Altered sperm DNA methylation provides an epigenetic route from overweight and older men to intergenerational disease transmission

Significance

When it comes to the health of a pregnancy and the child, the mother has long been the focus. However, an accumulation of evidence from animal models indicates that a father’s health and exposures are equally important. For humans, it remains unknown how sperm carries information on a father’s health and environment to influence embryo development. It is thought that such information may be transmitted through heritable information carried by sperm, including epigenetic modifications established during spermatogenesis and transmitted at fertilization. Epigenetic mechanisms control how genes are used. Here, we showed that the epigenetic information in sperm was influenced by paternal health and age and that these changes may alter embryonic gene expression, placenta development, and child health.

Abstract

The occurrence of childhood neurodevelopmental disorders has been steadily increasing for decades, yet we have little understanding of modes of inheritance implicated in these diseases. Epidemiology studies have revealed an association between an elevated paternal Body Mass Index (BMI) and age as factors implicated in increased risk for neurodevelopmental disorders in children. Epigenetic regulation of gene expression by DNA methylation at cytosines is susceptible to environmental, diet, and lifestyle changes. Yet, whether a man’s BMI and age influence the sperm methylome and potentially impact offspring development remains unresolved. Using MethylC capture sequencing, we identified over 38,000 differentially methylated CpGs in the sperm of men with an elevated BMI, with overlap occurring in regions associated with advanced paternal age. These were enriched for neural gene and disease ontologies. Differentially methylated regions localized to transposable elements that can act as active enhancers during human zygotic genome activation, and at gene promoters for early lineage specification. These results suggest that sperm methylome alterations that are linked to an elevated BMI may influence key embryonic transcriptional processes, notably those associated with trophectoderm and placental development.

Epigenetic modifications are dynamic gene regulators that are shaped by environmental influences (1). DNA methylation, which consists of cytosine methylation in CpG dinucleotides, is associated with gene repression, playing a pivotal role in regulating gametogenesis and early embryonic development. In mice, DNA methylation is a central regulator of spermatogenesis, male fertility, and early embryonic development. Namely, DNA-methylated imprinted genes, characterized by their monoallelic and parent-of-origin expression, have an extraordinary ability to resist the epigenetic reprogramming that occurs postfertilization in the early mammalian embryo (2). These genes have an unequal effect on embryonic development, with the paternal genome promoting the formation of extraembryonic tissues (3, 4). Along with the establishment of paternally methylated regions, DNA methylation at discrete regions serves essential roles in spermatogenesis, including silencing of transposons or regulation of gene expression (5–7). Moreover, the transmission of the paternal epigenome at fertilization is important for zygotic genome activation (8, 9). It is therefore plausible that errors in paternally inherited DNA methylation could alter embryo transcription programs and lead to higher risk for disease phenotypes in children.

Indeed, abnormalities in the establishment of DNA methylation in the developing germline have been associated with epigenetic inheritance between generations, mostly through imprinted genes. For example, Prader–Willi syndrome, a complex disorder characterized by hyperphagia and problems in neurodevelopment can be attributed to abnormal imprinting near the paternally expressed gene Small Nuclear Ribonucleoprotein Polypeptide (SNRPN) (10). Similarly, differential sperm DNA methylation in the SNORD gene family found within the Prader–Willi gene cluster is associated with an increased risk of autism in children (11). In fact, and despite high variabilities across continents and regions, a global rise in the rates of autism can be observed worldwide (i.e., 63.7% increase in 2016 vs. 2008 in the United States) (12). Notably, recent work estimates that the prevalence of autism in eight years olds across the United States is 1 in 36 (13). These staggering increases also coincide with rising rates of paternal obesity and the societal trend of delaying parenthood. Notably, paternal obesity increases the risk of having a child with autism (odds ratio of 1.73) (14), as does advanced paternal age (55% increased risk) (15, 16). Taken together, these findings suggest that advanced paternal age, metabolic health, and being overweight may be implicated in the rising rates of neurodevelopmental disorders in children. How these paternal factors affect the sperm methylome and in turn the next generation remains poorly understood (17).

Epidemiological studies support the notion of paternal epigenetic inheritance. However, there is a lack of studies using sperm and health data from men, that have associated a common paternal health factor, being overweight, to changes in the sperm methylome that could influence gene regulation in the early embryo with long-term consequences for child health. Recently we demonstrated that the sperm DNA methylome is perturbed in association with altered folate metabolism or when exposed to the environmental toxicant DDT (18, 19). Another recent study showed that paternal exposure to air pollution also alters the sperm methylome and is associated with lower birth weight (20). Little is known about the mechanisms by which the paternal sperm methylome could influence embryonic gene expression. Largely due to limitations in obtaining paired sperm-embryo samples. One route of transmission may involve transposable elements. Indeed, DNA methylation is a well-studied silencer of transposable element activity in the germline, which directly affects fertility and early human development (21). Namely, the silencing of LTR-transposable elements in early human development is critical for zygotic genome activation (22). These elements are particularly interesting candidates for bridging generations epigenetically; some are known to evade the global DNA methylation erasure that rewires the methylome of the embryo in its first days of development (23, 24). Moreover, they constitute central components of gene regulatory networks as they bear diverse transcription factor binding motifs, with some elements having evolved specific regulatory functions for the emergence of cell lineage in humans (25–27).

To date, studies of the mammalian sperm methylome have mostly focused on gene promoters and imprinted regions, overlooking the dynamic and underexplored intergenic regions that comprise a significant portion of the genome. Addressing these gaps in knowledge is essential toward uncovering the mechanisms underlying the effects of paternal health on offspring development.

