 57. Dowling DJ, Levy O. Ontogeny of early life immunity. Trends Immunol. 2014;35(7):299-310. 58. Mold JE, Michaelsson J, Burt TD, et al. Maternal alloantigens promote the development of tolerogenic fetal regulatory T cells in utero. Science. 2008;322(5907):1562-1565. 59. Gasparoni A, Ciardelli L, Avanzini A, et al. Age-related changes in intracellular TH1/TH2 cytokine production, immunoproliferative T lymphocyte response and natural killer cell activity in newborns, children and adults. Biol Neonate. 2003;84(4):297-303. 60. Antin JH, Emerson SG, Martin P, Gadol N, Ault KA. Leu-1+ (CD5+) B cells. A major lymphoid subpopulation in human fetal spleen: Phenotypic and functional studies. J Immunol. 1986;136(2):505-510. 61. Upham JW, Rate A, Rowe J, Kusel M, Sly PD, Holt PG. Dendritic cell immaturity during infancy restricts the capacity to express vaccine-specific T-cell memory. Infect Immun. 2006;74(2):1106-1112. 62. Palis J. Primitive and definitive erythropoiesis in mammals. Front Physiol. 2014;5:3. 63. Liu J, Zhang J, Ginzburg Y, et al. Quantitative analysis of murine terminal erythroid differentiation in vivo: Novel method to study normal and disordered erythropoiesis. Blood. 2013;121(8):e43-9. 64. Chen K, Liu J, Heck S, Chasis JA, An X, Mohandas N. Resolving the distinct stages in erythroid differentiation based on dynamic changes in membrane protein expression during erythropoiesis. Proc Natl Acad Sci U S A. 2009;106(41):17413-17418. 65. Keerthivasan G, Wickrema A, Crispino JD. Erythroblast enucleation. Stem Cells Int. 2011;2011:139851. 66. Shearstone JR, Pop R, Bock C, Boyle P, Meissner A, Socolovsky M. Global DNA demethylation during mouse erythropoiesis in vivo. Science. 2011;334(6057):799-802. 67. Yu Y, Mo Y, Ebenezer D, et al. High resolution methylome analysis reveals widespread functional hypomethylation during adult human erythropoiesis. J Biol Chem. 2013;288(13):8805-8814. 123  68. Ji P, Murata-Hori M, Lodish HF. Formation of mammalian erythrocytes: Chromatin condensation and enucleation. Trends Cell Biol. 2011;21(7):409-415. 69. Hanion-Lundberg KM, Kirby RS, Gandhi S, Broekhuizen FF. Nucleated red blood cells in cord blood of singleton term neonates. Am J Obstet Gynecol. 1997;176(6):1149-54; discussion 1154-6. 70. Snijders RJ, Abbas A, Melby O, Ireland RM, Nicolaides KH. Fetal plasma erythropoietin concentration in severe growth retardation. Am J Obstet Gynecol. 1993;168(2):615-619. 71. Baschat AA, Gungor S, Kush ML, Berg C, Gembruch U, Harman CR. Nucleated red blood cell counts in the first week of life: A critical appraisal of relationships with perinatal outcome in preterm growth-restricted neonates. Am J Obstet Gynecol. 2007;197(3):286.e1-286.e8. 72. Aali BS, Malekpour R, Sedig F, Safa A. Comparison of maternal and cord blood nucleated red blood cell count between pre-eclamptic and healthy women. J Obstet Gynaecol Res. 2007;33(3):274-278. 73. Yeruchimovich M, Dollberg S, Green DW, Mimouni FB. Nucleated red blood cells in infants of smoking mothers. Obstet Gynecol. 1999;93(3):403-406. 74. Yeruchimovich M, Mimouni FB, Green DW, Dollberg S. Nucleated red blood cells in healthy infants of women with gestational diabetes. Obstet Gynecol. 2000;95(1):84-86. 75. Stallmach T, Karolyi L, Lichtlen P, et al. Fetuses from preeclamptic mothers show reduced hepatic erythropoiesis. Pediatr Res. 1998;43(3):349-354. 76. Baschat AA. Fetal responses to placental insufficiency: An update. BJOG. 2004;111(10):1031-1041. 77. Constantino BT, Cogionis B. Nucleated RBCs - significance in the peripheral blood film. Laboratory Medicine. 2000;31:223-229. 78. Green DW, Hendon B, Mimouni FB. Nucleated erythrocytes and intraventricular hemorrhage in preterm neonates. Pediatrics. 1995;96(3 Pt 1):475-478. 79. Buonocore G, Perrone S, Gioia D, et al. Nucleated red blood cell count at birth as an index of perinatal brain damage. Am J Obstet Gynecol. 1999;181(6):1500-1505. 80. Silva AM, Smith RN, Lehmann CU, Johnson EA, Holcroft CJ, Graham EM. Neonatal nucleated red blood cells and the prediction of cerebral white matter injury in preterm infants. Obstet Gynecol. 2006;107(3):550-556. 124  81. Steurer MA, Berger TM. Massively elevated nucleated red blood cells and cerebral or pulmonary hemorrhage in severely growth-restricted infants--is there more than coincidence? Neonatology. 2008;94(4):314-319. 82. Wirbelauer J, Thomas W, Speer CP. Response of leukocytes and nucleated red blood cells in very low-birth weight preterm infants after exposure to intrauterine inflammation. J Matern Fetal Neonatal Med. 2011;24(2):348-353. 83. Redline RW. Elevated circulating fetal nucleated red blood cells and placental pathology in term infants who develop cerebral palsy. Hum Pathol. 2008;39(9):1378-1384. 84. Morera D, Roher N, Ribas L, et al. RNA-seq reveals an integrated immune response in nucleated erythrocytes. PLoS One. 2011;6(10):e26998. 85. Rodriguez MF, Wiens GD, Purcell MK, Palti Y. Characterization of toll-like receptor 3 gene in rainbow trout (oncorhynchus mykiss). Immunogenetics. 2005;57(7):510-519. 86. St Paul M, Paolucci S, Barjesteh N, Wood RD, Sharif S. Chicken erythrocytes respond to toll-like receptor ligands by up-regulating cytokine transcripts. Res Vet Sci. 2013;95(1):87-91. 87. Elahi S. New insight into an old concept: Role of immature erythroid cells in immune pathogenesis of neonatal infection. Front Immunol. 2014;5:376. 88. de Macario EC, Macario AJL. A new kind of immunosuppression associated with erythropoiesis. Immunology Letters. 1979;1:23-26. 89. Seledtsov VI, Seledtsova GV, Samarin DM, Taraban VY, Sennikov SV, Kozlov VA. Characterization of erythroid cell-derived natural suppressor activity. Immunobiology. 1998;198(4):361-374. 90. Fonseca AM, Porto G, Uchida K, Arosa FA. Red blood cells inhibit activation-induced cell death and oxidative stress in human peripheral blood T lymphocytes. Blood. 2001;97(10):3152-3160. 91. Arosa FA, Pereira CF, Fonseca AM. Red blood cells as modulators of T cell growth and survival. Curr Pharm Des. 2004;10(2):191-201. 92. Aoshiba K, Nakajima Y, Yasui S, Tamaoki J, Nagai A. Red blood cells inhibit apoptosis of human neutrophils. Blood. 1999;93(11):4006-4010. 93. Elahi S, Ertelt JM, Kinder JM, et al. Immunosuppressive CD71+ erythroid cells compromise neonatal host defence against infection. Nature. 2013;504(7478):158-162. 125  94. Mitasov AV, Tsyrlova IG, Kiselev SV, Kozlov VV. Species-nonspecific action of soluble immunosuppressive factor produced by murine immature erythroid cells. Ann N Y Acad Sci. 1991;628:399-409. 95. Seledtsova GV, Seledtsov VI, Samarin DM, et al. Erythroid cells in immunoregulation: Characterization of a novel suppressor factor. Immunol Lett. 2004;93(2-3):171-178. 96. Schakel K, von Kietzell M, Hansel A, et al. Human 6-sulfo LacNAc-expressing dendritic cells are principal producers of early interleukin-12 and are controlled by erythrocytes. Immunity. 2006;24(6):767-777. 97. Sennikov SV, Krysov SV, Injelevskaya TV, Silkov AN, Kozlov VA. Production of cytokines by immature erythroid cells derived from human embryonic liver. Eur Cytokine Netw. 2001;12(2):274-279. 98. Sennikov SV, Injelevskaya TV, Krysov SV, et al. Production of hemo- and immunoregulatory cytokines by erythroblast antigen+ and glycophorin A+ cells from human bone marrow. BMC Cell Biol. 2004;5(1):39. 99. Wynn JL, Scumpia PO, Stocks BT, et al. Neonatal CD71+ erythroid cells do not modify murine sepsis mortality. J Immunol. 2015;195(3):1064-1070. 100. Aagaard K, Ma J, Antony KM, Ganu R, Petrosino J, Versalovic J. The placenta harbors a unique microbiome. Sci Transl Med. 2014;6(237):237ra65. 101. Cao B, Mysorekar IU. Intracellular bacteria in placental basal plate localize to extravillous trophoblasts. Placenta. 2014;35(2):139-142. 102. Zhang X, Mozeleski B, Lemoine S, et al. CD4 T cells with effector memory phenotype and function develop in the sterile environment of the fetus. Sci Transl Med. 2014;6(238):238ra72. 103. Griffin DO, Holodick NE, Rothstein TL. Human B1 cells in umbilical cord and adult peripheral blood express the novel phenotype CD20+ CD27+ CD43+ CD70-. J Exp Med. 2011;208(1):67-80. 104. Mold JE, Venkatasubrahmanyam S, Burt TD, et al. Fetal and adult hematopoietic stem cells give rise to distinct T cell lineages in humans. Science. 2010;330(6011):1695-1699. 105. Sanjeevi CB, Vivekanandan S, Narayanan PR. Fetal response to maternal ascariasis as evidenced by anti Ascaris lumbricoides IgM antibodies in the cord blood. Acta Paediatr Scand. 1991;80(12):1134-1138. 106. Elbou Ould MA, Luton D, Yadini M, et al. Cellular immune response of fetuses to cytomegalovirus. Pediatr Res. 2004;55(2):280-286. 