Epigenetic scores indicate differences in the proteome of preterm infants

1

Epigenetic scores (EpiScores), reflecting DNA methylation (DNAm)-based surrogates for 2 complex traits, have been developed for multiple circulating proteins. EpiScores for pro-inflammatory 3 proteins, such as C-reactive protein (DNAm CRP), are associated with brain health and cognition in 4 adults and with inflammatory comorbidities of preterm birth in neonates. Social disadvantage can 5 become embedded in child development through inflammation, and deprivation is over-represented in 6 preterm infants. We tested the hypotheses that preterm birth and socioeconomic status (SES) are 7 associated with alterations in a set of EpiScores enriched for inflammation-associated proteins. 8 9

104 protein EpiScores were derived from saliva samples of 332 neonates born at gestational 10 age (GA) 22.14 to 42.14 weeks. Saliva sampling was between 36.57 and 47.14 weeks. Forty-three 11 (41%) EpiScores were associated with low GA at birth (standardised estimates |0.14 to 0.88|, 12 Bonferroni-adjusted p-value <8.3x10-3). These included EpiScores for chemokines, growth factors, 13 proteins involved in neurogenesis and vascular development, cell membrane proteins and receptors, 14 and other immune proteins. Three EpiScores were associated with SES, or the interaction between 15 birth GA and SES: afamin, intercellular adhesion molecule 5 and hepatocyte growth factor-like protein 16 (standardised estimates |0.06 to 0.13|, Bonferroni-adjusted p-value <8.3x10-3). In a preterm sub-group 17 (n=217, median [range] GA 29.29 weeks [22.14 to 33.0 weeks]), SES-EpiScore associations did not 18 remain statistically significant after adjustment for sepsis, bronchopulmonary dysplasia, necrotising 19 enterocolitis, and histological chorioamnionitis. 20 21

Low birth GA is substantially associated with a set of EpiScores. The set was enriched 22 for inflammatory proteins, providing new insights into immune dysregulation in preterm infants. SES 23 had fewer associations with EpiScores; these tended to have small effect sizes and were not 24 statistically significant after adjusting for inflammatory comorbidities. This suggests that inflammation 25 is unlikely to be the primary axis through which SES becomes embedded in the development of 26 preterm infants in the neonatal period. 27 28

Epigenetic, Inflammation, Neonatal, Preterm birth, Socioeconomic status. 29 30 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint 3 1.0 Background 1 Preterm birth (delivery at less than 37 weeks of gestation) affects around 10% of births worldwide and 2 is closely associated with increased likelihood of cerebral palsy, neurocognitive impairment, 3 behavioural, social and communication difficulties, and mental and cardiometabolic health diagnoses 4 across the life course [1–5]. These adverse outcomes can be explained, in part, by deleterious effects 5 of early exposure to extrauterine life on brain and cardiac development, and they are often 6 accompanied by changes in blood proteins, including those reflecting the perinatal innate and 7 adaptive immune response [6–8]. 8 9 Socioeconomic status (SES) is also associated with the adverse neurodevelopmental and health 10 outcomes listed above [9–12], and social deprivation is consistently over-represented among preterm 11 children and their families [13,14]. In a meta-analysis of 43 studies (n=111,156 individuals), low SES 12 associated with increased inflammatory markers of disease risk (C-reactive protein [CRP] and 13 interleukin-6 [IL6]), which suggests that pro-inflammatory pathways may be important mechanisms for 14 translating social inequalities into health disparities [15]. However, only four studies included 15 participants under 10 years of age, leaving uncertainty about SES-inflammation correlations in early 16 life [16–19]. 17 18 Although protein levels are commonly used as biomarkers of exposure and disease risk, they are 19 limited because they are often phasic in the systemic circulation, rely on venepuncture, and may not 20 capture baseline status or chronicity. For example, inflammation is often measured using acute-phase 21 inflammatory proteins such as CRP [20,21], but it is not always reliable [22], particularly in neonates, 22 and a single-time-point measure may not reflect baseline inflammation or capture chronic 23 inflammation [23]. These challenges have been addressed by the development of DNA methylation 24 (DNAm) markers of protein expression (EpiScores), which are derived from a linear weighted sum of 25 DNAm sites that are correlated with protein levels. Several EpiScores are associated with magnetic 26 resonance imaging (MRI) measures of brain health, cognition, child mental health, stroke, ischaemic 27 heart disease Alzheimer's disease, lung cancer [24–31]. In neonates, DNAm CRP is associated with 28 birth gestational age (GA), perinatal inflammatory processes, and MRI features of encephalopathy of 29 prematurity [32]. Childhood SES is associated with differential DNAm in inflammation-related genes 30 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint 4 [33,34] and at CpG sites that correlate with an inflammation index [35]. Adult SES and social mobility 1 are associated with variations in DNAm in inflammation-related genes [33,36]. Importantly, SES-2 related DNAm variations are associated with differences in gene expression so may have functional 3 consequences [33,36]. 4 5 We investigated relationships between preterm birth, SES, and 104 EpiScores enriched for 6 inflammation-related proteins [26–28,31,37]. We tested the following hypotheses: first, low GA is 7 associated with differences in EpiScores; second, SES is correlated with EpiScores, and interacts 8 with birth GA, but the relationship is attenuated by inflammatory disease burden in preterm infants. 9 10 2.0 Methods 11 2.1 Participants 12 Participants were preterm infants (born /g340933 weeks’ gestation) and term-born infants born at the Royal 13 Infirmary of Edinburgh, UK. These infants were recruited to a longitudinal cohort study designed to 14 investigate the effect of preterm birth on brain development and outcomes with multimodal data 15 collection [38]. Infants were recruited between February 2012 and December 2021. 16 17 Exclusion criteria were congenital malformation, chromosomal abnormality, congenital infection, cystic 18 periventricular leukomalacia, haemorrhagic parenchymal infarction, and post-haemorrhagic 19 ventricular dilatation. These criteria mean the cohort is representative of the majority of survivors of 20 modern intensive care practices [38]. 21 22 Final participants included were 217 preterm infants (born /g340933 weeks’ GA) and 115 term-born infants, 23 with median birth GA of 29.29 weeks and 39.71 weeks respectively. Their demographic 24 characteristics are shown in Table 1. The three SES measures (Scottish Index of Multiple Deprivation 25 (SIMD 2016) [39], maternal education and maternal occupation) differed between the preterm and 26 term groups (Cohen’s d effect sizes 0.52-0.68). Ethnicity did not differ between groups and is 27 representative of the Edinburgh area [40]. 