Background
Despite many attempts to understand the genetic architecture of depression, progress has been slow, particularly when compared with other complex disorders. The Psychiatric Genomics Consortium (PGC) recently identified 108 risk loci for schizophrenia [
1], whereas corresponding analyses performed for major depressive disorder (MDD) failed to find any robustly associated loci [
2]. More recently, however, studies have utilised data from the UK Biobank and 23andMe which resulted in discovery samples of over 100,000 participants of European ancestry [
3,
4]. The China, Oxford and Virginia Commonwealth University Experimental Research on Genetic Epidemiology (CONVERGE) consortium applied sparse whole genome sequencing to a sample of over 5000 cases of major depression and 5000 controls amongst Han Chinese women [
5]. These studies report finding several genetic variants associated with depressive symptoms measured in adults.
Both phenotypic and/or genetic heterogeneity may contribute to the slow progress in identifying genetic loci robustly associated with depression [
6]. Work has focused on dissecting the phenotype, with studies focusing on age at onset or subgroups including early onset or recurrent depression, in an attempt to understand the genetic architecture [
6‐
8].
Heritability is the proportion of variation in a phenotype attributable to genetic differences. A meta-analysis of twin studies estimated the heritability of MDD in adults to be 0.37 (95% CI: 0.31, 0.42) [
9], while recent work by the PGC used SNP-based methods implemented by the Genome-wide Complex Trait Analysis (GCTA) software [
10] to estimate a common SNP-heritability (h
2
SNP) of 0.21 [
11]. Although these estimates may at first appear inconsistent, the estimate from twin studies is expected to be higher as they take into account variation across the entire genome, whereas h
2
SNP estimates incorporate only common variants captured by SNPs included on genotyping platforms.
The majority of these studies of depression have been carried out in adults. However, a recent publication by Nivard et al. [
12] investigated symptoms of anxiety and depression across the lifecourse using data from the Netherlands Twin Registry (NTR), suggesting heritability might vary according to age. The study combined data from 49,524 twins whose ages ranged from 3 to 63. Heritability estimates calculated in this study showed fluctuations from early life until the early 20s (highest at 0.69, aged 3), after age 20 heritability remained relatively stable at around 0.5.
Nivard et al. [
12] found a substantial genetic component to childhood depression, with a distinct drop in heritability from 0.58 to 0.37 at age 12. Estimates increased again throughout adolescence and reached a peak of 0.53 at age 18. This is an interesting contrast to findings from many other twin studies which suggest a minor genetic component to depression in early life and increasing heritability with age, the heritability in late adolescence being comparable with that of adult depression [
13‐
17]
. However, the drop in heritability found in the Nivard study might be confounded by a switch in reporting method, as depression measures switched from parent- to self-reported questionnaires.
In addition to coinciding with a switch in reporting method, the timing of this change in heritability is interesting as it coincides with pubertal onset. At age 12, many participants are likely to be nearing the onset of puberty, if they have not entered it already. Pubertal status and depression are highly associated, with a 1 year prevalence of depression of around 1% in pre-pubertal children, rising to 4–5% in mid to late adolescence [
18]. Prior to puberty the prevalence of depression is similar in both males and females, however after pubertal onset this rises more rapidly among females. This has prompted some to hypothesise that changes during puberty, or the timing of pubertal onset, could be related to the onset of depressive symptoms [
19‐
21]. Although mechanisms are unclear, biological theories suggest earlier puberty and the hormonal changes associated with this, such as increasing levels of dehydroepiandrosterone (DHEA), may act as a trigger for some underlying vulnerability [
22,
23]. Alternatively, psychosocial theories hypothesise that depression is a social consequence of the early onset of puberty compared to peers [
24].
We aim to investigate patterns of heritability across adolescence using data from the Avon Longitudinal Study of Parents and Children (ALSPAC), which has both parent- and self-reported measures of depressive symptoms available at regular intervals during this important developmental period.
ALSPAC is a prospective birth cohort started in the early 1990s, which holds information on a number of measures, including depression and pubertal status. This includes a combination of parent- and self-reported depression measures at a variety of time points, which enabled us to investigate the change in heritability found by Nivard et al. [
12]. Genotyping data are available on a large number of ALSPAC participants, enabling us to estimate h
2
SNP directly at different ages using restricted maximum likelihood (REML) methods and to attempt to identify genetic variants associated with depression measures across a range of ages.
Discussion
This study aimed to build on work by Nivard et al. [
12] by attempting to unpick the heritability of depression across adolescence and investigating the role of puberty. Unlike the Nivard study, which found that heritability decreased substantially at age 12, our estimates, although lower, appeared to peak close to this time at age 13 (h
2
twin: 0.37 vs h
2
SNP:0.17). Although Nivard et al. [
12] offered several explanations for this, including puberty, they noted that it coincided with a switch in questionnaires from parent to self-report, acknowledging that the decrease may be an artefact of the change in reporting method.