126  107. Gibbons D, Fleming P, Virasami A, et al. Interleukin-8 (CXCL8) production is a signatory T cell effector function of human newborn infants. Nat Med. 2014;20(10):1206-1210. 108. Zhao J, Kim KD, Yang X, Auh S, Fu YX, Tang H. Hyper innate responses in neonates lead to increased morbidity and mortality after infection. Proc Natl Acad Sci U S A. 2008;105(21):7528-7533. 109. Joyner JL, Augustine NH, Taylor KA, La Pine TR, Hill HR. Effects of group B streptococci on cord and adult mononuclear cell interleukin-12 and interferon-y mRNA accumulation and protein secretion. J Infect Dis. 2000;182:974-977. 110. La Pine TR, Joyner JL, Augustine NH, Kwak SD, Hill HR. Defective production of IL-18 and IL-12 by cord blood mononuclear cells influences the T helper-1 interferon gamma response to group B streptococci. Pediatr Res. 2003;54(2):276-281. 111. Lee SM, Suen Y, Chang L, et al. Decreased interleukin-12 (IL-12) from activated cord versus adult peripheral blood mononuclear cells and upregulation of interferon-gamma, natural killer, and lymphokine-activated killer activity by IL-12 in cord blood mononuclear cells. Blood. 1996;88(3):945-954. 112. Dhakal M, Miller MM, Zaghouani AA, Sherman MP, Zaghouani H. Neonatal basophils stifle the function of early-life dendritic cells to curtail Th1 immunity in newborn mice. J Immunol. 2015;195(2):507-518. 113. Levy O, Coughlin M, Cronstein BN, Roy RM, Desai A, Wessels MR. The adenosine system selectively inhibits TLR-mediated TNF-alpha production in the human newborn. J Immunol. 2006;177(3):1956-1966. 114. Vanden Eijnden S, Goriely S, De Wit D, Goldman M, Willems F. Preferential production of the IL-12(p40)/IL-23(p19) heterodimer by dendritic cells from human newborns. Eur J Immunol. 2006;36(1):21-26. 115. Strunk T, Doherty D, Richmond P, et al. Reduced levels of antimicrobial proteins and peptides in human cord blood plasma. Arch Dis Child Fetal Neonatal Ed. 2009;94(3):F230-1. 116. Prosser A, Hibbert J, Strunk T, et al. Phagocytosis of neonatal pathogens by peripheral blood neutrophils and monocytes from newborn preterm and term infants. Pediatr Res. 2013;74(5):503-510. 117. Filias A, Theodorou GL, Mouzopoulou S, Varvarigou AA, Mantagos S, Karakantza M. Phagocytic ability of neutrophils and monocytes in neonates. BMC Pediatr. 2011;11:29-2431-11-29. 127  118. Levy O. Impaired innate immunity at birth: Deficiency of bactericidal/permeability-increasing protein (BPI) in the neutrophils of newborns. Pediatr Res. 2002;51(6):667-669. 119. Levy O. Antimicrobial proteins and peptides: Anti-infective molecules of mammalian leukocytes. J Leukoc Biol. 2004;76(5):909-925. 120. Yost CC, Cody MJ, Harris ES, et al. Impaired neutrophil extracellular trap (NET) formation: A novel innate immune deficiency of human neonates. Blood. 2009;113(25):6419-6427. 121. Leiber A, Graf B, Spring B, et al. Neonatal monocytes express antiapoptotic pattern of Bcl-2 proteins and show diminished apoptosis upon infection with Escherichia coli. Pediatr Res. 2014;76(2):142-149. 122. Roth I, Corry DB, Locksley RM, Abrams JS, Litton MJ, Fisher SJ. Human placental cytotrophoblasts produce the immunosuppressive cytokine interleukin 10. J Exp Med. 1996;184(2):539-548. 123. Piccinni MP, Giudizi MG, Biagiotti R, et al. Progesterone favors the development of human T helper cells producing Th2-type cytokines and promotes both IL-4 production and membrane CD30 expression in established Th1 cell clones. J Immunol. 1995;155(1):128-133. 124. Svensson-Arvelund J, Mehta RB, Lindau R, et al. The human fetal placenta promotes tolerance against the semiallogeneic fetus by inducing regulatory T cells and homeostatic M2 macrophages. J Immunol. 2015;194(4):1534-1544. 125. Badurdeen S, Mulongo M, Berkley JA. Arginine depletion increases susceptibility to serious infections in preterm newborns. Pediatr Res. 2015;77(2):290-297. 126. Yu HR, Kuo HC, Huang LT, et al. L-arginine modulates neonatal lymphocyte proliferation through an interleukin-2 independent pathway. Immunology. 2014;143(2):184-192. 127. Trend S, Strunk T, Hibbert J, et al. Antimicrobial protein and peptide concentrations and activity in human breast milk consumed by preterm infants at risk of late-onset neonatal sepsis. PLoS One. 2015;10(2):e0117038. 128. Trend S, de Jong E, Lloyd ML, et al. Leukocyte populations in human preterm and term breast milk identified by multicolour flow cytometry. PLoS One. 2015;10(8):e0135580. 129. Fulde M, Hornef MW. Maturation of the enteric mucosal innate immune system during the postnatal period. Immunol Rev. 2014;260(1):21-34. 128  130. Martin JA, Hamilton BE, Ventura SJ, Osterman MJ, Mathews TJ. Births: Final data for 2011. Natl Vital Stat Rep. 2013;62(1):1-69, 72. 131. Sharma AA, Jen R, Butler A, Lavoie PM. The developing human preterm neonatal immune system: A case for more research in this area. Clin Immunol. 2012;145(1):61-68. 132. Henderson JJ, McWilliam OA, Newnham JP, Pennell CE. Preterm birth aetiology 2004-2008. Maternal factors associated with three phenotypes: Spontaneous preterm labour, preterm pre-labour rupture of membranes and medically indicated preterm birth. J Matern Fetal Neonatal Med. 2012;25(6):642-647. 133. Zeitlin J, Saurel-Cubizolles MJ, De Mouzon J, et al. Fetal sex and preterm birth: Are males at greater risk? Hum Reprod. 2002;17(10):2762-2768. 134. Marchant EA, Boyce GK, Sadarangani M, Lavoie PM. Neonatal sepsis due to coagulase-negative staphylococci. Clin Dev Immunol. 2013;2013:586076. 135. Stoll BJ, Hansen N, Fanaroff AA, et al. Late-onset sepsis in very low birth weight neonates: The experience of the NICHD neonatal research network. Pediatrics. 2002;110(2 Pt 1):285-291. 136. Takala TI, Makela E, Suominen P, et al. Blood cell and iron status analytes of preterm and full-term infants from 20 weeks onwards during the first year of life. Clin Chem Lab Med. 2010;48(9):1295-1301. 137. Lavoie PM, Lavoie JC, Watson C, Rouleau T, Chang BA, Chessex P. Inflammatory response in preterm infants is induced early in life by oxygen and modulated by total parenteral nutrition. Pediatr Res. 2010;68(3):248-251. 138. Institute of Medicine (US) Committee on Understanding Premature Birth and Assuring Healthy Outcomes. Preterm Birth: Causes, Consequences, Prevention. 2007. 139. Bernsen RM, de Jongste JC, Koes BW, Aardoom HA, van der Wouden JC. Perinatal characteristics and obstetric complications as risk factors for asthma, allergy and eczema at the age of 6 years. Clin Exp Allergy. 2005;35(9):1135-1140. 140. Mai XM, Gaddlin PO, Nilsson L, et al. Asthma, lung function and allergy in 12-year-old children with very low birth weight: A prospective study. Pediatr Allergy Immunol. 2003;14(3):184-192. 141. Yuan W, Basso O, Sorensen HT, Olsen J. Indicators of fetal growth and infectious disease in childhood--a birth cohort with hospitalization as outcome. Eur J Epidemiol. 2001;17(9):829-834. 129  142. Bhutta AT, Cleves MA, Casey PH, Cradock MM, Anand KJ. Cognitive and behavioral outcomes of school-aged children who were born preterm: A meta-analysis. JAMA. 2002;288(6):728-737. 143. Cnattingius S, Hultman CM, Dahl M, Sparen P. Very preterm birth, birth trauma, and the risk of anorexia nervosa among girls. Arch Gen Psychiatry. 1999;56(7):634-638. 144. Monfils Gustafsson W, Josefsson A, Ekholm Selling K, Sydsjo G. Preterm birth or foetal growth impairment and psychiatric hospitalization in adolescence and early adulthood in a Swedish population-based birth cohort. Acta Psychiatr Scand. 2009;119(1):54-61. 145. Koupil I, Leon DA, Lithell HO. Length of gestation is associated with mortality from cerebrovascular disease. J Epidemiol Community Health. 2005;59(6):473-474. 146. Kajantie E, Osmond C, Barker DJ, Eriksson JG. Preterm birth--a risk factor for type 2 diabetes? The Helsinki birth cohort study. Diabetes Care. 2010;33(12):2623-2625. 147. Barker DJ, Gelow J, Thornburg K, Osmond C, Kajantie E, Eriksson JG. The early origins of chronic heart failure: Impaired placental growth and initiation of insulin resistance in childhood. Eur J Heart Fail. 2010;12(8):819-825. 148. Palfi M, Selbing A. Placental transport of maternal immunoglobulin G. Am J Reprod Immunol. 1998;39(1):24-26. 149. Malek A, Sager R, Kuhn P, Nicolaides KH, Schneider H. Evolution of maternofetal transport of immunoglobulins during human pregnancy. Am J Reprod Immunol. 1996;36(5):248-255. 150. Sharma AA, Jen R, Kan B, et al. Impaired NLRP3 inflammasome activity during fetal development regulates IL-1beta production in human monocytes. Eur J Immunol. 2015;45(1):238-249. 