Sex-based differences in placental DNA methylation profiles related to gestational age: an NIH ECHO meta-analysis

Figures & data

Table 1. A summary of key information from participating cohorts.

	Females			Males			Gestational Age Data Source	Methylation Array
Cohort	N	Gestational Age (weeks)		N	Gestational Age (weeks)
		Median	Range		Median	Range
Early Autism Risk Longitudinal Investigation (EARLI)	37	40.0	36.6–41.6	54	39.3	36.3–41.6	Reported by the labour and delivery nurse present at delivery; if unavailable, calculated as: (date of birth – (estimated delivery date – 280))/7	450 K
Extremely Low Gestational Age Newborns (ELGAN)	112	26.4	23.0–27.9	130	26.0	23.0–27.9	Determined by foetal ultrasound or dates of embryo retrieval or intrauterine insemination; if unavailable, calculated as: (date of birth – last menstrual period date)/7	EPIC
Healthy Start	47	39.7	36.1–41.7	51	39.3	35.1–41.1	Obtained from the medical record at delivery	EPIC
New Hampshire Birth Cohort Study (NHBCS)	159	39.0	31.8–41.9	184	39.0	31.0–41.9	Determined by foetal ultrasound; if unavailable, calculated as: (date of birth – last menstrual period date)/7	450 K
Overall	355	36.3	23.0–42.0	419	35.9	23.0–41.9

Figure 1. Manhattan and volcano plots of the inverse-variance fixed meta-analysis results for gestational age and placental CpG methylation among EARLI, ELGAN, healthy start, and NHBCS.

A. Manhattan plots of placental CpGs associated with gestational age among females (N = 355; top) and males (N = 419; bottom); B. Volcano plots of placental CpGs associated with gestational age among females (top) and males (bottom). In the Manhattan plots, the x-axes represent genomic locations. In the volcano plots, the x-axes show the estimated mean difference in methylation β-values for a 1-week increase in gestational age. Y-axes for both plots show -log₁₀(p-values) with the solid horizontal black line indicating the false discovery rate (q < 0.05) and the dashed horizontal black line indicating the more conservative Bonferroni threshold (1.41 × 10⁻⁷) for statistical significance.

Figure 2. Overlap of FDR-significant CpGs across the meta-analyses of gestational age and placental methylation among females and males from EARLI, ELGAN, healthy start, and NHBCS, with and without cell-type adjustment.

Female-specific results without cell-type adjustment are shown in light pink and with cell-type adjustment are shown in dark pink; male-specific results without cell-type adjustment are shown in light blue and with cell-type adjustment are shown in dark blue.

Table 2. A summary of the number of identified CpG sites from each EWAS meta-analysis of gestational age.

	Females (N = 355)		Males (N = 419)
Model	FDR-significant CpGs	Survived leave-one-out	FDR-significant CpGs	Survived leave-one-out
Unadjusted	407	382	794	598
Cell-type–adjusted	55	3	826	679

The false discovery rate (FDR) threshold for statistical significance was q < 0.05. Leave-one-out refers to performing the meta-analyses excluding one cohort each time. When any single cohort was omitted, ‘survival’ was defined as: (1) having a p-value < 0.05, (2) maintaining an association in the same direction as the overall meta-analysis; (3) maintaining a magnitude of association within 50% of the overall meta-analysis. EWAS, epigenome-wide association study.

Table 3. Top 10 FDR-significant CpGs for the meta-analysis of gestational age and placental methylation among females and males from the EARLI, ELGAN, Healthy Start, and NHBCS cohorts.