To investigate the association between paternal BMI and sperm DNA methylation, we analyzed sperm from men with BMIs ranging from normal to overweight and obese using MethylC Capture (MCC)-seq, a high-resolution method targeting over three million CpG sites, including environmentally sensitive regions (18). Specifically, this panel interrogates over three million CpGs in the genome, including the > 850,000 sites present on the Infinium MethylationEPIC BeadChip, an array-based technique targeting commonly assessed gene promoter, CpG island regions, and enhancers, widely used in DNA methylation analyses. In addition, the human sperm MethylC-Capture sequencing (MCC-seq) panel targets approximately one million CpGs with “dynamic” levels of DNA methylation. These are environmentally sensitive sequences which demonstrate intermediate DNA methylation levels (typically between 20 to 80%) mostly found at intergenic regions. Thus, using this approach, we gained insight into underexplored regions of the sperm methylome, while retaining our ability to compare our epigenetic signature with previously published datasets. In this study, our analysis identified over 38,000 differentially methylated cytosines (DMCs) in sperm from men with elevated BMI, that showed a trend for hypermethylation. These DMCs were significantly enriched in intergenic regions, highlighting the importance of these often-overlooked regions. Gene ontology analysis revealed that many of these regions were associated with neural development, suggesting potential impacts on offspring brain function and developmental processes.

Methods

Detailed experimental procedures, analytical parameters, quality-control metrics, statistics, software settings, and visualization packages used are described in SI Appendix.

Ethical Statements.

This study was approved by the Faculty of Medicine Institutional Review Board (McGill University), the Health Sciences Research Ethics Board (A09-M57-15B), the University of Toronto Research Ethics Board (IRB No. 30089), and the Health Canada–Public Health Agency of Canada Research Ethics Board (REB2019-0006). All participants provided written informed consent prior to enrollment through the CReATe Fertility Centre Biobank protocol. Consent included the use of excess biological materials and associated deidentified clinical data for research purposes. Additional details regarding ethical approval, participant consent, and data handling are provided in SI Appendix.

Participants and Sample Collection.

In this prospective study, 71 men were recruited at the CReATe Fertility Centre (Toronto, Canada) between January 2016 and December 2017. Semen and blood samples were collected on the same day following a minimum abstinence period of 2 d. Motile spermatozoa were isolated by density gradient centrifugation to remove debris, dead cells, and immature germ cells. Participants reporting tobacco use, recreational drug use, steroid exposure, or carrying the MTHFR 677C→T polymorphism were excluded (28).

Fifty-two samples were selected for downstream analyses. Participants were classified according to WHO BMI criteria as normal weight, overweight, or obese (29, 30). Clinical, fertility, and biochemical characteristics are summarized in SI Appendix, Table S1.

Analysis of Plasma and Serum Samples.

Biochemical analyses included serum folate, red blood cell folate, vitamin B12, homocysteine, 25-hydroxyvitamin D, and plasma C-reactive protein (CRP). Measurements were performed using validated clinical laboratory assays and standardized quality-control procedures (31, 32). CRP concentrations were determined by ELISA. These measurements were used as biological covariates in downstream methylation analyses.

Sperm DNA Isolation.

Genomic DNA was isolated from purified sperm samples as previously described (19). DNA quality and purity were assessed by bisulfite pyrosequencing of imprinted loci to exclude somatic cell contamination (28). (SI Appendix, Fig. S1 and Table S2).

Sperm DNA Methylation Capture (MCC-seq).

Genome-wide sperm DNA methylation was profiled using our previously developed human sperm MCC-seq (18, 19, 33). Captured libraries were sequenced on an Illumina NovaSeq 6000 platform.

MCC-seq Data Preprocessing.

Sequencing data were processed using the GenPipes methylation workflow (34). CpG sites overlapping known SNPs, blacklisted genomic regions, or failing coverage thresholds were excluded from downstream analyses (35–39). To assess genetic variation within the cohort, variants were inferred directly from MCC-seq data using Bis-SNP (40).

Differentially Methylated CpG (DMC) and Region (DMR) Identification.

Associations between sperm DNA methylation and BMI were evaluated using generalized linear regression models adjusted for age, serum folate, red blood cell folate, vitamin B12, homocysteine, and 25-hydroxyvitamin D. Statistical significance was assessed using false discovery rate correction (41). Significant CpGs were defined as differentially methylated cytosines (DMCs, q-values ≤ 0.01), and consecutive DMCs (at least 3) spanning 500 bp were merged to define DMRs. Additional analyses examined the relationship between sperm DNA methylation, circulating CRP levels, and paternal age using the same statistical framework.

Enrichment and Gene Ontology Analyses.

Enrichment analyses were performed using permutation-based approaches to determine whether DMCs and DMRs were overrepresented in specific genomic features or transposable element families (39). Gene and disease ontology analyses were performed using ClusterProfiler with multiple-testing correction (42).

Integration of Publicly Available Datasets.

To investigate the potential developmental relevance of BMI-associated sperm methylation changes, DMRs were integrated with publicly available human sperm epigenomic datasets, embryonic chromatin accessibility datasets, enhancer annotations, and transcriptomic datasets spanning early embryonic development (33, 43–47).

Transposable Element Insertion Age Analysis.

To investigate the relationship between DNA methylation changes and transposable element evolution, BMI-associated DMCs overlapping transposable elements were analyzed according to RepeatMasker divergence scores as a proxy for insertion age. Transcription factor motif enrichment analyses were performed on DMRs overlapping transposable elements, embryonic enhancers, and gene promoters using the monaLisa framework and motifs from the JASPAR database (48–50).

Results

Men with an Elevated BMI Showed Differential Sperm Methylation.

Extensive epigenetic remodeling occurs in germ cell development in utero and continues in adulthood throughout spermatogenesis (7, 51). Notably, an altered sperm methylome is also implicated in male infertility (52). To investigate whether there is an association between a man’s BMI and sperm DNA methylation, we collected sperm from men with normal semen parameters according to the WHO (29) as controls, and a range in BMI from overweight to obese (25 ≤ BMI and above). Clinical characteristics are detailed in Fig. 1 A and B. We integrated serum parameters for each individual to control for potential covariates. Notably, RBC folate range was adequate (661.50 to 1,879.06 nmol/L, mean: 1,182.54 nmol/L), along with serum folate (mean: 42.62 nmol/L), serum B12 (mean: 384.89 pmol/L), homocysteine (mean: 11.19 µmol/L), and 25-OHD (mean: 54.578 nmol/L). These methods indicate a mean BMI of 27.3 ± 4.5 kg/m² (Fig. 1A). While the average age of the patients is 37.3 y old, we observed no significant differences in the ages of individuals across different BMI groups, except for a slightly younger age in obese patients compared to overweight patients (P-value = 0.03, Fig. 1B). All parameters previously mentioned can be found in SI Appendix, Table S1.