151. Forster-Waldl E, Sadeghi K, Tamandl D, et al. Monocyte toll-like receptor 4 expression and LPS-induced cytokine production increase during gestational aging. Pediatr Res. 2005;58(1):121-124. 152. Weatherstone KB, Rich EA. Tumor necrosis factor/cachectin and interleukin-1 secretion by cord blood monocytes from premature and term neonates. Pediatr Res. 1989;25(4):342-346. 153. Strunk T, Prosser A, Levy O, et al. Responsiveness of human monocytes to the commensal bacterium Staphylococcus epidermidis develops late in gestation. Pediatr Res. 2012;72(1):10-18. 130  154. Currie AJ, Curtis S, Strunk T, et al. Preterm infants have deficient monocyte and lymphocyte cytokine responses to group B streptococcus. Infect Immun. 2011;79(4):1588-1596. 155. Lavoie PM, Huang Q, Jolette E, et al. Profound lack of interleukin (IL)-12/IL-23p40 in neonates born early in gestation is associated with an increased risk of sepsis. J Infect Dis. 2010;202(11):1754-1763. 156. Cypowyj S, Picard C, Marodi L, Casanova JL, Puel A. Immunity to infection in IL-17-deficient mice and humans. Eur J Immunol. 2012;42(9):2246-2254. 157. Sharma AA, Jen R, Brant R, et al. Hierarchical maturation of innate immune defences in very preterm neonates. Neonatology. 2014;106(1):1-9. 158. Forestier F, Daffos F, Catherine N, Renard M, Andreux JP. Developmental hematopoiesis in normal human fetal blood. Blood. 1991;77(11):2360-2363. 159. Christensen RD, Del Vecchio A, Henry E. Expected erythrocyte, platelet and neutrophil values for term and preterm neonates. J Matern Fetal Neonatal Med. 2012;25(Suppl 5):77-79. 160. Ziller MJ, Gu H, Muller F, et al. Charting a dynamic DNA methylation landscape of the human genome. Nature. 2013;500(7463):477-481. 161. Hackett JA, Surani MA. DNA methylation dynamics during the mammalian life cycle. Philos Trans R Soc Lond B Biol Sci. 2013;368(1609):20110328. 162. Illingworth R, Kerr A, Desousa D, et al. A novel CpG island set identifies tissue-specific methylation at developmental gene loci. PLoS Biol. 2008;6(1):e22. 163. Stadler MB, Murr R, Burger L, et al. DNA-binding factors shape the mouse methylome at distal regulatory regions. Nature. 2011;480:490-495. 164. Schmidl C, Klug M, Boeld TJ, et al. Lineage-specific DNA methylation in T cells correlates with histone methylation and enhancer activity. Genome Res. 2009;19(7):1165-1174. 165. Song F, Smith JF, Kimura MT, et al. Association of tissue-specific differentially methylated regions (TDMs) with differential gene expression. Proc Natl Acad Sci U S A. 2005;102(9):3336-3341. 166. Weber M, Davies JJ, Wittig D, et al. Chromosome-wide and promoter-specific analyses identify sites of differential DNA methylation in normal and transformed human cells. Nat Genet. 2005;37(8):853-862. 131  167. Maunakea AK, Nagarajan RP, Bilenky M, et al. Conserved role of intragenic DNA methylation in regulating alternative promoters. Nature. 2010;466(7303):253-257. 168. Deaton AM, Bird A. CpG islands and the regulation of transcription. Genes Dev. 2011;25(10):1010-1022. 169. Deaton AM, Webb S, Kerr AR, et al. Cell type-specific DNA methylation at intragenic CpG islands in the immune system. Genome Res. 2011;21(7):1074-1086. 170. Shukla S, Kavak E, Gregory M, et al. CTCF-promoted RNA polymerase II pausing links DNA methylation to splicing. Nature. 2011;479(7371):74-79. 171. Huh I, Zeng J, Park T, Yi SV. DNA methylation and transcriptional noise. Epigenetics Chromatin. 2013;6(1):9-8935-6-9. 172. Brenet F, Moh M, Funk P, et al. DNA methylation of the first exon is tightly linked to transcriptional silencing. PLoS One. 2011;6(1):e14524. 173. Bock C. Analysing and interpreting DNA methylation data. Nat Rev Genet. 2012;13(10):705-719. 174. Broske AM, Vockentanz L, Kharazi S, et al. DNA methylation protects hematopoietic stem cell multipotency from myeloerythroid restriction. Nat Genet. 2009;41(11):1207-1215. 175. Wilson CB, Rowell E, Sekimata M. Epigenetic control of T-helper-cell differentiation. Nat Rev Immunol. 2009;9(2):91-105. 176. Lee DU, Agarwal S, Rao A. Th2 lineage commitment and efficient IL-4 production involves extended demethylation of the IL-4 gene. Immunity. 2002;16(5):649-660. 177. Yu Q, Thieu VT, Kaplan MH. Stat4 limits DNA methyltransferase recruitment and DNA methylation of the IL-18Ralpha gene during Th1 differentiation. EMBO J. 2007;26(8):2052-2060. 178. Mullen AC, Hutchins AS, High FA, et al. Hlx is induced by and genetically interacts with T-bet to promote heritable T(H)1 gene induction. Nat Immunol. 2002;3(7):652-658. 179. Floess S, Freyer J, Siewert C, et al. Epigenetic control of the foxp3 locus in regulatory T cells. PLoS Biol. 2007;5(2):e38. 180. Young HA, Ghosh P, Ye J, et al. Differentiation of the T helper phenotypes by analysis of the methylation state of the IFN-gamma gene. J Immunol. 1994;153(8):3603-3610. 181. Valapour M, Guo J, Schroeder JT, et al. Histone deacetylation inhibits IL4 gene expression in T cells. J Allergy Clin Immunol. 2002;109(2):238-245. 132  182. Sun JC, Beilke JN, Lanier LL. Adaptive immune features of natural killer cells. Nature. 2009;457(7229):557-561. 183. Sun JC, Beilke JN, Lanier LL. Immune memory redefined: Characterizing the longevity of natural killer cells. Immunol Rev. 2010;236:83-94. 184. Kaikkonen MU, Spann NJ, Heinz S, et al. Remodeling of the enhancer landscape during macrophage activation is coupled to enhancer transcription. Mol Cell. 2013;51(3):310-325. 185. Ostuni R, Piccolo V, Barozzi I, et al. Latent enhancers activated by stimulation in differentiated cells. Cell. 2013;152(1-2):157-171. 186. Kleinnijenhuis J, Quintin J, Preijers F, et al. Bacille calmette-guerin induces NOD2-dependent nonspecific protection from reinfection via epigenetic reprogramming of monocytes. Proc Natl Acad Sci U S A. 2012;109(43):17537-17542. 187. Pacis A, Tailleux L, Morin AM, et al. Bacterial infection remodels the DNA methylation landscape of human dendritic cells. Genome Res. 2015;25(12):1801-1811. 188. Schroeder JW, Conneely KN, Cubells JC, et al. Neonatal DNA methylation patterns associate with gestational age. Epigenetics. 2011;6(12):1498-1504. 189. Novakovic B, Yuen RK, Gordon L, et al. Evidence for widespread changes in promoter methylation profile in human placenta in response to increasing gestational age and environmental/stochastic factors. BMC Genomics. 2011;12:529-2164-12-529. 190. Parets SE, Conneely KN, Kilaru V, et al. Fetal DNA methylation associates with early spontaneous preterm birth and gestational age. PLoS One. 2013;8(6):e67489. 191. Cruickshank MN, Oshlack A, Theda C, et al. Analysis of epigenetic changes in survivors of preterm birth reveals the effect of gestational age and evidence for a long term legacy. Genome Med. 2013;5(10):96. 192. Burris HH, Rifas-Shiman SL, Baccarelli A, et al. Associations of LINE-1 DNA methylation with preterm birth in a prospective cohort study. J Dev Orig Health Dis. 2012;3(3):173-181. 193. Bibikova M, Barnes B, Tsan C, et al. High density DNA methylation array with single CpG site resolution. Genomics. 2011;98(4):288-295. 194. Reinius LE, Acevedo N, Joerink M, et al. Differential DNA methylation in purified human blood cells: Implications for cell lineage and studies on disease susceptibility. PLoS One. 2012;7(7):e41361. 133  195. Houseman EA, Accomando WP, Koestler DC, et al. DNA methylation arrays as surrogate measures of cell mixture distribution. BMC Bioinformatics. 2012;13:86-2105-13-86. 196. Koestler DC, Christensen B, Karagas MR, et al. Blood-based profiles of DNA methylation predict the underlying distribution of cell types: A validation analysis. Epigenetics. 2013;8(8):816-826. 197. Houseman EA, Molitor J, Marsit CJ. Reference-free cell mixture adjustments in analysis of DNA methylation data. Bioinformatics. 2014;30(10):1431-1439. 198. Aryee MJ, Jaffe AE, Corrada-Bravo H, et al. Minfi: A flexible and comprehensive Bioconductor package for the analysis of Infinium DNA methylation microarrays. Bioinformatics. 2014. 199. Dedeurwaerder S, Defrance M, Calonne E, Denis H, Sotiriou C, Fuks F. Evaluation of the Infinium methylation 450K technology. Epigenomics. 2011;3(6):771-784. 200. Touleimat N, Tost J. Complete pipeline for Infinium((R)) Human Methylation 450K BeadChip data processing using subset quantile normalization for accurate DNA methylation estimation. Epigenomics. 2012;4(3):325-341. 201. Maksimovic J, Gordon L, Oshlack A. SWAN: Subset-quantile within array normalization for Illumina Infinium HumanMethylation450 BeadChips. Genome Biol. 2012;13(6):R44-2012-13-6-r44. 202. Teschendorff AE, Marabita F, Lechner M, et al. A beta-mixture quantile normalization method for correcting probe design bias in Illumina Infinium 450K DNA methylation data. Bioinformatics. 