28 29 [Table 1 near here] 30 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint 5 1 2.2 DNA methylation 2 Saliva samples for DNAm were collected at term-equivalent age using Oragene OG-575 Assisted 3 Collection kits (DNA Genotek, ON, Canada) and DNA was extracted using prepIT.L2P reagent (DNA 4 Genotek, ON, Canada). Saliva sampling was used due to accessibility and the non-invasiveness of 5 the method; DNAm patterns measured via saliva samples correlate with brain and other tissue DNAm 6 patterns [41,42]. We chose to sample at the term-corrected gestation timepoint to include the 7 allostatic load of both prenatal and early postnatal exposures. 8 9 DNA was bisulphite converted and methylation levels were measured using Illumina 10 HumanMethylationEPIC BeadChip (Illumina, San Diego, CA, USA) at the Edinburgh Clinical 11 Research Facility (Edinburgh, UK). The arrays were imaged on the Illumina iScan or HiScan platform, 12 and genotypes were called automatically using GenomeStudio Analysis software version 2011.1 13 (Illumina). DNAm was processed in four batches. 14 15 Raw intensity (.idat) files were read into the R environment using minfi. wateRmelon and minfi were 16 used for preprocessing, quality control and normalisation [43]. The pfilter function in wateRmelon was 17 used to exclude samples with 1% of sites with a detection p-value >0.05, sites with beadcount 0.05. Cross hybridising probes, 19 probes targeting single nucleotide polymorphisms with overall minor allele frequency ≥ 0.05, and 20 control probes were also removed. Samples were removed if there was a mismatch between 21 predicted sex (minfi) and recorded sex (n=3), or if samples did not meet preprocessing quality control 22 criteria (n=29). Data were danet normalised, which includes background correction and dye bias 23 correction [43]. Saliva contains different cells types, including buccal epithelial cells. Epithelial cell 24 proportions were estimated with epigenetic dissection of intra-sample heterogeneity with the reduced 25 partial correlation method implemented in the R package EpiDISH [44]. Probes located on sex 26 chromosomes were removed before analysis. The cohort includes twins (n=32); these were randomly 27 removed leaving one participant per twin pair. This left a final sample size of n=332. 28 29 2.3 EpiScore calculation 30 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint 6 The 104 protein EpiScores included 100 EpiScores from Gadd et al [26], a study enriched for 1 inflammatory-related proteins, excluding those where the required CpGs were not available, owing to 2 differences in assay platform CpG coverage relative to Gadd et al [26]. For duplicate proteins, those 3 developed using Olink platform-identified proteins (antibody-based assays) were prioritised over those 4 from SOMAscan platforms (aptamer-based assays), due to specificity and reproducibility, and the 5 variable correlation between the two methods [45–48] (for details see Supplementary eMethods, 6 Additional File 1). In addition, we included EpiScores for growth and differentiation factor 15 (GDF15) 7 and N-terminal-pro B-type natriuretic peptide (NTproBNP) [37], and CRP. The CRP EpiScore used 8 was Barker et al’s seven-CpG variation of Ligthart et al’s CRP EpiScore [27,31], as this is known to 9 correlate with birth GA, perinatal pro-inflammatory exposures, and neonatal brain development [32]. 10 11 For each individual, EpiScores were obtained by multiplying the methylation proportion at a given 12 CpG by the effect size from previous studies. This was performed using the MethylDetectR platform 13 [49] for those inflammatory proteins currently included and using R for those not currently included 14 (CRP, GDF15, IL6, NTproBNP). All CpG sites and coefficients required to calculate the 104 15 EpiScores are in Supplementary Table 1 (Additional File 2). 16 17 2.4 Statistics 18 The predictor variables were SES and birth GA. SES was operationalised in three ways: 19 neighbourhood-level SES using the Scottish Index of Multiple Deprivation (SIMD) [39], and two 20 measures of family-level SES, which were maternal education (highest educational qualification) and 21 maternal occupation (current or most recent occupation). For further details, see Supplementary 22 eMethods, Additional File 1. Birth GA was a continuous variable to maximise statistical power [50,51]. 23 We adjusted for GA at saliva sampling, DNAm batch, infant sex, and birthweight z-score. 24 25 All statistical analyses were performed in R (version 4.3.1) and were preregistered [52]. 26 27 Principal component analysis (PCA) was used to determine the significance threshold for controlling 28 type 1 error in analyses of multiple EpiScores [53]. We began with a correlation analysis, which 29 showed correlation coefficients between EpiScores of |0.01 to 0.93| (Figure 1A). To determine the 30 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint 7 number of statistical “families” among the 104 EpiScores, PCA was performed. This yielded two 1 principal components with eigenvalues >1, our pre-specified threshold, which explained 59.5% and 2 17.2% of variance, respectively (Figure 1B). Standardised component loadings are provided in 3 Supplementary Table 3 (Additional File 1). In all subsequent analyses, we corrected for multiple 4 comparisons across EpiScores and SES measures using a Bonferroni-adjusted p-value threshold of 5 8.3x10-3. This is 0.05/(2 x 3), with two reflecting the two principal components for Episcores and three 6 reflecting the number of SES measures used. 7 8 [Figure 1 near here] 9 10 We constructed generalised linear regression models for each EpiScore as outcome measure to 11 assess associations between GA, each of the three SES measures (separate models for each of 12 SIMD, maternal education, and maternal occupation), and the product interaction term SES*birth GA 13 (removing the term if not significant), and adjusting for GA at sampling, sex, and batch. 14 15 For the preterm sub-group, we then additionally adjusted for perinatal inflammatory exposures known 16 to be associated with the CRP EpiScore as, to our knowledge, this is the only DNAm proxy of an 17 inflammatory protein that has been studied in this context [32]. These were histological 18 chorioamnionitis (HCA), sepsis, bronchopulmonary dysplasia (BPD) and necrotising enterocolitis 19 (NEC). For definitions see Supplementary eMethods (Additional File 1) and for frequencies see 20 Supplementary Table 2 (Additional File 1). 21 22 3.0 Results 23 3.1 Associations between gestational age and EpiScores 24 Gestational age associated with 43 of the 104 EpiScores after adjustment for SIMD, maternal 25 education, or maternal occupation (Figure 2A-C). The proteins represented by the 43 EpiScores are 26 listed in Table 2 categorised by functional annotation adapted from the STRING database[54], and 27 their broader roles in immune processes and inflammation, and the pathogenesis of neonatal 28 diseases, where known, are described in Supplementary Table 4 (Additional File 1). 29 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint 8 29 (67%) EpiScores negatively associated with birth GA (standardised estimates |0.14-0.76|, adjusted 1 p-value <8.3x10-3), and 14 (33%) EpiScores positively associated with birth GA (standardised 2 estimates 0.14-0.88, adjusted p-value <8.3x10-3). 3 33 EpiScores associated with low GA irrespective of SES measure used in the model. The results for 4 all 104 EpiScores are provided in Supplementary Figures 1-9 (Additional File 1). 