Fig. 1.

BMI contributes to variability observed in sperm methylome. (A) Summary data of the study cohort. RBC Folate (red blood cell folate). 25OHD (25-hydroxy Vitamin D). All participants presented normal semen parameters according to WHO standards. (B) Barplot of BMI (Top) and Age (Bottom) repartition across three groups of participants. Grouping was made according to BMI; Obese (OB) ≥ 35, Overweight (OW) 25 ≤ BMI ≤ 29.9 and Normal (NOR) BMI ≤ 24.9. Data displayed are mean with SD; statistical test is ANOVA with Kruskall–Wallis. (C) Principal component analysis showing contribution to variability of BMI and Age (Top) and Genotype (Bottom) in top 1% differentially methylated cytosines.

To investigate the sperm DNA methylation landscape, we applied a customized MethylC-Capture followed by sequencing (18, 19) (SI Appendix, Fig. S1). Overall, the average of read coverage of CpGs per sample is 30×. After filtering CpGs with low read coverage (≤20×), a total number of 2,404,135 CpGs covered by at least one sample were retained. To further remove CpGs with low sample coverage, we obtained 1,835,671 CpGs with 20× coverage and covered by more than 30 samples (i.e., 75.35% of total CpGs were retained after removing low sample coverage CpGs). Furthermore, 1,937 CpGs showed nonvariable methylation (0.08% of total CpGs, or 0.11% after removing low sample coverage CpGs) (SI Appendix, Fig. S1 A–C) (18).

First, epigenetic variation in global DNA methylation was assessed by principal component analysis (PCA) to determine whether samples would be discriminated in relation to age, BMI, and genotype. While the first two principal components accounted for 17.59% and 10.25% of the variance (PC1 and PC2, respectively) in our dataset, there was no clustering by BMI or age (Fig. 1C). Meanwhile, using inferred genotype data, our genetic PCA analysis revealed minimal variance among individuals (Fig. 1C), indicating that there are no substantial genetic differences in our cohort and that global genetic effects are unlikely to drive the observed DNA methylation dynamics.

To identify whether there were DMCs in sperm that were associated with increased BMI, a quantitative association between sperm DNA methylation and BMI was established using a generalized linear regression model adjusted for potential covariates (age, folate, B12, homocysteine, and 25-OHD). Methylation gains and losses at DMCs were defined as a CpG showing a significant (q-value < 0.01) increase or decrease in DNA methylation level associated with BMI. We detected 38,719 DMCs, with 88% gaining methylation (34,234 DMCs) and 12% losing methylation (4,485 DMCs) (Fig. 2A). A genome-wide volcano plot summarizing effect size against statistical significance revealed that a subset of CpGs exhibits moderate to large methylation shifts with high significance, whereas most CpGs cluster near zero effect size (SI Appendix, Fig. S2 A–C and Table S3). We examined the genome-wide distribution of effect sizes across all CpGs to further evaluate the robustness of our findings relative to technical and biological variability. BMI-associated DMCs exhibited significantly larger absolute effect sizes compared with background non-DMCs (SI Appendix, Fig. S2D), indicating that detected associations were not driven by stochastic variation. We also calculated the signal-to-noise ratio (SNR), as defined by Czanner et al. (53), across all tested CpGs and found that BMI-associated DMCs showed consistently higher SNR values than non-DMCs (SI Appendix, Fig. S2E). Together, these analyses demonstrate that the magnitude of BMI-associated methylation changes exceed technical noise and random measurement variability, supporting the biological relevance of the observed signals.

Fig. 2.

BMI mostly associates with hypermethylated DMCs at intergenic regions in the sperm and with neural gene ontologies and disorders. (A) Manhattan plot showing the repartition of DNA methylation (DNAme) gains and losses in sperm BMI-DMC (n = 38,719) alongside chromosomes, according to their P-value (−log10). BMI-associated DMCs were defined using the General Linearized Model, correcting for all covariates displayed in Fig. 1, and showing q-value < 0.01. Arrows associate genes bearing most significant hypo- or hyper-DMCs. (B) Barplot of Zscore enrichment of BMI-DMCs (n = 38,719) on key genomic features over background. Scores were obtained using n = 1,000 random permutation t test using R package RegioneR. (C) Barplot of proportions of BMI-DMCs (n = 38,719) on key genomic features shown in percentage. (D and E) Bubble plot of Gene/Disease ontologies enrichment analysis (R package ClusterProfiler) for BMI-DMRs located at gene promoters, introns, or exons. Statistics are t test for enrichment over background with BH correction. DMRs were defined using 500 bp tile with 3 or more DMCs.

Analysis of CpG enrichment over the MCC-seq panel background revealed DMCs were primarily enriched within intergenic regions and displayed intermediate levels of DNA methylation (Z-score > 100 and DNA methylation level is, in average, 45% ± 6%). Some DMCs occurred in “genic context” regions (13.5% on exons and around 5.8% considering both 5′UTR and 3′UTR) (Fig. 2 B and C and SI Appendix, Fig. S3A). Genes of interest gaining methylation include Myelin Transcription Factor 1 Like (MYT1L), that has neuroendocrine functions and has been implicated in syndromic obesity, and neurodevelopmental disorders including autism spectrum disorder and schizophrenia (54–56) (Fig. 2A). Likewise, a member of the neuronal specific catenin family, CTNNA2, involved in complex cortical dysplasia, showed a loss of methylation in sperm from men with elevated BMI (57). While the density analysis suggested there might be an accumulation of DMCs at telomeric and subtelomeric regions, a hot-spot analysis revealed no clear telomeric regional bias (Fig. 2A and SI Appendix, Fig. S3B). Likewise, probing specific transposable element families that could accumulate DMCs at these regions revealed no clear overrepresentation (SI Appendix, Fig. S3C). Overall, our MCC-seq analysis reveals that men with an elevated BMI demonstrate a trend for sperm DNA methylation gains genome-wide, with DMCs mostly found close to genes with known association to neural disorders.

