DNA methylation (DNAm) plays an important role in a diverse range of epigenetically regulated processes in mammals including cell differentiation, X-chromosome inactivation, genomic imprinting, and the silencing of transposable elements (TEs) (Smith and Meissner, 2013). DNAm can influence gene expression and can in turn be influenced by molecular processes including differential action of methyltransferases and transcription factor (TF) binding (Jeltsch, 2006; Feldmann et al., 2013).

The human methylome is extensively remodelled in the very early embryo when parental gametic methylation marks are largely erased before acquisition of lineage and tissue-specific marks at implantation, gastrulation, and beyond (Guo et al., 2014). The days following conception may therefore offer a window of heightened sensitivity to external environmental influences, potentially stretching back to the period before conception coinciding with late maturation of oocytes and spermatozoa at loci that (partially) evade early embryonic reprogramming (Fleming et al., 2018).

The effects of early exposures on the mammalian methylome have been widely studied in animal models and periconceptional and early gestational factors including maternal folate and exposure to famine have been associated with DNAm changes in humans (Gonseth et al., 2015; Tobi et al., 2015; Finer et al., 2016). However, causal pathways are difficult to elucidate in human observational studies, and even randomised experimental designs are prone to confounding due to reverse causation from exposure-related effects (Birney et al., 2016).

Here, we address these limitations by exploiting a natural experiment in rural Gambia where conceptions occur against a background of repeating annual patterns of dry (‘harvest’) and rainy (‘hungry’) seasons with accompanying significant changes in energy balance, diet composition, nutrient status, and rates of infection (Moore et al., 1999; Dominguez-Salas et al., 2013). We assess the influence of seasonality on DNAm in two Gambian child cohorts (Moore et al., 2012; Chandak et al., 2017), one with longitudinal data, enabling robust identification of loci showing consistent effects at the ages of 2 years and in mid-childhood (Figure 1). Through prospective study designs we capture conceptions throughout the year and, in contrast to previous analyses in this population (Waterland et al., 2010; Dominguez-Salas et al., 2014; Silver et al., 2015), we use statistical models that make no prior assumptions about specific seasonal windows driving DNAm changes in offspring.

Figure 1

Download asset Open asset

We probe potential connections between season of conception (SoC)-associated loci and early embryonic events by leveraging published data on loci with evidence for the establishment of variable methylation states in the early embryo that persist in post-gastrulation and postnatal tissues; namely loci demonstrating systemic interindividual variation (SIV) (Kessler et al., 2018; Van Baak et al., 2018) and/or epigenetic supersimilarity (ESS) (Van Baak et al., 2018; Figure 1). These loci bear the hallmarks of metastable epialleles (MEs), loci with methylation states that vary between individuals that were first identified in isogenic mice. MEs exhibit stable patterns of SIV indicating stochastic establishment of methylation marks prior to gastrulation when tissue differentiation begins (Rakyan et al., 2002), and several MEs have been shown to be sensitive to periconceptional nutrition in mice (Anderson et al., 2012). These loci thus serve as useful tools for studying the effects of early environment on DNAm by enabling the use of accessible tissues (such as blood) that can serve as a proxy for systemic (cross-tissue) methylation, and by pinpointing the window of exposure to the periconceptional period (Gunasekara and Waterland, 2019). We also investigate links with TEs and TFs associated with the establishment of methylation states in the early embryo, and assess the influence of genetic variation and gene-environment interactions. Finally, by comparing our results with public DNAm data obtained from sperm, oocytes, and multi-stage human embryos, we investigate links between SoC-associated loci, histone marks, gametic, and parent-of-origin-specific methylation (PofOm), and the establishment of DNAm states in early embryonic development.

Our identification of hotspots in the postnatal methylome that retain a record of environmental conditions during gametic maturation and/or in the very early embryo provides a valuable resource for the investigation of epigenetic mechanisms linking early-life nutritional and other exposures to lifelong health and disease.

We have utilised a natural, seasonal experiment in rural Gambia whereby human conceptions are ‘randomised’ to contrasting environmental (especially dietary) conditions to examine whether this differential exposure leaves a discernible signature on the offspring methylome. We analysed methylation data from two independent, different-aged cohorts and identified 259 ‘SoC-CpGs’ with evidence of sensitivity to SoC in infancy. We found evidence of SoC effect attenuation with age, but SoC-related temporal patterns were nonetheless strikingly similar, suggestive of a common effect of periconceptional environment. Importantly, the analysed datasets have contrasting confounding structures, notably with regard to the timing of sample collection, the latter eliminating potential confounding due to seasonal differences in leukocyte composition. These results, derived from analysis of Illumina array data with limited coverage, suggest there may be many more hotspots sensitive to the periconceptional environment across the human methylome.

This analysis builds on previous epigenetic studies in this setting that have focussed on single cohorts and analysed methylation differences between individuals conceived at the peaks of the Gambian dry and rainy seasons only (Waterland et al., 2010; Dominguez-Salas et al., 2014; Silver et al., 2015; Van Baak et al., 2018; Kühnen et al., 2016). Increased methylation in offspring conceived at the peak of the Gambian rainy season is consistent with previous findings and this observation is now greatly strengthened by the application of Fourier regression to model year-round conceptions, an approach that makes no prior assumption of when methylation peaks and nadirs may occur. The number of identified SoC-CpGs is also substantially increased in this study. Comparisons with array background and control CpGs matching SoC-CpG methylation distributions increase confidence that these findings are not statistical artefacts.

Triangulation with other public data provides multiple lines of evidence supporting the notion that methylation states at these loci are established in the periconceptional period. First, they are highly enriched for putative MEs and related loci identified in other studies with characteristic methylation signatures suggestive of establishment early in embryonic development (Kessler et al., 2018; Van Baak et al., 2018). Second, like MEs, season-associated loci exhibit unusual methylation dynamics in early stage embryos (Kessler et al., 2018). Third, they have distinctive gametic methylation patterns, notably hypomethylation in sperm (in common with putative MEs; Kessler et al., 2018), and differential gametic and PofOm in a subset. Increased incidence of hypomethylated states in sperm at SoC-CpGs may reflect their enrichment at CpG islands (Molaro et al., 2011), sequence features that are largely refractory to protamine exchange, with the possibility for retaining epigenetic function associated with histone modifications into the early embryo (Hammoud et al., 2009). Fourth, many overlap H3K4me1 and H3K27me3 marks which coordinate transient gene expression and early lineage commitment in human ESCs, in part through demarking poised enhancers (Rada-Iglesias et al., 2011).

A large majority of SoC-CpGs have not previously been identified as MEs, but given the supporting evidence described above, we speculate that many are likely to be so. Indeed, evidence of attenuation of SoC effects with age suggests that, to the extent that interindividual variation is driven by periconceptional environmental factors, screens for putative MEs (including ESS and SIV) in adult tissues used as a reference in this analysis may be missing metastable regions which are more pronounced earlier in the life course. Evidence of much larger SoC effect sizes at known MEs in both Gambian cohorts supports this. Furthermore, identification of putative human SIV-MEs requires analysis of datasets with tissues derived from different germ layers for each individual. Such datasets are expensive to acquire and rare, meaning that the few published ME screens are likely to be underpowered due to small sample sizes.

The observed attenuation of SoC effects with age has implications for detecting the effect of periconceptional exposures on DNAm in samples collected beyond the neonatal and early childhood periods, an important consideration for epigenetic epidemiological studies since non-persisting methylation differences could still have a significant impact on early developmental trajectories with lifelong consequences (Vukic et al., 2019; Simpkin et al., 2015). The methylome changes with age, a phenomenon sometimes referred to as ‘epigenetic drift’ (Teschendorff et al., 2013). Attenuation of SoC effects could be a consequence of localised or global epigenetic drift or other processes such as clonal selection (Teschendorff and Relton, 2018).

Interestingly, more SoC-CpGs replicated in the smaller ENID (5–7 yr) dataset than in the older EMPHASIS (7–9 yr) cohort (n = 157 vs. 61, respectively). This could be due to age or other cohort-specific differences, or it might reflect differences in seasonality such as secular changes in the intensity and duration of the rainy season.

Intra-individual methylation states at SoC-associated loci are highly correlated across loci despite being distributed throughout the genome, suggesting that a common mechanism is at play. MEs located within intracisternal A particle (IAP) TEs have been the focus of many studies of SIV in mice, the Agouti viable yellow locus being a paradigm example. While one study found that methylation does not covary across different murine IAP MEs in the absence of environmental perturbations (Kazachenka et al., 2018), another recent study identified several IAP MEs located on different chromosomes which are modified by the same cluster of KRAB zinc finger proteins (KZFP) (Bertozzi et al., 2020), demonstrating that genetic mechanisms for the simultaneous epigenetic regulation of multiple loci do exist.

Potential insights into mechanisms linking periconceptional environment to DNAm changes in postnatal tissues come from our investigations of the methylation status and genomic context of SoC-CpGs.

First, we observed strong enrichment for gDMRs, with 14% of SoC-CpGs overlapping 19 gDMRs hypomethylated in sperm and hypermethylated in oocytes. A minority of these show evidence of PofOm persisting in postnatal tissues. This observation aligns with a growing body of evidence linking early environment, notably nutritional factors involved in one-carbon (C1) metabolism, with methylation at imprinted regions (Monk et al., 2019; James et al., 2018b). Indeed we have previously noted an association between SoC and several C1 metabolites at a maternally imprinted region at the small non-coding RNA VTRNA2-1 (Silver et al., 2015), consistent with evidence of ‘polymorphic imprinting’ linked to prenatal environment at this locus (Zink et al., 2018; Carpenter et al., 2018). Furthermore, we previously found strong enrichment for proximal binding sites of several TFs associated with the maintenance of PofOm in the early embryo at MEs detected in a WGBS screen (Kessler et al., 2018). We were only able to replicate enrichment for one of these, ZFP57, at SoC-CpGs in this study. This may reflect the relatively small proportion of PofOm loci in the set of SoC-CpGs, or factors related to the limited methylome coverage of Illumina arrays. More targeted experimental work is required to determine the extent of SoC effects at imprinted loci, especially given our observation that SoC-CpGs are often proximal to ERV TEs that have recently been shown to drive the establishment of germline-derived maternal PofOm (Bogutz et al., 2019). Hotspots with evidence of PofOm could be driven by an environmentally sensitive gain of methylation on the paternal allele that is propagated through development, incomplete reprogramming on the maternal allele leaving residual traces of methylated cytosines, or modest de novo methylation at some later point. A deeper understanding of mechanisms will require further investigation in cell and animal models.

Second, our observation of enrichment for proximity to ERV1 TEs at SoC-CpGs aligns with our previous finding at MEs (Silver et al., 2015; Kessler et al., 2018), and is notable since most environment-sensitive mouse MEs are associated with IAPs (which are rodent-specific ERVs) (Kazachenka et al., 2018). KZFP-mediated repression of TEs including ERVs has also been proposed as a driver of the rapid evolution of gene regulation (Cavalli and Heard, 2019). The KZFP ZFP57 is particularly interesting in this respect since its binding to DNA is linked both to repression of TEs and to the maintenance of genomic imprints in the pre-implantation embryo (Imbeault et al., 2017; Shi et al., 2019). We previously identified a putative SoC-associated DMR in the ZFP57 promoter in blood from younger Gambian infants (mean age 3.6 months) (Silver et al., 2015). It is possible that non-replication of the SoC association at ZFP57 in the older Gambian cohort (and of the VTRNA2-1 SoC-DMR mentioned above which was also identified in younger infants) reflects the more general attenuation of SoC effects described above. Interestingly there is some evidence that the ZFP57 DMR, which lies 3 kb upstream of the transcription start site, is established in the early embryo (Van Baak et al., 2018; Monteagudo-Sánchez et al., 2020), and that DNAm at this locus measured in neonatal blood is associated with maternal folate levels (Gonseth et al., 2015; Amarasekera et al., 2014; Irwin et al., 2019). Given the important function of ZFP57 in pre-implantation methylation dynamics, its potential role as an environmentally sensitive regulator of multi-locus DNAm effects remains an open question.

Third, DNAm at SoC-CpGs is enriched for intermediate methylation states. Intermediate methylation has also been observed at MEs in Gambians and in non-Africans (Finer et al., 2016; Waterland et al., 2010; Dominguez-Salas et al., 2014; Kühnen et al., 2016), and this coincides with a similar observation at MEs in post-gastrulation embryonic tissues (Kessler et al., 2018). Intermediate methylation appears to be predominantly driven by variegated (intercellular) differences in methylation status within a sampled tissue, rather than by allele-specific methylation (Elliott et al., 2015). Observed differences in aggregate methylation at SoC-CpGs could thus reflect a direct influence of periconceptional environment on the establishment and maintenance of DNAm states in the early embryo at the cellular level, or ‘epigenetic selection’ whereby epigenetically distinct cells form a substrate for clonal selection during development, for example, as a potential adaptation to differential metabolic exposures (Tobi et al., 2018). Stronger enrichment for putative MEs exhibiting SIV compared to those identified through ESS supports establishment in the post-implantation embryo, since methylation differences at ESS loci are presumed to originate in the pre-implantation embryo prior to embryo cleavage in MZ twins (Van Baak et al., 2018).

