Transcriptome-wide profiling of the neonatal monocyte response to E. coli and S. epidermidis

Emma de Jong, B.Sc. (Hons)

This thesis is presented for the degree of Doctor of Philosophy

School of Veterinary & Life Sciences

2015

Declaration

I declare that this thesis is my own account of my research, and contains as its main content work that has not previously been submitted for a degree at any other tertiary education institution.

...... (Emma de Jong)

i

ii Abstract Preterm infants are extremely vulnerable to life-threatening invasive infections (particularly with Staphylococcus epidermidis and Escherichia coli) however our understanding of their innate immune defences is limited. Furthermore, prenatal exposure to histologic chorioamnionitis (HCA) complicates 40–70% of preterm births and is known to modulate the risk for sepsis, yet the impact of HCA on the development of innate immunity is largely unknown. We hypothesised that inadequate monocyte activation by neonatal pathogens results in impaired innate immune responses thereby increasing preterm infants’ susceptibility to invasive infection, and that prenatal exposure to HCA overrides these impairments.

Cell sorting and bacterial stimulation methodologies were developed and optimised specifically for working with human infant cord blood samples. RNA-sequencing was performed on purified cord blood monocytes from very preterm (≤31 weeks gestational age (GA)) and term infants (37–40 weeks GA) following challenge with live S. epidermidis or E. coli to identify gene/pathway differences specific to the preterm infant. Protein levels of inflammatory cytokines and chemokines were measured in paired monocyte culture supernatants.

Preterm infants displayed a quantitative monocyte deficiency compared to term infants, manifesting as reduced frequencies of classical monocytes with significantly reduced CD14 expression. However monocytes from preterm infants did not exhibit an intrinsically deficient transcriptional or protein response to stimulation with either pathogen. Prenatal exposure to HCA resulted in the transcriptional reprograming of a subset of genes towards a hyporesponsive phenotype in response to S. epidermidis, but not E. coli. The major transcriptional changes induced by either pathogen were highly conserved across infant groups and between stimuli, highlighting a conserved neonatal monocyte response to infection that was largely mediated by pattern recognition receptor/NF-B signalling. In addition, we observed an interferon/anti-viral immune signature that was specific to monocyte stimulation with E. coli.

This is the first transcriptome-wide analysis of the neonatal monocyte response to E. coli and S. epidermidis. This data provides novel insights into the functionality of preterm and term infant monocytes and confirms that exposure to HCA may impact on the development on neonatal immunity.

iii

iv Table of contents

Declaration………………………..……………………………………………………...i Abstract…………………………………………………………………………………iii Acknowledgements……………………………………………………………………..ix Abbreviations…………………………………………………………………………...xi List of Figures……………………………………………………………………….…xiv List of Tables……………………...……………………………………………….….xvii

CHAPTER 1 LITERATURE REVIEW ...... 1

1.1 AN INTRODUCTION TO PRETERM BIRTH ...... 1 1.1.1 Epidemiology ...... 1 1.1.2 The significance of preterm birth ...... 2 1.1.3 Risk factors for preterm birth...... 3 1.2 INFECTIONS AND PRETERM BIRTH ...... 3 1.2.1 Intrauterine infection and inflammation ...... 4 1.2.2 Early-onset neonatal sepsis ...... 6 1.2.3 Late-onset neonatal sepsis ...... 6 1.3 PATHOGENESIS OF NEONATAL SEPSIS ...... 8 1.3.1 Regulation of inflammation in neonatal sepsis ...... 9 1.3.2 Pathogen-specific responses to sepsis pathogens ...... 11 1.4 DEVELOPMENT OF NEONATAL INNATE IMMUNE DEFENCES ARE CRITICAL FOR PROTECTION AGAINST BACTERIAL PATHOGENS ...... 11 1.4.1 Maturation of innate immunity in neonates ...... 13 1.4.2 Monocytes and immune-regulation ...... 17 1.5 HOW DOES EXPOSURE TO HISTOLOGIC CHORIOAMNIONITIS IMPACT NEONATAL MONOCYTE FUNCTION? ...... 27 1.6 TRANSCRIPTIONAL PROFILING OF IMMUNE RESPONSES IN SEPSIS ...... 28 1.6.1 DNA microarrays and RNA sequencing ...... 28 1.6.2 Relevant transcriptomics studies in the field of sepsis and innate immunity ...... 30 1.7 SUMMARY ...... 32 1.8 AIMS AND HYPOTHESES OF THIS THESIS...... 33

CHAPTER 2 MATERIALS AND METHODS ...... 35

2.1 STUDY PARTICIPANTS ...... 35 2.2 BLOOD SAMPLE COLLECTION AND PROCESSING ...... 35 2.3 MICROBIOLOGY ...... 36 2.3.1 Strain selection ...... 36 2.3.2 Preparation of bacterial broths ...... 36 2.3.3 Growth and storage of live mid-log bacterial stocks ...... 36 v 2.3.4 Determining bacterial viability of frozen stocks ...... 37 2.4 CELL CULTURE METHODOLOGY ...... 37 2.4.1 Thawing of cryopreserved mononuclear cells ...... 37 2.4.2 CBMC surface receptor staining ...... 38 2.4.3 Purification of monocytes by cell sorting ...... 39 2.4.4 Monocyte stimulation with live E. coli or S. epidermidis ...... 41 2.4.5 Quantitative detection of cytokines/chemokines in culture supernatants ...... 41 2.5 CELL DEATH ASSAYS ...... 43 2.5.1 LDH assay ...... 43 2.5.2 Apoptosis assessment by flow cytometry ...... 43 2.6 MOLECULAR STUDIES ...... 44 2.6.1 Purification of total RNA from cultured monocytes ...... 44 2.6.2 cDNA synthesis ...... 44 2.6.3 Primer design ...... 44 2.6.4 Conventional PCR for primer optimisation ...... 45 2.6.5 Sequencing of PCR products ...... 45 2.6.6 Real-time polymerase chain reaction ...... 47 2.7 NEXT-GENERATION SEQUENCING OF MRNA ...... 48 2.7.1 Experimental design ...... 48 2.7.2 RNA sequencing by AGRF ...... 48 2.7.3 Quality assessment of raw sequencing data ...... 49 2.7.4 Read alignment to the human genome and count summarisation ...... 49 2.7.5 Differential expression analysis ...... 49 2.8 BIOINFORMATICS AND STATISTICS ...... 50 2.8.1 Ingenuity® Pathway Analysis...... 50 2.8.2 Network analysis ...... 51 2.8.3 Statistical analyses ...... 52

CHAPTER 3 DEVELOPMENT OF METHODOLOGY ...... 53

3.1 INTRODUCTION ...... 53 3.2 OPTIMISATION OF MONOCYTE PURIFICATION ...... 53 3.2.1 Preliminary experiments using negative selection techniques ...... 54 3.2.2 Comparison between positive and negative cell sorting methods ...... 55 3.2.3 The optimal method for monocyte purification from cord blood: cell sorting using a combination of positive and negative markers ...... 59 3.2.4 Summary ...... 60 3.3 DETERMINING RNA INPUT REQUIREMENTS FOR RNA-SEQ ...... 61 3.4 CULTURE OF E. COLI AND S. EPIDERMIDIS ...... 63 3.5 OPTIMISATION OF MONOCYTE CULTURE CONDITIONS ...... 64 3.5.1 Resting cells prior to bacterial stimulation ...... 64 3.5.2 Optimisation of monocyte stimulation with live E. coli and S. epidermidis ...... 68 3.6 RNA-SEQ EXPERIMENTAL DESIGN ...... 76 vi 3.7 DISCUSSION ...... 79

CHAPTER 4 MONOCYTES FROM PRETERM INFANTS ARE NOT INTRINSICALLY DEFICIENT, BUT ARE TRANSCRIPTIONALLY DISTINCT FOLLOWING PRENATAL EXPOSURE TO HISTOLOGIC CHORIOAMNIONITIS ...... 83

4.1 INTRODUCTION ...... 83 4.2 CHARACTERISATION OF THE INFANTS AND SAMPLES USED FOR RNA-SEQ ...... 85 4.2.1 Demographic information ...... 85 4.2.2 Phenotyping lymphocyte and monocyte populations ...... 86 4.2.3 Phenotyping of monocyte subsets ...... 88 4.3 BACTERIAL STIMULATION INDUCES A DISTINCT MONOCYTE TRANSCRIPTIONAL PROFILE ...... 91 4.4 HYPOTHESIS-DRIVEN INTERROGATION OF RNA-SEQ DATA ...... 95 4.5 DIRECT STATISTICAL COMPARISONS OF UNSTIMULATED, E. COLI- OR S. EPIDERMIDIS-STIMULATED MONOCYTES BETWEEN INFANT GROUPS ...... 100 4.5.1 Comparisons across culture conditions ...... 109 4.6 COMPARISONS OF THE E. COLI OR S. EPIDERMIDIS-INDUCED MONOCYTE TRANSCRIPTIONAL RESPONSE BETWEEN INFANT GROUPS ...... 111 4.6.1 Differential expression analysis ...... 112 4.6.2 Bioinformatic analysis of differentially expressed genes ...... 118 4.6.3 Comparisons of differentially expressed genes using a direct statistical approach ...... 133 4.7 DO PRETERM MONOCYTES EXHIBIT A DISTINCT CD16+ MONOCYTE TRANSCRIPTOME SIGNATURE?...... 135 4.8 ANALYSIS OF INFLAMMATORY PROTEINS IN MONOCYTE AND CBMC CULTURE SUPERNATANTS 139 4.8.1 Correlations with RNA-seq data ...... 143 4.9 DISCUSSION ...... 146

CHAPTER 5 CHARACTERISATION OF THE CONSERVED AND PATHOGEN-SPECIFIC NEONATAL MONOCYTE TRANSCRIPTIONAL RESPONSES TO E. COLI AND S. EPIDERMIDIS… ...... 153

5.1 INTRODUCTION ...... 153 5.2 IDENTIFYING THE GLOBAL NEONATAL MONOCYTE TRANSCRIPTIONAL RESPONSES TO E. COLI AND S. EPIDERMIDIS ...... 154 5.3 CHARACTERISING THE NEONATAL MONOCYTE TRANSCRIPTIONAL RESPONSE SPECIFIC TO E. COLI… ...... 157 5.3.1 Over-represented canonical pathways, diseases or bio-functions ...... 158 5.3.2 Upstream regulator analysis ...... 160 5.3.3 Network analysis ...... 165 5.4 CHARACTERISING THE NEONATAL MONOCYTE TRANSCRIPTIONAL RESPONSE SPECIFIC TO S. EPIDERMIDIS ...... 167 5.4.1 Upstream regulator analysis ...... 168 5.4.2 Network analysis ...... 169 5.5 CHARACTERISING THE CONSERVED NEONATAL MONOCYTE TRANSCRIPTIONAL RESPONSE TO E. COLI AND S. EPIDERMIDIS ...... 171 vii 5.5.1 Over-represented canonical pathways, diseases or bio-functions...... 171 5.5.2 Upstream regulator analysis ...... 176 5.5.3 Network analysis ...... 178 5.6 DISCUSSION ...... 184

CHAPTER 6 GENERAL DISCUSSION ...... 190

6.1 SUMMARY AND DISCUSSION OF KEY FINDINGS ...... 191 6.2 LIMITATIONS OF THIS STUDY ...... 197 6.3 DIRECTIONS FOR FUTURE RESEARCH ...... 198 6.4 CONCLUDING REMARKS ...... 200

REFERENCES………………………………………………………………………………………… 200

APPENDIX…………………………………………………………………………………………….. 220

viii Acknowledgements

To my principle supervisor Dr Andrew Currie, thank you so much for your continued support and mentorship, and for encouraging me to achieve above and beyond what I thought was possible.

To my co-supervisor Dr Tobias Strunk, thank you for always taking the time to meet and provide feedback and valuable clinical expertise, and for all your encouragement mentorship.

To my co-supervisor Associate Professor Peter Richmond, thank you for your support and for offering insightful perspectives.

To Dr David Hancock, I am eternally thankful for your generosity and patience in guiding me through the world of transcriptomics and R programming, thank you for all of your help with the data analysis.

I would also like to sincerely thank Associate Professor Christine Wells and Othmar Korn from The University of Queensland for their generous advice and assistance with the RNA-seq data processing.

To Dr Lea-Ann Kirkham thank you for your unofficial mentorship and support.

Thank you to Dr Janessa Pickering and Dr Stephanie Trend for making me feel welcome when I first started, for training me in microbiological and PCR techniques, and for your ongoing support and friendship.

To Caitlyn Granland, thank you so much for all of your help in the lab, especially with bioplex and cell culture experiments. To Dr Bree Foley thank you for performing the cell sorting and being so accommodating, especially during the optimisation. A special thank you to Lisa Montgomery and Jan Jones for providing laboratory support wherever possible.

To my wonderful friends Lilian, Sara, Sonia and Dino who have supported me from the beginning, thank you.

ix

To my family Willemina, Ron, Penny and Julian thank you for your unwavering belief and support. To my sister Sonja, I am so grateful to have had your unconditional support, whether it was just listening to me or making me my nth cup of tea, thank you. To my partner Harry, thank you for being a constant source of strength and encouragement, and for helping me to keep life in perspective.

Finally, I would like to gratefully acknowledge the families and infants who have participated in this study, and the support I received from the Australian Postgraduate Awards, Murdoch University and the Preterm Birth Centre of Research Excellence that has made this thesis possible.

A preterm infant in the King Edward Memorial Hospital NICU, Perth WA

x Abbreviations

Δ Delta °C Degrees Celsius g Microgram L Microlitre 7-AAD 7-amino-actinomycin D Ab Antibody AGRF Australian Genome Research Facility APP Antimicrobial proteins and peptides ATCC American Type Culture Collection ATP Adenosine triphosphate bp Base pair BPI Bactericidal/permeability increasing protein BSA Bovine serum albumin CB Cord blood CBMC Cord blood mononuclear cell CD Cluster of differentiation cDNA Complementary DNA CFU Colony-forming units CL097 2-(ethoxymethyl)-1H-imidazo[4,5-c]quinolin-4-amine CoNS Coagulase-negative staphylococci CPM Count per million CRP C-reactive protein CS Cesarean section Ct Cycle threshold DE Differentially expressed DEG Differentially expressed gene DNA Deoxyribonucleic acid EC Escherichia coli EOS Early-onset sepsis FACS Fluorescence-activated cell sorting FC Fold change FCS Fetal calf serum FDR False discovery rate

xi FIRS Fetal inflammatory response syndrome fMLP N-Formylmethionine-leucyl-phenylalanine FSC Forward scatter GA Gestational age GBS Group B Streptococcus G-CSF Granulocyte colony-stimulating factor gDNA Genomic DNA GM-CSF Granulocyte-macrophage colony-stimulating factor GOI Gene of interest HCA Histologic chorioamnionitis HI-FCS Heat-inactivated fetal calf serum HINT High-quality INTeractomes IFN Interferon alpha IFN Interferon gamma Ig Immunoglobulin IL Interleukin IPA Ingenuity® Pathway Analysis IRF3 Interferon Regulatory Factor 3 IvIg Intravenous immunoglobulin LDH Lactate dehydrogenase LOS Late-onset sepsis LPS Lipopolysaccharide LTA Lipoteichoic acid MBL Mannose-binding lectin M-CSF Macrophage colony-stimulating factor MDP Muramyl dipeptide MFI Median fluorescence intensity MHC II Major histocompatibility complex class II MOI Multiplicity of infection mRNA Messenger RNA NOD Nucleotide-binding oligomerisation domain OD Optical density PAMP Pathogen-associated molecular pattern PBMC Peripheral blood mononuclear cells PBS Phosphate buffered saline xii PCR Polymerase chain reaction PMA Phorbol myristate acetate PPROM Preterm premature rupture of the membranes PRR Pattern recognition receptor PS Phosphatidylserine R-848 Resiquimod RIG-1 Retinoic acid-inducible gene 1 RIN RNA integrity number RNA Ribonucleic acid RNA-seq RNA sequencing ROS Reactive oxygen species RPKM Reads Per Kilo-base of exon Per Million mapped reads rpm Revolutions per minute rRNA Ribosomal RNA RT-PCR Real-time polymerase chain reaction SE Staphylococcus epidermidis SEM Standard error of the mean SSC Side scatter SVD Spontaneous vertex delivery TLR Toll-like receptor TGF- Transforming growth factor beta TNF Tumour necrosis factor alpha TREM-1 Triggering Receptor Expressed on Myeloid cells 1 TriDAP L-Ala-gamma-D-Glu-mDAP UN Unstimulated UV Ultraviolet WB Whole blood WT Wild type

xiii List of Figures Figure 1.1 The potential sites of intrauterine bacterial infection...... 5

Figure 2.1 Sequential cell sort gating strategy for monocyte purification...... 40

Figure 3.1 Sequential cell sort gating strategies for monocyte purification by positive and negative selection...... 56

Figure 3.2 Comparison of IL-6 and TNF gene expression by positively and negatively sorted monocytes...... 58

Figure 3.3 Comparison of cytokine production by positively and negatively sorted monocytes...... 59

Figure 3.4 Post-sort monocyte sample purities and yields using the optimised cell sorting protocol...... 60

Figure 3.5 Growth curves and viability of S. epidermidis and E. coli...... 63

Figure 3.6 Post-sort measurements of monocyte RNA quantity and purity over time...... 65

Figure 3.7 Post-sort levels of IL-6 and TNF gene and protein expression by monocytes of over time. .. 66

Figure 3.8 Post-sort measurements of LDH release over time...... 67

Figure 3.9 IL-6 and TNF gene expression following stimulation with increasing MOIs of E. coli or S. epidermidis...... 69

Figure 3.10 IL-6 and TNF protein expression following stimulation with increasing MOIs of E. coli or S. epidermidis...... 70

Figure 3.11 Monocyte lysis following stimulation with increasing MOIs of E. coli or S. epidermidis...... 71

Figure 3.12 Frequencies of apoptotic monocytes following stimulation with increasing MOIs of E. coli or S. epidermidis...... 73

Figure 3.13 Measures of monocyte RNA quantity and purity following stimulation with increasing MOIs of E. coli or S. epidermidis...... 74

Figure 3.14 Bioanalyzer electropherogram summaries...... 79

Figure 4.1 Proportions and absolute counts of lymphocyte and monocyte populations within CBMC. .... 87

Figure 4.2 Proportions and absolute counts of infant monocyte subsets...... 89

Figure 4.3 Median fluorescence intensity of CD14 on classical monocytes...... 89

Figure 4.4 HLA-DR staining on infant monocyte subsets...... 90

Figure 4.5 RNA-seq read and mapping statistics...... 92

Figure 4.6 Cluster analysis of all monocyte samples...... 93

Figure 4.7 Principal component analysis...... 94 xiv Figure 4.8 Normalised log2 gene expression values for key genes involved in TLR signalling...... 97

Figure 4.9 Normalised log2 gene expression values for key monocyte function genes...... 98

Figure 4.10 Genes differentially expressed by preterm HCA+ infant monocytes...... 99

Figure 4.11 Diagramatic representation of the direct sample comparisons performed...... 101

Figure 4.12 Diagrammatic representation of statistic and bioinfomatic analyses performed...... 111

Figure 4.13 Differentially expressed genes following monocyte stimulation with E. coli...... 114

Figure 4.14 Differentially expressed genes following monocyte stimulation with S. epidermidis...... 115

Figure 4.15 P-value correlations for significant upstream transcriptional regulators of the E. coli-induced monocyte response...... 121

Figure 4.16 Upstream regulator analysis of the top 2000 monocyte differentially expressed genes induced by E. coli...... 122

Figure 4.17 IPA generated networks for ERK, IL13 and IRF1...... 124

Figure 4.18 Upstream regulator analysis of unique monocyte responses to E. coli...... 125

Figure 4.19 The downstream targets of IFNG...... 125

Figure 4.20 P-value correlations for significant upstream transcriptional regulators of the S. epidermidis- induced monocyte response...... 127

Figure 4.21 Upstream regulator analysis of the top 2000 monocyte differentially expressed genes induced by S. epidermidis...... 128

Figure 4.22 IPA generated networks for Lh, SGK1 and CEBPE...... 129

Figure 4.23 Normalised log2 fold change values for genes identified through delta-fold change differential expression analysis...... 134

Figure 4.24 Comparisons of CD14 and CD16 gene expression between preterm and term infant unstimulated monocytes...... 135

Figure 4.25 Preterm and term infant monocytes express similar levels of genes associated with CD16+ or CD16- monocytes...... 136

Figure 4.26 Log2 expression values for genes most highly associated with CD16+ and CD16- monocyte subsets...... 137

Figure 4.27 Protein levels of inflammatory cytokines and chemokines in CBMC culture supernatants. 141

Figure 4.28 Protein levels of inflammatory cytokines and chemokines in monocyte culture supernatants...... 142

Figure 4.29 Monocyte gene and protein expression is highly correlated after 2 hours of culture...... 144

xv Figure 4.30 Early monocyte gene expression (2 hours) highly correlates with late protein expression (24 hours)...... 145

Figure 5.1 Differentially expressed genes specific and common to each stimuli...... 155

Figure 5.2 Cluster analysis and heatmaps based on EC- and SE-specific monocyte response genes...... 156

Figure 5.3 Upstream regulator analysis of the E. coli-specific response genes...... 162

Figure 5.4 Networks for the most significant E. coli-specific upstream transcriptional regulators...... 163

Figure 5.5 IRF3 is activated by multiple upstream transcriptional regulators in the monocyte response to E. coli...... 164

Figure 5.6 The most significant sub-network within the E. coli-specific neonatal monocyte response. .. 166

Figure 5.7 Networks for the significant S. epidermidis-specific upstream transcriptional regulators...... 168

Figure 5.8 The most significant sub-network within the S. epidermidis-specific neonatal monocyte response...... 170

Figure 5.9 Upstream regulator analysis of the conserved neonatal monocyte response to E. coli and S. epidermidis...... 177

Figure 5.10 The most significant sub-network within the conserved neonatal monocyte response to E. coli and S. epidermidis...... 179

Figure 5.11 The top 20 nodes common across the most significant E. coli and S. epidermidis response sub- networks...... 182

Figure 5.12 Nodes unique to the E. coli or S. epidermidis response...... 183

xvi List of Tables Table 1.1 A summary of studies assessing immune molecule expression by preterm infant monocytes ... 23

Table 1.2 The advantages of RNA-seq compared with DNA microarray (232) ...... 30

Table 2.1 Broth compositions (g/L) ...... 36

Table 2.2 Details of monoclonal antibodies used for monocyte purification and phenotyping ...... 39

Table 2.3 Antibody and recombinant standard information ...... 42

Table 2.4 Characteristics of in-house designed primers ...... 46

Table 2.5 Sanger sequencing of PCR amplicons showing alignment of primers ...... 46

Table 2.6 Biological replicates ...... 48

Table 3.1 Positive and negative cell sort staining panels ...... 56

Table 3.2 Comparison of positive and negative cell sorting outcomes ...... 57

Table 3.3 Measurements of RNA quantity, purity and integrity comparing total cell numbers and RNA extraction kit...... 62

Table 3.4 Raw read statistics from a preliminary RNA-seq experiment ...... 78

Table 3.5 E. coli genome mapping statistics ...... 78

Table 4.1 Cohort demographics ...... 85

Table 4.2 Numbers of differentially expressed genes between unstimulated monocytes from each infant group ...... 102

Table 4.3 Differentially expressed genes (p<0.05) in unstimulated monocytes that are specific to preterm HCA- infants (2 genes) ...... 102

Table 4.4 Differentially expressed genes (p<0.05) in unstimulated monocytes that are specific to preterm HCA+ infants (17 genes)* ...... 102

Table 4.5 Differentially expressed genes (p<0.05) in unstimulated monocytes that are specific to term infants (7 genes)* ...... 103

Table 4.6 Numbers of differentially expressed genes between E. coli-stimulated monocytes from each infant group ...... 104

Table 4.7 Differentially expressed genes (p<0.05) in E. coli-stimulated monocytes that are specific to preterm HCA+ infants (18 genes)* ...... 104

Table 4.8 Differentially expressed genes (p<0.05) in E. coli-stimulated monocytes that are specific to term infants (10 genes)* ...... 105

xvii Table 4.9 Numbers of differentially expressed genes between S. epidermidis-stimulated monocytes from each infant group ...... 106

Table 4.10 Differentially expressed genes (p<0.05) in S. epidermidis-stimulated monocytes that are specific to preterm HCA+ infants (159 genes)* ...... 106

Table 4.11 Differentially expressed genes (p<0.05) in S. epidermidis-stimulated monocytes that are specific to term infants (1 gene) ...... 109

Table 4.12 Genes expressed in the opposite direction by monocytes from each infant group ...... 116

Table 4.13 Overlap between genes significantly differentially expressed in response to stimulation and those identified in direct sample comparisons (section 4.5) ...... 117

Table 4.14 Significantly over-represented canonical pathways associated with the unique term infant monocyte response to E. coli ...... 119

Table 4.15 Significantly over-represented diseases or bio-functions associated with the unique preterm HCA- infant monocyte response to E. coli or S. epidermidis ...... 119

Table 4.16 The twenty most significant upstream transcriptional regulators associated with the top 2000 differentially expressed genes induced by E. coli for each infant group ...... 121

Table 4.17 Upstream transcriptional regulators significantly associated with the unique monocyte response to E. coli for each infant group ...... 123

Table 4.18 The twenty most significant upstream transcriptional regulators associated with the top 2000 differentially expressed genes induced by S. epidermidis for each infant group ...... 126

Table 4.19 Upstream transcriptional regulators significantly associated with the unique monocyte response to S. epidermidis for each infant group ...... 129

Table 4.20 Sub-network analyses for unique monocyte responses to E. coli ...... 131

Table 4.21 Sub-network analyses for unique monocyte responses to S. epidermidis ...... 132

Table 4.22 Numbers of differentially expressed genes (Venn approach) that are significantly differentially expressed between infant groups ...... 133

Table 4.23 Comparisons between monocyte subset specific genes, and genes significantly differentially expressed between unstimulated monocytes from each infant group ...... 138

Table 5.1 The top twenty most significantly differentially expressed genes specific to the neonatal monocyte response to E. coli ...... 157

Table 5.2 Significantly over-represented canonical pathways associated with the E. coli-specific response genes ...... 159

Table 5.3 Significantly over-represented diseases and bio-functions associated with the E. coli-specific response genes ...... 159 xviii Table 5.4 Upstream transcriptional regulators significantly associated with the E. coli-specific response genes ...... 161

Table 5.5 The top ten sub-network hubs regulating the neonatal monocyte response specific to E. coli. 165

Table 5.6 The top twenty most significantly differentially expressed genes specific to the neonatal monocyte response to S. epidermidis ...... 167

Table 5.7 Upstream transcriptional regulators significantly associated with the S. epidermidis-specific neonatal monocyte response ...... 168

Table 5.8 The top ten sub-network hubs regulating the neonatal monocyte response specific to S. epidermidis ...... 169

Table 5.9 The top thirty most significantly differentially expressed genes within the conserved neonatal monocyte response to E. coli and S. epidermidis ...... 172

Table 5.10 The top twenty significantly over-represented canonical pathways associated with the conserved neonatal monocyte response to E. coli and S. epidermidis ...... 173

Table 5.11 The top twenty significantly over-represented diseases and bio-functions associated with the conserved neonatal monocyte response to E. coli and S. epidermidis ...... 174

Table 5.12 The top twenty significant upstream transcriptional regulators associated with the conserved neonatal monocyte response to E. coli and S. epidermidis ...... 176

Table 5.13 The top twenty sub-network hubs regulating the conserved neonatal monocyte response to E. coli and S. epidermidis ...... 178

Table 5.14 The top 20 nodes conserved across all infant groups in the E. coli response network ...... 181

Table 5.15 The top 20 nodes conserved across all infant groups in the S. epidermidis response network ...... 181

xix

xx Chapter 1 Literature review

1.1 An introduction to preterm birth

1.1.1 Epidemiology Preterm birth occurs prior to 37 weeks of completed gestation. The incidence of preterm birth varies among developed countries, with recent reports indicating that rates of preterm deliveries are ~6–11% across Australia and Europe, and close to 13% in the United States (1-3). This equates to ~25,000 preterm births each year in Australia, and close to 15 million preterm births globally (1, 4). Despite exceptional advances in perinatal care over the past two decades the incidence of preterm birth has increased by up to 20% in developed countries, making preterm birth a significant public health concern (5).

Preterm births are classified according to gestational age (GA) at the time of delivery. Late-preterm infants are born at 34–36 weeks GA and account for the majority of preterm births (60-70%). Moderately preterm infants (31–33 weeks GA) account for ~20%, while very preterm infants (28–30 weeks GA) account for ~15% and extremely preterm infants (<28 weeks GA) make up only ~5% of all preterm births (2). The causes of preterm birth can be broadly classified into three clinical presentations: 1) medically induced preterm birth due to maternal or fetal indications (~30% of all preterm deliveries), 2) spontaneous onset of preterm labour with intact fetal membranes (~40%), and 3) spontaneous onset of preterm labour following premature rupture of membranes (PPROM; rupture of membranes before 37 weeks gestation and more than 1 hour prior to the onset of labour; ~30%) (2, 6).

Several factors contribute to the overall increase in the incidence of preterm births, including increases in maternal age, use of assisted reproductive technologies, rates of multiple gestations (twins, triplets etc.), and higher numbers of indicated preterm deliveries (6-9). However these observations are not entirely responsible for the increasing trend as the cause of preterm birth is unknown in ~50% of cases, highlighting a vast knowledge gap in the mechanisms underlying the overall increased rates of preterm birth (4, 9).

1 1.1.2 The significance of preterm birth Preterm infants are extremely vulnerable to life-threatening infections and account for 75% of all neonatal morbidity and mortality, translating to more than 1 million deaths each year globally (10, 11). Prematurity is the second leading cause of death in children under 5 years old (following pneumonia) and those who survive, experience high rates of morbidity resulting in extensive human and health care costs (10, 12).

Aside from the high emotional costs for parents associated with preterm birth, the total economic burden resulting from immediate medical care services, early intervention services, special education and lost productivity costs are $26.2 billion per year in the United States alone (13) . Birth weight and GA are inversely correlated with infant mortality and morbidity, and extremely preterm infants (<28 weeks GA) cause more than one third of the total costs associated with preterm birth (14). The most common complications of extreme prematurity are respiratory distress syndrome (93%), patent ductus arteriosus (46%), late-onset sepsis (36%), severe intraventricular haemorrhage (10%) and necrotizing enterocolitis (7.5%) (12, 15). Infections such as sepsis and necrotizing enterocolitis are leading causes for mortality in the neonatal intensive care units in Australia (15). The consequences of infection however are not limited to the direct mortality, as harmful effects of the induced inflammation contribute to increased risks of long-term complications such as bronchopulmonary dysplasia and central nervous system injury (16).

The negative outcomes associated with prematurity do not only affect those born extremely preterm, but also affect those born close to term (late preterm infants). The mortality and morbidity rates observed among late preterm infants is still several times higher than in term infants (1, 17), and contributes substantially to the health care burden. Specifically, late preterm infants are at higher risk of respiratory distress, intensive and prolonged hospitalisation, hypoglycaemia, hyperbilirubinaemia, hypothermia, feeding problems and invasive infection, as well as adverse neurocognitive outcomes (17-19). Importantly, the negative consequences of preterm birth persist beyond the neonatal period and extend even into adulthood (20).

2 1.1.3 Risk factors for preterm birth A number of genetic and environmental factors affecting the mother, fetus or placenta are associated with an increased risk of preterm birth. Maternal genetics in particular contribute significantly to the risk of preterm birth; women whose mothers or sisters gave birth to preterm infants, or who were born preterm themselves have an increased risk of preterm labour (risk ratios 1.4–1.95) (21, 22). The molecular mechanisms that govern these broad observations however, have not been determined. There is a body of evidence (although somewhat inconsistent) to suggest that genetic variations in specific genes of innate immune and inflammatory biological pathways (TNF, IL4, IL6, IL10, LTA, MBL2, TLR2, TLR4, NOD, MMP1 and MMP9) are associated with heightened risk of preterm birth, as these pathways are known to play a role in the onset of labour (23).

Maternal risk factors for preterm birth include maternal smoking, low socioeconomic status, multiple gestation pregnancy, previous preterm delivery, low or high maternal age, periodontitis, preeclampsia, stress and bacterial vaginosis (2, 4, 24-27). The placenta plays an indispensible role in fetal growth and development, and placental abnormalities are associated with an increased risk of preterm birth. In particular infection and inflammation of the placenta is very highly associated with preterm birth (affecting up to 85% of deliveries <28 weeks) (28, 29). The biological processes and consequences of placental infection and inflammation are discussed in detail in section 1.2.1. Other placental complications associated with preterm birth are haemorrhage, ischemia, placental abruption, and placental previa (placental obstruction of the cervix) (30, 31). Fetal intrauterine growth restriction is an important risk factor for preterm birth, and is closely linked to placental pathologies such as the presence of fibrosis and avascular placental villi (32).

1.2 Infections and preterm birth

Perinatal and postnatal infections result in systemic inflammation that is central to many adverse outcomes of extreme prematurity. The following sections will summarise the significance of antenatal and postnatal infection and inflammation in preterm infants.

3 1.2.1 Intrauterine infection and inflammation The spectrum of potential sites for intrauterine infection and inflammation are presented in Figure 1.1. The most common type of intrauterine infection and inflammation associated with preterm birth is histologic chorioamnionitis (HCA), which complicates 40-70% of preterm deliveries, most commonly following PPROM (33). HCA is defined as inflammation of the umbilical cord and/or the placental disc and membranes. This is characterised by neutrophil infiltration, amnion basement membrane thickening and the development of necrosis in more severe cases (34, 35). Bacterial invasion of the membranes activates the innate immune system, triggering an inflammatory cascade that culminates in uterine contractions (through release of prostaglandins) and rupture of the chorioamniotic membranes (through metalloproteases produced by neutrophils) (33). Ureaplasma urealyticum and Mycoplasma hominis, commensal inhabitants of the lower genital tract, are the predominant causative organisms of HCA and ascend from the vagina and cervix (36). These organisms are of relatively low virulence and cause low- grade chronic infection that remains undetected until the resulting inflammation leads to the onset of labour (33). Recently, the use of non-culture based microbiological techniques suggests that the placenta is not a sterile site, even among healthy term pregnancies (37). Advancing our understanding of the placental microbiome will in turn aid our understanding of the pathogenesis of HCA, and potentially direct development of new treatment strategies.

There are several important consequences of exposure to HCA, especially among preterm infants. The fetal inflammatory response syndrome (FIRS) is believed to be an extension of placental/amniotic fluid infection, and is characterised by increased levels of the inflammatory cytokine IL-6 in fetal circulation. Preterm infants exposed to HCA and those with FIRS experience heightened susceptibility to early-onset sepsis, respiratory disorders, intraventricular haemorrhage, and death (38-40) Interestingly, exposure to HCA is associated with a decreased risk of developing late-onset sepsis (41). The fact that preterm infants exposed to HCA exhibit an altered risk of acquiring invasive infections suggests that prenatal exposure to infection may modulate the developing innate immune system. Our current limited understanding of how exposure to HCA may modulate innate immunity (specifically the monocyte) is presented in section 1.5.

4

Figure 1.1 The potential sites of intrauterine bacterial infection. Adapted from Goldenberg et. al. (33). (A) Bacterial species are thought to ascend the urogenital tract to infect the choriodecidual space (B). Infection may then spread throughout the amnion and chorion (chorioamnionitis) (C) to infect the amniotic cavity (D). Fetal infection is then possible via bacterial migration into the umbilical cord (funisitis) (E) or by direct inhalation (F).

5 1.2.2 Early-onset neonatal sepsis The clinical signs and symptoms of neonatal sepsis are non-specific and include temperature instability, respiratory distress, apnoea, lethargy or irritability, jaundice and feed intolerance (42). Sepsis is confirmed by microbial growth in a normally sterile bodily fluid (blood, urine or cerebrospinal fluid culture) (43).

Early-onset sepsis (EOS) is an important cause of neonatal morbidity and mortality, especially among very low birth weight preterm infants (44). EOS is defined as culture- proven blood, urine or cerebrospinal fluid bacterial infection that occurs within 72 hours of life, however upwards of 90% of cases manifest within 24 hours (45-48). EOS is the result of fetal colonisation and infection prior to or during delivery, and may be an extension of FIRS (49). Mortality rates associated with EOS are as high as 38% among the preterm population, with most deaths occurring between 0–3 days of life (44).

Group B Streptococcus (GBS) and Escherichia coli are the leading causes of EOS in newborn infants (44, 50). Importantly, there has been a significant reduction in the overall incidence of EOS since the implementation of guidelines in the late 1990s for universal maternal GBS screening and the introduction of intrapartum antibiotic prophylaxis (from 5.9 to 1.7/1000 live births) (51). In contrast, the rates of E. coli EOS have remained stable or may have increased slightly in the era of GBS prophylaxis, particularly in preterm infants where E. coli is now the most common cause of EOS (~40% of episodes) (44, 50- 55).

Aside from prematurity, the most significant risk factors for EOS are linked to maternal factors such as prolonged premature rupture of membranes (>24 hours prior to delivery) and clinical or histologic chorioamnionitis, particularly when both are present (53).

1.2.3 Late-onset neonatal sepsis Late-onset sepsis (LOS) is distinct from EOS in several aspects. LOS occurs after 72 hours of life, with the peak incidence occurring between the 10th and 22nd day of life (56). The decreasing incidence of EOS observed over the last few decades has been paralleled by an increased incidence of LOS, largely due the improved survival of preterm infants (57, 58). Unlike the involvement of perinatal colonisation in EOS, LOS is typically the result of nosocomial infection by organisms that colonise the infant postnatally. LOS

6 predominately affects extremely preterm infants, and is several times more common among this population than EOS, affecting up to one third of extremely preterm infants in Australia (15).

Late-onset sepsis is most frequently caused by Gram-positive organisms, particularly coagulase-negative staphylococci (CoNS), which account for about two thirds of LOS in developed countries (57, 59-62). The most prevalent clinical isolate of CoNS is Staphylococcus epidermidis, which is major constituent of the normal skin and mucosa flora (63). However the biofilm forming capacity among other virulence factors of S. epidermidis allows for efficient adaptation to the nosocomial environment, and S. epidermidis is now widely recognised as an important opportunistic pathogen (64). The remainder of LOS infections are due to other CoNS species (S. haemolyticus, S. lugdunensis, S. schleiferi etc.)(65), and Staphylococcus aureus, E. coli, Klebsiella, Enterobacter, Pseudomonas, and Candida species (all associated with less than 10% of cases in Australia) (56).

Fewer infants die of LOS (1.5–10.2%)(66) compared to EOS, except for cases of Gram- negative or fungal LOS, which carry a 23% mortality rate (67). Preterm infants who survive LOS are at increased risk of several long-term adverse outcomes, as sepsis- induced inflammation can result in brain injury and neurodevelopmental impairment (68, 69). Therefore, while LOS caused by S. epidermidis does not carry mortality rates as high as other LOS-causing species, S. epidermidis infections are four to five times more common and contribute to the majority of LOS-related morbidity (70, 71).

In contrast to EOS, maternal factors do not play a significant role in the risk of LOS. The improved survival of preterm infants has lead to increased duration of neonatal intensive care hospitalisation, the length of which is significantly prolonged in infants who develop LOS (72). This may be linked to the increased use of indwelling devices such as central or peripheral venous catheters that are frequently required for administration of medication and parenteral nutrition in preterm infants, or mechanical ventilation via endotracheal or nasal tubes, which are significantly associated with an increased risk of LOS (73, 74). Intrinsic or acquired immunological factors also contribute to an infant’s risk of sepsis. Preterm infants who develop LOS receive significantly less breast milk, depriving them of a variety of immune factors that have antimicrobial activity against sepsis-causing pathogens (75). In addition, single nucleotide polymorphisms affecting

7 innate immune genes (particularly CD14 and IL8) are associated with increased severity of LOS (76). The administration of parenteral nutrition is also an independent risk factor for LOS (73), and is associated with impairments in infant innate immunity; reduced whole blood TNF production in response to CoNS, and reduced CoNS phagocytosis and killing (77, 78).

Importantly, up to half of very preterm infants who develop LOS have no known risk factors other than prematurity, further highlighting the importance of immune defences in these infants. Interestingly, prenatal exposure to histologic chorioamnionitis is associated with a reduced risk of developing LOS in preterm infants suggesting that exposure to infection and inflammation in utero may impact development of neonatal immunity (41). The immunological risk factors that may predispose preterm infants to sepsis are discussed further in section 1.4.

1.3 Pathogenesis of neonatal sepsis

The pathophysiology of sepsis is complex and incompletely understood. Early-onset sepsis develops as a result of perinatal colonisation and infection, evidenced by the strong association between chorioamnionitis and the risk of EOS (53). Systemic fetal infection may occur in utero via inhalation of infected amniotic fluid, which is often contaminated by GBS and E. coli (79). E. coli strains containing the K1 capsular polysaccharide are particularly virulent (through the degradation of complement proteins), and are frequently isolated in neonatal meningitis (80-83). The pathogenesis of CoNS sepsis has been linked to the use of colonised intravenous catheters, however not all infants who develop LOS are exposed to intravenous catheters, and molecular typing has only confirmed strain similarity between blood isolates and those from intravenous catheters in only two thirds of cases (84). An additional route of infection is via translocation of bacterial species from the gastrointestinal tract into the blood stream. Molecular similarity between CoNS isolated from the gastrointestinal tract and blood at the time of LOS was confirmed in more than 50% of cases (65, 85). Gut colonisation with invasive E. coli at birth has also been demonstrated and linked to neonatal sepsis (86). While the precise routes of E. coli or S. epidermidis infection in the neonate are not always known, it is the invasion of the bloodstream and subsequent activation of the immune system that ultimately leads to sepsis.

8 Recognition of bacterial products such as lipoteichoic acid (LTA) and lipopolysaccharide (LPS) initiates the wide-spread production of pro-inflammatory mediators by the innate immune system, characteristic of the initial hyperinflammatory phase of sepsis (87). The persistence of these inflammatory mediators is associated with severe sepsis in neonates (88). In addition, the production of pro-inflammatory cytokines is responsible for perturbing the coagulation system during adult sepsis, inducing a pro-coagulative state that contributes to vascular occlusion and organ failure, the main cause for sepsis-induced mortality (89).

Innate immunity is the first line defence against infection, and preterm infants’ immaturity of innate immunity likely is central to their unique susceptibility to invasive infections (reviewed in section 1.4). Monocytes are especially important as they are present in the blood and recognise a diverse range of pathogens through pattern recognition receptors, and are potent producers of inflammatory cytokines (90). Indeed, the expansion of the pro-inflammatory CD16+ monocyte numbers in the blood is observed in neonatal, paediatric and adult sepsis (91, 92). Inflammation clearly plays an important role in the early pathogenesis of sepsis, so much so that the term “cytokine storm” was coined in the 1990s to explain the immune response to sepsis (93). This terminology has since been replaced with the “Systemic Inflammatory Response Syndrome” to encompass clinical signs (fever or hypothermia, tachycardia, tachypnoea and changes in the blood leukocyte count) (94). Importantly, a range of anti-inflammatory mediators are also induced in sepsis, and it is becoming increasingly apparent that the dysregulation of inflammation and the innate immune response plays a significant role in the pathogenesis of sepsis (95).

1.3.1 Regulation of inflammation in neonatal sepsis Inflammation is regulated by various immune cell types, and is largely mediated through the effects of anti-inflammatory cytokines such as IL-4, IL-10 and TGF- that are rapidly produced alongside pro-inflammatory cytokines during neonatal sepsis (88). In vitro stimulation of whole cord blood with LPS revealed that neonates not only have a reduced capacity for IL-10 and TGF- production, but that the inhibitory effects of IL-10 and TGF- were diminished in neonates compared to adults (smaller reductions in LPS- induced IL-1, IL-6, IL-8, and TNF production were observed), an effect that may predispose neonates to the harmful affects of inflammation during sepsis (96). In preterm infants with LOS, high IL-6/IL-10 ratios were associated with severe sepsis, suggesting

9 that an inadequate anti-inflammatory response may be contribute to the pathogenesis of sepsis in neonates (97). Furthermore, the timing and duration of immunosuppression during the course of sepsis may be important, as systemic immunosuppression has been associated with sepsis related mortality among adults (98). In addition, a murine leukocyte transcriptome profiling study revealed that significant attenuation of the neonatal inflammatory response following the initial hyper-inflammatory phase of sepsis was associated with greater mortality in an experimental model of polymicrobial sepsis (99). The expansion of immune-regulatory cell types such as T regulatory cells has been associated with the early stages of adult sepsis (100), and increased frequencies of myeloid-derived suppressor cells (which supress CD8+ T cell IFN production) was observed in a polymicrobial murine model of sepsis (101). The frequency of T regulatory cells is inversely correlated with gestational age with significantly higher frequencies in preterm infants compared to term infants, which may contribute to differential immune regulation during invasive infection (102). However, no studies comparing preterm and term infant T regulatory cell function have been performed, which would be required to determine the potential clinical impact of the increased frequencies of these cells in preterm infants. The frequency of myeloid-derived suppressor cells is also significantly increased in human neonates (103, 104), however this has not been investigated in the context of preterm birth, and the potential role for these cells in neonatal sepsis has not been determined.

The pathogenesis and outcome of neonatal sepsis is likely to be much more complex than a simple balance between pro- and anti-inflammatory immune mechanisms. More than 100 phase II and III clinical trials (predominantly in adults) have aimed to modify systemic inflammation during sepsis either non-selectively (e.g. via corticosteroids, ibuprofen, polyvalent immunoglobulin) or by selective neutralisation of microbial or host inflammatory mediators (e.g. via anti-LPS, anti-TNF, IL-1 receptor agonists) (105). However overall, no specific strategy has lead to improved survival, indicating that we have an incomplete understanding of the complex molecular and biological mechanisms involved in the pathogenesis of the disease. Understanding the earliest events in host- pathogen interactions may provide greater insight into the pathogenesis of sepsis.

10 1.3.2 Pathogen-specific responses to sepsis pathogens Sepsis evolves through host recognition of pathogens and their products, suggesting that different invasive organisms may elicit distinct host responses during sepsis, as well as common activation pathways. Gram-negative, Gram-positive and fungal organisms express a diverse range of virulence factors and other molecules capable of interacting with innate and adaptive immune defences (106). The fact that patient prognosis is linked to causative organism (i.e. mortality rates are higher for Gram-negative sepsis) also suggests there are pathogen-dependent factors involved in the pathogenesis and outcome of sepsis. Innate immune responses to E. coli primarily occur through Toll-like receptor (TLR)-4 recognition of LPS (dependent on the co-receptor MD-2) (107), whereas the recognition and clearance of S. epidermidis is mediated (at least in part) by TLR2 (108, 109). Both Gram-negative and Gram-positive pathogens are also recognised by an overlapping repertoire of TLRs through the shared expression of several pathogen- associated molecular patterns (bacterial DNA, lipoproteins, flagellin and peptidoglycan) (110). Interestingly, serum levels of IL-6, TNF, C- reactive protein (CRP) and fibronectin were similar in neonatal sepsis caused by either Gram-positive or Gram- negative bacteria, but the levels of soluble CD14 were significantly increased in Gram- negative infections, likely reflecting the role of CD14 as a co-receptor for LPS (111). Furthermore, whole-transcriptome profiling of peripheral blood mononuclear cells in critically ill adult sepsis patients revealed no difference in gene expression profiles between Gram-positive and Gram-negative septic patients (112). These results suggest that a non-specific, innate immune response is prominent during sepsis, and is activated by structurally distinct pathogens. To date, no studies have identified or characterised this type of conserved innate response in neonates.

1.4 Development of neonatal innate immune defences are critical for protection against bacterial pathogens

The neonatal immune system is adequate for protection against infection in most term infants; only ~1% of term infants develop EOS or LOS in Australia (15). In contrast, preterm infants experience substantially higher rates of infection (~33% of extremely preterm infants develop LOS)(15), and it is increasingly recognised that immature immune defences contribute to their susceptibility. Identifying areas of greatest disparity

11 between preterm and term infant immune function will highlight the immune defences that are most vital for protection against sepsis.

Mammalian host defences are comprised of the innate and adaptive immune systems. The human adaptive immune system is critical for mounting an antigen-specific lymphocyte response and for long-lasting protection against pathogens through maintenance of memory lymphocytes (including B cells, CD4 helper T cells and CD8 cytotoxic T cells). However the role of adaptive immunity in early life is less prominent. While neonates generally display immature adaptive immunity compared to adults, preterm infants exhibit similar adaptive immune profiles to term infants.

Preterm (≥31 weeks) and term infants have similar frequencies and absolute counts of circulating CD19+ B cells, CD4+ helper T cells and CD8+ cytotoxic T cells (113, 114). The majority (>90%) of these lymphocyte subpopulations in preterm and term infants maintain a naïve phenotype even up to three months of age (114). Furthermore, the splenic marginal zone (an important site of IgM responses and B cell maturation) is not fully developed in humans until 2 years of age (115). These findings demonstrate that the cellular composition of adaptive immunity is similar between preterm and term infants, and is predominantly naïve until well past the time when preterm infants are most at risk of sepsis (2–3 weeks for LOS).

To counter this adaptive immune immaturity, maternal antibodies are passively transferred to the fetus in the form of maternal immunoglobulin G (IgG) via the placenta as early as 16 weeks gestation, and continues to increase in a linear fashion with increasing gestational age (116). Therefore, preterm infants (especially those born at <33 weeks gestation) have significantly lower levels of IgG compared to term infants, and may lack protection provided by pathogen-specific maternal IgG (117). However clinical trials of the administration of intravenous polyclonal or anti-staphylococcal IgG as preventative or adjunct therapy for sepsis in preterm infants showed no significant reduction in the incidence of disease caused by Staphylococci or other species, or overall mortality (118-121).

Preterm infants also respond equally as well to vaccination as term infants and are routinely vaccinated on the same schedule, providing further evidence that adaptive immune defences are comparable between these two populations. A comprehensive

12 review of preterm infant responses to multiple T helper cell-dependent vaccines (tetanus, diphtheria, meningococcal C, pneumococcal, pertussis and polio) has been conducted (122). This review demonstrates that the majority (80–98%) of preterm infants born at ≤32 weeks produce similar levels of antigen specific antibodies to term infants, which reach protective titres within 1–2 months of vaccination.

In conclusion, both preterm and term infants have naïve and developing adaptive immune defences, and therefore deficiencies in innate defences may account for the risk disparity observed between preterm and term infants in their susceptibility to sepsis. Herein we review aspects of the innate immune system most critical for defence against bacterial pathogens.

1.4.1 Maturation of innate immunity in neonates The innate immune system is present at birth, and represents the first line of defence against microbial invasion. Many soluble and cellular elements of the innate immune system are present in the blood (the site of sepsis), and these are the focus of this review.

1.4.1.1 Soluble innate immune factors There are a wide variety of non-cellular elements of innate immunity present in circulation including complement proteins, mannose-binding lectins and antimicrobial proteins and peptides. Complement proteins function to aid opsonisation and killing of pathogens, initiate the production of pro-inflammatory cytokines and the coagulation cascade, and leukocyte recruitment (123). The complement system is comprised of more than 30 proteins, many of which are quantitatively and functionally impaired in preterm infants (124-126). In both preterm and term infants with early-onset sepsis, the alternative pathway (triggered by direct interaction of C3 with foreign carbohydrates, lipids or proteins) is the primary route of complement activation, whereas markers of the classical pathway are not elevated (127, 128). This suggests that antibody-mediated complement activation (classical pathway) is not critical during neonatal sepsis, and perhaps partially explains the failure of intravenous IgG to prevent infection in neonates or improve outcome. Mannose-binding lectin (MBL) is a circulating pathogen recognition molecule produced by hepatocytes that recognises a wide range of microorganisms, and is a key inducer of the third complement activation pathway (lectin pathway) (123, 129). In addition to complement activation, MBL also provides host defence through interaction

13 with TLR2 and TLR6 within the phagosome, modulating the production of inflammatory cytokines (130). Plasma concentrations of MBL are reduced in preterm infants, an effect that is associated with an increased risk of sepsis (131, 132).

Antimicrobial proteins and peptides (APPs) are cationic oxygen-independent molecules primarily produced by neutrophils and monocytes (to a smaller extent) within circulation, and are also present human breast milk (133). APPs play an important role in innate immunity via direct killing of susceptible pathogens, as well as modulating immune responses via leukocyte activation and chemotaxis (134, 135). A variety of plasma APPs including lactoferrin, bactericidal/permeability increasing protein (BPI), LL-37 and secretory phospholipase A2 show gestational age-dependent increases in concentration within cord blood (independent of white cell or neutrophil count) (136, 137). Whether this GA-dependent deficiency of plasma APPs is due to diminished neutrophil granule stores remains to be determined. The most abundant APP in human breast milk is lactoferrin (in mothers of both preterm and term infants), which exerts significant direct microbicidal activity in vitro against several neonatal sepsis-causing pathogens including S. epidermidis and E. coli (75). Preterm infants who develop LOS however, consume significantly smaller volumes of breast milk (and therefore lower levels of lactoferrin and other APPs) than those who don’t develop sepsis (75). Several clinical trials of the administration of oral lactoferrin supplementation to preterm infants have shown a decrease in the incidence of LOS, suggesting that replenishment of APPs in preterm infants can be beneficial in a clinical setting (138). Together, these studies suggest that the availability of APPs is limited in preterm infants compared to term infants, whether through reduced levels in plasma (potentially through decreased numbers of neutrophils/monocytes) or via decreased consumption of APP-rich breast milk.

A collective gestational age-dependent deficiency in these non-cellular elements of innate immunity is likely to contribute to the increased risk of invasive bacterial infection observed among preterm infants. However none of these factors act alone, and are either produced by or require interactions with immune cells to competently provide defence against invading pathogens.

1.4.1.2 Innate immune cells in circulation Neutrophils, dendritic cells and monocytes are the significant phagocytic innate immune cell types in circulating blood. These cells develop from a common myeloid progenitor

14 in the bone marrow and differentiate into distinct cell types that perform specialised functions in response to infection (139). Neutrophils clear invading bacterial pathogens through phagocytosis, degranulation resulting in release of a range of bactericidal agents, and the formation of neutrophil extracellular traps (140). Neutrophils are present at relatively low frequencies early in gestation (~6% at ≤20 weeks), and exponentially expand after 31 weeks gestation to become the major leukocyte at term birth (~72% of all leukocytes) (141). Absolute counts of neutrophils in neonates (even in preterm infants) exceed those in adults (142), however neonates (particularly preterm infants) have significantly reduced neutrophil bone marrow pools, often resulting in neutropenia during the course of sepsis (143-145). Investigations of phagocytic capacity in whole blood assays have shown a significant impairment of preterm neutrophils to phagocytose E. coli compared to term infants (146). However this has been attributed to deficient opsonisation by complement and immunoglobulins, as others report competent bacterial phagocytosis by preterm infant neutrophils in vitro following supplementation with adult serum or immunoglobulin (147, 148). Studies of neutrophil oxidative burst (a mediator of bacterial killing) have yielded inconsistent results, with some reports of significantly impaired oxidative burst in preterm infants following whole blood stimulation with phorbol myristate acetate or S. epidermidis compared to term infants (149, 150), while others report competent oxidative burst by preterm neutrophils in response to E. coli compared to term infants (151). All of these studies had similar sensitivity to detect neutrophil oxidative burst (by flow cytometry), and therefore differences in study population or choice of stimuli may account for the inconsistencies; Drossou et. al. included preterm infants born earlier than Habermehl et. al. (<31 weeks vs. ≥34 weeks gestation), and perhaps a neutrophil oxidative burst deficiency is more strongly associated with very preterm rather than late preterm birth.

Impaired formation of neutrophil extracellular traps and a deficiency in killing of extracellular bacteria has been reported in neonates, however this effect was observed in both preterm and term infants, suggesting that preterm infant susceptibility to sepsis is not due to these deficiencies alone (152).