To assess the reproducibility of our findings, we compared BMI-associated sperm DNA methylation changes identified in this study with those reported by Donkin et al. (58), where sperm DNA methylation was profiled using reduced representation bisulfite Seq (RRBS) and compared between obese and lean men. Restricting the analysis to CpG sites measured in both datasets (99,207 CpG sites), we identified 1,449 BMI-associated DMCs in our cohort and 4,715 DMCs in the Donkin dataset. Of these, 281 CpGs overlapped between studies, representing approximately a fourfold enrichment over random expectation (Fisher’s exact test, P = 2.2 × 10⁻⁹¹; odds ratio = 4.08). Notably, this level of overlap is observed despite differences in cohort composition, analytical frameworks, and methylation profiling approaches, supporting the reproducibility of BMI-associated epigenetic signatures across independent human studies. Functional enrichment analysis of genes associated with the overlapping DMCs revealed enrichment for biological processes related to membrane potential regulation, cardiac development, and glutamatergic neuron differentiation—pathways consistent with previously reported observations and supporting the biological relevance of these epigenetic associations.

Elevated BMI Associated with DMCs Found at Neural Gene and Disease Ontologies.

To probe how differential DNA methylation may impact genome functionality, we merged differentially methylated CpGs into differentially methylated regions (DMRs). DMRs are more likely to exert regulatory functions than single DMCs. We observed over 3,828 sperm BMI-DMRs (SI Appendix, Table S3) among which, we found 1,373 unique genes bearing sperm BMI-DMRs (Fig. 2 D and E and SI Appendix, Table S4). Using ClusterProfiler, we revealed the gene and disease ontologies most enriched at these DMRs were associated with neural development including autism spectrum disorder (Fig. 2 D and E and SI Appendix, Table S5). This observation resonates with epidemiological studies that showed a higher risk for children of fathers with and elevated BMI and neural disorders (14, 59), although the underlying mechanisms and whether this involves epigenetic inheritance remains unclear.

Although epigenetic modifications in sperm are frequently studied alone, there are suggestions of functional interaction between DNA methylation and chromatin marks (notably histone modifications) in the establishment and role of the sperm epigenome across generations. In general, promoters with high CpG content contain higher levels of nucleosomes than CpG-poor intergenic and intronic space and are generally hypomethylated in sperm (60, 61). However, breaking the general misconception DNA methylation and methylation at lysine 4 of histone 3 (H3K4ME) are mutually exclusive in sperm, over 24,000 H3K4me3 peaks overlap regions marked by intermediate (between 20% and 80%) and high (over 80%) DNA methylation (43) at developmental loci. Indeed, among our 3,818 BMI-DMRs in sperm, we observed 594 DMRs overlapping with a significant H3K4me3 peaks (log2 enrichment >3) (SI Appendix, Fig. S3D), most of them being significantly enriched at intronic (60%) and genic regions (13% at 1 to 5 kb from TSS and exons) (SI Appendix, Fig. S3E and Table S6). Given the antagonism between DNA methylation and H3K4me3, and the strong association of the latter with active regulation of gene expression, this suggests that DMRs in the sperm of men with increased BMI could associate with gene misregulation in spermatogenesis, or even embryonic development.

BMI Impacts Methylation at Young LINE and LTR Elements in Human Sperm.

Transposable elements (TE) are drivers of genome organization and act as regulators of gene expression. In spermatogenesis, the disruption of DNA methylation at TE occurs in male infertility (21, 52). In addition, the expression of specific transposable elements is required for early embryo development at zygotic genome activation (ZGA) (22, 47), and lineage specification (26, 62). Interestingly, a significant proportion of BMI-DMCs (n = 5,920) were overrepresented compared to the MCC-seq background at retrotransposons (LINEs and LTR elements like ERVL, ERVL-MaLR, and ERVK) (z-score > 45; P-value < 1 × 10⁻³, n = 1,000 permutations) (Fig. 3 A and B and SI Appendix, Table S7), and these tended to gain methylation in individuals with increased BMI (Fig. 3C). Given that the insertion age of a TE is associated with its activity, our next step was to query TE insertion age in relation to changes in DNA methylation associated with increased BMI. To do so, we applied a linear model to associate significant changes in DNA methylation due to BMI with transposable element insertion age (SI Appendix, Fig. S4 and Table S8). We checked at the level of TE families then subfamilies. With the exception of ERVK and centromeric elements, no families appeared to be more likely enriched for BMI-DMCs compared to other families (Fig. 3D). However, some subfamilies demonstrated a strong correlation between their insertion age and the degree of DNA methylation variation associated with BMI. Notably, these included younger hominoid-specific subfamilies like HERVK elements (HERVK13-int), L1HS, L1PA14, or L1M subfamilies, as well as Alu elements (AluJr4, AluSc, AluY) (Fig. 3E and SI Appendix, Fig. S5A). This suggests that cytosines at young elements are more likely to be mismethylated in men with increased BMI. In the human genome, it is estimated that on average, 20% of all binding events of a transcription factor fall within a transposable element (63). Young LINE1 elements (L1), HERVK, and Alu elements have been identified as potential long-range regulators of gene expression in human cells, playing a key role in shaping early development by influencing regulatory networks (62, 64). Given the known role of TEs as platforms for transcription factor binding motifs, and their critical role in early development, we examined transcription factor binding motifs found at DMRs located in LINE (n = 504) and LTR (n = 288) elements. Many DMRs found at these elements were associated with known regulators of early development (HOX and ONECUT factors, ZFP42, ZIC4, KLF14, ETV5). We also noticed an association with brain developmental factors (ARNT, GSX1) [P-value < 0.05 and log10(enrichment) >2] (SI Appendix, Fig. S5B and Table S9).