The largest SoC-CpG cluster with 7 CpGs is on chromosome 15 and falls within the second intron of IGF1R. This bears the hallmarks of being a promoter or active/poised enhancer in multiple cell lines (Appendix 1—figure 14). Six CpGs in the IGF1R cluster have evidence of PofOm and show a relatively large SoC effect size (median SoC methylation amplitude) of 6.1%. Zink et al., 2018, were unable to demonstrate PofO allele-specific expression in this region although others have found evidence of maternal imprinting of an intronic lncRNA at this gene in cancerous cells (Kang et al., 2015; Sun et al., 2014). Loss of IGF1 receptors gives rise to a major decrease in expression at multiple imprinted genes in mice, suggesting a pathway by which IGF1R might regulate growth and metabolism during early development (Boucher et al., 2014). IGF1R signalling is implicated in fetal growth, glucose metabolism and cancer (Randhawa and Cohen, 2005; Aguirre et al., 2016; Larsson et al., 2005), and DNAm differences at IGF1R have been observed in birthweight-discordant adult twins (Tsai et al., 2015). Another SoC-CpG with evidence of PofOm, also on chromosome 15, falls within the known Prader-Willi syndrome-associated paternally expressed imprinted region.

Our observation that 109 out of 259 SoC-CpGs have been associated with sex in previous studies is intriguing. Given the relatively small number (compared to array size) of autosomal sex-linked loci identified in large studies on the Illumina 450k array (Singmann et al., 2015; Shah et al., 2014), this represents a very strong enrichment. We replicated a significant sex association at these loci in the ENID cohort analysed in this study. Our regression analyses were adjusted for sex, and additional sensitivity analyses with DNAm pre-adjusted for sex strongly suggest that our main findings are not confounded by sex. Interestingly, sex-associated loci are enriched at imprinted regions and sex discordance at autosomal CpGs has been linked to androgen exposures in utero (Singmann et al., 2015; Suderman et al., 2017). There is also evidence of sex differences in methylation at DNMT3A/B and TET1 genes involved in de novo methylation and de-methylation pathways (Singmann et al., 2015; Suderman et al., 2017), suggesting a possible interaction between sex-linked epigenetic changes and periconceptional environment during reprogramming in the early embryo. A deeper understanding of potential mechanisms underpinning the observed enrichment of sex effects at loci associated with periconceptional environment requires further functional analysis in cell and/or animal models.

DNAm is influenced by genotype and the latter is therefore a potential confounder when studying the effects of environmental exposures in human populations. A strength of our quasi-randomised Gambian seasonal model is that it minimises the potential for genetic confounding of modelled seasonal DNAm patterns, on the assumption that the timing of conceptions is not linked to genetic variants influencing DNAm. However, it is still possible that such variants might confound our observations, for example, if they promote embryo survival under conditions of environmental stress. We tested this possibility using genetic data available for the EMPHASIS cohort and found no evidence of SoC-associated genetic variants driving interindividual methylation differences at SoC-associated loci in cis or trans.

However, we did find that half of SoC-CpGs had at least one associated mQTL, indicating the presence of independent additive effects of environment and genetics at these loci, as has been suggested previously at other loci sensitive to pre/periconceptional nutritional exposures (Tobi et al., 2012; Saffari et al., 2020). We have previously argued that the definition of MEs should be extended to include genomic regions whose DNAm state is under partial but non-deterministic genetic influence in genetically heterogeneous human populations (Kessler et al., 2018), and we would argue that the above observations at SoC-CpGs that exhibit many of the characteristics of MEs support this. Further analysis in larger datasets with genome sequencing data combined with functional analysis using cell models will be required to fully understand the relative contributions of environment and genetics to DNAm variation at regions of the type highlighted in this study.

Further work is required to investigate the functional relevance of DNAm changes at SoC-CpGs, some of which are relatively small (around 4% SoC amplitude). However, we can speculate that observed methylation changes, which may reflect changes in the chromatin landscape, have the potential to influence gene expression and early development, since our chromatin state analysis predicts that many overlap regions with functional significance in H1 ESCs and fetal tissues. A similarly modest DNAm change at a locus in the POMC gene that is associated with SoC in blood has been linked to obesity risk in German children and adults (Kühnen et al., 2016; Kuehnen et al., 2012), and to differential TF binding and differences in POMC expression (Kuehnen et al., 2012). Furthermore, blood DNAm at a putative ME within the PAX8 gene has been linked to SoC in Gambian infants (Dominguez-Salas et al., 2014) and is associated with thyroid volume and function in Gambian children, and with certain maternal nutrition biomarkers at conception (Candler et al., 2021). PAX8 DNAm is also associated with expression of the anti-sense gene PAX8-AS1 (alternatively known as LOC654433) in thyroid tissue (Candler et al., 2021). No loci from either study overlap array background in this study. More generally, DNAm at SIV loci associated with periconceptional environment has been associated with a number of diseases including Alzheimer’s, cancer, rheumatoid arthritis, and schizophrenia (Gunasekara and Waterland, 2019).

We have previously shown that several metabolites involved in C1 methylation pathways show significant seasonal variation in maternal blood plasma in this population which is largely dependent on subsistence farming (Dominguez-Salas et al., 2013; Dominguez-Salas et al., 2014). While we suspect that seasonal differences in nutrition are likely drivers of the effects that we have observed, analysis of the links between maternal nutritional biomarkers and DNAm is challenging due to the complex interdependence of C1 metabolites acting as substrates and cofactors within C1 pathways (James et al., 2018a). Furthermore, it is possible that DNAm differences are linked to other aspects of Gambian seasonality, such as variation in pesticide use or infectious disease burden (Moore et al., 1999; Hernandez-Vargas et al., 2015).

Measured DNAm values at SoC-CpGs and matched controls have previously been reported to be reliable, in strong contrast to the majority of probes overlapping the 450k and EPIC arrays that have been found to have low test-retest reliability (Sugden et al., 2020). This likely reflects increased interindividual variability and/or intermediate DNAm at SoC-CpGs and controls (Logue et al., 2017; Xu and Taylor, 2021).

Limitations of this analysis include a lack of genetic data for the ENID cohort, so that our genetic analyses are restricted to the older EMPHASIS cohort where SoC effects are smaller. Furthermore, comparator data on MEs and gamete, embryo and PofO methylation are derived from non-African individuals so we cannot assess the influence of ethnicity on these analyses. We also note that there is some inter-relatedness in this study population which practices polygamy. However, of the 199 individuals with available data on paternal identity in the ENID cohort, 2% have a shared father suggesting that this is unlikely to influence the main findings.

There is increasing interest in the phenomenon of methylation variability as a marker of disease and of prenatal adversity, and in genetic variation as a potential driver of methylation variance (Ek et al., 2018). This raises the possibility that certain genetic variants could have been selected through their ability to enable graded, environmentally responsive methylation patterns at MEs and SoC-associated loci that are able to sense the periconceptional environment, record the information, and adapt the phenotype accordingly. Our gene-environment interaction analysis likely lacked power to detect gene-environment (SoC) interactions, so that it was not possible to investigate the relative contributions of stochastic, environmental, and genetically mediated variance effects on the establishment of periconceptional SoC-associated methylation states in this study. Nonetheless, the proposition that environmentally sensitive epigenetic signals are selected through their ability to direct phenotypic development to better fit the anticipated future environment, leading to maladaptation and future disease if the environment changes, is intriguing and worthy of further investigation (Fleming et al., 2018).

Illumina 450k methylation array data generated from Gambian 2 year olds from the ENID trial is deposited in GEO (GSE99863). Requests to access and analyse the other Gambian methylation datasets (ENID 5-7yr and EMPHASIS 7-9yr) should be submitted to the corresponding author in the first instance. An application would then need to be made to MRC Unit The Gambia's Scientific Coordinating Committee and the Joint MRC/Gambia Government Ethics Committee. Sources and locations of other publicly available data used in this analysis are described in Methods. Bespoke code used in the analysis is available at https://zenodo.org/record/5801480.

1. 1. Van Baak T

   (2018) NCBI Gene Expression Omnibus

   ID GSE99863. DNA methylation in children from The Gambia.

   https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE99863

1. 1. Aguirre GA
   2. De Ita JR
   3. de la Garza RG
   4. Castilla-Cortazar I

   (2016) Insulin-like growth factor-1 deficiency and metabolic syndrome

   Journal of Translational Medicine 14:1–23.

   https://doi.org/10.1186/s12967-015-0762-z

   • PubMed
   • Google Scholar
2. 1. Amarasekera M
   2. Martino D
   3. Ashley S
   4. Harb H
   5. Kesper D
   6. Strickland D
   7. Saffery R
   8. Prescott SL

   (2014) Genome-wide DNA methylation profiling identifies a folate-sensitive region of differential methylation upstream of ZFP57-imprinting regulator in humans

   FASEB Journal 28:4068–4076.

   https://doi.org/10.1096/fj.13-249029

   • PubMed
   • Google Scholar
3. 1. Anderson OS
   2. Sant KE
   3. Dolinoy DC

   (2012) Nutrition and epigenetics: an interplay of dietary methyl donors, one-carbon metabolism and DNA methylation

   The Journal of Nutritional Biochemistry 23:853–859.

   https://doi.org/10.1016/j.jnutbio.2012.03.003

   • PubMed
   • Google Scholar
4. 1. Ba Y
   2. Yu H
   3. Liu F
   4. Geng X
   5. Zhu C
   6. Zhu Q
   7. Zheng T
   8. Ma S
   9. Wang G
   10. Li Z
   11. Zhang Y

   (2011) Relationship of folate, vitamin B12 and methylation of insulin-like growth factor-II in maternal and cord blood

   European Journal of Clinical Nutrition 65:480–485.

   https://doi.org/10.1038/ejcn.2010.294

   • PubMed
   • Google Scholar
5. 1. Bertozzi TM
   2. Elmer JL
   3. Macfarlan TS
   4. Ferguson-Smith AC

   (2020) KRAB zinc finger protein diversification drives mammalian interindividual methylation variability

   PNAS 117:31290–31300.

   https://doi.org/10.1073/pnas.2017053117

   • PubMed
   • Google Scholar
6. 1. Birney E
   2. Smith GD
   3. Greally JM

   (2016) Epigenome-wide Association Studies and the Interpretation of Disease -Omics

   PLOS Genetics 12:e1006105.

   https://doi.org/10.1371/journal.pgen.1006105

   • PubMed
   • Google Scholar
7. 1. Bogutz AB
   2. Brind’Amour J
   3. Kobayashi H
   4. Jensen KN
   5. Nakabayashi K
   6. Imai H
   7. Lorincz MC
   8. Lefebvre L

   (2019) Evolution of imprinting via lineage-specific insertion of retroviral promoters

   Nature Communications 10:1–14.

   https://doi.org/10.1038/s41467-019-13662-9

   • PubMed
   • Google Scholar
8. 1. Boucher J
   2. Charalambous M
   3. Zarse K
   4. Mori MA
   5. Kleinridders A
   6. Ristow M
   7. Ferguson-Smith AC
   8. Kahn CR

   (2014) Insulin and insulin-like growth factor 1 receptors are required for normal expression of imprinted genes

   PNAS 111:14512–14517.

   https://doi.org/10.1073/pnas.1415475111

   • PubMed
   • Google Scholar
9. 1. Buniello A
   2. MacArthur JAL
   3. Cerezo M
   4. Harris LW
   5. Hayhurst J
   6. Malangone C
   7. McMahon A
   8. Morales J
   9. Mountjoy E
   10. Sollis E
   11. Suveges D
   12. Vrousgou O
   13. Whetzel PL
   14. Amode R
   15. Guillen JA
   16. Riat HS
   17. Trevanion SJ
   18. Hall P
   19. Junkins H
   20. Flicek P
   21. Burdett T
   22. Hindorff LA
   23. Cunningham F
   24. Parkinson H

   (2019) The NHGRI-EBI GWAS Catalog of published genome-wide association studies, targeted arrays and summary statistics 2019

   Nucleic Acids Research 47:D1005–D1012.

   https://doi.org/10.1093/nar/gky1120

   • PubMed
   • Google Scholar
10. 1. Candler T
    2. Kessler NJ
    3. Gunasekara CJ
    4. Ward KA
    5. James P
    6. Laritsky E
    7. Baker MS
    8. Dyer R
    9. Elango R
    10. Jeffries D
    11. Waterland RA
    12. Moore SE
    13. Ludgate M
    14. Prentice AM
    15. Silver MJ

    (2021) DNA methylation at a nutritionally sensitive region of the PAX8 gene is associated with thyroid volume and function in Gambian children

    Science Advances 7:eabj1561.

    https://doi.org/10.1126/sciadv.abj1561

    • PubMed
    • Google Scholar
11. 1. Carpenter BL
    2. Zhou W
    3. Madaj Z
    4. DeWitt AK
    5. Ross JP
    6. Grønbæk K
    7. Liang G
    8. Clark SJ
    9. Molloy PL
    10. Jones PA

    (2018) Mother-child transmission of epigenetic information by tunable polymorphic imprinting

    PNAS 115:E11970–E11977.

    https://doi.org/10.1073/pnas.1815005115

    • PubMed
    • Google Scholar
12. 1. Cavalli G
    2. Heard E

    (2019) Advances in epigenetics link genetics to the environment and disease

    Nature 571:489–499.

    https://doi.org/10.1038/s41586-019-1411-0

    • PubMed
    • Google Scholar
13. 1. Chandak GR
    2. Silver MJ
    3. Saffari A
    4. Lillycrop KA
    5. Shrestha S
    6. Sahariah SA
    7. Di Gravio C
    8. Goldberg G
    9. Tomar AS
    10. Betts M
    11. Sajjadi S
    12. Acolatse L
    13. James P
    14. Issarapu P
    15. Kumaran K
    16. Potdar RD
    17. Prentice AM
    18. Fall CH