Neutrophil development and maturation is supported by the hematopoietic growth factors G-CSF and GM-CSF (153, 154). Several clinical trials have tested the use of these growth factors in the prevention or treatment of neonatal sepsis, but despite improvements in neutrophil counts and function (bacterial phagocytosis and oxidative burst) in treated

15 infants, the incidence and survival rates of sepsis were unchanged (155-157). Similarly, direct granulocyte transfusions for the treatment of neonatal sepsis had no effect on neonatal mortality (158). These findings suggest that neutrophil deficiencies alone are not responsible for preterm infant susceptibility to sepsis, however as previously mentioned, no studies to date have assessed granule formation and composition in preterm neutrophils. This type of study would be of interest, as neutrophil granules are a major source of APPs, the concentrations of which are significantly reduced in preterm plasma (136).

Circulating human dendritic cells are phenotypically and functionally similar to tissue dendritic cells (159), and play an important role during infection through their ability to present foreign antigens to T cells and orchestrate immune responses through the release of cytokines (159, 160). In contrast to neutrophils, dendritic cells (broadly classified as classical or plasmacytoid) comprise a very small proportion of circulating leukocytes, representing less than 1% of total peripheral blood leukocytes in healthy adults (161). The very limited literature on neonatal blood dendritic cells suggests that they are phenotypically immature compared to adults; a higher frequency of neonatal dendritic cells express CD34 (a marker of immature precursor cells), and fewer neonatal dendritic cells express receptors required for co-stimulation of T cells or IgG receptors (required for antibody-dependent uptake of antigen) (162). The study by Holloway et. al. analysed cord blood samples from a range of gestational ages (28–37 weeks) and described similar phenotypes in preterm and term infants. Others also report similar frequencies and absolute counts of dendritic cell subsets between preterm and term infants, with similar expression of co-stimulation (CD86) and pathogen recognition receptors (TLR2, TLR4) (113, 163). The only phenotypic dendritic cell deficiency associated with low gestational age was reduced expression of the co-stimulatory receptor CD80 on both dendritic cell subsets (113). Functional studies however demonstrate a significant impairment in the ability of preterm dendritic cells to produce IL-12/IL-23p40 (a cytokine critical for T cell activation) and IFN (a major cytokine produced by plasmacytoid dendritic cells) in response to a broad range of TLR agonists (163, 164). Increased serum concentrations of IFN is more often associated with viral than bacterial infection in infants and children (77% vs. 8.6%) (165), and while plasmacytoid dendritic cell production of IFN does play a role in the immune response to some Gram-positive bacteria (e.g. Staphylococcus aureus) (166), S. epidermidis does not induce this response (167). The role of plasmacytoid dendritic cell production of IFN during E. coli infection has not been 16 documented. However adult sepsis patients (28% with confirmed Gram-negative infection) exhibit significantly decreased absolute counts of peripheral blood plasmacytoid dendritic cells compared to healthy controls, with even lower counts in patients who died as a result of sepsis (168). Therefore, an impaired IFN response by preterm infant plasmacytoid dendritic cells is unlikely to fully explain the increased susceptibility of these infants to S. epidermidis infection, but may be associated with E. coli infection. To date there is no literature on the phagocytic capacity of preterm dendritic cells, most likely due to practical limitations of obtaining large enough samples from preterm infants.

Monocytes are key effector cells of innate immunity through their capacity to detect, phagocytose and kill foreign pathogens, as well as activate the adaptive immune system and trigger inflammation through the release of chemoattractant and inflammatory cytokines (90). Furthermore, monocytes are unique in their plasticity and are established precursors for macrophages and dendritic cells (distinct from myeloid or plasmacytoid dendritic cells) under inflammatory conditions (169). The literature concerning monocyte phenotype and function in preterm infants is reviewed in the following section and summarised in Table 1.1. Of note, the process of monocyte differentiation into macrophages in preterm infants has been investigated by one study, which reports reduced basal expression of HLA-DR and CD16 compared to term infants, but lacks functional data (170). No studies have compared monocyte-derived dendritic cells between preterm and term infants.

1.4.2 Monocytes and immune-regulation Monocytes appear early in gestation (prior to 20 weeks) and represent 7–38% of all circulating mononuclear cells in cord blood, and ~10–15% of peripheral blood mononuclear cells in healthy adults (141, 171-174). There are at least three phenotypically and functionally distinct human monocyte subsets (175), defined by their expression of CD14 (co-receptor for LPS) and CD16 (low affinity receptor for Fc region of immunoglobulin). The majority (~90%) of circulating monocytes are CD14++ CD16- , termed classical monocytes. The remaining monocyte pool is composed of intermediate (CD14++ CD16+) and non-classical (CD14+ CD16+) monocytes. Prior to the identification of these two distinct CD16+ monocyte subsets, they were collectively referred to as pro-inflammatory monocytes due their capacity to produce higher levels of

17 inflammatory cytokines compared to CD16- monocytes (176, 177). CD16-positive monocytes are also expanded in neonates, children and adults with sepsis suggesting an active role for these cells during invasive infection (91, 92). However intermediate monocytes are increasingly appreciated as a functionally distinct subset that may represent cells in transition between a classical and non-classical phenotype, with a higher capacity for antigen presentation and angiogenesis (178, 179). Non-classical monocytes are closely related to intermediate monocytes at a molecular level, but producer higher levels of TNF and IL-1 following LPS stimulation (180).

Preterm infants have similar frequencies of monocytes compared to term infants, but absolute counts of monocytes significantly increase with increasing gestational age (113). The limited literature on monocyte subsets in neonates is conflicting, with one report that the intermediate monocyte subset is significantly elevated in very preterm infants (24–32 weeks) compared to term infants (181), while others report similar frequencies of all three monocyte subsets between preterm (24–32 weeks) and term infants (182). These inconsistencies may stem from differences in flow cytometry data analysis, either due to the inclusion or exclusion of HLA-DR staining to define the monocyte population, or due to the subjective nature of manual gating strategies.

1.4.2.1 Expression and function of pattern recognition receptors Monocytes are equipped with a range of pattern recognition receptors (PRRs); ancient proteins critical for the recognition of pathogen associated molecular patterns that are highly conserved within a class of microbes. PRRs are expressed on the cell surface and intracellularly within vesicles or the cytosol. Examples of PRRs include the TLRs, nucleotide-binding oligomerisation domain (NOD)-like receptors, retinoic-acid- inducible protein I (RIG-1)-like receptors, and C-type Lectin receptors (CLRs) (183). CLRs recognise a variety of both self, and non-self ligands and can initiate both activatory and inhibitory inflammatory signalling pathways (184). To date, no studies have assessed the expression of CLRs on preterm monocytes. RIG-1-like receptors are best characterised for their role in viral detection, and have not been characterised in preterm infant monocytes specifically. However cultures of mixed cord blood mononuclear cells have demonstrated impaired RIG-1 dependent cytokine responses in preterm compared to term infants (185). The vast majority of the literature on PRRs in preterm monocytes has investigated the expression of TLRs, particularly TLR2 and TLR4 as the primary receptors for recognition of Gram-positive and Gram-negative bacteria respectively.

18 These investigations however have yielded inconsistent results (Section 1, Table 1.1), with both decreased and equivalent TLR2 and TLR4 protein expression reported in preterm and term infant monocytes, which may be accounted for by differing experimental protocols. The only study to detect mRNA levels of TLR2 and TLR4 in purified monocytes demonstrated increased basal levels of both receptors in preterm infants relative to term infants (186).

CD14 and MD-2 are co-receptors for the detection of both Gram-negative and Gram- positive bacterial cell wall constituents alongside TLR2 and TLR4, and in fact MD-2 is required for TLR4-mediated responsiveness to LPS and enhances TLR2 responses (187). Despite the importance of MD-2 in the monocyte response to bacteria, only one study has assessed the expression of MD-2 on preterm monocytes and found significantly reduced surface expression compared to term infants (188). Studies on preterm monocyte surface expression of CD14 are also inconsistent, with both equivalent (189, 190), and reduced expression compared to term infants reported (188).

The NOD-like receptors NOD1 and NOD2 are expressed in the cytosol and recognise distinct structural motifs of peptidoglycan present on all Gram-positive and Gram- negative bacteria (183). Only one study has assessed the expression of NOD-like receptors in preterm infant monocytes, and report equivalent expression of both NOD1 and NOD2 by preterm (27–30 weeks) and term infants at both the mRNA and protein level (191).

The ligation of PRRs with microbial agonists initiates specific signalling cascades that orchestrate an effective innate immune response largely through the transcription of inflammatory cytokine and chemokine genes (183). These signalling pathways require the sequential phosphorylation of several downstream molecules for the activation and nuclear translocation of transcription factors (183). Three studies have collectively reported on the phosphorylation of eight PRR signalling molecules by preterm monocytes in response to a wide range of stimuli (specific TLR agonists, cytokines and growth factors, and whole Gram-negative and Gram-positive bacteria) (190, 192, 193). Overall preterm infants display equivalent or increased phosphorylation of downstream PRR signalling molecules (Section 4, Table 1.1).

19 1.4.2.2 Inflammatory responses Several studies have investigated the ability of preterm infants monocytes to produce inflammatory cytokines and chemokines, mainly in response to TLR agonists and a few utilising bacterial stimuli (Section 7, Table 1.1). While the results are somewhat conflicting, overall, preterm monocytes appear to have an impaired ability to produce pro- inflammatory cytokines (TNF, IL-6, IL-8), which may contribute to persistent bacterial infection. Preterm monocyte production of IL-10 and IL-12p35/p40 appears to be equivalent to term infants, however these cytokines have each only been assessed in response to LPS by one study, and therefore additional studies are required to confirm these results (194, 195). Importantly, almost all of the studies reviewed here have measured the production of these cytokines at the protein level. Given that preterm monocytes display normal phosphorylation of important TLR signalling molecules (including NF-B), an impaired inflammatory protein response may be due to a deficiency at the level of transcription. Furthermore, most studies have measured monocyte cytokine responses to purified bacterial products (mainly LPS), which induce very specific signalling cascades and may be poor representatives of in vivo infection.

1.4.2.3 Phagocytosis and killing mechanisms The phagocytosis and subsequent degradation of bacterial pathogens is a key mechanism by which monocytes control infection (196, 197). Collectively, investigations of the phagocytic capacity of preterm infant monocytes have reported equivalent phagocytosis of S. epidermidis, E. coli (± opsonisation), GBS and Staphylococcus aureus to term infant monocytes (Section 8, Table 1.1). In addition, monocyte expression of CD64 (a high affinity IgG Fc receptor involved in antibody-mediated phagocytosis) was similar between preterm and term infants, and was even increased at the earliest gestation assessed (24–31 weeks) (198). However as preterm infants exhibit efficient phagocytosis in the absence of endogenous immunoglobulin, the role of CD64 in this process may be negligible. One limitation of fluorescence based phagocytosis assays is the requirement to use killed bacterial preparations for effective fluorescent labelling. Killed preparations of S. epidermidis induce distinct effects on innate immune responses by human peripheral blood mononuclear cells compared to live S. epidermidis (cytokine production, phosphorylation of TLR pathway molecules) dependent on the method of bacterial killing (199). Therefore, the development of phagocytic assays utilising live bacteria would provide valuable validation that preterm infant monocytes exhibit competent phagocytosis. 20

There is very limited data on the intracellular killing capacity of preterm monocytes (Section 2, Table 1.1). Two studies have measured the production of reactive oxygen species (superoxide) by preterm monocytes and both report similar levels to that of term infants following stimulation with either phorbol myristate acetate or E. coli, however they reported conflicting findings following monocyte stimulation with N- Formylmethionine-leucyl-phenylalanine (fMLP, inducer of superoxide) (200, 201). The study by Kaufman et al also report equivalent lysozyme release by preterm and term monocytes in response to fMLP (201). Collectively these studies demonstrate that preterm monocytes are capable of producing bactericidal products, however additional studies are required to draw an appropriate conclusion on whether preterm monocytes are equally able to kill E. coli or S. epidermidis compared to term infants.

1.4.2.4 Antigen presentation and adhesion While macrophages and dendritic cells are considered to be the predominant antigen presenting cells, monocytes are equipped with machinery for antigen processing and presentation, and are capable of inducing T cell responses in the periphery (202). Furthermore, monocytes are considered to be a reservoir of myeloid precursors capable of differentiating into macrophages and dendritic cells during inflammation, and it is unknown whether an intrinsic monocyte defect in antigen presentation would be retained upon differentiation. Literature on the expression of molecules involved in antigen presentation has yielded inconsistent results (Section 9, Table 1.1), however the consensus suggests that preterm monocytes exhibit reduced surface expression of HLA- DR. There are fewer reports on the expression of T cell co-stimulation molecules (CD40, CD80, CD86) by preterm monocytes, and these are also inconsistent. Surface expression of adhesion molecules (CD11c, CD11b and CD18) required for monocyte migration from the periphery into inflamed tissues has generally been reported as significantly reduced on preterm monocytes (Section 5, Table 1.1). All of these studies have assessed the expression of these molecules at the protein level by flow cytometry, and differences in staining protocols or gating strategies may contribute the inconsistency of the findings.

1.4.2.5 Apoptosis and inflammasome formation The clearance of apoptotic cells is important for the resolution of infection-driven inflammation (203). Several reports indicate that neonatal monocytes exhibit significantly reduced apoptosis following E. coli or GBS infection compared to adults, which has been

21 associated with the up-regulated transcription of anti-apoptotic proteins (204-206). No studies have specifically investigated cell death in preterm infant monocytes, however this would be informative as decreased apoptosis may be a mechanism for persistent inflammation. One study has recently investigated formation of the NLRP3 inflammasome in preterm monocytes (Section 6, Table 1.1)(182). Activation of the NLRP3 inflammasome is triggered via multiple PRRs and has multiple functions including the regulation of caspase-1 (which is required for apoptotic and pyroptotic cell death), as well as secretion of active IL-1 (207, 208). Sharma et al reported a trend toward reduced NLRP3 expression with lower gestational age, to which they attribute the significantly reduced expression of caspase-1 observed prior to 29 weeks gestation (182). The primary motive for this study was to the investigate possible mechanisms for the impaired production of IL-1 by preterm monocytes, but these results also provide the first evidence that preterm monocytes may exhibit reduced apoptosis via decreased caspase-1 expression.

22 Table 1.1 A summary of studies assessing immune molecule expression by preterm infant monocytes GA of preterm Observation compared to term Parameter assessed Protein or mRNA assessed? *Sample type (stimulus) population (weeks) infants 1) Expression of pattern recognition receptors Peripheral WB 32–34 Similar (189) CD14 Surface protein (MFI) Whole CB ≤30 Similar (190) Whole CB 24–27 and 29–32 Decreased (188) MD-2 Surface protein (MFI) Whole CB 24–27 and 29–32 Decreased (188) mRNA Purified monocytes 27–30 Similar (191) NOD1 Surface protein (% positive, MFI) Purified monocytes 27–30 Increased % positive, similar MFI (191) mRNA Purified monocytes 27–30 Similar (191) NOD2 Surface protein (% positive, MFI) Purified monocytes 27–30 Similar (191) mRNA Purified monocytes 24–36 Increased (186) Surface protein (% positive) CBMC 24–36 Decreased (186) Surface protein (MFI, % positive) CBMC ≤30 Similar (193) TLR2 Surface protein (MFI) Whole CB ≤30 Similar (190) Surface protein (MFI) Whole CB 24–27 and 29–32 Decreased (188) Surface protein (% positive) Whole CB 30–34 and 34–<37 Similar (113) mRNA Purified monocytes 24–36 Increased (186) Surface protein (MFI) Peripheral WB (± LPS) 32–34 Similar (189) Surface protein (MFI, % positive) CBMC ≤30 Similar (193) Surface protein (% positive) CBMC 24–36 Decreased (186) TLR4 Whole CB ≤30 Decreased (190) Surface protein (MFI) Whole CB 24–27 and 29–32 Decreased 24–27 group only (188) Whole CB ≤30 Decreased (209) Surface protein (% positive) Whole CB 30–34 and 34–<37 Similar (113) 2) Intracellular killing Whole CB (fMLP) 21–32 Decreased (200) Intracellular (% positive) Whole CB (opsonised E. coli) 21–32 Similar (200) Production of ROS Similar, but decreased when monocytes Superoxide release Purified monocytes (PMA, fMLP) 26.5–32.5 were adhered for 18H (201) 3) Degranulation Similar, but decreased when monocytes Lysozyme release - Purified monocytes (fMLP) 26.5–32.5 were adhered for 18H (201)

23 GA of preterm Observation compared to term Parameter assessed Protein or mRNA assessed? *Sample type (stimulus) population (weeks) infants 4) Phosphorylation of downstream TLR effectors Whole CB (LTA) ≤30 Decreased (190) ERK1/2 - Whole CB (LPS) ≤30 Similar (190) Whole CB (PMA + A23187) ≤36 Similar (192) CBMC (S. epidermidis) ≤30 Similar (193) Whole CB (LPS) ≤30 Similar (190) p38 - Whole CB (TNF, LPS, MDP, live ≤36 Similar (192) S. epidermidis, S. aureus & E. coli) Whole CB (LTA) ≤30 Decreased (190) p65 - CBMC (S. epidermidis) ≤30 Similar (193) JNK - CBMC (S. epidermidis) ≤30 Similar (193) STAT1 and STAT3 - Whole CB (IFN, IL-6) ≤36 Similar (192) STAT5 - Whole CB (GM-CSF) ≤36 Similar (192) STAT6 - Whole CB (IL-4) ≤36 Similar (192) Whole CB (TNF, LPS, MDP, live NF-B - ≤36 Increased (192) S. epidermidis, S. aureus & E. coli) 5) Adhesion Purified monocytes (± LPS, ± Decreased in ≤30 group only for all ≤30 and 31–36 CD11c expression Surface protein (MFI) IFN) conditions (195) CBMC 26–32 Similar (182) CD11b expression Surface protein (MFI) Whole CB 26.5–32.5 Decreased (201) CD18 expression Surface protein (MFI) Whole CB 26.5–32.5 Decreased (201) 6) Inflammasome formation mRNA Purified monocytes (LPS) 24–28 Similar (182) NLRP3 expression Protein (cell lysate) Purified monocytes (LPS) <28 Similar (182) Pro-caspase-1 expression Protein (cell lysate) Purified monocytes (LPS) 24–30 Similar (182) Caspase-1 expression Protein (% positive) Purified monocytes (LPS+ATP) 24–27, 27–29 and 29–33 Decreased in ≤29 v 29–33 (182) ASC Protein (cell lysate) Purified monocytes (LPS) <28 Similar (182) P2X7R expression Surface protein (MFI) Purified monocytes (LPS) 24–30 Similar (182)

24

GA of preterm Observation compared to term Parameter assessed Protein or mRNA assessed? *Sample type (stimulus) population (weeks) infants 7) Cytokine responses G-CSF Secreted protein Purified monocytes (IL-1) 23–34 Decreased (210) IL-1 Secreted protein Purified monocytes (LPS) 29–34 Similar (211) Protein (cell lysate) Purified monocytes (± LPS) 24–30 Similar (182) Pro-IL-1 Intracellular protein (MFI) Purified monocytes 24–27, 27–29 and 29–33 Decreased in 24–27 group only (182) Purified monocytes (LPS+ATP) 26–30 Decreased (182) Purified monocytes (TriDAP, IL-1 Secreted protein 27–30 Similar (191) MDP) Purified adherent monocytes (LPS) 26.5–32.5 Similar (201) Purified monocytes (LPS, TriDAP, Decreased in response to LPS, similar in Secreted protein 27–30 MDP) response to TriDAP, MDP (191) Secreted protein Purified adherent monocytes (LPS) 26.5–32.5 Similar (201) Intracellular protein (% positive) Peripheral WB (LPS, R-848) ≤30 Decreased (212) IL-6 Secreted protein CBMC (live GBS) ≤30 Decreased (213) Intracellular protein (% positive) CBMC (heat-killed S. epidermidis) ≤30 Decreased (193)

Intracellular protein (% positive) CBMC (Pam3CSK4, R-FSL, LPS) ≤29 Decreased (164) Intracellular protein (% positive) CBMC (3M-013, 3M-003, 3M-002) ≤29 Similar (164) Intracellular protein (% positive) Whole CB (LPS) 26–32 and 32–<37 Similar (214) mRNA Purified monocytes (LPS) 23–32 Decreased (215) Secreted protein Purified monocytes (IL-1) 23–32 Decreased (215) IL-8 Secreted protein Purified monocytes (LPS) 24–36 Decreased (186) Intracellular protein (% positive) Whole CB (LPS) 26–32 and 32–<37 Similar (214) IL-10 Secreted protein Purified monocytes (LPS) 23–34 Similar (194) IL-12/IL-23 p40 Secreted protein Purified monocytes (LPS ± IFN) ≤30 Similar (195) IL-12 p35 mRNA Purified monocytes (LPS ± IFN) ≤30 Similar (195) Secreted protein Purified monocytes (LPS) 29–34 Decreased (211) Secreted protein Purified monocytes (LPS) 24–36 Decreased (186) Purified monocytes (TriDAP, Secreted protein 27–30 Similar (191) MDP) TNF Secreted protein Purified adherent monocytes (LPS) 26.5–32.5 Decreased (201) Intracellular protein (% positive) Peripheral WB (LPS, R-848) ≤30 Decreased (212) Secreted protein CBMC (live GBS) ≤30 Decreased (213) Intracellular protein (% positive) CBMC (heat-killed S. epidermidis) ≤30 Decreased (193)

25

GA of preterm Observation compared to term Parameter assessed Protein or mRNA assessed? *Sample type (stimulus) population (weeks) infants 8) Phagocytic capacity Decreased frequency of +ve monocytes, Peripheral WB ≤30 Heat-killed S. epidermidis - but similar pHrodo MFI (216) CBMC ≤30 Similar (193) Decreased frequency of +ve monocytes, Heat-killed E. coli - Peripheral WB ≤30 but increased pHrodo MFI (216) Non-opsonised E. coli - CBMC 24–31 and 32–36 Increased in 24–31 group only (198) Peripheral WB 32–37 Similar (217) Opsonised E. coli - CBMC 32–37 Similar (217) Whole CB 21–32 Similar (200) Heat-killed GBS - CBMC 26–32 Similar (213) Similar frequency of +ve monocytes, and Heat-killed S. aureus - Peripheral WB ≤30 increased pHrodo MFI (216) CD64 expression Surface protein (MFI) CBMC 24–31 and 32–36 Increased in 24–31 group only (198) 9) Antigen presentation Purified monocytes ≤30 and 31–36 Decreased in ≤30 group only (195) Peripheral WB ≤32 Decreased (218) Whole CB 24–27 and 29–32 Similar (188) HLA-DR expression Surface protein (MFI) CBMC 26–32 Similar (182) CBMC ≤30 Decreased (216) CBMC 24–31 and 32–36 Decreased in 24–31 group only (198) CD40 expression Surface protein (% positive) Purified monocytes ≤30 and 31–36 Similar (195) Decreased in ≤30 group only following Purified monocytes (± LPS) ≤30 and 31–36 CD80 expression Surface protein (% positive) LPS exposure (195) Whole CB 30–34 and 34–<37 Similar (113) Surface protein (MFI) CBMC ≤30 Similar (216) CD86 expression Purified monocytes ≤30 and 31–36 Decreased in≤30 group only (195) Surface protein (% positive) Whole CB 30–34 and 34–<37 Similar (113) *Studies utilising whole blood or CBMC were included where monocytes could be identified within mixed cultures by flow cytometry. ATP, adenosine triphosphate; CB, cord blood; CFU, colony forming units; fMLP, N-Formylmethionyl-leucyl-phenylalanine; LPS, lipopolysaccharide; LTA, lipoteichoic acid; MDP, muramyl dipeptide; MFI, median/mean fluorescence intensity; PMA, phorbol myristate acetate; R-848, resiquimod; ROS, reactive oxygen species; TriDAP, L-Ala-gamma-D-Glu-mDAP; WB, whole blood.

26

1.5 How does exposure to histologic chorioamnionitis impact neonatal monocyte function?

Despite the fact that histologic chorioamnionitis is present in 40–70% of preterm births, and is known to impact the risk of developing sepsis, there is very limited data on how exposure to histologic chorioamnionitis impacts immune development, and the monocyte in particular. Most human studies have focused on assessing the levels of inflammatory mediators in cord blood or cerebrospinal fluid, and collectively indicate an increase in IL- 6, IL-8, procalcitonin and CRP in preterm infants exposed to chorioamnionitis (219-223). Only two studies have assessed the effect of HCA exposure on preterm monocytes in cord and peripheral blood, and both demonstrated a significant reduction in the expression of MHC class II/HLA-DR, suggesting that HCA may result in a reduced capacity for antigen presentation (182, 219).

Monocyte function in the context of chorioamnionitis has been more extensively studied in animal models, particularly in sheep as they provide a more relevant model for human gestation than other animals; the developmental biology of ovine and human fetuses is similar, along with similar gestation length and number of offspring (224, 225). In these models, uterine inflammation is induced by LPS injection into the amniotic cavity and lambs are delivered preterm by cesarean section. Cord blood monocytes from preterm lambs exposed to intra-amniotic LPS produce significantly lower levels of hydrogen peroxide and IL-6 upon re-challenge with endotoxin in vitro, and exhibit significantly lower CD14 and MHC class II expression compared to preterm lambs without LPS exposure in utero (226, 227). The authors attribute these observations to endotoxin tolerance, an affect that wanes over 7–14 days in animals subject to a single LPS injection. Interestingly, repeated exposure to intra-amniotic LPS also had a cross-tolerance effect, attenuating the monocyte inflammatory response to TLR2, TLR5 and TLR9 agonists (228). The down-regulation of IRAK-4 and up-regulation of IRAK-M (positive and negative regulators of TLR4 signalling respectively) by blood monocytes was also observed, and was considered to be responsible for the tolerised monocyte phenotype (227, 228). It is important to note that following the initial period of monocyte hypo- responsiveness observed in these studies, blood monocytes exhibited a hyper-responsive or mature phenotype by 1–2 weeks, producing significantly increased levels if inflammatory cytokines and hydrogen peroxide (226-228). In perhaps a more clinically relevant sheep model, chronic exposure to Ureaplasma parvum (commonly isolated from

27 human placentae with chorioamnionitis) also resulted in attenuated blood monocyte production of IL-6 following in vitro stimulation with LPS in preterm lambs at birth (229). Despite the limitations of any animal model, these findings have important implications. Exposure to histologic chorioamnionitis is a strong risk factor for preterm infant development of early-onset sepsis, and in contrast is associated with a decreased risk of late-onset sepsis. The immunological mechanisms underlying this observation have not been elucidated, and there is a large knowledge gap regarding the impact of histologic chorioamnionitis on monocyte development in human neonates.

1.6 Transcriptional profiling of immune responses in sepsis

The review of the literature comparing preterm and term monocyte responses to various stimuli reveals inconsistent findings, but overall suggests a reduced capacity for the production of inflammatory cytokines to bacterial stimuli that is not due to a deficiency in phagocytic capacity, or in the phosphorylation of downstream TLR pathway molecules (Table 1.1). Furthermore, the deficiency for cytokine production is primarily seen at the protein level, suggesting there is an underlying transcriptional deficiency in preterm monocytes. Impressive developments in genomic technologies over the past decade have enabled researchers to interrogate transcriptional responses across the entire genome, rather than being restricted to assessing a handful of transcripts through polymerase chain reaction (PCR). The following sections provide a review of key transcriptomics studies in the field of innate immunity and sepsis, including recent advances in infant sepsis, highlighting the potential for this technology to advance diagnostic and therapeutic approaches.

1.6.1 DNA microarrays and RNA sequencing The two main technologies available for measurement of the entire transcriptome are DNA microarrays and RNA sequencing (RNA-seq). The development and application of DNA microarrays began to expand in the mid 1990s, and since then hundreds of publications utilising microarrays have emerged across a diverse range of biological systems and diseases in a testament to the success of this technology. However DNA microarrays are not without limitations, and the development of RNA-seq within the last ten years has provided researchers with an attractive alternative for analysis of the transcriptome. The two basic aims of transcriptomics studies are firstly to identify which

28 parts of the genome are being transcribed in a sample, and secondly to quantify the expression of these transcripts. DNA microarrays achieve the first aim by providing a dense array of oligonucleotide probes that hybridise to fluorescently labelled complimentary DNA (cDNA) sequences within a given sample (230). The relative abundance of each transcript is then determined based on the strength of fluorescent signal given by each oligonucleotide probe. However DNA microarrays are not fully quantitative, and the validation of gene expression results by quantitative real-time PCR is typically required (231). High background noise due to non-specific or cross hybridisation of transcripts limits the accuracy of expression measurements, particularly for transcripts present at very low abundance (232). The dynamic range for detection is also limited for highly expressed transcripts due to probe saturation (232). In addition, transcriptome analysis using microarray is limited by its reliance on pre-existing knowledge about genome sequence.

Sequencing based approaches overcome some of these limitations, with RNA-seq being the first sequencing based method that allows interrogation of the entire transcriptome in a high-throughput manner. RNA-seq identifies transcripts within a sample by direct sequencing of cDNA fragments (reverse transcribed from mRNA) followed by alignment of these sequences to a reference genome or by de novo assembly for species without a reference genome (232). This approach allows for sensitive measures of gene expression as RNA-seq does not have an upper limit for quantitation (limit correlates with the number of sequences obtained) and as transcripts can be unambiguously mapped to unique regions of the human genome there is little to no background signal (232). The resulting dynamic range of RNA-seq for gene expression is far greater than that of DNA microarrays, and the accuracy of RNA-seq for quantitation has been confirmed through comparisons with quantitative PCR and the use of spike-in RNA controls (233, 234). Importantly, RNA-seq is not limited to the analysis of transcripts according to pre-defined probe sequences, and therefore has the potential to discover novel transcripts and gene isoforms (235). Furthermore, a valuable advantage of RNA-seq is the much smaller requirement for input RNA compared to microarray, making RNA-seq ideal for studies utilising precious samples. A summary of the advantages of RNA-seq compared with DNA microarray is presented in Table 1.2.

RNA-seq however is not a mature technology, and currently there are no consensus guidelines for RNA-seq experimental design or data analysis, most likely due to the wide-

29 ranging applications for RNA-seq. The vast amount of sequencing data produced by RNA-seq also presents significant bioinformatic and data storage challenges. Nevertheless, RNA-seq is a superior technology to DNA microarray, the advantages of which have been demonstrated in the context of human immunity in a recent study of T cell activation (236).

Table 1.2 The advantages of RNA-seq compared with DNA microarray (232) Technology DNA Microarray RNA-seq Technology specifications High-throughput Principle Hybridisation sequencing Resolution Several–100 bp Single base Throughput High High Reliance on genomic sequence Yes In some cases Background noise High Low Application Dynamic range to quantify gene Up to a few- >8000-fold expression level hundredfold Ability to distinguish different isoforms Limited Yes Ability to distinguish allelic expression Limited Yes Practical issues Required amount of RNA High Low Cost for mapping transcriptomes of large High Relatively low genomes

1.6.2 Relevant transcriptomics studies in the field of sepsis and innate immunity The ability to simultaneously detect the global set of genes being transcribed in a cell through transcriptomics has significantly improved our understanding of many diseases and aspects of cellular immune function. In the context of monocyte biology, analysis of the transcriptome has helped to confirm and characterise the functional divergence between monocyte subsets, which was initially only attributed to differences in the surface expression of CD14 and CD16. A recent review and meta-analysis of five independent microarray studies comparing transcriptional profiles between CD16+ and CD16- monocytes revealed a core set of 182 genes that were differentially expressed between the two subsets (237). Functional characterisation of these genes showed that in 30 steady state, CD16- monocytes are geared toward anti-bacterial and inflammatory responses, and CD16+ monocytes are involved in patrolling and infiltration (237). Furthermore, distinct transcriptional programs have been identified between fetal and adult bone marrow monocytes using microarray, highlighting gene expression differences at baseline and post-stimulation with recombinant cytokines (238).

Importantly, transcriptomics has allowed for detailed characterisation of the monocyte response to various pathogenic stimuli including cytomegalovirus and the human immunodeficiency virus, as well as monitoring the effects of hypoxia and even exposure to cigarette smoke on monocyte function (239-242). Responses to bacterial stimuli have mainly been characterised in monocyte-derived macrophages. Nau and colleagues presented a comprehensive analysis of the human monocyte-derived macrophage transcriptional response to eight different pathogens including Gram-negative, Gram- positive and mycobacterial species (243). Interestingly, a conserved macrophage activation program was conserved across all stimuli, despite the diversity of bacterial species included. This study also provided insights into mechanisms of immune evasion strategies used by Mycobacterium tuberculosis.

Finally, transcriptomics methods are also being applied to the study of sepsis. A meta- analysis of ten independent microarray studies collectively containing close to 600 adult patients revealed an endotoxin tolerance signature was significantly associated with early sepsis (244). The authors validated this observation in a separate cohort and demonstrated that detection of the endotoxin tolerance signature could become a useful tool in the diagnosis of adult sepsis. Transcriptional profiling of peripheral whole blood from neonates with sepsis using microarray identified 52 conserved genes involving innate, adaptive and metabolic pathways that could accurately predict bacterial infection (245). This study also highlighted that neonatal sepsis was broadly associated with amplification of innate immune pathways and suppression of adaptive immunity, providing insights into the pathogenesis of neonatal sepsis. Furthermore, transcriptome-wide analysis of peripheral whole blood from very low birth weight infants with sepsis using microarray was able to distinguish those with Gram-positive and Gram-negative infection, and again identified innate immune pathways as significantly over-represented in the septic response (246). A novel study of the transcriptome of the fetal inflammatory response syndrome also identified immune and metabolic similarities to that observed in adults, and identified a set of 25 genes that were conserved across FIRS, paediatric sepsis and

31 adult models of endotoxin challenge (247). This suggests that clinically distinct causes of systemic infection and inflammation induce common immune and metabolic changes.

In summary, the development of transcriptomics has facilitated novel insights into monocyte biology and the role of innate immunity in infection and inflammation. RNA- seq is emerging as the tool of choice for transcriptome profiling due to its advantages over microarray technology. Evidence suggests that preterm infant monocytes harbour an underlying transcriptional deficiency, and at this time no studies have compared the transcriptomes of circulating monocytes between preterm and term infants using RNA- seq-based methodologies.

1.7 Summary

Preterm birth is a significant global health concern, and the incidence of preterm birth continues to rise in developed countries. Invasive infection remains an important cause or preterm morbidity and mortality, with E. coli and S. epidermidis the most common causes of early- and late-onset sepsis in preterm infants respectively. Our understanding of the immunological mechanisms underlying preterm infant susceptibility to sepsis is incomplete, however it is increasingly recognised that innate immune defences are critical for protection against infection in early life. As sentinels of peripheral blood, monocytes represent a key effector cell of innate immunity, and provide defence against invading bacterial pathogens through multiple specialised functions; pathogen recognition, phagocytosis, bacterial killing and initiation of inflammation. Evidence suggests that preterm infant monocytes are capable of bacterial phagocytosis, are not strikingly deficient in their expression of TLR2 or TLR4 and exhibit equivalent phosphorylation of downstream TLR pathway molecules to term infants. Yet preterm infant monocytes display significantly diminished expression of pro-inflammatory cytokines at the protein level, suggesting an underlying transcriptional deficiency. In addition, a large knowledge gap exists concerning the impact of prenatal exposure to histologic chorioamnionitis (present in 40–70% of preterm births) on monocyte development. The development of sophisticated next generation sequencing technologies such as RNA-seq has allowed us to study the global transcriptional responses of preterm monocytes to the most relevant neonatal pathogens in a universal and un-biased manner. Furthermore, the development of publicly available databases containing thousands of experimentally observed

32 molecular interactions allows us to identify the most important bacterial response genes through network analysis.

1.8 Aims and hypotheses of this thesis

The specific aims of thesis were to: 1) Compare the global transcriptional responses of preterm and term monocytes to live E. coli and S. epidermidis challenge, and identify transcriptional deficiencies specific to the preterm infant. 2) Determine the impact of prenatal exposure to histologic chorioamnionitis on the preterm monocyte transcriptional response to E. coli and S. epidermidis. 3) Identify and characterise the pathogen-specific, and core neonatal monocyte transcriptional response to E. coli and S. epidermidis.

We hypothesised that: 1) Monocytes from preterm infants would exhibit diminished transcription of TLR pathway, inflammatory cytokine and chemokine and antigen presentation genes, and exhibit an anti-apoptotic pattern of gene expression following stimulation with E. coli and S. epidermidis. 2) Prenatal exposure to histologic chorioamnionitis would alter the preterm monocyte transcriptional phenotype to more closely resemble that of term infants. 3) Neonatal monocytes would share an overlapping transcriptional response to both E. coli and S. epidermidis alongside more selective pathogen-specific responses.

33

34

Chapter 2 Materials and methods

2.1 Study participants

The collection of infant cord blood samples for gene expression analysis was approved by the King Edward Memorial Hospital Ethics Committee (Perth, Australia; Study ID - 814/EW). Written informed consent was obtained from parents or guardians prior to sample collection from study participants. Sixteen neonates born at King Edward Memorial Hospital between May 2004 and June 2005 were retrospectively chosen for inclusion in this study from a larger cohort of neonates enrolled in the Study of Postnatal Immunity in the Neonate. Neonates were selected for this study based on gestational age (GA) at birth, and the presence/absence of histologic chorioamnionitis (HCA), and were classified into three groups: (i) term infants 37–39 weeks GA (ii) very preterm infants 29–31 weeks GA with and (iii) without HCA. The presence of HCA was determined by semi-quantitative histologic scoring of the umbilical cord, chorionic plate, placenta and extra-placental membranes. Histologic scoring was performed according to standard guidelines (248) by an experienced perinatal histopathologist. Demographic information for each infant group is presented in (Table 4.1).

2.2 Blood sample collection and processing

Cord blood was collected into pre-heparinised syringes from cord blood vessels. Where necessary, blood was also collected from vessels immediately adjacent to the base of the cord (on the fetal placental surface) that feed into cord vessels. The collection sites were cleaned with alcohol swabs prior to sampling to prevent maternal blood contamination. Whole blood samples were mixed 1:1 with RPMI 1640 (Gibco®; Life Technologies, Victoria, Australia), and cord blood mononuclear cells (CBMC) isolated using lymphoprep gradient centrifugation, according to the manufacturer instructions (Axis- Shield PoC, Oslo, Norway). CBMC were cryopreserved using an established method that preserves cell viability and function (249).

35

2.3 Microbiology

2.3.1 Strain selection Escherichia coli (E. coli) O1:K1:H7 strain 11775 from the American Type Culture Collection (ATCC; Virginia, USA), and Staphylococcus epidermidis (S. epidermidis) strain wild type (WT) 1457 were chosen as representative strains for this study. E. coli 11775 is an extra-intestinal pathogenic strain isolated from the urine of a patient with cystitis. S. epidermidis WT 1457 was isolated from a patient with an infected central venous catheter, and was kindly provided by Dr Michael Otto (National Institute of Allergy and Infectious Diseases, MT) (250).

2.3.2 Preparation of bacterial broths Luria Bertani and Heart Infusion broths (Oxoid; Thermo Fisher Scientific, Adelaide, Australia) were reconstituted in sterile water for irrigation (Baxter; NSW, Australia) according to the manufacturer's instructions. The broths were sterilised by autoclaving for 15 minutes at 121°C. Sterile broths were stored at 4°C and used within 1 month. The compositions of each broth are described in Table 2.1.

Table 2.1 Broth compositions (g/L) Luria Bertani Heart Infusion Pancreatic digest of casein: 10 Beef heart infusion solids: 17.5 Yeast extract: 5 Proteose peptone: 10 Sodium chloride: 10 Glucose: 2 Final pH: 7 ± 0.2 at 25°C Sodium chloride: 5 Di-sodium phosphate: 2.5 Final pH: 7.4 ± 0.2 at 25°C

2.3.3 Growth and storage of live mid-log bacterial stocks The protocol for the cultivation and storage of mid-log bacterial preparations was based on an established method developed in-house (251). E. coli or S. epidermidis were streaked out for single colonies on blood agar plates (PathWest Media, Western Australia) from high-density frozen glycerol stocks, and the plates incubated overnight (37°C, 5%

CO2). The following day, two single colonies of E. coli or S. epidermidis were used to inoculate 15 mL (in a 50 mL tube) of Luria Bertani or Heart Infusion broth, respectively. 36

These broths were incubated with shaking overnight (37°C, 120 rpm). The following day, fresh broths were prepared by adjusting the optical density at 600 nanometers (OD600) of the overnight inoculations to ~0.05 in a final volume of 50 mL of their respective broths in a sterile 250 mL conical flask. The flasks were sealed with a sterile lid that allowed gas exchange. Cultures were prepared from each inoculation by incubating these broths with shaking (37°C, 80 rpm) until mid-log phase was reached; OD600 of 0.4 for E. coli, and 0.6 for S. epidermidis (see Figure 3.5). The mid-log cultures were gently centrifuged (2 minutes, 60g) to remove any clumped bacterial cells, and the supernatant containing suspended bacteria was transferred into a new sterile tube and supplemented with either sterile glycerol (Sigma-Aldrich, NSW, Australia) for E. coli or filter-sterile heat- inactivated fetal calf serum (HI-FCS; Sigma-Aldrich) for S. epidermidis at final concentration of 20% for both. The mid-log stocks were stored at -80°C in 1 mL aliquots in cryovials.

2.3.4 Determining bacterial viability of frozen stocks The concentrations of viable bacteria (colony-forming units (CFU)/mL) in bacterial stocks were calculated by performing six, ten-fold serial dilutions of each stock in phosphate buffered saline (PBS; Gibco®) after the stocks had been thawed and washed once in PBS to remove remaining broth. Three, 20 L samples of each dilution were spotted onto a single sterile blood agar plate and left to dry in a laminar flow cabinet.

Following overnight incubation (37°C, 5% CO2) agar plates with triplicate spots containing ~20-80 single colonies were counted to determine the mean CFU/mL of the original stock.

2.4 Cell culture methodology

2.4.1 Thawing of cryopreserved mononuclear cells Vials of cryopreserved CBMC were transferred to the laboratory in liquid nitrogen, then thawed rapidly in a 37°C water bath and transferred into a sterile 15 mL polypropylene tube containing 10 mL warm culture medium, consisting of RPMI 1640 (Gibco®), 2 mM glutamax (Gibco®), 10 mM HEPES buffer (Gibco®), 1 mM sodium pyruvate (Gibco®) and 0.05 mM 2-Mercaptoethanol (Gibco®). The CBMC were then pelleted by centrifugation (300g, 6 minutes), resuspended in 4 mL of culture medium (pre-warmed to 37°C) and passed through a 35 m nylon filter (Corning, Massachusetts, USA) to 37 remove any cell clumps. A sample of the filtered CBMC was stained with 0.4% Trypan Blue (Sigma-Aldrich), and visualised using a Neubauer haemocytometer (BOECO, Hamburg, Germany) to determine total cell number and viability. The CBMC were then pelleted (300g, 6 minutes) and washed with 2 mL sterile flow cytometry buffer containing PBS (Gibco®), 2% w/v bovine serum albumin (BSA; Sigma-Aldrich), and 2% v/v HI- FCS to prepare for staining.

2.4.2 CBMC surface receptor staining Once thawed, washed and filtered, the CBMC were then resuspended in flow cytometry buffer at a concentration of ~5x107/mL, and stained with monoclonal antibodies directed against monocyte (CD14, CD16) and non-monocyte (CD3, CD19, CD56) lineage markers. Antibody titrations were performed to determine the optimal staining concentrations (that which gives the highest stain index) for each antibody. Details of the monoclonal antibodies used for monocyte purification and phenotyping are presented in Table 2.2. Following incubation for 30 minutes at 4°C, the CBMC-antibody suspension was washed twice with 2 mL flow cytometry buffer (pelleted at 300g for 3 minutes in between each wash) and then resuspended in flow cytometry buffer at a concentration of ~1x107 cells/mL for cell sorting. A 10 L (~1x105 cells) aliquot of each stained CBMC sample was transferred into a BD Trucount™ tube (BD Biosciences) for quantitation of total cells/mL following sample acquisition on a BD FACSCanto™ II (BD Biosciences).

38

Table 2.2 Details of monoclonal antibodies used for monocyte purification and phenotyping Optimal Marker Fluorochrome Isotype Clone Supplier concentration Mouse BD CD14 PE-Cy7 4.5 g/mL M5E2 IgG2a Biosciences* Mouse BD CD16 FITC 1.4 g/mL 3G8 IgG1  Biosciences* Mouse BD CD3 APC 2.2 g/mL SK7 IgG1  Biosciences* Mouse CD19 AF647 4.5 g/mL HIB19 BioLegend* IgG1  Mouse N901 Beckman CD56 PE 0.17 g/mL IgG1 (NKH-1) Coulter† Mouse BD HLA-DR BV421 2.5 g/mL G46-6 IgG2a  Biosciences* HLA-DR Mouse BD BV421 2.5 g/mL G155-178 isotype IgG2a  Biosciences* Supplier locations: *California, USA; †NSW, Australia

2.4.3 Purification of monocytes by cell sorting All monocyte populations (classical: CD14++CD16-, intermediate: CD14++CD16+, and non-classical: CD14+CD16+) were purified from the stained CBMC sample using a BD FACSAria™ III (BD Biosciences, California, USA). The machine was set to “purity” mode and an 85 m nozzle was used. The gating strategy used (Figure 2.1) was optimised to deliver a pure sample of sorted monocytes, whereby debris and doublet cells were eliminated, followed by positive selection of CD14+ and CD16+ events, and subsequent exclusion of B cells, T cells and NK cells. The purified monocytes were collected into culture medium containing 30% filter-sterilised HI-FCS. For each sample, ~2.5x104 sorted monocytes were analysed using a BD FACSCanto™ II (BD Biosciences) flow cytometer to determine purity, with an additional stain for HLA-DR. A sample of sheath fluid was collected prior to each cell sort and spread onto two blood agar plates to check for sterility. An additional experiment confirmed that exposure to sheath fluid did not induce cellular activation; there was no change in TNF or IL-6 gene or protein expression by CBMC following 24 hour culture with sheath fluid (data not shown).

39

A B

C D

E

Figure 2.1 Sequential cell sort gating strategy for monocyte purification. (A) Elimination of non-cellular debris was based on forward- and side-scatter properties. (B) Elimination of doublet events to ensure only single cell were sorted. (C) Selection of CD14+ and CD16+ events. (D) Exclusion gate for NK cells (CD56+), B cells (CD19+) and T cells (CD3+). This gate was set using the blue contour plot, which represents cells positive for the relevant lineage markers. (E) A representation of the final sorted population using this gating strategy.

40

2.4.4 Monocyte stimulation with live E. coli or S. epidermidis

Purified monocytes were centrifuged (300g, 6 minutes) and resuspended in 500 L culture medium (pre-warmed to 37°C). A sample of the monocytes was stained with 0.4% Trypan Blue and counted using a Neubauer haemocytometer to determine total cell number and viability. The cells were resuspended at 5x105 cells/mL in culture medium and 100 L aliquots (5x104 cells) were dispensed into separate wells of a sterile 96-well polypropylene plate. An additional 15 L culture medium was added to each well, and the plate was incubated (37°C, 5% CO2) for 15 hours to allow the cells to return to basal, pre-sort gene expression levels. This is discussed in greater detail in section 3.5.1.

Frozen mid-log stocks of E. coli and S. epidermidis were thawed, centrifuged (13,000 rpm, 3 minutes), washed twice with sterile PBS to remove broth, and resuspended in PBS at a concentration of 5x107 CFU/mL. Following 15 hours of rest, the monocytes were stimulated in sterile 96-well polypropylene plates with either 10 L of E. coli or S. epidermidis preparations, (for a final MOI of 10:1 bacterial cells to monocytes), or with 10 L of sterile PBS (negative control) for a final volume of 125 L (5x104 monocytes/well, cultures performed in triplicate). Dilutions of each bacterial stock were spotted onto blood agar plates and incubated overnight to confirm estimated CFU/mL. 5 After 2 hours incubation (37°C, 5% CO2), the triplicate cultures were pooled (1.5x10 monocytes), supernatants harvested and stored -80°C, and cells resuspended in 300 L RNAprotect Cell Reagent (Qiagen, Victoria, Australia) and stored at -20°C before batch analysis. Each donor was processed on a separate day.

2.4.5 Quantitative detection of cytokines/chemokines in culture supernatants An in-house multiplex fluorescent-bead immunoassay was used to quantify levels of eleven cytokines, chemokines and growth factors in monocyte culture supernatants (213)(Table 2.3). Primary antibodies directed against each cytokine/chemokine were washed to remove sodium azide using Vivaspin 500 centrifugal concentrators (Sigma- Aldrich), and covalently conjugated to carboxylated microspheres (Bio-Rad Laboratories Inc., California, USA). Samples and protein standards were diluted in PBS containing 0.05% Tween 20 (Sigma-Aldrich) and 2% FCS, and incubated with antibody-conjugated microspheres in a 96-well Multiscreen Filter Plate (Merk Millipore, Victoria, Australia) for 30 minutes with shaking (room temp, 500 rpm, protected from light). Biotinylated 41 secondary antibodies were then added and the plate incubated for a further 30 minutes. The wells were washed (PBS, 1% BSA, 0.25% Tween 20, 0.001% sodium azide; Sigma- Aldrich) and 5 g/mL streptavidin-PE conjugate (BD Biosciences) was added to develop for 15 minutes. Excess streptavidin-PE was washed off, and the samples acquired on a BioPlex® 200 System (Bio-Rad) to determine fluorescence in each specific bead region. Analyte concentrations were determined from a 5-PL standard curve of median fluorescence intensity, generated from recombinant protein standards using BioPlex Manager 5.0 software. Sample concentrations below the limit of detection were assigned a value equal to half that of the lowest standard.

Table 2.3 Antibody and recombinant standard information Biotinylated Recombinant Lower limit Analyte Primary Ab Secondary Ab Standard of detection IL-1 Clone 2805 Polyclonal - 1.22 pg/mL Supplier R&D Systems* R&D Systems R&D Systems IL-6 Clone MQ2-13A5 MQ2-39C3 - 1.22 pg/mL Supplier BD Biosciences BD Biosciences BD Biosciences IL-8 Clone G265-5 G265-8 - 2.44 pg/mL Supplier BD Biosciences BD Biosciences R&D Systems IL-10 Clone JES3-9D7 JES3-12G8 - 1.22 pg/mL Supplier BD Biosciences BD Biosciences BD Biosciences IL-12p70 Clone 24945 Polyclonal - 1.22 pg/mL Supplier R&D Systems R&D Systems R&D Systems CXCL10 Clone 4D5/A7/C5 6D4/D6/G2 - 0.61 pg/mL Supplier BD Biosciences BD Biosciences BD Biosciences MCP1 Clone 23007 Polyclonal - 1.22 pg/mL (CCL2) Supplier R&D Systems R&D Systems R&D Systems MIP-1 Clone Polyclonal Polyclonal - 0.61 pg/mL (CCL3) Supplier R&D Systems R&D Systems R&D Systems M-CSF Clone 1113 Polyclonal - 1.22 pg/mL Supplier R&D Systems R&D Systems R&D Systems G-CSF Clone 3316 Polyclonal - 1.22 pg/mL Supplier R&D Systems R&D Systems R&D Systems TNF Clone MAb1 MAb11 - 1.22 pg/mL Supplier BD Biosciences BD Biosciences R&D Systems Supplier location: * Minnesota, USA

42

2.5 Cell death assays

2.5.1 LDH assay Lactate dehydrogenase (LDH) was measured in monocyte culture supernatants as an indicator of lytic cell death using the CytoTox 96 Non-Radioactive Cytotoxicity Assay (Promega, Madison, WI, USA), according to manufacturer’s instructions. LDH present in culture supernatants acted as an enzyme to convert a tetrazolium salt into a red formazan product during a 30-minute reaction. Values of absorbance at 492 nanometers were used to compare samples where monocytes exposed to 3% Triton X-100 (Sigma- Aldrich) in culture represented 100% cell lysis, and culture media without monocytes represented 0% cell lysis. Each donor had a matched 3% Triton X-100 positive control.

The formula used to calculate % cell lysis in each sample was: (A492 sample – A492 culture medium) / (A492 Triton X-100 – A492 culture medium) X 100. Flow cytometric analysis on the Triton X-100 treated cultures confirmed the absence of cells; forward and side scatter parameters did not visualise any discernable cell populations where monocytes would typically be present.

2.5.2 Apoptosis assessment by flow cytometry The presence of apoptosis in unstimulated and stimulated monocyte cultures was detected using the PE Annexin V Apoptosis Detection Kit I (BD Biosciences), according to manufacturer’s instructions. Samples were acquired on a BD FACSCanto™ II (BD Biosciences) flow cytometer within one hour of staining, and analysed using FlowJo software (vX, FlowJo LLC, Oregon, USA). Viable cells were defined as those with negative staining for both Annexin V (binds membrane phosphatidylserine) and 7-amino- actinomycin D (7-AAD; intercalates DNA), cells in the early stage of apoptosis were positive for Annexin V staining alone, and cells in late apoptosis or already dead were positive for both Annexin V and 7-AAD staining. CBMC that were heat shocked (56°C for 1 minute, then 37°C for 1 hour), or UV exposed (placed on a transilluminator for 10 minutes, then incubated at 37°C for 30 minutes) were used as positive controls for non- apoptotic and apoptotic death and to set fluorescence compensation.

43

2.6 Molecular studies

2.6.1 Purification of total RNA from cultured monocytes Batch RNA extractions (12 samples/batch) were performed on all cultured monocyte samples (stored in RNAprotect Cell Reagent) using the AllPrep DNA/RNA Micro kit (Qiagen), within 3 months of initial storage. Extractions were performed according to the manufacturer’s instructions. An RNase inhibitor (SUPERase In™, Ambion®, Life Technologies) was added to each RNA sample at a final concentration of 1 U/L to prevent RNA degradation. RNA quantity was assessed using a Qubit® 2.0 Fluorometer (Life Technologies), and purity assessed using a NanoDrop 2000 (Thermo Scientific, Delaware, USA). At least 100 ng total RNA was purified from each culture of 1.5x105 monocytes. RNA samples were stored at -20°C for no longer than two months prior to analysis by PCR, or two weeks prior to preparation for RNA-seq.

2.6.2 cDNA synthesis Total RNA was reverse transcribed into cDNA using the QuantiTect Reverse Transciption Kit (Qiagen), according to the manufacturer’s instructions. Briefly, contaminating DNA was eliminated from RNA samples by incubation with gDNA wipeout buffer. RNA samples were then primed with a mix of Oligo-dT and random primers, and reverse transcribed using Omniscript and Semsiscript reverse transcriptases. cDNA samples were stored at -20°C.

2.6.3 Primer design Forward and reverse primers for human 18S were purchased from Qiagen (QuantiTect Primer Assays) and used as a reference gene for all real-time PCR (RT-PCR) experiments. Forward and reverse primer sequences for IL-6 and TNF were designed in-house using Primer Express® software (v3, Life Technologies). The primers were designed to span intron junctions to ensure detection of gene transcripts only (avoiding detection of genomic DNA) and were synthesised by Sigma-Aldrich. Molecular characteristics of each primer are presented in Table 2.4. Nucleotide BLAST searches confirmed that the primer sequences were specific to the transcripts of interest (http://blast.ncbi.nlm.nih.gov/Blast.cgi).