Fig. 3.

BMI may influence methylation at young retrotransposons. (A) Barplot of Zscore enrichment of BMI-DMCs on TE Families over background (n = 5,920). Scores were obtained using n = 1,000 random permutation t test using R package RegioneR. (B) Barplot of proportions of BMI-DMCs on TE Families in percentage (n = 5,920). (C) Histogram of DNA methylation changes in OB. OW vs. NOR individuals at DMCs present on Transposable elements. (D) Matrix comparison of TE families ordered by their average age comparing their association with BMI-DMCs changes. Pairwise comparisons were made using the Wilcoxon rank test. (E) Bubble plot showing the transposable elements with most significant changes in BMI-DMC relative to the element age. Pearson R coefficient was calculated using the LM model.

Next, we sought to determine whether BMI-associated sperm DNA methylation changes were more pronounced in individuals with evidence of higher systemic inflammation. To address this, we stratified the 52 men into two equal groups based on circulating C-reactive protein (CRP) levels, using a threshold of 1.3 µg/mL to generate a high-CRP group (>1.3 µg/mL; n = 26) and a low-CRP group (<1.3 µg/mL; n = 26) (SI Appendix, Table S1). We first observed that individuals in the high-CRP group showed a broader BMI distribution than those in the low-CRP group (SI Appendix, Fig. S5C). We then repeated the BMI association analysis in the high-CRP subgroup only. While the overall direction of BMI-associated methylation changes remained highly consistent with that observed in the full cohort (Pearson Correlation Coefficient r = 0.81), effect sizes were globally increased in the high-CRP model (SI Appendix, Fig. S5D).

To further assess how systemic inflammation influences sperm DNA methylation, we included CRP as a continuous covariate and performed additional association analysis to identify CpGs linked to inflammatory status. This analysis identified 59,588 CRP-associated CpGs forming 5,955 DMRs, of which 10,295 DMCs and 1,409 DMRs overlapped with BMI-related loci. In agreement with the main analysis, genes associated with DMRs identified in the CRP model remained enriched for neural developmental and synaptic ontologies (SI Appendix, Fig. S5E and Table S5). Together, these findings suggest that BMI-associated sperm methylation signatures are strengthened in individuals with elevated inflammatory status, while preserving the same broad functional associations identified in the full cohort.

Given that both BMI and age are associated with increased inflammation, a known driver of DNA methylation alterations (65–68), we explored whether there was a connection between BMI, age, and altered sperm methylation. To test this potential association, we applied an adjusted generalized linear regression model to measure quantitative association between sperm DNA methylation and age. Indeed, we found 97,412 DMCs associated with age. Among them, 9,837 (SI Appendix, Table S10) sperm DMCs were found to be associated with both BMI and age (Shared Age/BMI DMCs) (q-value < 0.01) (SI Appendix, Fig. S6A). Like BMI-DMCs, Age-DMCs overlapped with previously identified dynamic CpGs (18) and found enriched in intergenic regions (Z-score > 100), but also strongly enriched at Introns and upstream gene promoters (Z-score > 200, and Z-score > 100 respectively) (SI Appendix, Fig. S6 A and B). Corresponding Age-DMRs at gene features were also associated with neural gene ontologies (P-value < 2.5 × 10⁻⁵) (SI Appendix, Fig. S6C and Table S11). This is consistent with previous work associating increased paternal age with increased risk of autism/ADHD syndrome in children (15). Moreover, Age-DMCs were enriched at retrotransposons (notably LINE and ERV elements, Z-score > 100) (SI Appendix, Fig. S6D). Again, shared BMI/Age-DMCs demonstrated strong association with subfamilies of transposable elements that were identified at BMI-DMCs like LTR33, SST1, and L1M3D elements (SI Appendix, Fig. S6E). This suggests that these subfamilies are more likely to be significantly affected by DNA methylation changes in association with both age and BMI. The consistent enrichment of both BMI-DMCs and Age-DMCs at intergenic regions that display dynamic methylation levels suggests a potential function as environment/metabolic biosensors. Indeed, after merging Age-DMCs into Age-DMRs (n = 10,329) (SI Appendix, Fig. S6F), we found some overlap between Age-DMRs with previously reported DMRs involved in published sperm epigenetic clocks (33, 44). Moreover, some DMRs influenced by both age and BMI (n = 1,304) showed overlaps with sperm epigenetic clocks DMR from Cao et al. and Pilsner et al. [(33) and (44), respectively]. More saliently, 136 and 83 BMI-DMR were detected to overlap with Cao’s and Pilsner’s epiclocks, respectively. This finding highlights BMI as a potential confounding factor when considering the effects of age on the sperm methylome (SI Appendix, Fig. S6F).

Sperm BMI-DMRs Occurred in Embryonic Enhancers Within Transposable Elements.

