This is the first study to comparatively examine DNAm signatures of GA in two ethnic populations (living in the same country) using cord blood. We found consistent DNAm signatures of gestational age at individual CpG levels and confirmed that the Bohlin's epigenetic GA clock correlated well with chronological GA in white Europeans and can be generalized to South Asians. In both populations, the GA clock was positively linked to newborn weight and length, and negatively to gestational diabetes, newborn female sex, and parity. Unique to South Asians, the GA clock was also associated with a higher newborn PI. We confirmed the associations of parity and newborn sex with GAA in both populations, and discovered that white European newborn males exhibited twice as much accelerated GA as compared to South Asians.
Previous EWASs of GA have found overlapping signals with other outcomes, such as birth weight, while EWASs of birthweight were enriched for CpGs previously associated with prenatal smoking, folic acid intake, and maternal hypertension or pre-eclampsia (26, 53, 80). Since the strongest association we found with epigenetic GA clock was GDM, we were interested in whether GDM or T2D genetics could be implicated in the epigenetics of GA. We found two CpGs significantly associated with GA in START that were present in the TCF7L2 gene, whose risk alleles increase the risk of T2D by reducing insulin secretion and are also associated with a lower BMI (81, 82). This lends support to the hypothesis that GDM in South Asians is predominantly due to insulin deficiency, which is consistent with their lower birth weight (83), whereby comparatively white Europeans may have a stronger insulin resistance component. Meanwhile, GLIS3 is another recognized gene for diabetes that is associated with the development of beta cells (84) and neonatal diabetes (85). Further, ADA gene, which has been linked to T2D (86) and several serum metabolites (87), was also mapped to CpGs significantly associated in START but was just below statistical significance in CHILD. Overall, there was an overrepresentation of genes that were known to associate with T2D in both START and CHILD, but with a more pronounced enrichment in START. The shared genes could be part of the mechanisms through which metabolic risks are transmitted from mothers to offspring, contributing to a transgenerational cycle of metabolic disorders (88–90).
Despite widespread differences in DNAm patterns between South Asians and white Europeans in an adult population (91), at an aggregate level, the Bohlin clock consisting of 132 CpGs produced a robust measure of biological GA with comparable performance in both cohorts. While the epigenetic GA clocks can adequately capture biological GA in both populations, they displayed subtle yet differential associations with maternal exposures and effects on offspring. The multitude of maternal exposures, ranging from nutrition, substance use, and metabolic factors to chronic health conditions, presented a complex tapestry of influences on GA, epigenetic GA and GAA. In general, South Asian women had low to no smoking and a diet that was mostly plant-based, and their newborns often exhibited a lower birthweight and a lower GA as compared to white Europeans. In this particular population, we found evidence of Western dietary pattern to influence GA and GAA. But this should be interpreted in the context of differences in the correlation between two populations: a plant-based diet score was positively correlated with Western and Health Conscious diets in South Asian women (r = 0.35 and 0.32), but a choice of plant-based diet implied a reduced Western diet (r = -0.21) and a strong preference for Health Conscious diet (r = 0.64) in white European women (Suppl. Figure 1).
In both cohorts, parity and newborn female sex were found to be negatively associated with epigenetic GA and GAA. While both maternal age and parity have been shown to associate with GAA (18, 49), our univariate and multivariate analyses pointed to parity being independently associated with epigenetic GA and GAA, but not maternal age. Consistent with this claim, a previous study showed a negative association between parity and maternal telomere length, which is an indicator of cellular aging. Advanced maternal age (e.g. >35) is typically considered a risk factor for pregnancy-related outcomes (92, 93). Here we showed that maternal age was negatively associated with GA and epigenetic GA in the univariate analysis, but only in the South Asian cohort, suggesting it could be capturing the effect of parity. This was supported by the observation that parity was more strongly correlated with maternal age in START (Pearson’s r = 0.45) than CHILD (r = 0.27), reflecting a higher proportion of older mothers being multiparous among South Asians. This could also explain why maternal age has the most significant impact on negative outcomes in South Asian women (94). On the other hand, our finding that newborn males have accelerated GA in both CHILD and START is consistent with established literature that male sex was associated with accelerated aging in newborns (43) and in adult populations (95).
In maternal health outcomes, the observed prevalence of GDM was consistent with those reported for the two populations in general (96, 97). We observed a negative relationship between GDM and epigenetic GA in both cohorts, as well as GDM and GAA in CHILD. The direction of associations was in line with a previous report (48). However, there was no evidence to support the role of GDM on GAA in the South Asian cohort despite much higher prevalence and sample size. We were able to replicate previous findings of pre-pregnancy weight and BMI having an impact on epigenetic GA in CHILD (43), but there was a coherent indication that gestational weight gain might be the more relevant exposure to GA and epigenetic GA in both cohorts.
The findings of this study should be interpreted in the light of several limitations. Cord blood DNAm in newborns reflects the regulation of gene activity during embryonic development, cellular differentiation, and response to environmental factors, and is thus a valuable source of information for assessing maternal risk factors and predicting future health outcomes. However, depending on the tissue types, resolution of the epigenetic data (e.g. single cell vs. tissue), technology (HM450K vs. EPIC), and method of GA clock construction, the performance of existing GA clocks varied substantially. The main observation is that the barriers to trans-ethnic EWAS analysis lie in various choices of technical and study design. For example, different tissue types might reveal tissue-specific responses, a placental clock could produce an elevated association with maternal exposures vs. offspring outcomes (6), which is complementary to existing studies of GA using cord blood. We have shown that DNAm clock, while universal in its foundational concept, can exhibit specific patterns and variations when analyzed across different ethnic groups. Factors contributing to these disparities might range from genetic backgrounds to varied environmental and socio-cultural practices that might not be readily available, such as alcohol consumption (34, 98) and air pollution (99, 100), both are known to alter epigenetics. These discrepancies, while enriching our understanding of epigenetic variability between ethnic groups, also serve as a clarion call for tailored healthcare and research approaches across populations.
In summary, this paper comparatively examined the transferability and health implications of epigenetic GA and GAA in white European and South Asian birth cohorts living in the same country. Two key findings are the consistent associations of parity, GDM, and newborn sex with epigenetic GA in both South Asian and white European cohorts, but with weaker effects observed in the South Asian cohort; and the positive association between epigenetic GA and newborn ponderal index that is exclusive to South Asian newborns. The epigenetic GA clock presents a promising avenue to bridge our understanding of maternal exposures and offspring health outcomes, and these differentiating characteristics of associations across the cohorts, enrich this narrative. Further research in the South Asian population, both adult and pediatric, is needed to strengthen these findings.