44

2.6.4 Conventional PCR for primer optimisation

IL-6- and TNF-specific PCR products were amplified from cDNA (from LPS- stimulated PBMC) using a reaction mixture containing 1X iTaq buffer, 1.5 mM MgCl2, 1.25 U iTaq DNA Polymerase (all from Bio-Rad), 0.02 mM dideoxynucleotide triphosphates (dNTPs, Fisher Biotech, WA, Australia), 400 nM of each primer and 1 L of cDNA in a final volume of 25 L. PCRs for both IL-6 and TNFwere conducted on a LightCycler® (Bio-Rad) using a temperature gradient to optimise primer annealing temperatures. No-template negative controls were included in each run to confirm no contamination of reagents. Reactions were held at 94°C for 5 minutes followed by forty cycles of 94°C for 30s, 50°C–68°C for 30s (twelve point gradient), 72°C for 30s and finally 72°C for 10 minutes. PCR products (10 L) were run on a 2% agarose gel stained with SYBR® Safe alongside a TrackIt™ DNA ladder (both from Invitrogen™, Life Technologies) for 40 minutes at 90V. Bands were visualised under UV light using a Gel Doc™ (Bio-Rad) and confirmed to match the expected PCR product length. The optimal annealing temperature for both IL-6 and TNF primers was 58°C.

2.6.5 Sequencing of PCR products PCR products were cleaned using the PureLink® PCR Purification Kit (Life Technologies) according to the manufacturer’s instructions, and quantified using a NanoDrop 2000. Eight nanograms of purified IL-6 and TNF PCR product were analysed by Sanger sequencing (performed by the Australian Genome Research Facility; AGRF) (Table 2.5). Nucleotide BLAST searches were performed on product sequences to confirm amplification of the transcripts of interest (http://blast.ncbi.nlm.nih.gov/Blast.cgi).

45

Table 2.4 Characteristics of in-house designed primers Assay 5’-3’ bp Position within mRNA Melting temp. (°C)* % GC Amplicon length (bp) FWD CACCGGGAACGAAAGAGAAG 20 Exon 1 65.8 55 IL-6 97 REV CCCAGGGAGAAGGCAACTG 19 Exon 2 67.1 63.2 FWD TCTTCTCGAACCCCGAGT 18 Exon 3 62.5 55.6 TNF 107 REV ATTGGCCAGGAGGGCATT 18 Exon 4 66.7 55.6 *Determined using the Sigma-Aldrich design tool: (https://www.sigmaaldrich.com/configurator/ servlet/DesignTool). FWD, forward; REV, reverse; bp, base pair

Table 2.5 Sanger sequencing of PCR amplicons showing alignment of primers Assay Amplicon sequencing results 5’-3’ FWD primer CACCGGGAACGAAAGAGAAG Seq 1 CACCGGGAACGAAAGAGAAGCTCTATCTCCCCTCCAGGAGCCCAGCTATGAA IL-6 Seq 2 ACTCCTTCTCCACAAGCGCCTTCGGTCCAGTTGCCTTCTCCCTGGG REV primer GTCAACGGAAGAGGGACCC FWD primer TCTTCTCGAACCCCGAGT Seq 1 TCTTCTCGAACCCCGAGTGACAAGCCTGTAGCCCATGTTGTAGCAAACCCTCAAGC TNF Seq 2 CCTCAAGCTGAGGGGCAGCTCCAGTGGCTGAACCGCCGGGCCAATGCCCTCCTGGCCAAT REV primer TTACGGGAGGACCGGTTA FWD, forward; REV, reverse; Seq, sequencing result.

46

2.6.6 Real-time polymerase chain reaction

For optimisation of monocyte culture conditions, 18S, IL-6 and TNF gene expression was measured by RT-PCR using a Rotor-Gene® Q (Qiagen). cDNA templates (1 L) were amplified in 1X Rotor-Gene SYBR Green PCR Master Mix containing HotStarTaq® Plus DNA Polymerase, SYBR Green PCR buffer and dNTP mix (Qiagen) and 300 mM primers in a final volume of 10 L. All assays were held at 95°C for 5 minutes, followed by forty cycles of 95°C for 10s, 58°C for 30s and 72°C for 10s. All reactions were performed in duplicate alongside no-template negative controls. A melt curve analysis was performed at the end of each run to confirm the specificity of amplicons.

2.6.6.1 Calculating gene fold change Fold change in IL-6 and TNF gene expression between control and experimental conditions was calculated using the standard 2-ΔΔCt method with 18S as a reference gene (252). The average cycle threshold (Ct) values were determined for all gene replicates, and the difference in Ct values (ΔCt) between genes of interest (IL-6 or TNF) and the reference gene (18S) were calculated for each sample individually. The difference in ΔCt values between control and experimental conditions (ΔΔCt) was then used to calculate fold change in gene expression using 2-ΔΔCt.

2.6.6.2 Amplification efficiency testing An important assumption when using the Delta-Delta Ct method for calculating fold change in gene expression is that the amplification efficiency of each reaction is approximately the same. Qiagen guarantees ~100% reaction efficiency for their QuantiTect Primer Assays (the 18S we used as a reference gene). To validate the performance of IL-6 and TNF in-house designed primers, amplification efficiencies were determined using five 1 in 10 serial dilutions of cDNA (from TLR7/8-stimulated PBMC). Reaction mixes and amplification cycles were as described above. Amplification efficiencies were 96% for IL-6 and 100% for TNF

47

2.7 Next-generation sequencing of mRNA

2.7.1 Experimental design Five biological replicates were included from each infant group and for each culture condition. Each sample comprised of 3 technical replicates (3 culture wells in a 96-well plate). Fourteen of sixteen infants had paired samples across all culture conditions, whereas two infants had samples for one bacterial stimulation only due to lower monocyte yields (Table 2.6). We estimated that five biological replicates per group would allow for detection of >2.3 fold-changes in gene expression with 90% power (253). In order to achieve a sequencing depth of ≥10 million mapped reads per sample (required for differential gene expression analysis (254)), all forty-six samples were multiplexed into a large pool and sequenced over 5 lanes of an Illumina HiSeq flow cell (to minimise technical variation). Further details of the RNA-seq experimental design are discussed in Chapter 3.

Table 2.6 Biological replicates Unstimulated E. coli-stimulated S. epidermidis-stimulated Term infants n=5 n=5 n=5 Preterm infants (HCA-) n=6* n=5 n=5 Preterm infants (HCA+) n=5 n=5 n=5 Total samples n=46 *4 samples have paired E. coli- and S. epidermidis-stimulated cultures, and 2 samples only have one paired bacterial stimulation sample (E. coli or S. epidermidis).

2.7.2 RNA sequencing by AGRF All RNA sequencing was outsourced to the Australian Genome Research Facility (AGRF, Melbourne, Australia). RNA samples from cultured monocytes were quantified using an Agilent 2100 Bioanalyser (Agilent Technologies, California, USA) prior to processing. cDNA libraries were constructed from a standard input of at least 100 ng total RNA using the TruSeq Stranded mRNA Sample Prep Kit (Illumina, Inc., California, USA). Samples were normalised to the lowest total weight sample, where E. coli-stimulated samples were assigned an arbitrary RNA concentration of 6 ng/L prior to normalisation, due to complications discussed in section 3.6.1.1. The cDNA library construction process involves an enrichment step for polyadenylated mRNA prior to cDNA synthesis eliminating rRNA (>80% of total RNA). Each sample was indexed with a unique barcode

48 adaptor sequence, allowing all forty-six samples to be pooled and sequenced over five lanes. RNA sequencing (50 bp, single-end) was performed on a HiSeq 2000 (Illumina, Inc.).

2.7.3 Quality assessment of raw sequencing data All forty-six Illumina HiSeq 2000 50 bp single-end FASTQ raw read libraries supplied by AGRF were quality assessed using the FASTC Toolkit suite (v0.0.13), with the -Q33 flag supplied for input using Sanger Phred score format (quality score offset of 33). Forty- two read libraries demonstrated Phred quality scores >33 across all 50 read bases, translating to ≥99.95% base call accuracy. The remaining four libraries were all from individual preterm donors (HCA+ unstimulated, HCA+ S. epidermidis-stimulated, HCA- E. coli-stimulated, and HCA- unstimulated) and required trimming of the first 3, 2, 1 and 1 bases respectively from the start of each 50 bp sequence (these bases had <99.9% base call accuracy). Post-trimming, all remaining bases across the four samples demonstrated Phred scores >33.

2.7.4 Read alignment to the human genome and count summarisation After quality assessment and base trimming were performed, all forty-six read libraries were aligned to the GRCh37 (hg19) Human genome using the R/Subread suite (v1.14.2) with default options (255). Read counts were summarised by subjecting aligned reads to Subread feature counting against Ensembl (v69) gene and transcript annotations, with a minimum Mapping Quality Score of 30 (256). Prior to visualisation of gene expression, annotated raw read counts were gene-length and library-size normalised to Reads Per Kilo-base of exon Per Million reads (RPKM) and then log2 transformed.

2.7.5 Differential expression analysis Genes that were differentially expressed following stimulation with E. coli or S. epidermidis compared to PBS were analysed for each infant group separately (term, preterm HCA-, preterm HCA+). Similarly, differentially expressed genes between infant groups for each culture condition were also calculated separately. Raw aligned read counts were analysed using the R/limma Voom-normalised differential gene expression analysis method (R software v3.1.1, limma v3.20.9) (257, 258). Voom normalisation converts raw read counts to normalised log 2-transformed (with a global offset of 0.5 to 49 avoid taking the log of 0) counts per million (CPM) using lowess regression-based precision weights (257, 258). Prior to Voom normalisation, read counts were filtered on the requirement that retained genes had at least five CPM in at least four samples (the minimum group size in the experiment). Gene filtering was performed independently of the experimental variables to avoid the introduction of bias. Library sizes were scale- normalized using the Trimmed Mean of M-values (TMM) method from the edgeR package to correct for differences in sequencing depth between libraries during Voom normalization (259). To calculate lists of differentially expressed genes between monocyte groups, limma topTable analyses were performed using the Voom-normalised data as input. Empirical Bayes moderation of the standard errors in T-statistics was applied. As the experiment followed a block design with multiple samples from the same infant (unstimulated and stimulated), we accounted for "within infant correlation" in the linear models using blocking variables and the standard limma duplicateCorrelation function.

2.8 Bioinformatics and statistics

2.8.1 Ingenuity® Pathway Analysis Analysis of over-represented canonical pathways, disease classifications and bio- functions was performed through QIAGEN’s Ingenuity® Pathways Analysis (IPA, QIAGEN Redwood City, California, USA). Using the Ingenuity® Knowledge Base as reference, the right-tailed Fisher’s exact test was used to estimate the probability that a function or pathway was significantly over-represented in our designated list of differentially expressed genes. P-values were corrected for multiple comparisons using the Benjamini-Hochberg method (260). Corrected p-values <0.05 were considered significant.

Upstream regulator analysis was performed within IPA to determine the cascade of upstream transcriptional regulators that could explain the observed changes in gene expression in our designated list of differentially expressed genes. Results were based on prior knowledge of known effects between transcriptional regulators and target genes (stored in the Ingenuity® Knowledge Base). Two statistical measures were used to predict significant transcriptional regulators. First, an overlap p-value was calculated using a right-tailed Fisher’s exact test to determine whether the known targets of each transcriptional regulator were over-represented in our dataset. An overlap p-value <0.01 50 was considered significant as recommended by IPA. Second, activation z-scores were calculated to infer the activation state of predicted transcriptional regulators by comparing their known effect on downstream targets with observed changes in gene expression. Transcriptional regulators with activation z-scores ≥2 or ≤2 were considered “activated” or “inhibited”, respectively.

2.8.2 Network analysis Network analysis was performed on differentially expressed genes from multiple analyses in Chapters 4 and 5. The lists of differentially expressed genes were used to generate first- order protein-protein interaction networks using the HINT (High-quality INTeractomes) database (261). First-order protein-protein interaction networks contain the differentially expressed genes, and all literature curated interactions involving these genes and other non-differentially expressed genes. In all analyses, genes involved in ubiquitination, sumoylation, and neddylation, as well as the amyloid precursor protein (APP), were removed prior to sub-network enrichment. While these genes can play an important role in immune function, they form a disproportionate number of interactions in the input HINT database, and are therefore frequently identified as Hubs irrespective of the input gene lists. Removing these genes generally allows for a better assessment of subset- specific network analysis, at the cost of potentially losing information about the role of ubiquitination/sumoylation/neddylation/APP.

Sub-network analysis was then performed on the filtered first-order interaction networks using the jActiveModules Cytoscape plugin to identify the most biologically significant network nodes/interactions (262). The jActiveModules plugin uses the p-values from the input gene list in a scoring function that identifies highly interconnected modules/sub- networks that maximise the score (i.e. containing the largest number of differentially expressed genes and the fewest number of non-differentially expressed genes). Where necessary, sub-networks were merged to obtain the single most significant sub-network containing ~150–250 nodes. Highly interconnected nodes represent the key molecules within each sub-network, and are referred to as Hubs. The comparisons of Hub genes between sub-networks derived from different lists of differentially expressed genes was the end point of the network analysis as it can suggest functional differences between networks.

51

A second method for network analysis was kindly performed by Dr David Hancock using the BioNet R package to identify the conserved monocyte response networks to E. coli and S. epidermidis stimulation presented in Chapter 5 (263). The Bionet package's heinz algorithm performs a similar test to the jActiveModules plugin (maximises the number of highly differentially expressed genes in the resultant sub-network), but generates this sub-network based on the p-values from all genes in the differential expression test, rather than a small number of input genes. As above, the filtered HINT curated human interactome was used as the base network (261). The Bionet package's heinz algorithm was run using settings to generate sub-networks containing ~150 nodes.

All networks from both methods were visualised using the Cytoscape software (v3.2.0), while network metrics were calculated using the built in NetworkAnalyzer plugin (264).

2.8.3 Statistical analyses Statistical analyses were performed using Prism (v5 for Max OS X; GraphPad Software, Inc. California, USA) using tests for non-parametric data, as the data was not normally distributed determined by the D'Agostino-Pearson omnibus test. Friedman tests with Dunn’s multiple comparisons tests were used for all statistics pertaining to Chapter 3, as comparative data were from paired observations across 3 or more conditions. Kruskal- Wallis tests with Dunn’s multiple comparisons tests were used to compare observations across the three infant groups in Chapter 4. Spearman’s correlations for non-parametric data were used for all correlations. P-values less than 0.05 were considered significant.

52

Chapter 3 Development of methodology

3.1 Introduction

A major aim of this thesis was to characterise the transcriptional response of infant monocytes to challenge with live pathogens using RNA-seq-based transcriptomics. Monocytes represent only one of the major immune cell types in cord blood, the frequencies of which are highly variable between individuals (7–38%) (141, 171-173). To account for this variation, and to effectively examine the intrinsic monocyte transcriptional response to pathogenic stimulation without influence from extrinsic cell factors, the purification of these cells from cord blood was essential. Furthermore, while it is possible to sequence the transcriptome of heterogeneous cell populations and then derive a cell-type-specific gene expression signature through complex computational modelling, more accurate transcriptional profiling data are derived from pure cell populations (265, 266). Several studies investigating monocyte function through transcriptional profiling have performed monocyte purification prior to analysis (180, 267-270). These studies have used multiple approaches for monocyte purification, mostly from adult peripheral blood mononuclear cells, and none have purified monocytes from cord blood. The cellular composition of cord blood is distinct from adult peripheral blood. Cord blood is enriched for stem and progenitor cells that lack the lineage markers of mature lymphocyte populations, and this can present a challenge for purification systems that rely on the expression of such lineage markers (271-273). Therefore, it was imperative that we evaluated the best method for monocyte purification from cord blood. This chapter describes the development of methods for the purification and subsequent stimulation of cord blood monocytes with live neonatal pathogens, such that the cells can be used for transcriptional profiling by RNA-seq.

3.2 Optimisation of monocyte purification

The two most important aims of cell separation techniques are to achieve a high purity and high yield of the required cell type. It is also ideal to preserve the natural activation state of the target cell population, so as not to influence the outcomes of downstream experiments. There are multiple methods available for the successful purification of monocytes. Examples include adherence-based techniques, separation-based methods using antibody-conjugated magnetic beads and fluorescence-activated cell sorting (FACS). Purification of monocytes by adherence is a commonly used technique as it is

53 easy and inexpensive to perform, however purity can be low (~60%) and adherence can result in cellular activation, rendering this method unsuitable for the purposes of this thesis (274-276). Magnetic bead separations have the advantage of simple and quick procedures, but are limited to the design and availability of commercial kits; pre-prepared antibody-bead cocktails are supplied for depletion/enrichment of particular cell types based on their expression of lineage markers. On the other hand, cell sorting has the advantage of being highly flexible, allowing users to distinctively target populations of interest through customisable panels of antibodies. However cell sorting can be more expensive and laborious than bead-based enrichment due to the requirement of specialised cell sorting equipment. The specific objectives of the monocyte purification approach used in this thesis to maximise sensitivity in detecting transcriptional differences and to reduce variability were to (i) achieve >90% purity, and (ii) obtain at least 4.5x105 viable cells from 1–2 vials of stored CBMCs (see section 3.3). All optimisation experiments were performed using term infant CBMC.

3.2.1 Preliminary experiments using negative selection techniques The principal advantage of monocyte isolation through negative selection is that the cells of interest remain unlabelled during selection and are therefore more representative of their in vivo state. Both magnetic-bead enrichment and cell sorting techniques were tested for their ability to negatively select CD14+ and CD16± monocytes from cord blood.

3.2.1.1 Magnetic bead-based enrichment A preliminary separation experiment was performed using the EasySep™ human monocyte enrichment kit without CD16 depletion (STEMCELL™ Technologies, Victoria, Australia) according to the manufacturer’s instructions. Briefly, non-monocytes were labelled with monoclonal antibodies conjugated to magnetic particles, allowing the monocytes to be freely poured from the cell suspension when adjacent to a magnet. The post-enrichment yield was acceptable (7.8x105 cells, ~6% of total starting CBMC) however monocyte purity was very low (~24%), despite the kit containing antibodies directed against lineage markers for the major non-monocyte cell populations; T cells, B cells, NK cells, and dendritic cells (CD2, CD3, CD19, CD20, CD56, and CD123). Flow cytometric analysis revealed that the contaminating lineage-negative cells were lymphocyte-sized based on forward- and side-scatter. As there was no way to eliminate

54 these lineage-negative lymphocyte-sized cells using the EasySep™ magnetic enrichment kit, this method was not pursued further.

3.2.1.2 Negative selection by cell sorting A negative selection cell sorting approach was evaluated as an alternative method, which has the added advantage of effective cell discrimination based on their size properties. Monocytes have a distinct forward-scatter (FSC) and side-scatter (SSC) profile compared to that of lymphocytes (277). Using this principle, the optimal negative cell sort gating strategy defined monocytes as large FSC and SSC events that were negative for T cell, B cell, NK cell, and dendritic cell lineage markers (CD3, CD19, CD56, CD123, and BDCA- 1). Three separate cell-sorting experiments were performed on a FACSAria III (BD Biosciences), two of which failed to yield the required numbers of monocytes (<4.5x105 cells, from ~10–12x106 CBMC). Post-sort purity was defined as the percentage of CD14+ cells (± CD16 expression) after further staining of sorted cells with both antibodies. Despite being able to effectively eliminate the lineage-negative lymphocyte-sized cells that were problematic during EasySep™ magnetic enrichment, the highest purity achieved through negative cell sorting was 86%. A distinct population of CD14- CD16- cells (also negative for T, B, NK and dendritic cell markers) were a consistent and significant source of contamination (15–25%). Additional post-sort staining revealed that <1% of the contaminating cells were CD34+ (indicating they were not stem cells), however they were mostly (~65%) HLA-DR+. As the negative cell sort method could not deliver the monocyte purity required for downstream experiments, and with no candidate markers that could distinguish the contaminating cells from monocytes, positive selection by cell sorting was explored.

3.2.2 Comparison between positive and negative cell sorting methods In order to determine whether positive monocyte selection could improve sorting outcomes, a direct comparison of positive and negative monocyte sorting methods was performed. A term infant CBMC sample was thawed, washed, counted and prepared for cell sorting as described in section 2.4.1. The sample was divided; one half was stained with the positive cell sort panel, and the other half stained with the negative cell sort panel (Table 3.1) for 30 minutes at 4°C. Both samples were then sorted using a FACSAria III; gating strategies for each sort are presented in Figure 3.1.

55

Table 3.1 Positive and negative cell sort staining panels Fluorochrome Panel FITC PE PerCP-Cy5.5 PE-Cy7 APC BV421 Positive sort CD16 - CD14 HLA-DR - - Negative sort CD19 BDCA-1 - CD123 CD3 CD56

Figure 3.1 Sequential cell sort gating strategies for monocyte purification by positive and negative selection. Monocytes were positively selected by first gating on HLA-DR+ cells with high side-scatter (A), of which only CD14+/CD16+ cells were included for sorting (B). Negative selection was achieved by first gating on CD3- and high side-scatter cells (C), followed by exclusion of CD56+ and CD19+ cells (D), and finally exclusion of BDCA-1+ and CD123+ cells (E).

56

3.2.2.1 Comparison of cell sorting outcomes A direct comparison of cell sorting outcomes is presented in Table 3.2. Positive monocyte sorting resulted in a higher yield, but slightly lower purity than negative monocyte sorting (due to an inability to exclude non-monocytes). Post-sort staining of each sample revealed that a lower frequency of lineage-negative cells was found in the sorted cell fraction of positively selected monocytes. In addition, fewer monocytes were lost to the non-sorted fraction using positive selection. Further examination of the positively sorted sample revealed contamination by T cells (2.1%) and NK cells (7.2%), whereas contamination in the negatively sorted sample was exclusively due to lineage-negative cells. Overall, positive selection performed better than negative selection in reducing contamination by lineage-negative cells.

Table 3.2 Comparison of positive and negative cell sorting outcomes Positive sort Negative sort Post-sort monocyte yield 1.4 x106 0.8 x106 Purity (CD14+, CD16±) 81.5% 85.1% Post-sort viability (% trypan blue negative) 98.1% 98.5% % Lineage negative cells in sorted fraction 9.3% 14.9% % CD14+ cells in the non-sorted fraction 3.2% 6.8% 2.1% CD3+ (T cells) 14.9% lineage- Phenotype of contaminating cells 7.2% CD56+ (NK cells) negative

3.2.2.2 Analysis of inflammatory gene and protein expression to determine the effects of positive selection The main concern of using positive selection to purify cell populations is the possibility of activating the cells of interest due to direct labelling. To test whether positive selection of monocytes by cell sorting would induce cell activation, gene and protein expression of inflammatory cytokines was measured post-sort in both positively and negatively sorted monocytes. IL-6 and TNF gene expression were measured by RT-PCR (see section 2.6.6) immediately post-sort (time zero), and following 6 hours of culture with media alone, with LPS (TLR4 agonist, 10 ng/mL) or with CL097 (TLR7/8 agonist, 1 g/mL). LPS and CL097 stimulations were included to test whether positive selection would alter TLR-induced monocyte inflammatory responses. Compared to time zero samples, IL-6

57 time zero samples, IL-6 and TNFα gene expression was upregulated relative to 18S to a similar extent in both positively and negatively sorted monocytes following 6 hours of culture without additional stimulation (Figure 3.2A). Furthermore, similar fold change increases in IL-6 and TNFα gene expression were observed between positively and negatively sorted monocytes following stimulation with LPS or CL097, indicating that positive selection does not interfere with TLR-induced production of these cytokines (Figure 3.2B-C).

A

25

20

15

10

5 Delta Ct (GOI - 18S) 0

T0 T6 T0 T6

IL-6 TNFα

B C IL-6 TNFα

15000 500

400 10000 300

200 5000 100

0 0 Fold change in gene expression gene in change Fold expression gene in change Fold LPS CLO97 LPS CLO97

Figure 3.2 Comparison of IL-6 and TNFα gene expression by positively and negatively sorted monocytes. (A) IL-6 and TNFα gene expression relative to 18S immediately post-sort (T0) or following 6 hours of unstimulated culture (T6). Fold change in IL-6 (B) and TNFα (C) gene expression following 6 hours of LPS or CL097 stimulation, relative to unstimulated cultures (calculated using the standard 2-ΔΔCt method with 18S as a reference gene). White and grey bars represent positively and negatively sorted monocytes, respectively. GOI, gene of interest. Data represents results from duplicate cultures from a single experiment.

58

The protein levels of IL-6, IL-8, IL-10 and TNFα in culture supernatants were measured using an in-house multiplex fluorescent-bead immunoassay (see section 2.4.5). Levels of all cytokines were higher post-stimulation with LPS or CL097, but were similar between positively and negatively sorted monocytes (Figure 3.3).

A B Positive sort IL-6 IL-8 Negative sort 80000 10000

8000 60000 6000 40000 4000 IL-8 pg/mL IL-8 IL-6 pg/mL IL-6 20000 2000

0 0

Unstimulated LPS CLO97 Unstimulated LPS CLO97

C D IL-10 TNFα

80 15000

60 10000 pg/mL 40 α

IL-10 pg/mL IL-10 5000 20 TNF

0 0

Unstimulated LPS CLO97 Unstimulated LPS CLO97

Figure 3.3 Comparison of cytokine production by positively and negatively sorted monocytes. Levels of IL-6 (A), IL-8 (B), IL-10 (C) and TNFα (D) in monocyte culture supernatants following 6 hours without stimulation, or after LPS or CL097 stimulation. Data are presented as mean ± SEM from duplicate cultures.

3.2.3 The optimal method for monocyte purification from cord blood: cell sorting using a combination of positive and negative markers

The data presented so far indicated that neither negative nor positive selection alone were effective in purifying monocytes from CBMC to a purity of >90%. Positive selection offered the advantage of reducing the amount of contamination by lineage- negative cells as well as reducing monocyte loss to the non-sorted fraction, without activating the monocytes. However without the inclusion of non-monocyte lineage markers, T cells and NK cells were sources of contamination in the post-sort sample of positively selected monocytes. In light of this, a novel cell sorting panel was designed

59 that included antibodies directed against both monocyte and non-monocyte lineage markers; CD3, CD19, CD56, CD14 and CD16. The gating strategy for monocyte purification using this optimised panel is presented in section 2.4.3. Sample purities obtained using this method were consistently >95%, and all samples used in the final RNA-seq experiment were >90% pure (Figure 3.4A). Monocyte yields obtained using this method were often in excess of the required 4.5x105 cells, depending on the frequency of monocytes in the unsorted CBMC sample and the absolute numbers of starting CBMC (Figure 3.4B). In addition, the proportions of monocyte subsets were similar between pre- and post-sort samples (presented in Chapter 4, Figure 4.2).

A B 100 1×107 p<0.0001 Spearman r = 0.93 95% CI = 0.83-0.97 95 1×106

% purity 90 1×105

Post-sort monocyte yield Post-sort monocyte 85 1×104 4 5 6 7 Preterm Preterm Term 1×10 1×10 1×10 1×10 HCA - HCA + Estimated available monocytes (frequency X total numbers of CBMC)

Figure 3.4 Post-sort monocyte sample purities and yields using the optimised cell sorting protocol. (A) Sample purities for the final samples used for RNA-seq, defined as the percentage of cells within the sample that were CD14+ and CD16±. All samples were >90% pure (represented by the dashed line). There were no significant differences in sample purity between preterm and term infants (Kruskal-Wallis test with Dunn’s multiple comparisons test). Data are presented for individual donors (symbols) with bars showing mean ± SEM. (B) Examples of post-sort monocyte yields obtained using the optimised cell sorting protocol, dependent on the frequency of monocytes within the CBMC sample and the total numbers of starting CBMC (Spearman correlation). The dashed line represents the minimum yield requirement for downstream culture and RNA-seq. HCA, histologic chorioamnionitis.

3.2.4 Summary

Effective isolation of monocytes from cord blood is critical to this thesis. The two criteria of effective isolation were to: (i) achieve a minimum purity of 90% and (ii) obtain a minimum yield of 4.5x105 cells. Negative selection of monocytes from cord blood using either magnetic-bead enrichment or cell sorting failed to achieve the required purity (maximum purity 86%). The failure of negative selection in this case can be attributed to the presence of lineage-negative cells in cord blood that are otherwise missing in adult blood, for which these methods may be optimal. Positive selection offered a solution to this problem and did not lead to activation of relevant transcripts in

60 the sorted monocytes, in line with the findings of other studies (278-280). By targeting both monocytes and non-monocytes, cell sorting proved to be an effective method for the purification of monocytes from cord blood.

3.3 Determining RNA input requirements for RNA-seq

RNA-seq has fast become the most comprehensive and accurate next-generation sequencing technology for analysing the entire set of transcripts in a cell. Effective preparation of complimentary DNA (cDNA) libraries from each RNA sample is critical to successful sequencing, and depends entirely on the quantity and quality of RNA derived from the sample in question. The AGRF (our chosen facility for RNA-seq) prepares cDNA libraries using the TruSeq Stranded mRNA Library Prep Kit (Illumina, Inc., California, USA), for which the RNA input recommendations are 0.1–4 µg of total RNA. The quality of input RNA is equally important to derive accurate gene expression data, and is determined by sample purity and integrity. A 260/280 absorbance ratio of >1.8 is generally indicative of pure RNA (determined by spectrophotometry). RNA integrity can be measured by microcapillary electrophoretic separation of RNA using an Agilent 2100 Bioanalyzer (Agilent Technologies), which employs a patented algorithm to derive an RNA integrity number (RIN) (281). The AGRF guidelines for sample submission recommend that samples have RIN values ≥8.

Based on these quality control parameters, only samples containing ≥100 ng total RNA, with an A260/A280 ratio of >1.8, and a RIN ≥8 would be considered eligible for RNA- seq. To determine how many cells would be required to fulfil these requirements, monocytes were purified from term infant CBMC using the optimised cell sorting protocol, and RNA extractions were performed on increasing numbers of cells (5x104 – 1.5x105, Table 3.3). In addition, the performances of two Qiagen RNA extraction kits were compared; the AllPrep DNA/RNA Micro kit and the RNeasy Micro kit. RNA quantity, purity and integrity were determined using a Qubit® 2.0 Fluorometer (Life Technologies), a NanoDrop 2000 (Thermo Scientific) and an Agilent 2100 Bioanalyzer (Agilent Technologies), respectively.

The results from six extractions are summarised in Table 3.3. Regardless of the kit used, extractions from 5x104 monocytes failed to yield a detectable quantity of total RNA (below the limit of detection; 25 ng/mL). Extractions from 1x105 monocytes produced

61

RNA of sufficient purity, but not sufficient quantity or integrity using either kit. The Qiagen AllPrep DNA/RNA micro kit performed better than the RNeasy micro kit, and yielded ≥100 ng total RNA from 1.5x105 monocytes with sufficient purity and integrity, whereas the RNeasy micro kit could only satisfy the purity requirement using this number of cells. Based on these results, 1.5x105 monocytes were used for each culture condition, and all future RNA extractions were performed using the Qiagen AllPrep DNA/RNA kit.

Table 3.3 Measurements of RNA quantity, purity and integrity comparing total cell numbers and RNA extraction kit. Qubit NanoDrop Bioanalyzer (quantity) (purity) (integrity) No. of Qiagen extraction Total RNA 260/280 RIN monocytes kit AllPrep DNA/RNA Undetectable 1.8 - 5x104 RNeasy Undetectable 1.9 - AllPrep DNA/RNA 85.7 ng 2.1 7 1x105 RNeasy 57.6 ng 1.9 - AllPrep DNA/RNA 147.6 ng 2.1 9.5 1.5x105 RNeasy 78.9 ng 1.9 - RIN, RNA integrity number.

62

3.4 Culture of E. coli and S. epidermidis

Live preparations of E. coli (strain ATCC 11775) and S. epidermidis (strain WT 1457) were chosen as stimuli instead of killed preparations or purified bacterial products as they provide a better representation of in vivo neonatal infection. In addition, heat- or ethanol-killed preparations of S. epidermidis elicit distinct PMBC cytokine responses compared to live preparations in vitro, and monocyte viability is significantly reduced upon stimulation with heat-killed S. epidermidis (199). In an effort to reduce technical variability between monocyte stimulation experiments, frozen mid-logarithmic stocks of E. coli and S. epidermidis were cultivated and stored as described in section 2.3.3. Figure 3.5A and C depict growth curves for S. epidermidis and E. coli respectively, indicating the point of mid-logarithmic growth. The viability of these frozen mid- logarithmic stocks was tested five times over a period of 104 days as described in section 2.3.4, and was stable for both S. epidermidis and E. coli (Figure 3.5B and D). All monocyte stimulations using these stocks were performed within this time frame.

A B S. epidermidis growth curve S. epidermidis viability 10 10

8 1 6 600 OD 0.1 4 Log CFU/mL

2 0.01 0 100 200 300 400 0 50 100 150 Time (minutes) Days frozen

C E. coli growth curve D E. coli viability 10 10

8 1 6 600 OD 0.1 4 Log CFU/mL

2 0.01 0 100 200 300 0 50 100 150 Time (minutes) Days frozen

Figure 3.5 Growth curves and viability of S. epidermidis and E. coli. Growth curves showing measurements of optical density at 600nm over time to determine mid- logarithmic phase of growth for S. epidermidis (A) and E. coli (C). Viability (CFU/mL) of frozen mid- logarithmic stocks of S. epidermidis (B) and E. coli (D) over time.

63

3.5 Optimisation of monocyte culture conditions

3.5.1 Resting cells prior to bacterial stimulation

Cell sorting proved to be the most effective method for purifying monocytes from cord blood, however the mechanics of this process leads to dramatic changes in the cellular environment. Cells are exposed to high pressure (45 PSI), rapidly accelerated (to 20 meters/second), are charged to a few hundred volts and then passed through an electric field before hitting a liquid surface at high speed. Murine monocytes exposed to external pressure exhibit upregulated gene expression of the potent monocyte chemoattractant protein MCP-1, via increased phosphorylation of MAP kinases (282). This raises important questions regarding the impact of the cell sorting process on monocyte activation, especially when aiming to measure transcriptional responses that may be sensitive to perturbations induced by this process.

To determine whether the cell sorting process lead to monocyte activation (induction of protein secretion or gene expression), purified cord blood monocytes from four term infants were rested by incubation with media alone in sterile polypropylene (low- binding) 96-well plates for 0, 5, 10, 15 and 23 hours post-sort (duplicate cultures of 5x104 monocytes/well). The duplicate cultures were then pooled, and monocytes pelleted and stored in 300 µL RNAprotect Cell Reagent (Qiagen) at -20°C prior to batch RNA extractions using the Qiagen AllPrep DNA/RNA Micro kit (see section 2.6.1). Total RNA was then analysed using a Qubit® 2.0 Fluorometer (Life Technologies) and a NanoDrop 2000 (Thermo Scientific), to assess any changes in RNA quantity and purity over time. Complimentary DNA was synthesised from extracted RNA (see section 2.6.2), and RT-PCR was performed for detection of IL-6 and TNFα gene expression as described in section 2.6.6 (reported as fold-change in gene expression compared to time zero). Monocyte culture supernatant was harvested at each time point (with the exception of time zero), and stored at -80°C prior to quantitation of IL-6 and TNFα protein levels (see section 2.4.5). LDH release was also measured in culture supernatants as an indicator of cell death (see section 2.5.1).

64

3.5.1.1 Post-sort assessment of RNA quantity and quality There was a trend towards decreasing RNA quantity over time, however there were no statistically significant differences in RNA quantity (p=0.08) or purity (p=0.48) between time points (Friedman test with Dunn’s multiple comparisons test; Figure 3.6).

A B 150 2.5

2.0 100 1.5

1.0 50 Total RNA (ng) (ng) RNA Total A260/A280 ratio A260/A280 0.5

0 0.0 0 5 10 15 23 0 5 10 15 23 Hours resting post-sort Hours resting post-sort

Figure 3.6 Post-sort measurements of monocyte RNA quantity and purity over time. Term infant monocytes were purified by cell sorting and then rested for 0, 5, 10, 15 and 23 hours. (A) Total RNA and (B) purity of RNA extracted from 1x105 monocytes at each time point. Dashed line represents the minimum threshold for pure RNA. Data is presented as mean ± SEM, n=4 donors.

65

3.5.1.2 Post-sort expression of inflammatory cytokines by unstimulated monocytes An increase in IL-6 and TNFα gene expression was observed in unstimulated purified monocyte cultures, with the highest fold change at 5 hours post-sort (Figure 3.7). Expression of both genes steadily declined to the minimum at 15 hours post-sort, before increasing to intermediate levels at 23 hours post-sort. These changes in gene expression were mirrored by similar changes in IL-6 and TNFα protein levels measured in paired culture supernatant (Figure 3.7).

A Protein

Gene Fold change IL-6 gene expression

4000 600

3000 400

2000

IL-6 pg/mL IL-6 200 1000

0 0 0 5 10 15 20 25 Hours resting post-sort B Fold change TNF

500 10

400 8

300 6

α pg/mL gene expression α 200 4 TNF 100 2

0 0 0 5 10 15 20 25 Hours resting post-sort

Figure 3.7 Post-sort levels of IL-6 and TNFα gene and protein expression by monocytes of over time. Term infant monocytes were purified by cell sorting and then rested for 5, 10, 15 and 23 hours. Left axes depict protein levels of IL-6 (A) and TNFα (B) in culture supernatants, and right axes indicate corresponding fold changes in gene expression compared to time zero samples (calculated using the standard 2-ΔΔCt method with 18S as a reference gene). Data are presented as mean ± SEM, n=4 donors. A Wilcoxon matched-pairs test revealed no significant differences between time points for either IL-6 or TNFα.

66

3.5.1.3 Post-sort assessment of cell death The percentage of monocyte lysis in each culture was quantified from LDH release into the culture supernatant, using cells treated with 3% Triton X-100 or cell culture medium alone as 100% and 0% lysis controls, respectively. There was a trend toward increased monocyte lysis over time in post-sort cultures, with a maximum of 19.5% at 23 hours post-sort (Figure 3.8), however there were no statistically significant differences between time points (p=0.075; Friedman test with Dunn’s multiple comparisons test).

100

80

60 40

% Monocyte lysis 20

0 5 10 15 23

Hours resting post-sort

Figure 3.8 Post-sort measurements of LDH release over time. Term infant monocytes were purified by cell sorting and then rested for 0, 5, 10, 15 and 23 hours. Cell lysis was determined using measurements of LDH in monocyte culture supernatant. Data are presented are mean ± SEM, n=4 donors.

3.5.1.4 Summary These results indicate that the cell sorting process induces monocyte activation, evidenced by increased gene expression of IL-6 and TNFα in unstimulated post-sort cultures, accompanied by increased protein secretion of these cytokines into the culture supernatant. However by 15 hours post-sort, gene expression of these cytokines had returned to levels comparable to those seen at time zero (Figure 3.7), suggesting that the monocytes had returned to steady-state. Separate analysis of RNA quantity and purity indicated that there was a trend toward decreased quantity of extracted RNA over time (Figure 3.6). This was paralleled by a slight increase in monocyte lysis over time (Figure 3.8), which may in part explain the reduction in total RNA observed. However there were no significant differences in RNA quantity, purity or the percentage of monocyte lysis at 15 hours post-sort compared with the earliest time points. Therefore in all future experiments, monocytes were rested for 15 hours following cell sorting prior to stimulation with E. coli and S. epidermidis. 67

3.5.2 Optimisation of monocyte stimulation with live E. coli and S. epidermidis

The following experiment was designed to determine the optimal stimulating dose of E. coli and S. epidermidis, defined as that which would induce a detectable monocyte inflammatory response (gene and protein level), without causing excessive cell death or affecting the quantity or purity of extracted RNA. A stimulation time of 2 hours was selected to capture the initial events in pathogen recognition, and is in line with other studies that have used TLR-agonists or live E. coli as stimuli to characterise human monocyte inflammatory responses (such as IL-6, TNF and IL-1β) (283-286).

Purified monocytes from three term infant donors were rested for 15 hours by incubation with media alone in sterile polypropylene 96-well plates, before a 2-hour stimulation with three doses of live E. coli or S. epidermidis, or negative control (PBS). Doses were calculated as the multiplicity of infection (MOI; the ratio of bacterial CFU to one monocyte) starting at 1:1, increasing to 10:1 with the highest dose at 100:1. All cultures were performed in triplicate, with 5x104 monocytes/culture. Following stimulation, two of the triplicate cultures were pooled, and the monocytes and culture supernatants were stored and analysed as described in section 3.5.1. The remaining triplicate culture was used to assess cell death by apoptosis through flow cytometric analysis of Annexin V and 7-AAD staining (see section 2.5.2).

3.5.2.1 Inflammatory cytokine analysis Gene expression All three doses of E. coli induced 100 to 1000-fold higher levels of IL-6 and TNFα gene expression compared with PBS-stimulated monocytes. S. epidermidis was a less potent stimulator of gene expression for both cytokines especially at the lowest dose (MOI 1:1), which induced the lowest fold-changes of 14.6 and 11.3 for IL-6 and TNFα respectively (Figure 3.9A and C). It should be noted that IL-6 was undetectable by RT- PCR in two of the three infant’s negative control samples (PBS-stimulated monocytes); therefore fold-change analysis for IL-6 gene expression could only be performed for one infant. However ΔCt values (which are an internal measure of IL-6 gene expression relative to 18S within individual samples) were similar across all three infants in all cultures (Figure 3.9B), and followed the same pattern as seen with fold-change for IL-6. This was also true for TNFα (Figure 3.9D).

68

A B IL-6 IL-6 10000 25 (n=1) 20 1000 15 100 10 10 5 Delta Ct (IL-6 - 18S)

1 0

expression gene in change Fold 1:1 10:1 100:1 1:1 10:1 100:1 PBS 1:1 10:1 100:1 1:1 10:1 100:1 E. coli S. epidermidis E. coli S. epidermidis

C D TNFα TNFα 1000 20

15

100 18S) - α 10 TNF 10 5

Delta Ct ( 1 0 expression gene in change Fold 1:1 10:1 100:1 1:1 10:1 100:1 PBS 1:1 10:1 100:1 1:1 10:1 100:1 E. coli S. epidermidis E. coli S. epidermidis

Figure 3.9 IL-6 and TNFα gene expression following stimulation with increasing MOIs of E. coli or S. epidermidis. Term infant monocytes were purified by cell sorting and stimulated with increasing MOIs of E. coli or S. epidermidis for 2 hours, following a 15-hour rest. Fold change in IL-6 (A) and TNFα (C) gene expression relative to unstimulated monocytes (calculated using the standard 2-ΔΔCt method with 18S as a reference gene; n=1 for IL-6). The relative expression of IL-6 (B) and TNFα (D) to 18S was calculated within each individual sample (ΔCt), where a lower value indicates increased expression of the gene of interest. Data are presented as mean ± SEM, n=3 donors.

69

Protein expression Considerable levels of IL-6 protein (415–1908 pg/mL) were measured in the culture supernatant of all samples, including supernatant from PBS-stimulated monocytes (Figure 3.10A), with no statistically significant differences in IL-6 secretion between culture conditions (p=0.0929; Friedman test with Dunn’s multiple comparisons test). The relatively high level of IL-6 protein measured in the supernatant of PBS-stimulated monocytes is comparable to those observed in unstimulated monocytes post-15 hour rest (Figure 3.7). Modest increases (compared to PBS-stimulated monocytes) in IL-6 secretion were detected following stimulation with E. coli, with the greatest mean increase being 464 pg/mL at MOI 10:1. However the effect of S. epidermidis stimulation was more subtle, with the greatest mean increase in IL-6 secretion being only 113 pg/mL compared to PBS-stimulated monocytes at the highest MOI.

In contrast to IL-6, monocyte secretion of TNFα was strongly induced by both E. coli and S. epidermidis stimulation, and was almost undetectable in PBS-stimulated cultures. E. coli was a potent inducer of TNFα secretion at all doses, whereas S. epidermidis- induced TNFα secretion followed a dose-dependent pattern (Figure 3.10B). These results closely mirror the changes in monocyte TNFα gene expression induced by bacterial stimulation observed in Figure 3.9.

A IL-6 B TNFα

2000 2500 *

2000 1500

1500

1000 pg/mL α 1000 IL-6 pg/mL IL-6 TNF 500 500

0 0 PBS 1:1 10:1 100:1 1:1 10:1 100:1 PBS 1:1 10:1 100:1 1:1 10:1 100:1 E. coli S. epidermidis E. coli S. epidermidis

Figure 3.10 IL-6 and TNFα protein expression following stimulation with increasing MOIs of E. coli or S. epidermidis. Term infant monocytes were purified by cell sorting and then stimulated with increasing MOIs of E. coli or S. epidermidis for 2 hours following a 15-hour rest. Levels of IL-6 (A) and TNFα (B) measured in monocyte culture supernatants. Data are presented as mean ± SEM, n=3 donors. Statistical analysis performed using the Friedman test with Dunn’s multiple comparisons test. *p<0.05.

70

3.5.2.2 Cell death assays A combination of established methods for the detection of necrosis/pyroptosis and apoptosis were used to gain a global picture of the effects of E. coli and S. epidermidis stimulation on monocyte cell death (287). LDH was measured in monocyte cultures supernatants as an indicator of cell lysis resulting from necrosis or pyroptosis. Cultured monocytes were stained with a combination of fluorescently labelled Annexin V and 7- AAD for detection of necrosis and apoptosis.

LDH release Stimulation with E. coli induced greater monocyte lysis compared to S. epidermidis and followed a dose-dependent pattern, with the most lysis observed at MOI 100:1 (mean 38.4%, Figure 3.11). S. epidermidis stimulation induced more monocyte lysis compared to PBS-stimulate monocytes only at the highest dose. Friedman test with Dunn’s multiple comparisons test revealed no statistically significant differences between culture conditions.

100

80

60 40

% Monocyte lysis 20

0 PBS 1:1 10:1 100:1 1:1 10:1 100:1

E. coli S. epidermidis

Figure 3.11 Monocyte lysis following stimulation with increasing MOIs of E. coli or S. epidermidis. Term infant monocytes were purified by cell sorting and then stimulated with increasing MOIs of E. coli or S. epidermidis for 2 hours following a 15-hour rest. Cell lysis was determined from measurements of LDH in monocyte culture supernatant against a 100% lysis control (treatment with 3% Triton-X 100). Data are presented as mean ± SEM, n=3 donors.

71

Annexin V and 7-AAD staining The majority of monocytes were live in all cultures, indicated by the absence of both Annexin V and 7-AAD staining and the retention of a normal FSC and SSC profile (Figure 3.12A and F). Of the remaining monocytes, most were in the early stage of apoptosis (Annexin V positive, 7-AAD negative), however there were no statistically significant differences between cultures (p=0.92, Figure 3.12B). On average, fewer than 12% of monocytes were in late-apoptosis across all cultures (Annexin V positive, 7- AAD positive), with the exception of monocytes stimulated with the highest dose of E. coli, which contained a significantly higher proportion (mean 19%) of late-apoptotic cells compared with PBS-stimulate cultures (Figure 3.12C). All cultures contained less than 2% necrotic monocytes, with the majority containing less than 1% (Annexin V negative, 7-AAD positive) (Figure 3.12D). Representative flow cytometry plots of Annexin V and 7-AAD staining are shown in Figure 3.12E-F and demonstrate that as the cells progress from early to late apoptosis, their SSC and FSC profiles ultimately decrease until they enter necrosis and can no longer be identified using size parameters alone as their plasma membranes have broken down.

72

A B Live (Q1) Early apoptotic (Q2)

100 100

80 80

60 60

40 40

20 20 % Annexin V- / 7-AAD- % Annexin V+ / 7-AAD- 0 0 PBS 1:1 10:1 100:1 1:1 10:1 100:1 PBS 1:1 10:1 100:1 1:1 10:1 100:1

E. coli S. epidermidis E. coli S. epidermidis

C D Late apoptotic (Q3) Necrotic (Q4) 100 100 80 10 60 40 * 1 Q4! Q3! Q1! 20 Q2! Q3! Q4! % Annexin V- / 7-AAD+

% Annexin V+ / 7-AAD+ 0 0.1 PBS 1:1 10:1 100:1 1:1 10:1 100:1 PBS 1:1 10:1 100:1 1:1 10:1 100:1 E. coli S. epidermidis E. coli S. epidermidis

Q1! Q2!

E F G H

Q4! Q3! Q1! Q2! Q3! Q4!

Q1! Q2!

Representative flow plots Heat shock UV exposure

Figure 3.12 Frequencies of apoptotic monocytes following stimulation with increasing MOIs of E. coli or S. epidermidis. Term infant monocytes were purified by cell sorting and then stimulated with increasing MOIs of E. coli or S. epidermidis for 2 hours following a 15-hour rest. Frequencies of live (A), early apoptotic (B), late apoptotic (C), and necrotic (D) cells were determined by flow cytometry using Annexin V and 7-AAD staining (E-F). Heat shocked (G) and UV exposed (H) CMBC were included as positive apoptotic controls. Data are presented as mean ± SEM, n=3 donors. Statistical analysis performed using the Friedman test with Dunn’s multiple comparisons test. *p<0.05.

73

3.5.2.3 RNA assessment To identify whether bacterial stimulation had an impact on the quantity or purity of RNA extracted from cultured monocytes, these parameters were assessed as described in section 3.3. The results from these assessments are presented in Figure 3.13, where dashed lines on each graph depict the minimum requirements for RNA yield and purity. In section 3.3 it was determined that 1.5x105 monocytes were required to yield the minimum quantity of RNA for transcriptome sequencing (100 ng), however in these experiments only 1x105 monocytes were used per culture, therefore the minimum requirement for RNA quantity has been adjusted proportionally to ~66 ng. All cultures met this requirement, with the exception of monocytes stimulated with S. epidermidis at the highest dose, which did not meet the threshold (Figure 3.13A). In contrast, monocytes stimulated with the highest dose of E. coli yielded significantly higher quantities of RNA compared with PBS-stimulated monocytes, with values ~4 times greater than that required. RNA extracted from all cultures was considered pure, with A260/A80 ratios of at least 1.8 (Figure 3.13B).

A B 400 2.5

* 2.0 300

1.5 200 1.0 Total RNA (ng) (ng) RNA Total A260/A280 ratio A260/A280 100 0.5

0 0.0 PBS 1:1 10:1 100:1 1:1 10:1 100:1 PBS 1:1 10:1 100:1 1:1 10:1 100:1 E. coli S. epidermidis E. coli S. epidermidis

Figure 3.13 Measures of monocyte RNA quantity and purity following stimulation with increasing MOIs of E. coli or S. epidermidis. Term infant monocytes were purified by cell sorting and then stimulated with increasing MOIs of E. coli or S. epidermidis for 2 hours following a 15-hour rest. (A) Total quantities of RNA extracted from 1x105 monocytes and (B) purities of extracted RNA. Dashed lines represent minimum thresholds for RNA quantity and purity. Data are presented as mean ± SEM, n=3 donors. Statistical analysis performed using the Friedman test with Dunn’s multiple comparisons test. *p<0.05

74

3.5.2.4 Summary The aims of these experiments were to determine the optimal stimulating dose of E. coli and S. epidermidis that would induce a detectable monocyte inflammatory response (gene and protein level) without causing excessive cell death or affecting the quantity or purity of extracted RNA. E. coli was a potent monocyte stimulus, eliciting strong increases in IL-6 and TNFα at both the gene and protein level even at the lowest dose (Figure 3.9, Figure 3.10). TNFα production followed a dose-dependent pattern, whereas the greatest increases in IL-6 (gene and protein) were observed following stimulation with E. coli at the intermediate dose (MOI 10:1). The quantity and purity of RNA extracted from E. coli-stimulated monocytes at all doses met the requirements for RNA- seq (Figure 3.13). However cell death assays revealed that monocyte stimulation with the highest dose of E. coli (100:1) led to noticeable increases in the percentage of monocyte lysis, and a significant increase in the proportion of late-apoptotic cells (Figure 3.11, Figure 3.12). Therefore an MOI of 10:1 was chosen as the optimal stimulating dose of E. coli.

In general, S. epidermidis was a less potent monocyte stimulus than E. coli, with the lowest dose inducing less than one tenth the increase in IL-6 and TNFα mRNA seen with the higher doses and effectively no TNFα protein (Figure 3.9, Figure 3.10). The middle and highest doses of S. epidermidis both induced substantial increases in IL-6 and TNFα mRNA as well as TNFα protein. In contrast, IL-6 protein levels (whilst >1000 pg/mL) remained similar to those seen in PBS-stimulated monocyte cultures with all doses of S. epidermidis. Similar to the effects seen with E. coli, the highest dose of S. epidermidis (MOI 100:1) induced notably higher levels of monocyte lysis (Figure 3.11). In addition, the quantity of RNA extracted from these cultures did not meet the minimum requirements for RNA-seq (Figure 3.13). Therefore an MOI of 10:1 was chosen as the optimal stimulating dose of S. epidermidis.

75

3.6 RNA-seq experimental design

There are currently no standard guidelines for RNA-seq experimental design, as the requirements for each experiment will be unique depending on the biological question and the nature of the RNA sample. However in general there are two main points to consider: 1) How much sequencing depth is required? and 2) How many biological replicates are needed to see a significant change in gene expression? There is often a trade-off between sequencing depth and the number of biological replicates due to budget constraints. Therefore, it is imperative that the RNA-seq experimental design balances these two parameters in an effort to maximise statistical power for accurate detection of differentially expressed genes. In this case, the final RNA-seq experimental design targeted a sequencing depth of 10–20 million mapped reads per sample with five biological replicates per group as outlined below.

In recent years a number of original and simulation studies have indicated that approximately 10–20 million mapped reads per sample provides sufficient sequencing depth for the detection of differentially expressed genes, and in fact increasing sequencing depth beyond this threshold provides minimal gains in statistical power (253, 288, 289). On the other hand, increasing the number of biological replicates bears much more weight and significantly increases the number of differentially expressed genes identified regardless of sequencing depth (290, 291). Using a mathematical model developed by Hart et al. (based on gene expression distributions from 127 RNA-seq experiments), we estimated that five biological replicates per group were required to detect >2.3 fold-changes in gene expression with 90% power (253). The model takes into account Type 1 error rate (set at 0.05), sequencing depth (gene counts, set at 200) and estimated biological variation (set at 0.4).

3.6.1.1 A preliminary RNA-seq experiment to validate sequencing depth and sample processing methodology Sequencing is charged on a per lane basis, therefore sequencing multiple samples within the same lane (multiplexing) is an ideal way to minimise costs, and reduce variation. Having estimated that approximately 10–20 million mapped reads were required per sample for differential expression analysis, a preliminary RNA-seq experiment was conducted to determine the level of sample multiplexing that would achieve the required sequencing depth. In addition, the success of this experiment would validate the optimised monocyte purification and stimulation methodology. The transcriptomes of 76 nine samples from three infants were multiplexed into one pool and sequenced (50 bp single-end) over two lanes (to assess lane/lane variability). One infant from each group was represented (term, preterm HCA-, preterm HCA+) and all infants had one sample each of unstimulated, E. coli- and S. epidermidis-stimulated monocytes.

Both sequencing lanes generated well above the expected 150 million reads, with an average of ~192 million reads across both lanes (with minimal lane/lane variation) resulting in an average of ~21 million raw reads per sample (Table 3.4). On average ~80% of these raw reads aligned to the human genome (see section 2.7.4) resulting in 16.5–19.9 million mapped reads per sample (directly within the required range) with the exception of one sample; an E. coli-stimulated sample from the preterm HCA+ infant (in red, Table 3.4).

Only 34.5% of the reads from this sample mapped to the human genome resulting in only 6.6 million mapped reads (about half of what is required). Inspection of the bioanalyzer report for this revealed an extra peak on the electropherogram that was also present in the other E. coli-stimulated samples (Figure 3.14), but was missing in the S. epidermidis-stimulated and unstimulated samples. In all cases this extra peak was incorrectly designated as either the 18S or 28S human RNA subunit by the bioanalyzer software resulting in an overestimation of RNA concentration (peak height correlated with the estimated sample RNA concentration). To confirm that E. coli genomic content was the source of contamination, reads from all E. coli-stimulated monocyte samples were mapped to the E. coli genome, along with reads from an unstimulated monocyte sample as a negative control (Table 3.5). The E. coli-stimulated sample that had the lowest proportion of reads mapping to the human genome had the highest proportion of reads mapping to the E. coli genome (54.6%), which corresponded with the highest extra peak observed in the bioanalyzer electropherogram (Figure 3.14A). The implications of this finding were that this sample in particular was effectively under- sequenced due to an overestimation of human RNA concentration (as all samples are normalised to lowest total RNA weight). To correct for this in the principle RNA-seq experiment, all E. coli-stimulated monocyte samples were arbitrarily assigned an RNA concentration equivalent to the average of all unstimulated monocyte samples.

