The impact of early adversity and education on genetic and brain morphological predictors of cognitive ability

Abstract Cognitive ability is a strong predictor of occupational achievement, quality of life and physical health. While variation in cognition is strongly heritable and has been robustly associated with early environment and brain morphology, little is known about how these factors combine and interact to explain this variation in cognition. To address this, we modelled the relationship between common genetic variation, grey matter volume, early life adversity and education and cognitive ability in a UK Biobank sample of N = 5237 individuals using structural equation modelling. We tested the hypotheses that total grey matter volume would mediate the association between genetic variation and cognitive ability, and that early life adversity and educational attainment would moderate this relationship. Common genetic variation, grey matter volume and early life adversity were each significant predictors in the model, explaining ~15% of variation in cognitive ability. Contrary to our hypothesis, grey matter volume did not mediate the relation between genetic variation and cognition performance. Neither did early life adversity or educational attainment moderate this relation, although educational attainment was observed to moderate the relationship between grey matter volume and cognitive performance. We interpret these findings in terms of the modest explanatory value of currently estimated polygenic scores accounting for variation in cognitive performance (~5%), making potential mediating and moderating variables difficult to confirm.

tested the hypotheses that total grey matter volume would mediate the association between genetic variation and cognitive ability, and that early life adversity and educational attainment would moderate this relationship. Common genetic variation, grey matter volume and early life adversity were each significant predictors in the model, explaining $15% of variation in cognitive ability. Contrary to our hypothesis, grey matter volume did not mediate the relation between genetic variation and cognition performance. Neither did early life adversity or educational attainment moderate this relation, although educational attainment was observed to moderate the relationship between grey matter volume and cognitive performance. We interpret these findings in terms of the modest explanatory value of currently estimated polygenic scores accounting for variation in cognitive performance ($5%), making potential mediating and moderating variables difficult to confirm.

K E Y W O R D S
childhood adversity, childhood trauma, cognition, education, genome-wide association, MRI brain volume

| INTRODUCTION
Cognitive ability is a strong predictor of mental and physical health, as well as mortality. 1 In the last decade, knowledge about how biological and environmental factors influence cognition has grown rapidly. This includes knowledge about the contribution of genetic variation and environmental factors, as well as the impact of both on brain morphological differences associated with cognition. In particular, the recent availability of large-scale data, including the UK Biobank, has propelled exciting developments in the area. Despite the significant advances in our understanding of the genetics of cognition, brain structure and environment, we remain at an early stage of modelling how these factors combine and interact. Addressing this gap in knowledge is critical because maintaining cognitive ability in the general population is important for reducing disability across the lifespan. In addition, a substantial body of literature suggests that variability in cognitive functioning can be explained by modifiable risk factors, including that of early life stress and educational attainment. 2 At a genetic level, twin and family studies have confirmed the heritability of cognitive ability, accounting for about 50% of the total variance in cognition. The contribution of genetics to explaining variation in cognitive ability increases during childhood and adolescence and remains high throughout adulthood. 3,4 Results from genome-wide association studies (GWAS) have showed that cognition is highly polygenic, with hundreds of genetic loci of small effect 4 and that these loci cluster in genes involved in regulating brain-specific gene expression. 1 While these effects are small and difficult to detect, the use of polygenic scores (PGS) has made it possible to model the cumulative effects of individual genetic variants and facilitate detection of geneenvironment interactions associated with cognition. In the largest study to date, 5 an intelligence-based PGS was found to explain up to 5.2% of the variability in intelligence.
A positive association between cognition and brain size has also been robustly identified by multiple studies, 6-8 with both phenotypes sharing a common genetic origin. 9,10 Indeed, Jansen and colleagues 9 identified an overlap of 67 genes between both traits with a genetic correlation of 0.23, and the association between genetic variants and cognition has been observed to be partly mediated by grey matter volume (GMV). 10 In a recent study by Cox et al., 7 examining differences in global and regional measures of grey and white matter volume, the strongest contribution to the variance explained in cognitive ability was found for total GMV.
It is widely hypothesised that genetic variation and brain volume both interact with environmental factors to explain variation in cognitive ability. 11,12 Early life adversity (ELA), including abuse, neglect, witnessing domestic or other violence and chronic poverty, 13,14 has received a great deal of attention in the literature. Exposure to ELA during periods of heightened plasticity may alter developmental trajectories via structural neurobiological mechanisms that in turn, increase the risk of cognitive impairments in adulthood. 15 Indeed, in both clinical and non-clinical samples, ELA has been found to be associated with variability in brain structure, including reductions in total GMV, 16,17 and across limbic and prefrontal regions [18][19][20] and several large prospective and retrospective studies have documented an association between ELA and poorer cognitive outcomes in adulthood. 14,21-24 Further, in a recent moderated mediation study by Wang et al., 25 GMV was found to mediate the association between an intelligence based PGS and cognitive function; this relationship was in turn moderated via ELA, based on data from the Adolescent Brain Cognitive Development study.
In addition to detrimental environmental factors that may moderate the association between genetic variation and cognition, the potential for mitigating factors to buffer the relationship between adverse early experiences and cognitive ability in adulthood, is also poorly understood.
In particular, greater years in education has been observed to have positive effects on both cognitive and general health outcomes in individuals exposed to early adverse experiences. [26][27][28][29] Consequently, the relationship between educational experience and cognitive ability may be causally bi-directional. 30,31 Confirming this, however, is made difficult by other genetic and environment factors that likely confound the relationship, including ELA, making it important to model these factors together as potential moderators in the relationship between genetic variation and cognition. Furthermore, understanding how ELA is associated with neurobiological mechanisms underlying cognitive ability and whether these effects can be targeted, that is, via educational attainment, has the potential to inform interventions for individuals exposed to ELA.
The purpose of the present cross sectional study was to examine the moderating role of ELA on the association between genetic variation, total GMV and cognitive ability using the UK Biobank. For this moderated mediation analysis, we used structural equation modelling and generated a latent factor of cognitive ability that was representative of cognitive domains of reasoning, processing speed, working memory and executive function. We hypothesised that the mediating effects of total GMV on the association between genetic variation and cognitive ability would vary depending on the severity of ELA experienced. Genetic variation was indexed using a PGS of Verbal-Numeric Reasoning, derived from a GWAS we carried out using a non-overlapping sample of 89,748 UK Biobank participants. Finally, we tested the hypothesis that greater educational attainment would at least in part, the effects of ELA on the association between total GMV, genetic variability and cognitive ability.