Nivard et al. [
12] found increased heritability estimates when using parent-reported depressive symptoms, however, in ALSPAC these estimates were consistently lower than for self-reported symptoms. Regardless of rater, and in both raw and transformed data, we found a peak in heritability around age 13. With raters only overlapping at one time point, it seems likely that the drop in heritability seen in the Nivard paper is related to change in reporting method.
Within ALSPAC both parent and self-reported depressive symptoms are available, enabling us to remove this source of confounding from estimates, however, confounding according to timing of puberty remains. We investigated this by stratifying on pubertal status, enabling us to explore patterns of heritability according to age and pubertal status. However, our sample size was reduced somewhat when running these exploratory analyses and restricting to those with data on puberty, and no firm conclusions could be made.
We dichotomised pubertal stage into pre- and post-pubertal onset according to Tanner stage. At age 11 around 30% of males had reached puberty, compared to approximately 40% of females. At this age, heritability estimates were highest among males who had reached puberty (h2 = 0.81, p = 0.156) and pre-pubertal females (h2 = 0.39, p = 0.134). Given that males generally experience puberty later than females, it is possible that these differences in heritability are related to differences in timing of puberty and changes in environmental rather than genetic risk. One explanation would be that the social consequences of earlier puberty are opposite for males and females, with earlier puberty being protective in boys reducing environmental risk and thereby increasing the relative contribution of genetic risk. While earlier puberty could increase environmental risk in females, hence reducing the relative contribution of genetic risk at this age.
When stratifying analyses on sex, regardless of pubertal status, heritability was increased for males at age 13 (h2 = 0.30) and females at 18 (h2 = 0.18). This increased male estimate appears to be driving the overall peak in heritability (h2 = 0.17) at 13, although a sex-stratified GWAS found no SNPs achieving genome-wide significance.
We had good power when using the entire sample (power at age 11:84%; 13:78%; 18:42%) to detect a true heritability of 17%. Despite little statistical evidence for non-zero heritability at most ages, the confidence intervals estimated for age 11 exclude 21%, suggesting that our results demonstrate h2
SNP at this age is lower than that previously reported for adult depression. Though this could be explained by phenotype definition, rather than biological differences per se, as we used self-reported measures of depressive symptoms rather than detailed diagnostic interviews. When performing stratified analyses, power will be diminished further, however, the pattern of results observed generate interesting hypotheses that could be followed up in larger datasets.
While Nivard et al. [
12] found high heritability estimates among their younger age groups, estimates at age 11 in our study do not overlap with these. However, it is important to remember SNP-based and twin heritability are different. We would expect twin estimates to be higher simply due to the study design, which relies on closely related individuals. Some of this inflation in h
2 is real, due to shared rare genetic variants, whereas some is due to shared common environment, and therefore false inflation of the estimates. Our analyses are restricted to unrelated individuals in an attempt to prevent inflation of our h
2
SNP estimates due to shared environment. Additionally, our analyses incorporate only variants captured on the genotyping platform used, and is an estimate of the heritability we could potentially estimate through GWAS.
Although we expect to see higher estimates from Nivard et al. [
12] in general, we can compare the patterns of heritability across time in the two studies. Nivard et al. [
12] report a decline in heritability during childhood and adolescence, however the results shown in ALSPAC do not support this. Our results are relatively stable from age 11–18, with the exception of a peak at around age 13.
It is also important to consider selective drop out when thinking about these patterns of heritability. Those with increased depressive symptoms may be more likely to drop out, leading to declines in heritability. However, no robust evidence of association between a genetic risk score for depression and dropout was found in our sample (p = 0.67).
A GWAS in ALSPAC found evidence of association between a variant on chromosome 7 (rs138191010) and depressive symptoms at age 13 (when we also see greatest heritability). However, this was not replicated in TRAILS (which found weak evidence of an association in the opposite direction) or TEDS (which found no robust evidence of an association). Both ALSPAC and TEDS used the SMFQ to assess depression, while the Youth Self Report was used in TRAILS. When restricting to just ALSPAC and TEDS, which used a consistent measure of depression, there remained no strong evidence of an association. It is therefore possible that this is a false positive.
Limitations
Although our analyses, in particular those stratifying on pubertal status, may lack power, the patterns observed provide interesting hypotheses surrounding the heritability of depression at different developmental stages, and provide pilot data that can be utilised to design well-powered studies to investigate these hypotheses.
Acknowledgements
We are extremely grateful to all the families who took part in this study, the midwives for their help in recruiting them, and the whole ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists and nurses.
We gratefully acknowledge the ongoing contribution of the participants in the Twins Early Development Study (TEDS) and their families.
This research is part of the TRacking Adolescents’ Individual Lives Survey (TRAILS). Participating centers of TRAILS include various departments of the University Medical Center and University of Groningen, the Erasmus University Medical Center Rotterdam, the University of Utrecht, the Radboud Medical Center Nijmegen, and the Parnassia Bavo group, all in the Netherlands. We are grateful to everyone who participated in this research or worked on this project to and make it possible.