77

Table 3.4 Raw read statistics from a preliminary RNA-seq experiment Number of reads – Number of reads – % Mapped to Total # of reads mapped to Infant group Sample Lane 1 Lane 2 human genome human genome (per lane) Unstimulated 23,487,243 21,429,899 87.6% 19,671,463 Term E. coli-stimulated 20,475,295 18,365,594 85.4% 16,588,944

S. epidermidis-stimulated 22,233,385 20,335,100 89.4% 18,181,000 Unstimulated 21,921,744 19,826,943 89.8% 18,753,510 Preterm HCA- E. coli-stimulated 23,643,675 21,430,395 84.9% 19,127,182 S. epidermidis-stimulated 23,852,767 21,641,869 87.6% 19,919,827 Unstimulated 22,517,049 20,243,175 85.4% 18,262,892 Preterm HCA+ E. coli-stimulated 20,062,007 18,468,282 34.5% 6,654,181 S. epidermidis-stimulated 23,182,013 21,114,702 85.4% 18,919,127 Total lane reads 201,375,178 182,855,959 Average reads/sample 22,375,019 20,317,328 Standard deviation 1,364,101 1,242,048

Table 3.5 E. coli genome mapping statistics Sample Reads mapped to E. coli genome Reads unmapped Term: E. coli-stimulated 9.5% 90.5% Preterm HCA-: E. coli-stimulated 2.2% 97.8% Preterm HCA+: E. coli-stimulated 54.6% 45.4% Preterm HCA+: Unstimulated 0.01% 99.98%

78

A E. coli 222 ng/uL

B E. coli 27 ng/uL C E. coli D Unstimulated 14 ng/uL 8 ng/uL

Reads mapped to E. coli genome 54.6% 9.5% 2.2% 0.02%

Figure 3.14 Bioanalyzer electropherogram summaries. (A-C) E. coli-stimulated samples from the preterm HCA+, term, and preterm HCA- infants respectively and (D) the unstimulated monocyte sample from the preterm HCA+ infant for comparison. The estimated RNA concentrations are presented for each sample in ng/L, with the corresponding proportions of reads that mapped to the E. coli genome below.

3.7 Discussion

The aim of these experiments was to design an in vitro system that allows for interrogation of the intrinsic monocyte transcriptional response to live E. coli or S. epidermidis and to compare this response between preterm and term infants. The optimisation of multiple cell culture and molecular conditions was required to achieve this aim as it relies on a number of variables: 1) obtaining highly pure monocyte samples, 2) minimising variation between bacterial stimulations, and 3) identifying an optimal stimulatory dose that induces detectable changes in gene expression without causing excessive cell death, while 4) generating RNA samples that meet the quality requirements for successful transcriptome sequencing.

The purification of monocytes from cord blood using negative selection techniques was complex due to the presence of a distinct population of lineage negative cells. Interestingly, limited characterisation of these cells revealed that the majority expressed HLA-DR but lacked surface expression of CD34, suggesting they were not stem cells. However a more thorough investigation may be required to rule out this possibility, as it has been shown that immature cord blood stem cells internally express CD34 prior to 79 external expression (292). Nevertheless, the presence of these cells directed development of the optimal monocyte purification method, as the only way to effectively exclude them was by positive selection of monocyte populations. Cell sorting using a combination of monocyte, T cell, B cell and NK cell lineage markers proved to be optimal (delivering purities >90%). However the sorting process itself appeared to induce monocyte activation, evidenced by increased inflammatory cytokine production at the gene and protein level in unstimulated cultures post-sort (highest at 5 hours). It is feasible that an artificially high level of inflammatory gene expression (induced by the purification process) in unstimulated samples may limit the power to detect differentially expressed genes following bacterial stimulation. Therefore monocytes were rested for 15 hours post- sort prior to bacterial stimulation as gene and protein levels of IL-6 and TNF were lowest at this time point. The practice of resting purified monocytes prior to bacterial stimulation is consistent with other studies using adult monocytes (293, 294).

The optimal stimuli for examining host-pathogen interactions are live bacteria. Not only are they the closest representation of infectious agents in vivo, they are more likely to elicit immune responses encompassing all forms of pattern recognition receptor interactions rather than discrete TLR pathways (e.g. LPS/TLR4) targeted by purified bacterial products. The E. coli strain selected for this study (ATCC 11775) contains the K1 capsular polysaccharide, which is strongly associated with neonatal infection and is a known virulence factor for bacteraemia through the degradation of complement proteins (80-83). S. epidermidis strain WT 1457 (a biofilm-forming strain isolated from a patient with an infected central venous catheter) has been used previously to investigate human neonatal innate immune responses and has also been used in a murine model of neonatal sepsis, making it an ideal representative strain (109, 193, 295). In order to minimise variability between bacterial stimulations (to effectively compare responses between infants) the use of frozen stocks generated from the same batch was an ideal alternative to cultivating fresh bacterial stocks for each stimulation (which practically occur over many days). Viability testing demonstrated that mid-logarithmic stocks of both E. coli and S. epidermidis survived well following long term storage at -80°C. Similar methods have been described for long term storage of other Gram-positive and Gram-negative species (251).

Having optimised methods for monocyte purification and bacterial cultivation and storage, the next aim was to determine the optimal stimulation dose of each pathogen.

80

Three doses of E. coli and S. epidermidis were tested with read-outs of inflammatory cytokine production, cell death and RNA quantity/purity. The middle dose for both pathogens proved to be optimal (MOI 10:1), by inducing high levels of IL-6 and TNF gene expression without causing a significant increase in either monocyte apoptosis or necrosis, and maintaining the RNA quality requirements for RNA-seq. Interestingly, the quantities of RNA extracted following stimulation with E. coli increased with dose and was significantly higher than PBS-stimulated monocytes at the highest dose. This could be partially due to increased monocyte transcription following stimulation, however it is more likely the result of contaminating bacterial RNA, supported by the fact that this observation was exclusive to E. coli and not S. epidermidis; Gram-negative species are more vulnerable to lysis than Gram-positives (296). This is further evidenced by the fact that a large proportion of transcripts from monocytes stimulated with E. coli mapped to the E. coli genome, an effect that correlated with the degree of E. coli contamination.

The primary aim of this thesis is to characterise the transcriptional response of infant monocytes to E. coli and S. epidermidis using RNA-seq by identifying and characterising genes that are differentially expressed following stimulation. Having an appropriate RNA-seq experimental design is critical to this aim, without which the benefits of using optimised cell culture and molecular methods are lost. RNA-seq is rapidly replacing traditional microarrays for transcriptome analysis, and even though currently there are no standardised guidelines for RNA-seq experimental design, a general consensus from the literature indicates that a sequencing depth of 10–20 million mapped reads is required for effective differential gene expression analysis. A preliminary RNA-seq experiment using cord blood monocytes from three infants (purified and stimulated using the optimised protocols) demonstrated that this depth of sequencing could easily be achieved when all nine samples were multiplexed and sequenced within the same lane. Importantly, this trial run also demonstrated the complexity of working with live pathogens and directed crucial changes to the experimental design in terms of E. coli-stimulated sample RNA inputs. Overall, the success of the preliminary RNA-seq experiment provides validation of the optimised methods presented in this chapter, and used in subsequent analyses.

81

82

Chapter 4 Monocytes from preterm infants are not intrinsically deficient, but are transcriptionally distinct following prenatal exposure to histologic chorioamnionitis

4.1 Introduction

Newborn infants are faced with significant immunological challenges during their transition from intra- to extra-uterine life, as they encounter a plethora of microorganisms for the first time. These challenges are greatest for those born prematurely, with an increasing prevalence of infection being a major consequence of preterm birth. Preterm infants are particularly susceptible to invasive infections, with E. coli and S. epidermidis being the leading causes of early- and late-onset sepsis in preterm infants, respectively (44, 56). Early-onset sepsis (occurs at <72 hours of age) is associated with high preterm infant mortality (up to 38% of E. coli cases) (44). In contrast, late-onset sepsis predominantly occurs between days 10 and 16 of life, and while the associated mortality is lower (1.5–10.2% of all cases) the overall incidence is several times higher than that of early-onset sepsis, affecting up to one third of extremely preterm infants (15, 59, 66). Early- and late-onset sepsis are also associated with considerable morbidity including increased risk of severe intraventricular haemorrhage, chronic lung disease and adverse neurological outcomes (69, 297). The immunological mechanisms underlying the heightened susceptibility of preterm infants to sepsis are not completely understood, but are critical for developing effective treatments and preventative interventions.

Preterm infants have deficient innate immune function, with particular deficits in monocyte responses (reviewed in section 1.4.2). The primary monocyte deficit appears to be in the significantly reduced expression of inflammatory cytokines following TLR stimulation (TNF, IL-1, IL-6 and IL-8) at the protein level, suggesting an underlying transcriptional deficiency (164, 185, 193, 209, 213). However, despite these clear deficits, preterm infant monocytes express comparable levels of core immune cell-surface receptors (e.g. TLRs) and have no clear deficits in the phosphorylation cascades downstream of the receptors (190, 192, 193). Therefore, a global transcriptional approach is likely best suited for elucidating the mechanisms underlying the differences between preterm and term infant immune function.

One major limitation of the literature to date is that very few studies have assessed the monocyte response(s) to live neonatal pathogens, and have instead focused on responses 83 to purified TLR agonists (LPS, R-848) or killed bacterial preparations. Furthermore, most studies have investigated preterm infant monocyte responses as part of mixed mononuclear cell cultures or within whole blood, rather than pure monocyte cultures. In addition, despite the fact that histologic chorioamnionitis (HCA) complicates 40–70% of preterm births, the consequences of exposure to HCA on human immune development are largely unknown, especially as placental histology is not routinely available and infants affected by clinical chorioamnionitis are actively excluded in many studies. This is the first study to use RNA-seq to assess the transcriptome of purified infant monocytes pre- and post-challenge with live, prototypical neonatal pathogens (E. coli and S. epidermidis) to identify deficiencies associated with preterm birth and the impact of HCA on monocyte function. This approach allows us to study the global, intrinsic monocyte transcriptional response following direct interactions with the most relevant preterm pathogens.

This chapter illustrates that preterm infants (± exposure to HCA) and term infant monocytes are not transcriptionally distinct based on assessment of the entire transcriptome. Using hypothesis-driven analyses we show that overall gene expression of TLR, inflammatory, antigen presentation and apoptosis pathway genes are similar between term and preterm infants (without exposure to HCA), indicating that preterm monocytes are not intrinsically deficient. In parallel assessments of monocyte culture supernatants we show high correlations between gene and protein expression of several inflammatory cytokines and chemokines, further demonstrating the competent functionality of preterm monocytes. Both targeted and unbiased expression analyses revealed that monocytes of preterm HCA+ infants displayed a significant hypo- responsive phenotype particularly in response to S. epidermidis, demonstrating that prenatal exposure to infection and inflammation may alter the risk for sepsis at least partly via modulation of monocyte responses. Finally, we show that while preterm infants exhibit higher frequencies of CD16+ monocytes in cord blood compared to term infants, this does not translate to an over-represented CD16+ monocyte gene signature.

84

4.2 Characterisation of the infants and samples used for RNA- seq

4.2.1 Demographic information Demographic information for the infants included in this study is presented in Table 4.1. Details for two cohorts are presented; the ‘Full’ cohort of infants that were subject to cord blood mononuclear cell (CBMC) and monocyte phenotyping by flow cytometry, and the ‘RNA-seq’ cohort of infants for which RNA-seq data was obtained. The ‘RNA-seq’ cohort is a subset of the ‘Full’ cohort as not all infants yielded adequate cell numbers for culture/sequencing. One term infant in the 'RNA-seq' cohort was subsequently excluded from downstream analysis due to sample contamination. A power calculation confirmed that the exclusion of one sample would only marginally decrease the sensitivity for differential expression analysis (see section 2.7.1 for methodology). Preterm infants were paired according to gestational age and birth weight. All preterm HCA- and term infants were born by cesarean section, and six out of seven preterm HCA+ infants were born by spontaneous vertex delivery (consistent with normal birth outcomes for these populations).

Table 4.1 Cohort demographics

Term Preterm infants Cohort infants HCA – HCA + p-value* Full (n=6) (n=9) (n=7) - Size RNA-seq (n=4) (n=6) (n=5) - Gestational age† Full 39 (± 1) 30.2 (± 0.4) 30 (± 1.3) ns (weeks) RNA-seq 38 (± 0.8) 30.3 (± 0.5) 30.4 (± 1.1) ns Birth weight† Full 3153 (± 372) 1511 (± 247) 1591 (± 93) ns (grams) RNA-seq 3103 (± 400) 1577 (± 114) 1595 (± 101) ns Full 3/3 7/2 4/3 ns Gender (M/F) RNA-seq 2/2 5/1 2/3 ns Mode of Full 6/0 9/0 1/6 0.0004 delivery RNA-seq 4/0 5/0 0/5 0.0016 (CS/SVD) *Statistical comparisons between preterm infant groups only using Chi squared test or Mann- Whitney test for continuous data; ns denotes non-significant (p>0.05). †Data presented as mean (± standard deviation). HCA, histologic chorioamnionitis; CS, cesarean section; SVD, spontaneous vertex delivery.

85

4.2.2 Phenotyping lymphocyte and monocyte populations A sample of CBMC from each infant was stained with monoclonal antibodies directed against CD14, CD16, CD56, CD3 and CD19 to identify major lymphocyte and monocyte populations (see section 2.4.2). Preterm HCA+ infants exhibited significantly decreased frequencies of B and T cells compared to preterm HCA- infants (Figure 4.1A) however this did not translate into significantly lower absolute counts (Figure 4.1D). All infant groups displayed similar frequencies and absolute counts of NK cells (Figure 4.1B and E). Preterm HCA- infants had significantly lower frequencies of total monocytes compared to preterm HCA+ and term infants (p<0.05, Figure 4.1C), which did not translate to significantly lower absolute counts of monocytes in preterm HCA- infants (Figure 4.1F).

86

A B C B and T cells NK cells Monocytes

100 100 100

C C C *

M

M M

B B B *

C

C C

l

l l

a

a a

t

t t

o * o o

t t t

f 10 f 10 f 10

o

o o

y

y y

c

c c

n n n

e

e e

u

u u

q

q q

e

e e

r r r

F F F 1 1 1 Preterm Preterm Term Preterm Preterm Term Preterm Preterm Term HCA- HCA+ HCA- HCA+ HCA- HCA+

D E F B and T cells NK c ells Mono cytes

10000 10000 10000

1000 1000 1000

L L

L

u u u / / /

s

s s

l

l l

l l l

e e

e

C C 100 C 100 100

10 10 10 Preterm Preterm Term Preterm Preterm Term Preterm Preterm Term HCA- HCA+ HCA- HCA+ HCA- HCA+

Figure 4.1 Proportions and absolute counts of lymphocyte and monocyte populations within CBMC. Flow cytometric analysis of CBMC samples was performed prior to cell sorting for monocyte purification . The frequencies of B and T cells combined (A), NK cells (B) and monocytes (C) with corresponding absolute counts within CBMC (D-F). Data are presented for individual donors (symbols) with bars showing mean ± SEM. Statistical analysis was performed using the Kruskal-Wallis test with Dunn’s multiple comparisons test. *p<0.05. Black data points indicate infants within the RNA-seq cohort, orange data points indicate infants exclusive to the Full cohort.

87

4.2.3 Phenotyping of monocyte subsets Further analysis of monocyte subsets based on their relative expression of CD14 and CD16 revealed that preterm infants have significantly lower frequencies of classical monocytes (CD14++ CD16-) and significantly increased frequencies of intermediate monocytes (CD14++ CD16+) compared to term infants (pre-sort analysis Figure 4.2D and E). Again, this did not translate to significant differences in absolute counts of each subset however the sample sizes may be underpowered to detect any differences (Figure 4.2A-C). Preterm HCA- infants also exhibited a significantly lower median fluorescence intensity (MFI) of CD14 on their classical monocytes compared to the other infant groups (Figure 4.3).

Differences in monocyte subset frequencies between preterm HCA- and term infants were also observed following analysis of purified monocytes from each infant (Figure 4.2G-I). These results confirmed that the composition of monocyte subsets was not changed following cell sorting. While cell sorting served to normalise differences in total monocyte frequencies, relative frequencies of the classical and intermediate subsets remained significantly different between preterm and term infants. No differences in the frequencies or absolute counts of non-classical monocytes were observed between infant groups (Figure 4.2C, F and I).

Additional staining for HLA-DR (MHC class II receptor involved in antigen presentation) on purified monocytes revealed no significant difference in the total frequency of HLA- DR+ monocytes or the intensity of HLA-DR staining on these cells between infant groups (Figure 4.4A and B). This result was unchanged when monocyte subsets were analysed individually (Figure 4.4C and D).

88

A B C Classical Intermediate Non-classical 1000 1000 1000

t

r

o 100 100 100

s

-

L L L

u u u

/ / /

e

s s s

l l l

r l 10 l 10 l 10

e e e

P C C C 1 1 1

0.1 0.1 0.1 Preterm Preterm Term Preterm Preterm Term Preterm Preterm Term HCA- HCA+ HCA- HCA+ HCA- HCA+

D E F Classical Intermediate Non-classical

s 100 s 100 ** s 100

e e e

t t t

t

y y y

c c c r *

o o o

n n n

o

o o o

s

m m 10 m 10

-

l l l

a a a

t t t e * o o o

t 50 t t

r

f f f

o o o

P y ** y 1 y 1

c c c

n n n

e e e

u u u

q q q

e e e

r r r

F 0 F 0.1 F 0.1 Preterm Preterm Term Preterm Preterm Term Preterm Preterm Term HCA- HCA+ HCA- HCA+ HCA- HCA+

G H I Classical Intermediate Non-classical s 100 s 100 ** s 100

e e e

t t t

y y y

t c c * c

o o o

r

n n n

o o o o

m m 10 m 10

s

l l l

-

a a a

t t t

t **

o o o

t 50 t t

f f f

s

o o o

o y y 1 y 1

c c c

n n n

P

e e e

u u u

q q q

e e e r r r F 0 F 0.1 F 0.1 Preterm Preterm Term Preterm Preterm Term Preterm Preterm Term HCA- HCA+ HCA- HCA+ HCA- HCA+

Figure 4.2 Proportions and absolute counts of infant monocyte subsets. Flow cytometric analysis of monocyte subsets pre- and post-sort. (A-C) Absolute counts of classical (CD14++ CD16-), intermediate (CD14++ CD16+) and non-classical (CD14+CD16+) monocyte subsets respectively pre-sort. Pre-sort (D-F) and post-sort (G-I) frequencies of each monocyte subset across infant groups. Data are presented for individual donors (symbols) with bars showing mean ± SEM. Statistical analysis was performed using the Kruskal-Wallis test with Dunn’s multiple comparisons test. *p<0.05, **p≤0.01. Black data points indicate infants within the RNA-seq cohort, orange data points indicate infants exclusive to the Full cohort.

Classical monocytes 20000 **** * * 15000

I

F

M

4 10000

1

D

C 5000

0 Preterm Preterm Term HCA - HCA +

Figure 4.3 Median fluorescence intensity of CD14 on classical monocytes. Flow cytometric analysis of the intensity of CD14 staining on infant classical monocytes (CD14++ CD16-

) was compared across infant groups. Data are presented for individual donors (symbols, duplicate measurements for each infant) with bars showing mean ± SEM. Statistical analysis was performed using the Kruskal-Wallis test with Dunn’s multiple comparisons test. *p<0.05, ****p≤0.0001. Black data points indicate infants within the RNA-seq cohort, orange data points indicate infants exclusive to the Full cohort.

89

A B Total monocytes Total monocytes 105 15000 * 100

+ R 10000

R

D 95 - D

-

A

L

A

L

H

I

H 90

F 5000

%

M 85 80 0 Preterm Preterm Term Preterm Preterm Term HCA- HCA+ HCA- HCA+

C

100

80

+

R 60

D

-

A

L

H 40

%

20 Preterm HCA- 0 Classical Interme diate Non-classical Preterm HCA+ Term

D

50000

40000

R

D 30000

-

A

L

H

I

F 20000 M 10000

0 Classical Interme diate Non-classical

Figure 4.4 HLA-DR staining on infant monocyte subsets. The frequencies of preterm and term infant monocytes expressing HLA-DR (A) and the corresponding MFI of HLA-DR on these cells (B). The frequencies of preterm and term classical (CD14++ CD16-), intermediate (CD14++ CD16+) and non-classical (CD14+ CD16+) monocytes expressing HLA-DR (C) and paired measurements of HLA-DR MFI on these cells (D). Data are presented for individual donors (symbols) with bars showing mean ± SEM. Statistical analysis was performed using the Kruskal-Wallis test with Dunn’s multiple comparisons test. *p<0.05, ***p≤0.001. Black data points indicate infants within the RNA-seq cohort, orange data points indicate infants exclusive to the Full cohort.

90

4.3 Bacterial stimulation induces a distinct monocyte transcriptional profile

Initial assessment of the raw RNA-seq data indicated that an average sequencing depth of 27 million reads was achieved across all samples (Figure 4.5A). On average, 88% of the raw reads from the unstimulated and S. epidermidis-stimulated samples mapped to the human genome (blue and grey data points, Figure 4.5B). Consistent with our observations in the trial run (see section 3.6.1.1), samples stimulated with E. coli had lower mapping statistics, due to residual contamination with bacterial RNA (average 77%, red data points Figure 4.5B). However all samples still achieved our predefined threshold of 10 million mapped reads, and there were no significant differences in mapping statistics between preterm and term infants (Figure 4.5C).

Unsupervised hierarchical clustering was performed on normalised RNA-seq data from all monocyte samples using Euclidean distance and the ‘complete’ linkage method to visualise global transcriptome similarities between samples. Hierarchical clustering measures dissimilarity between samples, therefore closely related samples cluster together. Monocytes stimulated with either E. coli or S. epidermidis formed a distinct cluster from unstimulated monocyte samples, however the two stimuli did not form distinct clusters within the larger ‘stimulation’ branch (Figure 4.6). Furthermore, within each major branch of the cluster dendrogram the infant groups did form unique clusters, suggesting that inter-infant variation was greater than inter-group variation. This was further supported by the pattern of clustering observed for E. coli-stimulated and S. epidermidis-stimulated monocyte samples from the same infant, which clustered closest together in eight out of fifteen individuals (infants 1, 4, 7, 8, 9, 10, 11 and 12; Figure 4.6).

Principal component analysis was performed to further investigate patterns of variation among the monocyte transcriptome profiles. Principal component analysis identifies strong patterns within the data, and arranges the samples to visualise the principle directions in which the data varies. The first three principle components (which accounted for the 76% of variation within the data) were compared. Consistent with the hierarchical cluster analysis, bacterial-stimulated monocytes separated from unstimulated monocytes based on the first and second principal components (Figure 4.7A). The E. coli- and S. epidermidis-stimulated monocyte samples separated from each other when the first and third principle components were compared, suggesting that differences between stimuli accounted for the third largest source of variation within the data (Figure 4.7B). 91

Comparisons of the second and third principle components did not yield relevant sample groupings (Figure 4.7C), but the overall distribution of samples suggested that the second principle component was likely to be related to inter-infant variability. However almost all S. epidermidis-stimulated monocyte samples were present in the top two quadrants, and all of the E. coli-stimulated samples were present in the bottom two quadrants, again supporting bacterial stimuli as an important source of variation in the data. Importantly, the principal component analysis showed no clear separation between preterm and term infants.

In contrast to previous data suggesting significant immune deficits in monocyte function between preterm and term infants (164, 186, 193, 212, 213), neither hierarchical clustering nor principal component analysis showed a clear distinction between groups. However, both clustering approaches have limitations based on their analysis of the full transcriptome, therefore minor trends in the data may be obscured. We next aimed to analyse the expression patterns of select genes involved in key pathways hypothesised to be different between preterm and term infants.

A B

4×107 100

7 s 80

d s 3×10

a

d

e

a

r

e 60

r

d

7 e

w 2×10

p

a

p

r

40

a

f

o

m

7 # 1×10

% 20

0 0 Preterm Preterm Term Preterm Preterm Term HCA- HCA+ HCA- HCA+

C

3×107

s

d

a

e

r

d 2×107

e

p

p

a

m

f 7

o

1×10

#

l

a

n

i

F 0 Preterm Preterm Term HCA- HCA+

Figure 4.5 RNA-seq read and mapping statistics. (A) Numbers of raw reads obtained from each sample. (B) The percentage of reads mapped to the human genome for each sample. (C) The final numbers of mapped reads for each sample where the dashed line represents the threshold required for differential expression analysis. Data are presented for individual samples (symbols) with bars showing mean ± SEM. Grey, orange and blue data points represent unstimulated, E. coli-stimulated and S. epidermidis-stimulated monocyte samples, respectively.

92

Figure 4.6 Cluster analysis of all monocyte samples. Unsupervised hierarchical clustering on normalised RNA-seq data revealed distinct clusters of bacterial- stimulated, and unstimulated monocytes. Infant groups did not form unique clusters within either major branch of the dendrogram, whereas E. coli- and S. epidermidis-stimulated monocytes from the same individual clustered together for 8 out of 15 infants. Individual infants are numbered; preterm HCA- infants 1–6, preterm HCA+ infants 7–11 and term infants 12–15. UN, unstimulated; EC, E. coli; SE, S. epidermidis; HCA, histologic chorioamnionitis.

93

Figure 4.7 Principal component analysis. (A) Comparison of principal components 1 and 2 revealed two distinct sample clusters; unstimulated and bacterial-stimulated monocytes. (B) The different stimuli were discernable as distinct sample clusters upon comparison of principal components 1 and 3. (C) This effect was still partially evident when principal components 2 and 3 were compared, but was much less distinct. UN, unstimulated; EC, E. coli; SE, S. epidermidis; HCA, histologic chorioamnionitis.

94

4.4 Hypothesis-driven interrogation of RNA-seq data

We defined a set of hypotheses based on available data on preterm and term infant monocyte responses to bacterial challenge (see section 1.4.2). Specifically, we hypothesised that compared to term infants, preterm HCA- infant monocytes: 1) Display diminished gene expression of TLR pathway receptors, adaptors, kinases and transcription factors (basal levels and post-stimulation). 1a) With reduced signalling through TLR pathways and diminished gene expression of inflammatory cytokines following stimulation with E. coli or S. epidermidis. 2) Exhibit decreased expression of genes involved in antigen processing and presentation following stimulation with E. coli or S. epidermidis. 3) Suppress stimulation-induced cell death by expressing an anti-apoptotic pattern of genes following stimulation with E. coli or S. epidermidis.

We also hypothesised that: 4) Patterns of gene expression in monocytes from preterm HCA+ infants would align more closely with that of term infant monocytes (through accelerated monocyte maturation due to prenatal exposure to inflammation).

To test these hypotheses, the expression values for seventy key genes in TLR, inflammatory, antigen presentation and apoptosis pathways were compared between preterm and term infants (basal levels and post-stimulation). This included a total of thirty-seven genes involved in TLR signalling; eight receptor genes (TLR1, TLR2, TLR4, TLR5, TLR6, TLR7, TLR8 and CD14), twenty-two downstream TLR pathway genes (TRAF6, TRAF3, TBK1, TAB2, TRIF, TRAM, TIRAP, ERK1, ERK2, NEMO, JNK, RIPK2, JUN, FOS, REL, IRAK2, IRAK4, MYD88, MAP2K4, MAP2K3, HSPD1 and HSPA1A) and seven TLR pathway transcription factor genes (IRF7, IRF3, NFKB1, EIF2AK2, NFKB1A, RELA and CREB1) (Figure 4.8). The expression of thirteen cytokine, chemokine and growth factor genes (IL1A, IL1B, IL6, IL8, TNF, CXCL10, LTA, CCL2, CCL3, PTGS2, CSF1, CSF2 and CSF3), nine genes involved in antigen presentation (CD40, CD74, CD80, CD86, CIITA, HLA-DRA, HLA-DRB1, HLA-DQB1 and HLA-DMB) and eleven apoptosis genes (pro-apoptotic: CASP1, CASP8, CASP9, NLRP3, BID, BIM, BAX, BAD, FADD and anti-apoptotic: MCL1, BCL-XL) were also compared (Figure 4.9).

The majority of genes included in our predefined hypotheses were up- or down-regulated following monocyte stimulation with either E. coli or S. epidermidis in all infant groups 95

(Figure 4.8 and Figure 4.9). Interestingly, genes involved in antigen processing and presentation showed a much higher degree of infant-to-infant variation with no clear up- or down- regulation following stimulation (Figure 4.9B). However, no clear separation between infant groups was observed in any group of genes.

We also performed differential expression analysis using a limma-Voom pipeline (see section 2.7.5 for methodology) on pairwise comparisons between monocytes from the three infant groups over each of the three culture conditions (unstimulated, E. coli- or S. epidermidis-stimulated). P-values were then extracted for each of the seventy genes relevant to our predefined hypotheses.

There were no significant differences between preterm HCA- and term infant monocytes in any of the seventy genes tested, under any culture condition. In contrast, preterm HCA+ infant monocytes displayed significantly reduced expression of twelve genes compared to either preterm HCA- or term infant monocytes, primarily among the S. epidermidis- stimulated samples (Figure 4.10). The significantly differentially expressed genes were from the TLR pathway (TLR4, TLR5, IRAK2, TKB1, NFKBIA), antigen presentation (CD74, HLA-DRA, CD86), apoptosis (NLRP3, CASP1) and cytokine (CSF2, CSF3) groupings.

In summary, we found no evidence to support our first three predefined hypotheses, as preterm and term infants expressed similar levels of key genes involved in TLR, inflammatory, antigen presentation, and anti-apoptosis/pro-apoptosis pathways. In contrast to our fourth hypothesis, preterm HCA+ infants were the only group to display significant differences in gene expression, which did not suggest a closer relationship to term infants. Finally, consistent with the findings from the hierarchical cluster and principal component analyses presented in the previous section, a large degree of inter- infant variability was observed across the analysed subset of genes.

96

A Toll-like receptors

15 Unstimulated EC-stimulated SE-stimulated

n TLR1

o

i

s TLR7

s 10

e r TLR6

p

x

e TLR2 5

e

n TLR8

e

g TLR4* 2 0

g

o TLR5*

L CD14 -5 Preterm Preterm Term Preterm Preterm Term Preterm Preterm Term HCA- HCA+ HCA- HCA+ HCA- HCA+

B TLR pathway genes TRAF6 MyD88 15 IRAK4 TIRAP Unstimulated EC-stimulated SE-stimulated n MAP2K4 o TBK1*

i

s

s 10 NEMO HSPD1

e

r

p JNK RIPK2

x

e 5 ERK2 HSPA1A

e

n MAP2K3 e ERK1

g

2 0 TAB2 JUN

g

o TRAF3 IRAK2

L * -5 TRAM FOS Preterm Preterm Term Preterm Preterm Term Preterm Preterm Term HCA- HCA+ HCA- HCA + HCA- HCA+ TRIF REL

C TLR transcription factors 15 Unstimulated EC-stimulated SE-stimulated

n

o IRF7

i

s s 10 IRF3

e

r

p NFKB1 x

e 5 EIF2AK2

e

n e NFKB1A*

g

2 0 RELA g

o

L CREB1 -5 Preterm Preterm Term Preterm Preterm Term Preterm Preterm Term HCA- HCA+ HCA- HCA+ HCA- HCA+

Figure 4.8 Normalised log2 gene expression values for key genes involved in TLR signalling. Expression of toll-like receptor genes (A), downstream TLR kinase and adaptor molecule genes (B) and TLR pathway transcription factor genes (C) were compared across infant groups within unstimulated, E. coli- and S. epidermidis-stimulated monocyte samples. Stars indicate genes that were differentially expressed by preterm HCA+ infants in at least one culture condition (see Figure 4.10). EC, E. coli; SE, S. epidermidis; HCA, histologic chorioamnionitis.

97

A Cytokines, chemokines and growth factors 15 Unstimulated

n CSF1 IL8

o

i

s CCL3 IL6

s 10

e r CCL2 IL1B

p

x

e LTA 5 IL1A

e

n CXCL10 CSF3*

e

g PTGS2 CSF2 2 0 * g TNF

o

L EC-stimulated SE-stimulated -5 Preterm Preterm Term Preterm Preterm Term Preterm Preterm Term HCA- HCA+ HCA- HCA + HCA- HCA+

B Antigen presentation genes 15 Unstimulated EC-stimulated SE-stimulated CD74*

n

o

i CIITA

s s 10 HLA-DMB

e

r

p

x HLA-DRA*

e 5

e HLA-DQB1

n

e CD40

g

2 0 HLA-DRB1

g

o L CD86* -5 CD80 Preterm Preterm Term Preterm Preterm Term Preterm Preterm Term HCA- HCA+ HCA- HCA+ HCA- HCA+

C Apoptosis genes 15 Unstimulated EC-stimulated SE-stimulated

n CASP1* BCL-XL

o

i

s BAX s 10 NLRP3*

e r BAD

p CASP9

x

e 5 CASP8 FADD

e

n MCL1

e

g

2 0 BID

g

o BIM

L -5 Preterm Preterm Term Preterm Preterm Term Preterm Preterm Term HCA- HCA+ HCA- HCA+ HCA- HCA+

Figure 4.9 Normalised log2 gene expression values for key monocyte function genes. Expression of cytokine, chemokine and growth factor genes (A), antigen processing and presentation genes (B) and apoptosis pathway genes (C) were compared across infant groups within unstimulated, E. coli- and S. epidermidis-stimulated monocyte samples. Stars indicate genes that were differentially expressed by preterm HCA+ infants in at least one culture condition (see Figure 4.10). EC, E. coli; SE, S. epidermidis; HCA, histologic chorioamnionitis.

98 Preterm HCA- Preterm HCA+ Term

IRAK2 TLR4 TLR5 TBK1 Preterm HCA- 15 10 6 8 *

n n n n * *

o

o o o i i 9 ** i i Preterm HCA+

s s s s

s s s 4 s 7

e e e * e

r 10 r r r

p p 8 p p x x x x Term

e * e * e e 2 6

e e e e

n n 7 n n

e

e e e

g

g 5 g g

2 2 2 0 2 5 IRAK2 TLR4 TLR5 TBK1 g g 6 g g o o o o 15 10 6 8

L L L L *

n n n n * *

o

o o o 0 5 -2 4 i i 9 ** i i

s s s s

Unstimulated SE-stim ulated EC-stimulated Unstimulated SE-stim ulated EC-stimulated Unstimulated SE-stim ulated EC-stimulated Unstimulated SE-stim ulated EC-stimulated s s s 4 s 7

e e e * e

r 10 r r r

p p 8 p p

x x x x

e * e * e e 2 6

e e e e

n n 7 n n

e e e e

g

g 5 g g

2 2 2 0 2 5 g g 6 g g

o o o o

L L L L NFKBIA CD74 HLA-DRA CD86 0 5 -2 4 Unstimulated SE-stim ulated EC-stimulated Unstimulated SE-stim ulated EC-stimulated Unstimulated SE-stim ulated EC-stimulated Unstimulated SE-stim ulated EC-stimulated 15 13 14 9

n n n * * n *

o

o o o i * i 12 * i i

s s s s

s s s s 8

e e e e r r * r 12 * r p p 11 p p

x x x x

e e e e 10 7 NFKBIA CD74 HLA-DRA CD86

e e e e

n n 10 n n e e e e 15 13 14 9

g

g g g 10

2 2 2 2

6 n n n n *

g g g g * *

o

o o o

9 i * i * i i o o o o 12

s s s s

L L L L

s s s s 8

e e e e r r * r 12 * r 5 8 8 5 p p 11 p p

x x x x

e e e e Unstimulated SE-stim ulated EC-stimulated Unstimulated SE-stim ulated EC-stimulated Unstimulated SE-stim ulated EC-stimulated Unstimulated SE-stim ulated EC-stimulated 10 7

e e e e

n n 10 n n

e e e e

g

g g g 10

2 2 2 2 6 g g 9 g g

o o o o

L L L L 5 8 8 5 Unstimulated SE-stim ulated EC-stimulated Unstimulated SE-stim ulated EC-stimulated Unstimulated SE-stim ulated EC-stimulated Unstimulated SE-stim ulated EC-stimulated NLRP3 CASP1 CSF3 (G-CSF) CSF2 (GM-CSF) 15 10 10 ** 10 *

n * * n n n o o * o o i i 9 i i

s s s s

s s s 5 s 5

e e e e r 10 r r r NLRP3 CASP1 CSF3 (G-CSF) CSF2 (GM-CSF)

p p 8 p p x x x x 15 10 10 10 * e e e e ** 0 0

e e e e

n n n n

n n n n * * 7 o o * o o i i i i

e e e e 9

s s s s

g 5 g g g

s s s 5 s 5

e e e e

2 2 2 2

-5 -5 r 10 r r r

g g g g 6 p p 8 p p

o o o o

x x x x

e e e e

L L L L 0 0

e e e e

0 5 -10 -10 n n 7 n n

e e e e

g 5 g g g

Unstimulated SE-stim ulated EC-stimulated Unstimulated SE-stim ulated EC-stimulated Unstimulated SE-stim ulated EC-stimulated Unstimulated SE-stim ulated EC-stimulated 2 2 2 -5 2 -5 g g 6 g g o o o o

L L L L 0 5 -10 -10 Figure 4.10 Genes differentially expressed by preterm HCA+ infant monocytes. Unstimulated SE-stim ulated EC-stimulated Unstimulated SE-stim ulated EC-stimulated Unstimulated SE-stim ulated EC-stimulated Unstimulated SE-stim ulated EC-stimulated Normalised log2 gene expression for the twelve genes found to be differentially expressed by preterm HCA+ infant monocytes compared to either preterm HCA- or term infant monocytes. Expression values for all culture conditions are shown. Statistical analysis was performed using the limma method for differential expression with p-values corrected for multiple comparisons using the Benjamini-Hochberg method. *p<0.05, **p≤0.01. EC, E. coli; SE, S. epidermidis; HCA, histologic chorioamnionitis.

99

4.5 Direct statistical comparisons of unstimulated, E. coli- or S. epidermidis-stimulated monocytes between infant groups

Analysis of the RNA-seq data using unsupervised hierarchical clustering, principle component analysis and hypothesis-dependent analyses did not provide evidence that monocytes from preterm and term infants were transcriptionally distinct. Therefore, differential gene expression analysis was performed on all pair-wise group comparisons within unstimulated, E. coli- or S. epidermidis-stimulated monocytes to determine if any genes were significantly differentially expressed between preterm and term infants (see section 2.7.5 for methodology). An overview of all comparisons performed is presented in Figure 4.11. This analysis yielded a total of nine primary comparisons, the results of which are summarised in Table 4.2, Table 4.6 and Table 4.9 for analysis of unstimulated, E. coli-stimulated and S. epidermidis-stimulated monocytes respectively. In addition, nine secondary comparisons were performed to identify any infant-group specific genes (eg. genes present in both primary comparisons involving term infants). Infant-group specific genes were designated to a representative biological process using the functional annotation tool available through the Database for Annotation, Visualization and Integrated Discovery (DAVID).

Primary comparisons between preterm HCA- and term infants consistently yielded the lowest numbers of differentially expressed genes across all culture conditions; 23 between unstimulated monocytes, 21 between E. coli-stimulated monocytes and only 2 between S. epidermidis-stimulated monocytes. In addition, no genes were identified when a more stringent p-value cut-off was used (p<0.001) for comparisons between E. coli- and S. epidermidis-stimulated monocytes (Table 4.6 and Table 4.9). In contrast, primary comparisons involving preterm HCA+ infants consistently resulted in the largest numbers of differentially expressed genes, with the greatest numbers observed between comparisons of S. epidermidis-stimulated monocytes; 262 and 765 genes were significantly differentially expressed compared to preterm HCA- and term infants respectively (Table 4.9). Not surprisingly, the greatest numbers of infant-group specific differentially expressed genes belonged to preterm HCA+ infants, again with the highest number resulting from comparisons of S. epidermidis-stimulated monocytes. Interestingly, 91 of the 159 differentially expressed genes were down-regulated (Table 4.10). 100

Unstimulated ! EC-stimulated! monocytes! monocytes!

1a. Preterm HCA- v Preterm HCA+! 1a. Preterm HCA- v Preterm HCA+!

1b. Term v Preterm HCA-! Table 4.2! 1b. Term v Preterm HCA-! Table 4.6! 1c. Term v Preterm HCA+! 1c. Term v Preterm HCA+!

2a. Genes overlapping between 1a and 1b = Preterm HCA- specific genes : Table 4.3! 2a. Genes overlapping between 1a and 1b = Preterm HCA- specific genes : No table*!

2b. Genes overlapping between 1a and 1c = Preterm HCA+ specific genes : Table 4.4! 2b. Genes overlapping between 1a and 1c = Preterm HCA+ specific genes : Table 4.7!

2c. Genes overlapping between 1b and 1c = Term specific genes : Table 4.5! 2c. Genes overlapping between 1b and 1c = Term specific genes : Table 4.8!

SE-stimulated! Unstimulated ! EC-stimulatedUnstimulatedmonocytes! !! EC-stimulated! monocytes! monocytes! monocytes! monocytes! 1a. Preterm HCA- v Preterm HCA+! 1a. Preterm HCA- v Preterm HCA+! 1a.1a. PretermPreterm HCA-HCA- vv PretermPreterm HCA+HCA+!! 1a. Preterm HCA- v Preterm HCA+! 1b. Term v Preterm HCA-! Table 4.9! 1b. Term v Preterm HCA-! 1b. Term v Preterm HCA-! Table 4.2! 1b. Term v Preterm HCA-! TableTable 4.24.6!! 1b. Term v Preterm HCA-! Table 4.6! 1c. Term v Preterm HCA+! 1c. Term v Preterm HCA+! 1c.1c. TermTerm vv PretermPreterm HCA+HCA+!! 1c. Term v Preterm HCA+! 2a. Genes overlapping between 1a and 1b = Preterm HCA- specific genes : No table*! 2a. Genes overlapping between 1a and 1b = Preterm HCA- specific genes : Table 4.3! 2a.2a. GenesGenes overlappingoverlapping betweenbetween 1a1a andand 1b1b == PretermPreterm HCA-HCA- specificspecific genesgenes :: TableNo table 4.3*!! 2a. Genes overlapping between 1a and 1b = Preterm HCA- specific genes : No table*! 2b. Genes overlapping between 1a and 1c = Preterm HCA+ specific genes : Table 4.10! 2b. Genes overlapping between 1a and 1c = Preterm HCA+ specific genes : Table 4.4! 2b.2b. GenesGenes overlappingoverlapping betweenbetween 1a1a andand 1c1c == PretermPreterm HCA+HCA+ specificspecific genesgenes :: TableTable 4.44.7!! 2b. Genes overlapping between 1a and 1c = Preterm HCA+ specific genes : Table 4.7! 2c. Genes overlapping between 1b and 1c = Term specific genes : Table 4.11! 2c. Genes overlapping between 1b and 1c = Term specific genes : Table 4.5! 2c.2c. GenesGenes overlappingoverlapping betweenbetween 1b1b andand 1c1c == TermTerm specificspecific genesgenes :: TableTable 4.54.8!! 2c. Genes overlapping between 1b and 1c = Term specific genes : Table 4.8!

SE-stimulated! SE-stimulated! monocytes! monocytes!

1a. Preterm HCA- v Preterm HCA+! 1a. Preterm HCA- v Preterm HCA+!

1b. Term v Preterm HCA-! 1b. Term v Preterm HCA-! Table 4.9! Table 4.9! 1c. Term v Preterm HCA+! 1c. Term v Preterm HCA+!

2a. Genes overlapping between 1a and 1b = Preterm HCA- specific genes : No table*! 2a. Genes overlapping between 1a and 1b = Preterm HCA- specific genes : No table*! 2b. Genes overlapping between 1a and 1c = Preterm HCA+ specific genes : Table 4.10! 2b. Genes overlapping between 1a and 1c = Preterm HCA+ specific genes : Table 4.10! 2c. Genes overlapping between 1b and 1c = Term specific genes : Table 4.11 ! 2c. Genes overlapping between 1b and 1c = Term specific genes : Table 4.11!

Figure 4.11 Diagramatic representation of the direct sample comparisons performed. Primary comparisons (1a, 1b and 1c) were performed to identify any genes that were differentially expressed (p<0.05) between infant groups within the same monocyte culture condition (unstimulated monocytes, E. coli-stimulated monocytes, or S. epidermidis-stimulated monocytes). Secondary comparisons (2a, 2b and 2c) were performed to identify the infant-group specific genes. The result tables corresponding to each comparison is indicated. *No table indicates that no genes were found to overlap between the specified primary comparisons. EC, E. coli; SE, S. epidermidis; HCA, histologic chorioamnionitis.

101

Table 4.2 Numbers of differentially expressed genes between unstimulated monocytes from each infant group Significance threshold Pair-wise group comparison p<0.05 p<0.01 p<0.001 Preterm HCA- vs. Preterm HCA+ 169 20 1 Term vs. Preterm HCA- 23 10 5 Term vs. Preterm HCA+ 50 9 1

Table 4.3 Differentially expressed genes (p<0.05) in unstimulated monocytes that are specific to preterm HCA- infants (2 genes) Expression relative to Significance (p-value) Representative Gene symbol Preterm HCA+ and Compared to Compared biological process Term infants Preterm HCA+ to Term Nucleotide R3HCC1 Up 6.35E-03 1.10E-02 binding GPX3 Metabolism Up 4.77E-02 3.83E-03

Table 4.4 Differentially expressed genes (p<0.05) in unstimulated monocytes that are specific to preterm HCA+ infants (17 genes)* Expression relative to Significance (p-value) Representative Gene symbol Preterm HCA- and Compared to Compared biological process Term infants Preterm HCA- to Term FKBP2 Protein folding Down 9.86E-04 4.80E-03 Regulation of XBP1 Down 1.29E-03 4.73E-02 transcription Neurotrophic MANF Down 1.80E-03 1.89E-02 factor PDIA3 Protein transport Down 3.27E-03 4.73E-02 OLFM1 Protein assembly Down 1.01E-02 3.91E-02 PEA15 Apoptosis Down 1.01E-02 1.14E-02 DHRSX Metabolism Down 2.06E-02 5.86E-03 ALDH1A1 Metabolism Down 4.12E-02 5.86E-03 PFKFB3 Metabolism Down 4.12E-02 4.82E-02 Signal MAP4K2 Up 2.93E-03 4.73E-02 transduction FCGR1A Immune response Up 4.53E-03 3.73E-02 Type IV MMP2 Up 2.23E-02 4.82E-02 collagenase LAMC1 Cell motility Up 2.71E-02 1.50E-02 IMPA2 Metabolism Up 3.82E-02 4.82E-02 *Two non-coding RNA genes and one pseudogene are not shown

102

Table 4.5 Differentially expressed genes (p<0.05) in unstimulated monocytes that are specific to term infants (7 genes)* Representative Expression relative Significance (p-value) Gene symbol biological to Preterm HCA- Compared to Compared to process and HCA+ infants Preterm HCA- Preterm HCA+ IGF2BP1 mRNA binding Down 6.73E-05 2.50E-06 NDRG2 Cytoskeleton Down 4.82E-02 1.42E-04 Regulation of WLS signal Up 1.45E-03 1.94E-02 transduction Signal NRG1 Up 2.89E-02 3.76E-02 transduction CLDN23 Cell adhesion Up 4.82E-02 6.79E-03 Regulation of NENF signal Up 4.88E-02 1.83E-03 transduction *One pseudogene is not shown

103

Table 4.6 Numbers of differentially expressed genes between E. coli-stimulated monocytes from each infant group Significance threshold Pair-wise group comparison p<0.05 p<0.01 p<0.001 Preterm HCA- vs. Preterm HCA+ 43 3 0 Term vs. Preterm HCA- 21 0 0 Term vs. Preterm HCA+ 395 36 2

Table 4.7 Differentially expressed genes (p<0.05) in E. coli-stimulated monocytes that are specific to preterm HCA+ infants (18 genes)* Expression relative Significance (p-value) Representative Gene symbol to Preterm HCA- Compared to Compared biological process and Term infants Preterm HCA- to Term AGPAT6 Metabolism Down 4.55E-03 1.69E-03 FKBP2 Protein folding Down 1.62E-02 1.72E-02 mRNA SRSF5 Down 1.62E-02 5.40E-03 processing G-protein coupled GPR31 Down 2.33E-02 1.30E-02 receptor PDIA3 Protein transport Down 3.05E-02 7.62E-03 EPB41L3 Cytoskeleton Down 4.49E-02 7.08E-03 IDI1 Metabolism Down 4.49E-02 2.16E-02 MPHOSPH8 Methylation Down 4.49E-02 3.50E-02 IL13RA1 Immune response Down 4.57E-02 5.40E-03 MYADM Cell membrane Up 1.25E-02 3.27E-02 MGAT4B Metabolism Up 1.55E-02 1.72E-02 ZDHHC18 Ion binding Up 1.62E-02 2.13E-03 EIF6 Translation Up 2.87E-02 2.05E-03 ANP32A Multifunctional Up 3.21E-02 4.87E-02 KCNAB2 Ion transport Up 3.21E-02 3.54E-02 *One pseudogene is not shown

104

Table 4.8 Differentially expressed genes (p<0.05) in E. coli-stimulated monocytes that are specific to term infants (10 genes)* Representative Expression relative Significance (p-value) Gene symbol biological to Preterm HCA- Compared to Compared to process and HCA infants Preterm HCA- Preterm HCA+ IGF2BP1 mRNA binding Down 1.72E-02 6.35E-04 MED13 Transcription Up 1.17E-02 3.07E-03 SYNPO2 Cytoskeleton Up 1.72E-02 6.35E-04 Immune LAIR1 Up 3.60E-02 4.05E-03 response Regulator of SMAP2 Up 4.01E-02 3.43E-02 GTPase activity MPEG1 Uncharacterised Up 4.03E-02 3.71E-02 Regulation of PTEN signal Up 4.39E-02 1.06E-02 transduction Regulation of WLS signal Up 4.39E-02 1.69E-03 transduction ZFYVE26 Ion binding Up 4.39E-02 4.88E-02 *One non-coding RNA gene is not shown

105

Table 4.9 Numbers of differentially expressed genes between S. epidermidis- stimulated monocytes from each infant group Significance threshold Pair-wise group comparison p<0.05 p<0.01 p<0.001 Preterm HCA- vs. Preterm HCA+ 262 34 1 Term vs. Preterm HCA- 2 1 0 Term vs. Preterm HCA+ 765 173 24

Table 4.10 Differentially expressed genes (p<0.05) in S. epidermidis-stimulated monocytes that are specific to preterm HCA+ infants (159 genes)* Expression relative to Significance (p-value) Representative Gene symbol Preterm HCA- and Compared to Compared biological process Term infants Preterm HCA- to Term AGPAT6 Cell development Down 3.16E-03 8.08E-04 IL13RA1 Immune response Down 4.80E-03 1.69E-03 Endosome/ PLEKHF2 Down 5.56E-03 8.07E-04 apoptosis FYN Immune response Down 5.58E-03 2.85E-02 FKBP2 Protein folding Down 5.79E-03 2.46E-02 CHORDC1 Ion binding Down 7.28E-03 7.92E-03 DNAJA4 Protein binding Down 8.11E-03 1.85E-02 CMIP T-cell signalling Down 8.21E-03 3.33E-02 ANKHD1- Uncharacterised Down 8.51E-03 1.42E-02 EIF4EBP3 ANKHD1 Uncharacterised Down 8.76E-03 1.66E-02 EPB41L3 Cytoskeleton Down 9.06E-03 3.27E-03 IDI1 Metabolism Down 9.96E-03 1.04E-02 Regulation of SCML1 Down 1.09E-02 8.33E-03 transcription UBB Apoptosis Down 1.16E-02 1.03E-03 HNRNPA3 RNA splicing Down 1.30E-02 2.66E-02 TMEM2 Uncharacterised Down 1.30E-02 3.08E-03 NLRP3 Immune response Down 1.35E-02 2.60E-02 ADIPOR2 Multifunctional Down 1.44E-02 2.40E-02 GPR132 Cell cycle Down 1.44E-02 2.41E-02 MN1 Uncharacterised Down 1.44E-02 3.70E-03 DNAJB1 Protein binding Down 1.47E-02 1.41E-02 RFC2 DNA replication Down 1.47E-02 1.79E-03 Chromatin ING3 Down 1.51E-02 1.74E-02 organisation RCSD1 Uncharacterised Down 1.51E-02 4.64E-02 DNAJB6 ATPase activity Down 1.53E-02 1.69E-03 SIPA1L3 Uncharacterised Down 1.56E-02 4.11E-02 ANO6 Ion binding Down 1.72E-02 7.59E-03 HLA-DRA Immune response Down 1.72E-02 3.84E-02 PLXNA2 Cell motility Down 1.72E-02 7.92E-03

106

RAB33A GTP-binding Down 1.72E-02 1.79E-03 TXNDC16 Uncharacterised Down 1.72E-02 2.02E-02 WNT5A WNT-signalling Down 1.72E-02 3.48E-03 PDIA3 Protein transport Down 1.78E-02 1.41E-02 Regulation of RFX5 Down 1.78E-02 4.09E-02 transcription Regulation of XBP1 Down 1.78E-02 2.61E-02 transcription Amino-acid SLC36A4 Down 1.79E-02 1.44E-02 transport ADAM28 Multifunctional Down 1.85E-02 9.34E-03 Regulation of TAF7 Down 1.85E-02 3.09E-02 transcription CD4 Immune response Down 1.88E-02 2.14E-02 HNRNPA2B1 RNA splicing Down 1.90E-02 2.03E-02 Regulation of DDX5 Down 1.94E-02 5.39E-03 transcription LRRD1 Uncharacterised Down 1.94E-02 1.41E-02 PDSS1 Metabolism Down 1.94E-02 3.23E-03 SPCS2 Uncharacterised Down 1.94E-02 1.41E-02 CYP51A1 Metabolism Down 2.02E-02 1.58E-02 HMGCS1 Metabolism Down 2.02E-02 1.68E-02 UBAC2 Uncharacterised Down 2.02E-02 3.78E-02 WWC3 Uncharacterised Down 2.02E-02 1.85E-02 Regulation of ZNF33A Down 2.06E-02 1.69E-03 transcription ELOVL5 Metabolism Down 2.08E-02 1.63E-02 RBM23 RNA processing Down 2.08E-02 1.13E-02 SRSF5 RNA splicing Down 2.08E-02 3.48E-03 Mg(2+) MAGT1 Down 2.18E-02 5.04E-03 transporter Cell proliferation/ BTN2A2 Down 2.19E-02 1.69E-03 development OGDH Metabolism Down 2.24E-02 7.59E-03 Regulation of HES1 Down 2.35E-02 9.13E-03 transcription AHRR Apoptosis Down 2.37E-02 2.49E-02 FHAD1 Uncharacterised Down 2.41E-02 2.97E-02 Heat-shock HSP90AA1 Down 2.76E-02 4.55E-02 response Regulation of ZNF136 Down 2.97E-02 5.60E-03 transcription ABR Apoptosis Down 3.22E-02 5.39E-03 FOXO3 Apoptosis Down 3.24E-02 5.04E-03 NFKBIL1 Immune response Down 3.24E-02 7.94E-03 C6orf223 Uncharacterised Down 3.33E-02 1.71E-02 SFR1 Uncharacterised Down 3.33E-02 3.08E-03 Regulation of RNF141 Down 3.58E-02 1.30E-02 transcription 107

C1orf61 FOS signalling Down 3.66E-02 2.02E-02 MAPK1IP1L Uncharacterised Down 3.66E-02 3.26E-02 EIF1 Translation Down 3.69E-02 3.21E-02 PDGFB Growth factor Down 3.69E-02 3.56E-02 MYO9B Metabolism Down 3.73E-02 6.07E-03 UBQLN2 Proteolysis Down 3.84E-02 8.53E-03 Heat-shock HSPA6 Down 3.85E-02 3.21E-02 response BIRC2 Apoptosis Down 3.91E-02 4.74E-03 WTAP RNA processing Down 3.98E-02 9.29E-03 PNRC1 Transcription Down 4.02E-02 1.93E-02 CLIP2 Cytoskeleton Down 4.04E-02 2.60E-02 SAV1 Apoptosis Down 4.04E-02 3.46E-03 TNS3 Actin-binding Down 4.05E-02 3.52E-02 TP53INP1 Apoptosis Down 4.12E-02 3.33E-02 TBK1 Immune response Down 4.31E-02 1.66E-02 NADH MT-ND2 Down 4.47E-02 1.66E-02 dehydrogenase EDEM1 Proteolysis Down 4.49E-02 3.08E-03 Regulation of ELL3 Down 4.49E-02 2.40E-02 transcription KANK1 Cytoskeleton Down 4.49E-02 1.99E-02 NADH MT-ND4 Down 4.49E-02 2.11E-02 dehydrogenase MTRNR2L8 Uncharacterised Down 4.49E-02 2.95E-02 SLC4A7 Ion transport Down 4.49E-02 3.70E-02 TCTEX1D4 TGF-β signalling Down 4.49E-02 4.01E-02 SELK Immune response Down 4.53E-02 2.60E-02 Regulation of ATF6 Down 4.69E-02 1.43E-02 transcription UBXN7 Ubiquitin binding Up 4.31E-03 8.08E-04 Insulin/growth- GRB10 Up 4.80E-03 3.52E-02 factor signalling Myeloid MYADM Up 5.56E-03 3.46E-02 differentiation MGAT4B Metabolism Up 5.79E-03 1.41E-02 TXNL4A Cell cycle Up 7.28E-03 2.46E-02 Nucleic-acid HNRNPUL2 Up 1.09E-02 3.51E-03 binding WIPI2 Multifunctional Up 1.51E-02 3.56E-02 EIF6 Translation Up 1.72E-02 4.74E-03 FAM134A Uncharacterised Up 1.72E-02 1.34E-02 PHF23 Ion binding Up 1.72E-02 3.70E-02 DDHD2 Vesicle transport Up 1.85E-02 1.22E-02 PFKP Metabolism Up 1.88E-02 1.41E-02 ZDHHC18 Ion binding Up 1.88E-02 3.70E-03 PEX26 Protein transport Up 1.90E-02 3.78E-02 PRR12 DNA binding Up 2.02E-02 7.69E-03

108

Endoprotease PCSK6 Up 2.06E-02 2.28E-02 activity LAMC1 Cell motility Up 2.08E-02 3.08E-03 RNA binding/ PPP1R8 Up 2.22E-02 1.41E-02 degradation RMI1 DNA replication Up 2.58E-02 3.15E-02 ZBED5 Ion binding Up 3.24E-02 2.66E-02 GNPAT Metabolism Up 3.68E-02 3.79E-03 PEX19 Protein transport Up 4.04E-02 1.41E-02 ATRAID Apoptosis Up 4.18E-02 3.29E-02 IL17RA Immune response Up 4.31E-02 3.60E-02 ATG14 Autophagy Up 4.47E-02 4.26E-02 TACO1 COX1 translation Up 4.48E-02 2.84E-02 ANP32A Apoptosis Up 4.49E-02 7.92E-03 CMAS Metabolism Up 4.49E-02 2.23E-02 SUMF1 Sulfatase catalyst Up 4.49E-02 1.23E-02 TEX264 Uncharacterised Up 4.49E-02 3.25E-02 TKT Metabolism Up 4.95E-02 3.77E-03 GPI Metabolism Up 4.96E-02 3.08E-03 *Twelve pseudogenes and twenty-four non-coding RNA genes are not shown

Table 4.11 Differentially expressed genes (p<0.05) in S. epidermidis-stimulated monocytes that are specific to term infants (1 gene) Expression relative to Significance (p-value) Representative Gene symbol Preterm HCA- and Compared to Compared to biological process HCA+ infants Preterm HCA- Preterm HCA+ IGF2BP1 mRNA binding Down 2.30E-02 6.87E-04

4.5.1 Comparisons across culture conditions

The lists of infant-group specific differentially expressed genes were then compared across culture conditions to identify genes that were consistently associated with a single infant group. These genes may be critically related to differences in immune capacity between monocytes from the preterm HCA-, preterm HCA+, and term infant groups. This comparison was not performed for preterm HCA- infants as there were no differentially expressed genes specific to preterm HCA- monocytes following stimulation with either pathogen. Culture-wide comparisons of differentially expressed genes specific to preterm HCA+ infants revealed two genes were differentially expressed in both unstimulated and S. epidermidis-stimulated monocytes; XBP1 and LAMC1 with decreased and increased expression respectively (relative to preterm HCA-

109 and term infants). This pattern of gene expression was also observed for E. coli- stimulated monocytes, but was not significantly different (p>0.05) compared to both preterm HCA- and term infants. Ten genes were differentially expressed in response to both stimuli; five with increased expression (MYADM, MGAT4B, ZDHHC18, EIF6, ANP32A) and five with decreased expression (AGPAT6, SRSF5, IDI1, EPB41L3, IL13RA1) relative to preterm HCA- and term infants. Again, the same pattern of expression was observed for these ten genes in unstimulated monocytes, but did not reach significance compared to both preterm HCA- and term infants (p>0.05). Two genes were consistently differentially expressed across all culture conditions; FKBP2 and PDIA3 both with decreased expression within preterm HCA+ infant monocytes.