Transposable elements can evade epigenetic reprogramming in the early embryo (23, 24, 69). To investigate the potential of sperm BMI-DMRs to impact early human development, we overlapped sperm BMI-DMRs with active 8 Cell-stage (8C) enhancers, a critical time point of human zygote genome activation (47). Enhancers were defined by the presence of Histone 3 Lysine 27 acetylation (H3K27ac), with and/or accessible chromatin marked by ATAC-seq, and showing HiC interaction with gene promoters at a minimum distance of 3 kb from the nearest TSS (47). We identified 197 BMI-DMR enhancers that associated with 123 genes (Fig. 4A and SI Appendix, Table S12). Some DMR associated enhancers overlapped TEs (n = 79), that were primarily ERV, LINE elements (Fig. 4B), aligning with previous findings showing that LTR, LINE elements to contribute to enhancer regulation in early development (70, 71). Most DMRs localized to 8C enhancers tended to gain methylation (n = 159), while 38 lost methylation (Fig. 4 A and C and SI Appendix, Fig. S7D). Given the relatively short list of genes, we detected no enriched ontologies associated with these regions. To gain insight into how DMRs associated with enhancers may influence gene expression and embryo quality, we leveraged RNA-seq data profiled in 8C human embryos over arrested blastomeres (SI Appendix, Fig. S7 A–C) (46, 47). We observed that genes associated to sperm BMI-DMR at 8C enhancers were active at ZGA (Fig. 4D). To further characterize these overlaps between 8C Enhancers and sperm BMI-DMRs, we looked for enriched transcription factor binding motifs. Identified binding motifs included important regulators of early development, most notably embryonic and brain development factors (SOX, ASCL1, FOS:JUN, TFAP factors) (Fig. 4E and SI Appendix, Fig. S7 D and E and Table S13). Interestingly, several of these factors are involved in inflammation or are clinical targets used to reduce inflammation like MAFG or REL (72, 73). This may reflect an impact of BMI that is related to the chronic inflammation state of being overweight. In addition, some DMR/ZGA enhancer elements associated with placental development–associated factors SNAI1 and FOSL1 (74–76) (Fig. 4E and SI Appendix, Table S13). These associations suggest that if these DMRs are paternally transmitted and escape reprogramming, this may impact ZGA, initiation of embryogenesis, and/or placentation.

Fig. 4.

A subset of sperm DMRs overlaps with 8-cell enhancers that are expressed in 8C embryos. (A) Scatter plot showing the level of DNA methylation, and ATAC/H3K27ac enrichment (Average log2 enrichment) at BMI-8C enhancers DMRs. (B) Proportion of Transposable elements (n = 79) found overlapping at BMI-Enhancers DMR. (C) Signal track showing the ∆DNA methylation at DMRs overlapping with 8C enhancers marked with H3K27ac and ATAC. (D) Volcano plots showing the expression of BMI-DMR and 8C-enhancers associated genes in normal 8C human embryos over arrested blastomeres. Thresholds were set up at FDR < 5% and Log2(FC) > 1 (two-sided t test). Data obtained from biological duplicate (46, 47). (E) Transcription factor binding motif enrichment analysis on lost and gained (L/G) BMI-DMR 8C enhancer. Enrichment stats were calculated using the one-sided Fisher test.

A Subset of BMI-DMRs Overlapped with Promoters of Active Genes in Lineage Specification at the Blastocyst Stage.

It has long been established that the paternal epigenome is a key regulator of the first lineage segregation and the development of the extraembryonic tissues through imprinted genes (3, 77). In pregnancies where the father is overweight, there is an increased incidence of miscarriage, placental developmental abnormalities, and obstetrical complications such as preeclampsia (78). To assess whether DMRs in sperm from men with elevated BMI could persist through development and impact trophectoderm gene expression, we queried whether 8C enhancer regions overlapped with sperm-DMRs that remain active at the blastocyst stage. Interestingly, 8C enhancers marked with sperm-BMI DMRs not only retained accessible chromatin at the ICM stage compared to the 8C stage, but even demonstrated a slight increase in H3K27ac (Fig. 5A). This suggests a potentially persistent effect of the altered epigenomic landscape at regions overlapping with sperm BMI-DMRs.Next, to investigate the potential impact of sperm BMI-DMRs on cell fate decisions at the blastocyst stage, we looked at the overlap of DMRs with gene promoters. We found 477 DMRs overlapping with gene promoters and 5′UTRs (SI Appendix, Fig. S8A and Table S14), some of which were associated with enriched H3K4me3 overlapping with ATAC peaks at the ICM stage, regardless of whether there was a gain or loss of DNA methylation level. These DMRs were associated with 186 genes. Consistent with our findings at the 8C stage, these also included inflammation genes, or genes that are clinical targets to reduce inflammation like ACKR3, S1PR3, and GAS6 (Fig. 5 B and C and SI Appendix, Table S14) (79–81). Interestingly, we also found regulators of spermatogenesis like TEX19, which is involved in the piRNA pathway (82), and STOX2 whose expression in the placenta is associated with preeclampsia (83) (Fig. 5C and SI Appendix, Table S14). Despite the convergent roles in inflammation and placental development found between gene promoter overlapping with sperm BMI-DMR and gene regulated with ZGA enhancer bearing a DMR, they were not enriched gene ontology terms given the limited number of genes with DMRs involved in these processes.

Fig. 5.

A subset of sperm DMRs overlaps with gene promoters that show a Trophectoderm-biased expression in human blastocysts. (A) Violin plots of enrichment for ATAC and H3K27ac at 8C enhancer at 8C stage and in the ICM at blastocyst stage. (B) Volcano plots showing the expression of BMI-DMR at gene promoter and their associated expression ICM vs. TE human blastocysts. Thresholds were set up at FDR < 5% and Log2(FC) > 1 (two-sided t test). Data obtained from five biological replicate (45). (C) Signal tracks in human ICM showing the ∆DNA methylation at DMRs overlapping with gene promoters (ACKR3 inflammation-related gene and STOX2 expressed in placenta of preeclamptic women). (D) Transcription factor motif enrichment analysis on lost and gained (L/G) BMI-DMR at gene promoters. Enrichment stats were calculated using the one-sided Fisher test.

We reanalyzed RNA-seq data of human early lineages, the epiblast/ICM and trophectoderm/TE (45) (SI Appendix, Fig. S8 B and C) to examine the expression patterns of these genes marked by DMRs in the sperm of men with elevated BMI. We found that sperm BMI-DMR occurred at gene promoters in both ICM and TE lineages and were expressed in both lineages. A larger proportion (61%) of DMRs were associated with developing trophectoderm at the blastocyst stage (Fig. 5B and SI Appendix, Table S14). Binding motifs at these promoters revealed transcription factors associated with neurodevelopment (ASCL2, ETV5), as well as early cell fate decision transcription factors (RUNX1, TFAPs, FOX factors). In addition, TFAP factors have been shown to play an important role in regulating the emergence of the trophectoderm lineage in mice (84). Interestingly, these motifs were also associated with behavioral genes, which links to our disease ontology analysis (notably through MEIS factors) (Fig. 5D and SI Appendix, Fig. S8D and Table S15). This supports the notion that BMI-sensitive DMRs at promoters could influence genes expressed in the blastocyst, and their epigenetic alteration could lead to impaired embryonic development. Moreover, the higher proportion of sperm-DMR-marked genes being expressed in the trophectoderm resonates with the known function of the paternally inherited epigenome in the development of the placenta (3, 4, 77). The placenta, a known neuroendocrine tissue, is critical in the development of the embryo’s nervous system (85). Hence, our data highlight the possibility of a sperm–placenta–brain axis as a potential route for paternally transmitted epigenetic insults to affect early embryo development and neural development.