| Participants
The current study used data from the UK Biobank; a large epidemio-  Figure S1 ( Table 1).

| Cognitive measures
Cognitive tests were administered online on the same day as the MRI scan. As a measure of cognitive ability, we selected four cognitive tests: Verbal-Numerical Reasoning, Symbol-Digit Substitution, Matrix Reasoning and Trail Making to maximally capture important domains of cognitive ability including reasoning, processing speed, working memory and executive function. From this we constructed a latent variable of cognitive ability as previous work has shown improvement when combining these tests into a latent variable. 7,[32][33][34] The Verbal-Numerical reasoning test involved a series of 13 items assessing verbal and arithmetical deduction (Cronbach α reliability = 0.62). 35 The Symbol-Digit test, which is similar in format to the Symbol Digit modalities test, 36 involved matching symbols to single-digit integers and is a well validated measure of processing speed. The score was based on the number of correct Symbol-Digit matches made in 60 s. For the Numerical Memory test-a measure of working memory-participants were shown a two-digit number which they had to recall after a short pause. Numbers increased by one until the participant made an error or until they reached the maximum number of 12 digits. In the trail-making test part B, participants were presented with the numbers 1-13 and the letters A-L arranged pseudo-randomly on the screen. They were instructed to alternate between touching the numbers in numeric order and letters in alphabetical order (i.e., 1-A-2-B-3-C). As detailed by Salthouse 37

| Education
As a measure of education, UK Biobank participants were asked which of the following qualifications applied to them (with the option of selecting more than one), (1) 'college or university degree; (2) A levels or AS levels or equivalent; (3) O levels or GCSE or equivalent; (4) CSEs or equivalent; (5) NVQ or HND or HNC or equivalent; (6) Other professional qualifications, for example, nursing, teaching/ none of the above; (7) prefer not to answer'. Following the approach described by Rietveld et al., 39 we created a binary variable for education to index whether participants had obtained a college or university-level degree.