Comparisons of term-infant specific differentially expressed genes revealed one gene was differentially expressed in both unstimulated and E. coli-stimulated term infant monocytes (WLS; up-regulated by term infant monocytes), which was also up-regulated in response to S. epidermidis, but was not significantly different compared to both preterm HCA- and HCA+ infants. The only gene that was specifically differentially expressed in S. epidermidis-stimulated monocytes (IGF2BP1) was also differentially expressed in the other culture conditions (consistently down-regulated by term infant monocytes).

110

4.6 Comparisons of the E. coli or S. epidermidis-induced monocyte transcriptional response between infant groups

A diagrammatic representation of the analyses performed in this section are presented in Figure 4.12, where a combination of statistical and bioinformatic approaches were used to characterise the infant-group specific monocyte response to bacterial stimulation.

A! 1. Identify genes differentially expressed following monocyte stimulation!

EC response! SE response!

Comparison between groups: Comparison between groups: Figure 4.13! Figure 4.14!

Identification of DEG unique to each infant group!

Genes exclusively up- or down-regulated Compared to those identified in section 4.4: by each infant group: Table 4.12! Table 4.13!

B! 2. Bioinformatic analysis of DEG!

Over-represented canonical Over-represented diseases or Upstream regulator analysis! Network analysis! pathways! bio-functions! ! !

EC response! EC response! EC response ! EC response all groups : Table 4.20! Preterm HCA- genes: No results! Preterm HCA- genes: Table 4.15! Top 2000 DEG all groups: ! SE response all groups: Table 4.21! Preterm HCA+ genes: No results! Preterm HCA+ genes: No results! Table 4.16! Term genes: Table 4.14! Term genes: No results! Figure 4.15 and Figure 4.16! ! ! ! SE response! SE response! Unique DEG all groups: ! Preterm HCA- genes: No results! Preterm HCA- genes: Table 4.15! Table 4.17! Preterm HCA+ genes: No results! Preterm HCA+ genes: No results! Figure 4.17–Figure 4.19! Term genes: No results! Term genes: No results! ! SE response! Top 2000 DEG all groups: ! Table 4.18 ! Figure 4.20 and Figure 4.21! ! Unique DEG all groups: ! Table 4.19! Figure 4.22!

C! 3. Delta-fold change analysis!

# of DEG that are significantly differentially expressed between groups: Table 4.22 !

Fold change for significant genes: Figure 4.23 !

Figure 4.12 Diagrammatic representation of statistic and bioinfomatic analyses performed. (A) Infant-group comparisons of genes differentially expressed following monocyte stimulation with E. coli or S. epidermidis. (B) Bioinformatic analyses performed on the lists of differentially expressed genes unique to each infant group. (C) Delta-fold change analysis on all genes to identify those that were significantly differentially expressed between infant groups in response to E. coli or S. epidermidis. DEG, differentially expressed genes; EC, E. coli; SE, S. epidermidis.

111

4.6.1 Differential expression analysis

In the previous section, differential expression analysis was used to directly compare unstimulated monocytes, or those stimulated with E. coli or S. epidermidis. However this does not directly identify differences in expression induced by the stimuli relative to the resting state. Therefore differential expression analysis was used to determine the sets of genes that were differentially expressed in response to E. coli or S. epidermidis for each infant group (see section 2.7.5 for methodology). The gene lists generated were then compared to identify stimulus-induced transcriptional changes specific to monocytes from each infant group. This chapter focuses on the differences between infants groups. A comparison of E. coli and S. epidermidis responses is presented in Chapter 5.

E. coli-stimulation induced the statistically significant differential expression (p<0.05) of 8857, 8755 and 8495 genes by monocytes from preterm HCA-, preterm HCA+ and term infants respectively. Comparably high numbers of significantly differentially expressed genes (p<0.05) were also observed in S. epidermidis-stimulated monocytes, with 9157, 9094, and 8778 genes differentially expressed by monocytes from preterm HCA-, preterm HCA+ and term infants respectively. For both stimuli, the vast majority of these genes were consistently differentially expressed by all groups; 7461 genes in response to E. coli and 7850 genes in response to S. epidermidis (Figure 4.13A and Figure 4.14A, respectively). On average less than 5% of differentially expressed genes were unique to each infant group. In response to E. coli this equated to 412 genes for preterm HCA- infants, 372 genes for preterm HCA+ infants and 405 genes for term infants (Figure 4.13A). In response to S. epidermidis there were 364 genes for preterm HCA- infants, 480 genes for preterm HCA+ infants and 291 genes for term infants (Figure 4.14A). Interestingly, preterm HCA- and preterm HCA+ showed the highest degree of overlap in differentially expressed genes following stimulation with either bacterium (Figure 4.13A and Figure 4.14A).

Heatmaps were generated to visualise the log2 fold-change expression values for uniquely differentially expressed genes, comparing all infant groups (Figure 4.13B-D and Figure 4.14B-D for E. coli and S. epidermidis responses, respectively). Interestingly, in response to both stimuli, the majority of genes (>90%) that were statistically-determined to be unique, actually showed a highly similar pattern of expression across all infant groups, indicating a potential statistical artefact. Indeed, 112 only a small number of genes (1–16) were expressed in the opposite direction by each infant group. These genes (summarised in Table 4.12) were of particular interest as they may indicate points of differential monocyte regulation in response to bacterial stimulation between infant groups.

Unsupervised hierarchical clustering was also performed on all E. coli- or S. epidermidis-stimulated monocyte samples based on the collection of differentially expressed genes unique to each infant group, totalling 1189 genes in response to E. coli, and 1135 genes in response to S. epidermidis. This analysis however, did not result in separation of the infant groups into distinct clusters for either stimulus, suggesting this gene list does not represent a robust infant group-specific gene signature (Figure 4.13E and Figure 4.14E for E. coli and S. epidermidis, respectively).

113

Figure 4.13 Differentially expressed genes following monocyte stimulation with E. coli. (A) Venn diagram representing the number of differentially expressed genes (DEG) common and unique to each infant group. Heatmaps visualising the normalised log2 fold change in gene expression for all DEG unique to preterm HCA- infants (B), preterm HCA+ infants (C) and term infants (D). Unsupervised heirarchical clustering of all E. coli-stimulated monocyte samples based on the collective lists of DEG unique to each infant group (E). HCA, histologic chorioamnionitis; FC, fold change.

114

Figure 4.14 Differentially expressed genes following monocyte stimulation with S. epidermidis. (A) Venn diagram representing the number of differentially expressed genes (DEG) common and unique to each infant group. Heatmaps visualising the normalised log2 fold change in gene expression for all DEG unique to preterm HCA- infants (B), preterm HCA+ infants (C) and term infants (D). Unsupervised heirarchical clustering of all S. epidermidis-stimulated monocyte samples based on the collective lists of DEG unique to each infant group (E). HCA, histologic chorioamnionitis; FC, fold change.

115

Table 4.12 Genes expressed in the opposite direction by monocytes from each infant group Monocyte response to EC Genes

Increased: DUSP12, C6orf211, TPK1, GATC, ZNF814, RAE1, RGL4, SRRD Unique to preterm HCA-

infants Decreased: BCL2L11, MIER3, FBXO30, SCRN1, RP11-77K12.1, YPEL2

Increased: HDAC1, COX7A2L, UGDH, INPP5E, NUDT14, ATP5H, Unique to preterm HCA+ TPGS2 infants Decreased: CCNH, DTX4, SC5D

Increased: EGLN1, HLX, SOS1, XRN1, RSRC1, C3orf38, ETFDH, MAT2B, MYLIP, GPR107, MARCH5, UBAC2, KCTD12, ARRDC4, Unique to term infants FMR1, GAB3

Decreased: UBIAD1, SLC11A1, CLPTM1L, INO80, FXR2, KXD1

Monocyte response to SE Genes

Increased: RAB28, RANBP9, SLC18A2, RELT, CARHSP1 Unique to preterm HCA-

infants Decreased: STARD10

Increased: TAF1A, CNST, PSMD1, QPCT, SLC25A36, SLC51A, YIPF5, SFXN1, ZC3HC1, FAM109A, BAHD1, ANKRD40, C19orf12, GPI, OLIG2, PRPS1 Unique to preterm HCA+

infants Decreased: ADAMTSL4, RCSD1, ANKRD44, KIAA1109, EIF4EBP3, C6orf211, MAN1A1, OXR1, TP53INP1, C11orf71, EML3, MEFV, BAIAP2, RTTN, MYO9B

Increased: EGLN1, NIPAL1, RGS10, UBXN1, TMEM9B, CAMKK2, GATM, GAA, C19orf25 Unique to term infants Decreased: PELO, CPSF1, STOML2, MGST1, COPS7A, MRPL42, LOH12CR1, LARP4

EC, E. coli; SE, S. epidermidis; HCA, histologic chorioamnionitis.

116

We next sought to compute the overlap between the stimulus-induced differentially expressed genes above, with the infant-group specific differentially expressed genes identified in section 4.5. Genes identified in both analyses are more likely to represent true infant-group specific genes as they would be differentially regulated upon stimulation and differentially expressed between monocytes at rest or following stimulation. Table 4.13 summarises the genes that were identified as significantly differentially expressed in both analyses; MMP2, MANF, MYADM, HSPA6, CMIP, SIPA1L3, RCSD1, TP53INP1, PCSK6, GPI, SIPA1L3, MYO9B, MYADM specific to preterm HCA+ infants, and NENF and PTEN specific to term infants.

Table 4.13 Overlap between genes significantly differentially expressed in response to stimulation and those identified in direct sample comparisons (section 4.5) Preterm HCA- specific DEG DEG unique to preterm Unstimulated HCA- infants monocytes EC-stimulated SE-stimulated (Table 4.3) monocytes monocytes EC response (412 genes) 0 - - SE response (364 genes) 0 - - Preterm HCA+ specific DEG DEG unique to preterm Unstimulated EC-stimulated SE-stimulated HCA+ infants monocytes monocytes monocytes (Table 4.4) (Table 4.7) (Table 4.10) 3 genes (HSPA6, CMIP, EC response (372 genes) 1 gene (MMP2) 0 SIPA1L3) 7 genes (RCSD1, TP53INP1, PCSK6, SE response (480 genes) 1 gene (MANF) 1 gene (MYADM) GPI, SIPA1L3, MYO9B, MYADM) Term specific DEG DEG unique to term Unstimulated EC-stimulated SE-stimulated infants monocytes monocytes monocytes (Table 4.5) (Table 4.8) (Table 4.11) EC response (406 genes) 1 gene (NENF) 1 gene (PTEN) 0 SE response (291 genes) 0 0 0 DEG, differentially expressed gene; EC, E. coli; SE, S. epidermidis; HCA, histologic chorioamnionitis.

117

4.6.2 Bioinformatic analysis of differentially expressed genes 4.6.2.1 Over-represented canonical pathways, diseases and bio-functions To further characterise the infant group-specific monocyte response to E. coli or S. epidermidis, Ingenuity Pathway Analysis® (IPA) was performed to assess the biological significance of the unique gene lists identified in section 4.6.1. Using the Ingenuity knowledge base as a reference, IPA analyses canonical pathways, disease classifications and bio-functions that are significantly over-represented in each list of differentially expressed genes.

There were no significantly over-represented canonical pathways in the uniquely differentially expressed genes from any infant group following monocyte stimulation with S. epidermidis, or within the preterm HCA- or HCA+ monocyte response to E. coli. Four canonical pathways were significantly over-represented in the term infant monocyte response to E. coli (Table 4.14). These canonical pathways were categorised as significant based on the differential expression of 12–17 genes, eight of which were present in three or four of these pathways (up-regulated: GNB4, HDAC6, PLCB1, PPP3CB, SOS1; down-regulated: HDAC2, PPP3R1, PRKCA).

IPA did not detect any disease classifications or bio-functions significantly over- represented within the E. coli or S. epidermidis monocyte response unique to preterm HCA+ or term infants.

In contrast, two disease classifications were over-represented within the unique preterm HCA- infant monocyte response to E. coli; “Neurological Disease” and “Hereditary Disorder” (Table 4.15). Both disease classifications included the same nine genes (up- regulated: HSPD1, MARS2, PMP22, VPS37A; down-regulated: L1CAM, LMNA, NT5C2, PRPS1, SLC12A6). In addition, five disease classifications and bio-functions were over-represented in the unique preterm HCA- infant monocyte response to S. epidermidis (Table 4.15), the most significant of which was “RNA Post-transcriptional Modification” based on the differential expression of 14 genes (up-regulated: DHX38, GEMIN7, PRPF19, RSRC1; down-regulated: INTS12, LSM3, PIN1, PLRG1, SF3A3, SNRPC, SPI1, USB1, USP49, WDR83). Three of the remaining over-represented disease classifications and bio-functions (“Tumor Morphology”, “Cell-to-Cell Signalling and Interaction” and “Cellular Compromise”) were of similar significance and included the same five genes (up-regulated: PTK2, SELL; down-regulated: AKT2, 118

AMICA1, THBS1). The bio-function “Cell Cycle” was called over-represented based on the differential expression of ten genes (up-regulated: CREG1, DTD1, MXD1, PLK2, RCC2, TFDP1; down-regulated: BAP1, ESRRA, FZR1, RB1CC1).

Table 4.14 Significantly over-represented canonical pathways associated with the unique term infant monocyte response to E. coli p-score* Canonical pathway Preterm HCA- Preterm HCA+ Term unique unique genes unique genes genes Role of NFAT in Cardiac ns ns 1.5 Hypertrophy T Huntington's Disease ns ns 1.4 Signalling T Phospholipase C Signalling ns ns 1.4 Protein Kinase A Signalling ns ns 1.4

*p-score = -log10 (p-value). Values above 1.3 are considered statistically significant (FDR adjusted p-value <0.05). HCA, histologic chorioamnionitis; ns, non-significant.

Table 4.15 Significantly over-represented diseases or bio-functions associated with the unique preterm HCA- infant monocyte response to E. coli or S. epidermidis p-score* Disease or Bio-function Preterm HCA- Preterm HCA+ Term unique (EC response) unique genes unique genes genes Hereditary Disorder 1.4 ns ns Neurological Disease 1.4 ns ns p-score* Disease or Bio-function Preterm HCA- Preterm HCA+ Term unique (SE response) unique genes unique genes genes RNA Post-Transcriptional 3.2 ns ns Modification Cell-To-Cell Signalling and 1.7 ns ns Interaction Cellular Compromise 1.7 ns ns Tumor Morphology 1.7 ns ns Cell Cycle 1.7 ns ns

*p-score = -log10 (p-value). Values above 1.3 are considered statistically significant (FDR adjusted p-value <0.05). ns denotes non-significant. EC, E. coli; SE, S. epidermidis; HCA, histologic chorioamnionitis; ns, non-significant. .

119

4.6.2.2 Upstream regulator analysis IPA upstream regulator analysis was performed to identify which transcriptional regulators were significantly associated with the monocyte response to bacterial stimulation in each infant group. This analysis compares the overlap between known downstream targets of each transcriptional regulator with the genes differentially expressed upon bacterial stimulation, and is thus predicted to identify molecules regulating the observed differential expression pattern. IPA also predicts whether the transcriptional regulators are activated or inhibited, based on known relationships between each regulator and the expression (up- or down-regulation) of the downstream targets within our dataset. These predictions are reported as activation z-scores, and are a second method for ranking predicted upstream transcriptional regulators. Positive z- scores indicate activation and negative z-scores indicate inhibition of the upstream regulator. Absolute activation z-scores ≥2 and p-values <0.01 were considered significant as recommended by IPA.

In addition to assessment of the differentially expressed genes unique to each infant group, IPA upstream regulator analysis was performed on the top 2000 (input limit for IPA analysis) most significantly differentially expressed genes for each infant group in response to E. coli or S. epidermidis, to gain an understanding of the major transcriptional changes induced by stimulation. This analysis was performed on each infant group separately, and therefore the lists of top 2000 differentially expressed genes were variable between infant groups.

4.6.2.2.1 Infant monocyte responses to E. coli

Upstream regulator analysis performed on the top 2000 genes differentially expressed by monocytes in response to E. coli resulted in >350 significant upstream regulators for each infant group, the top twenty of which are displayed in Table 4.16. The p-values for these upstream transcriptional regulators were highly similar across all infant groups (Figure 4.15), and the same ten upstream transcriptional regulators were predicted to be the most activated or inhibited across all infant groups (Figure 4.16). These results indicate that similar pathways were predicted to regulate the major monocyte transcriptional changes induced by E. coli in all infant groups, consistent with the high degree of overlap in induced gene expression (Figure 4.13A).

120

Table 4.16 The twenty most significant upstream transcriptional regulators associated with the top 2000 differentially expressed genes induced by E. coli for each infant group p-score* Upstream transcriptional Preterm HCA- top Preterm HCA+ top Term top 2000 regulator 2000 genes 2000 genes genes NR3C1 24.5 24.5 25.1 TREM1 25.4 24.6 23.7 TNF 19.4 22.1 22.1 PDGF BB 20.5 19.5 16.5 NUPR1 16.4 18.8 13.7 NFkB (complex) 14.7 13.5 17.2 F7 12.1 12.1 13.5 TP53 12.1 12.5 11.7 FSH 11.3 10.3 11.3 PGR 10.5 9.2 9.2 TCR 9.6 9.6 9.6 RELA 9.6 9.0 9.0 P38 MAPK 7.8 10.2 8.9 JUN 7.0 10.2 9.5 CD40LG 10.0 7.9 8.6 SELPLG 8.7 8.8 8.7 KRAS 8.1 9.9 8.1 IL1A 8.0 9.5 8.0 TLR7 8.7 7.9 8.7 Lh 9.0 7.1 8.5 *p-score = -log10 (p-value). Values above 2 are considered statistically significant (p<0.01). HCA, histologic chorioamnionitis.

A B

30 30 p<0.0001 p<0.0001 TREM1 NR3C1 TREM1 TNF NR3C1 TNF 20 PDGF BB 20 NUPR1 NFkB (complex) PDGF BB TP53 NFkB (complex) F7 NUPR1 F7 TP53 10 FSH 10 FSH

infants Term P-score: 0 0 infants HCA+ Preterm P-score: 0 10 20 30 0 10 20 30 P-score: Preterm HCA- infants P-score: Preterm HCA- infants

C

30 p<0.0001 NR3C1 TREM1 TNF 20 NFkB (complex) PDGF BB F7 NUPR1 TP53 10 FSH P-score: Term infants Term P-score: 0 0 10 20 30 P-score: Preterm HCA+ infants

Figure 4.15 P-value correlations for significant upstream transcriptional regulators of the E. coli- induced monocyte response. (A-C) Pairwise correlations were performed to compare p-scores (-log10 p- values) for all upstream transcriptional regulators significantly associated with the top 2000 differentially expressed genes induced by E. coli in each infant group. Statistical analysis performed using Spearman correlation.

121

TNF

NUPR1

IL1B

SMARCA4

TREM1

JAG2

IL1RN

COL18A1

Term IgG Preterm HCA+

mir-155-5p Preterm HCA-

-8 -6 -4 -2 0 2 4 6 8

Activation z-score

Figure 4.16 Upstream regulator analysis of the top 2000 monocyte differentially expressed genes induced by E. coli. The top five most activated (yellow bars) or inhibited (blue bars) upstream transcriptional regulators prediced by IPA. Absolute z-scores ≥2 were considered significant.

IPA upstream regulator analysis was then performed on the lists of differentially expressed genes unique to each infant group in response to E. coli, identifying 35 unique significant upstream transcriptional regulators across the three infant groups (Table 4.17). However the majority of these (80%) were linked to less than five downstream targets and therefore are more likely to be statistical artefacts and less likely to be biologically meaningful. Excluding those with fewer than five downstream targets, the most significant predicted upstream regulators were ERK (preterm HCA+ response) and IL13 (term response). The networks of predicted downstream targets for these transcriptional regulators are depicted in Figure 4.17, along with the network for IRF1 (the top upstream regulator of the preterm HCA- unique response). The activation z-scores calculated for ERK and IL13 were -0.6 and -0.8 respectively, suggesting they are modestly inhibited.

Ranking of upstream transcriptional regulators based on activation z-scores identified one significantly inhibited (IL1R1) and three significantly activated (IFNG, TNF and 122

TGM2) regulators of the term infant monocyte response to E. coli (Figure 4.18A). Two regulators were identified as significantly activated (HSF1, IL6) within the preterm HCA+ response, and TP73 and IFNG were identified as activated and inhibited regulators respectively within the preterm HCA- response to E. coli (Figure 4.18A). This last result was particularly interesting as IFNG was identified as activated within the unique term monocyte response, suggesting a potential point of differential regulation between preterm and term infants. Comparative networks of the downstream targets of IFNG within the E. coli responses unique to term and preterm infants are illustrated in Figure 4.19.

Table 4.17 Upstream transcriptional regulators significantly associated with the unique monocyte response to E. coli for each infant group p-score* Upstream transcriptional Preterm HCA- Preterm HCA+ Term unique regulator unique genes unique genes genes IRF1† 2.2 ns ns ZAP70† 2.1 ns ns ERK ns 3.9 ns HDAC10† ns 3.7 ns PTGS1† ns 3.7 ns SENP1† ns 3.7 ns GIT1† ns 3.0 ns FGF16† ns 3.0 ns LDL† ns 2.8 ns miR-491-5p† ns 2.7 ns PROC† ns 2.7 ns Fcgr2† ns 2.7 ns EHF ns 2.6 ns Jnk ns 2.6 ns L1CAM† ns 2.6 ns AXL† ns 2.6 ns SCUBE3† ns 2.3 ns miR-451a† ns 2.3 ns CEBPA ns 2.3 ns DEF6† ns 2.1 ns CRP† ns 2.1 ns Lh ns 2.1 ns BSG† ns 2.0 ns PTP4A3† ns ns 2.9 RAE1† ns ns 2.7 Immunoglobulin† ns ns 2.6 ANG† ns ns 2.6 IL13 ns ns 2.5 RARRES3† ns ns 2.4 MAP2K6† ns ns 2.4 MAP2K3† ns ns 2.4 IL1RN ns ns 2.3 LILRB4† ns ns 2.3 mir-183† ns ns 2.1 RNF216† ns ns 2.0 *p-score = -log10 (p-value). Values above 2 are considered statistically significant (p<0.01). †Regulator linked to less than five downstream targets. HCA, histologic chorioamnionitis; ns, non-significant. 123

A B

C

Figure 4.17 IPA generated networks for ERK, IL13 and IRF1. (A) The ERK network identified in the preterm HCA+ infant E. coli response. (B) The IL13 network identified in the term infant E. coli response. (C) The IRF1 network identified in the preterm HCA- infant E. coli response. The magnitude of up- or down-regulation for each downstream target is indicated by the intensity of red or green colouring, respectively. Dashed and solid lines represent indirect and direct relationships respectively, where orange leads to activation, blue leads to inhibition, yellow indicates that the finding is inconsistent with the state of the downstream molecule and grey indicates that the effect was not predicted. Refer to appendix Figure 7.1 for the full network legend.

124

A B

IL1RN

TGM2 Term unique genes TNF

IFNG

HSF1 Preterm HCA+ unique genes IL6

IFNG Preterm HCA- unique genes TP73

-4 -2 0 2 4 0 1 2 3

Activation z-score -log10 (p-value)

Figure 4.18 Upstream regulator analysis of unique monocyte responses to E. coli. The upstream transcriptional regulators with significant activation z-scores (absolute score ≥2) for each infant-group unique monocyte response to E. coli. Activated (yellow) and inhibited (blue) transcriptional regulators have positive and negative scores, respectively.

A Term response B Preterm HCA- response

Figure 4.19 The downstream targets of IFNG. Networks of the downstream targets of IFNG within the term infant (A) and preterm HCA- infant (B) monocyte response to E. coli. IPA predicted IFNG to be an activated (z-score 2.5) and inhibited (z-score - 2.7) upstream transcriptional regulator within the term and preterm HCA- infant responses to E. coli, respectively. The magnitude of up- or down-regulation for each downstream target is indicated by the intensity of red or green colouring, respectively. Dashed lines represent indirect relationships where orange leads to activation, blue leads to inhibition and grey indicates that the effect was not predicted. Refer to appendix Figure 7.1 for the full network legend.

125

4.6.2.2.2 Infant monocyte responses to S. epidermidis

Similar to the results observed for the major E. coli-induced monocyte transcriptional changes, upstream regulator analysis on the top 2000 differentially expressed genes by monocytes in response to S. epidermidis also resulted in >350 significant upstream regulators for each infant group, the top twenty of which are displayed in Table 4.17. Again the p-values for these transcriptional regulators were highly similar across all infant groups (Figure 4.20) and the same ten upstream transcriptional regulators were predicted to be the most activated or inhibited across all infant groups (Figure 4.21). These results indicate that similar pathways were predicted to regulate the major monocyte transcriptional changes induced by S. epidermidis in all infant groups. Interestingly many of these transcriptional regulators were associated with both the E. coli and S. epidermidis response; this is explored in more detail in Chapter 5.

Table 4.18 The twenty most significant upstream transcriptional regulators associated with the top 2000 differentially expressed genes induced by S. epidermidis for each infant group p-score* Upstream transcriptional Preterm HCA- top Preterm HCA+ top Term top regulator 2000 genes 2000 genes 2000 genes TREM1 25.2 25.3 28.8 NR3C1 26.8 22.6 22.2 PDGF BB 22.6 22.6 23.8 TNF 23.6 20.9 22.1 NUPR1 16.2 20.6 17.3 FSH 14.9 12.8 13.4 NFkB (complex) 15.2 13.4 12.3 TP53 12.7 13.6 12.9 F7 12.1 10.8 12.1 KRAS 10.8 10.9 9.9 Lh 11.0 8.5 10.6 RELA 11.4 8.9 9.6 Gsk3 10.8 7.4 9.7 IL1A 9.4 9.5 8.8 P38 MAPK 8.3 9.5 9.6 Pkc(s) 9.7 8.1 8.9 CD40LG 9.2 8.5 8.6 PGR 8.5 9.1 8.5 TP63 8.4 8.4 8.5 MYC 10.0 8.5 6.1 *p-score = -log10 (p-value). Values above 2 are considered statistically significant (p<0.01). HCA, histologic chorioamnionitis.

126

A B

30 30 p<0.0001 p<0.0001 NR3C1 TREM1 NR3C1 TREM1 TNF PDGF BB PDGF BB TNF 20 NUPR1 20 NUPR1 TP53 NFkB (complex) FSH NFkB (complex) FSH TP53 F7 10 F7 10 P-score: Term infants Term P-score: 0 0

P-score: Preterm HCA+ infants HCA+ Preterm P-score: 0 10 20 30 40 0 10 20 30 40 P-score: Preterm HCA- infants P-score: Preterm HCA- infants

C

30 p<0.0001 NR3C1 TREM1 TNF PDGF BB 20

NUPR1 FSH NFkB (complex) F7 TP53 10

P-score: Term infants Term P-score: 0 0 10 20 30 P-score: Preterm HCA+ infants

Figure 4.20 P-value correlations for significant upstream transcriptional regulators of the S. epidermidis-induced monocyte response. (A-C) Pairwise correlations were performed to compare p- scores (-log10 p-values) for all upstream transcriptional regulators significantly associated with the top 2000 differentially expressed genes induced by S. epidermidis in each infant group. Statistical analysis performed using Spearman correlation.

127

TNF NUPR1

TREM1

SMARCA4

PDGF BB

RBM5

IgG

Estrogen receptor Term JAG2 Preterm HCA+

Preterm HCA- COL18A1

-8 -6 -4 -2 0 2 4 6 8 Activation z-score

Figure 4.21 Upstream regulator analysis of the top 2000 monocyte differentially expressed genes induced by S. epidermidis. The top five most activated (yellow) or inhibited (blue) upstream transcriptional regulators prediced by IPA. Absolute z-scores ≥2 were considered significant.

Analysis of the differentially expressed genes unique to each infant group in response to S. epidermidis revealed thirteen unique significant upstream transcriptional regulators across all three infant groups (Table 4.19). Only three (all associated with the preterm HCA- unique genes) were linked to more than five downstream targets. Luteinizing hormone (Lh) was the most significant of these, the network for which is presented in Figure 4.22 along with the networks for SGK1 and CEBPE which were the top upstream regulators associated with the preterm HCA+ and term infant unique responses to S. epidermidis, respectively. No upstream transcriptional regulators were associated with significant activation z-scores.

128

Table 4.19 Upstream transcriptional regulators significantly associated with the unique monocyte response to S. epidermidis for each infant group p-score* Upstream transcriptional Preterm HCA- Preterm HCA+ Term unique regulator unique genes unique genes genes Lh 4.1 ns ns FSH 3.1 ns ns IL15† 2.5 ns ns SPI1† 2.3 ns ns IL13 2.2 ns ns miR-218-5p † 2.1 ns ns MAPK9† 2.1 ns ns SGK1† ns 2.0 ns CEBPE† ns ns 3.9 FES† ns ns 2.9 CEBPD† ns ns 2.7 KITLG† ns ns 2.5 CSF3† ns ns 2.0

*p-score = -log10 (p-value). Values above 2 are considered statistically significant (p<0.01). †Regulator linked to less than five downstream targets. HCA, histologic chorioamnionitis; ns, non-significant.

A B

C

Figure 4.22 IPA generated networks for Lh, SGK1 and CEBPE. (A) The Lh network identified in the preterm HCA- infant S. epidermidis response. (B) The SGK1 network identified in the preterm HCA+ infant S. epidermidis response. (C) The CEBPE network identified in the term infant S. epidermidis response. The magnitude of up- or down-regulation for each downstream target is indicated by the intensity of red or green colouring, respectively. Dashed and solid lines represent indirect and direct relationships respectively, where grey indicates that the effect (activation or inhibition) was not predicted. Refer to appendix Figure 7.1 for the full network legend.

129

4.6.2.3 Analysis of biological networks Network analysis helps to explain the biological significance of a collection of genes by linking them to all molecules with which they are known to interact. Identification of the genes that are the most highly interconnected (considered the most significant) is a key step toward understanding the biological relevance of a network. Network analysis was performed independently on each list of differentially expressed genes unique to each infant group following monocyte stimulation with E. coli or S. epidermidis (six separate analyses, see section 2.8.2 for methodology).

The HINT (High-quality INTeractomes) database was used to build the initial networks based on known molecular interactions involving the genes present in the unique differentially expressed gene lists. These networks contained >1500 nodes (genes or molecules) and therefore sub-networks (~150–250 nodes) were enriched to identify clusters containing nodes with the most significant changes in gene expression. The top ten most highly interconnected nodes for each sub-network are presented in Table 4.20 and Table 4.21 for analysis of the uniquely differentially expressed genes in response to E. coli and S. epidermidis, respectively. In all analyses, the majority (80–100%) of the top ten sub-network hubs were not present within the original lists of uniquely differentially expressed genes. This demonstrates that the differentially expressed genes unique to each infant group do not form highly interconnected sub-networks themselves, requiring other non-differentially expressed genes to connect them. The only hubs present with the lists of differentially expressed genes were PTK2 (cytoplasmic protein tyrosine kinase involved in many cellular processes) as the seventh most interconnected hub within the term infant unique response to E. coli, and GOLGA2 (a putative Golgi apparatus protein) and PIN1 (an isomerase involved in post- phosphorylation protein regulation) as the eighth and tenth most interconnected hubs within the preterm HCA- infant unique response to S. epidermidis respectively.

130

Table 4.20 Sub-network analyses for unique monocyte responses to E. coli Present in the Up- or down- Unique gene list # of nodes Top hubs Hub degree* DEG list? regulated GRB2 36 No - ATXN1 11 No - YWHAG 10 No - GSK3B 10 No - Preterm HCA- EEF1G 10 No - 183 (412 genes) COPS6 10 No - CHD3 9 No - GADD45G 9 No - KAT5 8 No - WNK1 8 No - EGFR 29 No - GRB2 25 No - SRC 19 No - ERBB2 16 No - Preterm HCA+ CDC37 14 No - 196 (372 genes) A2M 11 No - MAPK9 10 No - TK1 10 No - CDC42 10 No - ABL1 10 No - EGFR 30 No - ATXN1 16 No - ESR2 11 No - CDC37 11 No - ERBB3 9 No - Term (406 genes) 156 ERBB4 9 No - PTK2 8 Yes Up PSEN1 7 No - SYK 7 No - MAPT 7 No - *Degree is the number of connections (indirect and direct) to other genes/molecules DEG, differentially expressed gene; HCA, histologic chorioamnionitis.

131

Table 4.21 Sub-network analyses for unique monocyte responses to S. epidermidis Unique gene Present in the Up- or down- # of nodes Top hubs Hub degree* list DEG list? regulated CSNK2B 18 No - CREB3 13 No - APPBP2 10 No - LUC7L2 9 No - 162 (top 2 Preterm HCA- UNC119 8 No - sub-networks (364 genes) COPS6 7 No - merged) LSM8 7 No - GOLGA2 7 Yes Up WNK1 6 No - PIN1 6 Yes Down VCAM1 29 No - ZDHHC17 18 No - NOTCH2NL 14 No - ESR2 13 No - 200 (top 2 Preterm HCA+ KRTAP10-7 10 No - sub-networks (480 genes) HNRNPA1 9 No - merged) XRCC6 9 No - ZNF512B 9 No - HNRNPD 9 No - STAT1 9 No - EGFR 31 No - MYC 17 No - AKT1 13 No - SYK 10 No - Term (291 YBX1 10 No - 150 genes) GSK3B 9 No - TERF2IP 9 No - TERF1 8 No - SH3GL2 7 No - CRKL 7 No - *Degree is the number of connections (indirect and direct) to other genes/molecules DEG, differentially expressed gene; HCA, histologic chorioamnionitis.

132

4.6.3 Comparisons of differentially expressed genes using a direct statistical approach

In contrast to many studies assessing transcriptional differences between two biological states (e.g. disease vs. healthy, unstimulated vs. stimulated), we aimed to compare the monocyte transcriptional response to bacterial stimulation, and then compare this response between preterm and term infants (a two-fold comparison). This can either be analysed in two, distinct steps using the Venn diagram-based approach utilised in section 4.6.1 or using a "delta-fold change" approach that directly performs this comparison in a single statistical step. Both methods have their inherent limitations. The two-step approach employed in section 4.6.1 is more commonly used due to its ease of interpretation, explanation, and visualisation but can allow for false positives due to the use of an arbitrary, set p-value cut-off (i.e. p<0.05). In contrast the delta-fold change approach does not suffer from these limitations, but can be more difficult to interpret and is generally overly restrictive. To provide a more robust analysis of differential infant group responses to bacterial stimulation, we chose to supplement the standard analysis in section 4.6.1 with a delta fold-change analysis.

Results from the delta-fold change differential expression analysis are summarised in Table 4.22. None of the monocyte genes differentially expressed following stimulation with E. coli were significantly differentially expressed between infant groups. Seventeen differentially expressed genes induced by S. epidermidis stimulation were significantly differentially expressed between preterm HCA+ and preterm HCA- infants, and seventy-seven were differentially expressed between preterm HCA+ and term infants. Interestingly, no genes that were differentially expressed following monocyte stimulation with S. epidermidis were significantly differentially expressed between preterm HCA- and term infants, similar to the results from direct comparisons of S. epidermidis-stimulated monocytes in section 4.5, Table 4.9.

Table 4.22 Numbers of differentially expressed genes (Venn approach) that are significantly differentially expressed between infant groups EC v Unstimulated SE v Unstimulated Preterm HCA- vs. Preterm HCA+ 0 17 Term vs. Preterm HCA- 0 0 Term vs. Preterm HCA+ 0 77 EC, E. coli; SE, S. epidermidis; HCA, histologic chorioamnionitis.

133

Additionally, ten genes overlapped between the two lists of differentially expressed genes identified in the delta-fold change analysis; SPRYD3, PPP1R8, HNRNPUL2, UBXN7, NFKBIL1, RMI1, FAM134A, RFC2, PLEKHF2 and PHF23. All of these genes except for SPRYD3 were identified as significantly differentially expressed in direct comparisons of S. epidermidis-stimulated monocytes specific to preterm HCA+ infants (section 4.5), suggesting they represent robust differences. Analysis of the expression patterns for these ten genes revealed that the magnitude of up- or down-regulation was significantly less in monocytes from preterm HCA+ infants (not shown), suggesting a dampened response to S. epidermidis-stimulation. Interestingly, this same pattern of expression was observed for the majority (88%) of the genes identified in the delta-fold change analysis (Figure 4.23).

Next, the differentially expressed genes identified through delta-fold change differential expression analysis were compared to the lists of differentially expressed genes unique to each group generated using the Venn approach to identify the most robust differences. Only one out of the 17 genes differentially expressed between preterm HCA+ and preterm HCA- monocytes was found using the Venn approach (KCTD11, unique to preterm HCA- infants). Six out of the 77 genes differentially expressed between preterm HCA+ and term infants were found using the Venn approach; ELK3 and TMEM263 unique to preterm HCA+ infants, and TP53INP1, STOML2, RGS10 and C19orf12 unique to both preterm HCA+ and term infants as they were differentially expressed in opposite directions.

A B

6 6

4 4

2 2

0 0

Log fold change -2 Log fold change -2

-4 -4 Preterm Preterm Term Preterm Preterm Term HCA- HCA+ HCA- HCA+

Figure 4.23 Normalised log2 fold change values for genes identified through delta-fold change differential expression analysis. A total of 84 unique genes that were differentially expressed following stimulation with S. epidermidis were identified as differentially expressed between infant groups. Preterm HCA+ infant monocytes exhibit significantly decreased magnitudes of up-regulation (A) or down- regulation (B) of the majority (88%) of these genes. 134

4.7 Do preterm monocytes exhibit a distinct CD16+ monocyte transcriptome signature?

It is well established that there are at least three phenotypically and functionally distinct subsets of human monocytes (175). Characterisation of the samples used for RNA-seq in this thesis showed that preterm infants had significantly lower frequencies of classical monocytes (CD14++ CD16-), and significantly higher frequencies of CD16+ monocytes compared to term infants, an effect that was most distinct for preterm HCA- infants (Figure 4.2). Given that the monocyte subsets are functionally distinct, perhaps these differences in monocyte subset frequencies were obscuring the true transcriptional differences between preterm and term infants or alternatively introducing transcriptional differences between samples. To explore this hypothesis, we attempted to detect CD16+ and CD16- monocyte gene signatures within the RNA-seq data, which may correlate with a particular infant group. Initial analysis revealed that there were no significant differences between infant groups in expression of the CD14 gene, or the two genes encoding for CD16; FCGR3A and FCGR3B (Figure 4.24).

A CD14 B FCGR3A (CD16a) C FCGR3B (CD16b) 10 10 10 8 8 8 6 6 6 4 4 4 2 2 2 gene expression gene expression gene expression gene 2 2 2 0 0 0 Log Log Log -2 -2 -2 Preterm Preterm Term Preterm Preterm Term Preterm Preterm Term HCA - HCA + HCA - HCA + HCA - HCA +

Figure 4.24 Comparisons of CD14 and CD16 gene expression between preterm and term infant unstimulated monocytes. Normalised log2 expression values for CD14 (A) and the two genes encoding for CD16 (B-C), between unstimulated monocytes from preterm and term infants. Data are presented for individual donors (symbols) with bars showing mean ± SEM. HCA, histologic chorioamnionitis.

135

We then generated subset specific signatures using a recent meta-analysis of five microarray studies comparing the transcriptome profiles of CD16+ and CD16- human monocytes (237). This meta-analysis identified 108 genes upregulated by CD16- monocytes (84 present in our dataset) and 74 genes upregulated by CD16+ monocytes (51 present in our dataset). The genes not present in our analysis were expressed below the expression threshold of this experiment and were therefore removed prior to RNA- seq analysis (see section 2.7.5). Box plots representing the levels of gene expression for these 84 and 51 genes are presented for each infant in Figure 4.25. Similar levels of expression were observed between preterm and term infants for both lists of subset specific genes (Figure 4.25). Patterns of expression were also similar between infant groups when the analysis was narrowed down to the top five genes most differentially expressed by the CD16+ and CD16- monocyte subsets (Figure 4.26).

A B

15

10

5

0

gene expression gene 2 -5 Log -10 Preterm Preterm Term Preterm Preterm Term HCA- HCA+ HCA- HCA+

Figure 4.25 Preterm and term infant monocytes express similar levels of genes associated with CD16+ or CD16- monocytes. Normalised log2 expression values for 51 genes more highly expressed by CD16+ monocytes (A), and 84 genes more highly expressed by CD16- monocytes (B).

136

A 15 Top genes upregulated in CD16+ monocytes

10 CKB 5 CDKN1C FMNL2

0 VMO1 gene expression gene 2 SH2D1B -5 Log -10 Preterm HCA- Preterm HCA+ Term

B

15 Top genes upregulated in CD16- monocytes

10

VCAN 5 NRG1 CD163 0 CD14 gene expression gene 2 -5 S100A12 Log -10 Preterm HCA- Preterm HCA+ Term

Figure 4.26 Log2 expression values for genes most highly associated with CD16+ and CD16- monocyte subsets. Expression values for the five genes most upregulated by CD16+ monocytes (A) and CD16- monocytes (B) were compared across infant groups using RNA-seq data from unstimulated monocytes. Genes were selected using a meta-analysis of five microarray studies comparing human monocyte subsets as a reference (237).

137

All 135 subset-specific genes were then compared to the lists of genes identified as significantly differentially expressed between unstimulated monocytes of each infant group (see section 4.5) to determine whether variations in CD16+ subset frequencies between infant groups were a contributing factor to these differences. The results from these comparisons are summarised in Table 4.23. These results provide no evidence that CD16+ monocytes were over-represented transcriptionally in the preterm infant groups, as suggested by differences in cell frequencies.

Table 4.23 Comparisons between monocyte subset specific genes, and genes significantly differentially expressed between unstimulated monocytes from each infant group Genes overlapping with Pair-wise group comparison of CD16+ monocyte CD16- monocyte unstimulated monocytes (# of DEG) subset genes subset genes 5 genes (ALDH1A1, Preterm HCA- vs. Preterm HCA+ 2 genes (RYR1, FMNL2) ADAM28, CYP27A1, (169 genes) QPCT, SCARB1) Term vs. Preterm HCA- (23 genes) 0 1 gene (NRG1) 2 genes (NRG1, Term vs. Preterm HCA+ (50 genes) 0 ALDH1A1) Colours indicate significantly increased (red) or decreased (blue) expression by the first infant group relative to the second infant group listed in the comparison. DEG, differentially expressed genes; HCA, histologic chorioamnionitis.

138

4.8 Analysis of inflammatory proteins in monocyte and CBMC culture supernatants

The literature has demonstrated that whole blood or mixed CBMC cultures from preterm infants exhibit significantly deficient pro-inflammatory cytokine responses following stimulation with TLR agonists, compared to term infants (164, 185, 209, 212, 298). The RNA-seq data thus far is not consistent with these findings. Therefore we repeated the mixed CBMC cultures (1x105 cells) with a two-hour stimulation by live E. coli or S. epidermidis instead of TLR agonists, to see if we could recapitulate the deficiency associated with preterm infants. We also measured the levels of cytokines in culture supernatants from isolated monocytes (paired to RNA-seq data), to determine whether there was a deficiency at the protein level. In addition, we performed live E. coli or S. epidermidis stimulations on CBMC that were rested for fifteen hours, as a control for the post cell-sort monocyte rest period (see section 2.4.4). A lack of sample precluded CBMC analysis for some infants with RNA-seq data, therefore additional infants were included to provide even numbers across groups. The final cohort for the CBMC stimulations included five term infants (four with RNA-seq data), six preterm HCA- infants (two with RNA-seq data) and seven preterm HCA+ infants (none with RNA-seq data). The preterm infants were age-matched to those with RNA-seq data. A panel of eleven cytokines (IL-1β, IL-6, IL-8, IL-10, IL-12p70, CXCL10, TNFα, CCL2, CCL3, G-CSF and M-CSF) were measured in all culture supernatants using a multiplex bead-based immunoassay (see section 2.4.5).

Levels of IL-12p70, M-CSF and G-CSF were undetectable in all CBMC culture supernatants. Levels of IL-10, CXCL10 and CCL2 fell below the limit of detection in two-hour stimulation supernatants (Figure 4.27D-E and G). Term infant CBMC produced the highest levels of all remaining cytokines (IL-1β, IL-6, IL-8, TNFα and CCL3) following two-hour stimulation with E. coli or S. epidermidis compared to preterm infants (HCA- or HCA+), with significantly higher levels of IL-1β in response to E. coli compared to preterm HCA+ infants (p<0.05, Figure 4.27A) and significantly higher levels of IL-8 in response to S. epidermidis compared to preterm HCA- infants (p<0.05, Figure 4.27C). Despite the trend toward higher levels IL-6, TNFα and CCL3 in term infant CBMC supernatants, comparisons between preterm and term infants for these cytokines were non-significant, in contrast to our previous work using heat-killed bacteria or TLR agonists (193, 213).

139

Interestingly, CBMC rested for fifteen hours prior to stimulation generally produced lower levels of the inflammatory cytokines IL-1β and IL-6 (Figure 4.27A and B) and higher levels of the anti-inflammatory and chemoattractant cytokines IL-8, IL-10, CXCL10, CCL2 and CCL3 (Figure 4.27C-E and G-H) compared to CBMC stimulated for two hours without resting. Furthermore, the significant differences in IL-1β and IL-8 observed between term and preterm CBMC became non-significant when the cells were rested for fifteen hours prior to stimulation, suggesting alterations in the culture dynamics during the rest period.

Overall, monocytes from all infant groups produced higher levels of all measured cytokines compared to CBMC with the exception of IL-12p70 and M-CSF, which were also undetectable in monocyte culture supernatants. There were no significant differences in the levels of IL-1β, IL-6, IL-8, IL-10, CXCL10, TNFα, CCL2, CCL3 or G-CSF between preterm (HCA- or HCA+) and term infant monocyte culture supernatants from monocytes at rest, or following bacterial stimulation (Figure 4.28). Preterm HCA- monocytes produced significantly higher levels of IL-1β in response to S. epidermidis compared to preterm HCA+ monocytes (Figure 4.27A). Monocytes from preterm and term infants also produced similar levels of all detectable analytes following stimulation with the direct TLR agonists LPS (TLR4 agonist) or Pam3CSK4 (TLR2 agonist) for 2 or 24 hours (not shown).

140

A B Preterm HCA- IL-1β IL-6 80 500 Preterm HCA+ * 400 Term 60 300 40 pg/mL pg/mL 200 20 100 0 0 UN EC SE UN EC SE UN EC SE UN EC SE CBMC CBMC CBMC CBMC post 15-hour rest post 15-hour rest

C IL-8 D IL-10 5000 6 4000 4 3000 * pg/mL 2000 pg/mL 2 1000 0 0 UN EC SE UN EC SE UN EC SE UN EC SE CBMC CBMC CBMC CBMC post 15-hour rest post 15-hour rest

E CXCL10 F TNFα 20 600

15 400

10 pg/mL pg/mL 200 5

0 0 UN EC SE UN EC SE UN EC SE UN EC SE

CBMC CBMC CBMC CBMC post 15-hour rest post 15-hour rest

G CCL2 H CCL3 300 1500

200 1000 pg/mL pg/mL 100 500

0 0 UN EC SE UN EC SE UN EC SE UN EC SE CBMC CBMC CBMC CBMC post 15-hour rest post 15-hour rest Figure 4.27 Protein levels of inflammatory cytokines and chemokines in CBMC culture supernatants. CBMC from preterm HCA- (n=6), HCA+ (n=7) and term infants (n=5) were stimulated with E. coli or S. epidermidis immediately for 2 hours (left panels), or rested for 15 hours prior to a 2- hour stimulation (right panels). Data are presented as mean ± SEM. Statistical analysis was performed using the Kruskal-Wallis test with Dunn’s multiple comparisons test. *p<0.05. UN, unstimulated; EC, E. coli; SE, S. epidermidis; HCA, histologic chorioamnionitis.

141

Preterm HCA- Preterm HCA+

Term

A IL-1β B IL-6 C IL-8

400 * 2500 25000

300 2000 20000 1500 15000 200 pg/mL pg/mL pg/mL 1000 10000 100 500 5000 0 0 0 UN EC SE UN EC SE UN EC SE

D E F TNF IL-10 CXCL10 α 100 1000 1500

80 800 1000 60 600 pg/mL pg/mL 40 pg/mL 400 500 20 200 0 0 0 UN EC SE UN EC SE UN EC SE

G CCL2 H CCL3 I G-CSF 4000 4000 15

3000 3000 10 2000 2000 pg/mL pg/mL pg/mL 5 1000 1000

0 0 0 UN EC SE UN EC SE UN EC SE

Figure 4.28 Protein levels of inflammatory cytokines and chemokines in monocyte culture supernatants. Unstimulated, E. coli-stimulated and S. epidermidis-stimulated monocytes from preterm and term infants were cultured for 2 hours following a 15-hour rest post-sort. Data are presented as mean ± SEM. Statistical analysis was performed using the Kruskal-Wallis test with Dunn’s multiple comparisons test. *p<0.05. UN, unstimulated; EC, E. coli; SE, S. epidermidis; HCA, histologic chorioamnionitis.

142

4.8.1 Correlations with RNA-seq data

Correlation tests between monocyte gene and protein expression were performed on all monocyte samples using Spearman correlation, and significant positive correlations between gene and protein levels were observed for all cytokines and chemokines. This was true whether the protein analysis was performed on supernatants following 2 hours or 24 hours of monocyte culture (Figure 4.29 and Figure 4.30 respectively). For several cytokines (IL-1β, IL-6, IL-8, IL-10 and G-CSF), the correlation coefficient was higher when gene expression (at 2 hours) was correlated with protein levels at 24 hours, suggesting that protein expression for these cytokines may lag behind gene expression. Overall these high correlations validate the robustness of the RNA-seq data as an accurate measure of monocyte transcriptional responses, in the context of this model.