Probe (Gene)	Coefficient (95% CI)	Q-value	Heterogeneity I^2 (P-value)	Direction in Each Cohort	Survived leave-one-out
Females (N = 355)
cg06677013 (CYB5R4)	0.015 (0.011, 0.019)	9.60E-08	0.0 (0.39)	++++	Yes
cg03094728 (AKAP9)	−0.017 (−0.022, −0.012)	1.76E-04	41.2 (0.18)	– ?	Yes
cg23657718 (AX748388^+)	−0.014 (−0.018, −0.009)	2.58E-04	34.4 (0.21)	– -	Yes
cg15271616 (RUSC2)	0.015 (0.010, 0.021)	6.23E-04	0.0 (0.65)	++++	Yes
cg10649156 (RPL3^+)	−0.010 (−0.014, −0.007)	7.12E-04	0.0 (0.68)	– -	Yes
cg09290175 (RPL3^+)	−0.014 (−0.019, −0.009)	3.37E-03	0.0 (0.60)	– -	Yes
cg22821677 (LOC646214^+)	0.012 (0.007, 0.016)	3.57E-03	0.0 (0.75)	++++	Yes
cg10821976 (TSNAX-DISC1)	0.013 (0.009, 0.018)	3.78E-03	24.2 (0.27)	-+++	Yes
cg03982074 (MUC12^+)	−0.013 (−0.017, −0.008)	9.60E-08	48.5 (0.12)	– -	Yes
cg08995609 (RIN1)	0.007 (0.005, 0.010)	2.58E-04	24.6 (0.26)	++++	Yes
Males (N = 419)
cg05169312 (GEMIN5)	0.010 (0.007, 0.013)	4.82E-05	0.0 (0.43)	++++	Yes
cg05875982 (PRCD)	0.009 (0.006, 0.012)	4.82E-05	41.9 (0.16)	++++	Yes
cg12447744 (MDM4)	0.008 (0.005, 0.011)	3.94E-04	0.0 (0.52)	++++	Yes
cg14462758 (EXOSC2)	0.008 (0.005, 0.011)	5.07E-04	0.0 (0.80)	++++	Yes
cg20285026 (PRKCZ)	0.012 (0.008, 0.016)	6.62E-04	0.0 (0.49)	++++	Yes
cg11373252 (NHEJ1^+)	0.004 (0.003, 0.006)	7.67E-04	0.0 (0.53)	++++	Yes
cg08827358 (BEAN1)	0.008 (0.005, 0.011)	7.99E-04	0.0 (0.54)	++++	Yes
cg04503912 (DSPP^+)	0.011 (0.007, 0.015)	1.05E-03	0.0 (0.51)	++++	Yes
cg07642463 (ZNF264)	0.010 (0.006, 0.014)	5.07E-04	0.0 (0.43)	++++	Yes
cg07954075 (LOC107984507^+)	0.006 (0.004, 0.009)	6.62E-04	24.2 (0.27)	++++	Yes

Coefficients correspond to the expected change in placental DNA methylation per additional week of gestation. Gene annotations were verified in the UCSC Genome Browser (hg19) with ⁺ indicating the closest gene. The direction of coefficients is ordered by cohort as follows: EARLI, ELGAN, Healthy Start, NHBCS. A ‘?’ mark indicates the respective cohort was missing methylation data for a given probe.

Figure 3. Comparison of meta-analysis model coefficients for gestational age and placental CpG methylation among EARLI, ELGAN, Healthy Start, and NHBCS before and after cell-type adjustment.

A. Scatter plot of unadjusted and cell-type adjusted model coefficients for a 1-week increase in gestational age among females (N = 355); B. Scatter plot of unadjusted and cell-type adjusted model coefficients for a 1-week increase in gestational age among males (N = 419).

Figure 4. Manhattan and volcano plots of the inverse-variance fixed meta-analysis results for gestational age and placental CpG methylation after cell-type adjustment among EARLI, ELGAN, healthy start, and NHBCS.

A. Manhattan plots of placental CpGs associated with gestational age after adjustment for cell-type composition among females (N = 355; top) and males (N = 419; bottom); B. Volcano plots of placental CpGs associated with gestational age after adjustment for cell-type composition among females (top) and males (bottom). In the Manhattan plots, the x-axes represent genomic locations, whereas in the volcano plots, the x-axes show the estimated mean difference in methylation β-values for a 1-week increase in gestational age. Y-axes for both plots show -log₁₀(p-values) with the solid horizontal black line indicating the false discovery rate (q < 0.05) and the dashed horizontal black line indicating the more conservative Bonferroni threshold (1.41 × 10⁻⁷) for statistical significance.

Table 4. Top 10 FDR-significant CpGs for the meta-analysis of gestational age and placental methylation with adjustment for cell-type proportions among females and males from the EARLI, ELGAN, healthy start, and NHBCS cohorts.