| Early life adversity
Childhood adversity items were based on the short version of the Childhood Trauma Questionnaire Short Form (CTQ-SF). 40 The CTQ-SF is a self-report questionnaire measuring physical abuse, emotional abuse, sexual abuse, physical neglect and emotional neglect. Items

| MRI Acquisition and analysis
MRI data were collected in a single Siemens Skyra 3 T scanner with a standard 32-channel hear coil located at UK Biobank's recruitment centre. T1-weighted MPRAGE data was acquired in the sagittal plane using a three-dimensional magnetization-prepared rapid gradientecho sequence at a resolution of 1 Â 1 Â 1 mm, with a 208 Â 256 Â 256 field of view. Global and regional brain imagingderived phenotypes (IDPs) were processed by the UK Biobank team and made available to approved researchers. Full details of the brain imaging protocols, and quality control (QC) measures have been made available. 41 For our study, we used a global brain IDP of total GMV, which had been extracted using FMRIB's Automated Segmentation Tool (FAST). 42 Scans of individuals with severe and visual normalisation problems were excluded by the UK Biobank through manual inspection (as noted in Alfaro-Almagro et al., 41 ).

| Genome-wide association analysis of intelligence
We performed a GWAS of intelligence using the Verbal-Numerical reasoning test from the UK Biobank (discovery sample N = 89,748).
We chose to conduct our own GWAS primarily to avoid sample T A B L E 1 Descriptive characteristics of the UK Biobank sample. We first assessed the mediating role of total GMV on the association between IQ-PGS and cognitive ability. Cognitive ability was defined as a latent construct using the four cognitive tests (see Section 2.2). We further examined whether this association was better captured by the cognitive latent variable or the Verbal-Numerical

Mean (SD) Range
Numerical Reasoning test alone. We next evaluated whether ELA would moderate (a) the association between IQ-PGS and cognitive ability, (b) the association between total GMV and cognitive ability and/or (c) the mediation effect. Finally, we investigated whether the structural model differed across educational groups. See Figure 1 for a schematic overview of the models.
The moderating effect of education (college/university degree vs. no college/university degree) was carried out using a multi-group SEM, which allowed for the estimation of measurement invariance (factor loadings and path coefficients) across the two groups. Given the sensitivity of the Δχ 2 to sample sizes, differences to reject invariance across specifications were based on the following indices and values only: ΔCFI ≤ 0.01 and ΔRMSEA ≤ 0.015 for both factor loadings and intercepts and ΔSRMR ≤ 0.03 for factor loadings and ≤0.01 for intercepts. 50,51 Throughout models tested, we corrected for age, sex and total intracranial volume (TIV).

| Is the relationship between the IQ-PGS and cognitive ability mediated by total GMV?
Testing our first hypothesis, which examined the mediating role of total GMV on the association between IQ-PGS and cognitive ability, we found that the data had two fit indices (TLI and SRMR) outside our pre- and without (13.5%) a college/university degree. Before testing for group differences, we tested for measurement invariance between the educational groups by (1) constraining the loadings (metric invariance) and (2) the loadings and intercepts (scalar invariance) to equality (tested with analysis of variance, Table 2). In all models, acceptable fit indices were observed and we found that metric and scalar invariance were held across groups. Thus, any observed differences in structural relations were not because of differences or errors in measurement and we proceeded to the multigroup analysis using our baseline model. Table 3, IQ-PGS and total GMV were directly associated with cognitive ability, while the relationship between IQ-PGS and GMV was nonsignificant in both groups. A significant moderating effect of ELA on the association between GMV and cognitive ability was observed in the group who had been to college/university (indirect effect β = 0.015, 95% CIs = À0.022; À0.008, SE = 0.004, p < 0.001). However, there was no evidence of a moderated mediation effect of ELA on the association between total GMV, IQ-PGS and cognitive ability in either group (see Table 3).

| DISCUSSION
Based on rich multivariate data from 5273 individuals from the UK Biobank, our study sought to model the complex relationship between genetic variation, GMV, ELA and education on cognitive ability. Consistent with previous studies, we found that genetic variation, as measured by IQ-PGS and total GMV were each independently predictive of cognitive ability. We did not find evidence that GMV mediated the association between genetic variation and cognitive performance. We further found that ELA was also significantly associated with cognitive performance, but that neither ELA nor years in education moderated the relationship between genetic variation and cognition, either directly or indirectly via GMV.