143

IL-1β IL-6 IL-8 100000 p<0.0001 100000 p<0.0001 100000 p<0.0001 Spearman r = 0.83 Spearman r = 0.56 Spearman r = 0.65 10000 95% CI = 0.704-0.909 10000 95% CI = 0.29-0.74 10000 95% CI = 0.41-0.79 1000 1000 1000 pg/mL 100 pg/mL 100 pg/mL 100

10 10 10

1 1 1 -10 0 10 20 -10 0 10 20 -10 0 10 20 Log gene expression Log gene expression Log gene expression 2 2 2

IL-10 CXCL10 TNFα

100000 p<0.0001 100000 p<0.0001 100000 p<0.0001 Spearman r = 0.71 Spearman r = 0.85 Spearman r = 0.85 10000 10000 95% CI = 0.509-0.84 95% CI = 0.72-0.92 10000 95% CI = 0.73-0.92 1000 1000 1000 100 pg/mL pg/mL pg/mL 100 100 10 10 10 1 1 0.1 1 -10 0 10 20 -10 0 10 20 -10 0 10 20

Log2 gene expression Log2 gene expression Log2 gene expression

CCL2 CCL3 G-CSF

100000 p<0.0001 100000 p<0.0001 100000 p=0.0032 Spearman r = 0.81 Spearman r = 0.76 Spearman r = 0.45 10000 95% CI = 0.67-0.89 10000 95% CI = 0.59-0.87 10000 95% CI = 0.15-0.67 1000 1000 1000 pg/mL pg/mL 100 100 pg/mL 100 10 10 10 1 1 1 -10 0 10 20 -10 0 10 20 -10 0 10 20 Log2 gene expression Log2 gene expression Log2 gene expression

Preterm HCA- UN Preterm HCA- EC Preterm HCA- SE Preterm HCA+ UN Preterm HCA+ EC Preterm HCA+ SE Term UN Term EC Term SE

Figure 4.29 Monocyte gene and protein expression is highly correlated after 2 hours of culture. Gene and protein expression of IL-1β, IL-6, IL-8, IL-10, CXCL10, TNFα, CCL2, CCL3 and G-CSF was measured in unstimulated, E. coli-stimulated and S. epidermidis-stimulated monocytes following 2 hours of culture. Gene expression was derived from normalised log2 RNA-seq data, and protein expression was measured in infant-paired culture supernatants. There was a significant positive correlation between gene and protein expression for all analytes (Spearman correlation). P-values, correlation coefficients, and 95% confidence intervals are indicated for each analyte. UN, unstimulated; EC, E. coli; SE, S. epidermidis; HCA, histologic chorioamnionitis.

144

IL-1β IL-6 IL-8 100000 p<0.0001 100000 p=0.0009 100000 p<0.0001 Spearman r = 0.88 Spearman r = 0.66 Spearman r = 0.84 10000 95% CI = 0.73-0.95 10000 95% CI = 0.31-0.85 10000 95% CI = 0.57-0.94 1000 1000 1000 pg/mL pg/mL 100 100 pg/mL 100 10 10 10 1 1 1 -10 0 10 20 -10 0 10 20 -10 0 10 20 Log2 gene expression Log2 gene expression Log2 gene expression

IL-10 CXCL10 TNFα 100000 p<0.0001 100000 p<0.0001 100000 p<0.0001 Spearman r = 0.78 Spearman r = 0.79 Spearman r = 0.82 10000 10000 95% CI = 0.56-0.89 95% CI = 0.59-0.91 10000 95% CI = 0.62-0.93 1000 1000 1000 100 pg/mL pg/mL 100 pg/mL 100 10

10 1 10

1 0.1 1 -10 0 10 20 -10 0 10 20 -10 0 10 20 Log gene expression Log gene expression Log gene expression 2 2 2

CCL2 CCL3 G-CSF 100000 p<0.0001 100000 p=0.0006 100000 p=0.0003 Spearman r = 0.76 Spearman r = 0.62 Spearman r = 0.65 10000 95% CI = 0.52-0.88 10000 95% CI = 0.29-0.81 10000 95% CI = 0.34-0.83 1000 1000 1000 pg/mL pg/mL pg/mL 100 100 100

10 10 10

1 1 1 -10 0 10 20 -10 0 10 20 -10 0 10 20 Log gene expression Log gene expression Log gene expression 2 2 2 Preterm HCA- UN Preterm HCA- EC Preterm HCA- SE Preterm HCA+ UN Preterm HCA+ EC Preterm HCA+ SE Term UN Term EC Term SE

Figure 4.30 Early monocyte gene expression (2 hours) highly correlates with late protein expression (24 hours). Gene and protein expression of IL-1β, IL-6, IL-8, IL-10, CXCL10, TNFα, CCL2, CCL3 and G-CSF was measured in unstimulated, E. coli-stimulated and S. epidermidis-stimulated monocytes following 2 hours (for gene expression) or 24 hours (for protein expression) of culture. Gene expression was derived from normalised log2 RNA-seq data, and protein expression was measured in infant-paired culture supernatants. There was a significant positive correlation between gene and protein expression for all analytes (Spearman correlation). P-values, correlation coefficients, and 95% confidence intervals are indicated for each analyte. UN, unstimulated; EC, E. coli; SE, S. epidermidis; HCA, histologic chorioamnionitis.

145

4.9 Discussion

It has been suggested that preterm infant monocytes are functionally deficient in their responses to TLR agonists and killed bacterial preparations, which may contribute to their unique susceptibility to invasive infection (164, 182, 185, 190, 193, 209, 213, 298). These findings are based primarily on protein-level assessments of inflammatory cytokine and chemokine production, utilising monocytes within mixed mononuclear cells cultures or whole blood. For the first time, we have sequenced the entire transcriptome of purified cord blood monocytes from preterm and term infants before, and following stimulation with live E. coli or S. epidermidis. We demonstrate that monocytes from preterm infants are not intrinsically deficient compared to term infant monocytes, but that exposure to HCA results in the transcriptional re-programming of preterm monocytes toward a hyporesponsive phenotype when challenged with S. epidermidis.

Unsupervised hierarchical clustering and principle component analysis were used to identify structure within the RNA-seq data and revealed distinct separation between unstimulated and E. coli-stimulated and S. epidermidis-stimulated monocytes. However the individual infant groups did not form unique clusters in either analysis. This suggests that a higher amount of variation in the transcriptional profiles was due to inter-infant variability, rather than gestational age-dependent variability.

We then used a targeted approach to analysis based on previously described deficiencies in innate immunity in preterm infants, to narrow the analysis to seventy key genes from TLR, inflammatory, antigen presentation and apoptosis pathways. The TLR2 and TLR4 pathway genes were of particular interest as these pathogen recognition receptors are central to the immune responses to S. epidermidis and E. coli, respectively (109, 299). We found no significant differences in the expression of TLR2 or TLR4-related receptor, adaptor molecule or transcription factor genes between preterm HCA- and term infant monocytes at rest, or post-stimulation. This finding is contradictory to several reports of significantly reduced surface expression of TLR4 by preterm monocytes (186, 190, 209), but consistent with other findings of similar TLR4 protein expression between preterm and term infant monocytes (113, 189, 193). Literature on TLR2 protein expression by neonatal monocytes is also inconsistent, with reports of similar (113, 190, 193) and decreased (186, 188) surface expression, and one report of increased TLR2 mRNA (186) by preterm monocytes. Importantly, preterm infant 146 monocytes also expressed comparable levels of IRAK4 and MyD88 compared to term monocytes. These two genes are critical for effective TLR2 and TLR4 signalling, evidenced by the heightened susceptibility to bacterial infections experienced by those with genetic deficiencies affecting these genes (300, 301). In contrast, monocytes from preterm infants exposed to HCA expressed significantly lower levels of several TLR pathway genes including TLR4, TLR5, TBK1 and NFKBIA. These differences were predominantly in comparison to term infant monocytes following stimulation with S. epidermidis, suggesting this is not an intrinsic deficiency (normal levels at rest) but a general dampening of TLR pathway signalling upon stimulation.

Defective TLR signalling may contribute to preterm infants’ impaired inflammatory cytokine and chemokine production (302). Here we show that preterm monocytes are highly capable of producing several inflammatory cytokines and chemokines at the gene and protein level in response to E. coli or S. epidermidis. The high correlations between gene and protein expression observed suggests the machinery required for the translation of these genes in response to E. coli or S. epidermidis is intact in preterm monocytes, in keeping with the expression of TLR pathway genes. Interestingly, monocytes from preterm HCA+ infants produced the lowest levels of IL-1β, IL-6, IL-10 and CXCL10 protein in response to bacterial stimulation, suggesting that prenatal exposure to HCA may attenuate the monocyte inflammatory response. Impaired blood monocyte production of IL-6, TNFα and hydrogen peroxide following TLR ligation is also seen at early postnatal time-points in fetal sheep models of LPS-induced chorioamnionitis (226-228). Overall this suggests that prenatal exposure to infection and inflammation may alter the risk of sepsis at least partly via modulation of monocyte responses to infection.

The presentation of foreign antigens via MHC class II by monocytes and monocyte- derived cells provides an important link between innate and adaptive immunity (169). We showed similar HLA-DR surface expression on monocytes between preterm and term infants, however preterm HCA+ infants had significantly reduced frequencies of HLA-DR+ monocytes with reduced staining intensity, particularly on their non-classical monocyte subset. Azizia et al. also reported similar frequencies of MHC class II cord blood monocytes between preterm and term infants, but significantly reduced frequencies in preterm infants with HCA (219). Analysis of nine genes involved in antigen presentation revealed that exposure to HCA was associated with significantly 147 lower expression of CD74, CD86 and HLA-DRA following stimulation with S. epidermidis, with CD74 and HLA-DRA also intrinsically lower in unstimulated preterm HCA+ infant monocytes. Preterm infants with fetal inflammatory response syndrome also down-regulate several genes critical for antigen processing and presenting including CD74 and HLA-DRA (247). A reduced capacity for antigen presentation is associated with sepsis in the first week of life, which may partly explain the heightened susceptibility for early-onset sepsis of HCA-exposed neonates (39, 219).

The clearance of apoptotic cells is important for the resolution of infection (203). Neutrophils of septic patients display reduced apoptosis via the up-regulation of MCL1, and gut pathogens including E. coli supress death receptor signalling in order to establish infection (303, 304). The limited data on neonatal monocyte apoptosis has focused on differences between term infants and adults and has suggested that neonates suppress infection-induced apoptosis (204-206, 305). Here we report no significant differences in the expression of pro- or anti-apoptotic genes between preterm and term infants. However, we did observe significantly reduced expression of CASP1 and NLRP3 in preterm HCA+ infant monocytes following stimulation with S. epidermidis. The NLRP3 inflammasome is activated through a variety of PAMPs and regulates the activation of caspase-1 (encoded by the CASP1 gene), required for apoptotic and pyroptotic cell death, as well as the secretion of active IL-1β (207, 208). Reduced expression of CASP1 and NLRP3 by preterm HCA+ infant monocytes may contribute to the decreased production of IL-1β observed by these cells. This is in line with recent findings of a significant reduction in the frequencies of Caspase-1 positive monocytes in preterm infants exposed to HCA (182). Our findings provide further evidence that exposure to HCA does modulate monocyte responses to sepsis-causing pathogens.

We also performed unbiased differential expression analysis to compare the infant groups using two methods; i) direct statistical comparisons of the transcriptome profiles from monocytes with or without stimulation with E. coli or S. epidermidis (section 4.5) and ii) bioinformatic and statistical comparisons of the genes differentially expressed by monocytes from each group following stimulation with E. coli or S. epidermidis (section 4.6). Collectively the results illustrate that the major monocyte transcriptional changes induced by E. coli or S. epidermidis are similar between preterm and term infants and that preterm HCA+ infants displayed a hyporesponsive transcriptional profile. We found relatively low numbers of genes (≤23) to be differentially expressed 148 between preterm HCA- and term monocytes across all culture conditions, providing further evidence that preterm monocytes are not intrinsically deficient (section 4.5). In contrast, a recent study comparing the basal transcriptional states of fetal (18–22 weeks GA) and adult bone marrow monocytes identified >2000 significantly differentially expressed genes, including several genes involved in pathogen recognition and inflammation (238). Three of these genes (MMP9, SOCS3 and IGF2BP1) were also identified in our differential expression analysis of unstimulated monocytes. Together these findings suggest that there are major transcriptional differences between fetal and adult monocytes, and relatively small transcriptional differences between neonatal monocytes of preterm and term gestational ages, indicating that continued maturation of innate immunity occurs in early life.

Analysis of the genes that were differentially expressed upon stimulation revealed that the vast majority (>90%) were common to all infant groups, suggesting there is a core neonatal monocyte response to E. coli and S. epidermidis that is not affected by gestational age or exposure to HCA. The differentially expressed genes that were identified as unique to each infant group generally had similar patterns of expression across all groups and were generally not associated with any canonical signalling pathways, diseases or bio-functions. Furthermore, network analysis on the differentially expressed genes unique to each infant group did not reveal highly interconnected sub- networks, as a large number of non-differentially expressed genes were required to connect the differentially expressed genes. In addition, upstream regulator analysis illustrated that the major transcriptional changes (top 2000 differentially expressed genes) induced by each stimuli were likely to regulated by highly similar pathways across all groups where many of the top regulators were involved in TLR signalling (TREM1, TNF, NF-κB complex, RELA, P38 MAPK, JUN, IL1A, TLR7). In general, upstream regulator analysis of the differentially expressed genes unique to each group revealed relatively few transcriptional regulators of much lower significance that affected a small number (<12) of E. coli or S. epidermidis response genes. This analysis did suggest that IFNG may differentially regulate a subset of the differentially expressed genes unique to preterm HCA- and term infant monocytes in response to E. coli. However the differential expression of these genes (downstream of IFNG) was not statistically significantly different between groups. In fact, delta-fold change analysis (which is more stringent than the Venn approach) did not identify any genes that were

149 differentially expressed following monocyte stimulation with E. coli or S. epidermidis to be differentially expressed between preterm HCA- and term infants.

One potential reason for the observed lack of differences between preterm and term infant groups was due to the gestational ages of the preterm infant cells used in this study. Maturation of several innate immune parameters is dependent on gestational age, with the greatest disparity observed between extremely preterm (<28 weeks GA) and term infants (190, 193, 306). In this study, I only had access to sufficient cell quantities of preterm infants who were 29–32 weeks gestational age, which could mean that any differences between preterm infants at this gestation and term infants are similar, and require greater sensitivity to detect. In support of this, while Strunk et. al. demonstrated a gestational-age dependent increase in gene and protein expression of IL-1β, IL-6, IL-8 and TNFα by mononuclear cells in response to S. epidermidis, the difference between preterm infants at 30–33 weeks and term infants was not significant (193). Additionally, by studying monocytes in isolation we may have minimised any quantitative monocyte deficiencies due to extrinsic factors that would be apparent in whole blood or mixed mononuclear cell cultures. Differences in the frequencies of monocyte subsets between preterm and term infants is unlikely to have obscured large transcriptional differences, as we found no distinct monocyte sub-population gene signatures specific to either infant group.

Contrary to our hypothesis, the infant group with the most distinct transcriptional profile were preterm HCA+ infants, where the majority of the genes that were differentially expressed by this group were down-regulated or supressed following S. epidermidis- stimulation compared to preterm HCA- or term infants. The only clinical parameter that was significantly different between preterm infants with or without exposure to HCA was mode of delivery, which is inherent to the underlying pathophysiology of preterm birth. However data on term infants (without evidence of HCA) shows that mode of delivery does not impact on innate immune responses, suggesting another factor is responsible for the differences observed (307, 308). Consistent with data from sheep models of intra-uterine infection and inflammation we observed a hyporesponsive monocyte phenotype associated with prenatal exposure to HCA (227, 228). These sheep studies also observe a reverse in this phenotype toward hyper-responsiveness at later time points (1–2 weeks), therefore it would be of interest to determine monocyte responsiveness from preterm HCA+ infants at later time points, especially as these 150 infants are known have a decreased risk of late-onset sepsis (41). Given the prominent involvement of HCA with preterm birth, understanding the mechanisms responsible for this hyporesponsive phenotype may provide insight into the altered risk of sepsis observed in this population. We have identified a number of genes that were identified as differentially expressed between preterm HCA+ and preterm HCA- and term infants in response to S. epidermidis across multiple analyses including PPP1R8, HNRNPUL2, UBXN7, NFKBIL1, RMI1, FAM134A, RFC2, PLEKHF2 and PHF23. These genes represent candidates for future study.

In conclusion, we have demonstrated for the first time that preterm infant monocyte responses to bacterial challenge are not transcriptionally deficient compared to term infants, but that prenatal exposure to HCA alters the monocyte transcriptional response to S. epidermidis resulting in a tolerised phenotype affecting a subset of genes.

151

152

Chapter 5 Characterisation of the conserved and pathogen- specific neonatal monocyte transcriptional responses to E. coli and S. epidermidis

5.1 Introduction

Understanding how the host responds to a specific pathogen provides insight into the mechanisms required for effective protection against infection. E. coli and S. epidermidis are two distinct pathogens that inhabit distinct niches within the host (E. coli is part of the normal gut flora whereas S. epidermidis colonises the skin) but which can both cause sepsis in neonates (44, 56, 309, 310). A fundamental difference between E. coli and S. epidermidis is evident in their distinct structures as Gram-negative and Gram-positive microorganisms, respectively. Innate immune responses are triggered by the recognition of the conserved structures present on the bacterial surface by pattern recognition receptors such as TLRs, and the responses to E. coli and S. epidermidis have generally been attributed to TLR4 and TLR2 signalling respectively (109, 299, 311). While TLR4 and TLR2 recognise distinct bacterial structures, the signalling pathways induced by these receptors share many common downstream molecules, and a certain level of redundancy may exist among cell surface TLRs (300, 312). Microarray analysis of human monocyte-derived macrophages challenged with a range of structurally diverse bacterial species (including E. coli, Staphylococcus aureus and Mycobacterium tuberculosis) revealed a conserved transcriptional response, highlighting conserved macrophage responses to bacterial infection (243).

This is the first study to use RNA-seq-based analysis of neonatal monocyte transcriptional responses following challenge with live E. coli or S. epidermidis, providing an opportunity to identify both shared and unique defence pathways to these distinct pathogens. In this chapter, we illustrate that E. coli and S. epidermidis both induce the differential expression of an exclusive subset of genes by both preterm and term infant monocytes. The E. coli-specific genes were characterised by interferon signalling and an anti-viral immune signature. In contrast, the S. epidermidis-specific genes were not collectively associated with any specific pathways identified to-date. Additionally, this chapter illustrates that the major transcriptional changes induced by both pathogens were highly conserved across all infant groups. This response was strongly associated with inflammatory and TLR signalling pathways and illustrates that

153 a conserved neonatal monocyte response to bacterial infection is already present early in gestation.

5.2 Identifying the global neonatal monocyte transcriptional responses to E. coli and S. epidermidis

To determine the pathogen-specific and conserved neonatal monocyte transcriptional responses to E. coli and S. epidermidis, the global sets of transcripts that were differentially expressed upon stimulation were compared between stimuli for each infant group (Figure 5.1A). Across all infant groups the majority of transcripts were differentially expressed in response to both stimuli (74–78%) equating to 7329–7882 genes, whereas only 9–15% of differentially expressed transcripts were specific to either stimuli (Figure 5.1A). Venn diagrams were used to compare the lists of genes that were specifically differentially expressed in response to E. coli or S. epidermidis between infant groups, to identify the genes that represent the common neonatal monocyte response specific to each pathogen (irrespective of gestational age or exposure to HCA). This analysis identified 327 genes specific to the E. coli response (Figure 5.1B) and 407 genes specific to the S. epidermidis response (Figure 5.1C). A third Venn diagram (Figure 5.1D) was used to compare lists of genes that were differentially expressed in response to both stimuli between infant groups, identifying 6511 genes in common that represent the conserved neonatal monocyte response to E. coli and S. epidermidis.

Unsupervised hierarchical cluster analysis was performed on all E. coli- or S. epidermidis-stimulated monocyte samples based on the expression of 734 genes that were specifically differentially expressed in response to either stimuli (327 E. coli- specific and 407 S. epidermidis-specific genes). The resulting dendrogram contained two major branches separating E. coli-stimulated from S. epidermidis-stimulated monocyte samples, indicating that the 734 differentially expressed genes represent pathogen-specific monocyte responses (Figure 5.2A). A heatmap visualising the normalised log2 fold-change expression for these 734 genes illustrates distinct patterns of expression between E. coli- and S. epidermidis-stimulated monocytes, but similar patterns between infant groups within each culture condition (Figure 5.2B).

154

Figure 5.1 Differentially expressed genes specific and common to each stimuli. (A) Venn diagrams comparing the global set of differentially expressed genes in response to EC and SE for each infant group. Numbers in parentheses provide references to B-D. (B) Venn diagram comparing the EC-specific differentially expressed genes across infant groups, identifying 327 genes that represent the neonatal monocyte response to EC. (C) Venn diagram comparing the SE-specific differentially expressed genes across infant groups, identifying 407 genes that represent the neonatal monocyte response to SE. (D) Venn diagram comparing the differentially expressed genes common to EC and SE stimulation across infant groups, identifying 6511 genes that represent the conserved neonatal monocyte response to bacterial stimulation. EC, E. coli; SE, S. epidermidis. 155

Figure 5.2 Cluster analysis and heatmaps based on EC- and SE-specific monocyte response genes. (A) Unsupervised hierarchical clustering was performed on normalised RNA-seq data from EC- and SE- stimulated monocytes based on expression of the 327 and 407 differentially expressed genes specific to each stimulus. The resulting dendrogram contained two major branches that distinctly separated EC- and SE-stimulated monocyte samples. (B) Heatmaps visualizing the log2 fold change in gene expression for the 327 and 407 genes specifically differentially expressed by neonatal monocytes in response to SE and EC, respectively. EC, E. coli; SE, S. epidermidis.

156

5.3 Characterising the neonatal monocyte transcriptional response specific to E. coli

The top twenty most significantly differentially expressed genes specific to the neonatal monocyte response to E. coli are presented in Table 5.1, the top 19 of which were up- regulated. The most significantly differentially expressed gene was PELI1, a known mediator of TLR3/4 signalling and the subsequent production of proinflammatory cytokines (313). Other biological functions associated with the top twenty differentially expressed genes were cell death (RRAGC, RIPK1, SEMA3A), regulation of transcription (ELK, ERF, KLF4) and interferon signalling (ZC3HAV1, OASL).

Table 5.1 The top twenty most significantly differentially expressed genes specific to the neonatal monocyte response to E. coli Significance (p-value)* Up- or down- Gene symbol Preterm HCA- Preterm HCA+ Term infants regulated infants infants PELI1 5.41E-22 4.42E-22 1.79E-20 Up KDELC2 2.73E-14 6.16E-14 7.80E-14 Up ZC3HAV1 3.04E-14 7.77E-14 6.72E-13 Up REV3L 2.39E-14 3.10E-12 2.00E-13 Up GPR31 3.26E-14 1.35E-10 6.21E-13 Up TMOD3 7.89E-11 5.99E-12 1.66E-10 Up JADE3 2.83E-10 2.80E-10 5.96E-11 Up RRAGC 4.21E-10 5.74E-10 8.51E-10 Up RIPK1 1.66E-09 4.18E-09 6.27E-10 Up FBXO34 7.06E-11 5.37E-09 3.74E-09 Up DAGLA 1.63E-10 8.06E-11 1.13E-08 Up NANS 2.22E-09 1.80E-08 5.83E-09 Up ELK3 1.46E-09 5.92E-08 5.85E-12 Up TRIM26 2.75E-08 4.77E-08 1.89E-08 Up RPS6KC1 1.07E-07 1.42E-08 1.72E-09 Up KLF4 1.18E-07 8.88E-09 3.28E-09 Up SEMA3A 9.99E-08 2.27E-08 1.54E-08 Up OASL 1.50E-09 6.02E-09 2.25E-07 Up EEA1 1.91E-07 1.11E-09 4.61E-08 Up ERF 2.51E-07 5.84E-07 2.85E-09 Down *P-values (adjusted for multiple comparisons) for differential expression between unstimulated and E. coli-stimulated monocytes. HCA, histologic chorioamnionitis.

157

5.3.1 Over-represented canonical pathways, diseases or bio-functions

Over-representation analysis was performed within Ingenuity® Pathway Analysis (IPA) using the 327 E. coli-specific response genes to help illuminate the biological processes involved. Twelve canonical pathways (Table 5.2) and twenty-two disease classifications or bio-functions (Table 5.3) were over-represented within the list of E. coli-specific response genes. Interestingly, pathways typically associated with anti-viral immunity were among the most significant, with “Role of RIG1-like Receptors in Antiviral Innate Immunity” classified as significant based on the up-regulation of DHX58, IFIH1, DDX58, IFNB1, RIPK1, TRAF6. The same six genes were also associated with “Activation of IRF by Cytosolic Pattern Recognition Receptors” along with the up- regulation of ISG15 and IFIT2.

Perhaps not surprisingly, the most significant bio-functions associated with the E. coli- specific response genes were “Antimicrobial Response” and “Inflammatory Response” both based on the up-regulation of CREB3, DDX58, DDX60, IFI44, IFIH1, IFIT1, IFIT2, IFIT3, ISG15, MICA, OASL, PML, ZC3HAV1. The “Inflammatory Response” bio-function was also associated with the up-regulation of NCF2, PLSCR1, TRIM21 and TRIM26, and the down-regulation of TRIM28.

158

Table 5.2 Significantly over-represented canonical pathways associated with the E. coli-specific response genes p-score* Canonical pathway EC-specific SE-specific genes genes Activation of IRF by Cytosolic Pattern Recognition 3.0 ns Receptors Role of RIG1-like Receptors in Antiviral Innate 2.3 ns Immunity Huntington's Disease Signalling 1.9 ns Actin Nucleation by ARP-WASP Complex 1.9 ns Phospholipase C Signalling 1.9 ns ERK/MAPK Signalling 1.5 ns Acute Phase Response Signalling 1.3 ns Synaptic Long Term Potentiation 1.3 ns ATM Signalling 1.3 ns TGF-b Signalling 1.3 ns Phagosome maturation 1.3 ns ERK5 Signalling 1.3 ns

Table 5.3 Significantly over-represented diseases and bio-functions associated with the E. coli-specific response genes p-score* Disease or Bio-function EC-specific genes SE-specific genes Antimicrobial Response 3.4 ns Inflammatory Response 3.4 ns Gene Expression 3.1 ns Developmental Disorder 2.1 ns Hereditary Disorder 2.1 ns Metabolic Disease 2.1 ns Cell Signalling 2.1 ns Cell Cycle 2.1 ns Infectious Disease 2.1 ns Cell Death and Survival 2.0 ns Cellular Compromise 1.9 ns Cellular Function and Maintenance 1.9 ns Cellular Development 1.9 ns Cellular Growth and Proliferation 1.9 ns Hematological System Development and Function 1.9 ns Cellular Movement 1.7 ns Neurological Disease 1.5 ns Cancer 1.5 ns Organismal Injury and Abnormalities 1.5 ns Reproductive System Disease 1.5 ns DNA Replication, Recombination, and Repair 1.5 ns Dermatological Diseases and Conditions 1.3 ns

*p-score = -log10 (p-value). Values above 1.3 were considered statistically significant (FDR adjusted p-value <0.05). EC, E. coli; SE, S. epidermidis; ns, non-significant.

159

5.3.2 Upstream regulator analysis

IPA upstream regulator analysis was performed to identify which transcriptional regulators were significantly associated with the neonatal monocyte response specific to E. coli, by comparing the overlap between known downstream targets of each regulator and the 327 differentially expressed genes specific to E. coli-stimulation. Additionally, activation z-scores were calculated to infer the activation state of predicted transcriptional regulators by comparing their known effect on downstream targets with observed changes in gene expression. Transcriptional regulators with activation z-scores ≥2 or ≤2 were considered “activated” or “inhibited”, respectively.

A total of 38 upstream transcriptional regulators were significantly associated with the E. coli-specific response genes (Table 5.4). Only one of these (mir-10) was also significantly associated with the S. epidermidis-specific response genes. Interestingly, the top two most significant upstream transcriptional regulators were IFNL1 (encodes the cytokine IL-29) and TLR3, which are both known for their roles in anti-viral immunity. IFNL1 was also predicted to be the most activated upstream regulator (z- score 4.2), and MAPK1 was predicted to be the most inhibited (z-score -4.1) (Figure 5.3). Networks of predicted downstream targets for the top four most significant upstream transcriptional regulators are presented in Figure 5.4. All of the downstream targets were up-regulated, and there was a high degree of overlap between the downstream targets of the top four transcriptional regulators (Figure 5.4). Many of these genes were associated with the over-represented canonical pathways and bio-functions described in the previous section (DDX58, IFIH1, DDX58, IFNB1, RIPK1, ISG15, IFIT2, DDX60, IFI44, IFIT1, IFIT3, OASL, PML, ZC3HAV1, PLSCR1 and TRIM21).

Analysis of directional networks involving the significant upstream transcriptional regulators revealed that IRF3 was involved in mechanistic networks with four of the predicted upstream regulators (TLR3, RNF216, DDX58 and Interferon alpha). The interaction networks between these transcriptional regulators and IRF3, and the affected downstream targets are presented in Figure 5.5. The activation of IRF3 by the other upstream regulators directly lead to the activation of IFNB1, ISG15, IFIT1 and IFIT2, whereas TLR3, RNF216, DDX58 and Interferon alpha were associated with indirect activation of these targets.

160

Table 5.4 Upstream transcriptional regulators significantly associated with the E. coli-specific response genes p-score* Upstream regulator EC-specific genes SE-specific genes IFNL1 18.8 ns TLR3 18.7 ns IFNA2 14.7 ns MAPK1 13.9 ns TCR 10.9 ns IFNG 7.9 ns PAF1 7.9 ns IL1RN 7.4 ns Interferon alpha 7.0 ns TGM2 6.7 ns CNOT7 6.7 ns SOCS1 5.5 ns RNF216† 5.3 ns EIF2AK2 4.9 ns DDX58† 4.2 ns IFNB1† 4.1 ns MTORC2† 3.7 ns MSR1† 3.7 ns MAVS† 3.5 ns TICAM1† 3.3 ns HERC5† 3.2 ns IRF3† 3.0 ns ARF6† 2.9 ns mir-10† 2.8 2.5 TIRAP† 2.7 ns RARRES3† 2.5 ns TRAF3IP2† 2.5 ns IFN type 1† 2.5 ns TAB1† 2.5 ns SREBF1† 2.3 ns ERK1/2† 2.3 ns IFI16† 2.3 ns IFNA1/IFNA13† 2.3 ns ELF1† 2.2 ns TRAF3† 2.1 ns miR-100-5p† 2.1 ns GAPDH† 2.1 ns TNF 2.0 ns

*p-score = –log10 (p-value). Values above 2 are considered statistically significant (p<0.01). †Regulator affects less than five downstream targets. EC, E. coli; SE, S. epidermidis; ns, non-significant.

161

A B IFNL1 IFNA2

IFNG Interferon alpha PAF1 TGM2

TNF IFNB1

NFkB (complex) TLR3 RNF216 TREM1 SOCS1 IL1RN MAPK1 -6 -4 -2 0 2 4 6 0 5 10 15 20

Activation z-score -log10 (p-value)

Figure 5.3 Upstream regulator analysis of the E. coli-specific response genes. Upstream transcriptional regulators with significant activation z-scores (absolute score ≥2) associated with the neonatal monocyte response specific to E. coli. Regulators with negative scores were classified as inhibited (blue) and regulators with positive scores were classified as activated (yellow).

162

B A

D C

Figure 5.4 Networks for the most significant E. coli-specific upstream transcriptional regulators. Networks of the downstream targets of IFNL1 (z-score 4.2) (A), TLR3 (z-score 2.2) (B), IFNA2 (z-score 4) (C) and MAPK1 (z-score -4.1) (D). The magnitude of up-regulation for each downstream target is indicated by the intensity of red colouring. Dashed lines represent indirect relationships where orange leads to activation, yellow indicates that the finding is inconsistent with the state of the downstream molecule, and grey indicates that the effect was not predicted. Refer to appendix Figure 7.1 for the full network legend.

163

A

B C

Figure 5.5 IRF3 is activated by multiple upstream transcriptional regulators in the monocyte response to E. coli. IPA-generated mechanistic networks illustrating the relationships between IRF3 and TLR3/RNF216 (A), DDX58 (B), Interferon alpha (C) and the up-regulated downstream targets of these transcriptional regulators, specific to the neonatal response to E. coli. The magnitude of up-regulation for each downstream target is indicated by the intensity of red colouring. Dashed and solid lines represent indirect and direct relationships respectively; where orange leads to activation and grey indicates that the effect was not predicted. Refer to appendix Figure 7.1 for the full network legend.

164

5.3.3 Network analysis

To further characterise the E. coli-specific response genes, network analysis was performed to identify the most biologically significant network nodes/interactions. A protein-protein interaction network containing the 327 E. coli-specific response genes and their first-order interacting partners (based on experimental evidence) was generated using the HINT (High-quality INTeractomes) database. HINT is a regularly updated database that is filtered following curation to remove inaccurate or low-quality interactions The first-order interaction network contained ~2000 nodes and therefore sub-network analysis was performed to enrich for interactions between the most significantly differentially expressed genes (see section 2.8.2 for methodology). The resulting sub-network contained 166 nodes, and the top ten most interconnected nodes (hub genes) were identified (Table 5.5). Only one of the top ten hubs (PRKCD) was from the list of E. coli-specific response genes. A diagrammatic representation of this sub-network is presented in Figure 5.6.

Table 5.5 The top ten sub-network hubs regulating the neonatal monocyte response specific to E. coli Network Present within the Average Up- or down- Hub degree* hub DEG list? p-value† regulated EGFR 28 No - - NR3C1 13 No - - CEBPB 13 No - - STAT1 12 No - - SP1 11 No - - CDC37 10 No - - NFKB1 10 No - - MAPK14 9 No - - PRKCD 9 Yes 1.55E-03 Up PRKCZ 9 No - - *Degree is the number of connections (indirect and direct) to other genes/molecules. †Average p-value (adjusted for multiple comparisons) for differential expression between unstimulated and E. coli-stimulated monocytes across the three infant groups. DEG, differentially expressed gene.

165

Figure 5.6 The most significant sub-network within the E. coli-specific neonatal monocyte response. A first-order protein-protein interaction network was generated from the 327 E. coli-specific response genes using the HINT database. The most significant sub-network (presented here) was then enriched to identify the most biologically significant network nodes/interactions. Node size is relative to hub degree (connectivity). For ease of visualisation only nodes with hub degree ≥2 are labelled. Nodes are coloured based on gene expression where red and blue indicate up- and down-regulation respectively (colour intensity is propotional to magnitude of gene expression). White nodes represent genes that were not included in the initial list of 327 differentially expressed genes.

166

5.4 Characterising the neonatal monocyte transcriptional response specific to S. epidermidis

The top twenty most significantly differentially expressed genes of the 407 S. epidermidis-specific response genes are presented in Table 5.6. In contrast to the E. coli-specific monocyte response, the majority of the most significantly differentially expressed S. epidermidis-specific response genes were down-regulated. Surprisingly, analysis using the full list of 407 S. epidermidis-specific response genes using IPA did not identify any significantly over-represented canonical pathways, disease classifications or bio-functions.

Table 5.6 The top twenty most significantly differentially expressed genes specific to the neonatal monocyte response to S. epidermidis Significance (p-value)* Up- or down- Gene symbol Preterm HCA- Preterm HCA+ Term infants regulated infants infants KCTD12 4.42E-17 3.27E-18 1.09E-13 Down HNRNPLL 3.23E-15 1.85E-12 1.90E-12 Down ANGEL2 4.54E-12 7.79E-12 3.26E-11 Down ORAI2 1.98E-14 7.04E-13 7.74E-11 Down HSPA8 1.05E-14 3.29E-12 1.93E-10 Up UBN2 1.68E-11 1.54E-11 1.86E-09 Down RPS6KA1 1.83E-11 3.62E-09 4.77E-11 Down ADNP 8.47E-13 2.76E-09 3.94E-09 Up KAT7 1.83E-09 1.14E-10 5.05E-09 Down BTBD10 9.09E-11 1.03E-08 1.29E-08 Down NRDE2 2.54E-10 7.28E-09 1.65E-08 Down TICAM2 1.55E-10 9.67E-09 3.09E-08 Down FASTKD5 1.35E-09 2.80E-09 3.70E-08 Up TMED7- 2.08E-10 1.21E-08 3.44E-08 Down TICAM2 RAB9A 9.78E-09 1.89E-08 2.88E-08 Down TNS3 3.12E-10 6.84E-08 1.41E-08 Up DPH1 4.52E-11 1.48E-07 1.84E-10 Up TRIM8 4.85E-10 1.76E-07 2.26E-08 Down NBR1 1.18E-09 4.14E-10 2.26E-07 Down FBXO8 1.71E-07 9.11E-09 2.29E-07 Down *P-values (adjusted for multiple comparisons) for differential expression between unstimulated and S. epidermidis-stimulated monocytes. HCA, histologic chorioamnionitis.

167

5.4.1 Upstream regulator analysis

Only three upstream transcriptional regulators were significantly associated with the 407 S. epidermidis-specific response genes, one of which (mir-10) was also associated with the E. coli-specific monocyte response (Table 5.7). The most significant upstream regulator (mir-223) was only linked to three downstream targets (FBXW7, SP3 and KIF1BP) and therefore is unlikely to be of prominent biological significance as it affects <1% of the S. epidermidis-specific response genes. Only one transcriptional regulator was predicted to be inhibited (HSF1, z-score -2) and none were predicted to be activated. The networks of downstream targets of SUZ12 (the only S. epidermidis- specific significant regulator associated with >5 downstream targets) and HSF1 are presented in Figure 5.7.

Table 5.7 Upstream transcriptional regulators significantly associated with the S. epidermidis-specific neonatal monocyte response p-score* Upstream regulator SE-specific genes EC-specific genes mir-223† 2.6 ns mir-10† 2.4 2.8 SUZ12 2.0 ns

*p-score = –log10 (p-value). Values above 2 are considered statistically significant (p<0.01). †Regulator affects less than five downstream targets. SE, S. epidermidis; EC, E. coli; ns, non- significant.

A B

Figure 5.7 Networks for the significant S. epidermidis-specific upstream transcriptional regulators. Networks of the downstream targets of SUZ12 (A) and HSF1 (z-score -2) (B). The magnitude of up- or down-regulation for each downstream target is indicated by the intensity of red or green colouring respectively. Dashed and solid lines represent indirect and direct relationships respectively, where orange leads to activation and grey indicates that the effect was not predicted. Refer to appendix Figure 7.1 for the full network legend.

168

5.4.2 Network analysis

To further characterise the S. epidermidis-specific response genes, network analysis was performed to identify the most biologically significant network nodes/interactions. A protein-protein interaction network containing the 407 S. epidermidis-specific response genes and their first-order interacting partners (based on experimental evidence) was generated using the HINT database. This network contained ~2000 nodes and therefore sub-network analysis was performed to enrich for interactions between the most significantly differentially expressed genes (see section 2.8.2 for methodology). The resulting sub-network contained 181 nodes, and the top ten most interconnected nodes (hub genes) were identified (Table 5.8). None of the top ten hubs were from the list of S. epidermidis-specific response genes. A diagrammatic representation of this sub-network is presented in Figure 5.8.

Table 5.8 The top ten sub-network hubs regulating the neonatal monocyte response specific to S. epidermidis Network Hub Present within the DEG Average Up- or down- hub degree* list? p-value† regulated GRB2 33 No - - ESR1 16 No - - ESR2 16 No - - ERBB2 12 No - - KIT 12 No - - ZDHHC17 12 No - - PLCG1 10 No - - CRKL 9 No - - HDAC4 9 No - - CDC37 8 No - - *Degree is the number of connections (indirect and direct) to other genes/molecules. †Average p-value (adjusted for multiple comparisons) for differential expression across the three infant groups. DEG, differentially expressed gene.

169

Figure 5.8 The most significant sub-network within the S. epidermidis-specific neonatal monocyte response. A first-order protein-protein interaction network was generated fromthe 407 S. epidermidis- specific response genes using the HINT database. The most significant sub-network (presented here) was then enriched to identify the most biologically significant network nodes/interactions. Node size is relative to hub degree (connectivity). For ease of visualisation only nodes with hub degree ≥2 are labelled. Nodes are coloured based on gene expression where red and blue indicate up- and down- regulation respectively (colour intensity is propotional to magnitude of gene expression). White nodes represent genes that were not included in the initial list of 407 differentially expressed genes.

170

5.5 Characterising the conserved neonatal monocyte transcriptional response to E. coli and S. epidermidis

Comparisons of the genes differentially expressed in response to both stimuli identified 6511 conserved pathogen response genes across all infant groups (Figure 5.1D). The thirty most significant differentially expressed genes (average across all groups) are presented in Table 5.9, twenty-eight of which were up-regulated.

5.5.1 Over-represented canonical pathways, diseases or bio-functions

IPA analyses are optimal for gene lists between 100–2000 entries. As the top 2000 most significantly differentially expressed genes within the larger list of 6511 genes was variable across infant groups and between stimuli, all IPA analyses were performed independently on the top 2000 differentially expressed genes by preterm HCA-, preterm HCA+ and term monocytes in response to E. coli or S. epidermidis (six independent analyses). The majority of the top 2000 differentially expressed genes were conserved across all infant groups in response to both stimuli (1214 genes).

There were 105 over-represented canonical pathways, and 51 over-represented disease classifications and bio-functions with an average FDR adjusted p-value <0.05 across all infant groups. The top twenty of which are presented in Table 5.10 and Table 5.11 respectively. Many of the top canonical pathways are central to the monocyte inflammatory response to infection (IL-6, TLR, IL-10, TNF, NF-κb, CD40 and iNOS signalling). In contrast, many of the top over-represented bio-functions were generally related cellular response and maintenance mechanisms.

171

Table 5.9 The top thirty most significantly differentially expressed genes within the conserved neonatal monocyte response to E. coli and S. epidermidis Significance (p-value adjusted for multiple comparisons) Gene symbol E. coli S. epidermidis Up- or down- Average Preterm HCA- Preterm HCA+ Term Preterm HCA- Preterm HCA+ Term regulated MAP2K3 8.04E-29 4.40E-28 4.31E-28 6.14E-30 4.98E-29 2.54E-29 1.72E-28 Up ATP2B1 1.07E-26 3.03E-26 1.25E-26 4.41E-28 6.42E-27 8.31E-28 1.02E-26 Up USP12 2.30E-27 2.60E-27 6.42E-26 3.57E-28 4.48E-27 1.11E-26 1.42E-26 Up PPP1R15A 2.30E-27 2.79E-27 1.53E-25 4.41E-28 2.28E-26 8.51E-26 4.45E-26 Up DUSP1 1.64E-27 6.25E-28 2.75E-25 7.55E-29 2.54E-27 5.32E-26 5.55E-26 Up KDM6B 1.03E-26 7.60E-27 3.51E-25 2.39E-29 3.80E-28 6.34E-28 6.16E-26 Up BHLHE40 1.03E-26 6.19E-27 3.51E-25 3.21E-27 1.02E-26 8.76E-26 7.80E-26 Up SLC7A5 1.03E-26 3.60E-25 1.53E-25 3.41E-28 6.08E-26 1.11E-26 9.94E-26 Up DDIT4 1.31E-27 8.02E-28 1.32E-25 5.63E-27 4.75E-26 7.52E-25 1.56E-25 Up SERPINB8 3.38E-26 2.86E-25 1.07E-24 1.46E-27 6.08E-26 1.91E-26 2.46E-25 Up PLK3 5.80E-26 6.49E-26 1.28E-24 8.23E-27 1.21E-25 1.30E-25 2.78E-25 Up MAFF 2.84E-26 6.44E-26 2.34E-24 1.14E-28 2.67E-27 1.38E-26 4.07E-25 Up TNFAIP8 9.14E-25 2.32E-26 1.11E-24 3.21E-25 1.45E-25 1.07E-24 5.98E-25 Up RABGEF1 1.27E-26 5.95E-26 5.26E-24 3.17E-28 4.48E-27 1.10E-25 9.07E-25 Up SNX9 2.08E-25 1.72E-24 9.08E-24 3.33E-26 3.83E-25 2.62E-24 2.34E-24 Up EIF1 8.59E-26 4.34E-25 2.16E-23 6.77E-29 6.42E-27 7.79E-27 3.69E-24 Up NDEL1 1.48E-25 2.11E-24 1.96E-23 2.43E-26 8.75E-25 1.67E-24 4.07E-24 Up CSK 5.95E-25 1.34E-23 3.29E-24 5.35E-25 5.00E-24 3.28E-24 4.35E-24 Down TMEM2 1.27E-24 1.29E-23 2.98E-23 1.90E-25 1.27E-23 5.71E-24 1.04E-23 Up ETS2 1.73E-24 3.39E-24 5.58E-23 8.16E-26 1.13E-24 3.28E-24 1.09E-23 Up NR3C1 6.07E-25 1.85E-25 2.34E-24 1.39E-23 1.88E-23 5.41E-23 1.50E-23 Up SDC4 1.33E-23 3.39E-24 8.19E-23 1.61E-25 8.72E-25 9.91E-25 1.68E-23 Up B4GALT5 1.73E-24 6.33E-25 1.21E-23 7.10E-24 3.95E-23 4.51E-23 1.77E-23 Up TBC1D14 7.38E-25 2.30E-24 1.26E-23 5.82E-24 1.27E-23 9.11E-23 2.09E-23 Down NFKB1 5.08E-23 3.39E-24 4.68E-23 6.64E-24 4.29E-24 1.77E-23 2.16E-23 Up SPAG9 4.07E-24 2.13E-23 1.10E-22 8.76E-25 2.45E-23 2.42E-23 3.08E-23 Up LDLR 2.46E-25 1.63E-24 1.86E-22 1.68E-26 1.74E-24 1.60E-23 3.44E-23 Up NFKBIA 1.78E-23 2.11E-24 2.98E-23 1.31E-23 9.75E-23 8.83E-23 4.14E-23 Up BCL2A1 4.16E-24 2.09E-23 2.44E-22 2.08E-25 1.29E-23 3.76E-23 5.32E-23 Up PLAUR 2.10E-24 2.14E-24 3.51E-22 4.49E-27 4.05E-26 6.80E-25 5.93E-23 Up 172

Table 5.10 The top twenty significantly over-represented canonical pathways associated with the conserved neonatal monocyte response to E. coli and S. epidermidis p-score* EC SE Canonical pathway Preterm Preterm Preterm Preterm Average Term Term HCA- HCA+ HCA- HCA+ IL-6 Signalling 11.8 8.1 10.3 9.9 7.8 9.4 9.5 PI3K/AKT Signalling 11.4 9.7 8.9 8.2 9.3 9.6 9.5 Toll-like Receptor 9.9 8.1 9.8 11.2 7.8 9.2 9.3 Signalling IL-10 Signalling 9.9 7.5 9.8 9.8 6.4 9.6 8.8 TNFR2 Signalling 7.8 6.9 7.8 7.9 8.7 9.0 8.0 TNFR1 Signalling 6.7 7.1 6.4 8.0 9.3 9.6 7.8 Molecular Mechanisms 7.3 6.9 6.2 7.3 7.9 8.4 7.3 of Cancer Glucocorticoid Receptor 7.3 5.8 5.5 7.3 7.4 7.7 6.8 Signalling Unfolded protein 5.3 4.5 5.1 7.1 7.8 7.8 6.3 response B Cell Receptor 6.6 5.5 5.5 7.3 5.5 6.9 6.2 Signalling Apoptosis Signalling 6.7 5.0 5.9 6.7 5.5 7.2 6.2 HMGB1 Signalling 7.0 5.5 5.9 6.6 4.8 7.0 6.1 Induction of Apoptosis 5.8 5.0 5.0 6.4 7.6 6.9 6.1 by HIV1 RANK Signalling in 6.9 4.6 5.6 6.9 5.7 6.9 6.1 Osteoclasts PPAR Signalling 6.6 5.2 5.7 6.5 5.3 6.5 6.0 NF-B Signalling 6.7 6.0 5.9 5.4 5.3 5.7 5.8 p38 MAPK Signalling 7.1 6.5 6.4 5.4 5.2 4.4 5.8 TREM1 Signalling 7.4 5.0 5.6 5.8 5.2 5.3 5.7 CD40 Signalling 6.5 4.5 5.1 6.4 5.1 6.4 5.7 Role of IL-17A in 7.0 4.4 5.0 6.4 4.5 6.4 5.6 Arthritis Death Receptor 5.0 4.7 4.8 5.4 6.9 6.4 5.5 Signalling Acute Phase Response 6.7 4.5 5.3 5.8 5.0 5.4 5.5 Signalling iNOS Signalling 6.0 5.0 5.1 7.3 3.9 5.2 5.4 Ceramide Signalling 6.9 4.6 4.2 5.4 5.7 5.3 5.3 LPS-stimulated MAPK 6.1 4.7 5.2 6.0 3.3 5.4 5.1 Signalling

*p-score = –log10 (corrected p-value). Values above 1.3 are considered statistically significant (corresponding to an FDR adjusted p-value <0.05). EC, E. coli; SE, S. epidermidis; HCA, histologic chorioamnionitis.

173

Table 5.11 The top twenty significantly over-represented diseases and bio- functions associated with the conserved neonatal monocyte response to E. coli and S. epidermidis p-score* Disease or Bio- EC SE function Preterm Preterm Preterm Preterm Average Term Term HCA- HCA+ HCA- HCA+ Gene Expression 25.4 23.7 27.4 20.6 24.9 25.8 24.7 Cell Death and 20.4 22.7 19.0 21.4 25.0 25.8 22.4 Survival Cellular Growth 16.1 12.3 13.7 19.7 18.7 18.9 16.6 and Proliferation Cell Cycle 13.6 13.0 12.3 13.6 12.7 13.3 13.1 Cellular 10.1 8.9 8.8 11.1 12.5 13.7 10.9 Development Neurological 8.0 9.6 8.4 13.3 13.4 11.6 10.7 Disease Infectious Diseases 12.4 8.9 9.9 12.7 7.7 12.2 10.6 Embryonic 6.1 8.3 7.1 7.5 8.8 11.8 8.3 Development Cellular Movement 8.6 8.2 7.1 6.3 8.1 7.6 7.6 Cell Morphology 7.0 8.1 8.6 5.9 6.3 7.0 7.1 Cell Signalling 8.3 6.0 6.7 6.6 5.6 8.3 6.9 Dermatological Diseases and 5.6 6.7 6.8 6.7 6.8 6.9 6.6 Conditions Cellular Assembly 6.4 6.7 7.1 4.8 5.6 7.5 6.4 and Organization Post-Translational 7.6 7.0 6.7 5.4 5.9 5.1 6.3 Modification Cellular Function 6.4 5.9 6.3 5.2 6.3 5.0 5.9 and Maintenance Cell-To-Cell Signalling and 6.7 5.8 6.5 4.8 4.7 6.0 5.8 Interaction Cardiovascular System 5.8 5.9 4.7 4.5 6.7 4.5 5.3 Development and Function Organismal 5.8 5.9 4.7 4.5 6.7 4.5 5.3 Development Tissue 4.8 3.4 4.2 6.7 5.2 6.2 5.1 Development Inflammatory 4.8 4.5 4.8 6.5 4.3 5.4 5.0 Disease

*p-score = –log10 (p-value). Values above 1.3 are considered statistically significant (corresponding to an FDR adjusted p-value <0.05). EC, E. coli; SE, S. epidermidis; HCA, histologic chorioamnionitis.

174

5.5.2 Upstream regulator analysis

Upstream regulator analysis performed within IPA identified a total of 347 transcriptional regulators with an average p-value <0.01 (threshold for significance) across all groups. The top twenty are presented in Table 5.12. This analysis differs from that performed in Chapter 4 (section 4.6.2.2) in that this analysis was restricted to the top 2000 most significantly differentially expressed genes within the larger pool of 6511 conserved pathogen response genes. The vast majority of the transcriptional regulators with a significant activation z-score were predicted to be activated (113 out of 130), whereas only 17 were predicted to be inhibited (Figure 5.9).

Table 5.12 The top twenty significant upstream transcriptional regulators associated with the conserved neonatal monocyte response to E. coli and S. epidermidis p-score* Upstream EC SE regulator Preterm Preterm Preterm Preterm Average Term Term HCA- HCA+ HCA- HCA+ TREM1 26.2 24.6 24.6 29.8 25.3 25.3 26.0 NR3C1 24.4 23.9 25.2 22.8 22.1 26.2 24.1 TNF 19.3 21.6 21.6 22.7 20.9 24.2 21.7 PDGF BB 20.5 19.5 17.5 22.7 21.6 22.6 20.7 NUPR1 15.4 18.8 13.7 16.9 20.2 16.2 16.9 NFκB (complex) 14.6 14.1 17.2 12.4 13.4 15.2 14.5 TP53 12.4 12.5 12.5 13.0 13.7 12.8 12.8 F7 12.1 12.1 13.5 12.1 10.8 12.1 12.1 FSH 11.8 10.3 11.3 12.4 12.3 14.4 12.1 RELA 9.5 9.6 9.0 9.6 8.9 11.4 9.7 KRAS 8.1 9.9 8.1 9.9 10.9 10.9 9.6 PGR 10.5 9.2 9.2 8.6 9.1 8.5 9.2 IL1A 8.7 10.3 8.0 8.8 9.5 9.5 9.1 CD40LG 10.0 7.9 9.3 8.6 8.5 9.2 8.9 P38 MAPK 7.8 10.2 9.0 9.0 9.5 7.7 8.8 Lh 9.5 7.1 8.5 9.6 8.0 10.0 8.8 SELPLG 8.7 8.8 8.8 8.8 7.7 8.7 8.6 Pkc(s) 8.1 7.3 8.9 8.9 8.1 9.7 8.5 TLR7 8.7 8.7 8.8 7.9 7.9 8.7 8.5 JUN 6.9 10.2 9.5 7.0 7.6 8.8 8.3

*p-score = –log10 (p-value). Values above 2 are considered statistically significant (p<0.01). EC, E. coli; SE, S. epidermidis; HCA, histologic chorioamnionitis.

175

A B TNF NUPR1 TREM1 SMARCA4 IL1B IL1A PDGF BB NFkB (complex) CD40L Mek CSF2 TLR7 Cg F7 RELA Jnk PPRC1 IL2 MAPK1/2 CCL5 FOXL2 EGFR SYVN1 IL3 A2M ERK EZH2 TLR2 PTGS2 TLR4 Fcer1 IL17A

CEBPA IFNG IL5 ERBB2 CD40 AR PAF1 SELPLG P38 MAPK TLR3 ELF4 FOXO4 STAT3 FOXO1 NFKB1 MGEA5 LY6E NEUROG1 mir-122-5p SCD CD3 PPP2R5C NPPB CD28 IL1RN TAB1 RBM5 JAG2 mir-155-5p Estrogen receptor IgG COL18A1 -8 -6 -4 -2 0 2 4 6 8 0 10 20 30

Activation z-score -log10 (p-value)

Figure 5.9 Upstream regulator analysis of the conserved neonatal monocyte response to E. coli and S. epidermidis. Upstream regulator analysis was performed on the 2000 most significant differentially expressed genes within the 6511 conserved genes commonly differentially expressed in response to E. coli and S. epidermidis across all infant groups (analysis repeated six times as the top 2000 genes varied slightly across stimuli and infant group). The seventeen inhibited (blue) and the top forty-seven activated (yellow) transcriptional regulators are shown. Data are presented as mean ± SEM.

176

5.5.3 Network analysis

As in sections 5.3.3 and 5.4.2, network analysis was performed on the 6511 conserved pathogen response genes by generating a first-order protein-protein interaction network (containing ~9000 nodes) using the HINT database (see section 2.8.2 for methodology). Sub-network analysis was performed to enrich for interactions between the most significantly differentially expressed genes, resulting in a network containing 268 nodes. The top twenty most interconnected nodes (hub genes) are presented in Table 5.13. A diagrammatic representation of this sub-network is presented in Figure 5.10.

Table 5.13 The top twenty sub-network hubs regulating the conserved neonatal monocyte response to E. coli and S. epidermidis Network Hub Present within the DEG Average Up- or down- hub degree* list? p-value† regulated TP53 36 Yes 6.82E-13 Down EGFR 35 No - - JUN 26 Yes 1.55E-05 Up IKBKG 24 No - - VCAM1 23 No - - RELA 21 Yes 1.58E-13 Up MAPK1 21 Yes 7.40E-10 Down SRC 18 Yes 1.92E-13 Up YWHAZ 18 Yes 4.20E-13 Up HDAC2 17 No - - MAPK3 16 Yes 4.73E-09 Down ESR2 16 No - - YWHAG 16 No - - MAP3K7 15 No - - TRAF1 15 Yes 2.78E-19 Up BCL2 12 Yes 4.03E-15 Up ZDHHC17 12 No - - NR3C1 12 Yes 1.50E-23 Up MAPK8 12 Yes 9.66E-13 Up RIPK1 12 No - - *Degree is the number of connections (indirect and direct) to other genes/molecules. †Average p-value (adjusted for multiple comparisons) for differential expression across the three infant groups. DEG, differentially expressed gene.