    (2017) Protocol for the EMPHASIS study; epigenetic mechanisms linking maternal pre-conceptional nutrition and children’s health in India and Sub-Saharan Africa

    BMC Nutrition 3:81.

    https://doi.org/10.1186/s40795-017-0200-0

    • PubMed
    • Google Scholar
14. 1. Chen Y
    2. Lemire M
    3. Choufani S
    4. Butcher DT
    5. Grafodatskaya D
    6. Zanke BW
    7. Gallinger S
    8. Hudson TJ
    9. Weksberg R

    (2013) Discovery of cross-reactive probes and polymorphic CpGs in the Illumina Infinium HumanMethylation450 microarray

    Epigenetics 8:203–209.

    https://doi.org/10.4161/epi.23470

    • PubMed
    • Google Scholar
15. 1. Comuzzie AG
    2. Cole SA
    3. Laston SL
    4. Voruganti VS
    5. Haack K
    6. Gibbs RA
    7. Butte NF

    (2012) Novel genetic loci identified for the pathophysiology of childhood obesity in the Hispanic population

    PLOS ONE 7:e51954.

    https://doi.org/10.1371/journal.pone.0051954

    • PubMed
    • Google Scholar
16. 1. Crider KS
    2. Yang TP
    3. Berry RJ
    4. Bailey LB

    (2012) Folate and DNA methylation: A review of molecular mechanisms and the evidence for folate’s role

    Advances in Nutrition (Bethesda, Md.) 3:21–38.

    https://doi.org/10.3945/an.111.000992

    • PubMed
    • Google Scholar
17. 1. Do C
    2. Shearer A
    3. Suzuki M
    4. Terry MB
    5. Gelernter J
    6. Greally JM
    7. Tycko B

    (2017) Genetic–epigenetic interactions in cis: A major focus in the post-GWAS era

    Genome Biology 18:1–22.

    https://doi.org/10.1186/s13059-017-1250-y

    • PubMed
    • Google Scholar
18. 1. Dominguez-Salas P
    2. Moore SE
    3. Cole D
    4. da Costa K-A
    5. Cox SE
    6. Dyer RA
    7. Fulford AJC
    8. Innis SM
    9. Waterland RA
    10. Zeisel SH
    11. Prentice AM
    12. Hennig BJ

    (2013) DNA methylation potential: dietary intake and blood concentrations of one-carbon metabolites and cofactors in rural African women

    The American Journal of Clinical Nutrition 97:1217–1227.

    https://doi.org/10.3945/ajcn.112.048462

    • PubMed
    • Google Scholar
19. 1. Dominguez-Salas P
    2. Moore SE
    3. Baker MS
    4. Bergen AW
    5. Cox SE
    6. Dyer RA
    7. Fulford AJ
    8. Guan Y
    9. Laritsky E
    10. Silver MJ
    11. Swan GE
    12. Zeisel SH
    13. Innis SM
    14. Waterland RA
    15. Prentice AM
    16. Hennig BJ

    (2014) Maternal nutrition at conception modulates DNA methylation of human metastable epialleles

    Nature Communications 5:1–7.

    https://doi.org/10.1038/ncomms4746

    • PubMed
    • Google Scholar
20. 1. Du P
    2. Zhang X
    3. Huang C-C
    4. Jafari N
    5. Kibbe WA
    6. Hou L
    7. Lin SM

    (2010) Comparison of Beta-value and M-value methods for quantifying methylation levels by microarray analysis

    BMC Bioinformatics 11:587.

    https://doi.org/10.1186/1471-2105-11-587

    • PubMed
    • Google Scholar
21. 1. Ek WE
    2. Rask-Andersen M
    3. Karlsson T
    4. Enroth S
    5. Gyllensten U
    6. Johansson Å

    (2018) Genetic variants influencing phenotypic variance heterogeneity

    Human Molecular Genetics 27:799–810.

    https://doi.org/10.1093/hmg/ddx441

    • PubMed
    • Google Scholar
22. 1. Elliott G
    2. Hong C
    3. Xing X
    4. Zhou X
    5. Li D
    6. Coarfa C
    7. Bell RJA
    8. Maire CL
    9. Ligon KL
    10. Sigaroudinia M
    11. Gascard P
    12. Tlsty TD
    13. Harris RA
    14. Schalkwyk LC
    15. Bilenky M
    16. Mill J
    17. Farnham PJ
    18. Kellis M
    19. Marra MA
    20. Milosavljevic A
    21. Hirst M
    22. Stormo GD
    23. Wang T
    24. Costello JF

    (2015) Intermediate DNA methylation is a conserved signature of genome regulation

    Nature Communications 6:6363.

    https://doi.org/10.1038/ncomms7363

    • PubMed
    • Google Scholar
23. 1. Ernst J
    2. Kellis M

    (2012) ChromHMM: automating chromatin-state discovery and characterization

    Nature Methods 9:215–216.

    https://doi.org/10.1038/nmeth.1906

    • PubMed
    • Google Scholar
24. 1. Evans DS
    2. Avery CL
    3. Nalls MA
    4. Li G
    5. Barnard J
    6. Smith EN
    7. Tanaka T
    8. Butler AM
    9. Buxbaum SG
    10. Alonso A
    11. Arking DE
    12. Berenson GS
    13. Bis JC
    14. Buyske S
    15. Carty CL
    16. Chen W
    17. Chung MK
    18. Cummings SR
    19. Deo R
    20. Eaton CB
    21. Fox ER
    22. Heckbert SR
    23. Heiss G
    24. Hindorff LA
    25. Hsueh W-C
    26. Isaacs A
    27. Jamshidi Y
    28. Kerr KF
    29. Liu F
    30. Liu Y
    31. Lohman KK
    32. Magnani JW
    33. Maher JF
    34. Mehra R
    35. Meng YA
    36. Musani SK
    37. Newton-Cheh C
    38. North KE
    39. Psaty BM
    40. Redline S
    41. Rotter JI
    42. Schnabel RB
    43. Schork NJ
    44. Shohet RV
    45. Singleton AB
    46. Smith JD
    47. Soliman EZ
    48. Srinivasan SR
    49. Taylor HA
    50. Van Wagoner DR
    51. Wilson JG
    52. Young T
    53. Zhang Z-M
    54. Zonderman AB
    55. Evans MK
    56. Ferrucci L
    57. Murray SS
    58. Tranah GJ
    59. Whitsel EA
    60. Reiner AP
    61. Sotoodehnia N
    62. CHARGE QRS Consortium

    (2016) Fine-mapping, novel loci identification, and SNP association transferability in a genome-wide association study of QRS duration in African Americans

    Human Molecular Genetics 25:4350–4368.

    https://doi.org/10.1093/hmg/ddw284

    • PubMed
    • Google Scholar
25. 1. Feldmann A
    2. Ivanek R
    3. Murr R
    4. Gaidatzis D
    5. Burger L
    6. Schübeler D

    (2013) Transcription factor occupancy can mediate active turnover of DNA methylation at regulatory regions

    PLOS Genetics 9:e1003994.

    https://doi.org/10.1371/journal.pgen.1003994

    • PubMed
    • Google Scholar
26. 1. Finer S
    2. Iqbal MS
    3. Lowe R
    4. Ogunkolade BW
    5. Pervin S
    6. Mathews C
    7. Smart M
    8. Alam DS
    9. Hitman GA

    (2016) Is famine exposure during developmental life in rural Bangladesh associated with A metabolic and epigenetic signature in young adulthood? A historical cohort study

    BMJ Open 6:e011768.

    https://doi.org/10.1136/bmjopen-2016-011768

    • PubMed
    • Google Scholar
27. 1. Fleming TP
    2. Watkins AJ
    3. Velazquez MA
    4. Mathers JC
    5. Prentice AM
    6. Stephenson J
    7. Barker M
    8. Saffery R
    9. Yajnik CS
    10. Eckert JJ
    11. Hanson MA
    12. Forrester T
    13. Gluckman PD
    14. Godfrey KM

    (2018) Origins of lifetime health around the time of conception: causes and consequences

    Lancet (London, England) 391:1842–1852.

    https://doi.org/10.1016/S0140-6736(18)30312-X

    • PubMed
    • Google Scholar
28. 1. Fortin J-P
    2. Labbe A
    3. Lemire M
    4. Zanke BW
    5. Hudson TJ
    6. Fertig EJ
    7. Greenwood CM
    8. Hansen KD

    (2014) Functional normalization of 450k methylation array data improves replication in large cancer studies

    Genome Biology 15:0–42.

    https://doi.org/10.1186/s13059-014-0503-2

    • PubMed
    • Google Scholar
29. 1. Gaunt TR
    2. Shihab HA
    3. Hemani G
    4. Min JL
    5. Woodward G
    6. Lyttleton O
    7. Zheng J
    8. Duggirala A
    9. McArdle WL
    10. Ho K
    11. Ring SM
    12. Evans DM
    13. Davey Smith G
    14. Relton CL

    (2016) Systematic identification of genetic influences on methylation across the human life course

    Genome Biology 17:61.

    https://doi.org/10.1186/s13059-016-0926-z

    • PubMed
    • Google Scholar
30. 1. Gonseth S
    2. Roy R
    3. Houseman EA
    4. de Smith AJ
    5. Zhou M
    6. Lee S-T
    7. Nusslé S
    8. Singer AW
    9. Wrensch MR
    10. Metayer C
    11. Wiemels JL

    (2015) Periconceptional folate consumption is associated with neonatal DNA methylation modifications in neural crest regulatory and cancer development genes

    Epigenetics 10:1166–1176.

    https://doi.org/10.1080/15592294.2015.1117889

    • PubMed
    • Google Scholar
31. 1. Gunasekara CJ
    2. Waterland RA

    (2019) A new era for epigenetic epidemiology

    Epigenomics 11:1647–1649.

    https://doi.org/10.2217/epi-2019-0282

    • PubMed
    • Google Scholar
32. 1. Guo H
    2. Zhu P
    3. Yan L
    4. Li R
    5. Hu B
    6. Lian Y
    7. Yan J
    8. Ren X
    9. Lin S
    10. Li J
    11. Jin X
    12. Shi X
    13. Liu P
    14. Wang X
    15. Wang W
    16. Wei Y
    17. Li X
    18. Guo F
    19. Wu X
    20. Fan X
    21. Yong J
    22. Wen L
    23. Xie SX
    24. Tang F
    25. Qiao J

    (2014) The DNA methylation landscape of human early embryos

    Nature 511:606–610.

    https://doi.org/10.1038/nature13544

    • PubMed
    • Google Scholar
33. 1. Hammoud SS
    2. Nix DA
    3. Zhang H
    4. Purwar J
    5. Carrell DT
    6. Cairns BR

    (2009) Distinctive chromatin in human sperm packages genes for embryo development

    Nature 460:473–478.

    https://doi.org/10.1038/nature08162

    • PubMed
    • Google Scholar
34. 1. Hennig BJ
    2. Unger SA
    3. Dondeh BL
    4. Hassan J
    5. Hawkesworth S
    6. Jarjou L
    7. Jones KS
    8. Moore SE
    9. Nabwera HM
    10. Ngum M
    11. Prentice A
    12. Sonko B
    13. Prentice AM
    14. Fulford AJ

    (2017) Cohort Profile: The Kiang West Longitudinal Population Study (KWLPS)-a platform for integrated research and health care provision in rural Gambia

    International Journal of Epidemiology 46:e13.

    https://doi.org/10.1093/ije/dyv206

    • PubMed
    • Google Scholar
35. 1. Hernandez-Vargas H
    2. Castelino J
    3. Silver MJ
    4. Dominguez-Salas P
    5. Cros M-P
    6. Durand G
    7. Le Calvez-Kelm F
    8. Prentice AM
    9. Wild CP
    10. Moore SE
    11. Hennig BJ
    12. Herceg Z
    13. Gong YY
    14. Routledge MN

    (2015) Exposure to aflatoxin B1 in utero is associated with DNA methylation in white blood cells of infants in The Gambia

    International Journal of Epidemiology 44:1238–1248.

    https://doi.org/10.1093/ije/dyv027

    • PubMed
    • Google Scholar
36. 1. Imbeault M
    2. Helleboid PY
    3. Trono D

    (2017) KRAB zinc-finger proteins contribute to the evolution of gene regulatory networks

    Nature 543:550–554.

    https://doi.org/10.1038/nature21683

    • PubMed
    • Google Scholar
37. 1. Irwin RE
    2. Thursby S-J
    3. Ondičová M
    4. Pentieva K
    5. McNulty H
    6. Richmond RC
    7. Caffrey A
    8. Lees-Murdock DJ
    9. McLaughlin M
    10. Cassidy T
    11. Suderman M
    12. Relton CL
    13. Walsh CP

    (2019) A randomized controlled trial of folic acid intervention in pregnancy highlights A putative methylation-regulated control element at ZFP57

    Clinical Epigenetics 11:31.

    https://doi.org/10.1186/s13148-019-0618-0

    • PubMed
    • Google Scholar
38. 1. Jaffe AE
    2. Irizarry RA

    (2014) Accounting for cellular heterogeneity is critical in epigenome-wide association studies

    Genome Biology 15:R31.

    https://doi.org/10.1186/gb-2014-15-2-r31

    • PubMed
    • Google Scholar
39. 1. James P
    2. Dominguez-Salas P
    3. Hennig BJ
    4. Moore SE
    5. Prentice AM
    6. Silver MJ

    (2018a) Maternal One-Carbon Metabolism and Infant DNA Methylation between Contrasting Seasonal Environments: A Case Study from The Gambia

    Current Developments in Nutrition 3:e82.