2013;29(2):189-196. 203. Price ME, Cotton AM, Lam LL, et al. Additional annotation enhances potential for biologically-relevant analysis of the Illumina Infinium HumanMethylation450 BeadChip array. Epigenetics Chromatin. 2013;6(1):4-8935-6-4. 204. Leek JT, Storey JD. Capturing heterogeneity in gene expression studies by surrogate variable analysis. PLoS Genet. 2007;3(9):1724-1735. 205. Johnson WE, Li C, Rabinovic A. Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics. 2007;8(1):118-127. 206. Buhule OD, Minster RL, Hawley NL, et al. Stratified randomization controls better for batch effects in 450K methylation analysis: A cautionary tale. Front Genet. 2014;5:354. 207. Nygaard V, Rodland EA, Hovig E. Methods that remove batch effects while retaining group differences may lead to exaggerated confidence in downstream analyses. Biostatistics. 2016;17(1):29-39. 134  208. Chen L, Kostadima M, Martens JH, et al. Transcriptional diversity during lineage commitment of human blood progenitors. Science. 2014;345(6204):1251033. 209. Simon LM, Edelstein LC, Nagalla S, et al. Human platelet microRNA-mRNA networks associated with age and gender revealed by integrated plateletomics. Blood. 2014;123(16):e37-45. 210. Lam LL, Emberly E, Fraser HB, et al. Factors underlying variable DNA methylation in a human community cohort. Proc Natl Acad Sci U S A. 2012;109 Suppl 2:17253-17260. 211. Whitney AR, Diehn M, Popper SJ, et al. Individuality and variation in gene expression patterns in human blood. Proc Natl Acad Sci U S A. 2003;100(4):1896-1901. 212. Hunt JS, Beck ML, Hardman JT, Tegtmeier GE, Bayer WL. Characterization of human erythrocyte alloantibodies by IgG subclass and monocyte interaction. Am J Clin Pathol. 1980;74(3):259-264. 213. Fuss IJ, Kanof ME, Smith PD, Zola H. Isolation of whole mononuclear cells from peripheral blood and cord blood. Curr Protoc Immunol. 2009;Chapter 7:Unit7.1. 214. Du P, Kibbe WA, Lin SM. Lumi: A pipeline for processing Illumina microarray. Bioinformatics. 2008;24(13):1547-1548. 215. R Core Team. R: A language and environment for statistical computing. Vienna, Austria: the R Foundation for Statistical Computing; 2014. 216. Ritchie ME, Phipson B, Wu D, et al. Limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 2015;43(7):e47. 217. Benjamini Y, Hochberg Y. Controlling the false discovery rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society. 1995;57:289-300. 218. Accomando WP, Wiencke JK, Houseman EA, Nelson HH, Kelsey KT. Quantitative reconstruction of leukocyte subsets using DNA methylation. Genome Biol. 2014;15(3):R50-2014-15-3-r50. 219. Katsura Y. Redefinition of lymphoid progenitors. Nat Rev Immunol. 2002;2(2):127-132. 220. Mansson R, Hultquist A, Luc S, et al. Molecular evidence for hierarchical transcriptional lineage priming in fetal and adult stem cells and multipotent progenitors. Immunity. 2007;26(4):407-419. 221. Yuen RK, Avila L, Penaherrera MS, et al. Human placental-specific epipolymorphism and its association with adverse pregnancy outcomes. PLoS One. 2009;4(10):e7389. 135  222. Rodriguez P, Bonte E, Krijgsveld J, et al. GATA-1 forms distinct activating and repressive complexes in erythroid cells. EMBO J. 2005;24(13):2354-2366. 223. Tsang AP, Visvader JE, Turner CA, et al. FOG, a multitype zinc finger protein, acts as a cofactor for transcription factor GATA-1 in erythroid and megakaryocytic differentiation. Cell. 1997;90(1):109-119. 224. Varricchio L, Dell'Aversana C, Nebbioso A, et al. Identification of NuRSERY, a new functional HDAC complex composed by HDAC5, GATA1, EKLF and pERK present in human erythroid cells. Int J Biochem Cell Biol. 2014;50:112-122. 225. Lawrence T. The nuclear factor NF-kappaB pathway in inflammation. Cold Spring Harb Perspect Biol. 2009;1(6):a001651. 226. Fensterl V, Sen GC. Interferon-induced Ifit proteins: Their role in viral pathogenesis. J Virol. 2015;89(5):2462-2468. 227. de Back DZ, Kostova EB, van Kraaij M, van den Berg TK, van Bruggen R. Of macrophages and red blood cells; a complex love story. Front Physiol. 2014;5:9. 228. Hebbar S, Misha M, Rai L. Significance of maternal and cord blood nucleated red blood cell count in pregnancies complicated by preeclampsia. J Pregnancy. 2014;2014:496416. 229. Salafia CM, Ghidini A, Pezzullo JC, Rosenkrantz TS. Early neonatal nucleated erythrocyte counts in preterm deliveries: Clinical and pathologic correlations. J Soc Gynecol Investig. 1997;4(3):138-143. 230. Dulay AT, Buhimschi IA, Zhao G, et al. Nucleated red blood cells are a direct response to mediators of inflammation in newborns with early-onset neonatal sepsis. Am J Obstet Gynecol. 2008;198(4):426.e1-426.e9. 231. Phelan JP, Ahn MO, Korst LM, Martin GI. Nucleated red blood cells: A marker for fetal asphyxia? Am J Obstet Gynecol. 1995;173(5):1380-1384. 232. Liang P, Song F, Ghosh S, et al. Genome-wide survey reveals dynamic widespread tissue-specific changes in DNA methylation during development. BMC Genomics. 2011;12(1):231-2164-12-231. 233. Schroeder DI, Blair JD, Lott P, et al. The human placenta methylome. Proc Natl Acad Sci U S A. 2013;110(15):6037-6042. 234. Berman BP, Weisenberger DJ, Aman JF, et al. Regions of focal DNA hypermethylation and long-range hypomethylation in colorectal cancer coincide with nuclear lamina-associated domains. Nat Genet. 2011;44(1):40-46. 136  235. Hon GC, Hawkins RD, Caballero OL, et al. Global DNA hypomethylation coupled to repressive chromatin domain formation and gene silencing in breast cancer. Genome Res. 2012;22(2):246-258. 236. Reik W, Dean W, Walter J. Epigenetic reprogramming in mammalian development. Science. 2001;293(5532):1089-1093. 237. Wilson SL, Blair JD, Hogg K, Langlois S, von Dadelszen P, Robinson WP. Placental DNA methylation at term reflects maternal serum levels of INHA and FN1, but not PAPPA, early in pregnancy. BMC Med Genet. 2015;16(1):111-015-0257-z. 238. Heyn H, Li N, Ferreira HJ, et al. Distinct DNA methylomes of newborns and centenarians. Proc Natl Acad Sci U S A. 2012;109(26):10522-10527. 239. Lessard S, Beaudoin M, Benkirane K, Lettre G. Comparison of DNA methylation profiles in human fetal and adult red blood cell progenitors. Genome Med. 2015;7(1):1-014-0122-2. eCollection 2015. 240. Hanna CW, Penaherrera MS, Saadeh H, et al. Pervasive polymorphic imprinted methylation in the human placenta. Genome Res. 2016. 241. Peters TJ, Buckley MJ, Statham AL, et al. De novo identification of differentially methylated regions in the human genome. Epigenetics Chromatin. 2015;8:6-8935-8-6. eCollection 2015. 242. Gillis J, Mistry M, Pavlidis P. Gene function analysis in complex data sets using ErmineJ. Nat Protoc. 2010;5(6):1148-1159. 243. Price EM, Cotton AM, Penaherrera MS, McFadden DE, Kobor MS, Robinson W. Different measures of "genome-wide" DNA methylation exhibit unique properties in placental and somatic tissues. Epigenetics. 2012;7(6):652-663. 244. Liang G, Chan MF, Tomigahara Y, et al. Cooperativity between DNA methyltransferases in the maintenance methylation of repetitive elements. Mol Cell Biol. 2002;22(2):480-491. 245. Novakovic B, Saffery R. The ever growing complexity of placental epigenetics - role in adverse pregnancy outcomes and fetal programming. Placenta. 2012;33(12):959-970. 246. Fraga MF, Ballestar E, Paz MF, et al. Epigenetic differences arise during the lifetime of monozygotic twins. Proc Natl Acad Sci U S A. 2005;102(30):10604-10609. 247. Teschendorff AE, West J, Beck S. Age-associated epigenetic drift: Implications, and a case of epigenetic thrift? Hum Mol Genet. 2013;22(R1):R7-R15. 137  248. Shah S, McRae AF, Marioni RE, et al. Genetic and environmental exposures constrain epigenetic drift over the human life course. Genome Res. 2014;24(11):1725-1733. 249. Pastor WA, Pape UJ, Huang Y, et al. Genome-wide mapping of 5-hydroxymethylcytosine in embryonic stem cells. Nature. 2011;473(7347):394-397. 250. Li W, Liu M. Distribution of 5-hydroxymethylcytosine in different human tissues. J Nucleic Acids. 2011;2011:870726. 251. Haffner MC, Chaux A, Meeker AK, et al. Global 5-hydroxymethylcytosine content is significantly reduced in tissue stem/progenitor cell compartments and in human cancers. Oncotarget. 2011;2(8):627-637. 252. Madzo J, Liu H, Rodriguez A, et al. Hydroxymethylation at gene regulatory regions directs stem/early progenitor cell commitment during erythropoiesis. Cell Rep. 2014;6(1):231-244. 253. Hu J, Liu J, Xue F, et al. Isolation and functional characterization of human erythroblasts at distinct stages: Implications for understanding of normal and disordered erythropoiesis in vivo. Blood. 