| Brain morphology and the relationship between IQ-PGS and cognitive ability
Previous studies have found that structural brain metrics are positively associated with cognition, 52 and that both share a common genetic basis. [9][10][11] In our study, we focused on total GMV given its consistent, albeit modest, association ($0.15-0.35) with cognition, and because cognitive ability is likely to involve multiple brain areas rather than one specific region. 1,7 In line with previous studies, IQ-PGS and total GMV were both predictive of cognitive ability. However, total GMV was not a significant mediator of the relationship between IQ-PGS and cognitive ability either alone or when moderated by environmental variables.
One interpretation of these findings is that although significant direct associations were observed between cognitive performance and both genetic variation (as measured by the IQ-PGS) and GMV, the modest amount of variation in cognitive performance explained by genetic variation may have limited our power to detect the mediating role of brain volume. Despite the sample size (>5000 individuals) available for the analysis, a more sensitive measure of brain structure and/or volume may have been required to identify its mediating role. Alternatively, while cognitive variation and GMV were associated, the underlying genetic variation shared between these phenotypes may be insufficient to confirm GMV as a significant mediator of the relationship between an IQ-PGS and cognitive variation. 4.2 | Environmental exposure as a moderator of the genetic and brain-related underpinnings of cognitive ability In addition to the associations observed between cognitive performance and both polygenic variation and brain volume, exposure to adversity experienced in early life was also significantly associated with cognitive performance. Consistent with previous studies from our group and others, greater exposure to ELA was associated with lower cognitive performance. Contrary to our hypothesis however, ELA was not observed to moderate the relationship between polygenic variance and cognitive performance either directly or indirectly via moderated mediation of brain volume. Educational experience was also not observed to be a significant moderator in the relationship between polygenic variation and cognition, although it was observed to moderate the relationship between GMV and cognitive performance.
Previous studies from our group and others have found evidence of association between ELA, structural variation in GMV and cognitive functioning in both clinical and non-clinical samples. 17,53,54 The present study differed from these previous studies by testing whether exposure to ELA represented a potential moderator of genetic effects rather than as an independent variable in its own right. Although this issue might be mitigated by a larger sample size, a more parsimonious conclusion is the explanatory power of the model tested was limited. If true, future studies may benefit from testing alternative models, for example, testing whether polygenic variation might moderate the effects of early environmental exposure to cognitive performance, either directly, or as mediated by a measure of brain volume.

| Limitations and strengths
The cross-sectional design of this study prevents us from drawing any firm conclusions about causality. Indeed, for the models tested here, one could equally speculate about whether individuals with lower cognitive ability and/or socioeconomic status are at greater risk for exposure to ELA, and it will be important for future studies to examine these associations longitudinally. Another potential limitation is that ELA was measured using retrospective self-reports. However, it is noteworthy that retrospective reports of ELA are shown to correlate moderately well with prospective measures and show comparable effects on negative life outcomes. 55 Future studies will benefit from the inclusion of a prospective assessment of ELA, as well as considering the developmental timing and frequency of trauma exposure, which may mediate the effects on cognitive functioning. 56 We did not examine other environmental risk factors (e.g., low birth weight and family income) nor did we include regional specific brain MRI volumes, which may also contribute to variation in cognitive ability. As such, future research should implement a fully data driven approach to include such factors. tions are worthwhile domain to consider in relation to adversity as they are recognised as an important protective factor for individuals exposed to ELA, promoting better stress and emotion regulation as well cognitive functioning. 57 Finally, while we controlled for the effects of age, the UK Biobank sample is restricted to middle and older age adults which potentially limits the generalizability of the results.
Our investigation of the relationship between IQ-PGS, ELA, education and brain morphology on cognition has important strengths, including the relatively large sample size and the inclusion of imaging and genetic data. Further, the SEM approach taken to characterise these associations expands on prior research, which has mostly used simpler regression approaches to analyse the complex relationships between biological and environmental predictors of cognition. The main advantages of SEM are that (1)

| Conclusions
This study sought to model the relationship between polygenic variation, environmental exposure, brain volume and cognitive performance in a large non-clinical sample. While polygenic variation, early adversity and brain volume were each observed to be independently associated with cognitive performance, the hypothesised moderated mediation model was not supported. Given the modest explanatory value of currently estimated PGS, we conclude that future studies modelling the relationship between these variables may benefit from consideration of polygenic variation as itself a moderator of other (e.g., environmental) factors rather than as a predictor variable which is itself moderated.