    https://doi.org/10.1093/cdn/nzy082

    • Google Scholar
40. 1. James P
    2. Sajjadi S
    3. Tomar AS
    4. Saffari A
    5. Fall CHD
    6. Prentice AM
    7. Shrestha S
    8. Issarapu P
    9. Yadav DK
    10. Kaur L
    11. Lillycrop K
    12. Silver M
    13. Chandak GR
    14. EMPHASIS study group

    (2018b) Candidate genes linking maternal nutrient exposure to offspring health via DNA methylation: a review of existing evidence in humans with specific focus on one-carbon metabolism

    International Journal of Epidemiology 47:1910–1937.

    https://doi.org/10.1093/ije/dyy153

    • PubMed
    • Google Scholar
41. 1. Jeltsch A

    (2006) DNA methyltransferases: facts, clues, mysteries

    Basic Mechanisms 1:203–225.

    https://doi.org/10.1007/3-540-31390-7

    • PubMed
    • Google Scholar
42. 1. Kang L
    2. Sun J
    3. Wen X
    4. Cui J
    5. Wang G
    6. Hoffman AR
    7. Hu J-F
    8. Li W

    (2015) Aberrant allele-switch imprinting of a novel IGF1R intragenic antisense non-coding RNA in breast cancers

    European Journal of Cancer (Oxford, England 51:260–270.

    https://doi.org/10.1016/j.ejca.2014.10.031

    • PubMed
    • Google Scholar
43. 1. Kazachenka A
    2. Bertozzi TM
    3. Sjoberg-Herrera MK
    4. Walker N
    5. Gardner J
    6. Gunning R
    7. Pahita E
    8. Adams S
    9. Adams D
    10. Ferguson-Smith AC

    (2018) Identification, Characterization, and Heritability of Murine Metastable Epialleles: Implications for Non-genetic Inheritance

    Cell 175:1259–1271.

    https://doi.org/10.1016/j.cell.2018.09.043

    • PubMed
    • Google Scholar
44. 1. Kessler NJ
    2. Waterland RA
    3. Prentice AM
    4. Silver MJ

    (2018) Establishment of environmentally sensitive DNA methylation states in the very early human embryo

    Science Advances 4:eaat2624.

    https://doi.org/10.1126/sciadv.aat2624

    • PubMed
    • Google Scholar
45. 1. Kuehnen P
    2. Mischke M
    3. Wiegand S
    4. Sers C
    5. Horsthemke B
    6. Lau S
    7. Keil T
    8. Lee Y-A
    9. Grueters A
    10. Krude H

    (2012) An Alu element-associated hypermethylation variant of the POMC gene is associated with childhood obesity

    PLOS Genetics 8:e1002543.

    https://doi.org/10.1371/journal.pgen.1002543

    • PubMed
    • Google Scholar
46. 1. Kühnen P
    2. Handke D
    3. Waterland RA
    4. Hennig BJ
    5. Silver M
    6. Fulford AJ
    7. Dominguez-Salas P
    8. Moore SE
    9. Prentice AM
    10. Spranger J
    11. Hinney A
    12. Hebebrand J
    13. Heppner FL
    14. Walzer L
    15. Grötzinger C
    16. Gromoll J
    17. Wiegand S
    18. Grüters A
    19. Krude H

    (2016) Interindividual Variation in DNA Methylation at a Putative POMC Metastable Epiallele Is Associated with Obesity

    Cell Metabolism 24:502–509.

    https://doi.org/10.1016/j.cmet.2016.08.001

    • PubMed
    • Google Scholar
47. 1. Larsson O
    2. Girnita A
    3. Girnita L

    (2005) Role of insulin-like growth factor 1 receptor signalling in cancer

    British Journal of Cancer 92:2097–2101.

    https://doi.org/10.1038/sj.bjc.6602627

    • PubMed
    • Google Scholar
48. 1. Logue MW
    2. Smith AK
    3. Wolf EJ
    4. Maniates H
    5. Stone A
    6. Schichman SA
    7. McGlinchey RE
    8. Milberg W
    9. Miller MW

    (2017) The correlation of methylation levels measured using Illumina 450K and EPIC BeadChips in blood samples

    Epigenomics 9:1363–1371.

    https://doi.org/10.2217/epi-2017-0078

    • PubMed
    • Google Scholar
49. 1. Min JL
    2. Hemani G
    3. Davey Smith G
    4. Relton C
    5. Suderman M

    (2018) Meffil: efficient normalization and analysis of very large DNA methylation datasets

    Bioinformatics (Oxford, England) 34:3983–3989.

    https://doi.org/10.1093/bioinformatics/bty476

    • PubMed
    • Google Scholar
50. 1. Molaro A
    2. Hodges E
    3. Fang F
    4. Song Q
    5. McCombie WR
    6. Hannon GJ
    7. Smith AD

    (2011) Sperm methylation profiles reveal features of epigenetic inheritance and evolution in primates

    Cell 146:1029–1041.

    https://doi.org/10.1016/j.cell.2011.08.016

    • PubMed
    • Google Scholar
51. 1. Monk D
    2. Mackay DJG
    3. Eggermann T
    4. Maher ER
    5. Riccio A

    (2019) Genomic imprinting disorders: lessons on how genome, epigenome and environment interact

    Nature Reviews. Genetics 20:235–248.

    https://doi.org/10.1038/s41576-018-0092-0

    • PubMed
    • Google Scholar
52. 1. Monteagudo-Sánchez A
    2. Hernandez Mora JR
    3. Simon C
    4. Burton A
    5. Tenorio J
    6. Lapunzina P
    7. Clark S
    8. Esteller M
    9. Kelsey G
    10. López-Siguero JP
    11. de Nanclares GP
    12. Torres-Padilla M-E
    13. Monk D

    (2020) The role of ZFP57 and additional KRAB-zinc finger proteins in the maintenance of human imprinted methylation and multi-locus imprinting disturbances

    Nucleic Acids Research 48:11394–11407.

    https://doi.org/10.1093/nar/gkaa837

    • PubMed
    • Google Scholar
53. 1. Moore SE
    2. Cole TJ
    3. Collinson AC
    4. Poskitt EM
    5. McGregor IA
    6. Prentice AM

    (1999) Prenatal or early postnatal events predict infectious deaths in young adulthood in rural Africa

    International Journal of Epidemiology 28:1088–1095.

    https://doi.org/10.1093/ije/28.6.1088

    • PubMed
    • Google Scholar
54. 1. Moore SE
    2. Fulford AJ
    3. Darboe MK
    4. Jobarteh ML
    5. Jarjou LM
    6. Prentice AM

    (2012) A randomized trial to investigate the effects of pre-natal and infant nutritional supplementation on infant immune development in rural Gambia: the ENID trial: Early Nutrition and Immune Development

    BMC Pregnancy and Childbirth 12:107.

    https://doi.org/10.1186/1471-2393-12-107

    • PubMed
    • Google Scholar
55. 1. Okae H
    2. Chiba H
    3. Hiura H
    4. Hamada H
    5. Sato A
    6. Utsunomiya T
    7. Kikuchi H
    8. Yoshida H
    9. Tanaka A
    10. Suyama M
    11. Arima T

    (2014) Genome-wide analysis of DNA methylation dynamics during early human development

    PLOS Genetics 10:e1004868.

    https://doi.org/10.1371/journal.pgen.1004868

    • PubMed
    • Google Scholar
56. 1. Owens S
    2. Gulati R
    3. Fulford AJ
    4. Sosseh F
    5. Denison FC
    6. Brabin BJ
    7. Prentice AM

    (2015) Periconceptional multiple-micronutrient supplementation and placental function in rural Gambian women: A double-blind, randomized, placebo-controlled trial

    The American Journal of Clinical Nutrition 102:1450–1459.

    https://doi.org/10.3945/ajcn.113.072413

    • PubMed
    • Google Scholar
57. 1. Pan H
    2. Holbrook JD
    3. Karnani N
    4. Kwoh CKG

    (2016) Gene, Environment and Methylation (GEM): a tool suite to efficiently navigate large scale epigenome wide association studies and integrate genotype and interaction between genotype and environment

    BMC Bioinformatics 17:299.

    https://doi.org/10.1186/s12859-016-1161-z

    • PubMed
    • Google Scholar
58. 1. Rada-Iglesias A
    2. Bajpai R
    3. Swigut T
    4. Brugmann SA
    5. Flynn RA
    6. Wysocka J

    (2011) A unique chromatin signature uncovers early developmental enhancers in humans

    Nature 470:279–283.

    https://doi.org/10.1038/nature09692

    • PubMed
    • Google Scholar
59. 1. Rakyan VK
    2. Blewitt ME
    3. Druker R
    4. Preis JI
    5. Whitelaw E

    (2002) Metastable epialleles in mammals

    Trends in Genetics 18:348–351.

    https://doi.org/10.1016/s0168-9525(02)02709-9

    • PubMed
    • Google Scholar
60. 1. Randhawa R
    2. Cohen P

    (2005) The role of the insulin-like growth factor system in prenatal growth

    Molecular Genetics and Metabolism 86:84–90.

    https://doi.org/10.1016/j.ymgme.2005.07.028

    • PubMed
    • Google Scholar
61. 1. Rayco-Solon P
    2. Fulford AJ
    3. Prentice AM

    (2005) Differential effects of seasonality on preterm birth and intrauterine growth restriction in rural Africans

    The American Journal of Clinical Nutrition 81:134–139.

    https://doi.org/10.1093/ajcn/81.1.134

    • PubMed
    • Google Scholar
62. 1. Roadmap Epigenomics Consortium
    2. Kundaje A
    3. Meuleman W
    4. Ernst J
    5. Bilenky M
    6. Yen A
    7. Heravi-Moussavi A
    8. Kheradpour P
    9. Zhang Z
    10. Wang J
    11. Ziller MJ
    12. Amin V
    13. Whitaker JW
    14. Schultz MD
    15. Ward LD
    16. Sarkar A
    17. Quon G
    18. Sandstrom RS
    19. Eaton ML
    20. Wu Y-C
    21. Pfenning AR
    22. Wang X
    23. Claussnitzer M
    24. Liu Y
    25. Coarfa C
    26. Harris RA
    27. Shoresh N
    28. Epstein CB
    29. Gjoneska E
    30. Leung D
    31. Xie W
    32. Hawkins RD
    33. Lister R
    34. Hong C
    35. Gascard P
    36. Mungall AJ
    37. Moore R
    38. Chuah E
    39. Tam A
    40. Canfield TK
    41. Hansen RS
    42. Kaul R
    43. Sabo PJ
    44. Bansal MS
    45. Carles A
    46. Dixon JR
    47. Farh K-H
    48. Feizi S
    49. Karlic R
    50. Kim A-R
    51. Kulkarni A
    52. Li D
    53. Lowdon R
    54. Elliott G
    55. Mercer TR
    56. Neph SJ
    57. Onuchic V
    58. Polak P
    59. Rajagopal N
    60. Ray P
    61. Sallari RC
    62. Siebenthall KT
    63. Sinnott-Armstrong NA
    64. Stevens M
    65. Thurman RE
    66. Wu J
    67. Zhang B
    68. Zhou X
    69. Beaudet AE
    70. Boyer LA
    71. De Jager PL
    72. Farnham PJ
    73. Fisher SJ
    74. Haussler D
    75. Jones SJM
    76. Li W
    77. Marra MA
    78. McManus MT
    79. Sunyaev S
    80. Thomson JA
    81. Tlsty TD
    82. Tsai L-H
    83. Wang W
    84. Waterland RA
    85. Zhang MQ
    86. Chadwick LH
    87. Bernstein BE
    88. Costello JF
    89. Ecker JR
    90. Hirst M
    91. Meissner A
    92. Milosavljevic A
    93. Ren B
    94. Stamatoyannopoulos JA
    95. Wang T
    96. Kellis M

    (2015) Integrative analysis of 111 reference human epigenomes

    Nature 518:317–330.

    https://doi.org/10.1038/nature14248

    • PubMed
    • Google Scholar
63. 1. Saffari A
    2. Shrestha S
    3. Issarapu P
    4. Sajjadi S
    5. Betts M
    6. Sahariah SA
    7. Tomar AS
    8. James P
    9. Dedaniya A
    10. Yadav DK
    11. Kumaran K
    12. Prentice AM
    13. Lillycrop KA
    14. Fall CHD
    15. Chandak GR
    16. Silver MJ
    17. EMPHASIS Study Group

    (2020) Effect of maternal preconceptional and pregnancy micronutrient interventions on children’s DNA methylation: Findings from the EMPHASIS study

    The American Journal of Clinical Nutrition 112:1099–1113.

    https://doi.org/10.1093/ajcn/nqaa193

    • PubMed
    • Google Scholar
64. 1. Sanchez-Delgado M
    2. Court F
    3. Vidal E
    4. Medrano J
    5. Monteagudo-Sánchez A
    6. Martin-Trujillo A
    7. Tayama C
    8. Iglesias-Platas I
    9. Kondova I
    10. Bontrop R
    11. Poo-Llanillo ME
    12. Marques-Bonet T
    13. Nakabayashi K
    14. Simón C
    15. Monk D

    (2016) Human Oocyte-Derived Methylation Differences Persist in the Placenta Revealing Widespread Transient Imprinting

    PLOS Genetics 12:e1006427.

    https://doi.org/10.1371/journal.pgen.1006427

    • PubMed
    • Google Scholar
65. 1. Shah S
    2. McRae AF
    3. Marioni RE
    4. Harris SE
    5. Gibson J
    6. Henders AK
    7. Redmond P
    8. Cox SR
    9. Pattie A
    10. Corley J
    11. Murphy L
    12. Martin NG
    13. Montgomery GW
    14. Starr JM
    15. Wray NR
    16. Deary IJ
    17. Visscher PM