2013;121:3246-3253. 254. Fernando F, Keijser R, Henneman P, et al. The idiopathic preterm delivery methylation profile in umbilical cord blood DNA. BMC Genomics. 2015;16:736-015-1915-4. 255. Jones PA. Functions of DNA methylation: Islands, start sites, gene bodies and beyond. Nat Rev Genet. 2012;13(7):484-492. 256. Bussel JB, Cunningham-Rundles S, LaGamma EF, Shellabarger M. Analysis of lymphocyte proliferative response subpopulations in very low birth weight infants and during the first 8 weeks of life. Pediatr Res. 1988;23(5):457-462. 257. Peden DB, VanDyke K, Ardekani A, Mullett MD, Myerberg DZ, VanDyke C. Diminished chemiluminescent responses of polymorphonuclear leukocytes in severely and moderately preterm neonates. J Pediatr. 1987;111(6 Pt 1):904-906. 258. Tortorella C, Stella I, Piazzolla G, Simone O, Cappiello V, Antonaci S. Role of defective ERK phosphorylation in the impaired GM-CSF-induced oxidative response of neutrophils in elderly humans. Mech Ageing Dev. 2004;125(8):539-546. 259. Hakkim A, Fuchs TA, Martinez NE, et al. Activation of the Raf-MEK-ERK pathway is required for neutrophil extracellular trap formation. Nat Chem Biol. 2011;7(2):75-77. 260. Martin L, Pingle SC, Hallam DM, Rybak LP, Ramkumar V. Activation of the adenosine A3 receptor in RAW 264.7 cells inhibits lipopolysaccharide-stimulated tumor necrosis 138  factor-alpha release by reducing calcium-dependent activation of nuclear factor-kappaB and extracellular signal-regulated kinase 1/2. J Pharmacol Exp Ther. 2006;316(1):71-78. 261. Roger T, Schneider A, Weier M, et al. High expression levels of macrophage migration inhibitory factor sustain the innate immune responses of neonates. Proc Natl Acad Sci U S A. 2016;113(8):E997-E1005. 262. Jain R, Weninger W. Shedding light on cutaneous innate immune responses: The intravital microscopy approach. Immunol Cell Biol. 2013;91(4):263-270. 263. Morales J, Russell JE, Liebhaber SA. Destabilization of human alpha-globin mRNA by translation anti-termination is controlled during erythroid differentiation and is paralleled by phased shortening of the poly(A) tail. J Biol Chem. 1997;272(10):6607-6613. 264. Kabanova S, Kleinbongard P, Volkmer J, Andree B, Kelm M, Jax TW. Gene expression analysis of human red blood cells. Int J Med Sci. 2009;6(4):156-159. 265. Popova EY, Krauss SW, Short SA, et al. Chromatin condensation in terminally differentiating mouse erythroblasts does not involve special architectural proteins but depends on histone deacetylation. Chromosome Res. 2009;17(1):47-64.     139  Appendices Appendix A  : Supplementary material for Chapter 2 A.1 Supplementary methods Cord blood collection and hematopoietic cell sorting Cord blood was collected from term neonates (38-40 weeks GA) delivered by elective caesarean section in absence of labor at the Children’s and Women’s Health Centre of BC (Vancouver, Canada). Cord blood was collected in sodium heparin anti-coagulated Vacutainers (Becton Dickinson, ON, Canada). CBMCs were extracted using a Lymphoprep (StemCell Technologies Inc., BC, Canada) density gradient centrifugation, washed and resuspended in phosphate-buffered saline. B cells, T cells, monocytes, NK cells and nRBCs were purified by FACS. The following conjugated antibodies were used to sort these cells: anti-CD3 FITC (clone HIT3a; BD Bioscience), anti-CD3 PE (clone UCHT1; BD Bioscience), anti-CD4 BV605 (clone OKT4; BioLegend, CA, USA), anti-CD8 PerCP-eFluor710 (clone SK1; eBioscience, CA, USA), anti-CD14 PE-Cy7 (clone 61D3; eBioscience), anti-CD19 Alexa Fluor 700 (clone HIB19; eBioscience), anti-CD19 PE (clone HIB19; eBioscience), antiCD56 APC (clone MEM188; eBioscience), anti-CD71 APC (clone OKT9; eBioscience) and anti-CD235 FITC (clone 10F7MN; eBioscience). The following additional antibodies were used for cell sorting of naïve CD4 T cells for gene expression experiments: anti-CCR7 Alexa Fluor 647 (clone 3D12; BD Bioscience), anti-CD25 PE-Cy7 (clone M-A251; BD Bioscience) and anti-CD45RO Alexa Fluor700 (clone UCHL1; BioLegend). For DNAm studies, cells were sorted from a total of 12 subjects: whole (CD3+) T cells, monocytes, and nRBCs were collected from 5 individuals (2 female, 3 male) by the standard 140  sorting method; B cells, CD4 T cells, CD8 T cells, granulocytes, monocytes, NK cells, and nRBCs were collected from 7 individuals (5 female, 2 male) by the stringent sorting method. Unstained, single-stain compensation and fluorescence-minus one (FMO) controls were prepared for each sample run. Compensation for spectral overlaps between fluorophores was done before cell acquisition. Cells were sorted on the FACSAria III flow cytometer using FACSDiva Software (both Becton Dickinson). A summary of the sorting parameters for each cell type are provided in Supplementary Table A.1. Data analysis was performed with Flowjo software (TreeStar, Inc., OR, USA).  Granulocytes were obtained from the bottom fraction of the Lymphoprep gradient during CBMC purification, mixed with 3% dextran/0.9% saline solution to allow separation of granulocytes from erythrocytes by sedimentation, and followed by three steps of hypotonic lysis. Hypotonic lysis was achieved by incubation of the granulocyte fraction with ice cold 0.2% sodium chloride (NaCl) for 30 seconds to lyse remaining red blood cells. Following lysis, isotonicity was restored by adding an equal volume of 1.6% NaCl solution at room temperature.  DNAm quality control and probe filtering Sample identity and quality were evaluated in three ways: 1) clustering with the 65 SNP probes provided on the array, with samples from the same individual grouping together; 2) clustering with probes on the X and Y chromosomes, with samples grouping by known sex; 3) clustering based on all probes, producing groups based on cell type. Based on these checks, one NK cell sample was removed as an outlier. Probes were removed from analysis if they fell into any of the following categories: 1) probes that target SNPs (n = 65); 2) probes that target or cross-hybridize with sites on the sex chromosomes (n = 11,648 and 11,359, respectively); and 3) 141  probes that target CpGs which may also contain SNPs (n = 19,271)203. Probes that had a detection p-value >0.01 or under 3 bead replicates in more than one sample were also removed (n = 2,919), for a final dataset of 440,315 CpG sites.  142  A.2 Supplementary tables Supplementary Table A.1 Summary of surface antigens targeted to sort each cord blood hematopoietic cell type using the standard and stringent FACS protocols. Standard-sorted cells CD3 CD4 CD8 CD14 CD19 CD56 CD71 CD235 T cells + • • • - • • • Monocytes • • • + - • • • nRBCs • • • - - • + + Stringent-sorted cells         B cells - • • - + • • - CD4 T cells + + - - - • • - CD8 T cells + - + - - • • - Monocytes - • • + - • • - NK cells - • • - - + • - nRBCs - • • - - • + + “+” and “-” symbols indicate positive and negative gating for a specific antigen, respectively. “•” indicates that the antigen was not used to sort a given cell type.  Supplementary Table A.2 Six nRBC-DM sites (FDR <5%, |Δβ| >0.20) located in erythroid-specific (hemoglobin) genes. 450K array CpG identifier CpG location: chromosome, closest gene Location in gene Mean nRBC β (min., max.) Mean non-erythroid cell β (min., max.) cg18764164 11, HBB TSS200 0.478 (0.355, 0.582) 0.904 (0.843, 0.935) cg14544583 11, HBB TSS1500; enhancer* 0.709 (0.647, 0.733) 0.940 (0.912, 0.962) cg18768582 11, HBG1 Intron 0.570 (0.466, 0.666) 0.842 (0.788, 0.884) cg20896063 11, HBG2 Intron 0.371 (0.237, 0.474) 0.714 (0.662, 0.761) cg12559170 11, HBG2 Intron 0.404 (0.273, 0.511) 0.658 (0.570, 0.734) cg27009246 11, HBG2 TSS200 0.460 (0.241, 0.608) 0.897 (0.768, 0.954) TSS200: within 200 bp of TSS; TSS1500: within 1500 bp of TSS *Based on UCSC Genome Browser: ENCODE Enhancer- and promoter- associated histone mark (H3K4Me1) in K562 cells   143  A.3 Supplementary figures  Supplementary Figure A.1 Publicly available gene expression data (GEO dataset GSE24759) of hemoglobin alpha, beta, and gamma genes in hematopoietic cells isolated from peripheral and cord blood by flow cytometry. Expression reported on a log2 scale. 144   Supplementary Figure A.2 Scatter plots of major principal components following principal component analysis of T cells, monocytes, and nRBCs sorted by the standard and stringent FACS strategies. 145   Supplementary Figure A.3 DNAm of B cells, CD4 T cells, CD8 T cells, granulocytes, monocytes, NK cells, and nRBCs from cord blood at the top eight CpG sites at which nRBCs are significantly DM from each of the other cell types (FDR <5%, Δβ >0.50). 