5 6 [Table 2 near here] 7 [Figure 2 near here] 8 9 3.2 Associations between SES, EpiScores, and the effect of inflammatory comorbidities of preterm 10 birth. 11 Three out of 104 EpiScores associated with SES measures or the interaction between SES and GA 12 (Figure 3). There was a small effect size association between higher afamin EpiScore and higher 13 maternal occupation (standardised /g2010=0.06, 95% confidence interval (CI) 0.02-0.11, p=0.0082), and 14 DNAm afamin associated with the birth GA*maternal education interaction term such that afamin 15 positively correlated with birth GA among babies with mothers without university education, and 16 negatively correlated with birth GA among babies with mothers with university education 17 (undergraduate or postgraduate) (standardised /g2010=-0.12, 95% CI -0.20 to -0.04, p=0.0041, 18 Supplementary Figure 10, Additional File 1). 19 20 Higher intercellular adhesion molecule 5 (ICAM5) EpiScore associated with higher SIMD 21 (standardised /g2010=0.13, 95% CI 0.03-0.23, p=0.0079). Hepatocyte growth factor-like protein (HGFI) 22 associated with the birth GA*maternal occupation interaction term (standardised /g2010=-0.10, 95% CI -23 0.16- -0.04, p=0.0021, Supplementary Figure 11, Additional File 1), such that HGFI EpiScore 24 positively correlated with birth GA among babies with mothers who were unemployed, homemakers, 25 in full time education, or in unskilled or manual occupations, but negatively correlated with birth GA 26 among babies with mothers in partly skilled, non-manual skilled or professional occupations. 27 28 In the planned analysis of preterm-born babies only, when controlling for inflammatory exposures 29 (sepsis, HCA, NEC, and BPD), a greater proportion of R2 was explained (R2=0.053-0.28 in unadjusted 30 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint 9 models, R2=0.029-0.437 in adjusted models), but the EpiScores no longer met our statistical threshold 1 (p-values >0.045 with adjusted p-value threshold<8.3x10-3, see Supplementary Table 5, Additional 2 File 1). 3 4 [Figure 3 near here] 5 6 4.0 Discussion 7 In this study, we identified several associations between a set of EpiScores enriched for inflammatory 8 proteins and low GA at birth. Few EpiScores associated with SES within the whole sample and these 9 associations were partially attenuated in preterm infants who experienced inflammatory comorbidities. 10 This is the first study to assess the impact of preterm birth and social status using epigenetic 11 signatures designed to reflect the circulating proteome. 12 13 4.1 Associations between birth GA and EpiScores 14 43 EpiScores associated with preterm birth when sampled at term equivalent age. The EpiScores 15 reflect chemokines, growth factors, proteins required for neurogenesis and vascular development, cell 16 membrane proteins and receptors, and immune response proteins (Table 2). As well as having 17 specific immunoregulatory roles, in the neonatal setting or relevant animal models, these proteins are 18 associated with several comorbidities and developmental consequences of preterm birth. These 19 include lung development and disease such as BPD [55–67], in utero and postnatal growth failure 20 [68–71], HCA [7,72,73], patent ductus arteriosus [74–77], retinopathy of prematurity [78–82], NEC 21 [83–85], hyperglycaemia [86], sepsis [85,87–90], brain injury [32,91–94], and neurodevelopmental 22 outcomes [95–99]. 23 24 4.2 Associations between SES measures and EpiScores 25 SES appears to play a much smaller role in the patterning of EpiScores compared to birth GA. SES 26 measures, or interactions between birth GA and SES measures, were associated with only three of 27 the 104 EpiScores studied: afamin, ICAM5 and HGFI. Afamin and ICAM5 positively associated with 28 maternal occupation and SIMD, respectively. Afamin and HGFI both associated with the interaction 29 term between SES and birth GA. Afamin is involved in vitamin E transport [100], ICAM5 has a role in 30 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint 10 microglial regulation [101], and HGFI is a macrophage-stimulating protein [102]. The relationships 1 between these proteins and SES have not previously been investigated, although afamin is 2 associated with the development of metabolic syndrome [103], which varies with SES [104]. 3 4 Among preterm infants, no SES-EpiScore associations survived adjustment for inflammatory 5 exposures, which suggests that the weak effects of SES on the neonatal proteome that we observed 6 in a small number of EpiScores are at least partially accounted for by inflammatory pathologies in 7 early life. Taken together, the results suggest immune dysregulation, proxied by EpiScores, may not 8 be the primary axis through which SES becomes embedded in the development of preterm infants 9 during neonatal intensive care. 10 11 SES has been consistently associated with inflammation in adulthood, including in relation to 12 childhood deprivation [15,105], but less is known about the relationship between SES and 13 inflammation in the neonatal period. A longitudinal study by Leviton et al [106], with five sampling 14 timepoints during the first month of life after preterm birth, showed that maternal eligibility for Medicaid 15 associated with levels of 14 inflammatory proteins (IL6R, TNFR1, TNFR2, IL8, ICAM1, VCAM1, TSH, 16 EPO, bFGF, IGF1, VEGF, PIGF, Ang-1, Ang-2). However, only three were significant at more than 17 one timepoint during the month after preterm birth (TSH, bFGF, Ang-1), and none associated at all 18 five measurement timepoints. These studies, taken together with our results, suggest that the impact 19 of SES on immune regulation is relatively modest and inconsistent in the newborn period but accrues 20 through to adulthood. Further research is required to understand how and when SES becomes 21 embedded in child development and whether early life events such as preterm birth modify that 22 process; EpiScores could be a powerful tool for investigating the temporal dynamics of social 23 determinants of child health. 24 25 4.3 Strengths and limitations 26 Strengths of this study include the large sample of term and preterm neonates; to the best of our 27 knowledge, this is the first examination of multiple DNA methylation-based estimators of circulating 28 proteins in a neonatal sample. We derived EpiScores from minimally-invasive sampling (buccal cells 29 from saliva) which overcomes the ethical challenge of venepuncture for research in children. The 30 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint 11 EpiScores, serving as proxies of inflammatory proteins and sampled at term-equivalent age in 1 preterm infants, were selected for their potential to capture chronic, cumulative inflammation 2 associated with preterm birth and neonatal intensive care exposures [23–25]. We adjusted for 3 variables associated with DNAm, and additionally for inflammatory exposures to increase the clinical 4 validity of our results. 