    (2014) Genetic and environmental exposures constrain epigenetic drift over the human life course

    Genome Research 24:1725–1733.

    https://doi.org/10.1101/gr.176933.114

    • PubMed
    • Google Scholar
66. 1. Shi H
    2. Strogantsev R
    3. Takahashi N
    4. Kazachenka A
    5. Lorincz MC
    6. Hemberger M
    7. Ferguson-Smith AC

    (2019) ZFP57 regulation of transposable elements and gene expression within and beyond imprinted domains

    Epigenetics & Chromatin 12:e49.

    https://doi.org/10.1186/s13072-019-0295-4

    • PubMed
    • Google Scholar
67. 1. Silver MJ
    2. Kessler NJ
    3. Hennig BJ
    4. Dominguez-Salas P
    5. Laritsky E
    6. Baker MS
    7. Coarfa C
    8. Hernandez-Vargas H
    9. Castelino JM
    10. Routledge MN
    11. Gong YY
    12. Herceg Z
    13. Lee YS
    14. Lee K
    15. Moore SE
    16. Fulford AJ
    17. Prentice AM
    18. Waterland RA

    (2015) Independent genomewide screens identify the tumor suppressor VTRNA2-1 as a human epiallele responsive to periconceptional environment

    Genome Biology 16:118.

    https://doi.org/10.1186/s13059-015-0660-y

    • PubMed
    • Google Scholar
68. 1. Simpkin AJ
    2. Suderman M
    3. Gaunt TR
    4. Lyttleton O
    5. McArdle WL
    6. Ring SM
    7. Tilling K
    8. Davey Smith G
    9. Relton CL

    (2015) Longitudinal analysis of DNA methylation associated with birth weight and gestational age

    Human Molecular Genetics 24:3752–3763.

    https://doi.org/10.1093/hmg/ddv119

    • PubMed
    • Google Scholar
69. 1. Singmann P
    2. Shem-Tov D
    3. Wahl S
    4. Grallert H
    5. Fiorito G
    6. Shin SY
    7. Schramm K
    8. Wolf P
    9. Kunze S
    10. Baran Y
    11. Guarrera S
    12. Vineis P
    13. Krogh V
    14. Panico S
    15. Tumino R
    16. Kretschmer A
    17. Gieger C
    18. Peters A
    19. Prokisch H
    20. Relton CL
    21. Matullo G
    22. Illig T
    23. Waldenberger M
    24. Halperin E

    (2015) Characterization of whole-genome autosomal differences of DNA methylation between men and women

    Epigenetics & Chromatin 8:e43.

    https://doi.org/10.1186/s13072-015-0035-3

    • PubMed
    • Google Scholar
70. 1. Smith ZD
    2. Meissner A

    (2013) DNA methylation: roles in mammalian development

    Nature Reviews. Genetics 14:204–220.

    https://doi.org/10.1038/nrg3354

    • PubMed
    • Google Scholar
71. 1. Suderman M
    2. Simpkin A
    3. Sharp G
    4. Gaunt T
    5. Lyttleton O
    6. McArdle W
    7. Ring S
    8. Davey Smith G
    9. Relton C

    (2017) Associations between body mass index-related genetic variants and adult body composition: The Fenland cohort study

    Epidemiology (Cambridge, Mass.) 1:e265.

    https://doi.org/10.1101/118265

    • Google Scholar
72. 1. Sugden K
    2. Hannon EJ
    3. Arseneault L
    4. Belsky DW
    5. Corcoran DL
    6. Fisher HL
    7. Houts RM
    8. Kandaswamy R
    9. Moffitt TE
    10. Poulton R
    11. Prinz JA
    12. Rasmussen LJH
    13. Williams BS
    14. Wong CCY
    15. Mill J
    16. Caspi A

    (2020) Patterns of Reliability: Assessing the Reproducibility and Integrity of DNA Methylation Measurement

    Patterns (New York, N.Y.) 1:100014.

    https://doi.org/10.1016/j.patter.2020.100014

    • PubMed
    • Google Scholar
73. 1. Sun J
    2. Li W
    3. Sun Y
    4. Yu D
    5. Wen X
    6. Wang H
    7. Cui J
    8. Wang G
    9. Hoffman AR
    10. Hu J-F

    (2014) A novel antisense long noncoding RNA within the IGF1R gene locus is imprinted in hematopoietic malignancies

    Nucleic Acids Research 42:9588–9601.

    https://doi.org/10.1093/nar/gku549

    • PubMed
    • Google Scholar
74. 1. Teschendorff AE
    2. West J
    3. Beck S

    (2013) Age-associated epigenetic drift: implications, and a case of epigenetic thrift?

    Human Molecular Genetics 22:R7–R15.

    https://doi.org/10.1093/hmg/ddt375

    • PubMed
    • Google Scholar
75. 1. Teschendorff AE
    2. Relton CL

    (2018) Statistical and integrative system-level analysis of DNA methylation data

    Nature Reviews. Genetics 19:129–147.

    https://doi.org/10.1038/nrg.2017.86

    • PubMed
    • Google Scholar
76. 1. Tobi EW
    2. Slagboom PE
    3. van Dongen J
    4. Kremer D
    5. Stein AD
    6. Putter H
    7. Heijmans BT
    8. Lumey LH
    9. Lyko F

    (2012) Prenatal Famine and Genetic Variation Are Independently and Additively Associated with DNA Methylation at Regulatory Loci within IGF2/H19

    PLOS ONE 7:e37933.

    https://doi.org/10.1371/journal.pone.0037933

    • PubMed
    • Google Scholar
77. 1. Tobi EW
    2. Slieker RC
    3. Stein AD
    4. Suchiman HED
    5. Slagboom PE
    6. van Zwet EW
    7. Heijmans BT
    8. Lumey LH

    (2015) Early gestation as the critical time-window for changes in the prenatal environment to affect the adult human blood methylome

    International Journal of Epidemiology 44:1211–1223.

    https://doi.org/10.1093/ije/dyv043

    • PubMed
    • Google Scholar
78. 1. Tobi EW
    2. van den Heuvel J
    3. Zwaan BJ
    4. Lumey LH
    5. Heijmans BT
    6. Uller T

    (2018) Selective Survival of Embryos Can Explain DNA Methylation Signatures of Adverse Prenatal Environments

    Cell Reports 25:2660–2667.

    https://doi.org/10.1016/j.celrep.2018.11.023

    • PubMed
    • Google Scholar
79. 1. Tsai P-C
    2. Van Dongen J
    3. Tan Q
    4. Willemsen G
    5. Christiansen L
    6. Boomsma DI
    7. Spector TD
    8. Valdes AM
    9. Bell JT

    (2015) DNA Methylation Changes in the IGF1R Gene in Birth Weight Discordant Adult Monozygotic Twins

    Twin Research and Human Genetics 18:635–646.

    https://doi.org/10.1017/thg.2015.76

    • PubMed
    • Google Scholar
80. 1. Van Baak TE
    2. Coarfa C
    3. Dugué P-A
    4. Fiorito G
    5. Laritsky E
    6. Baker MS
    7. Kessler NJ
    8. Dong J
    9. Duryea JD
    10. Silver MJ
    11. Saffari A
    12. Prentice AM
    13. Moore SE
    14. Ghantous A
    15. Routledge MN
    16. Gong YY
    17. Herceg Z
    18. Vineis P
    19. Severi G
    20. Hopper JL
    21. Southey MC
    22. Giles GG
    23. Milne RL
    24. Waterland RA

    (2018) Epigenetic supersimilarity of monozygotic twin pairs

    Genome Biology 19:2.

    https://doi.org/10.1186/s13059-017-1374-0

    • PubMed
    • Google Scholar
81. 1. van der Harst P
    2. van Setten J
    3. Verweij N
    4. Vogler G
    5. Franke L
    6. Maurano MT
    7. Wang X
    8. Mateo Leach I
    9. Eijgelsheim M
    10. Sotoodehnia N
    11. Hayward C
    12. Sorice R
    13. Meirelles O
    14. Lyytikäinen L-P
    15. Polašek O
    16. Tanaka T
    17. Arking DE
    18. Ulivi S
    19. Trompet S
    20. Müller-Nurasyid M
    21. Smith AV
    22. Dörr M
    23. Kerr KF
    24. Magnani JW
    25. Del Greco M. F
    26. Zhang W
    27. Nolte IM
    28. Silva CT
    29. Padmanabhan S
    30. Tragante V
    31. Esko T
    32. Abecasis GR
    33. Adriaens ME
    34. Andersen K
    35. Barnett P
    36. Bis JC
    37. Bodmer R
    38. Buckley BM
    39. Campbell H
    40. Cannon MV
    41. Chakravarti A
    42. Chen LY
    43. Delitala A
    44. Devereux RB
    45. Doevendans PA
    46. Dominiczak AF
    47. Ferrucci L
    48. Ford I
    49. Gieger C
    50. Harris TB
    51. Haugen E
    52. Heinig M
    53. Hernandez DG
    54. Hillege HL
    55. Hirschhorn JN
    56. Hofman A
    57. Hubner N
    58. Hwang S-J
    59. Iorio A
    60. Kähönen M
    61. Kellis M
    62. Kolcic I
    63. Kooner IK
    64. Kooner JS
    65. Kors JA
    66. Lakatta EG
    67. Lage K
    68. Launer LJ
    69. Levy D
    70. Lundby A
    71. Macfarlane PW
    72. May D
    73. Meitinger T
    74. Metspalu A
    75. Nappo S
    76. Naitza S
    77. Neph S
    78. Nord AS
    79. Nutile T
    80. Okin PM
    81. Olsen JV
    82. Oostra BA
    83. Penninger JM
    84. Pennacchio LA
    85. Pers TH
    86. Perz S
    87. Peters A
    88. Pinto YM
    89. Pfeufer A
    90. Pilia MG
    91. Pramstaller PP
    92. Prins BP
    93. Raitakari OT
    94. Raychaudhuri S
    95. Rice KM
    96. Rossin EJ
    97. Rotter JI
    98. Schafer S
    99. Schlessinger D
    100. Schmidt CO
    101. Sehmi J
    102. Silljé HHW
    103. Sinagra G
    104. Sinner MF
    105. Slowikowski K
    106. Soliman EZ
    107. Spector TD
    108. Spiering W
    109. Stamatoyannopoulos JA
    110. Stolk RP
    111. Strauch K
    112. Tan S-T
    113. Tarasov KV
    114. Trinh B
    115. Uitterlinden AG
    116. van den Boogaard M
    117. van Duijn CM
    118. van Gilst WH
    119. Viikari JS
    120. Visscher PM
    121. Vitart V
    122. Völker U
    123. Waldenberger M
    124. Weichenberger CX
    125. Westra H-J
    126. Wijmenga C
    127. Wolffenbuttel BH
    128. Yang J
    129. Bezzina CR
    130. Munroe PB
    131. Snieder H
    132. Wright AF
    133. Rudan I
    134. Boyer LA
    135. Asselbergs FW
    136. van Veldhuisen DJ
    137. Stricker BH
    138. Psaty BM
    139. Ciullo M
    140. Sanna S
    141. Lehtimäki T
    142. Wilson JF
    143. Bandinelli S
    144. Alonso A
    145. Gasparini P
    146. Jukema JW
    147. Kääb S
    148. Gudnason V
    149. Felix SB
    150. Heckbert SR
    151. de Boer RA
    152. Newton-Cheh C
    153. Hicks AA
    154. Chambers JC
    155. Jamshidi Y
    156. Visel A
    157. Christoffels VM
    158. Isaacs A
    159. Samani NJ
    160. de Bakker PIW

    (2016) 52 Genetic Loci Influencing Myocardial Mass

    Journal of the American College of Cardiology 68:1435–1448.

    https://doi.org/10.1016/j.jacc.2016.07.729

    • PubMed
    • Google Scholar
82. 1. Vukic M
    2. Wu H
    3. Daxinger L

    (2019) Making headway towards understanding how epigenetic mechanisms contribute to early-life effects

    Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 374:20180126.

    https://doi.org/10.1098/rstb.2018.0126

    • PubMed
    • Google Scholar
83. 1. Waterland RA
    2. Jirtle RL

    (2003) Transposable elements: targets for early nutritional effects on epigenetic gene regulation

    Molecular and Cellular Biology 23:5293–5300.

    https://doi.org/10.1128/MCB.23.15.5293-5300.2003

    • PubMed
    • Google Scholar
84. 1. Waterland RA
    2. Kellermayer R
    3. Laritsky E
    4. Rayco-Solon P
    5. Harris RA
    6. Travisano M
    7. Zhang W
    8. Torskaya MS
    9. Zhang J
    10. Shen L
    11. Manary MJ
    12. Prentice AM

    (2010) Season of conception in rural gambia affects DNA methylation at putative human metastable epialleles

    PLOS Genetics 6:e1001252.

    https://doi.org/10.1371/journal.pgen.1001252

    • PubMed
    • Google Scholar
85. 1. Weale ME

    (2010) Quality control for genome-wide association studies

    Methods in Molecular Biology (Clifton, N.J.) 628:341–372.

    https://doi.org/10.1007/978-1-60327-367-1_19

    • PubMed
    • Google Scholar
86. 1. Xu Z
    2. Taylor JA

    (2021) Reliability of DNA methylation measures using Illumina methylation BeadChip

    Epigenetics 16:495–502.

    https://doi.org/10.1080/15592294.2020.1805692

    • PubMed
    • Google Scholar
87. 1. Zheng Y
    2. Joyce BT
    3. Liu L
    4. Zhang Z
    5. Kibbe WA
    6. Zhang W
    7. Hou L

    (2017) Prediction of genome-wide DNA methylation in repetitive elements

    Nucleic Acids Research 45:8697–8711.

    https://doi.org/10.1093/nar/gkx587

    • PubMed
    • Google Scholar
88. 1. Zink F
    2. Magnusdottir DN
    3. Magnusson OT
    4. Walker NJ
    5. Morris TJ
    6. Sigurdsson A
    7. Halldorsson GH
    8. Gudjonsson SA
    9. Melsted P
    10. Ingimundardottir H
    11. Kristmundsdottir S
    12. Alexandersson KF
    13. Helgadottir A
    14. Gudmundsson J
    15. Rafnar T
    16. Jonsdottir I
    17. Holm H
    18. Eyjolfsson GI
    19. Sigurdardottir O
    20. Olafsson I
    21. Masson G
    22. Gudbjartsson DF
    23. Thorsteinsdottir U
    24. Halldorsson BV
    25. Stacey SN
    26. Stefansson K

    (2018) Insights into imprinting from parent-of-origin phased methylomes and transcriptomes

    Nature Genetics 50:1542–1552.

    https://doi.org/10.1038/s41588-018-0232-7

    • PubMed
    • Google Scholar

Decision letter after peer review:

Thank you for submitting your article "Environmentally sensitive hotspots in the methylome of the early human embryo" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Jessica Tyler as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Toby Mansell (Reviewer #1).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

Below are first the most important points coming from the discussion between the reviewers, followed by additional points coming from the individual review reports.