146   Supplementary Figure A.4 DNAm in nRBCs is influenced by their proportion in whole cord blood, as measured by number of nRBCs/100 WBCs. (A) The 450K array-wide median methylation in nRBC samples shows decreasing DNAm with increasing nRBC count. (B) DNAm in nRBCs, CBMCs, whole cord blood (WB), and CD4 T cells at two of the CpG sites with the strongest association between nRBC DNAm and nRBC proportion (FDR <5%, magnitude of regression coefficient > 0.05). CBMCs and whole cord blood are included to display the potential impact these DNAm changes in nRBCs could have on cord blood cell mixtures; CD4 T cells are included as a reference for the other blood cell types, which do not show an association with nRBC count. (C) DNAm in nRBCs, CBMCs, whole cord blood, and CD4 T cells at the three of our eight identified erythroid DNAm marker CpGs that are significantly associated with nRBC proportion.   147  Appendix B  : Supplementary material for Chapter 3 B.1 Supplementary methods DNA methylation data quality control and probe filtering Once 450K array intensity data were converted into M values in R software, sample identity and quality were evaluated in three ways: 1) clustering with the 65 SNP probes provided on the array, with samples from the same individual grouping together; 2) clustering with probes on the X and Y chromosomes, with samples grouping by known sex; 3) clustering based on all probes, producing groups based on cell type. Based on these checks, one NK cell sample was removed as an outlier.  Probes were removed from analysis if they fell into any of the following categories: 1) probes that target SNPs (n = 65); 2) probes that target or cross-hybridize with sites on the sex chromosomes (n = 11,648 and 11,388, respectively); and 3) probes that target CpGs which may also contain SNPs (n = 19,474)203. Probes that had a detection p-value >0.01 or under 3 bead replicates in more than one sample were also removed (n = 1,400). Finally, additional filtering was performed to remove CpG sites that are suspected “epipolymorphisms”, in which changes in DNAm levels were associated with individuals rather than cell types. For each probe, the standard deviations in DNAm within each cell type and within each individual were calculated. If the mean cell type standard deviation was greater than the mean individual standard deviation, and the mean cell type standard deviation was greater than 0.05, that probe was removed from the dataset. 9,835 probes met these criteria and were filtered out, producing a final dataset of 431,767 probes. 148  B.2 Supplementary tables Supplementary Table B.1 Complete blood counts for each individual. All WBCs reported as x 109/L, %  Total WBC (x 109/L) Total RBC (x 1012/L) Lymphocytes (x 109/L, %) Neutrophils (x 109/L, %) Eosinophils (x 109/L, %) Basophils (x 109/L, %) Monocytes (x 109/L, %) nRBCs (x 109/L, /100 WBC) A 9.86 3.72 4.17, 42.3 4.59, 46.6 0.27, 2.7 0.03, 0.3 0.80, 8.1 0.09, 0.9 B 12.21 4.99 3.92, 32.1 6.81, 55.8 0.45, 3.7 0.16, 1.3 0.87, 7.1 0.14, 1.1 C 8.04 3.91 3.62, 45.0 3.70, 46.1 0.22, 2.7 0.01, 0.1 0.49, 6.1 0.13, 1.7 D 13.66 3.65 4.75, 34.8 6.69, 49.0 0.85, 6.2 0.08, 0.6 1.29, 9.4 0.54, 3.9 E 13.13 3.94 5.41, 41.2 6.27, 47.7 0.52, 4.0 0.12, 0.9 0.81, 6.2 0.76, 5.8 F 10.84 3.77 3.57, 32.9 6.16, 57.0 0.20, 1.8 0.07, 0.6 0.84, 7.7 0.80, 7.4 G 11.64 4.62 6.45, 55.4 4.27, 36.7 0.41, 3.5 0.10, 0.9 0.41, 3.5 1.22, 10.5    149  Supplementary Table B.2 Top 50 GO terms from ErmineJ analysis of cell-only nRBC DMRs, ordered by corrected p-value. GO term GO ID Number of genes Raw score Corrected p-value modulation of growth of symbiont involved in interaction with host GO:0044144 15 0.267 0.238 negative regulation of growth of symbiont involved in interaction with host GO:0044146 15 0.267 0.238 cardiac conduction GO:0061337 31 0.161 0.245 osteoblast development GO:0002076 18 0.222 0.256 regulation of growth of symbiont in host GO:0044126 14 0.286 0.266 negative regulation of growth of symbiont in host GO:0044130 14 0.286 0.266 defense response to Gram-positive bacterium GO:0050830 35 0.143 0.277 regulation of cardiac muscle contraction GO:0055117 53 0.113 0.295 defense response to fungus GO:0050832 17 0.235 0.303 calcium-mediated signaling GO:0019722 56 0.107 0.311 tumor necrosis factor receptor binding GO:0005164 22 0.182 0.348 regulation of cardiac muscle contraction by calcium ion signaling GO:0010882 13 0.308 0.385 regulation of defense response to virus by virus GO:0050690 27 0.148 0.410 Rac protein signal transduction GO:0016601 13 0.231 0.423 multicellular organismal signaling GO:0035637 67 0.090 0.431 response to fungus GO:0009620 27 0.148 0.433 regulation of striated muscle contraction GO:0006942 62 0.097 0.441 neutrophil mediated immunity GO:0002446 13 0.231 0.444 myeloid leukocyte activation GO:0002274 67 0.090 0.458 regulation of chemokine biosynthetic process GO:0045073 12 0.250 0.461 regulation of leukocyte migration GO:0002685 88 0.080 0.485 regulation of cardiac muscle contraction by regulation of the release of sequestered calcium ion GO:0010881 12 0.250 0.494 myeloid leukocyte mediated immunity GO:0002444 26 0.154 0.518 tumor necrosis factor receptor superfamily binding GO:0032813 30 0.133 0.531 hydrolase activity, acting on glycosyl bonds GO:0016798 100 0.070 0.535 150  GO term GO ID Number of genes Raw score Corrected p-value hydrolase activity, hydrolyzing N-glycosyl compounds GO:0016799 17 0.176 0.589 regulation of relaxation of cardiac muscle GO:1901897 5 0.400 0.597 regulation of relaxation of muscle GO:1901077 6 0.333 0.609 disruption by host of symbiont cells GO:0051852 5 0.400 0.616 killing by host of symbiont cells GO:0051873 5 0.400 0.616 sperm midpiece GO:0097225 6 0.333 0.622 cell killing GO:0001906 39 0.103 0.633 regulation of lymphocyte migration GO:2000401 20 0.150 0.634 regulation of mucus secretion GO:0070255 6 0.333 0.635 positive regulation of interleukin-8 biosynthetic process GO:0045416 5 0.400 0.636 macrophage activation GO:0042116 20 0.150 0.647 micropinocytosis GO:0044351 6 0.333 0.649 positive regulation of interferon-gamma production GO:0032729 40 0.100 0.654 cytokine production GO:0001816 91 0.066 0.654 microvillus assembly GO:0030033 5 0.400 0.657 purinergic receptor activity GO:0035586 18 0.167 0.659 regulation of bicellular tight junction assembly GO:2000810 7 0.286 0.663 regulation of glial cell apoptotic process GO:0034350 6 0.333 0.664 negative regulation of glial cell apoptotic process GO:0034351 6 0.333 0.664 killing of cells of other organism GO:0031640 16 0.188 0.669 disruption of cells of other organism GO:0044364 16 0.188 0.669 cell communication by electrical coupling involved in cardiac conduction GO:0086064 7 0.286 0.674 regulation of defense response to virus GO:0050688 67 0.075 0.676 positive regulation of cysteine-type endopeptidase activity GO:2001056 92 0.065 0.677 positive regulation of endopeptidase activity GO:0010950 95 0.063 0.677  151  Supplementary Table B.3 Top 50 GO terms from ErmineJ analysis of count-associated nRBC DMRs, ordered by corrected p-value. Significantly enriched terms (corrected p-value <0.10) are shaded. GO term GO ID Number of genes Raw score Corrected p-value Rho guanyl-nucleotide exchange factor activity GO:0005089 63 0.317 0.003 Ras guanyl-nucleotide exchange factor activity GO:0005088 100 0.250 0.010 RISC complex GO:0016442 6 0.833 0.062 RNAi effector complex GO:0031332 6 0.833 0.062 cytoskeletal adaptor activity GO:0008093 16 0.500 0.076 phosphatidylinositol 3-kinase binding GO:0043548 17 0.471 0.079 regulation of cell shape GO:0008360 95 0.221 0.138 peptidyl-tyrosine phosphorylation GO:0018108 88 0.216 0.197 peptidyl-tyrosine modification GO:0018212 89 0.213 0.211 amide transmembrane transporter activity GO:0042887 12 0.500 0.211 TAP1 binding GO:0046978 5 0.800 0.225 TAP2 binding GO:0046979 5 0.800 0.225 Rho GTPase binding GO:0017048 47 0.277 0.226 positive T cell selection GO:0043368 17 0.412 0.228 MHC class I peptide loading complex GO:0042824 5 0.800 0.250 response to hydrogen peroxide GO:0042542 64 0.234 0.258 actin filament bundle assembly GO:0051017 31 0.323 0.264 actin filament bundle organization GO:0061572 31 0.323 0.264 anchored component of external side of plasma membrane GO:0031362 18 0.389 0.267 ferric iron transport GO:0015682 25 0.320 0.280 transferrin transport GO:0033572 25 0.320 0.280 trivalent inorganic cation transport GO:0072512 25 0.320 0.280 iron ion homeostasis GO:0055072 67 0.224 0.281 cellular transition metal ion homeostasis GO:0046916 68 0.221 0.281 non-membrane spanning protein tyrosine kinase activity GO:0004715 36 0.278 0.282 myelination