5 6 The study has some limitations. The EpiScores used were developed in adult cohorts [26–28,37] and 7 have not been validated in neonatal populations with neonatal protein levels, although we have 8 previously established that neonatal DNAmCRP scores correlate with cumulative inflammatory 9 exposures [32]. Further studies that evaluate serial blood protein levels with EpiScores could be 10 informative but would need to rely on small volume samples taken during time of venepuncture for 11 clinical reasons, given the practical and ethical challenges of serial venepuncture for research 12 purposes in neonates. The EpiScores were also trained using blood samples [26–28,37], whereas we 13 have projected these scores into saliva samples. However, previous studies have successfully used 14 similar cross-tissue techniques [32,107,108], and in neonates saliva provides a non-invasive and 15 accessible sample method. Not all inflammatory-related proteins are represented, as we were limited 16 by available EpiScores, so we may have underestimated the full complexity of the relationship 17 between birth GA, SES, and inflammation. Longitudinal investigations are imperative for elucidating 18 whether the DNA methylation signatures associated with gestational age identified in this study exert 19 a causal influence on the inflammation-associated mechanisms in preterm birth. It remains crucial to 20 discern whether these signatures represent a direct downstream consequence of GA itself or are 21 induced by specific factors correlated with GA, yet not necessarily driven by chronic inflammation. 22 Mendelian randomization studies, integrating genomic and epigenomic determinants, are a promising 23 methodological approach to disentangle the directionality of these intricate relationships. The study 24 population is comparable to other neonatal populations in high-income, majority-white settings, but 25 these results may not generalise to settings with different socioeconomic or ethnicity profiles. We 26 studied several measures of SES but were not able to include all that could be relevant, such as 27 household income. 28 29 5.0 Conclusion 30 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint 12 We identified 43 EpiScores enriched for inflammatory proteins that associated with low birth GA. 1 These 43 proteins offer novel insights into the physiological response to preterm birth and warrant 2 further study to explore their role in the relationship between preterm birth, inflammation, and longer-3 term outcomes. We found only three EpiScores associated with SES in the neonatal period, none of 4 which survived adjustment for perinatal pro-inflammatory exposures, suggesting that inflammation is 5 unlikely to be the primary axis through which SES becomes embedded in the development of preterm 6 infants in the neonatal period. 7 8 6.0 Declarations 9 6.1 Ethics approval and consent to participate 10 Ethical approval for the study was obtained from the National Research Ethics Service, South-East 11 Scotland Research Ethics Committee (REC 11/55/0061, 13/SS/0143, 16/SS/0154), and NHS Lothian 12 Research and Development (2016/0255). Written informed consent was obtained from parents. 13 14 6.2 Consent for publication 15 Not applicable. 16 17 6.3 Availability of data and materials 18 DNA methylation data are available to researchers subject to the terms of the Data Access Policy: 19 https://www.ed.ac.uk/centre-reproductive-health/tebc/about-tebc/for-researchers/data-access-20 collaboration. 21 22 6.4 Competing interests 23 LM has received speaker and consultancy fees from Illumina. REM is a scientific advisor to the 24 Epigenetic Clock Development Foundation and to Optima Partners. All other authors declare that they 25 have no competing interests. 26 27 6.5 Funding 28 This work was supported by Theirworld (www.theirworld.org) and a UKRI MRC Programme Grant 29 MR/X003434/1. KM receives salary from NHS Scotland. GS is funded by an MRC Clinician Scientist 30 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint 13 Fellowship MR/X019535/1. SRC is supported by a Sir Henry Dale Fellowship jointly funded by the 1 Wellcome Trust and the Royal Society (221890/Z/20/Z). 2 3 6.6 Authors’ contributions 4 KM: Conceptualisation, methodology, formal analysis, writing (original draft), writing (review/editing), 5 visualization. ELSC: Formal analysis, writing (review/editing). KV: Formal analysis, writing 6 (review/editing), visualization. RH: Methodology, formal analysis, writing (review/editing). DG: 7 Methodology, formal analysis, writing (review/editing). JB: Formal analysis, writing (review/editing). 8 GS: Investigation, writing (review/editing). AJS: Methodology, writing (review/editing). AC: 9 Investigation, resources, data curation, writing (review/editing). LM: Investigation, resources, writing 10 (review/editing). HCW: Methodology, writing (review/editing). HR: Conceptualisation, writing (original 11 draft), writing (review/editing), supervision. REM: Conceptualisation, methodology, writing (original 12 draft), writing (review/editing), supervision. SRC: Conceptualisation, methodology, writing (original 13 draft), writing (review/editing), supervision. JPB: Conceptualisation, methodology, writing (original 14 draft), writing (review/editing), supervision, project administration, funding acquisition. 15 16 6.7 Acknowledgements 17 The authors are grateful to the families who consented to take part in the study. 18 19 20 21 22 23 24 25 26 27 28 29 30 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint 14 Figure legends 1 Figure 1. Determining significance threshold 2 Principal component analysis was used to determine the adjusted statistical significance threshold, 3 given multiple statistical comparisons. Figure 1A shows a correlation matrix of 104 EpiScores, 4 showing correlation coefficient as red for positive and blue negative associations when significant 5 (p<0.05). Figure 1B shows a scree plot of principal components, with the eigenvalues for each 6 component. Standardised component loadings for principal components one and two are provided in 7 Supplementary Table 3 (Additional File 1). 8 9 Figure 2. EpiScores associated with gestational age in regression models adjusted for socioeconomic 10 status 11 EpiScores associated with gestational age in regression models with (A) Scottish Index of Multiple 12 Deprivation, (B) maternal education, and (C) maternal occupation. Figure 2A (n=331) shows 39 13 associations, Figure 2B (n=323) shows 35 associations, and Figure 2C shows (n=328) shows 39 14 associations. Points and bars represent standardised beta and 95% confidence intervals, with red 15 indicating positive and blue negative associations. Covariates included in all models: age at sample, 16 birthweight z-score, sex, and methylation processing batch. Bonferroni-adjusted p-value <8.3x10-3. 17 CCL11 – C-C chemokine 11, CCL18 – C-C chemokine 18, CCL21 – C-C chemokine 21, CCL22 – C-18 C chemokine 22, CCL25 – C-C chemokine 25, CD5L – CD5 antigen-like protein, CD6 – T-cell 19 differentiation antigen, CD163 – scavenger receptor cysteine-rich type 1 protein M130, CI – 20 confidence interval, CRP – C-reactive protein, CRTAM – Cytotoxic and regulatory T-cell molecule, 21 CXCL9 – C-X-C motif chemokine 9, CXCL10 – C-X-C motif chemokine 10, FAP – Fibroblast 22 Activation Protein alpha, FCGR3B – Low affinity immunoglobulin gamma Fc region receptor III-B, 23 FcRL2 – Fc receptor-like protein 2, FGF21 – Fibroblast growth factor 21, GDF15 – 24 Growth/differentiation factor 15, GHR – Growth hormone receptor, HCII – Heparin cofactor II, HGF – 25 Hepatocyte growth factor alpha chain, IGFBP4 – Insulin-like growth factor-binding protein 4, MMP9 – 26 Matrix metalloproteinase-9, NCAM1 – Neural cell adhesion molecule 1, PIGR – Polymeric 27 immunoglobulin receptor, SCGF – stem cell growth factor, SIMD – Scottish Index of Multiple 28 Deprivation, SKR3 – Serine/threonine-protein kinase receptor R3, SLITRK5 – SLIT and NTRK-like 29 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint 15 protein 5, VCAM1 – Vascular cell adhesion protein 1, VEGFA – Vascular endothelial growth factor A, 1 WFIKKN2 – WAP Kazal immunoglobulin Kunitz and NTR domain-containing protein 2. 