Discussion

The Sperm Epigenome Reflects Elevated BMI and Age.

The literature highlights external factors as being influential on men’s preconceptional health, with elevated weight linked to impaired fertility and altered sperm methylome (86–89). Furthermore, the effects of paternal inheritance have been observed across generations in both human population studies and animal models (88–91), highlighting the consequences of men’s health on the following generation. The sperm epigenome, and DNA methylation in particular, may serve as a potential bridge between generations.

Overall, we identified over 38,000 BMI-sensitive DMCs and over 3,800 differentially BMI-sensitive DMRs, mostly located at intergenic regions in the sperm of men with elevated BMI. Supporting a role for paternal BMI influence on neurodevelopmental disorder in children, we observed DMRs at genes associated with neural developmental and disease ontologies. This finding is consistent with previous studies (92) and further implicates sperm DNA methylation as a potential mechanism underlying the epidemiological findings that overweight and obese men are at a greater risk of having a child with a neurological disorder including autism. This association has important ramifications for prevention of childhood disease and the need to include fathers’ health as a consideration for preconception intervention.

Interestingly, most of the sperm BMI-DMCs/DMRs displayed a trend for hypermethylation. As we observed enrichment of DMR at promoters of genes involved in inflammation (REL, ACKR3, MAFG) (SI Appendix, Table S14), one possible explanation can be drawn from the association between obesity and low-grade chronic inflammation in the testes (66, 93–95). Indeed, oxidative stress triggered by inflammation is known to impact the DNA methylation landscape through DNA damage repair pathways, or alteration of metabolism, and/or changes in the cell’s redox environment (67, 68). It was proposed that some superoxide reactive species could increase the likelihood of methylation deposition at cytosines (96). Furthermore, DNMTs are known to be recruited to DNA damage sites; specifically hypermethylation can be promoted via DNMT recruitment at promoters via oxidative stress (97, 98). This is worth considering as one molecular mechanism that could explain the preponderance of DMCs bearing increased methylation.

Here, we report that BMI and age may influence a common subset of DMCs and that sperm BMI-DMR overlapped those implicated in sperm epigenetic clock models (33, 44). Interestingly, inflammation is a common physiological response that also occurs with increasing age (99). This reinforces the idea that there may be a common subset of DMCs/DMRs in sperm that are responsive to multiple physiological and/or environmental conditions and diseases. Likewise, a recent study as also found that age-related change in sperm DNA methylation associate with neural ontologies, further highlighting the association between these cues (100). As open-source sperm DNA methylation data from men under various conditions becomes available, such a large-scale analysis using machine learning or AI-based approaches to probe for common functional regions of the genome will be warranted.

Moreover, in a mouse model paternal obesity, oxidative stress, and inflammation have also been linked to changes in other epigenetic marks, particularly H3K4me3 (101, 102). Our work here highlights that DMRs can occur in regions normally marked by H3K4me3. Interestingly, H3K4me3 and DNA methylation both rely on the availability of the methyl donor SAM, which is produced through the one-carbon cycle, highlighting a potentially shared metabolic dependency (103, 104). Due to limitations in sperm number, and the use of multi-omics methods not yet being optimized for sperm, we were unable to simultaneously probe other epigenetic read-outs like small noncoding RNA, chromatin accessibility or other histone modifications (105). In future work, a more comprehensive view of the sperm epigenome in men with elevated BMI should be pursued to explore their interconnective relationships.

Could a “Sperm–Placenta–Brain” Axis Link Paternal BMI to Offspring Disease Via Transposable Elements?

BMI-associated DMCs significantly overlap with some families of retrotransposons (mostly LTR/ERV and LINE1). These elements have the capacity to duplicate themselves with potential deleterious effects for the host genome, and DNA methylation is a critical repressor of their activity. Notably, the repression of young and potentially transcriptionally active transposable elements is of importance for spermatogenesis and fertility (5, 106, 107). Despite their potential negative effect on the genome, transposable elements are also known to act as platforms for the recruitment of transcription factors (TFs), and play the role of cis-regulatory elements, like enhancers. It is estimated that 20% of human transcription factor binding motifs are encoded in transposable element sequences (63). Remarkably, our binding motif enrichment analysis over BMI-DMRs at LTR and LINE1 elements indeed showed strong enrichment for transcription factors that are associated with embryonic and brain development. Additionally, we observed BMI-DMRs overlapping with transposable elements and 8-cell enhancers that are active at ZGA in human embryos up-until the blastocyst stage, a critical time point of cell fate decisions and embryo development. The consistent representation of neural pathways ontologies and neural factor motifs at DMRs over transposable elements or gene promoters is intriguing. This association between differences in DNA methylation and neural regulation has been recurrently observed in the literature both in the regulation of cell fate decisions in early development but also in men’s sperm (108–110). A recent publication demonstrated potential causal association between placental DNA methylation and neuropsychiatric disorders (111).