in peripheral nervous system GO:0022011 18 0.389 0.283 152  GO term GO ID Number of genes Raw score Corrected p-value peripheral nervous system axon ensheathment GO:0032292 18 0.389 0.283 transition metal ion homeostasis GO:0055076 90 0.200 0.288 cellular iron ion homeostasis GO:0006879 49 0.245 0.288 alpha-beta T cell differentiation GO:0046632 30 0.300 0.292 alpha-beta T cell activation involved in immune response GO:0002287 14 0.429 0.292 T cell differentiation involved in immune response GO:0002292 14 0.429 0.292 alpha-beta T cell differentiation involved in immune response GO:0002293 14 0.429 0.292 regulation of interleukin-4 production GO:0032673 19 0.368 0.293 establishment or maintenance of cell polarity GO:0007163 90 0.200 0.298 alpha-beta T cell lineage commitment GO:0002363 6 0.667 0.298 osteoblast development GO:0002076 18 0.389 0.301 Schwann cell development GO:0014044 20 0.350 0.301 negative regulation of myeloid leukocyte differentiation GO:0002762 30 0.300 0.304 iron ion transport GO:0006826 41 0.268 0.312 alpha-beta T cell activation GO:0046631 35 0.286 0.327 3',5'-cyclic-GMP phosphodiesterase activity GO:0047555 11 0.455 0.329 CD4-positive or CD8-positive, alpha-beta T cell lineage commitment GO:0043369 7 0.571 0.330 SH3 domain binding GO:0017124 100 0.190 0.336 MHC class II protein complex GO:0042613 11 0.455 0.338 peptide transporter activity GO:0015197 7 0.571 0.338 excretion GO:0007588 45 0.244 0.339 inorganic anion transmembrane transport GO:0098661 64 0.219 0.341 Rac guanyl-nucleotide exchange factor activity GO:0030676 11 0.455 0.347 response to transition metal nanoparticle GO:1990267 78 0.205 0.351   153  B.3 Supplementary figures  Supplementary Figure B.1 (A) Median absolute deviation distribution and (B) principal components analysis show that nRBCs are a distinct and highly variable blood cell population. 154   Supplementary Figure B.2 nRBC DNAm is similar to other blood cells at imprinted and placenta-specific DMRs, and does not change with nRBC proportion in cord blood. (A) Scatterplots of DNAm for exemplar known somatic imprinted DMRs. (B) Scatterplots of DNAm for exemplar placenta-specific imprinted DMRs. (C) DNAm by cell/tissue type at all 450K array probes located in known somatic imprinted DMRs240. (D) DNAm by cell/tissue type at all 450K array probes located in placenta-specific imprinted DMRs240. (E) Median DNAm across somatic and placenta-specific imprinted DMRs by nRBC proportion in cord blood.   155  Appendix C  : Supplementary material for Chapter 4 C.1 Supplementary methods Cord blood collection and hematopoietic cell sorting Cord blood was collected from neonates delivered by elective caesarean section in absence of labor at the Children’s and Women’s Health Centre of BC (Vancouver, Canada). Cord blood was collected in sodium heparin anti-coagulated Vacutainers (Becton Dickinson). CBMCs were extracted using a Lymphoprep (StemCell Technologies Inc.) density gradient centrifugation, washed and resuspended in phosphate-buffered saline. T cells, monocytes, and nRBCs were purified by FACS. The following conjugated antibodies were used to sort these cells: anti-CD3 PE (clone UCHT1; BD Bioscience), anti-CD14 PE-Cy7 (clone 61D3; eBioscience), anti-CD19 Alexa Fluor 700 (clone HIB19; eBioscience), anti-CD71 APC (clone OKT9; eBioscience) and anti-CD235 FITC (clone 10F7MN; eBioscience).  Cells were sorted from a total of 10 subjects: 5 preterm (GA range 26-30 weeks) and 5 term (GA range 38-40 weeks). Unstained, single-stain compensation and FMO controls were prepared for each sample run. Compensation for spectral overlaps between fluorophores was done before cell acquisition. Cells were sorted on the FACSAria III flow cytometer using FACSDiva Software (both Becton Dickinson). Data analysis was performed with Flowjo software (TreeStar, Inc.). T cells were sorted according to the following parameters: CD3+/CD19-/CD235-/CD14-. Monocytes were sorted according to the following parameters: CD3-/CD19-/CD14+/CD71-. nRBCs were sorted according to the following parameters: CD3-/CD19-/CD235+/CD71+. 156  Granulocytes were obtained from the bottom fraction of the Lymphoprep gradient during CBMC purification, mixed with 3% dextran/0.9% saline solution to allow separation of granulocytes from erythrocytes by sedimentation, and followed by three steps of hypotonic lysis. Hypotonic lysis was achieved by incubation of the granulocyte fraction with ice cold 0.2% sodium chloride (NaCl) for 30 seconds to lyse remaining red blood cells. Following lysis, isotonicity was restored by adding an equal volume of 1.6% NaCl solution at room temperature.  DNA methylation data quality control and probe filtering Once 450K array intensity data were converted into M values in R software, sample identity and quality were evaluated in three ways: 1) clustering with the 65 SNP probes provided on the array, with samples from the same individual grouping together as expected; 2) clustering with probes on the X and Y chromosomes, with samples grouping by known sex as expected; 3) clustering based on all probes, producing groups based on cell type. Probes were removed from analysis if they fell into any of the following categories: 1) probes that target SNPs (n = 65); 2) probes that target or cross-hybridize with sites on the sex chromosomes (n = 11,648 and 11,366, respectively); and 3) probes that target CpGs which may also contain SNPs (n = 19,283)203. Probes that had a detection p-value >0.01 or under 3 bead replicates in more than one sample were also removed (n = 2,943). Additional filtering was performed to remove CpG sites that are suspected epipolymorphisms: for each probe, the DNAm standard deviations within each cell type and within each individual were calculated. If the mean cell type standard deviation was both greater than the mean individual standard deviation and greater than 0.05, that probe was removed from the dataset. 10,507 probes met these criteria and were filtered out, producing a final dataset of 429,765 probes.157  C.2 Supplementary tables Supplementary Table C.1 Significantly enriched GO terms (corrected p-value <0.10) from ErmineJ analysis of prematurity-DM sites for each cell type, ordered by corrected p-value. GO term GO ID Number of genes Raw score Corrected p-value T cells embryonic placenta development GO:0001892 71 0.099 2.90E-02 Granulocytes negative regulation of ERK1 and ERK2 cascade GO:0070373 25 0.320 1.66E-02 Ras guanyl-nucleotide exchange factor activity GO:0005088 100 0.140 7.85E-02 Monocytes molting cycle GO:0042303 69 0.159 2.75E-03 hair cycle GO:0042633 69 0.159 2.75E-03 hair follicle development GO:0001942 65 0.169 2.94E-03 molting cycle process GO:0022404 65 0.169 2.94E-03 hair cycle process GO:0022405 65 0.169 2.94E-03 skin epidermis development GO:0098773 65 0.169 2.94E-03 hair follicle morphogenesis GO:0031069 24 0.250 4.25E-02 epidermis morphogenesis GO:0048730 28 0.214 8.21E-02 nRBCs forebrain generation of neurons GO:0021872 46 0.609 3.68E-03 regulation of fat cell differentiation GO:0045598 64 0.531 9.61E-03 Ras guanyl-nucleotide exchange factor activity GO:0005088 100 0.460 1.12E-02 actin filament binding GO:0051015 72 0.500 1.23E-02 negative regulation of fat cell differentiation GO:0045599 32 0.625 1.28E-02 Rho guanyl-nucleotide exchange factor activity GO:0005089 63 0.524 1.34E-02 forebrain neuron differentiation GO:0021879 36 0.611 1.39E-02 axon part GO:0033267 99 0.455 1.53E-02 renal system process GO:0003014 66 0.500 1.60E-02 sarcolemma GO:0042383 86 0.465 1.82E-02 158  GO term GO ID Number of genes Raw score Corrected p-value positive regulation of JNK cascade GO:0046330 84 0.464 2.18E-02 site of polarized growth GO:0030427 94 0.447 2.22E-02 neurotransmitter secretion GO:0007269 66 0.485 2.31E-02 regulation of dendritic spine morphogenesis GO:0061001 14 0.786 2.31E-02 positive regulation of synaptic transmission GO:0050806 60 0.500 2.41E-02 regulation of interleukin-2 production GO:0032663 34 0.588 2.43E-02 growth cone GO:0030426 91 0.451 2.43E-02 skeletal muscle tissue development GO:0007519 99 0.444 2.47E-02 negative regulation of fibroblast growth factor receptor signaling pathway GO:0040037 9 0.889 4.00E-02 digestive tract development GO:0048565 97 0.433 4.11E-02 T cell differentiation GO:0030217 97 0.433 4.31E-02 positive regulation of stress-activated MAPK cascade GO:0032874 91 0.440 4.36E-02 positive regulation of leukocyte differentiation GO:1902107 88 0.443 4.47E-02 canonical Wnt signaling pathway GO:0060070 69 0.464 4.63E-02 positive regulation of stress-activated protein kinase signaling cascade GO:0070304 92 