2 3 Figure 3. EpiScores associated with socioeconomic status or an interaction between socioeconomic 4 status and birth gestational age 5 EpiScores associated with socioeconomic status (Scottish Index of Multiple Deprivation, maternal 6 education, or maternal occupation), or with an interaction between socioeconomic status and birth 7 gestational age. Regression models with gestational age at birth, gestational age at sample, 8 birthweight z-score, sex, and methylation processing batch. Sample sizes: for the Scottish Index of 9 Multiple Deprivation n=331, for maternal education n=323, and for maternal occupation n=328. 3/104 10 EpiScores were significant (Bonferroni-adjusted p-value <8.3x10-3). Points and bars represent 11 standardised beta and 95% confidence intervals, with red indicating positive and blue negative 12 associations. 13 CI – confidence interval, GA – gestational age, HGFI – Hepatocyte growth factor-like protein alpha 14 chain, ICAM5 – Intercellular adhesion molecule 5, SIMD – Scottish Index of Multiple Deprivation. 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint 16 Tables 1 Table 1. Participant characteristics 2 Demographic measure Preterm (n=217) Term (n=115) Sex Male 114/217 (52.5%) 64/115 (55.7%) Female 103/217 (47.5%) 51/115 (44.3%) Birth GA (weeks) - median (range) 29.29 (22.14-33.0) 39.71 (37.0-42.14) Birthweight (g) - median (range) 1200 (370-2510) 3450 (2346-4670) Birthweight z-score - median (range) 0.10 (-3.13-2.07) 0.43 (-2.3-2.96) SIMD rank - median (range)A 3720 (6-6966) 5344 (267-6967) Maternal ethnicity African 1/217 (0.5%) 0/115 (0%) Bangladeshi 0/217 (0%) 1/115 (0.9%) Caribbean 0/217 (0%) 0/115 (0%) Chinese 0/217 (0%) 1/115 (0.9%) Indian 3/217 (2.4%) 1/115 (0.9%) Pakistani 4/214 (1.8%) 1/115 (0.9%) White 195/217 (89.9%) 106/115 (92.2%) White/Asian 2/217 (0.9%) 0/115 (0%) White/Black African 0/217 (0%) 1/115 (0.9%) White/Black Caribbean 1/217 (0.5%) 1/115 (0.9%) Other AsianB 1/217 (0.5%) 2/115 (1.7%) Other ethnic groupC 6/217 (2.8%) 0/115 (0%) Other mixed ethnic backgroundD 4/217 (1.8%) 1/115 (0.9%) Maternal education None 7/208 (3.4%) 0/115 (0%) Basic high school qualification (1-4) 5/208 (2.4%) 2/115 (1.7%) Basic high school qualification (>/=5) 8/208 (3.8%) 1/115 (0.9%) Advanced high school 32/208 (15.4%) 3/115 (2.6%) . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint 17 qualification College qualification 46/208 (22.1%) 8/115 (7.0%) University undergraduate 61/208 (29.3%) 50/115 (43.5%) University postgraduate 49/208 (23.6%) 51/115 (44.3%) Maternal occupation Unemployed 12/213 (5.6%) 1/115 (0.9%) Homemaker 10/213 (4.7%) 1/115 (0.9%) Still in full time education 9/213 (4.2%) 1/115 (0.9%) Sheltered employment 1/213 (0.5%) 0/115 (0%) Unskilled 11/213 (5.2%) 3/115 (2.6%) Partly skilled 7/213 (3.3%) 2/115 (1.7%) Manual skilled 25/213 (11.7%) 9/115 (7.8%) Non-manual skilled 44/213 (20.7%) 11/115 (9.6%) Professional 94/213 (44.1%) 87/115 (75.7%) Sample GA (weeks) - median (range) 40.57 (36.57-45.86) 42.0 (39.86-47.14) Batch 1 92/217 (42.4%) 46/115 (40.0%) 2 62/217 (28.6%) 56/115 (48.7%) 3 32/217 (14.7%) 1/115 (0.9%) 4 31/217 (14.3%) 12/115 (10.4%) A: Preterm n=216, term n=115 1 B: “Other Asian” includes Sri Lankan (n=1), Malay (n=1) 2 C: “Other ethnic group” includes Arab (n=1), Iraqi (n=1), white Bulgarian (n=1), Fijian (n=1), Japanese 3 (n=1), Hong Kong (n=1). 4 D: “Other mixed ethnic background” includes Pakistani/Scottish (n=2), British/Arab (n=1), Sri 5 Lankan/Black (n=1), Sri Lankan/Indian (n=1). 6 GA = gestational age, SIMD = Scottish index of multiple deprivation. 7 8 9 10 11 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint 18 Table 2. Protein EpiScores associated with birth gestational age 1 Chemokines Growth factors Neurogenesis Vascular development Cell membrane proteins / receptors Other immune response CCL11 CCL18 CCL21 CCL22 CCL25 CXCL9 CXCL10 FGF21 GDF15 GHR HGF IGFBP4 SCGF alpha SCGF beta WFIKKN2 NCAM1 Semaphorin 3E SLITRK5 SKR3 VCAM1 VEGFA Afamin CD5L CD6 CD48 CD163 CD209 Contactin-4 FcRL2 FcGR3B PIGR Thrombopoeitin receptor Complement C9 CRP CRTAM FAP HCII L-selectin Lactotransferrin MMP9 S100A9 Sialoadhesin Trypsin-2 43 EpiScores associated with birth gestational age in regression models adjusted for socioeconomic status. Roles adapted from the STRING database[54]. 2 See Supplementary Table 4, Additional File 1, for further details of the functional roles of each protein, including roles in immunity and inflammation, and in 3 preterm infants specifically. 4 CCL11 – C-C chemokine 11, CCL18 – C-C chemokine 18, CCL21 – C-C chemokine 21, CCL22 – C-C chemokine 22, CCL25 – C-C chemokine 25, CD5L – 5 CD5 antigen-like protein, CD6 – T-cell differentiation antigen, CD163 – scavenger receptor cysteine-rich type 1 protein M130, CRP – C-reactive protein, 6 CRTAM – Cytotoxic and regulatory T-cell molecule, CXCL9 – C-X-C motif chemokine 9, CXCL10 – C-X-C motif chemokine 10, FAP – Fibroblast Activation 7 Protein alpha, FCGR3B – Low affinity immunoglobulin gamma Fc region receptor III-B, FcRL2 – Fc receptor-like protein 2, FGF21 – Fibroblast growth factor 8 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint 19 21, GDF15 – Growth/differentiation factor 15, GHR – Growth hormone receptor, HCII – Heparin cofactor II, HGF – Hepatocyte growth factor alpha chain, 1 IGFBP4 – Insulin-like growth factor-binding protein 4, MMP9 – Matrix metalloproteinase-9, NCAM1 – Neural cell adhesion molecule 1, PIGR – Polymeric 2 immunoglobulin receptor, SCGF – stem cell growth factor, SIMD – Scottish Index of Multiple Deprivation, SKR3 – Serine/threonine-protein kinase receptor 3 R3, SLITRK5 – SLIT and NTRK-like protein 5, VCAM1 – Vascular cell adhesion protein 1, VEGFA – Vascular endothelial growth factor A, WFIKKN2 – WAP 4 Kazal immunoglobulin Kunitz and NTR domain-containing protein 2. 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint 20

1 1. Ohuma EO, Moller A-B, Bradley E, Chakwera S, Hussain-Alkhateeb L, Lewin A, et al. National, 2 regional, and global estimates of preterm birth in 2020, with trends from 2010: a systematic analysis. 3 Lancet. 2023;402:1261–71. 4 2. Johnson S, Marlow N. Early and long-term outcome of infants born extremely preterm. Archives of 5 Disease in Childhood. 2017;102:97–102. 6 3. Agrawal S, Rao SC, Bulsara MK, Patole SK. Prevalence of Autism Spectrum Disorder in Preterm 7 Infants: A Meta-analysis. Pediatrics. 2018;142:1–14. 8 4. Twilhaar ES, Wade RM, deKieviet JF, vanGoudoever JB, vanElburg RM, Oosterlaan J. Cognitive 9 Outcomes of Children Born Extremely or Very Preterm Since the 1990s and Associated Risk Factors: 10 A Meta-analysis and Meta-regression. Jama Pediatr. 2018;172:361. 11 5. Crump C. An overview of adult health outcomes after preterm birth. Early Hum Dev. 12 2020;150:105187. 13 6. Hagberg H, Mallard C, Ferriero DM, Vannucci SJ, Levison SW, Vexler ZS, et al. The role of 14 inflammation in perinatal brain injury. Nat Rev Neurol. 2015;11:192–208. 