177

Figure 5.10 The most significant sub-network within the conserved neonatal monocyte response to E. coli and S. epidermidis. A first-order protein-protein interaction network was generated from the 6511 conserved pathogen response genes using the HINT database. The most significant sub-network (presented here) was then enriched to identify the most biologically significant network nodes/interactions. Node size is relative to hub degree (connectivity). For ease of visualisation only nodes with hub degree ≥5 are labelled. Nodes are coloured based on gene expression where red and blue indicate up- and down-regulation respectively (colour intensity is propotional to magnitude of gene expression). White nodes represent genes that were not included in the initial list of 6511 differentially expressed genes.

178

An alternative approach to network analysis was also performed using the complete lists of differentially expressed genes in response to E. coli or S. epidermidis (see section 2.8.2 for methodology). Briefly, the global sets of differentially expressed genes by monocytes from each infant group in response to each stimulus were overlaid onto the entire HINT curated human interactome, and the network was filtered to only contain interactions involving the differentially expressed genes (261). From this larger network, highly interconnected sub-networks (~150 nodes) containing the most significantly differentially expressed genes were enriched using the BioNet R package and the full heinz algorithm (263). This analysis was performed independently on each infant group, and responses to E. coli and S. epidermidis were analysed separately. This approach provided an alternative method for comparisons between infant monocyte responses to E. coli and S. epidermidis allowing identification of the most significant conserved, and pathogen-specific response genes within one analysis.

The results between infant groups for each stimulus were highly similar (p<0.001 for all pair-wise comparisons using Spearman correlations, not shown) therefore node degrees were averaged across infant groups, and the top twenty most highly interconnected nodes were identified within the E. coli and S. epidermidis-induced sub-networks (Table 5.14 and Table 5.15 respectively). As the infant group responses were highly similar, the top sub-networks were merged across infant groups creating one representative sub- network for each stimulus. A comparison of the representative E. coli and S. epidermidis sub-networks revealed that 74 nodes were common to both (40 of which were also present within the sub-network generated in the previous analysis; Figure 5.10). The top twenty of these 74 nodes in order of decreasing hub degree were TRAF1, TP53, NFKB1, MAPK6, REL, CDKN1A, NR3C1, NFKB2, PPP1CC, FLNA, NFKBIA, TANK, RIPK2, TNIP2, TNFAIP3, TAB1, RIF1, PSEN1, FOXO3 and TAB2. A network representation of the top twenty common nodes is presented in Figure 5.11. This comparison also identified fourteen nodes that were unique to the representative E. coli sub-network, two of which were among the top twenty (JUNB and IRAK1) and eight nodes that were unique to the representative S. epidermidis sub-network, again two of which were among the top twenty (JUN and FBXW11). Network representations of the E. coli and S. epidermidis-unique sub-network nodes are presented in Figure 5.12A and B respectively. The first neighbours of each stimulus-unique node were also included to add biological context to each network.

179

Table 5.14 The top 20 nodes conserved across all infant groups in the E. coli response network Node degree Average Up- or down- Node Preterm HCA- Preterm HCA+ Term p-value* regulated TRAF1 9.3 8.7 10.1 3.59E-19 Up SRC 9.3 8.7 8.8 4.61E-14 Up TP53 6.6 8.1 8.8 1.46E-14 Down MAPK1 8.6 7.4 7.4 1.47E-09 Down MAPK14 6.6 6.0 7.4 4.46E-17 Down NFKB1 5.3 6.0 6.8 3.37E-23 Up MAPK6 5.3 5.4 6.1 1.91E-21 Up YWHAB 4.6 5.4 5.4 2.15E-11 Up REL 4.6 4.7 5.4 2.21E-21 Up †JUNB 4.0 4.7 4.1 2.09E-16 Up †IRAK1 4.0 4.0 4.1 2.48E-14 Up CDKN1A 3.3 4.0 4.1 2.18E-19 Up NR3C1 3.3 4.0 4.1 1.04E-24 Up FLNA 3.3 3.4 3.4 3.28E-16 Up PPP1CC 3.3 3.4 3.4 2.13E-14 Down NFKB2 3.3 2.7 4.1 1.14E-19 Up TNIP1 4.0 2.7 2.7 1.03E-17 Up RIPK2 2.6 2.7 3.4 7.42E-21 Up TANK 2.6 2.7 3.4 3.61E-13 Up TNIP2 2.6 2.7 3.4 1.33E-19 Up *Average p-value (adjusted for multiple comparisons) for differential expression across the three infant groups. †Unique to the E. coli response network.

Table 5.15 The top 20 nodes conserved across all infant groups in the S. epidermidis response network Node degree Average Up- or down- Node Preterm Preterm HCA- Term p-value* regulated HCA+ †JUN 10.6 10.1 9.5 9.54E-12 Up TP53 9.3 8.1 7.5 1.35E-12 Down TRAF1 8.6 6.8 8.8 1.98E-19 Up NFKB1 6.0 5.4 7.5 9.53E-24 Up CSNK2B 5.3 5.4 6.8 4.21E-06 Up REL 4.6 4.1 5.4 1.09E-22 Up MAPK6 5.3 4.7 4.1 5.31E-22 Up CDKN1A 3.3 4.7 4.1 8.14E-21 Up NR3C1 4.0 4.1 3.4 2.89E-23 Up NFKBIA 3.3 2.7 4.1 6.63E-23 Up NFKB2 4.0 2.7 3.4 1.96E-20 Up PPP1CC 3.3 2.7 3.4 1.12E-15 Down †FBXW11 3.3 3.4 2.7 1.19E-20 Up FLNA 2.0 2.7 3.4 5.27E-18 Up RIF1 2.6 2.7 2.7 2.79E-17 Up SETDB1 2.6 2.7 2.7 3.33E-15 Down TANK 2.6 2.0 2.7 2.75E-15 Up RIPK2 2.0 2.0 2.7 1.30E-20 Up TAB1 2.0 2.0 2.7 4.94E-16 Down TNFAIP3 2.0 2.0 2.7 5.68E-20 Up *Average p-value (adjusted for multiple comparisons) for differential expression across the three infant groups. †Unique to the S. epidermidis response network.

180

Figure 5.11 The top 20 nodes common across the most significant E. coli and S. epidermidis response sub-networks. Node size is relative to degree (interconnectivity). Nodes are coloured based on normalised log2 fold change in gene expression relative to unstimulated monocytes.

181

Figure 5.12 Nodes unique to the E. coli or S. epidermidis response. (A) Fourteen nodes (green) that are unique to the most significant E. coli response sub-network and their first neighbours. (B) Eight nodes (yellow) that are unique to the most significant S. epidermidis response sub-network and their first neighbours. Nodes labelled in black represent those that are common to both sub-networks. Nodes labelled in grey are not exclusive to either response, or common to both sub- networks. Node size is relative to degree (interconnectivity). Nodes are coloured based on log2 fold change in gene expression relative to unstimulated monocytes.

182

5.6 Discussion

This chapter aimed to identify and characterise the pathogen-specific, and conserved neonatal monocyte responses to E. coli and S. epidermidis; two of the most important sepsis-causing pathogens in preterm infants. We have demonstrated that monocytes from both preterm and term infants exhibit a largely conserved response to E. coli and S. epidermidis (6511 genes), with only a relatively small subset of differentially expressed genes specific to either pathogen (≤ 407 genes). The conserved pathogen response genes were associated with antimicrobial and inflammatory mechanisms, pattern-recognition receptor signalling and morphological processes. E. coli specifically elicited the transcription of genes typically involved in anti-viral immunity (type 1 interferons), which were not differentially expressed followings stimulation with S. epidermidis. In contrast, the subset of genes specific to S. epidermidis infection were not associated with any pre-identified canonical pathways or bio-functions.

Bioinformatic analyses on the top 2000 most significant conserved pathogen response genes revealed significant associations with antimicrobial/inflammatory canonical pathways (IL-6, IL-10, TLR, TNFR, NF-κb, p38 MAPK and iNOS signalling). This is perhaps not surprising given that engagement of TLR signalling and the subsequent transcription of inflammatory cytokines and chemokines is a major end point of bacterial recognition (300). In addition to antimicrobial/inflammatory pathways, several bio-functions associated with cell morphology were significantly over-represented within the top 2000 conserved pathogen response genes (cell cycle, movement, growth, proliferation, development, maintenance, assembly and organisation). Furthermore, sub- network analysis on the 6511 conserved pathogen response genes identified VCAM1 as the fifth most highly interconnect hub. VCAM1 is expressed by endothelial cells (and up-regulated during inflammation), and acts as a scaffold for monocytes to infiltrate inflamed/infected tissues (314). These findings suggest that neonatal monocytes may be primed to interact with the tissue microenvironment following bacterial challenge in the blood.

No other studies have assessed the neonatal monocyte transcriptional response to live E. coli or S. epidermidis, however in keeping with our findings, human monocyte-derived macrophages exhibit a conserved response to a range of eight bacterial species including Gram-positive, Gram-negative and mycobacterial organisms determined by microarray (243). This conserved response was considerably smaller (191 genes) than that observed 183 in our study (most likely reflecting the criteria to be differentially expressed in response to all eight pathogens) yet was still associated with an inflammatory response as well as genes involved in adhesion, signalling and transcription (243). Together these findings suggest that human monocytes exhibit a generalised inflammatory/morphological response to infection that is retained upon differentiation into mature macrophages.

Of note, alternative pattern recognition receptor (PPR) pathways to the TLRs were represented in the conserved neonatal monocyte response to live E. coli and S. epidermidis. The NOD family of receptors is crucial for the cytosolic detection of microbial components. In particular, NOD1 and NOD2 initiate inflammation through activation of NF-κB and MAP kinases following recognition of γ-d-glutamyl-meso- diaminopimelic acid primarily found in Gram-negative bacteria and muramyl dipeptide, a common peptidoglycan found in Gram-negative and Gram-positive bacteria (315). The serine/threonine kinase RIPK2 is an adaptor molecule required for both NOD1 and NOD2 activation of NF-κb (316). RIPK2 was significantly up-regulated by monocytes from all infant groups in response to both E. coli and S. epidermidis, and was among the top twenty most highly interconnected network nodes in the conserved monocyte response to both stimuli. Additionally BIRC3, which is required for the polyubiquitination of RIPK2 (317), was also significantly up-regulated by monocytes from all infant groups in response to both E. coli and S. epidermidis and present within the conserved network response (ranked 21 out of 74 nodes).

In support of this finding, Strunk et. al. found that the treatment of human whole blood with monoclonal antibodies that block PAMP binding to TLR2 and TLR4 reduced, but did not abolish the production of IL-6 in response to live S. epidermidis (WT 1457) or LPS (109). The authors suggest that cytosolic pathways such as NOD-like receptor signalling could mediate TLR-independent production of IL-6. There is limited data assessing NOD1 and NOD2 expression and function in preterm infant mononuclear cells but one study by Granland et. al. demonstrates that these receptors are expressed on monocytes early in gestation and are functional (191). Collectively our findings suggest that NOD signalling may contribute to the neonatal monocytes response to live E. coli and S. epidermidis, highlighting that there may be some redundancy between TLR and NLR signalling during Gram-positive or Gram-negative bacterial infection.

184

In order to gain insight into mechanisms that may govern the major changes in gene expression observed following monocyte stimulation with E. coli and S. epidermidis, we performed IPA upstream regulator analysis on the top 2000 most significantly differentially expressed genes in response to both stimuli. This analysis identified over 300 significant upstream transcriptional regulators, highlighting the complexity of the conserved monocyte response to bacterial infection. Triggering Receptor Expressed on Myeloid cells (TREM)-1 was the most significant of these, and was predicted to be activated. TREM-1 is a member of the Ig superfamily and is highly expressed by neonatal monocytes (318). Peptidoglycan recognition protein 1 (expressed on polymorphonuclear leukocytes) has recently been identified as a potent ligand for TREM-1 (either multimerised, or complexed with peptidoglycan), the activation of which results in enhanced inflammatory cytokine production (319). Bacterial ligands for TREM-1 have not yet been identified, however TREM-1 expression is up-regulated on human monocytes following stimulation with whole bacteria (S. aureus, Pseudomonas aeruginosa) (320), and the up-regulation of monocyte TREM-1 expression by S. aureus cell wall constituents is attenuated when TREM-1 ligands were bound by LP17 (synthetic peptide mimicking a highly conserved domain of TREM-1) (321), suggesting bacterial ligands for TREM-1 exist. While the precise signalling events initiated by TREM-1 ligation have not been elucidated, TREM-1 signalling alone can activate NF- κB to induce inflammatory cytokine production, but can also act synergistically with TLR and NLR signalling pathways to amplify inflammation (322). The role of TREM-1 in the amplification of inflammation is thought to contribute to the pathophysiology of sepsis. Human monocytes are major producers of the soluble form of TREM-1 (323), the peripheral blood plasma concentrations of which are significantly increased in neonates with sepsis (324), and significantly higher in neonates that progress to septic shock/death (325). Blockade of TREM-1 signalling with a soluble TREM-1-IgG fusion protein in mice is protective against LPS-induced septic shock and lethal E. coli peritonitis (320). Our finding that TREM-1 was the most significant upstream transcriptional regulator associated with the conserved pathogen response genes supports a role for TREM-1 signalling in neonatal monocytes during bacterial infection, and may be an attractive target for therapeutic intervention in sepsis.

The conserved neonatal monocyte response to E. coli and S. epidermidis infection also exhibited several mechanisms for the regulation of inflammation, which is relevant given that the dysregulation of inflammation may contribute to the pathogenesis of

185 sepsis (95). PI3K-AKT signalling was the second most significantly over-represented canonical pathway within the conserved neonatal monocyte response. This pathway affects a broad range of cellular functions but also plays a role in the regulation of TLR signalling in human monocytes via suppression of ERK1/2 and NF-κb p65 leading to decreased production of IL-12 and increased production of IL-10 (326). Network analysis also revealed that six of the top twenty conserved E. coli and S. epidermidis response network hubs were involved in negative regulation of TLR signalling (TRAF1, NFKBIA, TANK, TNFAIP3, TNIP2, and FOXO3, all up-regulated). In fact, TRAF1 was the most highly interconnected node within the conserved neonatal monocyte response network. TRAF1 negatively regulates TRIF-dependent TLR4 signalling by preventing TRIF-induced activation of NFκB and IRF3 (327, 328). In the context of E. coli infection in our model, the interaction between TRAF1 and TRIF may serve as a feedback inhibition loop for the production of inflammatory cytokines. S. epidermidis also induces up-regulation of TRAF1 in keratinocytes, which limits inflammation through inhibition of TLR3 (TRIF-dependent) signalling (329). This effect is mediated through TLR2 recognition of lipoteichoic acid on the cell surface of S. epidermidis, highlighting crosstalk between TLR signalling pathways (330).

In addition to characterising the conserved neonatal monocyte response to E. coli and S. epidermidis, we also identified a subset of genes that were specifically differentially expressed in response to each pathogen. Bioinformatic analyses of the E. coli-specific differentially expressed genes found repeated associations with anti-viral/type I interferon immune pathways, with IRF3 implicated in the transcriptional regulation of these genes. In contrast, the differentially expressed genes specific to S. epidermidis stimulation were not collectively associated with any canonical pathways, diseases or bio-functions, making it difficult to characterise this response in a biologically meaningful way. The association of an anti-viral gene program exclusively with E. coli stimulation is likely due to the induction of MyD88-independent TLR4 signalling, where TRIF-activation of IRF3 initiates the transcription of genes typically involved in anti-viral immunity such as IFNB, IFIT1 and ISG15 (331, 332). The marked lack of biological context pertaining to the S. epidermidis-specific differentially expressed genes may be reflective of the fact that S. epidermidis normally maintains a benign relationship with the host, and has not evolved to be a pathogen (63). Alternatively, as the responses induced by S. epidermidis are generally less well studied than those

186 induced by E. coli, this result may highlight a limitation of using databases such as IPA, which rely on pre-existing knowledge of molecular interactions.

Interestingly, sub-network analysis on the global set of differentially expressed genes induced by E. coli or S. epidermidis identified differences in some of the major hub genes. The most notable difference was the presence of JUN (c-JUN) as the most highly interconnected hub within the S. epidermidis-induced sub-network, whereas the paralog JUNB was among the top hubs within the E. coli-induced sub-network. JUN and JUNB are both components of the transcription factor AP-1 that regulates a broad range of cellular responses, however they exert opposing effects on the progression of cell cycle (JUN is a positive regulator and JUNB is a negative regulator), and the pro-apoptotic activity of JUN is not observed for JUNB (333). The suppression of cell cycle and apoptosis may be favourable to E. coli as a pathogen, and assist in evasion of the immune system. Indeed, cytolethal distending toxins (intracellular-acting bacterial protein toxins), which are produced by a range of Gram-negative pathogens including E. coli, and are best characterised for their ability to induce cell cycle arrest (334).

In conclusion, our results demonstrate that the neonatal monocyte transcriptional response to E. coli and S. epidermidis (two structurally and biologically diverse pathogens) is largely conserved, is present early in gestation and is not altered by prenatal exposure to HCA. This transcriptional response was significantly associated with PRR signalling and inflammation (as well as mechanisms to regulate these pathways) and cell morphology, demonstrating that neonatal monocytes are capable of responding to infection, and become primed to interact with/infiltrate into tissues. Contrary to previous ideology that healthy pregnancy is associated with a sterile intrauterine environment, evidence is emerging that the placenta hosts a range of microorganisms even in healthy term pregnancy (335). Therefore, a conserved monocyte response to infection (as identified here) may be established early in gestation to protect the developing fetus in utero from bacterial species that colonise the placenta, which can potentially invade the amniotic cavity. We also demonstrate distinct transcription of a subset of genes between E. coli- and S. epidermidis-stimulated monocytes, highlighting that neonatal monocytes elicit pathogen-specific immune responses.

187

188

Chapter 6 General discussion

Preterm birth is an increasing and significant global public health concern. Preterm infants are extremely vulnerable to life-threatening bacterial infections and collectively this population accounts for 75% of all neonatal morbidity and mortality. E. coli and S. epidermidis are the predominant causes of early- and late-onset sepsis among preterm infants in the developed world, respectively, and therefore have been the focus of this thesis. Low gestational age is the most important risk factor for sepsis and this has generally been attributed to an immaturity of immune defences. Indeed, evidence suggests that innate immunity matures continuously during gestation and hence, more extreme prematurity correlates inversely with the risk of both early and late-onset sepsis (reviewed in section 1.4.1). Neutrophils are the most abundant innate immune leukocyte, and several quantitative neutrophil deficiencies are associated with low gestational age (section 1.4.1.2). Attempts to restore the numbers of circulating neutrophils in very preterm infants have been the focus of several clinical trials aimed at preventing or limiting sepsis in recent times, yet neither the administration of growth factors such as G-CSF/GM-CSF (to promote neutrophil maturation) nor even granulocyte transfusions, has had any impact on neonatal sepsis (155-158). Therefore, the role of other innate leukocytes in defence against infection is warranted.

Although present at lower frequencies in circulation than neutrophils, monocytes represent a critical blood defence against bacterial infection. Monocytes from preterm infants exhibit impaired pro-inflammatory cytokine responses following whole bacterial or specific ligand stimulation, despite comparable levels of bacterial phagocytosis and phosphorylation of downstream pattern recognition receptor (TLR, NOD) signalling molecules, suggesting that preterm monocytes may harbour an underlying transcriptional deficiency. Preterm birth is frequently associated with histologic chorioamnionitis (HCA; present in 40–70% of preterm infant placentae), which is known to alter the risk for sepsis. However there is very limited data concerning the impact of HCA on human neonatal monocyte function. Earlier studies of preterm monocyte function have largely assessed responses to purified bacterial products (e.g. LPS) that may be poorly representative of in vivo infection. Furthermore, many studies have analysed monocyte responses within mixed cell cultures (whole blood, CBMC), which does not allow assessment of intrinsic monocyte function. Finally, very few studies have assessed transcriptional responses in preterm monocytes, and none have

189 done so at a transcriptome-wide scale. The aims of this thesis were to compare the global transcriptional responses of preterm and term monocytes to direct challenge with live E. coli and S. epidermidis to identify transcriptional deficiencies specific to preterm infants, and determine the impact of HCA on this response. We also aimed to characterise the pathogen-specific and core neonatal monocyte response to E. coli and S. epidermidis to enhance our understanding of the critical innate immune pathways involved in defence against invasive bacterial infections. An in vitro challenge model was developed and optimised (Chapter 3) to effectively examine the intrinsic transcriptional response of purified cord blood monocytes to live E. coli and S. epidermidis using RNA-seq (Chapters 4 and 5).

6.1 Summary and discussion of key findings

Preterm infant monocytes are not intrinsically deficient A number of statistical and bioinformatic strategies were applied to compare the monocyte transcriptional profiles between preterm and term infants and collectively, the results indicate that preterm infants >29 weeks GA do not exhibit a transcriptional deficiency following stimulation with live E. coli or S. epidermidis (Chapter 4). Instead, the vast majority (76%) of the variation within the RNA-seq data could be attributed to the changes induced by bacterial stimulation and inter-infant variability. Targeted analysis of TLR, inflammatory, antigen presentation and apoptosis pathway genes did not identify differences between preterm HCA- and term infants. Similarly, un-biased differentially expression analyses to directly compare unstimulated, E. coli- or S. epidermidis-stimulated monocytes identified very few differentially expressed genes between preterm HCA- and term infants.

Comparisons of the differentially expressed genes induced by bacterial stimulation identified ~300–400 genes that were unique to each infant group, however the vast majority of these followed similar expression patterns in all groups, and bioinformatic and network analyses suggested that overall these genes did not form biologically significant or highly interconnected gene networks. This was supported by the delta-fold change analysis, which did not identify any genes that were both significantly differentially expressed upon bacterial stimulation (either stimulus), and significantly differentially expressed between preterm HCA- and term infants. Furthermore, similar protein levels of inflammatory cytokines and chemokines were observed in monocyte

190 culture supernatants following stimulation with E. coli or S. epidermidis between preterm HCA- and term infants. Taken together, these results suggest that the intrinsic capability of monocytes to respond to infection at birth is established as early as 29 weeks gestation (two months prior to term birth). This finding falsifies our hypothesis that preterm infant monocytes exhibit a transcriptional deficiency and raises interesting questions about why maturation of monocyte responses early in gestation would be advantageous.

An intrinsic capacity to respond to bacterial infection early in gestation may provide fetal protection in utero against invading organisms and thereby offer a survival advantage during pregnancy. Passive immunity in the form of pathogen-specific maternal IgG is largely acquired during the last four weeks of gestation, and in the absence of antigen-specific T or B cells, fetuses are highly dependent on innate immune defences such as monocytes, earlier in gestation (113, 117). However, the fact that in utero and fetal inflammation is strongly linked to preterm labour suggests that responding to infection too early in gestation is detrimental, especially when this occurs at the limits of fetal viability (23–24 weeks gestation). Moreover, the immunological landscape of the maternal-fetal interface is programmed toward a tolerant/anti- inflammatory state to prevent graft rejection and spontaneous abortion/preterm birth (336, 337). These two concepts seem conflicting; on one hand fetal protection against infection requires functional innate immune defences, but an in utero inflammatory response to infection may lead to preterm birth and place the fetus at risk of adverse outcomes associated with prematurity and inflammation. Active regulation of fetal monocytes responses during pregnancy, through extrinsic factors that allow for appropriate monocyte development, may provide a mechanism to control and regulate the monocyte inflammatory response to infection.

Extrinsic factors were removed in our model, as we assessed monocytes in isolation, which may be why we did not observe major transcriptional differences between preterm and term infants. Monocytes in circulation are exposed to plasma, which hosts a range of soluble mediators capable of modulating immune responses (section 1.4.1.1). Extrinsic regulation of innate immune function in neonates has been demonstrated by plasma exchange experiments, whereby adult plasma significantly enhanced neonatal (term) leukocyte TNFα production in response to TLR1/2, TLR2/6, TLR4 and TLR7 agonists. Conversely, neonatal plasma significantly reduced the ability of adult

191 leukocytes to produce TNFα in response to the same agonists, suggesting that neonatal plasma either lacks an activating factor, or contains an inhibitor (338). However this study was limited to assessment of plasma from term infants, which does not account for potential changes in plasma composition associated with prematurity. In addition, while plasma from cord blood may be physiologically representative of peripheral blood plasma at the most clinically relevant time point of EOS (within 24 hours of birth), it may not be physiologically representative of plasma at the peak incidence of LOS (2–3 weeks post-birth). Therefore, these findings should be confirmed using peripheral blood plasma from preterm infants, at clinically relevant time points. Nevertheless, differential regulation of monocyte responses by extrinsic factors during gestation might explain why studies utilising whole blood demonstrate significant impairment of preterm monocytes to produce inflammatory cytokines (209, 212, 298), whereas we and others report similar cytokine production between preterm and term infants when monocytes are studied in isolation (191, 194, 195, 201, 211).

Various extracellular factors with monocyte modulating properties have been identified. For example, adenosine (an endogenous purine metabolite) induces cyclic adenosine monophosphate to inhibit neonatal monocyte TNFα production in response to TLR agonists or whole bacteria (339). Plasma concentrations of adenosine are increased in newborns compared to adults, but the relative plasma concentrations in preterm infants have not been determined (339). In addition, the host defence (antimicrobial) peptide LL-37 is a monocyte chemoattractant, activates ERK1/2 and p38 in adult human monocytes (dependent on additional serum factors, potentially apolipoproteins), and induces monocyte transcription of cytokine, chemokine and interferon-related genes (340, 341). The significantly reduced levels of LL-37 observed in preterm cord blood plasma (compared to term infants) may therefore result in reduced monocyte activation and chemotaxis in preterm infants (136). Furthermore, at least two distinct and as yet unidentified factors in term infant plasma polarise TLR4-mediated cytokine responses toward an anti-inflammatory environment through reduced IL-12p70 and increased IL- 10, this has not been confirmed in preterm infant plasma (342). A comprehensive literature review of extracellular mediators with immune-modulating properties is beyond the scope of this discussion, however the aforementioned studies provide evidence that extrinsic factors present in neonatal plasma at birth (cord blood) can modulate the monocyte inflammatory response. This highlights a potential mechanism for differential immune regulation during gestation, which may contribute to preterm 192 infant susceptibility to sepsis. Future studies should aim to assess the immunoregulatory properties of preterm infant peripheral whole blood plasma at the most clinically relevant time points (day 1 for EOS, days 10–22 for LOS).

Aside from their prominent role during infection, monocytes may play an important role in fetal development as embryonic hematopoietic precursors for tissue macrophages and dendritic cells. Recent murine studies have identified fetal liver monocytes as the predominant origin of adult tissue-resident macrophages in the kidney, liver, skin and lung, and dendritic cells in the epidermis (343, 344). Our finding that transcriptional maturation of circulating human fetal monocytes is established several months prior to birth may relate to their essential role in seeding populations of tissue macrophages and dendritic cells, however this concept has not been explored in humans. In fact, whether or not preterm and term infant monocytes in circulation are derived from a common hematopoietic origin remains unknown. Fetal bone marrow haematopoiesis begins late in the first trimester (11–12 weeks) of human gestation, concurrent with transitory haematopoiesis in the fetal liver and spleen, which suggests that monocytes from fetuses ~30 weeks (as in our study) and full term could be of common origin (345). A clearer picture of developmental haematopoiesis will certainly aid our understanding of preterm infant immunity, but due to ethical considerations and limited access to human embryos at the early stages of gestation, this remains challenging.

Histologic chorioamnionitis alters the transcriptional profile of preterm monocytes Given the prominent role of HCA in preterm birth, the established link between HCA and an altered risk of sepsis, and the evidence from sheep studies that HCA alters monocyte function, we aimed to determine the impact of HCA on preterm infant monocyte transcriptional responses to E. coli and S. epidermidis. Targeted gene expression analysis revealed that preterm HCA+ infants exhibited significantly reduced expression of several TLR pathway genes (TLR4, TLR5, TBK1, IRAK2, NFKBIA), inflammasome formation genes (NLRP3, CASP1), antigen presentation genes (CD74, HLA-DRA, CD86) and growth factor genes (G-CSF, GM-CSF) compared to either preterm HCA- or term infants. This reduction in gene expression was predominantly observed following S. epidermidis stimulation, indicating a dampened immune response to Gram-positive infection. In addition, delta-fold change analysis to identify any genes that were significantly differentially expressed between infant groups following bacterial stimulation, revealed a subset of genes that almost all followed a significant

193 hypo-responsive pattern of expression by preterm HCA+ monocytes. These findings are in line with observations of hypo-responsive monocyte responses (decreased IL-6 and hydrogen peroxide production, and lower surface expression of MHC class II and CD14) by preterm lambs following intra-amniotic exposure to LPS or Ureaplasma parvum (226-229), and may contribute to the increased risk of EOS observed for preterm HCA+ infants (38).

Our results suggest that an intrinsic mechanism capable of regulating gene transcription may be differentially regulated in infants exposed to HCA. Bacterial infections can induce a variety of epigenetic changes in the host, including histone modifications and DNA methylation, which are major mechanisms by which gene transcription can be regulated (346, 347). Several candidate genes that were identified as significantly differentially expressed by preterm HCA+ infant monocyte in multiple analyses were involved in DNA replication (RFC2), DNA/RNA binding (HNRNPUL2) and RNA degradation (PPP1R8), which may contribute to transcriptional regulation (348-350). Furthermore, HDCA10 (a histone deacetylase involved in transcriptional repression (351)) was predicted to be a significant upstream regulator of the preterm HCA+ infant unique response to E. coli. A recent study has reported a significant association between chorioamnionitis and fetal hyper-methylation of the imprint regulatory region PLAGL1, the dysregulation of which is associated with transient neonatal diabetes mellitus and tumour growth (352).

Our findings provide evidence that prenatal exposure to HCA may re-program the preterm infant monocyte transcriptional response toward a hypo-responsive phenotype involving a subset of well-studied immune genes, as well as other novel candidate genes. Epigenetic changes induced by exposure to bacterial species associated with HCA may provide a mechanism by which this occurs and warrants further investigation.

Neonatal monocytes exhibit a highly conserved transcriptional response to E. coli and S. epidermidis The final aim of this study was to identify and characterise the pathogen-specific, and core neonatal monocyte transcriptional response to E. coli and S. epidermidis. Infection- induced inflammation is responsible for much of the neonatal morbidity associated with sepsis, and characterisation of the inflammatory pathways induced by these common neonatal pathogens may inform the development of new inflammation-targeted

194 therapeutics. Across all infant groups the vast majority of transcripts were differentially expressed in response to both stimuli, resulting in a set of 6511 transcripts that were significantly differentially expressed regardless of gestational age or exposure to HCA (Chapter 5). We describe these genes as the conserved neonatal monocyte response to bacterial infection. The largely conserved transcriptional response of neonatal monocytes to E. coli and S. epidermidis reinforces the fundamental non-specific nature of innate immunity. Similar conserved transcriptional responses have been observed in human monocyte-derived dendritic cells and macrophages in response to a range of diverse bacterial pathogens including Gram-negative, Gram-positive and Mycobacterium species (243, 353). Bioinformatic and network analyses revealed that the core neonatal monocyte response to E. coli and S. epidermidis was largely characterised by pattern recognition receptor-driven inflammation. Importantly, several genes/pathways involved in the negative regulation of TLR signalling were also up- regulated within this conserved response, suggesting that neonatal monocytes exhibit mechanisms for the intrinsic regulation of inflammation. Again, these findings indicate that impaired or dysregulated inflammation in preterm infants (associated with the pathogenesis of sepsis) may not be due to an intrinsic monocyte deficiency, suggesting the potential role of extrinsic factors.

In addition to pattern recognition receptor signalling and inflammation, the conserved pathogen response genes were significantly associated with cell morphology, with VCAM1 (involved in monocyte transendothelial migration) one of the most highly interconnected sub-network hubs. A unique function of monocytes is their ability to migrate into infected/inflamed tissues and differentiate into macrophages and dendritic cells to clear pathogens and resolve inflammation (90, 169), and our results suggest that neonatal monocytes are primed to interact with the tissue microenvironment following infection with E. coli and S. epidermidis. The gut is presumed to be the initial site of infection in neonatal sepsis (leading to translocation of bacteria/bacterial products into the blood), however the role circulating monocytes in resolving gut infection in neonates is unknown. Indeed, only one study has assessed preterm monocyte differentiation in macrophages in vitro (but did not assess function) (170), and no studies have assessed the ability of preterm monocytes to differentiate into dendritic cells, these would be valuable future studies.

195

In addition, studies comparing the functions of tissue macrophages or dendritic cells between preterm and term infants are lacking, likely due to ethical and practical limitations of obtaining these samples. One study has performed immunohistochemistry on intestinal tissue from human fetuses and preterm infants (11–24 weeks GA) and term infants, and report TNFα and IL-8 expression by intestinal macrophages was exclusive to tissue from fetuses and preterm infants (354). The authors suggest that increased macrophage-induced intestinal inflammation may lead to mucosal injury and susceptibility to necrotising enterocolitis in preterm infants. More studies are required to characterise the functional changes in intestinal macrophages and dendritic cells over the course of gestation, to determine whether susceptibility to sepsis is associated with innate immune deficiencies at the presumed site of infection.

We also identified a collective set of 734 genes that were specifically differentially expressed in response to either E. coli (327 genes) or S. epidermidis (407 genes). The E. coli-specific response was characterised by an interferon/anti-viral immune signature, likely induced via TRIF-dependent TLR4 signal transduction leading to activation of the transcription factor IRF3. TRIF signalling is critical for survival of neonatal mice to E. coli sepsis, reinforcing the importance of this pathway in protection against invasive E. coli infection (355). In contrast, bioninformatic analyses of the S. epidermidis- specific response did not highlight the involvement of any particular biological functions, pathways or highly interconnected gene networks. This suggests that the S. epidermidis-specific response genes are either not well characterised and therefore could not be categorised via bioinformatics (which relies on pre-existing knowledge of molecular interactions), or that collectively these genes are genuinely not part of any canonical signalling pathways or cellular bio-functions associated with infection, possibly reflecting the commensal behaviour of S. epidermidis.

6.2 Limitations of this study

An important, but unavoidable limitation of this study was the use of umbilical cord blood rather than peripheral blood to source preterm and term infant monocytes. As the predominant site of systemic infection, peripheral blood provides a more relevant sample type than umbilical cord blood. However many immunological studies utilise cord blood as an easily accessible surrogate, due to the ethical and practical challenges associated with peripheral blood sampling of preterm infants. Positive correlations of E.

196 coli and S. epidermidis phagocytosis between cord blood monocytes and peripheral whole blood monocytes (from day 1) provide some evidence that cord blood is an appropriate surrogate (216). Studies utilising peripheral whole blood from preterm infants are ethically limited to very small samples (0.5–1 mL), which yield insufficient numbers of monocytes and extracted nucleic acids for the RNA-seq technology available in this study.

The use of in vitro cell culture models to study in vivo immune responses has a number of limitations including the simplification of complex biological interactions and the artificial nature of cell culture environments. However the primary aim of this study to characterise the intrinsic neonatal monocyte transcriptional response to live bacterial stimulation necessitated the use of in vitro methods. Our methodology was developed and optimised to minimise experimental variability by using the same stocks of live E. coli or S. epidermidis for each stimulation, and the same batches of cell culture reagents wherever possible. Live bacterial stimuli were used as the best representation of in vivo infection. The culture conditions were optimised to minimise gene expression changes induced by the monocyte purification process, and ensure that the extracted RNA was of the highest achievable standard prior to RNA-seq. While the overall experimental variability was minimised, inter-infant variability remained an important source of variation within the RNA-seq data. The biological variability observed in global gene expression profiles between individual infants potentially limited our sensitivity to detect significant differences between infant groups. The inclusion of more biological replicates in our study design would have increased our power to detect gene expression differences between infant groups, and we recognise this as a potential limitation. Being mindful of budget constraints, and in the absence of consensus guidelines for RNA-seq design, published guidelines for sequencing depth and numbers of biological replicates were used to aid our experimental design and maximise statistical power (253, 254). Efforts to reduce variability were made by matching preterm HCA- and term infants on mode of delivery (all cesarean section), and both preterm groups were matched for gestational age and birth weight. Both male and female participants were included in all groups however gender was more difficult to control for in such a small cohort.

Despite these limitations, the overall approach of this study was successful in characterising the neonatal monocyte transcriptional response to live E. coli and S.

197 epidermidis, and we were able to detect significantly differentially expressed genes with a fold change of at least 1.1 between unstimulated and bacterial-stimulated monocytes.

6.3 Directions for future research

Here we report no intrinsic transcriptionally deficiency in major immune defence pathways in preterm infant monocytes following stimulation with E. coli or S. epidermidis. Whether differential regulation of monocyte responses to infection occurs at different stages of gestation via extrinsic factors is unknown, but this could contribute to preterm infant susceptibility to infection. Future work to identify differences in immunoregulatory plasma proteins and metabolites between preterm and term infants will provide insight into potential mechanisms for extrinsic regulation during pregnancy. Protocols utilising mass spectrometry have been developed for analysis of the entire proteome and metabolome, allowing researchers to move beyond measurements of single proteins and metabolites (356, 357). These protocols require as little as 100–300 µL of human plasma, making it possible to utilise peripheral whole blood from preterm and term infants. A major benefit of whole proteome or metabolome analysis is the potential to identify novel target proteins and metabolites for further characterisation. Indeed, this holistic approach has been applied to comparative analysis of the urine metabolome between preterm and term infants, which identified several metabolites including several amino acids (phenylalanine, proline, tyrosine, tryptophan) that could discriminate infants at different stages of gestation (358). Whole proteome and metabolome analysis of peripheral blood plasma from preterm infants and term infants would be the next step toward identifying candidate proteins and metabolites that are differentially regulated over the course of gestation, to assess their potential for regulating innate immune responses to bacterial infection.

Exploration of the mechanisms by which prenatal exposure to HCA modulates infant immunity will enhance our understanding of how early infection/inflammation exposure alters postnatal immunity and the risk of and sepsis. Fetal epigenetic changes induced via bacterial exposure in utero may be one such mechanism, consistent with the established link between bacterial infection and changes to chromatin structure (i.e. histone modifications and DNA methylation) (347). DNA was extracted alongside RNA from all culture monocytes used in this study, which could be used for analysis of genome-wide DNA methylation via Illumina next generation sequencing to identify

198 differences between preterm and term infants. Integration of RNA-seq and genome- wide methylation data has successfully been applied to the study hematopoietic stem cell differentiation to identify coordinated changes in gene methylation and transcription (359). As we are limited to analysis of cord blood monocytes, additional studies would be required to determine whether any changes associated with prenatal exposure to HCA are transient, or persist over the first few weeks of life. This is important from a clinical perspective, as the risk of sepsis for preterm HCA+ infants changes over time from an increased risk of early-onset sepsis, to a reduced risk of late-onset sepsis (38, 41). Ideally, these studies would sample peripheral whole blood from preterm infants with and without exposure to HCA at time points corresponding to greatest risk of sepsis, and quantify the expression of the candidate genes identified in this study by PCR following challenge with live S. epidermidis.

6.4 Concluding remarks

We have used RNA-seq for the first transcriptome-wide profiling and network analysis of preterm and term infant monocyte responses to live E. coli and S. epidermidis. This novel data set provides a valuable resource for immunologists and monocyte biologists looking to compare monocyte transcriptional profiles from individuals at different stages of gestation or age, or in response to different stimuli. Our findings suggest that preterm infant susceptibility to sepsis is not due to an underlying transcriptional monocyte deficiency per se. However differential regulation of monocyte responses during gestation via extrinsic factors may still play a role, which warrants further investigation. We report that prenatal exposure to histologic chorioamnionitis results in the transcriptional re-programing of a subset of genes toward a hypo-responsive phenotype, which may contribute to the altered risk of sepsis in this population, at least at birth. We demonstrate that the neonatal monocyte transcriptional response to E. coli and S. epidermidis is highly conserved, suggesting that novel therapeutics specifically targeting shared inflammatory pathways should be applicable to Gram-negative and Gram-positive sepsis in preterm infants. Applying an integrated ‘omics’ approach to investigate the multiple cellular mechanisms that regulate peripheral blood immune cell responses (genome, epigenome, transcriptome, metabolome, proteome) preceding the time of sepsis will be the next step toward improving our understanding of preterm infant immunity, and the immunological mechanisms underlying their unique vulnerability to sepsis.

199

References

1. Lie, Z., Zeki, R., Hilder, L and Sullivan, E. A. 2013. Australia's mothers and babies 2011. Perinatal statistics series no. 28. Canberra: AIHW National Perinatal Epidemiology and Statistics Unit.

2. Goldenberg, R. L., Culhane, J. F., Iams, J. D., et al. 2008. Epidemiology and causes of preterm birth. Lancet 371: 75-84.

3. Zeitlin, J., Szamotulska, K., Drewniak, N., et al. 2013. Preterm birth time trends in europe: A study of 19 countries. Brit J Obstet Gynaec 120: 1356-1365.

4. Blencowe, H., Cousens, S., Chou, D., et al. 2013. Born too soon: The global epidemiology of 15 million preterm births. Reprod Health 10 Suppl 1: S2.

5. Blencowe, H., Cousens, S., Oestergaard, M. Z., et al. 2012. National, regional, and worldwide estimates of preterm birth rates in the year 2010 with time trends since 1990 for selected countries: A systematic analysis and implications. Lancet 379: 2162-2172.

6. Cheong, J. L., and Doyle, L. W. 2012. Increasing rates of prematurity and epidemiology of late preterm birth. J Paediatr Child Health 48: 784-788.

7. Jackson, R. A., Gibson, K. A., Wu, Y. W., et al. 2004. Perinatal outcomes in singletons following in vitro fertilization: A meta-analysis. Obstet Gynecol 103: 551-563.

8. Ananth, C. V., Joseph, K. S., Oyelese, Y., et al. 2005. Trends in preterm birth and perinatal mortality among singletons: United states, 1989 through 2000. Obstet Gynecol 105: 1084-1091.

9. Chang, H. H., Larson, J., Blencowe, H., et al. 2013. Preventing preterm births: Analysis of trends and potential reductions with interventions in 39 countries with very high human development index. Lancet 381: 223-234.

10. Howson, C. P. K., M.V, Lawn, J.E. 2012. Born too soon: The global action report on preterm birth. March of Dimes, PMNCH, Save the Children, WHO, Geneva.

11. McCormick, M. C. 1985. The contribution of low birth weight to infant mortality and childhood morbidity. New Engl J Med 312: 82-90.

12. Stoll, B. J., Hansen, N. I., Bell, E. F., et al. 2010. Neonatal outcomes of extremely preterm infants from the nichd neonatal research network. Pediatrics: peds. 2009-2959.

13. Butler, A. S., and Behrman, R. E. 2007. Preterm birth:: Causes, consequences, and prevention. National Academies Press.

14. Behrman, R. E., Butler, A.S. 2007. In Preterm birth: Causes, consequences, and prevention. R. E. Behrman, and A. S. Butler, eds, Washington (DC).

15. Chow, S., Le Marsney, R., Hossain, S., et al. 2015. Report of the australian and new zealand neonatal network 2013. ANZNN, Sydney.

16. Adams-Chapman, I. 2012. Long-term impact of infection on the preterm neonate. In Semin Perinatol. Elsevier. 462-470.

17. Kramer, M. S., Demissie, K., Yang, H., et al. 2000. The contribution of mild and moderate preterm birth to infant mortality. Fetal and infant health study group of the canadian perinatal surveillance system. JAMA 284: 843-849.

18. Petrini, J. R., Dias, T., McCormick, M. C., et al. 2009. Increased risk of adverse neurological development for late preterm infants. J Pediatr 154: 169-176.

200

19. Engle, W. A. 2011. Morbidity and mortality in late preterm and early term newborns: A continuum. Clin Perinatol 38: 493-516.

20. Moster, D., Lie, R. T., and Markestad, T. 2008. Long-term medical and social consequences of preterm birth. New Engl J Med 359: 262-273.

21. Winkvist, A., Mogren, I., and Hogberg, U. 1998. Familial patterns in birth characteristics: Impact on individual and population risks. Int J Epidemiol 27: 248-254.

22. Boyd, H. A., Poulsen, G., Wohlfahrt, J., et al. 2009. Maternal contributions to preterm delivery. Am J Epidemiol 170: 1358-1364.

23. Plunkett, J., and Muglia, L. J. 2008. Genetic contributions to preterm birth: Implications from epidemiological and genetic association studies. Ann Med 40: 167-179.

24. Dörtbudak, O., Eberhardt, R., Ulm, M., et al. 2005. Periodontitis, a marker of risk in pregnancy for preterm birth. J Clin Periodontol 32: 45-52.

25. Leitich, H., Bodner-Adler, B., Brunbauer, M., et al. 2003. Bacterial vaginosis as a risk factor for preterm delivery: A meta-analysis. Am J Obstet Gynecol 189: 139-147.

26. Sibai, B., Dekker, G., and Kupferminc, M. 2005. Pre-eclampsia. Lancet 365: 785-799.

27. Dole, N., Savitz, D. A., Hertz-Picciotto, I., et al. 2003. Maternal stress and preterm birth. Am J Epidemiol 157: 14-24.

28. DeFranco, E. A., Jacobs, T., Plunkett, J., et al. 2011. Placental pathologic aberrations in cases of familial idiopathic spontaneous preterm birth. Placenta 32: 386-390.

29. Epstein, F. H., Goldenberg, R. L., Hauth, J. C., et al. 2000. Intrauterine infection and preterm delivery. New Engl J Med 342: 1500-1507.

30. Ananth, C. V., and Vintzileos, A. M. 2006. Maternal-fetal conditions necessitating a medical intervention resulting in preterm birth. Am J Obstet Gynecol 195: 1557-1563.

31. Ananth, C. V., Demissie, K., Smulian, J. C., et al. 2001. Relationship among placenta previa, fetal growth restriction, and preterm delivery: A population-based study. Obstet Gynecol 98: 299-306.

32. Salafia, C. M., Minior, V. K., Pezzullo, J. C., et al. 1995. Intrauterine growth restriction in infants of less than thirty-two weeks' gestation: Associated placental pathologic features. Am J Obstet Gynecol 173: 1049-1057.

33. Goldenberg, R. L., Hauth, J. C., and Andrews, W. W. 2000. Intrauterine infection and preterm delivery. New Engl J Med 342: 1500-1507.

34. Czikk, M. J., McCarthy, F. P., and Murphy, K. E. 2011. Chorioamnionitis: From pathogenesis to treatment. Clin Microbiol Infect 17: 1304-1311.

35. Redline, R. W., Faye-Petersen, O., Heller, D., et al. 2003. Amniotic infection syndrome: Nosology and reproducibility of placental reaction patterns. Pediatr Dev Pathol 6: 435-448.

36. Murtha, A. P., and Edwards, J. M. 2014. The role of mycoplasma and ureaplasma in adverse pregnancy outcomes. Obstet Gynecol Clin North Am 41: 615-627.

37. Aagaard, K., Ma, J., Antony, K. M., et al. 2014. The placenta harbors a unique microbiome. Sci Transl Med 6: 237ra265-237ra265.

38. Kacerovsky, M., Musilova, I., Andrys, C., et al. 2014. Prelabor rupture of membranes between 34 and 37 weeks: The intraamniotic inflammatory response and neonatal outcomes. Am J Obstet Gynecol 210: 325 e321-325 e310.

201

39. Arayici, S., Kadioglu Simsek, G., Oncel, M. Y., et al. 2014. The effect of histological chorioamnionitis on the short-term outcome of preterm infants≤ 32 weeks: A single-center study. J Matern Fetal Neonatal Med 27: 1129-1133.

40. Hofer, N., Kothari, R., Morris, N., et al. 2013. The fetal inflammatory response syndrome is a risk factor for morbidity in preterm neonates. Am J Obstet Gynecol 209: 542 e541-542 e511.

41. Strunk, T., Doherty, D., Jacques, A., et al. 2012. Histologic chorioamnionitis is associated with reduced risk of late-onset sepsis in preterm infants. Pediatrics 129: e134-141.

42. Group, Y. I. C. S. S. 2008. Clinical signs that predict severe illness in children under age 2 months: A multicentre study. Lancet 371: 135-142.

43. Chirico, G., and Loda, C. 2011. Laboratory aid to the diagnosis and therapy of infection in the neonate. Pediatr Rep 3: e1.

44. Stoll, B. J., Hansen, N. I., Sanchez, P. J., et al. 2011. Early onset neonatal sepsis: The burden of group b streptococcal and e. Coli disease continues. Pediatrics 127: 817-826.

45. Bromberger, P., Lawrence, J. M., Braun, D., et al. 2000. The influence of intrapartum antibiotics on the clinical spectrum of early-onset group b streptococcal infection in term infants. Pediatrics 106: 244-250.

46. Escobar, G. J., Li, D. K., Armstrong, M. A., et al. 2000. Neonatal sepsis workups in infants >/=2000 grams at birth: A population-based study. Pediatrics 106: 256-263.

47. Zaleznik, D. F., Rench, M. A., Hillier, S., et al. 2000. Invasive disease due to group b streptococcus in pregnant women and neonates from diverse population groups. Clin Infect Dis 30: 276-281.

48. Schrag, S. J., Zywicki, S., Farley, M. M., et al. 2000. Group b streptococcal disease in the era of intrapartum antibiotic prophylaxis. N Engl J Med 342: 15-20.

49. Mukhopadhyay, S., and Puopolo, K. M. 2012. Risk assessment in neonatal early onset sepsis. Semin Perinatol 36: 408-415.

50. Stoll, B. J., Hansen, N. I., Higgins, R. D., et al. 2005. Very low birth weight preterm infants with early onset neonatal sepsis: The predominance of gram-negative infections continues in the national institute of child health and human development neonatal research network, 2002-2003. Pediatr Infect Dis J 24: 635-639.

51. Stoll, B. J., Hansen, N., Fanaroff, A. A., et al. 2002. Changes in pathogens causing early-onset sepsis in very-low-birth-weight infants. N Engl J Med 347: 240-247.

52. Bizzarro, M. J., Dembry, L. M., Baltimore, R. S., et al. 2008. Changing patterns in neonatal escherichia coli sepsis and ampicillin resistance in the era of intrapartum antibiotic prophylaxis. Pediatrics 121: 689-696.

53. Klinger, G., Levy, I., Sirota, L., et al. 2009. Epidemiology and risk factors for early onset sepsis among very-low-birthweight infants. Am J Obstet Gynecol 201: 38 e31-36.

54. Daley, A. J., and Isaacs, D. 2004. Ten-year study on the effect of intrapartum antibiotic prophylaxis on early onset group b streptococcal and escherichia coli neonatal sepsis in australasia. Pediatr Infect Dis J 23: 630-634.

55. Cohen-Wolkowiez, M., Moran, C., Benjamin, D. K., et al. 2009. Early and late onset sepsis in late preterm infants. Pediatr Infect Dis J 28: 1052-1056.

56. Dong, Y., and Speer, C. P. 2015. Late-onset neonatal sepsis: Recent developments. Arch Dis Child Fetal Neonatal Ed 100: F257-263.

202

57. Van Den Hoogen, A., Gerards, L. J., Verboon-maciolek, M. A., et al. 2009. Long-term trends in the epidemiology of neonatal sepsis and antibiotic susceptibility of causative agents. Neonatology 97: 22-28.

58. Bizzarro, M. J., Raskind, C., Baltimore, R. S., et al. 2005. Seventy-five years of neonatal sepsis at yale: 1928–2003. Pediatrics 116: 595-602.

59. Boghossian, N. S., Page, G. P., Bell, E. F., et al. 2013. Late-onset sepsis in very low birth weight infants from singleton and multiple-gestation births. J Pediatr 162: 1120-1124. e1121.

60. Lahra, M. M., Beeby, P. J., and Jeffery, H. E. 2009. Intrauterine inflammation, neonatal sepsis, and chronic lung disease: A 13-year hospital cohort study. Pediatrics 123: 1314-1319.

61. Troger, B., Gopel, W., Faust, K., et al. 2014. Risk for late-onset blood-culture proven sepsis in very-low-birth weight infants born small for gestational age: A large multicenter study from the german neonatal network. Pediatr Infect Dis J 33: 238-243.

62. Vergnano, S., Menson, E., Kennea, N., et al. 2011. Neonatal infections in england: The neonin surveillance network. Arch Dis Child Fetal Neonatal Ed 96: F9-F14.

63. Otto, M. 2009. Staphylococcus epidermidis - the 'accidental' pathogen. Nat Rev Micro 7: 555- 567.

64. Vuong, C., and Otto, M. 2002. Staphylococcus epidermidis infections. Microbes Infect 4: 481- 489.

65. Costa, S. F., Miceli, M. H., and Anaissie, E. J. 2004. Mucosa or skin as source of coagulase- negative staphylococcal bacteraemia? Lancet Infect Dis 4: 278-286.

66. Dong, Y., and Speer, C. P. 2014. The role of staphylococcus epidermidis in neonatal sepsis: Guarding angel or pathogenic devil? Int J Med Microbiol 304: 513-520.

67. Tsai, M.-H., Hsu, J.-F., Chu, S.-M., et al. 2014. Incidence, clinical characteristics and risk factors for adverse outcome in neonates with late-onset sepsis. Pediatr Infect Dis J 33: e7-e13.

68. Shah, D. K., Doyle, L. W., Anderson, P. J., et al. 2008. Adverse neurodevelopment in preterm infants with postnatal sepsis or necrotizing enterocolitis is mediated by white matter abnormalities on magnetic resonance imaging at term. J Pediatr 153: 170-175.e171.

69. Stoll, B. J., Hansen, N. I., Adams-Chapman, I., et al. 2004. Neurodevelopmental and growth impairment among extremely low-birth-weight infants with neonatal infection. JAMA 292: 2357- 2365.

70. Ozkan, H., Cetinkaya, M., Koksal, N., et al. 2014. Culture-proven neonatal sepsis in preterm infants in a neonatal intensive care unit over a 7 year period: Coagulase-negative staphylococcus as the predominant pathogen. Pediatr Int 56: 60-66.

71. Stoll, B. J., and Hansen, N. 2003. Infections in vlbw infants: Studies from the nichd neonatal research network. Semin Perinatol 27: 293-301.

72. Stoll, B. J., Hansen, N., Fanaroff, A. A., et al. 2002. Late-onset sepsis in very low birth weight neonates: The experience of the nichd neonatal research network. Pediatrics 110: 285-291.

73. Perlman, S. E., Saiman, L., and Larson, E. L. 2007. Risk factors for late-onset health care– associated bloodstream infections in patients in neonatal intensive care units. Am J Infect Control 35: 177-182.

74. Craft, A., and Finer, N. 2000. Nosocomial coagulase negative staphylococcal (cons) catheter- related sepsis in preterm infants: Definition, diagnosis, prophylaxis, and prevention. J Perinatol 21: 186-192.

203

75. Trend, S., Strunk, T., Hibbert, J., et al. 2015. Antimicrobial protein and peptide concentrations and activity in human breast milk consumed by preterm infants at risk of late-onset neonatal sepsis. PLoS One 10: e0117038.

76. Esposito, S., Zampiero, A., Pugni, L., et al. 2014. Genetic polymorphisms and sepsis in premature neonates. PLoS One 9: e101248.

77. Okada, Y., Klein, N. J., van Saene, H. K., et al. 2000. Bactericidal activity against coagulase- negative staphylococci is impaired in infants receiving long-term parenteral nutrition. Ann Surg 231: 276.

78. Okada, Y., Papp, E., Klein, N. J., et al. 1999. Total parenteral nutrition directly impairs cytokine production after bacterial challenge. J Pediatr Surg 34: 277-280.