0.435 4.64E-02 positive regulation of neuroblast proliferation GO:0002052 13 0.769 4.73E-02 regulation of cardiac muscle contraction GO:0055117 53 0.491 4.75E-02 antigen receptor-mediated signaling pathway GO:0050851 93 0.430 4.79E-02 regulation of protein kinase B signaling GO:0051896 87 0.437 4.80E-02 cell cortex part GO:0044448 75 0.453 4.91E-02 regulation of axonogenesis GO:0050770 87 0.437 4.96E-02 inositol phosphate metabolic process GO:0043647 42 0.524 5.06E-02 odontogenesis GO:0042476 87 0.437 5.13E-02 G-protein coupled chemoattractant receptor activity GO:0001637 20 0.650 5.14E-02 chemokine receptor activity GO:0004950 20 0.650 5.14E-02 positive regulation of alpha-beta T cell activation GO:0046635 40 0.525 5.39E-02 regulation of interferon-gamma production GO:0032649 65 0.462 5.40E-02 pallium development GO:0021543 91 0.429 5.40E-02 159  GO term GO ID Number of genes Raw score Corrected p-value embryonic digestive tract development GO:0048566 30 0.567 5.47E-02 regulation of alpha-beta T cell activation GO:0046634 51 0.490 5.49E-02 cardiac muscle cell differentiation GO:0055007 62 0.468 5.52E-02 regulation of reactive oxygen species biosynthetic process GO:1903426 38 0.526 5.72E-02 postsynaptic density GO:0014069 98 0.418 5.81E-02 regulation of glucose import GO:0046324 38 0.526 5.85E-02 adherens junction organization GO:0034332 57 0.474 5.88E-02 filopodium GO:0030175 57 0.474 6.02E-02 negative regulation of blood pressure GO:0045776 33 0.545 6.03E-02 regulation of alpha-beta T cell differentiation GO:0046637 34 0.529 6.22E-02 regulation of fibroblast migration GO:0010762 14 0.714 6.23E-02 positive regulation of blood circulation GO:1903524 56 0.464 6.26E-02 cardiocyte differentiation GO:0035051 83 0.422 6.26E-02 positive regulation of protein kinase B signaling GO:0051897 62 0.452 6.27E-02 ventricular cardiac muscle cell differentiation GO:0055012 19 0.632 6.30E-02 negative regulation of interferon-gamma production GO:0032689 24 0.583 6.31E-02 regulation of JUN kinase activity GO:0043506 65 0.446 6.32E-02 regulation of osteoblast differentiation GO:0045667 86 0.419 6.33E-02 regulation of synapse structure or activity GO:0050803 52 0.481 6.33E-02 negative regulation of vasculature development GO:1901343 56 0.464 6.34E-02 regulation of striated muscle contraction GO:0006942 62 0.452 6.35E-02 hindgut morphogenesis GO:0007442 8 0.875 6.37E-02 hindgut development GO:0061525 8 0.875 6.37E-02 regulation of catenin import into nucleus GO:0035412 19 0.632 6.39E-02 hair follicle morphogenesis GO:0031069 24 0.583 6.40E-02 Z disc GO:0030018 86 0.419 6.40E-02 hair follicle development GO:0001942 65 0.446 6.40E-02 molting cycle process GO:0022404 65 0.446 6.40E-02 hair cycle process GO:0022405 65 0.446 6.40E-02 160  GO term GO ID Number of genes Raw score Corrected p-value skin epidermis development GO:0098773 65 0.446 6.40E-02 regulation of dendrite development GO:0050773 64 0.453 6.41E-02 regulation of synapse organization GO:0050807 48 0.479 6.42E-02 digestive tract morphogenesis GO:0048546 45 0.489 6.42E-02 ventricular septum development GO:0003281 39 0.513 6.42E-02 positive regulation of stem cell proliferation GO:2000648 55 0.473 6.43E-02 positive regulation of alpha-beta T cell differentiation GO:0046638 29 0.552 6.44E-02 regulation of nitric oxide biosynthetic process GO:0045428 36 0.528 6.45E-02 regulation of heterotypic cell-cell adhesion GO:0034114 12 0.750 6.45E-02 negative regulation of glucose transport GO:0010829 10 0.800 6.46E-02 tissue remodeling GO:0048771 80 0.425 6.46E-02 phospholipid transporter activity GO:0005548 31 0.548 6.46E-02 regulation of dendrite morphogenesis GO:0048814 40 0.500 6.47E-02 negative regulation of angiogenesis GO:0016525 52 0.481 6.47E-02 regulation of interleukin-4 production GO:0032673 19 0.632 6.48E-02 M band GO:0031430 17 0.647 6.49E-02 stress-activated protein kinase signaling cascade GO:0031098 98 0.408 6.50E-02 odontogenesis of dentin-containing tooth GO:0042475 64 0.453 6.52E-02 myoblast differentiation GO:0045445 29 0.552 6.53E-02 stem cell division GO:0017145 26 0.577 6.54E-02 transmembrane receptor protein tyrosine phosphatase activity GO:0005001 17 0.647 6.57E-02 transmembrane receptor protein phosphatase activity GO:0019198 17 0.647 6.57E-02 T cell mediated immunity GO:0002456 19 0.632 6.57E-02 diencephalon development GO:0021536 61 0.459 6.57E-02 positive regulation of humoral immune response GO:0002922 10 0.800 6.58E-02 T cell receptor signaling pathway GO:0050852 70 0.443 6.62E-02 smooth muscle contraction GO:0006939 42 0.500 6.64E-02 muscle cell development GO:0055001 91 0.418 6.65E-02 161  GO term GO ID Number of genes Raw score Corrected p-value negative regulation of blood vessel morphogenesis GO:2000181 53 0.472 6.70E-02 stress-activated MAPK cascade GO:0051403 97 0.412 6.78E-02 photoreceptor outer segment GO:0001750 43 0.488 6.88E-02 regulation of neurotransmitter levels GO:0001505 97 0.412 6.89E-02 regulation of dendritic spine development GO:0060998 22 0.591 6.89E-02 cell proliferation in forebrain GO:0021846 22 0.591 6.96E-02 A band GO:0031672 27 0.556 7.00E-02 cerebral cortex development GO:0021987 63 0.444 7.12E-02 response to activity GO:0014823 35 0.514 7.15E-02 positive regulation of endothelial cell migration GO:0010595 46 0.478 7.26E-02 positive regulation of dendrite development GO:1900006 30 0.533 7.47E-02 regulation of nodal signaling pathway GO:1900107 5 1.000 7.50E-02 tumor necrosis factor receptor superfamily binding GO:0032813 30 0.533 7.54E-02 receptor localization to synapse GO:0097120 5 1.000 7.58E-02 cardiac muscle cell development GO:0055013 38 0.500 7.58E-02 neurotransmitter-gated ion channel clustering GO:0072578 5 1.000 7.65E-02 negative regulation of axonogenesis GO:0050771 38 0.500 7.65E-02 lens development in camera-type eye GO:0002088 49 0.469 7.70E-02 regulation of urine volume GO:0035809 13 0.692 7.71E-02 response to electrical stimulus GO:0051602 25 0.560 7.73E-02 hyperosmotic response GO:0006972 20 0.600 7.78E-02 oligodendrocyte differentiation GO:0048709 41 0.488 7.81E-02 signaling adaptor activity GO:0035591 55 0.455 7.82E-02 central nervous system neuron development GO:0021954 55 0.455 7.90E-02 positive regulation of behavior GO:0048520 91 0.407 8.11E-02 regulation of endothelial cell migration GO:0010594 73 0.425 8.13E-02 humoral immune response GO:0006959 85 0.412 8.16E-02 T cell differentiation in thymus GO:0033077 44 0.477 8.19E-02 thymocyte aggregation GO:0071594 44 0.477 8.19E-02 Golgi lumen GO:0005796 67 0.433 8.21E-02 positive regulation of excitatory postsynaptic GO:2000463 11 0.727 8.25E-02 162  GO term GO ID Number of genes Raw score Corrected p-value membrane potential axonal growth cone GO:0044295 11 0.727 8.32E-02 regulation of development, heterochronic GO:0040034 11 0.727 8.38E-02 regulation of timing of cell differentiation GO:0048505 11 0.727 8.38E-02 negative regulation of interleukin-2 production GO:0032703 11 0.727 8.45E-02 epidermis morphogenesis GO:0048730 28 0.536 8.62E-02 labyrinthine layer development GO:0060711 36 0.500 8.67E-02 regulation of N-methyl-D-aspartate selective glutamate receptor activity GO:2000310 9 0.778 8.69E-02 gas homeostasis GO:0033483 7 0.857 8.70E-02 negative regulation of developmental growth GO:0048640 53 0.453 8.75E-02 regulation of cardioblast differentiation GO:0051890 9 0.778 8.76E-02 ephrin receptor signaling pathway GO:0048013 23 0.565 8.78E-02 C-C chemokine receptor activity GO:0016493 9 0.778 8.83E-02 positive regulation of lymphocyte differentiation GO:0045621 56 0.446 8.84E-02 regulation of fibroblast growth factor receptor signaling pathway GO:0040036 23 0.565 8.85E-02 myotube differentiation GO:0014902 39 0.487 8.87E-02 regulation of gliogenesis GO:0014013 59 0.441 8.93E-02 regulation of synapse assembly GO:0051963 31 0.516 9.12E-02 regulation of wound healing GO:0061041 80 0.413 9.13E-02 main axon GO:0044304 42 0.476 9.14E-02 embryonic placenta development GO:0001892 71 0.423 9.18E-02 actin filament bundle assembly GO:0051017 31 0.516 9.19E-02 actin filament bundle organization GO:0061572 31 0.516 9.19E-02 glomerulus development GO:0032835 42 0.476 9.20E-02 negative regulation of axon extension GO:0030517 16 0.625 9.87E-02 second-messenger-mediated signaling GO:0019932 99 0.394 9.90E-02 hypothalamus development GO:0021854 16 0.625 9.94E-02 163  C.3 Supplementary figures  Supplementary Figure C.1 Overlap of prematurity-DM sites between each hematopoietic cell type. The 25 common prematurity-DM sites are highlighted with bold text. 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0228788/manifest

Comment

Related Items