We noted the presence of some placental genes marked by DMRs and/or regulated by ZGA enhancers overlapping with a DMR at a transposable element. Notably, STOX2 expression is detected in the placenta of preeclamptic women, and SNAI1, JUNB. This aligns with our observation that some sperm BMI-DMR marked gene promoters are active in the trophectoderm lineage at the blastocyst stage. Interestingly, the placenta, which emerges from the trophectoderm, is a neuroendocrine tissue that plays a critical role in the regulation of embryonic development. First identified in the context of the “Placenta-heart connection” (112), the role of the placenta in regulating fetal brain development has since been increasingly recognized (85). This connection has been recently referred to as the “Placenta–Brain axis,” or “neuroplacentology” and constitutes an emerging field of research (113, 114). Indeed, previous studies have emphasized the connection between placental dysfunction and obstetrical complications in pregnancies fathered by overweight men, as well as the increased placental serotonin output to the fetal brain in response to maternal inflammation during pregnancy (78, 115, 116). More recently, the placental origin of intergenerational effects has also been demonstrated in mice as sires with microbiome dysbiosis triggered reduced placental size and altered placental transcriptome in their offspring (117). Particularly pertinent for this study are our own previous findings that pregnancies sired by obese male mice result in abnormal placenta development and function (101, 102).

Hence, we propose that DMR carried by the sperm could affect this Placenta–Brain axis. Changes in the sperm methylome could impact a key regulator of extraembryonic tissue development and later indirectly impact embryonic brain development. From a developmental standpoint, our proposed “Sperm–Placenta–Brain” axis hypothesis is supported by the well-known unequal contribution of parental gametes, with the sperm contributing mostly to the emergence of extraembryonic tissues in the early embryo, notably through gene imprints (3, 4). This suggests that in mice and humans there are fewer regions that are paternally imprinted than maternally imprinted (118, 119). Notably, previous work has shown that there exist transient and tissue specific imprinted regions that have life-long consequence for the offspring (120, 121). The mechanism by which canonical and transient imprints evade epigenetic reprogramming in early mammalian development is well established, and usually involving targeted DNA methylation maintenance via KRAB-ZFPs (2, 122–124) or an epigenetic switch (125–128).

How sperm DMCs/DMRs are transmitted and maintained in early embryogenesis has not been mechanistically explored. However, we propose the following hypotheses: First, it has been shown that in some instances, the formation of maternal imprints can arise from the expression of LTR transposable elements (termed LTR-initiated transcriptional units; LITs) that promote the deposition of DNA methylation in sequences located downstream from these LITs (69, 129). Indeed, we observed sperm DMRs and DMCs occurring at LTR elements, some of which evade reprogramming (7, 24). Whether such events also occur in the sperm methylome remains to be investigated.

Second, reconstructive inheritance through an epigenetic change that triggers changes in downstream signaling is an interesting possibility for the transmission of an epimutation. For example, inheritance of transcription factors bound to the sperm chromatin and sperm chromatin genome organization through CTCF have been shown to be passed down from father to their offspring and influence development (130, 131). This resonates with our observations of large enrichment of TFs binding motifs at sperm-DMRs, including CTCFL, that we observed in DMR-marked 8C enhancers. Interestingly, a mouse model of paternal inheritance of epigenetically edited CGI suggests that DNA methylation could be both replicatively (for highly methylated CGI) and reconstructively (for lowly methylated CGI) inherited from father to offspring (132).

There are limitations of this study including the modest cohort size. In addition, the clinical nature of the cohort introduces variability in fertility status that was not resolved by selecting the vast majority of samples as having normal semen analysis and DFI. It is well established that normal semen analysis does not provide direct insight into fertility status (133). Although multiple covariates were incorporated into the model, these analyses should be interpreted with appropriate caution. Another important consideration is that a sperm population is inherently heterogenous. Despite our use of selection methods to isolate the most viable sperm by gradient centrifugation prior to analysis, there was likely sperm-to-sperm variability in the epigenome. As quantitative single cell sperm methylome profiling is not yet available, we performed bulk epigenetic profiling and therefore captured signals across the entire sperm population. It is worth considering that this epigenetic profiling may have included contributions from nonviable sperm or developmentally impaired sperm. Notably, our approach captured population-level epigenetic variation that may be relevant to early developmental processes, including those implicated in the proposed sperm–placenta–brain axis, although direct transmission to the embryo cannot be resolved in the current study.

To evaluate the robustness of our findings to potential confounding, we performed sensitivity analysis using a reduced model including age as the sole covariate. This analysis identified 36,514 DMCs (q-value < 0.01), of which 22,786 (~60%) overlapped with those identified in the full model (38,719 DMCs), all with consistent direction of effect in terms of gain or loss of methylation. These findings indicate that the majority of BMI-associated methylation signals are across modeling strategies and are not driven by overfitting or covariate selection.

Together, these observations support that the identified methylation patterns represent biological meaningful variation associated with paternal metabolic state rather than technical or analytical artifacts. More broadly, our findings support a model in which the sperm epigenome reflects both metabolic and inflammatory states of the father, with systemic inflammation (as measured by CRP) acting to modulate the magnitude of these effects. In this context, sperm DNA methylation pattern may serve as a sensitive biomarker -akin to a “canary in the coal mine” of men’s overall health, capturing environmental and physiological exposures with potential implications for early embryonic development and the proposed sperm–placenta–brain axis.

By revealing the extensive epigenetic alterations in the sperm of men with elevated BMI, our findings together with other studies in the field linking sperm BMI, paternal age, and BMI (92, 100, 134) provide a foundational framework for advancing our understanding on paternal factors that may influence early embryonic development and the long-term health outcomes in subsequent generations. This emphasizes the importance of integrating paternal health into reproductive and public health strategies to optimize both paternal and offspring well-being.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Dataset S05 (XLSX)

Dataset S06 (XLSX)

Dataset S07 (XLSX)

Dataset S08 (XLSX)

Dataset S09 (XLSX)

Dataset S10 (XLSX)

Dataset S11 (XLSX)

Dataset S12 (XLSX)

Dataset S13 (XLSX)

Dataset S14 (XLSX)

Dataset S15 (XLSX)

Data, Materials, and Software Availability

The supporting datasets can be referenced in the main text, supplementary figures, and tables. The raw data generated in this study have been deposited in the European Genome-phenome Archive (EGA) with dataset accession number EGAS50000001028 (135). No original code was written for bioinformatic analysis and visualization displayed in this manuscript; all information can be found in the listed packages in the Supplementary Materials.