1) The reviewers are worried that there is residual confounding impacting the analyses and that the authors only observe increases in DNA methylation. Regressing out sex is not sufficient to show there is no bias in the analyses. We therefore request a correlation matrix (containing rho's, not p-values) containing all batch effects, arrray covariates (e.g. sentrix row and column), biological covariates (e.g. sex), PCAs before and after functional normalisation, and study covariates (e.g. village, month of collection, month of birth, month of conception). This matrix should show if there is residual confounding remaining that should be taken into account in the analyses.

2) The authors should make clear why some decisions were made regarding certain methods (i.e. adjustment using PCA instead of Houseman estimates) and thresholds (i.e. SoC-associated CpGs with difference <4%). More on this in point 14 and 15.

3) The Discussion section should be rewritten to better reflect the relevance of the findings and the limitations of the study. For example, the actual number of unique seasonal MEs is modest and the discussion should reflect this. Moreover, the findings should be compared with previously published studies on, e.g., folic acid supplementation and famine (this last point also applies to the Introduction). More on this in point 12 and 13.

Additional points from individual review reports:

Results section:

4) Page 3, line 94: Have the authors access to longitudinal intra-individual methylation scores during different seasonal windows?

5) Page 4, line 141: Did the authors re-evaluate SoC CpG positions with alternative methods (e.g. pyrosequencing) to analyse larger groups/ cluster of CpG positions and to control the reliability of array data?

6) Line 238 onward: from 257 Soc-CpGs to the "full" set of 768 SoC-CpGs for enrichment tests out of "power" reasons ◊ not consistent and confusing. I would stick with your choice for the 257.

7) Page 7, line 301: The authors mentioned overlaps with previous studies. It would be helpful, if overlapping or related findings could be summarized (and maybe visualized) in more detail. This would facilitate the comparison of the current data set with previous publications.

8) I would strongly recommend reporting the relationship between methylation of these SoC sites and measures of growth or development in these two cohorts – it's not clear to me why this has not been done so, assuming such data is available under approved ethics. In the field there is an ever-growing body of papers linking various early development exposures to differences in methylation in childhood, but there is a dearth of reproduceable evidence for exposures associating with methylation differences that in turn associate with tangible differences in child health/development. Particularly since the manuscript already mentions things like the relationship between methylation and BMI – while adult BMI is different to child measures of body composition, any evidence for a relationship (or a lack of a relationship) with available phenotypes in these two child cohorts would be a logical conclusion that section of analysis and would provide much greater clarity for the relevant Discussion section.

9) Relatedly, if consistent with approved ethics, this paper would benefit from a table summarising the cohort characteristics for each cohort. If appropriate, stratifying the cohorts summaries by season of conception would also be of interest.

10) Rationale for not doing a meta-analysis for power reasons should be explained.

11) The Results section switches from past to present tense and back again. Please keep it to past tense.

Discussion section

12) From my point of view, the observation of an attenuation of DNA methylation levels until mid-childhood is of interest. It would be helpful, if the authors could discuss this point in more detail within the Discussion section. What is known about methylation changes during aging (epigenetic drift etc.) and are these observations relevant for SoC associated CpGs?

13) The authors might discuss in more detail, why these observations of SoC associated loci are disease relevant.

Methods section

14) As mentioned in the public review, I think there needs to be more clarity about why you do and do not use the SoC-associated CpGs with difference <4% – in the results it is stated that the full list is being used to 'maximise power', but it is unclear why maximising power is necessary/better for that analysis compared to the other analyses where only CpGs with difference >4% are used? A little more detail on why it is important/fine to use there would be helpful.

15) Another point that would benefit from additional clarity would be adjusting for cell composition differences in your models – it appears that you have adjusted for these by including the top 6 PCs, but you also have cell count estimates from the Houseman method. Was there are a particular reason for using the PCs when the Houseman estimates were available?

16) Methods does not read as one coherent story, but fragmented and will require some shuffling to make it easier to read.

17) Outlier removal is with 25th and 75th percentile very aggressive compared to what is standard in the field.

18) You can Check SNPs in 450k dataset for possibility to adjust for genetic background in the discovery cohort and 850k cohort to exclude sample swaps/mixtures etc.

19) Reading the methods you are worried by genome-wide effects on DNAm. Why not look at the LINES ALU effects via: Zheng Y Joyce BT Liu L Zhang Z Kibbe WA Zhang W Hou L. Prediction of genome-wide DNA methylation in repetitive elements. Nucleic Acids Res 2017;45:8697-8711 [works for 450k data]

20) Were African allele frequencies used to filter out CpG probes?

21) Was imputation done with AGVP?

22) Was genetic information used to exclude family relationships, mixtures and sample duplications? [knowing the fluidity of family structures and logistics of sample collection in the Gambia]

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Environmentally sensitive hotspots in the methylome of the early human embryo" for further consideration by eLife. Your revised article has been evaluated by Jessica Tyler (Senior Editor) and a Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below. Two of the three original reviewers were happy with the way you addressed their comments, but since the other reviewer was no longer available, I have asked an additional reviewer to specifically take a look at the applied adjustment for batch effects (which was one of the main points brought up by this original reviewer). As you can see below, the new reviewer was in general happy with the rebuttal, but brought up some additional points that need to be addressed.

Reviewer #4 (Recommendations for the authors):

Thank you for the opportunity to review the rebuttal to the article titled "Environmentally sensitive hotspots in the methylome of the early human embryo". I would like to say that reading through the rebuttal it appears the authors have been reactive and enthusiastic in their responses to the reviewers. Overall, it appears the changes introduced have help interpretation of the findings as far as I can tell.

In terms of the concerns reviewers have regarding batch effects, on the whole it appears that these have been these addressed as far as is possible within the confines of this particular experiment. The sensitivity analyses reported in table 1s suggest that controlling for the suspected technical confounders (e.g plate, slide, cell proportion estimates) does not significantly attenuate the association between SoC-CpG and doc. As noted, there does appear to be non-randomization with respect to Plate (and by extension, Slide) in the youngest ENID sample, which I suspect is due to assignment of samples to plates as they were collected (i.e in chronological order). What I cannot tell is whether the methylation arrays are also processed as samples are received (i.e. once a plate is filled), or all together at the end of the collection phase. If it is the former, that is a likely source of some variation between plates. Given the use of this cohort as the discovery sample, it would be useful if the authors could declare whether or not this is the case in the methods.

I am however suspicious that there is an alternative confounding factor that could have implications for the interpretation of the results within the paper, and that is the notion of DNA methylation data reliability (or lack of). Multiple studies (e.g. Logue et al. (2017); Epigenomics 9(11):1363-1371; Sugden et al. Patterns 1(2): 100014) have demonstrated that a large proportion of the CpG sites shared between the 450K and EPIC array (as used in this paper) are unreliable, so that it is unlikely they yield the same value when the same sample is measured twice. This of course has implications for determining 'significant associations', for assessing DNA methylation changes over time, and for replication. Because of what we know about reliability and how it manifests, the authors should address reliability in the context of these findings.

Reassuringly for this paper, the SoC-CpGs appear to be pretty robust – cross-referencing the subset of 259 CpGs against the list of reliabilities from Sugden et al., shows the mean ICC is over.7 (close to 'excellent') – this is not the case for the 509 CpGs not taken forward, who have a mean reliability of less than 0.3 ('poor'). The list of CpGs selected as matched and array background controls does not appear available, but I strongly suspect the array background CpGs (at the very least) will have a much lower reliability than the SoC-CpGs (just due to random chance). This clouds the interpretation of any tests that compare these sets of CpGs and should be taken into consideration.

Further, differential reliability should be considered as an alternative explanation for the reduction in mean SoC amplitudes over time – could it be that measurements made using 450K arrays (ENID 2yr) are different to those using EPIC arrays for reasons beyond age-specific effects (i.e. measurement error)?

Finally, the description of the properties of SoC-CpGs (lines 173 onwards) notes that they are more likely to be intermediately methylated – this is again a property of reliable CpGs since arrays do not have the resolution to measure hyper- and hypo- methylated sites very well, leading to unreliable measurements. In all, these points suggest that reliability of DNA methylation is important for interpretation of these results, and, given the high reliability of the SoC-CpGs, is actually a positive feature.

The reviewers were concerned that a large proportion of the discussion was dedicated to MEs and how these findings are related. I believe the authors frame this appropriately, and do not find the references to them excessive. MEs are of great interest, and using cheaper and easier methods to uncover them (such as here) will be beneficial.

Essential revisions:

Below are first the most important points coming from the discussion between the reviewers, followed by additional points coming from the individual review reports.

(1) The reviewers are worried that there is residual confounding impacting the analyses and that the authors only observe increases in DNA methylation. Regressing out sex is not sufficient to show there is no bias in the analyses. We therefore request a correlation matrix (containing rho's, not p-values) containing all batch effects, arrray covariates (e.g. sentrix row and column), biological covariates (e.g. sex), PCAs before and after functional normalisation, and study covariates (e.g. village, month of collection, month of birth, month of conception). This matrix should show if there is residual confounding remaining that should be taken into account in the analyses.

We have carried out additional analyses to check for potential residual confounding (see new Supplementary Tables ST1p-1r), including an additional longitudinal analysis with recently acquired methylation data measured on the EPIC array on a subset of individuals from the ENID cohort at age 5-7yr.

Please note that the majority of batch and biological covariates are categorical so that it was not possible to report correlation rho’s. We have instead reported p-values for corresponding association tests – see Supplementary Tables for further details of tests that were carried out. Also note that for simplicity season of conception is modelled as a binary variable (Dry: Jan-Jun; Rainy: July-Dec). We consider this to be a valid approximation to the main cosinor (Fourier) regression analysis since this showed a clear relationship between DNAm and dichotomised (Dry/Rainy) season of conception (Figures2D and 4A). Note that we have not included month of collection as this completely confounds season of conception in the main ENID (2yr) analysis and cannot confound the EMPHASIS (7-9yr) analysis, as discussed in the manuscript (lines 88-91, Figure 2A). This is a key reason why we compared SoC effects across these two cohorts. Note that the month of collection also cannot confound the ENID 5-7yr (longitudinal) analysis as all samples are collected in the rainy season (additional Supp. Figure 2).

The covariate correlation analysis confirms:

– No correlation between SoC and all considered batch and biological covariates including principal components across all three analysed datasets (Supp Table, ST1p-1r).

– No correlation between sex and all considered batch and biological covariates; weak correlations with PC4 and PC3 in EMPHASIS and ENID 5-7yr datasets respectively (ST1q,1r). Please also note also that the sex sensitivity analysis previously reported in the manuscript used methylation values that were pre-adjusted for sex using a regression model that included sex as the only adjustment covariate, alleviating concerns that there may be residual confounding due to strong correlations between technical/biological/sampling covariates and sex. We have added some additional comments on this to Results (lines 335-341).

– Expected strong correlations between SoC, month of conception and month of birth in all datasets (ST1p-1r).

– Functional Normalisation (FN) removed most but not all of the effects of technical batch effects (sample plate, slide etc) from the DNAm array data used in the main ENID analysis (ST1p and Supp Table for Reviewers STR1).

– Samples are not perfectly randomised across 450k sample plate (month of birth [mob] and conception [moc]) and slide (mob and village) for the ENID 2yr cohort (ST1p).

The last point raises the possibility of potential residual confounding due to array batch effects in the ENID analysis. We checked for this in two ways. First, we performed sensitivity analyses with batch and village ID variables included directly in the linear regression models, in addition to the PCs that served as proxies for batch variables in our original analysis. This suggested no residual confounding due to array batch or village ID effects (ST1s: ‘batch adjusted model’ and ‘village adjusted model’). Second, we confirmed that neither mob, moc nor village ID were associated with batch or any other covariates in the EMPHASIS or new ENID 5-7yr analyses (ST1q, ST1r). The tight correspondence of date of methylation maximum across all three datasets (cross-cohort and longitudinal analyses) (Figures 2C, 3A and 4A) with different confounding structures (ST1p-1r) strongly suggests that the reported SoC associations are not driven by residual confounding.