Probe (Gene)	Coefficient (95% CI)		Q-value	Heterogeneity I^2 (P-value)		Direction in Each Cohort	Survived leave-one-out
Females (N = 355)
cg11650925 (GNG12-AS1)	0.012 (0.008, 0.015)	3.18E-04			21.5 (0.28)	++++	No
cg06677013 (CYB5R4)	0.012 (0.008, 0.016)	4.19E-04			36.8 (0.19)	++++	No
cg15030415 (RAP1GAP2)	−0.012 (−0.016, −0.008)	4.19E-04			35.2 (0.20)	– -	No
cg04732596 (MEIS1)	−0.005 (−0.007, −0.003)	9.63E-03			0.0 (0.43)	– -	No
cg10113589 (OPALIN)	0.009 (0.005, 0.012)	9.63E-03			41.3 (0.16)	++++	No
cg26709300 (YPEL3)	−0.008 (−0.011, −0.005)	9.63E-03			38.7 (0.18)	– -	No
cg02959609 (C1orf230)	0.009 (0.006, 0.012)	9.93E-03			15.6 (0.31)	++++	No
cg15175129 (MSRB3)	0.009 (0.005, 0.012)	1.10E-02			43.8 (0.15)	++++	No
cg12617598 (CDC73)	−0.011 (−0.015, −0.007)	1.16E-02			26.3 (0.25)	– -	No
cg10821976 (TSNAX-DISC1)	0.014 (0.009, 0.02)	1.20E-02			0.0 (0.62)	++++	No
Males (N = 419)
cg12447744 (MDM4)	0.008 (0.006, 0.011)	7.32E-05			40.5 (0.17)	++++	Yes
cg12134349 (SLC45A4; LOC105375787)	0.008 (0.005, 0.011)	1.42E-04			27.9 (0.24)	++++	Yes
cg23323879 (CILP^+)	0.007 (0.005, 0.009)	1.42E-04			0.0 (0.82)	++++	Yes
cg05169312 (GEMIN5)	0.010 (0.006, 0.013)	2.37E-04			0.0 (0.53)	++++	Yes
cg17402397 (SMTN)	0.007 (0.005, 0.010)	2.37E-04			0.0 (0.76)	++++	Yes
cg09753772 (RPS7^+)	0.010 (0.007, 0.013)	3.57E-04			1.3 (0.39)	++++	Yes
cg05341549 (PIK3CD)	0.008 (0.005, 0.010)	4.21E-04			0.0 (0.75)	++++	Yes
cg11075561 (MAPK13^+)	0.010 (0.007, 0.014)	4.21E-04			33.4 (0.21)	++++	Yes
cg11585358 (EPHX4^+)	0.007 (0.005, 0.009)	5.35E-04			31.3 (0.22)	++++	Yes
cg19905757 (CORO2B)	0.007 (0.005, 0.010)	5.99E-04			0.0 (0.97)	++++	Yes

Coefficients correspond to the expected change in placental DNA methylation per additional week of gestation after adjustment for placental cell-type proportions. Gene annotations were verified in the UCSC Genome Browser (hg19) with ⁺ indicating the closest gene. The direction of coefficients is ordered by cohort as follows: EARLI, ELGAN, Healthy Start, NHBCS.

Figure 5. Sexually monomorphic associations of gestational age with placental methylation levels at cg17284609 (SOX6) before and after cell-type adjustment.

Forest plot depicting cohort-specific and meta-analysed associations of gestational age with placental methylation levels at cg17284609, the only probe that was FDR-significant in meta-analyses among both females and males before and after cell-type adjustment.

Figure 6. Sex-heterogeneous associations of gestational age with placental CpG methylation after cell-type adjustment.

Forest plot depicting the top 10 probes with methylation levels that differed most significantly between the two sexes. Gene annotations were verified in the UCSC Genome Browser (hg19) with + indicating the closest gene.

Figure 7. Gene Ontology (GO) enrichment analyses for the cell-type adjusted association of gestational age with placental CpG methylation among females and males.

A. The top 25 GO biological processes identified by enrichment p-value among female-associated CpG sites after undergoing REVIGO pruning. B. The top 25 GO biological processes identified by enrichment p-value among male-associated CpG sites after undergoing REVIGO pruning. GO term clusters are identified by rectangles with labels that reflect the least redundancy and highest degree of enrichment; colours represent related clustered GO terms, and the size of the rectangles is proportional to -log₁₀(p-value for enrichment).

Data availability statement

The individual-level datasets used for conducting EWASs in this manuscript are not publicly available because, per the NIH-approved ECHO Data Sharing Policy, ECHO-wide data have not yet been made available to the public for review/analysis. Requests to access the individual-level datasets should be directed to the ECHO Data Analysis Center, [email protected]. However, the EWAS summary statistics from each cohort that were used to conduct subsequent meta-analyses are available through the UNC Dataverse at https://doi.org/10.15139/S3/K73DN9.

All software and programs used to conduct these analyses are freely available and are cited in the manuscript. All code used to generate our results is available on the UNC Dataverse at https://doi.org/10.15139/S3/EEB75X.