15 7. Sullivan G, Galdi P, Blesa-Cábez M, Borbye-Lorenzen N, Stoye DQ, Lamb GJ, et al. Interleukin-8 16 dysregulation is implicated in brain dysmaturation following preterm birth. Brain Behav Immun. 17 2020;90:311–8. 18 8. Raisi-Estabragh Z, Cooper J, Bethell MS, McCracken C, Lewandowski AJ, Leeson P, et al. Lower 19 birth weight is linked to poorer cardiovascular health in middle-aged population-based adults. Heart. 20 2023;109:535–41. 21 9. Farah MJ. The Neuroscience of Socioeconomic Status: Correlates, Causes, and Consequences. 22 Neuron. 2017;96:56–71. 23 10. Kivimäki M, Batty GD, Pentti J, Shipley MJ, Sipilä PN, Nyberg ST, et al. Association between 24 socioeconomic status and the development of mental and physical health conditions in adulthood: a 25 multi-cohort study. Lancet Public Heal. 2020;5:e140–9. 26 11. Duncan GJ, Ziol‐ Guest KM, Kalil A. Early‐ Childhood Poverty and Adult Attainment, Behavior, and 27 Health. Child Dev. 2010;81:306–25. 28 12. Pillas D, Marmot M, Naicker K, Goldblatt P, Morrison J, Pikhart H. Social inequalities in early 29 childhood health and development: a European-wide systematic review. Pediatr Res. 2014;76:418–30 24. 31 13. Thomson K, Moffat M, Arisa O, Jesurasa A, Richmond C, Odeniyi A, et al. Socioeconomic 32 inequalities and adverse pregnancy outcomes in the UK and Republic of Ireland: a systematic review 33 and meta-analysis. Bmj Open. 2021;11:e042753. 34 14. Ruiz M, Goldblatt P, Morrison J, Kukla L, Švancara J, Riitta-Järvelin M, et al. Mother’s education 35 and the risk of preterm and small for gestational age birth: a DRIVERS meta-analysis of 12 European 36 cohorts. J Epidemiol Commun H. 2015;69:826–33. 37 15. Muscatell KA, Brosso SN, Humphreys KL. Socioeconomic status and inflammation: a meta-38 analysis. Mol Psychiatr. 2020;25:2189–99. 39 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint 21 16. Broyles ST, Staiano AE, Drazba KT, Gupta AK, Sothern M, Katzmarzyk PT. Elevated C-Reactive 1 Protein in Children from Risky Neighborhoods: Evidence for a Stress Pathway Linking Neighborhoods 2 and Inflammation in Children. Plos One. 2012;7:e45419. 3 17. Dowd JB, Zajacova A, Aiello AE. Predictors of Inflammation in U.S. Children Aged 3–16 Years. 4 Am J Prev Med. 2010;39:314–20. 5 18. Schmeer KK, Yoon AJ. Home sweet home? Home physical environment and inflammation in 6 children. Soc Sci Res. 2016;60:236–48. 7 19. Schmeer KK, Yoon A. Socioeconomic status inequalities in low-grade inflammation during 8 childhood. Arch Dis Child. 2016;101:1043. 9 20. Chiesa C, Natale F, Pascone R, Osborn JF, Pacifico L, Bonci E, et al. C reactive protein and 10 procalcitonin: Reference intervals for preterm and term newborns during the early neonatal period. 11 Clin Chim Acta. 2011;412:1053–9. 12 21. Brown JVE, Meader N, Cleminson J, McGuire W. C-reactive protein for diagnosing late-onset 13 infection in newborn infants. Group CN, editor. Cochrane Database of Systematic Reviews. 14 2019;10:16–78. 15 22. Bower JK, Lazo M, Juraschek SP, Selvin E. Within-Person Variability in High-Sensitivity C-16 Reactive Protein. Arch Intern Med. 2012;172:1519–21. 17 23. Dammann O, Leviton A. Intermittent or sustained systemic inflammation and the preterm brain. 18 Pediatr Res. 2014;75:376–80. 19 24. Stevenson AJ, McCartney DL, Harris SE, Taylor AM, Redmond P, Starr JM, et al. Trajectories of 20 inflammatory biomarkers over the eighth decade and their associations with immune cell profiles and 21 epigenetic ageing. Clin Epigenetics. 2018;10:159. 22 25. Stevenson AJ, McCartney DL, Hillary RF, Campbell A, Morris SW, Bermingham ML, et al. 23 Characterisation of an inflammation-related epigenetic score and its association with cognitive ability. 24 Clin Epigenetics. 2020;12:113. 25 26. Gadd DA, Hillary RF, McCartney DL, Zaghlool SB, Stevenson AJ, Cheng Y, et al. Epigenetic 26 scores for the circulating proteome as tools for disease prediction. Elife. 2022;11:e71802. 27 27. Ligthart S, Marzi C, Aslibekyan S, Mendelson MM, Conneely KN, Tanaka T, et al. DNA 28 methylation signatures of chronic low-grade inflammation are associated with complex diseases. 29 Genome Biol. 2016;17:255. 30 28. Stevenson AJ, Gadd DA, Hillary RF, McCartney DL, Campbell A, Walker RM, et al. Creating and 31 Validating a DNA Methylation-Based Proxy for Interleukin-6. Journals Gerontology Ser Biological Sci 32 Medical Sci. 2021;76:2284–92. 33 29. Conole ELS, Stevenson AJ, Maniega SM, Harris SE, Green C, Hernández M del CV, et al. DNA 34 Methylation and Protein Markers of Chronic Inflammation and Their Associations With Brain and 35 Cognitive Aging. Neurology. 2021;97:e2340–52. 36 30. Green C, Shen X, Stevenson AJ, Conole ELS, Harris MA, Barbu MC, et al. Structural brain 37 correlates of serum and epigenetic markers of inflammation in major depressive disorder. Brain 38 Behav Immun. 2021;92:39–48. 39 31. Barker ED, Cecil CAM, Walton E, Houtepen LC, O’Connor TG, Danese A, et al. Inflammation-40 related epigenetic risk and child and adolescent mental health: A prospective study from pregnancy to 41 middle adolescence. Dev Psychopathol. 2018;30:1145–56. 42 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint 22 32. Conole ELS, Vaher K, Blesa-Cábez M, Sullivan G, Stevenson AJ, Hall J, et al. Immuno-epigenetic 1 signature derived in saliva associates with the encephalopathy of prematurity and perinatal 2 inflammatory disorders. Brain Behav Immun. 2023;110:322–38. 3 33. Needham BL, Smith JA, Zhao W, Wang X, Mukherjee B, Kardia SLR, et al. Life course 4 socioeconomic status and DNA methylation in genes related to stress reactivity and inflammation: 5 The multi-ethnic study of atherosclerosis. Epigenetics. 2015;10:958–69. 6 34. Stringhini S, Polidoro S, Sacerdote C, Kelly RS, Veldhoven K van, Agnoli C, et al. Life-course 7 socioeconomic status and DNA methylation of genes regulating inflammation. Int J Epidemiol. 8 2015;44:1320–30. 9 35. McDade TW, Ryan C, Jones MJ, MacIsaac JL, Morin AM, Meyer JM, et al. Social and physical 10 environments early in development predict DNA methylation of inflammatory genes in young 11 adulthood. Proc National Acad Sci. 2017;114:7611–6. 12 36. Smith JA, Zhao W, Wang X, Ratliff SM, Mukherjee B, Kardia SLR, et al. Neighborhood 13 characteristics influence DNA methylation of genes involved in stress response and inflammation: The 14 Multi-Ethnic Study of Atherosclerosis. Epigenetics. 2017;12:662–73. 15 37. Gadd DA, Smith HM, Mullin D, Chybowska O, Hillary RF, Kimenai DM, et al. DNAm scores for 16 serum GDF15 and NT-proBNP levels associate with a range of traits affecting the body and brain. 17 medRxiv. 2023;2023.10.18.23297200. 18 38. Boardman JP, Hall J, Thrippleton MJ, Reynolds RM, Bogaert D, Davidson DJ, et al. Impact of 19 preterm birth on brain development and long-term outcome: protocol for a cohort study in Scotland. 20 Bmj Open. 2020;10:e035854. 21 39. Scottish National Statistics. SIMD - Scottish Index of Multiple Deprivation: SIMD16 technical 22 notes. Edinburgh: Scottish National Statistics; 2016 p. 1–69. 23 40. National Records of Scotland. Scotland’s Census [Internet]. 2020 [cited 2023 Nov 23]. Available 24 from: https://www.scotlandscensus.gov.uk/search-the-census/#/ 25 41. Lowe R, Gemma C, Beyan H, Hawa MI, Bazeos A, Leslie RD, et al. Buccals are likely to be a 26 more informative surrogate tissue than blood for epigenome-wide association studies. Epigenetics. 27 2013;8:445–54. 