In summary, this analysis provides strong reassurance that our main analysis is not confounded by residual associations with technical and/or biological covariates considered in this analysis, and that the observed enrichment for previously identified sex-associations amongst SoC-CpGs is not driven by residual confounding due to sex.

We have made multiple amendments to the manuscript to incorporate the longitudinal analysis; in the Introduction (lines 58-9); in the first section of Results (lines 87-156); and we have made particular reference to the alignment of SoC effects across 3 datasets with different confounding structures in lines 152-156. We have also amended several figure captions to distinguish the ENID 2yr and 5-7yr datasets and added the longitudinal dataset to Methods (lines 558-560) and to the study design schematic (revised Figure 1), and visualised key results from this additional analysis in Figures 3 and 4A. Finally we have added additional text on the sensitivity analyses in the main text (lines 131-133) and in Methods (lines 629-635).

(2) The authors should make clear why some decision were made regarding certain methods (i.e. adjustment using PCA instead of Houseman estimates) and thresholds (i.e. SoC-associated CpGs with difference <4%). More on this in point 14 and 15.

Regarding adjustment for cell count estimates using PCs as a proxy, we have performed an additional sensitivity analysis with Houseman estimated blood cell counts added directly to the linear regression model for the ENID cohort (see ST1s). 518 out of the 520 estimated Fourier regression coefficients from the main analysis (1 pair of sine and cosine terms for each of the 259 SoC-CpGs) fall within the 95% confidence interval obtained in the Houseman-adjusted analysis, confirming that cell composition effects did not unduly influence SoC effect estimates in the original analysis. As outlined in our response to the previous comment we have added a brief note on this and the other sensitivity analyses (batch, cell composition and village effects) in Results to the manuscript (lines 131-133), with more details in Methods (lines 629-635).

Regarding our use of different SoC amplitude thresholds for one analysis, our original motivation for analysing all 768 ‘SoC-associated CpGs’ with FDR<5% in the ENID 2yr analysis, including those with amplitude < 4%, was to explore the degree to which the strength / amplitude of SoC effects could be explained by proximity to ERV1 over the wider range of amplitudes represented by the larger set of loci. However we agree that this approach is open to question and have removed this analysis (previous Figure 6B and Supp. Figure 11, and text in section headed ‘Enrichment of transposable elements and transcription factors associated with genomic imprinting’). We have also removed the definition of ‘SoC-associated CpGs’ (which included CpGs with SoC amplitude < 4%) from Table 2 and Methods to aid clarity and avoid confusion.

(3) The Discussion section should be rewritten to better reflect the relevance of the findings and the limitations of the study. For example, the actual number of unique seasonal MEs is modest and the discussion should reflect this.

This reviewer (see additional comments in the public evaluation summary) acknowledges that the fold-enrichment for previously identified MEs is large, but they are concerned that MEs are the main focus of the Results and Discussion and are left wondering what could have been found in the large majority (approx. 90%) of SoC-CpGs that do not overlap known MEs.

Our strategy throughout the Results and Discussion was to focus on characteristics including metastability, parent of origin-specific methylation, histone modifications and gametic and early embryo methylation patterns that suggest a link to establishment of methylation states in the early embryo at SoC-CpGs. For these analyses all SoC-CpGs were considered at every stage and metastability was not the primary focus. However, we do repeatedly point out that many of the contextual genomic features associated with SoC-CpGs have also been associated with metastability which we consider to be worthy of note, in part because it suggests that many SoC-CpGs may in fact be MEs, despite not having been previously identified as such. We have further cause to believe this could be the case because of (i) the typically small sample size of multi-germ layer/multi-tissue datasets used in previous screens for MEs, meaning that published screens for human MEs are underpowered and will hence fail to capture most MEs; and (ii) the evidence that we present suggesting that environmentally-driven inter-individual variation at loci exhibiting ME-like properties may diminish with age, again suggesting that ME screens, which tend to analyse adult tissues, will miss metastable loci present in infancy and early childhood.

We had already made the point (ii) above in the Discussion (lines 383-389. However, given the reviewer’s concerns we have added an additional comment on point (i)) (lines 389-392).

Moreover, the findings should be compared with previously published studies on, e.g., folic acid supplementation and famine (this last point also applies to the Introduction). More on this in point 12 and 13.

We did in fact compare our results with previously published studies looking at both famine and folic acid supplementation that were included in a recent review of DNAm loci associated with maternal nutrient exposures (James et al. IJE 2018; see section titled ‘Overlap of SoC-CpGs with existing studies’). For ease of reference we have included an additional table summarising overlaps with loci highlighted in the James et al. review (ST1k). We have also added an explicit reference to the fact that folate supplementation was considered in this comparison to the main text (line 310) and have also cited existing evidence on links to periconceptional folate and famine in the Introduction (lines 50-51).

Additional points from individual review reports:

Results section:

(4) Page 3, line 94: Have the authors access to longitudinal intra-individual methylation scores during different seasonal windows?

As outlined above, we are pleased to report that we have now generated additional EPIC array data covering a subset of n=138 individuals from the ENID cohort included in the main analysis. This subset had methylation measured in blood at age 5-7yrs enabling us to conduct an investigation of longitudinal methylation changes in these individuals. This analysis strongly supports the evidence of SoC effect attenuation with age suggested by our previous comparison of the independent ENID (2yr) and EMPHASIS (7-9yr) cohorts, with:

(a) Strong correlation of conception date methylation maximum between age 2yrs and 5-7yrs at SoC-CpGs in these 138 individuals (Figures 3A, 4A); and

(b) Evidence of SoC effect size attenuation at the majority of SoC-CpGs (Figure 3B; Wilcoxon signed rank sum p=10^-12).

Note that as described above, this additional longitudinal dataset also has a different confounding structure with respect to biological and technical covariates (Supp Tables 1p-1r) and date of sample collection (Supp. Figure 2), strongly supporting our previous analysis.

Interestingly, a larger number of SoC-CpGs replicate in this smaller and slightly younger longitudinal dataset, suggesting a possible secular or cohort effect. We have added some additional comments on this to the Discussion (lines 400-403) and have highlighted the different years of conception between the two analysed cohorts in Methods (lines 554,563).

(5) Page 4, line 141: Did the authors re-evaluate SoC CpG positions with alternative methods (e.g. pyrosequencing) to analyse larger groups/ cluster of CpG positions and to control the reliability of array data?

We did not perform a technical validation using an alternative DNAm measurement method and unfortunately DNA stocks from ENID 2yr individuals studied in the main analysis are now exhausted. However we consider the strong concordance of our findings, which with the additional dataset now span 3 datasets with different confounding structures, to be a strong technical validation of our findings.

(6) Line 238 onward: from 257 Soc-CpGs to the "full" set of 768 SoC-CpGs for enrichment tests out of "power" reasons ◊ not consistent and confusing. I would stick with your choice for the 257.

Please see our response to point (2) above.

(7) Page 7, line 301: The authors mentioned overlaps with previous studies. It would be helpful, if overlapping or related findings could be summarized (and maybe visualized) in more detail. This would facilitate the comparison of the current data set with previous publications.

As outlined in a previous response above, we have listed regions associated with periconception and gestational exposures covered in the cited review (James et al. IJE, 2018) that overlap the array background analysed in our study in a new Supplementary Table (Supp. Table 1k).

(8) I would strongly recommend reporting the relationship between methylation of these SoC sites and measures of growth or development in these two cohorts – it's not clear to me why this has not been done so, assuming such data is available under approved ethics. In the field there is an ever-growing body of papers linking various early development exposures to differences in methylation in childhood, but there is a dearth of reproduceable evidence for exposures associating with methylation differences that in turn associate with tangible differences in child health/development. Particularly since the manuscript already mentions things like the relationship between methylation and BMI – while adult BMI is different to child measures of body composition, any evidence for a relationship (or a lack of a relationship) with available phenotypes in these two child cohorts would be a logical conclusion that section of analysis and would provide much greater clarity for the relevant Discussion section.

We are currently researching links between several SoC-CpGs and health-related outcomes including measures of growth, and we have prepared/submitted other papers with different groups of authors (e.g. the EMPHASIS team) relating to other phenotypes. We consider a detailed analysis of links between SoC-CpGs and diverse outcome measures in Gambian children to be beyond the scope of the current study and would argue that such an analysis would dilute the central focus of this paper that is already long and complex. In the current paper we do already refer to two existing studies linking Gambian SoC or nutrition-associated CpGs to health outcomes in non-Gambians (child and adult obesity/POMC, Kuhnen et al. Cell Metab 2016; cancer/VTRNA2-1, Silver et al., Gen Biol 2015) in the current manuscript (lines 513-515 and 419-422). The VTRNA2-1 locus does not overlap any SoC-CpGs and we already speculate that this may be due to SoC effect attenuation, since the previous association was observed in younger (3-9mth) infants (lines 443-445). We have additionally referenced a recently published paper linking another SoC-associated locus to thyroid volume and function in Gambian children (Candler et al. Sci Adv 2021; lines 515-522) and highlighted that neither this nor the POMC locus overlap the array background analysed in this study. Finally we have already included an analysis of overlaps between SoC-CpGs and traits in published EWAS and GWAS catalogues (lines 320-347).

(9) Relatedly, if consistent with approved ethics, this paper would benefit from a table summarising the cohort characteristics for each cohort. If appropriate, stratifying the cohorts summaries by season of conception would also be of interest.

As above, we think that inclusion of additional phenotypic characteristics for individuals with DNAm measures covering the two cohorts (and two timepoints) analysed is beyond the scope of this study. We strongly prefer to focus on the robust exposure – methylation associations across cohorts and timepoints identified in this paper.

(10) Rationale for not doing a meta-analysis for power reasons should be explained.

The rationale for not performing a meta-analysis is our observation, now supported by longitudinal data, that SoC effect sizes diminish with age. This means that a combined meta-analysis of all datasets would have reduced power to detect SoC-CpGs present at age 2yr. We have added a note to this effect in Methods (lines 625-628).

(11) The Results section switches from past to present tense and back again. Please keep it to past tense.

If the editor agrees we would prefer to stick to the current style of referring to analyses performed in the past tense (“we investigated…”, “we analysed…”, “we tested…”), with key findings in the present tense (”SoC-CpGs are distributed throughout the genome…”). The manuscript has been read by a large number of authors, colleagues and reviewers and this point has not been raised by anyone except this reviewer.

Discussion section

(12) From my point of view, the observation of an attenuation of DNA methylation levels until mid-childhood is of interest. It would be helpful, if the authors could discuss this point in more detail within the Discussion section. What is known about methylation changes during aging (epigenetic drift etc.) and are these observations relevant for SoC associated CpGs?

We have added some further comment on this in the Discussion (lines 397-399).

(13) The authors might discuss in more detail, why these observations of SoC associated loci are disease relevant.

Apart from our own studies (see response to point 8 above), Gunasakera et al. (Epigenomics 11, 2019) note that stochastic and/or environmentally influenced SIV (systemic interindividual variation) loci identified by our group and others have been associated with a number of diseases including Alzheimer’s, cancer, rheumatoid arthritis and schizophrenia. We have added some further comment on this to the Discussion (lines 520-522).

Methods section

(14) As mentioned in the public review, I think there needs to be more clarity about why you do and do not use the SoC-associated CpGs with difference <4% – in the results it is stated that the full list is being used to 'maximise power', but it is unclear why maximising power is necessary/better for that analysis compared to the other analyses where only CpGs with difference >4% are used? A little more detail on why it is important/fine to use there would be helpful.

We have now removed the ERV1 proximity vs amplitude analysis that included sub-threshold SoC-associated CpGs. See our response to (2) above for further details.

15) Another point that would benefit from additional clarity would be adjusting for cell composition differences in your models – it appears that you have adjusted for these by including the top 6 PCs, but you also have cell count estimates from the Houseman method. Was there are a particular reason for using the PCs when the Houseman estimates were available?

We have performed an additional sensitivity analysis to address this point. See our response to (2) above.

(16) Methods does not read as one coherent story, but fragmented and will require some shuffling to make it easier to read.

We have reviewed Methods and included additional material as outlined in our response to comments above. We hope it is now easier to follow.

(17) Outlier removal is with 25th and 75th percentile very aggressive compared to what is standard in the field.

In fact we used a common, much less aggressive threshold, sometimes referred to as ‘Tukey’s method’, with outliers defined as 3 x the inter-quartile range (IQR) in either direction (see Methods lines 621-624).

(18) You can Check SNPs in 450k dataset for possibility to adjust for genetic background in the discovery cohort and 850k cohort to exclude sample swaps/mixtures etc.

We acknowledge the potential for further refinement of our analysis using genotypes imputed from the methylation data. However we believe such an analysis is highly unlikely to alter the main findings. In particular we note (a) the strong evidence that major genetic PCA clusters are driven by ethnicity (Mandinka vs Fula; Supp. Figure 12), with village serving as a strong proxy for this; and (b) the strong concordance of SoC effects between ENID 2yr and EMPHASIS 7-9yr cohorts in the sensitivity analysis where we adjust for genetic background in the EMPHASIS cohort, and for village as a proxy for ethnicity in the ENID 2yr cohort (main text lines 293-307; Methods lines 782-789; Supp. Figure 13; sensitivity analysis described in ST1s). Finally, again given the strong concordance of SoC effects between the two cohorts described above, we think it highly improbable that sample swaps or mixtures revealed by an analysis of methylation-imputed genotypes in the EMPHASIS cohort would materially alter our main findings.