79. Nizet, V., and Klein, J. O. 2011. Bacterial sepsis and meningitis. Infectious diseases of the fetus and newborn.

80. Korhonen, T. K., Valtonen, M. V., Parkkinen, J., et al. 1985. Serotypes, hemolysin production, and receptor recognition of escherichia coli strains associated with neonatal sepsis and meningitis. Infect Immun 48: 486-491.

81. Robbins, J. B., McCracken, G. H., Jr., Gotschlich, E. C., et al. 1974. Escherichia coli k1 capsular polysaccharide associated with neonatal meningitis. N Engl J Med 290: 1216-1220.

82. Wooster, D. G., Maruvada, R., Blom, A. M., et al. 2006. Logarithmic phase escherichia coli k1 efficiently avoids serum killing by promoting c4bp-mediated c3b and c4b degradation. Immunology 117: 482-493.

83. Bingen, E., Bonacorsi, S., Brahimi, N., et al. 1997. Virulence patterns of escherichia coli k1 strains associated with neonatal meningitis. J Clin Microbiol 35: 2981-2982.

84. Garland, J. S., Alex, C. P., Sevallius, J. M., et al. 2008. Cohort study of the pathogenesis and molecular epidemiology of catheter-related bloodstream infection in neonates with peripherally inserted central venous catheters. Infect Control Hosp Epidemiol 29: 243-249.

85. Soeorg, H., Huik, K., Parm, U., et al. 2013. Genetic relatedness of coagulase-negative staphylococci from gastrointestinal tract and blood of preterm neonates with late-onset sepsis. Pediatr Infect Dis J 32: 389-393.

86. Carl, M. A., Ndao, I. M., Springman, A. C., et al. 2014. Sepsis from the gut: The enteric habitat of bacteria that cause late-onset neonatal bloodstream infections. Clin Infect Dis 58: 1211-1218.

87. Bosmann, M., and Ward, P. A. 2013. The inflammatory response in sepsis. Trends Immunol 34: 129-136.

88. Reis Machado, J., Soave, D. F., da Silva, M. V., et al. 2014. Neonatal sepsis and inflammatory mediators. Mediat Inflamm 2014: 269681.

89. Cohen, J. 2002. The immunopathogenesis of sepsis. Nature 420: 885-891.

90. Auffray, C., Sieweke, M. H., and Geissmann, F. 2009. Blood monocytes: Development, heterogeneity, and relationship with dendritic cells. Annu Rev Immunol 27: 669-692.

91. Skrzeczyñska, J., Kobylarz, K., Hartwich, Z., et al. 2002. Cd14+cd16+ monocytes in the course of sepsis in neonates and small children: Monitoring and functional studies. Scand J Immunol 55: 629-638.

92. Fingerle, G., Pforte, A., Passlick, B., et al. 1993. The novel subset of cd14+/cd16+ blood monocytes is expanded in. Blood 82.

93. Aikawa, N. 1996. [cytokine storm in the pathogenesis of multiple organ dysfunction syndrome associated with surgical insults]. Nihon Geka Gakkai zasshi 97: 771-777. 204

94. Comstedt, P., Storgaard, M., and Lassen, A. T. 2009. The systemic inflammatory response syndrome (sirs) in acutely hospitalised medical patients: A cohort study. Scand J Trauma Resusc Emerg Med 17: 67-67.

95. Pinsky, M. R. 2004. Dysregulation of the immune response in severe sepsis. Am J Med Sci 328: 220-229.

96. Schultz, C., Temming, P., Bucsky, P., et al. 2004. Immature anti-inflammatory response in neonates. Clin Exp Immunol 135: 130-136.

97. Ng, P., Li, K., Wong, R., et al. 2003. Proinflammatory and anti-inflammatory cytokine responses in preterm infants with systemic infections. Arch Dis Child Fetal Neonatal Ed 88: F209-F213.

98. Boomer, J. S., To, K., Chang, K. C., et al. 2011. Immunosuppression in patients who die of sepsis and multiple organ failure. JAMA 306: 2594-2605.

99. Gentile, L. F., Nacionales, D. C., Lopez, M. C., et al. 2014. Protective immunity and defects in the neonatal and elderly immune response to sepsis. J Immunol 192: 3156-3165.

100. Leng, F.-Y., Liu, J.-L., Liu, Z.-J., et al. 2013. Increased proportion of cd4+cd25+foxp3+ regulatory t cells during early-stage sepsis in icu patients. J Microbiol Immunol Infect 46: 338- 344.

101. Delano, M. J., Scumpia, P. O., Weinstein, J. S., et al. 2007. Myd88-dependent expansion of an immature gr-1(+)cd11b(+) population induces t cell suppression and th2 polarization in sepsis. J Exp Med 204: 1463-1474.

102. Luciano, A. A., Arbona-Ramirez, I. M., Ruiz, R., et al. 2014. Alterations in regulatory t cell subpopulations seen in preterm infants. PLoS One 9: e95867.

103. Gervassi, A., Lejarcegui, N., Dross, S., et al. 2014. Myeloid derived suppressor cells are present at high frequency in neonates and suppress in vitro t cell responses. PLoS One 9: e107816.

104. Rieber, N., Gille, C., Köstlin, N., et al. 2013. Neutrophilic myeloid-derived suppressor cells in cord blood modulate innate and adaptive immune responses. Clin Exp Immunol 174: 45-52.

105. Marshall, J. C. 2014. Why have clinical trials in sepsis failed? Trends Mol Med 20: 195-203.

106. Moine, P., and Abraham, E. 2004. Immunomodulation and sepsis: Impact of the pathogen. Shock 22: 297-308.

107. Oblak, A., and Jerala, R. 2015. The molecular mechanism of species-specific recognition of lipopolysaccharides by the md-2/tlr4 receptor complex. Mol Immunol 63: 134-142.

108. Bi, D., Qiao, L., Bergelson, I., et al. 2015. Staphylococcus epidermidis bacteremia induces brain injury in neonatal mice via toll-like receptor 2-dependent and -independent pathways. J Infect Dis.

109. Strunk, T., Power Coombs, M. R., Currie, A. J., et al. 2010. Tlr2 mediates recognition of live staphylococcus epidermidis and clearance of bacteremia. PLoS One 5: e10111.

110. van der Poll, T., and Opal, S. M. 2008. Host–pathogen interactions in sepsis. Lancet Infect Dis 8: 32-43.

111. Blanco, A., Solis, G., Arranz, E., et al. 1996. Serum levels of cd14 in neonatal sepsis by gram- positive and gram-negative bacteria. Acta Paediatr 85: 728-732.

112. Tang, B. M. P., McLean, A. S., Dawes, I. W., et al. 2008. Gene-expression profiling of gram- positive and gram-negative sepsis in critically ill patients. Crit Care Med 36: 1125-1128.

205

113. Quinello, C., Silveira-Lessa, A. L., Ceccon, M. E. J. R., et al. 2014. Phenotypic differences in leucocyte populations among healthy preterm and full-term newborns. Scand J Immunol 80: 57- 70.

114. Walker, J. C., Smolders, M. A. J. C., Gemen, E. F. A., et al. 2011. Development of lymphocyte subpopulations in preterm infants. Scand J Immunol 73: 53-58.

115. Timens, W., Boes, A., Rozeboom-Uiterwijk, T., et al. 1989. Immaturity of the human splenic marginal zone in infancy. Possible contribution to the deficient infant immune response. J Immunol 143: 3200-3206.

116. Saji, F., Samejima, Y., Kamiura, S., et al. 1999. Dynamics of immunoglobulins at the feto- maternal interface. Rev Reprod. 4: 81-89.

117. Palmeira, P., Quinello, C., Silveira-Lessa, A. L., et al. 2011. Igg placental transfer in healthy and pathological pregnancies. Clin Dev Immunol 2012.

118. Alejandria, M. M., Lansang, M. A. D., Dans, L. F., et al. 2013. Intravenous immunoglobulin for treating sepsis, severe sepsis and septic shock. The Cochrane Library.

119. DeJonge, M., Burchfield, D., Bloom, B., et al. 2007. Clinical trial of safety and efficacy of ihn- a21 for the prevention of nosocomial staphylococcal bloodstream infection in premature infants. J Pediatr 151: 260-265.e261.

120. Benjamin, D., Schelonka, R., White, R., et al. 2006. A blinded, randomized, multicenter study of an intravenous staphylococcus aureus immune globulin. J Perinatol 26: 290-295.

121. Patel, M., and Kaufman, D. A. 2015. Anti-lipoteichoic acid monoclonal antibody (pagibaximab) studies for the prevention of staphylococcal bloodstream infections in preterm infants. Expert Opin Biol Ther 15: 595-600.

122. Baxter, D. 2010. Vaccine responsiveness in premature infants. Hum Vaccin 6: 506-511.

123. Sarma, J. V., and Ward, P. A. 2011. The complement system. Cell Tissue Res 343: 227-235.

124. Wolach, B., Dolfin, T., Regev, R., et al. 1997. The development of the complement system after 28 weeks' gestation. Acta Paediatr 86: 523-527.

125. Notarangelo, L. D., Chirico, G., Chiara, A., et al. 1984. Activity of classical and alternative pathways of complement in preterm and small for gestational age infants. Pediatr Res 18: 281- 285.

126. Fleer, A., Gerards, L., Aerts, P., et al. 1985. Opsonic defense to staphylococcus epidermidis in the premature neonate. J Infect Dis 152: 930-937.

127. Zilow, E. P., Hauck, W., Linderkamp, O., et al. 1997. Alternative pathway activation of the complement system in preterm infants with early onset infection. Pediatr Res 41: 334-339.

128. Zilow, G., Zilow, E. P., Burger, R., et al. 1993. Complement activation in newborn infants with early onset infection. Pediatr Res 34: 199-203.

129. Eddie Ip, W., Takahashi, K., Alan Ezekowitz, R., et al. 2009. Mannose-binding lectin and innate immunity. Immunol Rev 230: 9-21.

130. Ip, W. E., Takahashi, K., Moore, K. J., et al. 2008. Mannose-binding lectin enhances toll-like receptors 2 and 6 signaling from the phagosome. J Exp Med 205: 169-181.

131. Dzwonek, A. B., Neth, O. W., Thiébaut, R., et al. 2008. The role of mannose-binding lectin in susceptibility to infection in preterm neonates. Pediatr Res 63: 680-685.

132. Frakking, F. N., Brouwer, N., Zweers, D., et al. 2006. High prevalence of mannose-binding lectin (mbl) deficiency in premature neonates. Clin Exp Immunol 145: 5-12. 206

133. Pettengill, M. A., van Haren, S. D., and Levy, O. 2014. Soluble mediators regulating immunity in early life. Front Immunol 5: 457.

134. Hölzl, M. A., Hofer, J., Steinberger, P., et al. 2008. Host antimicrobial proteins as endogenous immunomodulators. Immunol Lett 119: 4-11.

135. Levy, O. 2004. Antimicrobial proteins and peptides: Anti-infective molecules of mammalian leukocytes. J Leukoc Biol 76: 909-925.

136. Strunk, T., Doherty, D., Richmond, P., et al. 2009. Reduced levels of antimicrobial proteins and peptides in human cord blood plasma. Arch Dis Child Fetal Neonatal Ed 94: F230-231.

137. Scott, P. 1989. Plasma lactoferrin levels in newborn preterm infants: Effect of infection. Ann Clin Biochem 26: 412-415.

138. Pammi, M., and Abrams, S. A. 2015. Oral lactoferrin for the prevention of sepsis and necrotizing enterocolitis in preterm infants. The Cochrane Library.

139. Doulatov, S., Notta, F., Laurenti, E., et al. 2012. Hematopoiesis: A human perspective. Cell Stem Cell 10: 120-136.

140. Amulic, B., Cazalet, C., Hayes, G. L., et al. 2012. Neutrophil function: From mechanisms to disease. Annu Rev Immunol 30: 459-489.

141. Davies, N. P., Buggins, A. G., Snijders, R. J., et al. 1992. Blood leucocyte count in the human fetus. Arch Dis Child 67: 399-403.

142. Urlichs, F., and Speer, C. 2004. Neutrophil function in preterm and term infants. NeoReviews 5: e417-e430.

143. Ohls, R. K., Li, Y., Abdel-Mageed, A., et al. 1995. Neutrophil pool sizes and granulocyte colony-stimulating factor production in human mid-trimester fetuses. Pediatr Res 37: 806-811.

144. Gessler, P., Lüders, R., König, S., et al. 1995. Neonatal neutropenia in low birthweight premature infants. Am J Perinatol 12: 34-38.

145. Carr, R. 2000. Neutrophil production and function in newborn infants. Br J Haematol 110: 18- 28.

146. Falconer, A., Carr, R., and Edwards, S. 1995. Impaired neutrophil phagocytosis in preterm neonates: Lack of correlation with expression of immunoglobulin or complement receptors. Neonatology 68: 264-269.

147. Forman, M. L., Stiehm, E. R., and Meyer, J. 1969. Impaired opsonic activity but normal phagocytosis in low-birth-weight infants. New Engl J Med 281: 926-931.

148. Fujiwara, T., Taniuchi, S., Hattori, K., et al. 1997. Effect of immunoglobulin therapy on phagocytosis by polymorphonuclear leucocytes in whole blood of neonates. Clin Exp Immunol 107: 435-439.

149. Björkqvist, M., Jurstrand, M., Bodin, L., et al. 2004. Defective neutrophil oxidative burst in preterm newborns on exposure to coagulase-negative staphylococci. Pediatr Res 55: 966-971.

150. Drossou, V., Kanakoudi, F., Tzimouli, V., et al. 1997. Impact of prematurity, stress and sepsis on the neutrophil respiratory burst activity of neonates. Neonatology 72: 201-209.

151. Habermehl, P., Hauer, T., Mannhardt, W., et al. 1998. Granulocyte function in premature infants before the 34th week of pregnancy and in mature newborn infants. Klinische Padiatrie 211: 149- 153.

152. Yost, C. C., Cody, M. J., Harris, E. S., et al. 2009. Impaired neutrophil extracellular trap (net) formation: A novel innate immune deficiency of human neonates. Blood 113: 6419-6427. 207

153. Root, R. K., and Dale, D. C. 1999. Granulocyte colony-stimulating factor and granulocyte- macrophage colony-stimulating factor: Comparisons and potential for use in the treatment of infections in nonneutropenic patients. J Infect Dis 179: S342-S352.

154. Bober, L. A., Grace, M. J., Pugliese-Sivo, C., et al. 1995. The effect of gm-csf and g-csf on human neutrophil function. Immunopharmacology 29: 111-119.

155. Carr, R., Brocklehurst, P., Doré, C. J., et al. 2009. Granulocyte-macrophage colony stimulating factor administered as prophylaxis for reduction of sepsis in extremely preterm, small for gestational age neonates (the programs trial): A single-blind, multicentre, randomised controlled trial. Lancet 373: 226-233.

156. Carr, R., Modi, N., and Doré, C. J. 2003. G-csf and gm-csf for treating or preventing neonatal infections. The Cochrane Library.

157. Ahmad, M., Fleit, H. B., Golightly, M. G., et al. 2004. In vivo effect of recombinant human granulocyte colony-stimulating factor on phagocytic function and oxidative burst activity in septic neutropenic neonates. Biol Neonate 86: 48-54.

158. Pammi, M., and Brocklehurst, P. 2011. Granulocyte transfusions for neonates with confirmed or suspected sepsis and neutropenia. Cochrane Database Syst Rev 10.

159. Mittag, D., Proietto, A. I., Loudovaris, T., et al. 2011. Human dendritic cell subsets from spleen and blood are similar in phenotype and function but modified by donor health status. J Immunol 186: 6207-6217.

160. Collin, M., McGovern, N., and Haniffa, M. 2013. Human dendritic cell subsets. Immunology 140: 22-30.

161. Haller Hasskamp, J., Zapas, J. L., and Elias, E. G. 2005. Dendritic cell counts in the peripheral blood of healthy adults. Am J Hematol 78: 314-315.

162. Holloway, J. A., Thornton, C. A., Diaper, N. D., et al. 2009. Phenotypic analysis of circulating dendritic cells during the second half of human gestation. Pediatr Allergy Immunol 20: 119-125.

163. Schüller, S. S., Sadeghi, K., Wisgrill, L., et al. 2013. Preterm neonates display altered plasmacytoid dendritic cell function and morphology. J Leukoc Biol 93: 781-788.

164. Lavoie, P. M., Huang, Q., Jolette, E., et al. 2010. 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 202: 1754-1763.

165. Gendrel, D., Raymond, J., Coste, J., et al. 1999. Comparison of procalcitonin with c-reactive protein, interleukin 6 and interferon-alpha for differentiation of bacterial vs. Viral infections. Pediatr Infect Dis J 18: 875-881.

166. Bekeredjian-Ding, I., Greil, J., Ammann, S., et al. 2014. Plasmacytoid dendritic cells: Neglected regulators of the immune response to staphylococcus aureus. Front Immunol 5: 238.

167. Parcina, M., Wendt, C., Goetz, F., et al. 2008. Staphylococcus aureus-induced plasmacytoid dendritic cell activation is based on an igg-mediated memory response. J Immunol 181: 3823- 3833.

168. Guisset, O., Dilhuydy, M.-S., Thiébaut, R., et al. 2007. Decrease in circulating dendritic cells predicts fatal outcome in septic shock. Intens Care Med 33: 148-152.

169. Geissmann, F., Manz, M. G., Jung, S., et al. 2010. Development of monocytes, macrophages and dendritic cells. Science (New York, N.Y.) 327: 656-661.

170. Gille, C., Spring, B., Tewes, L. J., et al. 2006. Diminished response to interleukin-10 and reduced antibody-dependent cellular cytotoxicity of cord blood monocyte-derived macrophages. Pediatr Res 60: 152-157. 208

171. Reinhardt, P. P., Reinhardt, B., Lathey, J. L., et al. 1995. Human cord blood mononuclear cells are preferentially infected by non-syncytium-inducing, macrophage-tropic human immunodeficiency virus type 1 isolates. J Clin Microbiol 33: 292-297.

172. Hariharan, D., Ho, W., Cutilli, J., et al. 2000. C-c chemokine profile of cord blood mononuclear cells: Selective defect in rantes production. Blood 95: 715-718.

173. Kohler, C., Adegnika, A. A., van der Linden, R., et al. 2011. Phenotypic characterization of mononuclear blood cells from pregnant gabonese and their newborns. Trop Med Int Health 16: 1061-1069.

174. Gelinas, L., Falkenham, A., Oxner, A., et al. 2011. Highly purified human peripheral blood monocytes produce il-6 but not tnfalpha in response to angiotensin ii. J Renin Angiotensin Aldosterone Syst 12: 295-303.

175. Ziegler-Heitbrock, L. 2014. Monocyte subsets in man and other species. Cell Immunol 289: 135- 139.

176. Belge, K. U., Dayyani, F., Horelt, A., et al. 2002. The proinflammatory cd14+cd16+dr++ monocytes are a major source of tnf. J Immunol 168: 3536-3542.

177. Ziegler-Heitbrock, H. W. 2000. Definition of human blood monocytes. J Leukoc Biol 67: 603- 606.

178. Ziegler-Heitbrock, L., and Hofer, T. P. 2013. Toward a refined definition of monocyte subsets. Front Immunol 4: 23.

179. Zawada, A. M., Rogacev, K. S., Rotter, B., et al. 2011. Supersage evidence for cd14++ cd16+ monocytes as a third monocyte subset. Blood 118: e50-e61.

180. Wong, K. L., Tai, J. J., Wong, W. C., et al. 2011. Gene expression profiling reveals the defining features of the classical, intermediate, and nonclassical human monocyte subsets. Blood 118: e16-31.

181. Frankenberger, M., Kaßner, G., Oak, P., et al. 2014. Intermediate cd14++cd16+ blood monocytes are elevated in preterm neonates. Eur Respir J 44.

182. Sharma, A. A., Jen, R., Kan, B., et al. 2015. Impaired nlrp3 inflammasome activity during fetal development regulates il-1beta production in human monocytes. Eur J Immunol 45: 238-249.

183. Kawai, T., and Akira, S. 2009. The roles of tlrs, rlrs and nlrs in pathogen recognition. Int Immunol 21: 317-337.

184. Iborra, S., and Sancho, D. 2015. Signalling versatility following self and non-self sensing by myeloid c-type lectin receptors. Immunobiology 220: 175-184.

185. Sharma, A. A., Jen, R., Brant, R., et al. 2014. Hierarchical maturation of innate immune defences in very preterm neonates. Neonatology 106: 1-9.

186. Shen, C.-M., Lin, S.-C., Niu, D.-M., et al. 2013. Development of monocyte toll-like receptor 2 and toll-like receptor 4 in preterm newborns during the first few months of life. Pediatr Res 73: 685-691.

187. Dziarski, R., and Gupta, D. 2000. Role of md-2 in tlr2-and tlr4-mediated recognition of gram- negative and gram-positive bacteria and activation of chemokine genes. J Endotoxin Res 6: 401- 405.

188. Tissieres, P., Ochoda, A., Dunn-Siegrist, I., et al. 2012. Innate immune deficiency of extremely premature neonates can be reversed by interferon-gamma. PLoS One 7: e32863.

189. Levy, E., Xanthou, G., Petrakou, E., et al. 2009. Distinct roles of tlr4 and cd14 in lps-induced inflammatory responses of neonates. Pediatr Res 66: 179-184. 209

190. Sadeghi, K., Berger, A., Langgartner, M., et al. 2007. Immaturity of infection control in preterm and term newborns is associated with impaired toll-like receptor signaling. J Infect Dis 195: 296- 302.

191. Granland, C., Strunk, T., Hibbert, J., et al. 2014. Nod1 and nod2 expression and function in very preterm infant mononuclear cells. Acta Paediatr 103: e212-218.

192. Nupponen, I., Kuuliala, A., Siitonen, S., et al. 2013. Cord blood monocytes, neutrophils and lymphocytes from preterm and full-term neonates show multiple aberrations in signalling profiles measured using phospho-specific whole-blood flow cytometry. Scand J Immunol 78: 426-438.

193. Strunk, T., Prosser, A., Levy, O., et al. 2012. Responsiveness of human monocytes to the commensal bacterium staphylococcus epidermidis develops late in gestation. Pediatr Res 72: 10- 18.

194. Le, T., Leung, L., Carroll, W. L., et al. 1997. Regulation of interleukin-10 gene expression: Possible mechanisms accounting for its upregulation and for maturational differences in its expression by blood mononuclear cells. Blood 89: 4112-4119.

195. Perez, A., Bellon, J. M., Gurbindo, M. D., et al. 2010. Impairment of stimulation ability of very- preterm neonatal monocytes in response to lipopolysaccharide. Hum Immunol 71: 151-157.

196. Grunwald, U., Fan, X., Jack, R. S., et al. 1996. Monocytes can phagocytose gram-negative bacteria by a cd14-dependent mechanism. J Immunol 157: 4119-4125.

197. Karlsson, H., Larsson, P., Wold, A. E., et al. 2004. Pattern of cytokine responses to gram- positive and gram-negative commensal bacteria is profoundly changed when monocytes differentiate into dendritic cells. Infect Immun 72: 2671-2678.

198. Hallwirth, U., Pomberger, G., Pollak, A., et al. 2004. Monocyte switch in neonates: High phagocytic capacity and low hla-dr expression in vlbwi are inverted during gestational aging. Pediatr Allergy Immunol 15: 513-516.

199. Strunk, T., Richmond, P., Prosser, A., et al. 2011. Method of bacterial killing differentially affects the human innate immune response to staphylococcus epidermidis. Innate Immun 17: 508-516.

200. Strunk, T., Temming, P., Gembruch, U., et al. 2004. Differential maturation of the innate immune response in human fetuses. Pediatr Res 56: 219-226.

201. Kaufman, D., Kilpatrick, L., Hudson, R. G., et al. 1999. Decreased superoxide production, degranulation, tumor necrosis factor alpha secretion, and cd11b/cd18 receptor expression by adherent monocytes from preterm infants. Clin Diagn Lab Immunol 6: 525-529.

202. Tacke, F., Ginhoux, F., Jakubzick, C., et al. 2006. Immature monocytes acquire antigens from other cells in the bone marrow and present them to t cells after maturing in the periphery. J Exp Med 203: 583-597.

203. Briken, V. 2012. "With a little help from my friends": Efferocytosis as an antimicrobial mechanism. Cell Host Microbe 12: 261-263.

204. Gille, C., Leiber, A., Mundle, I., et al. 2009. Phagocytosis and postphagocytic reaction of cord blood and adult blood monocyte after infection with green fluorescent protein-labeled escherichia coli and group b streptococci. Cytometry B Clin Cytom 76: 271-284.

205. Gille, C., Leiber, A., Spring, B., et al. 2008. Diminished phagocytosis-induced cell death (picd) in neonatal monocytes upon infection with escherichia coli. Pediatr Res 63: 33-38.

206. Leiber, A., Graf, B., Spring, B., et al. 2014. Neonatal monocytes express antiapoptotic pattern of bcl-2 proteins and show diminished apoptosis upon infection with escherichia coli. Pediatr Res 76: 142-149. 210

207. Sollberger, G., Strittmatter, G. E., Garstkiewicz, M., et al. 2014. Caspase-1: The inflammasome and beyond. Innate Immun 20: 115-125.

208. Sagulenko, V., Thygesen, S. J., Sester, D. P., et al. 2013. Aim2 and nlrp3 inflammasomes activate both apoptotic and pyroptotic death pathways via asc. Cell Death Differ 20: 1149-1160.

209. Forster-Waldl, E., Sadeghi, K., Tamandl, D., et al. 2005. Monocyte toll-like receptor 4 expression and lps-induced cytokine production increase during gestational aging. Pediatr Res 58: 121-124.

210. Schibler, K. R., Liechty, K. W., White, W. L., et al. 1993. Production of granulocyte colony- stimulating factor in vitro by monocytes from preterm and term neonates. Blood 82: 2478-2484.

211. Weatherstone, K. B., and Rich, E. A. 1989. Tumor necrosis factor/cachectin and interleukin-1 secretion by cord blood monocytes from premature and term neonates. Pediatr Res 25: 342-346.

212. Marchant, E. A., Kan, B., Sharma, A. A., et al. 2015. Attenuated innate immune defenses in very premature neonates during the neonatal period. Pediatr Res.

213. Currie, A. J., Curtis, S., Strunk, T., et al. 2011. Preterm infants have deficient monocyte and lymphocyte cytokine responses to group b streptococcus. Infect Immun 79: 1588-1596.

214. Schultz, C., Rott, C., Temming, P., et al. 2002. Enhanced interleukin-6 and interleukin-8 synthesis in term and preterm infants. Pediatr Res 51: 317-322.

215. Schibler, K. R., Trautman, M. S., Liechty, K. W., et al. 1993. Diminished transcription of interleukin-8 by monocytes from preterm neonates. J Leukoc Biol 53: 399-403.

216. Prosser, A., Hibbert, J., Strunk, T., et al. 2013. Phagocytosis of neonatal pathogens by peripheral blood neutrophils and monocytes from newborn preterm and term infants. Pediatr Res 74: 503- 510.

217. Filias, A., Theodorou, G. L., Mouzopoulou, S., et al. 2011. Phagocytic ability of neutrophils and monocytes in neonates. BMC Pediatr 11: 29.

218. Birle, A., Nebe, C. T., and Gessler, P. 2003. Age-related low expression of hla-dr molecules on monocytes of term and preterm newborns with and without signs of infection. J Perinatol 23: 294-299.

219. Azizia, M., Lloyd, J., Allen, M., et al. 2012. Immune status in very preterm neonates. Pediatrics 129: e967-974.

220. Laborada, G., and Nesin, M. 2005. Interleukin-6 and interleukin-8 are elevated in the cerebrospinal fluid of infants exposed to chorioamnionitis. Biol Neonate 88: 136-144.

221. Janota, J., Stranak, Z., Belohlavkova, S., et al. 2001. Postnatal increase of procalcitonin in premature newborns is enhanced by chorioamnionitis and neonatal sepsis. Eur J Clin Invest 31: 978-983.

222. Yang, H. J., Romero, R., Park, C.-W., et al. 513: The relationship between the severity of histologic chorioamnionitis and the intensity of fetal inflammatory response. Am J Obstet Gynecol 201: S191.

223. Rogers, B. B., Alexander, J. M., Head, J., et al. 2002. Umbilical vein interleukin-6 levels correlate with the severity of placental inflammation and gestational age. Hum Pathol 33: 335- 340.

224. Bezold, K. Y., Karjalainen, M. K., Hallman, M., et al. 2013. The genomics of preterm birth: From animal models to human studies. Genome Med 5: 10.1186.

211

225. Wolfs, T. G., Jellema, R. K., Turrisi, G., et al. 2012. Inflammation-induced immune suppression of the fetus: A potential link between chorioamnionitis and postnatal early onset sepsis. J Matern Fetal Neonatal Med 25 Suppl 1: 8-11.

226. Kramer, B. W., Ikegami, M., Moss, T. J., et al. 2005. Endotoxin-induced chorioamnionitis modulates innate immunity of monocytes in preterm sheep. Am J Respir Crit Care Med 171: 73- 77.

227. Kallapur, S. G., Jobe, A. H., Ball, M. K., et al. 2007. Pulmonary and systemic endotoxin tolerance in preterm fetal sheep exposed to chorioamnionitis. J Immunol 179: 8491-8499.

228. Kramer, B. W., Kallapur, S. G., Moss, T. J., et al. 2009. Intra-amniotic lps modulation of tlr signaling in lung and blood monocytes of fetal sheep. Innate Immun 15: 101-107.

229. Kallapur, S. G., Kramer, B. W., Knox, C. L., et al. 2011. Chronic fetal exposure to ureaplasma parvum suppresses innate immune responses in sheep. J Immunol 187: 2688-2695.

230. Chaussabel, D., Pascual, V., and Banchereau, J. 2010. Assessing the human immune system through blood transcriptomics. BMC Biol 8: 84-84.

231. Morey, J. S., Ryan, J. C., and Van Dolah, F. M. 2006. Microarray validation: Factors influencing correlation between oligonucleotide microarrays and real-time pcr. Biol Proced Online 8: 175- 193.

232. Wang, Z., Gerstein, M., and Snyder, M. 2009. Rna-seq: A revolutionary tool for transcriptomics. Nat Rev Genet 10: 57-63.

233. Nagalakshmi, U., Wang, Z., Waern, K., et al. 2008. The transcriptional landscape of the yeast genome defined by rna sequencing. Science 320: 1344-1349.

234. Mortazavi, A., Williams, B. A., McCue, K., et al. 2008. Mapping and quantifying mammalian transcriptomes by rna-seq. Nat Methods 5: 621-628.

235. Trapnell, C., Williams, B. A., Pertea, G., et al. 2010. Transcript assembly and quantification by rna-seq reveals unannotated transcripts and isoform switching during cell differentiation. Nat Biotechnol 28: 511-515.

236. Zhao, S., Fung-Leung, W.-P., Bittner, A., et al. 2014. Comparison of rna-seq and microarray in transcriptome profiling of activated t cells. PLoS One 9: e78644.

237. Anbazhagan, K., Duroux-Richard, I., Jorgensen, C., et al. 2014. Transcriptomic network support distinct roles of classical and non-classical monocytes in human. Int Rev Immunol 33: 470-489.

238. Krow-Lucal, E. R., Kim, C. C., Burt, T. D., et al. 2014. Distinct functional programming of human fetal and adult monocytes. Blood 123: 1897-1904.

239. Van den Bergh, R., Florence, E., Vlieghe, E., et al. 2010. Transcriptome analysis of monocyte- hiv interactions. Retrovirology 7: 53-53.

240. Chan, G., Bivins-Smith, E. R., Smith, M. S., et al. 2008. Transcriptome analysis reveals human cytomegalovirus reprograms monocyte differentiation toward an m1 macrophage. J Immunol 181: 698-711.

241. Wright, W. R., Parzych, K., Crawford, D., et al. 2012. Inflammatory transcriptome profiling of human monocytes exposed acutely to cigarette smoke. PLoS One 7: e30120.

242. Bosco, M. C., Puppo, M., Santangelo, C., et al. 2006. Hypoxia modifies the transcriptome of primary human monocytes: Modulation of novel immune-related genes and identification of cc- chemokine ligand 20 as a new hypoxia-inducible gene. J Immunol 177: 1941-1955.

243. Nau, G. J., Richmond, J. F. L., Schlesinger, A., et al. 2002. Human macrophage activation programs induced by bacterial pathogens. Proc Natl Acad Sci U S A 99: 1503-1508. 212

244. Pena, O. M., Hancock, D. G., Lyle, N. H., et al. 2014. An endotoxin tolerance signature predicts sepsis and organ dysfunction at initial clinical presentation. EBioMedicine 1: 64-71.

245. Smith, C. L., Dickinson, P., Forster, T., et al. 2014. Identification of a human neonatal immune- metabolic network associated with bacterial infection. Nat Commun 5: 4649.

246. Cernada, M., Serna, E., Bauerl, C., et al. 2014. Genome-wide expression profiles in very low birth weight infants with neonatal sepsis. Pediatrics 133: e1203-1211.

247. Madsen-Bouterse, S. A., Romero, R., Tarca, A. L., et al. 2010. The transcriptome of the fetal inflammatory response syndrome. Am J Reprod Immunol 63: 73-92.

248. Redline, R. W. 2004. Placental inflammation. Seminars in Neonatology 9: 265-274.

249. Macaubas, Sly, Burton, et al. 1999. Regulation of t-helper cell responses to inhalant allergen during early childhood. Clin Exp Allergy 29: 1223-1231.

250. Mack, D., Siemssen, N., and Laufs, R. 1992. Parallel induction by glucose of adherence and a polysaccharide antigen specific for plastic-adherent staphylococcus epidermidis: Evidence for functional relation to intercellular adhesion. Infect Immun 60: 2048-2057.

251. Kirkham, L. A., Corscadden, K. J., Wiertsema, S. P., et al. 2013. A practical method for preparation of pneumococcal and nontypeable haemophilus influenzae inocula that preserves viability and immunostimulatory activity. BMC Res Notes 6: 522.

252. Schmittgen, T. D., and Livak, K. J. 2008. Analyzing real-time pcr data by the comparative c(t) method. Nat Protoc 3: 1101-1108.

253. Hart, S. N., Therneau, T. M., Zhang, Y., et al. 2013. Calculating sample size estimates for rna sequencing data. J Comput Biol 20: 970-978.

254. Liu, Y., Zhou, J., and White, K. P. 2014. Rna-seq differential expression studies: More sequence or more replication? Bioinformatics 30: 301-304.

255. Liao, Y., Smyth, G. K., and Shi, W. 2013. The subread aligner: Fast, accurate and scalable read mapping by seed-and-vote. Nucleic Acids Res 41: e108.

256. Liao, Y., Smyth, G. K., and Shi, W. 2014. Featurecounts: An efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics 30: 923-930.

257. Ritchie, M. E., Phipson, B., Wu, D., et al. 2015. Limma powers differential expression analyses for rna-sequencing and microarray studies. Nucleic Acids Res.

258. Law, C. W., Chen, Y., Shi, W., et al. 2014. Voom: Precision weights unlock linear model analysis tools for rna-seq read counts. Genome Biol 15: R29.

259. Robinson, M. D., and Oshlack, A. 2010. A scaling normalization method for differential expression analysis of rna-seq data. Genome Biol 11: R25.

260. Benjamini, Y., and Hochberg, Y. 1995. Controlling the false discovery rate: A practical and powerful approach to multiple testing. J R Stat Soc Series B 57: 289-300.

261. Das, J., and Yu, H. 2012. Hint: High-quality protein interactomes and their applications in understanding human disease. BMC Syst Biol 6: 92.

262. Ideker, T., Ozier, O., Schwikowski, B., et al. 2002. Discovering regulatory and signalling circuits in molecular interaction networks. Bioinformatics 18: S233-S240.

263. Beisser, D., Klau, G. W., Dandekar, T., et al. 2010. Bionet: An r-package for the functional analysis of biological networks. Bioinformatics 26: 1129-1130.

213

264. Shannon, P., Markiel, A., Ozier, O., et al. 2003. Cytoscape: A software environment for integrated models of biomolecular interaction networks. Genome Res 13: 2498-2504.

265. Li, Y., and Xie, X. 2013. A mixture model for expression deconvolution from rna-seq in heterogeneous tissues. BMC Bioinformatics 14 Suppl 5: S11.

266. Gong, T., and Szustakowski, J. D. 2013. Deconrnaseq: A statistical framework for deconvolution of heterogeneous tissue samples based on mrna-seq data. Bioinformatics 29: 1083-1085.

267. Dang, T. M., Wong, W. C., Ong, S. M., et al. 2015. Microrna expression profiling of human blood monocyte subsets highlights functional differences. Immunology.

268. Frankenberger, M., Hofer, T. P., Marei, A., et al. 2012. Transcript profiling of cd16-positive monocytes reveals a unique molecular fingerprint. Eur J Immunol 42: 957-974.

269. Ancuta, P., Liu, K. Y., Misra, V., et al. 2009. Transcriptional profiling reveals developmental relationship and distinct biological functions of cd16+ and cd16- monocyte subsets. BMC Genomics 10: 403.

270. Zhao, C., Zhang, H., Wong, W. C., et al. 2009. Identification of novel functional differences in monocyte subsets using proteomic and transcriptomic methods. J Proteome Res 8: 4028-4038.

271. Ingram, D. A., Mead, L. E., Tanaka, H., et al. 2004. Identification of a novel hierarchy of endothelial progenitor cells using human peripheral and umbilical cord blood. Blood 104: 2752- 2760.

272. Erices, A., Conget, P., and Minguell, J. J. 2000. Mesenchymal progenitor cells in human umbilical cord blood. Br J Haematol 109: 235-242.

273. Hordyjewska, A., Popiołek, Ł., and Horecka, A. 2015. Characteristics of hematopoietic stem cells of umbilical cord blood. Cytotechnology 67: 387-396.

274. Kelley, J. L., Rozek, M. M., Suenram, C. A., et al. 1987. Activation of human blood monocytes by adherence to tissue culture plastic surfaces. Exp Mol Pathol 46: 266-278.

275. Bennett, S., and Breit, S. N. 1994. Variables in the isolation and culture of human monocytes that are of particular relevance to studies of hiv. J Leukoc Biol 56: 236-240.

276. Zhou, L., Somasundaram, R., Nederhof, R. F., et al. 2012. Impact of human granulocyte and monocyte isolation procedures on functional studies. Clin Vaccine Immunol 19: 1065-1074.

277. Loken, M. R., Brosnan, J. M., Bach, B. A., et al. 1990. Establishing optimal lymphocyte gates for immunophenotyping by flow cytometry. Cytometry 11: 453-459.

278. Zhou, L., Somasundaram, R., Nederhof, R. F., et al. 2012. Impact of human granulocyte and monocyte isolation procedures on functional studies. Clin Vaccine Immunol 19: 1065-1074.

279. Rodenburg, R. J., van den Hoogen, F. H., van de Putte, L. B., et al. 1998. Peripheral blood monocytes of rheumatoid arthritis patients do not express elevated tnf alpha, il-1beta, and il-8 mrna levels. A comparison of monocyte isolation procedures. J Immunol Methods 221: 169-175.

280. Lyons, P. A., Koukoulaki, M., Hatton, A., et al. 2007. Microarray analysis of human leucocyte subsets: The advantages of positive selection and rapid purification. BMC Genomics 8: 64.

281. Schroeder, A., Mueller, O., Stocker, S., et al. 2006. The rin: An rna integrity number for assigning integrity values to rna measurements. BMC Mol Biol 7: 3.

282. Suda, T., Osajima, A., Tamura, M., et al. 2001. Pressure-induced expression of monocyte chemoattractant protein-1 through activation of map kinase. Kidney Int 60: 1705-1715.

214

283. Chhikara, M., Wang, S., Kern, S. J., et al. 2009. Carbon monoxide blocks lipopolysaccharide- induced gene expression by interfering with proximal tlr4 to nf-kappab signal transduction in human monocytes. PLoS One 4: e8139.

284. Subrahmanyam, Y. V., Yamaga, S., Prashar, Y., et al. 2001. Rna expression patterns change dramatically in human neutrophils exposed to bacteria. Blood 97: 2457-2468.

285. DeForge, L. E., and Remick, D. G. 1991. Kinetics of tnf, il-6, and il-8 gene expression in lps- stimulated human whole blood. Biochem Biophys Res Commun 174: 18-24.

286. Strieter, R. M., Remick, D. G., Ham, J. M., et al. 1990. Tumor necrosis factor-alpha gene expression in human whole blood. J Leukoc Biol 47: 366-370.

287. Kroemer, G., Galluzzi, L., Vandenabeele, P., et al. 2009. Classification of cell death: Recommendations of the nomenclature committee on cell death 2009. Cell Death Differ 16: 3- 11.

288. Tarazona, S., Garcia-Alcalde, F., Dopazo, J., et al. 2011. Differential expression in rna-seq: A matter of depth. Genome Res 21: 2213-2223.

289. Soneson, C., and Delorenzi, M. 2013. A comparison of methods for differential expression analysis of rna-seq data. BMC Bioinformatics 14: 91.

290. Liu, Y., Zhou, J., and White, K. P. 2014. Rna-seq differential expression studies: More sequence or more replication? Bioinformatics 30: 301-304.

291. Ching, T., Huang, S., and Garmire, L. X. 2014. Power analysis and sample size estimation for rna-seq differential expression. RNA 20: 1684-1696.

292. McGuckin, C. P., Pearce, D., Forraz, N., et al. 2003. Multiparametric analysis of immature cell populations in umbilical cord blood and bone marrow. Eur J Haematol 71: 341-350.

293. Rola-Pleszczynski, M., and Stankova, J. 1992. Leukotriene b4 enhances interleukin-6 (il-6) production and il-6 messenger rna accumulation in human monocytes in vitro: Transcriptional and posttranscriptional mechanisms. Blood 80: 1004-1011.

294. Fulton, S. A., Johnsen, J. M., Wolf, S. F., et al. 1996. Interleukin-12 production by human monocytes infected with mycobacterium tuberculosis: Role of phagocytosis. Infect Immun 64: 2523-2531.

295. Kronforst, K. D., Mancuso, C. J., Pettengill, M., et al. 2012. A neonatal model of intravenous staphylococcus epidermidis infection in mice <24 h old enables characterization of early innate immune responses. PLoS One 7: e43897.

296. Mahalanabis, M., Al-Muayad, H., Kulinski, M. D., et al. 2009. Cell lysis and DNA extraction of gram-positive and gram-negative bacteria from whole blood in a disposable microfluidic chip. Lab Chip 9: 2811-2817.

297. Klinger, G., Levy, I., Sirota, L., et al. 2010. Outcome of early-onset sepsis in a national cohort of very low birth weight infants. Pediatrics 125: e736-e740.

298. Yachie, A., Takano, N., Ohta, K., et al. 1992. Defective production of interleukin-6 in very small premature infants in response to bacterial pathogens. Infect Immun 60: 749-753.

299. Guha, M., and Mackman, N. 2001. Lps induction of gene expression in human monocytes. Cellular signalling 13: 85-94.

300. Gay, N. J., Symmons, M. F., Gangloff, M., et al. 2014. Assembly and localization of toll-like receptor signalling complexes. Nat Rev Immunol 14: 546-558.

215

301. Bustamante, J., Boisson-Dupuis, S., Jouanguy, E., et al. 2008. Novel primary immunodeficiencies revealed by the investigation of paediatric infectious diseases. Curr Opin Immunol 20: 39-48.

302. Sharma, A. A., Jen, R., Butler, A., et al. 2012. The developing human preterm neonatal immune system: A case for more research in this area. Clin Immunol 145: 61-68.

303. Giogha, C., Lung, T. W., Pearson, J. S., et al. 2014. Inhibition of death receptor signaling by bacterial gut pathogens. Cytokine Growth Factor Rev 25: 235-243.

304. Harter, L., Mica, L., Stocker, R., et al. 2003. Mcl-1 correlates with reduced apoptosis in neutrophils from patients with sepsis. J Am Coll Surg 197: 964-973.

305. Gille, C., Dreschers, S., Leiber, A., et al. 2013. The cd95/cd95l pathway is involved in phagocytosis-induced cell death of monocytes and may account for sustained inflammation in neonates. Pediatr Res 73: 402-408.

306. Härtel, C., Osthues, I., Rupp, J., et al. 2008. Characterisation of the host inflammatory response to staphylococcus epidermidis in neonatal whole blood. Arch Dis Child Fetal Neonatal Ed 93: F140-F145.

307. Yerkovich, S. T., Wikstrom, M. E., Suriyaarachchi, D., et al. 2007. Postnatal development of monocyte cytokine responses to bacterial lipopolysaccharide. Pediatr Res 62: 547-552.

308. Levy, O., Zarember, K. A., Roy, R. M., et al. 2004. Selective impairment of tlr-mediated innate immunity in human newborns: Neonatal blood plasma reduces monocyte tnf-alpha induction by bacterial lipopeptides, lipopolysaccharide, and imiquimod, but preserves the response to r-848. J Immunol 173: 4627-4634.

309. Kloos, W. E., and Musselwhite, M. S. 1975. Distribution and persistence of staphylococcus and micrococcus species and other aerobic bacteria on human skin. Applied microbiology 30: 381- 395.

310. Fanaro, S., Chierici, R., Guerrini, P., et al. 2003. Intestinal microflora in early infancy: Composition and development. Acta Paediatr 92: 48-55.

311. Hajjar, A. M., O’Mahony, D. S., Ozinsky, A., et al. 2001. Cutting edge: Functional interactions between toll-like receptor (tlr) 2 and tlr1 or tlr6 in response to phenol-soluble modulin. J Immunol 166: 15-19.

312. Barreiro, L. B., Ben-Ali, M., Quach, H., et al. 2009. Evolutionary dynamics of human toll-like receptors and their different contributions to host defense. PLoS Genetics 5: e1000562.

313. Jin, W., Chang, M., and Sun, S.-C. 2012. Peli: A family of signal-responsive e3 ubiquitin ligases mediating tlr signaling and t-cell tolerance. Cell Mol Immunol 9: 113-122.

314. Cook-Mills, J. M., Marchese, M. E., and Abdala-Valencia, H. 2011. Vascular cell adhesion molecule-1 expression and signaling during disease: Regulation by reactive oxygen species and antioxidants. Antioxid Redox Signal 15: 1607-1638.

315. Motta, V., Soares, F., Sun, T., et al. 2015. Nod-like receptors: Versatile cytosolic sentinels. Physiological reviews 95: 149-178.

316. Magalhaes, J. G., Lee, J., Geddes, K., et al. 2011. Essential role of rip2 in the modulation of innate and adaptive immunity triggered by nod1 and nod2 ligands. Eur J Immunol 41: 1445- 1455.

317. Bertrand, M. J. M., Doiron, K., Labbé, K., et al. 2009. Cellular inhibitors of apoptosis ciap1 and ciap2 are required for innate immunity signaling by the pattern recognition receptors nod1 and nod2. Immunity 30: 789-801.

216

318. Garofoli, F., Borghesi, A., Mazzucchelli, I., et al. 2010. Preterm newborns are provided with triggering receptor expressed on myeloid cells-1. Int J Immunopathol Pharmacol 23: 1297-1301.

319. Read, C. B., Kuijper, J. L., Hjorth, S. A., et al. 2015. Cutting edge: Identification of neutrophil pglyrp1 as a ligand for trem-1. J Immunol 194: 1417-1421.

320. Bouchon, A., Facchetti, F., Weigand, M. A., et al. 2001. Trem-1 amplifies inflammation and is a crucial mediator of septic shock. Nature 410: 1103-1107.

321. Yu, Z., Huang, H., Mao, P., et al. 2010. [study on the natural ligand (s) of triggering receptor expressed on myeloid cell-1 on bacteria cell wall]. Zhongguo wei zhong bing ji jiu yi xue= Chinese critical care medicine= Zhongguo weizhongbing jijiuyixue 22: 335-339.

322. Arts, R. J. W., Joosten, L. A. B., van der Meer, J. W. M., et al. 2013. Trem-1: Intracellular signaling pathways and interaction with pattern recognition receptors. J Leukoc Biol 93: 209- 215.

323. Gibot, S., Kolopp-Sarda, M.-N., Béné, M.-C., et al. 2004. A soluble form of the triggering receptor expressed on myeloid cells-1 modulates the inflammatory response in murine sepsis. J Exp Med 200: 1419-1426.

324. Saldir, M., Tunc, T., Cekmez, F., et al. 2015. Endocan and soluble triggering receptor expressed on myeloid cells-1 as novel markers for neonatal sepsis. Pediatrics & Neonatology.

325. Arízaga-Ballesteros, V., Alcorta-García, M. R., Lázaro-Martínez, L. C., et al. 2015. Can strem-1 predict septic shock & death in late-onset neonatal sepsis? A pilot study. Int J Infect Dis 30: 27- 32.

326. Martin, M., Schifferle, R. E., Cuesta, N., et al. 2003. Role of the phosphatidylinositol 3 kinase- akt pathway in the regulation of il-10 and il-12 by porphyromonas gingivalis lipopolysaccharide. J Immunol 171: 717-725.

327. Miggin, S. M., and O'Neill, L. A. 2006. New insights into the regulation of tlr signaling. J Leukoc Biol 80: 220-226.

328. Su, X., Li, S., Meng, M., et al. 2006. Tnf receptor-associated factor-1 (traf1) negatively regulates toll/il-1 receptor domain-containing adaptor inducing ifn-β (trif)-mediated signaling. Eur J Immunol 36: 199-206.

329. Lai, Y., Di Nardo, A., Nakatsuji, T., et al. 2009. Commensal bacteria regulate toll-like receptor 3-dependent inflammation after skin injury. Nat Med 15: 1377-1382.

330. Ghosh, T. K., Mickelson, D. J., Solberg, J. C., et al. 2007. Tlr–tlr cross talk in human pbmc resulting in synergistic and antagonistic regulation of type-1 and 2 interferons, il-12 and tnf-α. Int Immunopharmacol 7: 1111-1121.

331. Lu, Y.-C., Yeh, W.-C., and Ohashi, P. S. 2008. Lps/tlr4 signal transduction pathway. Cytokine 42: 145-151.

332. Doyle, S. E., Vaidya, S. A., O'Connell, R., et al. 2002. Irf3 mediates a tlr3/tlr4-specific antiviral gene program. Immunity 17: 251-263.

333. Mechta-Grigoriou, F., Gerald, D., and Yaniv, M. 2001. The mammalian jun proteins: Redundancy and specificity. Oncogene 20: 2378-2389.

334. Gargi, A., Reno, M., and Blanke, S. R. 2012. Bacterial toxin modulation of the eukaryotic cell cycle: Are all cytolethal distending toxins created equally? Front Cell Infect Microbiol 2: 124.

335. Wassenaar, T. M., and Panigrahi, P. 2014. Is a foetus developing in a sterile environment? Lett Appl Microbiol 59: 572-579.

217

336. Brown, M. B., von Chamier, M., Allam, A. B., et al. 2014. M1/m2 macrophage polarity in normal and complicated pregnancy. Front Immunol 5: 606.

337. Svensson-Arvelund, J., Ernerudh, J., Buse, E., et al. 2014. The placenta in toxicology. Part ii: Systemic and local immune adaptations in pregnancy. Toxicol Pathol 42: 327-338.

338. Levy, O., Zarember, K. A., Roy, R. M., et al. 2004. Selective impairment of tlr-mediated innate immunity in human newborns: Neonatal blood plasma reduces monocyte tnf-α induction by bacterial lipopeptides, lipopolysaccharide, and imiquimod, but preserves the response to r-848. J Immunol 173: 4627-4634.

339. Levy, O., Coughlin, M., Cronstein, B. N., et al. 2006. The adenosine system selectively inhibits tlr-mediated tnf-alpha production in the human newborn. J Immunol 177: 1956-1966.

340. Hilchie, A. L., Wuerth, K., and Hancock, R. E. W. 2013. Immune modulation by multifaceted cationic host defense (antimicrobial) peptides. Nat Chem Biol 9: 761-768.

341. Bowdish, D. M. E., Davidson, D. J., Speert, D. P., et al. 2004. The human cationic peptide ll-37 induces activation of the extracellular signal-regulated kinase and p38 kinase pathways in primary human monocytes. J Immunol 172: 3758-3765.

342. Belderbos, M. E., Levy, O., Stalpers, F., et al. 2012. Neonatal plasma polarizes tlr4-mediated cytokine responses towards low il-12p70 and high il-10 production via distinct factors. PLoS One 7: e33419.

343. Hoeffel, G., Chen, J., Lavin, Y., et al. 2015. C-myb(+) erythro-myeloid progenitor-derived fetal monocytes give rise to adult tissue-resident macrophages. Immunity 42: 665-678.

344. Hoeffel, G., Wang, Y., Greter, M., et al. 2012. Adult langerhans cells derive predominantly from embryonic fetal liver monocytes with a minor contribution of yolk sac–derived macrophages. J Exp Med 209: 1167-1181.

345. Tavian, M., and Peault, B. 2005. Embryonic development of the human hematopoietic system. Int J Dev Biol 49: 243-250.

346. Moore, L. D., Le, T., and Fan, G. 2013. DNA methylation and its basic function. Neuropsychopharmacology 38: 23-38.

347. Bierne, H., Hamon, M., and Cossart, P. 2012. Epigenetics and bacterial infections. Cold Spring Harb Perspect Med 2: a010272.

348. Claverie-Martin, F., Wang, M., and Cohen, S. N. 1997. Ard-1 cdna from human cells encodes a site-specific single-strand endoribonuclease that functionally resembles escherichia coli rnase e. J Biol Chem 272: 13823-13828.

349. Kubota, T., Nishimura, K., Kanemaki, Masato T., et al. 2013. The elg1 replication factor c-like complex functions in pcna unloading during DNA replication. Molecular Cell 50: 273-280.

350. Fackelmayer, F. O., and Richter, A. 1994. Purification of two isoforms of hnrnp-u and characterization of their nucleic acid binding activity. Biochemistry 33: 10416-10422.

351. Tong, J. J., Liu, J., Bertos, N. R., et al. 2002. Identification of hdac10, a novel class ii human histone deacetylase containing a leucine-rich domain. Nucleic Acids Res 30: 1114-1123.

352. Liu, Y., Hoyo, C., Murphy, S., et al. 2013. DNA methylation at imprint regulatory regions in preterm birth and infection. Am J Obstet Gynecol 208: 395. e391-395. e397.

353. Huang, Q., Liu, D., Majewski, P., et al. 2001. The plasticity of dendritic cell responses to pathogens and their components. Science 294: 870-875.

218

354. Maheshwari, A., Kelly, D. R., Nicola, T., et al. 2011. Tgf-β 2 suppresses macrophage cytokine production and mucosal inflammatory responses in the developing intestine. Gastroenterology 140: 242-253.

355. Cuenca, A. G., Joiner, D. N., Gentile, L. F., et al. 2015. Trif-dependent innate immune activation is critical for survival to neonatal gram-negative sepsis. J Immunol 194: 1169-1177.

356. Tang, H.-Y., Beer, L. A., and Speicher, D. W. 2011. In-depth analysis of a plasma or serum proteome using a 4d protein profiling method. Methods Mol Biol 728: 47-67.

357. Zou, W., She, J., and Tolstikov, V. V. 2013. A comprehensive workflow of mass spectrometry- based untargeted metabolomics in cancer metabolic biomarker discovery using human plasma and urine. Metabolites 3: 787-819.

358. Atzori, L., Antonucci, R., Barberini, L., et al. 2011. 1h nmr-based metabolomic analysis of urine from preterm and term neonates. Front Biosci (Elite Ed) 3: 1005-1012.

359. Cabezas-Wallscheid, N., Klimmeck, D., Hansson, J., et al. 2014. Identification of regulatory networks in hscs and their immediate progeny via integrated proteome, transcriptome, and DNA methylome analysis. Cell Stem Cell 15: 507-522.

219

Appendix

A B C

Figure 7.1 Reference legend for IPA networks. (A) Symbol representations of different types of molecules. (B) The types of relationships (edges) between two molecules and (C) the IPA prediction legend.

220