Within-subject changes in methylome profile identify individual signatures of early-life adversity, with a potential to predict neuropsychiatric outcome

Abstract

Background: Adverse early-life experiences (ELA), including poverty, trauma and neglect, affect a majority of the world's children. Whereas the impact of ELA on cognitive and emotional health throughout the lifespan is well-established, it is not clear how distinct types of ELA influence child development, and there are no tools to predict for an individual child their vulnerability or resilience to the consequences of ELAs. Epigenetic markers including DNA-methylation profiles of peripheral cells may encode ELA and provide a predictive outcome marker. However, the rapid dynamic changes in DNA methylation in childhood and the inter-individual variance of the human genome pose barriers to identifying profiles predicting outcomes of ELA exposure. Here, we examined the relation of several dimensions of ELA to changes of DNA methylation, using a longitudinal within-subject design and a high threshold for methylation changes in the hope of mitigating the above challenges.

Methods: We analyzed DNA methylation in buccal swab samples collected twice for each of 110 infants: neonatally and at 12 months. We identified CpGs differentially methylated across time, calculated methylation changes for each child, and determined whether several indicators of ELA associated with changes of DNA methylation for individual infants. We then correlated select dimensions of ELA with methylation changes as well as with measures of executive function at age 5 years. We examined for sex differences, and derived a sex-dependent 'impact score' based on sites that most contributed to the methylation changes.

Findings: Setting a high threshold for methylation changes, we discovered that changes in methylation between two samples of an individual child reflected age-related trends towards augmented methylation, and also correlated with executive function years later. Among the tested factors and ELA dimensions, including income to needs ratios, maternal sensitivity, body mass index and sex, unpredictability of parental and household signals was the strongest predictor of executive function. In girls, an interaction was observed between a measure of high early-life unpredictability and methylation changes, in presaging executive function.

Interpretation: These findings establish longitudinal, within-subject changes in methylation profiles as a signature of some types of ELA in an individual child. Notably, such changes are detectable beyond the age-associated DNA methylation dynamics. Future studies are required to determine if the methylation profile changes identified here provide a predictive marker of vulnerabilities to poorer cognitive and emotional outcomes.

Keywords: Adverse Childhood Experiences; Biomarkers; DNA methylation; early-life adversity; epigenetic clocks; executive control; methylomics; precision medicine; stress; within-subject design.

Figure 1.. Methylation is influenced by age.

A) Timeline of sample collection and assessments in 110 infants. B) Heatmap depicting distinct patterns of methylation distinguishing DNA methylation profiles from newborn and 1-year old children. C) Average percentage methylation at selected sites increases with age. D) Using PCA, the first principal component, explaining 25% of the variance, accounts for the age of sample collection. **p<0.001, bars represent mean, lines represent individual sites.

Figure 2:. localization and gene ontology of the sites differentially methylated (DMS) between neonatal and one year old samples.

(A). Chromosomal distribution of the DMS demonstrates that the reside on all autosomes. Numbers on the left denote the percentage of the overall DMS that localize to each chromosome. (B) Alignment of the DMS with genes and their structures: 13983 of the 14037 DMS localized to within 1000kb of transcription a start site (TSS), and these 13983 DMS associated with 2764 unique genes. (C) Gene ontology identified developmental processes as the key theme of genes associated with DMS between 10-day old and one year old samples of the same child. n=110 infants.

Figure 3.. Methylation-changes of individual infants between the ages of 10 days and one year predict effortful control at 5 years.

A) The first component of the principal component analysis (PCA) of methylation changes in the ~ 14,000 differentially methylated sites reflects the average change in methylation (n=110). B) Average methylation change from newborn to one year of age of an individual child predicts effortful control performance at five years of age (n=90). C) Average percent methylation in newborns does not predict outcome. D) Similarly, average percent methylation at one year of age does not predict outcome Note that analogous results were observed when using all 1.74 million methylated sites, as shown in the Supplemental Fig.S-3. Points represent individual samples, circles = females, triangles = males. Line represents linear regression.

Figure 4:. Unpredictability, assessed using the QUIC, portends functional outcomes at 5 years.

A) Income / needs ratio (INR) has a weak association with effortful control in our sample of individuals. B) Correlation of maternal depressive symptoms with effort control suggests a weak negative association in these individuals. C) Measures of maternal sensitivity do not correlate with effortful control at 54 months of age in this sample. D) Unpredictability measured using the Questionnaire of Unpredictability in Childhood (QUIC) is inversely correlated with effortful control score at 54 months. Points represent individual samples, circles = females, triangles = males. Line represents linear regression (n=90).

Figure 5.. Unpredictability portends child development and may interact with methylation changes over time.

A) Unpredictability assessed using the QUIC predicts effortful control at five years of age to a similar degree in both males and females. B) The change in methylation over the first year of life also predicts effortful control at the same age to the same degree in both sexes. n: females=41, males=49. C) There is an interaction between unpredictability and change in methylation in females only: for females who experience high unpredictability, the change in methylation over the first year of life predicts effortful control. In contrast, there is no such interaction observed in males. n: females low=28, females high=13, males low =16, males high=34. Points represent individual samples, purple = females, green = males. Line represents linear regression. F=females, M=Male.

Figure 6.. Impact scores identify individuals vulnerable to poor outcomes following the unpredictability dimension of early-life adversity.

A) Flow chart of the computation method used to select highly contributing sites. The clumping and thresholding method identifies sites that appear to have a significant interaction but do not correlate highly with other sites that have already been selected. B) Manhattan plot representing the differentially methylated sites (blue) and the distribution across the chromosomes and the corresponding significance score (−log10p) of each site interacting with QUIC to predict effortful control in females. Sites in red are those selected via the clumping and thresholding algorithm. Dotted line is at p=0.05. C) The significant interaction of QUIC and risk score calculated from top sites according to our model predicts effortful control at 5 years in females who have experienced more adversity n: low=28, high=13. Line represents linear regression.