(19) Reading the methods you are worried by genome-wide effects on DNAm. Why not look at the LINES ALU effects via: Zheng Y Joyce BT Liu L Zhang Z Kibbe WA Zhang W Hou L. Prediction of genome-wide DNA methylation in repetitive elements. Nucleic Acids Res 2017;45:8697-8711 [works for 450k data]

Thank you for this suggestion. We have now analysed potential SoC effects on global methylation at LINE1 and Alu elements using the suggested method published by Zheng et al. This confirmed a small but significant effect of increased methylation in rainy vs dry conceptions at both LINE1 and Alu elements (0.02% and 0.01% respectively, both Wilcoxon p<2.2x10^-16) suggesting a SoC effect on global methylation levels. This contrasts with a significant but extremely small effect in the opposite direction across array background (-5x10^-4%; p<2.2x10^-16). This supports evidence of a similar effect of peri/pre-conceptional folate (a one-carbon metabolite that varies significantly by season in our population; Dominguez-Salas et al. Nature Communications, 2014) on global methylation levels observed in previous studies.

We have updated the section on inflation in Methods to include this additional analysis (lines 647-655) and summarised the findings from the LINE1/ALU analysis in Supp. Figure 15.

(20) Were African allele frequencies used to filter out CpG probes?

Probes in or near SNPs were masked using the Africa-specific (GWD) list provided by Zhou et al., Comprehensive characterization, annotation and innovative use of Infinium DNA methylation BeadChip probes. Nucleic Acids Research 45, e22 (2017).

(21) Was imputation done with AGVP?

As outlined in Methods (lines 736-738), imputation was performed using 1000 Genomes phase 3 data as a reference.

(22) Was genetic information used to exclude family relationships, mixtures and sample duplications? [knowing the fluidity of family structures and logistics of sample collection in the Gambia]

See response to (18) above regarding mixtures/duplications. Regarding family relationships, there is indeed a relatively high frequency of half-sibs in our study population. We were unable to test this directly in the discovery (ENID 2yr) cohort as genetic data were not available. However, of the 199 ENID children with paternal ID data we identified 2 pairs of half-sibs each with a common father, corresponding to 2% of participants. This number is too small to influence our findings. We have added a note on this in the Discussion (lines 534-537).

Finally, please note that the age range of the EMPHASIS cohort was previously incorrectly written as 8-9yrs throughout the manuscript. This has been changed to 7-9yrs to reflect the fact that 6 out of the 233 individuals in this cohort were aged 7.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Reviewer #4 (Recommendations for the authors):

Thank you for the opportunity to review the rebuttal to the article titled "Environmentally sensitive hotspots in the methylome of the early human embryo". I would like to say that reading through the rebuttal it appears the authors have been reactive and enthusiastic in their responses to the reviewers. Overall, it appears the changes introduced have help interpretation of the findings as far as I can tell.

In terms of the concerns reviewers have regarding batch effects, on the whole it appears that these have been these addressed as far as is possible within the confines of this particular experiment. The sensitivity analyses reported in table 1s suggest that controlling for the suspected technical confounders (e.g plate, slide, cell proportion estimates) does not significantly attenuate the association between SoC-CpG and doc.

As noted, there does appear to be non-randomization with respect to Plate (and by extension, Slide) in the youngest ENID sample, which I suspect is due to assignment of samples to plates as they were collected (i.e in chronological order). What I cannot tell is whether the methylation arrays are also processed as samples are received (i.e. once a plate is filled), or all together at the end of the collection phase. If it is the former, that is a likely source of some variation between plates. Given the use of this cohort as the discovery sample, it would be useful if the authors could declare whether or not this is the case in the methods.

We can confirm that the samples were processed in two batches. Samples were not processed sequentially by date of collection and each processing batch covered conception dates throughout the year. We have added this information to Methods (lines 578-80):

“Samples from the ENID 2yr dataset were processed in two batches with each batch covering conception dates throughout the year. Samples from the ENID (5-7yr) and EMHASIS (7-9yr) datasets were processed in single batches.”

As the reviewer notes we previously took additional steps to check for the possibility of residual confounding due to batch/technical effects including carrying out sensitivity analyses with batch effects included directly in the model. We have additionally checked and confirmed that adding processing batch to these models had no significant effect on estimated Fourier regression coefficients in the main SoC association analysis.

Importantly, as previously noted we also observed a tight correspondence of date of methylation maximum across all three datasets considered (Figures 2C, 3A and 4A). Two of these datasets (longitudinal ENID dataset with DNAm measured at 5-7yrs and EMPHASIS dataset measured at 7-9yrs) had very different confounding structures with no associations between date of conception / birth / collection and batch/technical variables.

I am however suspicious that there is an alternative confounding factor that could have implications for the interpretation of the results within the paper, and that is the notion of DNA methylation data reliability (or lack of). Multiple studies (e.g. Logue et al. (2017); Epigenomics 9(11):1363-1371; Sugden et al. Patterns 1(2): 100014) have demonstrated that a large proportion of the CpG sites shared between the 450K and EPIC array (as used in this paper) are unreliable, so that it is unlikely they yield the same value when the same sample is measured twice. This of course has implications for determining 'significant associations', for assessing DNA methylation changes over time, and for replication. Because of what we know about reliability and how it manifests, the authors should address reliability in the context of these findings.

We agree that probe reliability is an important consideration, although we note that this is likely to increase the number of false negatives, since low reliability probes would be unlikely to be picked up in our discovery analysis, and methylation maxima would be unlikely to align across datasets.

Reassuringly for this paper, the SoC-CpGs appear to be pretty robust – cross-referencing the subset of 259 CpGs against the list of reliabilities from Sugden et al., shows the mean ICC is over.7 (close to 'excellent') – this is not the case for the 509 CpGs not taken forward, who have a mean reliability of less than 0.3 ('poor'). The list of CpGs selected as matched and array background controls does not appear available, but I strongly suspect the array background CpGs (at the very least) will have a much lower reliability than the SoC-CpGs (just due to random chance). This clouds the interpretation of any tests that compare these sets of CpGs and should be taken into consideration.

We confirm that the probe reliability as measured by Sugden et al. is ‘excellent’ for SoC-CpGs (median ICC = 0.76). The same measure for distribution-matched control CpGs is 0.68 (‘good’), providing reassurance that analysed DNAm values for both sets of probes are reliable. This aligns with observations by Xu and Taylor (Epigenetics, 2021 May;16(5):495-502) that low ICC scores are associated with sites with low variability.

We agree that array background and random control probes will have much lower reliability, as assessed by Sugden et al. In almost all cases for the present study, analyses that include random controls or array background also included a comparison with matched controls. In these cases we think it is still useful to include random controls and/or array background as it gives the reader useful information on how our analyses relate to properties of the 450k array. To take two examples, it is instructive to see 450k DNAm distributions in the embryonic inner cell mass and embryonic liver (Figure 4A); and also to see how 450k probes are distributed with respect to ERVs and ZFP57 binding sites (Figure 6A). There is one example where SoC-CpGs are compared with array background only (Figure 4D). Again we think the comparison is justified as it highlights array-wide associations between sperm and oocyte hypo/hyper methylation and hypo/hyper methylation in blood, rather than making any detailed quantitative comparisons.

We have added the following to Results (lines 184-186):

“SoC-CpG reliability is classified as ‘excellent’ (median ICC=0.76) using probe reliability estimates from a recent repeated measures study²³. Reliability of matched control CpGs is classified as ‘good’ (median ICC=0.68) using the same method.”

We have also added a comment on the issue of general unreliability of probes overlapping the 450k and EPIC arrays to the Discussion, including the additional point on intermediate methylation and probe reliability raised by the reviewer below (lines 535-538):

“Measured DNAm values at SoC-CpGs and matched controls have previously been reported to be reliable, in strong contrast to the majority of probes overlapping the 450k and EPIC arrays that have been found to have low test-retest reliability²³. This likely reflects increased inter-individual variability and/or intermediate DNAm at SoC-CpGs and controls^75,76.”

23: Sugden et al. Patterns (2020); 75: Logue et al. Epigenomics (2017); 76: Xu and Taylor Epigenetics (2021).

Further, differential reliability should be considered as an alternative explanation for the reduction in mean SoC amplitudes over time – could it be that measurements made using 450K arrays (ENID 2yr) are different to those using EPIC arrays for reasons beyond age-specific effects (i.e. measurement error)?

We think this is unlikely given the high reliability of these probes discussed above, and given the strong concordance between DNAm maxima further suggesting consistency between arrays.

Finally, the description of the properties of SoC-CpGs (lines 173 onwards) notes that they are more likely to be intermediately methylated – this is again a property of reliable CpGs since arrays do not have the resolution to measure hyper- and hypo- methylated sites very well, leading to unreliable measurements. In all, these points suggest that reliability of DNA methylation is important for interpretation of these results, and, given the high reliability of the SoC-CpGs, is actually a positive feature.

Thank you. We agree that increased reliability of intermediately methylated probes is reassuring. We have added the point about increased reliability of intermediately methylated probes – see added text above.

The reviewers were concerned that a large proportion of the discussion was dedicated to MEs and how these findings are related. I believe the authors frame this appropriately, and do not find the references to them excessive. MEs are of great interest, and using cheaper and easier methods to uncover them (such as here) will be beneficial.

We thank the reviewer for their comment which we certainly agree with.

Author details

1. Matt J Silver

   Medical Research Council Unit The Gambia at the London School of Hygiene and Tropical Medicine, Gambia, United Kingdom

   Contribution

   Conceptualization, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Software, Supervision, Visualization, Writing - original draft, Writing – review and editing

   For correspondence

   matt.silver@lshtm.ac.uk

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-3852-9677

2. Ayden Saffari

   Medical Research Council Unit The Gambia at the London School of Hygiene and Tropical Medicine, Gambia, United Kingdom

   Contribution

   Formal analysis, Methodology, Software, Writing – review and editing

   Competing interests

   No competing interests declared

3. Noah J Kessler

   Department of Genetics, University of Cambridge, Cambridge, United Kingdom

   Contribution

   Formal analysis, Methodology, Software, Visualization, Writing – review and editing

   Competing interests

   No competing interests declared

4. Gririraj R Chandak

   Genomic Research on Complex Diseases, CSIR-Centre for Cellular and Molecular Biology, Hyderabad, India

   Contribution

   Funding acquisition, Project administration, Supervision, Writing – review and editing

   Competing interests

   No competing interests declared

5. Caroline HD Fall

   MRC Lifecourse Epidemiology Unit, University of Southampton, Southampton General Hospital, Southampton, United Kingdom

   Contribution

   Funding acquisition, Project administration, Writing – review and editing

   Competing interests

   No competing interests declared

6. Prachand Issarapu

   Genomic Research on Complex Diseases, CSIR-Centre for Cellular and Molecular Biology, Hyderabad, India

   Contribution

   Investigation, Writing – review and editing

   Competing interests

   No competing interests declared

7. Akshay Dedaniya

   Genomic Research on Complex Diseases, CSIR-Centre for Cellular and Molecular Biology, Hyderabad, India

   Contribution

   Investigation, Writing – review and editing

   Competing interests

   No competing interests declared

8. Modupeh Betts

   Medical Research Council Unit The Gambia at the London School of Hygiene and Tropical Medicine, Gambia, United Kingdom

   Contribution

   Data curation, Writing – review and editing

   Competing interests

   No competing interests declared

9. Sophie E Moore

   1. Medical Research Council Unit The Gambia at the London School of Hygiene and Tropical Medicine, Gambia, United Kingdom
   2. Department of Women and Children's Health, King's College London, London, United Kingdom

   Contribution

   Data curation, Funding acquisition, Project administration, Writing – review and editing

   Competing interests

   No competing interests declared

10. Michael N Routledge

    1. School of Medicine, University of Leeds, Leeds, United Kingdom
    2. School of Food and Biological Engineering, Jiangsu University, Zhenjiang, China

    Contribution

    Data curation, Writing – review and editing

    Competing interests

    No competing interests declared

11. Zdenko Herceg

    Epigenomics and Mechanisms Branch, International Agency For Research On Cancer, Lyon, France

    Contribution

    Funding acquisition, Writing – review and editing

    Competing interests

    No competing interests declared

12. Cyrille Cuenin

    Epigenomics and Mechanisms Branch, International Agency For Research On Cancer, Lyon, France

    Contribution

    Data curation, Writing – review and editing

    Competing interests

    No competing interests declared

13. Maria Derakhshan

    Medical Research Council Unit The Gambia at the London School of Hygiene and Tropical Medicine, Gambia, United Kingdom

    Contribution

    Formal analysis, Methodology, Writing – review and editing

    Competing interests

    No competing interests declared

14. Philip T James

    Medical Research Council Unit The Gambia at the London School of Hygiene and Tropical Medicine, Gambia, United Kingdom

    Contribution

    Formal analysis, Writing – review and editing

    Competing interests

    No competing interests declared

15. David Monk

    1. Biomedical Research Centre, University of East Anglia, Norwich, United Kingdom
    2. Bellvitge Institute for Biomedical Research, Barcelona, Spain

    Contribution

    Data curation, Investigation, Writing – review and editing

    Competing interests

    No competing interests declared

16. Andrew M Prentice

    Medical Research Council Unit The Gambia at the London School of Hygiene and Tropical Medicine, Gambia, United Kingdom

    Contribution

    Conceptualization, Funding acquisition, Investigation, Writing – review and editing

    Competing interests

    No competing interests declared

• 1,041

  Page views

• 193

  Downloads

• 6

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.