28 42. Braun PR, Han S, Hing B, Nagahama Y, Gaul LN, Heinzman JT, et al. Genome-wide DNA 29 methylation comparison between live human brain and peripheral tissues within individuals. Transl 30 Psychiat. 2019;9:47. 31 43. Pidsley R, Wong CCY, Volta M, Lunnon K, Mill J, Schalkwyk LC. A data-driven approach to 32 preprocessing Illumina 450K methylation array data. Bmc Genomics. 2013;14:293. 33 44. Zheng SC, Webster AP, Dong D, Feber A, Graham DG, Sullivan R, et al. A novel cell-type 34 deconvolution algorithm reveals substantial contamination by immune cells in saliva, buccal and 35 cervix. Epigenomics-uk. 2018;10:925–40. 36 45. Katz DH, Robbins JM, Deng S, Tahir UA, Bick AG, Pampana A, et al. Proteomic profiling 37 platforms head to head: Leveraging genetics and clinical traits to compare aptamer- and antibody-38 based methods. Sci Adv. 2022;8:eabm5164. 39 46. Haslam DE, Li J, Dillon ST, Gu X, Cao Y, Zeleznik OA, et al. Stability and reproducibility of 40 proteomic profiles in epidemiological studies: comparing the Olink and SOMAscan platforms. 41 PROTEOMICS. 2022;22:e2100170. 42 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint 23 47. Raffield LM, Dang H, Pratte KA, Jacobson S, Gillenwater LA, Ampleford E, et al. Comparison of 1 Proteomic Assessment Methods in Multiple Cohort Studies. PROTEOMICS. 2020;20:1900278. 2 48. Joshi A, Mayr M. In Aptamers They Trust: The caveats of the SOMAscan biomarker discovery 3 platform from SomaLogic. Circulation. 2018;138:2482–5. 4 49. Hillary RF, Marioni RE. MethylDetectR: a software for methylation-based health profiling. 5 Wellcome Open Res. 2021;5:283. 6 50. Preacher KJ. Extreme Groups Designs. The Encyclopedia of Clinical Psychology. 2015;1–4. 7 51. Fisher JE, Guha A, Heller W, Miller GA. Extreme-Groups Designs in Studies of Dimensional 8 Phenomena: Advantages, Caveats, and Recommendations. J Abnorm Psychol. 2020;129:14–20. 9 52. Mckinnon K, Conole ELS, Vaher K, Binkowska J, Sullivan G, Hillary R, et al. The relationship 10 between socioeconomic status, preterm birth and systemic inflammation using DNA methylation 11 proxies. OSF. 2023; 12 53. Gao X, Becker LC, Becker DM, Starmer JD, Province MA. Avoiding the high Bonferroni penalty in 13 genome‐ wide association studies. Genet Epidemiol. 2010;34:100–5. 14 54. Szklarczyk D, Kirsch R, Koutrouli M, Nastou K, Mehryary F, Hachilif R, et al. The STRING 15 database in 2023: protein–protein association networks and functional enrichment analyses for any 16 sequenced genome of interest. Nucleic Acids Res. 2022;51:D638–46. 17 55. Kandasamy J, Roane C, Szalai A, Ambalavanan N. Serum eotaxin-1 is increased in extremely-18 low-birth-weight infants with bronchopulmonary dysplasia or death. Pediatr Res. 2015;78:498–504. 19 56. Almudares F, Hagan J, Chen X, Devaraj S, Moorthy B, Lingappan K. Growth and differentiation 20 factor 15 (GDF15) levels predict adverse respiratory outcomes in premature neonates. Pediatr 21 Pulmonol. 2023;58:271–8. 22 57. Al-Mudares F, Reddick S, Ren J, Venkatesh A, Zhao C, Lingappan K. Role of Growth 23 Differentiation Factor 15 in Lung Disease and Senescence: Potential Role Across the Lifespan. 24 Frontiers Medicine. 2020;7:594137. 25 58. Wan Y, Fu J. GDF15 as a key disease target and biomarker: linking chronic lung diseases and 26 ageing. Mol Cell Biochem. 2023;1–14. 27 59. Zhang Y, Jiang W, Wang L, Lingappan K. Sex-specific differences in the modulation of Growth 28 Differentiation Factor 15 (GDF15) by hyperoxia in vivo and in vitro: Role of Hif-1α . Toxicol Appl 29 Pharm. 2017;332:8–14. 30 60. Wang G, Wen B, Deng Z, Zhang Y, Kolesnichenko OA, Ustiyan V, et al. Endothelial progenitor 31 cells stimulate neonatal lung angiogenesis through FOXF1-mediated activation of BMP9/ACVRL1 32 signaling. Nat Commun. 2022;13:2080. 33 61. Collaco JM, McGrath-Morrow SA, Griffiths M, Chavez-Valdez R, Parkinson C, Zhu J, et al. 34 Perinatal Inflammatory Biomarkers and Respiratory Disease in Preterm Infants. J Pediatrics. 35 2022;246:34-39.e3. 36 62. Oak P, Hilgendorff A. The BPD trio? Interaction of dysregulated PDGF, VEGF, and TGF signaling 37 in neonatal chronic lung disease. Mol Cell Pediatrics. 2017;4:11. 38 63. Course CW, Lewis PA, Kotecha SJ, Cousins M, Hart K, Watkins WJ, et al. Characterizing the 39 urinary proteome of prematurity-associated lung disease in school-aged children. Respir Res. 40 2023;24:191. 41 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint 24 64. Bose CL, Dammann CEL, Laughon MM. Bronchopulmonary dysplasia and inflammatory 1 biomarkers in the premature neonate. Archives Dis Child - Fetal Neonatal Ed. 2008;93:F455. 2 65. Zasada M, Suski M, Bokiniec R, Szwarc-Duma M, Borszewska-Kornacka MK, Madej J, et al. 3 Comparative two time-point proteome analysis of the plasma from preterm infants with and without 4 bronchopulmonary dysplasia. Ital J Pediatr. 2019;45:112. 5 66. Lassus P, Heikkilä P, Andersson LC, Boguslawski K von, Andersson S. Lower concentration of 6 pulmonary hepatocyte growth factor is associated with more severe lung disease in preterm infants. J 7 Pediatr. 2003;143:199–202. 8 67. Wagner BD, Babinec AE, Carpenter C, Gonzalez S, O’Brien G, Rollock K, et al. Proteomic 9 Profiles Associated with Early Echocardiogram Evidence of Pulmonary Vascular Disease in Preterm 10 Infants. Am J Respir Crit Care Med. 2018;197:394–7. 11 68. Guasti L, Silvennoinen S, Bulstrode NW, Ferretti P, Sankilampi U, Dunkel L. Elevated FGF21 12 Leads to Attenuated Postnatal Linear Growth in Preterm Infants Through GH Resistance in 13 Chondrocytes. J Clin Endocrinol Metab. 2014;99:E2198–206. 14 69. Spencer R, Maksym K, Hecher K, Maršál K, Figueras F, Ambler G, et al. Ultrasound and 15 biochemical predictors of pregnancy outcome at diagnosis of early-onset fetal growth restriction. 16 Medrxiv. 2023;2023.01.27.23285087. 17 70. Qiu Q, Bell M, Lu X, Yan X, Rodger M, Walker M, et al. Significance of IGFBP-4 in the 18 Development of Fetal Growth Restriction. J Clin Endocrinol Metab. 2012;97:E1429–39. 19 71. Voller SB, Chock S, Ernst LM, Su E, Liu X, Farrow KN, et al. Cord blood biomarkers of vascular 20 endothelial growth (VEGF and sFlt-1) and postnatal growth: A preterm birth cohort study. Early Hum 21 Dev. 2014;90:195–200. 22 72. Kim MJ, Romero R, Kim CJ, Tarca AL, Chhauy S, LaJeunesse C, et al. Villitis of Unknown 23 Etiology Is Associated with a Distinct Pattern of Chemokine Up-Regulation in the Feto-Maternal and 24 Placental Compartments: Implications for Conjoint Maternal Allograft Rejection and Maternal Anti-25 Fetal Graft-versus-Host Disease. J Immunol. 2009;182:3919–27. 26 73. Romero R, Chaemsaithong P, Chaiyasit N, Docheva N, Dong Z, Kim CJ, et al. CXCL10 and IL‐ 6: 27 Markers of two different forms of intra‐ amniotic inflammation in preterm labor. Am J Reprod Immunol. 28 2017;78:e12685. 29 74. Sellmer A, Henriksen TB, Palmfeldt J, Bech BH, Astono J, Bennike TB, et al. The Patent Ductus 30 Arteriosus in Extremely Preterm Neonates Is More than a Hemodynamic Challenge: New Molecular 31 Insights. Biomol. 2022;12:1179. 32 75. Olsson KW, Larsson A, Jonzon A, Sindelar R. Exploration of potential biochemical markers for 33 persistence of patent ductus arteriosus in preterm infants at 22–27 weeks’ gestation. Pediatr Res. 34 2019;86:333–8. 35 76. Waleh N, Seidner S, McCurnin D, Giavedoni L, Hodara V, Goelz S, et al. Anatomic Closure of the 36 Premature Patent Ductus Arteriosus: The Role of CD14+/CD163+ Mononuclear Cells and VEGF in 37 Neointimal Mound Formation. Pediatr Res. 2011;70:332–8. 38 77. Xu C, Su X, Chen Y, Xu Y, Wang Z, Mo X. Proteomics analysis of plasma protein changes in 39 patent ductus arteriosus patients. Ital J Pediatr. 2020;46:64. 40 78. Cheng Y, Zhu X, Linghu D, Xu Y, Liang J. Serum levels of cytokines in infants treated with 41 conbercept for retinopathy of prematurity. Sci Rep-uk. 2020;10:12695. 42 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint 25 79. Sato T, Kusaka S, Shimojo H, Fujikado T. Simultaneous Analyses of Vitreous Levels of 27 1 Cytokines in Eyes with Retinopathy of Prematurity. Ophthalmology. 2009;116:2165–9. 2 80. Rathi S, Jalali S, Patnaik S, Shahulhameed S, Musada GR, Balakrishnan D, et al. Abnormal 3 Complement Activation and Inflammation in the Pathogenesis of Retinopathy of Prematurity. Front 4 Immunol. 2017;8:1868. 5 81. The Royal College of Ophthalmologists. Treating Retinopathy of Prematurity in the UK. The Royal 6 College of Ophthalmologists; 2022. 7 82. Rivera JC, Holm M, Austeng D, Morken TS, Zhou T (Ellen), Beaudry-Richard A, et al. Retinopathy 8 of prematurity: inflammation, choroidal degeneration, and novel promising therapeutic strategies. J 9 Neuroinflamm. 2017;14:165. 10 83. Klerk DH, Plösch T, Verkaik-Schakel RN, Hulscher JBF, Kooi EMW, Bos AF. DNA Methylation of 11 TLR4, VEGFA, and DEFA5 Is Associated With Necrotizing Enterocolitis in Preterm Infants. Frontiers 12 Pediatrics. 2021;9:630817. 13 84. Olaloye OO, Liu P, Toothaker JM, McCourt BT, McCourt CC, Xiao J, et al. CD16+CD163+ 14 monocytes traffic to sites of inflammation during necrotizing enterocolitis in premature infants. J Exp 15 Med. 2021;218:e20200344. 16 85. Pammi M, Suresh G. Enteral lactoferrin supplementation for prevention of sepsis and necrotizing 17 enterocolitis in preterm infants. Cochrane Db Syst Rev. 2020;3:CD007137. 18 86. Satrom KM, Ennis K, Sweis BM, Matveeva TM, Chen J, Hanson L, et al. Neonatal hyperglycemia 19 induces CXCL10/CXCR3 signaling and microglial activation and impairs long-term synaptogenesis in 20 the hippocampus and alters behavior in rats. J Neuroinflamm. 2018;15:82. 21 87. Groselj-Grenc M, Ihan A, Derganc M. Neutrophil and Monocyte CD64 and CD163 Expression in 22 Critically Ill Neonates and Children with Sepsis: Comparison of Fluorescence Intensities and 23 Calculated Indexes. Mediat Inflamm. 2008;2008:202646. 24 88. Kingsmore SF, Kennedy N, Halliday HL, Velkinburgh JCV, Zhong S, Gabriel V, et al. Identification 25 of Diagnostic Biomarkers for Infection in Premature Neonates. Mol Cell Proteomics. 2008;7:1863–75. 26 89. Zilow EP, Hauck W, Linderkamp O, Zilow G. Alternative Pathway Activation of the Complement 27 System in Preterm Infants with Early Onset Infection. Pediatr Res. 1997;41:334–9. 28 90. Mohammed A, Okwor I, Shan L, Onyilagha C, Uzonna JE, Gounni AS. Semaphorin 3E Regulates 29 the Response of Macrophages to Lipopolysaccharide-Induced Systemic Inflammation. J Immunol. 30 2020;204:128–36. 31 91. Leviton A, Allred EN, Fichorova RN, O’Shea TM, Fordham LA, Kuban KKC, et al. Circulating 32 biomarkers in extremely preterm infants associated with ultrasound indicators of brain damage. Eur J 33 Paediatr Neuro. 2018;22:440–50. 34 92. Schultz SJ, Aly H, Hasanen BM, Khashaba MT, Lear SC, Bendon RW, et al. Complement 35 component 9 activation, consumption, and neuronal deposition in the post-hypoxic–ischemic central 36 nervous system of human newborn infants. Neurosci Lett. 2005;378:1–6. 37 93. Morales DM, Townsend RR, Malone JP, Ewersmann CA, Macy EM, Inder TE, et al. Alterations in 38 Protein Regulators of Neurodevelopment in the Cerebrospinal Fluid of Infants with Posthemorrhagic 39 Hydrocephalus of Prematurity. Mol Cell Proteom. 2012;11:M111.011973. 40 94. Ochoa TJ, Sizonenko SV. Lactoferrin and prematurity: a promising milk protein? Biochem Cell 41 Biol. 2017;95:22–30. 42 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint 26 95. Che X, Hornig M, Bresnahan M, Stoltenberg C, Magnus P, Surén P, et al. Maternal mid-1 gestational and child cord blood immune signatures are strongly associated with offspring risk of ASD. 2 Mol Psychiatr. 2022;27:1527–41. 3 96. Leviton A, Joseph RM, Fichorova RN, Allred EN, Taylor HG, O’Shea TM, et al. Executive 4 Dysfunction Early Postnatal Biomarkers among Children Born Extremely Preterm. J Neuroimmune 5 Pharm. 2019;14:188–99. 6 97. Allred EN, Dammann O, Fichorova RN, Hooper SR, Hunter SJ, Joseph RM, et al. Systemic 7 Inflammation during the First Postnatal Month and the Risk of Attention Deficit Hyperactivity Disorder 8 Characteristics among 10 year-old Children Born Extremely Preterm. J Neuroimmune Pharm. 9 2017;12:531–43. 10 98. Aly H, Khashaba M, Nada A, Hasanen B, McCarter R, Schultz S, et al. The Role of Complement 11 in Neurodevelopmental Impairment following Neonatal Hypoxic-Ischemic Encephalopathy. Am J 12 Perinatol. 2009;26:659–65. 13 99. Limbrick DD, Morales DM, Shannon CN, Wellons JC, Kulkarni AV, Alvey JS, et al. Cerebrospinal 14 fluid NCAM-1 concentration is associated with neurodevelopmental outcome in post-hemorrhagic 15 hydrocephalus of prematurity. PLoS ONE. 2021;16:e0247749. 16 100. Voegele AF, Jerković L, Wellenzohn B, Eller P, Kronenberg F, Liedl KR, et al. Characterization 17 of the Vitamin E-Binding Properties of Human Plasma Afamin. Biochemistry. 2002;41:14532–8. 18 101. Paetau S, Rolova T, Ning L, Gahmberg CG. Neuronal ICAM-5 Inhibits Microglia Adhesion and 19 Phagocytosis and Promotes an Anti-inflammatory Response in LPS Stimulated Microglia. Front Mol 20 Neurosci. 2017;10:431. 21 102. Skeel A, Yoshimura T, Showalter SD, Tanaka S, Appella E, Leonard EJ. Macrophage stimulating 22 protein: purification, partial amino acid sequence, and cellular activity. J Exp Med. 1991;173:1227–34. 23 103. Kronenberg F, Kollerits B, Kiechl S, Lamina C, Kedenko L, Meisinger C, et al. Plasma 24 Concentrations of Afamin Are Associated With the Prevalence and Development of Metabolic 25 Syndrome. Circ: Cardiovasc Genet. 2018;7:822–9. 26 104. Montez JK, Bromberger JT, Harlow SD, Kravitz HM, Matthews KA. Life-Course Socioeconomic 27 Status and Metabolic Syndrome Among Midlife Women. J Gerontol: Ser B. 2016;71:1097–107. 28 105. Castagné R, Delpierre C, Kelly-Irving M, Campanella G, Guida F, Krogh V, et al. A life course 29 approach to explore the biological embedding of socioeconomic position and social mobility through 30 circulating inflammatory markers. Sci Rep-uk. 2016;6:25170. 31 106. Leviton A, Allred EN, Dammann O, Joseph RM, Fichorova RN, O’Shea TM, et al. Socioeconomic 32 status and early blood concentrations of inflammation-related and neurotrophic proteins among 33 extremely preterm newborns. Plos One. 2019;14:e0214154. 34 107. Suarez A, Lahti J, Lahti-Pulkkinen M, Girchenko P, Czamara D, Arloth J, et al. A polyepigenetic 35 glucocorticoid exposure score at birth and childhood mental and behavioral disorders. Neurobiology 36 Stress. 2020;13:100275. 37 108. Blostein FA, Fisher J, Dou J, Schneper L, Ware EB, Notterman DA, et al. Polymethylation scores 38 for prenatal maternal smoke exposure persist until age 15 and are detected in saliva in the Fragile 39 Families and Child Wellbeing cohort. Epigenetics. 2022;17:2223–40. 40 41 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint A B . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 19, 2023. ; https://doi.org/10.1101/2023.12.19.23300227doi: medRxiv preprint