Background
Depression and anxiety conditions are an increasing problem, leading to substantial economic and social consequences [
1]. The prevalence of depression in pre-adolescents is approximately 1% [
2] and there is an increase in rates of various psychiatric problems during adolescence and adulthood in individuals with a history of childhood depression [
3,
4]. Moreover, childhood depression has shown a risk factor pattern different from that of adolescent and adult onset depression, raising the question whether childhood depression is etiologically separate from the latter [
5,
6].
The longitudinal approach serves as an excellent way of studying risk factors, development and course of psychiatric disorders and symptoms. Many factors may elevate the risk for onset of depression and emotional problems in children. Postpartum depression has been shown to predict child behavior problems in numerous studies [
7,
8], but there is also support for the fact that on-going maternal depression exerts a risk factor for internalizing as well as externalizing symptoms in children [
9‐
11].
Among other known risk factors for depression and emotional problems are traumatic life events [
12] - children exposed to physical abuse, parental divorce and domestic violence are at increased risk of behavior problems and adult depression [
13‐
15]. Still, not everyone subjected to traumatic life events develops depressive symptoms. Hence, there may be individuals predisposed or more vulnerable to certain risk factors.
In recent years, much effort has been given to explaining depression in relation to genetic factors. Many candidate genes have been examined, of which the serotonin transporter gene (
5-HTT) is the most studied. The
5-HTT gene includes a functional polymorphism, the serotonin transporter gene-linked polymorphic region (
5-HTTLPR), which consists of two common alleles; short (s) and long (l). Less effective serotonin expression and availability have been shown in s-allele carriers, compared to individuals homozygous for the l allele [
16]. The gene-by-environment model put up by Caspi
et al.[
17], showing an association between the
5-HTTLPR, stressful life events and depression, has since publication been subjected to numerous attempts at replication. Some researchers have been able to reproduce the results partially or completely [
18‐
24], whereas others have not [
25‐
28], however between-study heterogeneity has to be taken into account. When looking particularly at
5-HTTLPR gene-by-environment studies on children, research is not that extensive and results have been conflicting. Araya
et al.[
7] found no association between
5-HTTLPR and emotional symptoms in seven-year-olds whereas associations have been shown by other researchers, although on smaller study samples [
20,
22,
29,
30]. Moreover, also direct association between
5-HTTLPR and depression has been found [
16,
31‐
34].
Given not only the inconsistency of single genetic explanatory models, but also the complex etiology of depression, researchers have lately been looking at gene-by-gene (−by-environment) models for an explanation of depressive vulnerability. Another candidate gene for depressive disorders is the Brain Derived Neurotrophic Factor (
BDNF), which has previously been shown to act in synergy with serotonin [
35].
BDNF is involved in reparation, plasticity and neurogenesis in the brain, and a single nucleotide polymorphism (SNP) G/A (Val66Met) in the
BDNF gene has been shown to affect levels of
BDNF in the brain [
36]. Similarly to
5-HTTLPR, studies regarding the association between
BDNF Val66Met and depression have been both confirmative [
37,
38] and negating [
39].
A candidate gene-by-gene-by-environment interaction effect of
5-HTTLPR-by-
BDNF Val66Met-by-childhood adversity on depression has been shown [
40] and to our knowledge four studies have attempted replication of this three-way interaction effect, however with contradictory results [
41‐
44]. The findings are inconsistent regarding both the presence of a three-way interaction effect and the risk genetic variants in the presence of childhood adversity. Two studies found evidence of a three-way interaction effect [
42,
44] whereas two other studies did not [
41,
43].
In view of this, the aim of the present paper was to study the gene-by-environment interaction on depressive symptoms in 12-year-old children. We hypothesized to find a gene-by-environment, and possibly gene-by-gene-by-environment interaction on depressive symptoms, including the genetic polymorphisms 5-HTTLPR and BDNF Val66Met, traumatic life events and maternal symptoms of depression.
Discussion
This study used data from a longitudinal birth cohort study in order to examine predictors of internalizing and externalizing symptoms in 12-year old children. Environmental as well as genetic factors were included in order to test for gene-by-environment and gene-by-gene-by-environment interactions. The results of the study can be summarized in three main findings.
First and foremost, maternal symptoms of depression and anxiety predicted child behavior problems at age 12. Numerous researchers have been able to show the importance of maternal mental health for child development and wellbeing. Our results indicate that in the present population, maternal symptoms of postpartum depression (analyzed with different cut off scores on the EPDS) did not have a long-term impact on child behavior, i.e., no increase of risk was seen at age 12. Symptoms of depression in mothers at the 12-year follow-up could possibly reflect recurrent or chronic depressive problems, which would put the child under greater stress than would be the case with a single episode depression. In clinical practice, we suggest that besides focusing on the child’s mental health, it is important also to ask about depressive symptoms in the mother.
Second, we found a main gene effect of
5- HTTLPR on child behavior at age 12. Since the finding of an interaction between
5-HTTLPR and stressful life events [
17], focus of research has shifted to gene-by-environment interaction effects. The present finding does have support in the literature [
16,
31‐
34], however most gene-by-environment studies have not shown a main effect of
5-HTTLPR on depression [
20,
21]. Recently, Duncan and Keller reviewed the literature on candidate gene-by-environment interaction (cGxE) studies in psychiatry over a decade (2000–2009), and based on the analyses of 103 studies suggested that “well-powered direct replications deserve more attention than novel cG×E findings and indirect replications” [
58]. The current study attempted testing a previous hypothesis and replicating previous findings in a large and longitudinal population-based independent study/sample. The set-up of the current study can help to understand the generalizability of previous findings since it used a virtually similar set-up compared to previous studies with regard to phenotypic variables, genetic polymorphisms, statistical model, environmental moderator, and inclusion of both sexes [
58]. This represents a major strength of the current study which might contribute to the drawing of clearer conclusions in the context of future meta-analytical studies. The present study did not find any GxE or GxGxE effects but rather a main effect of
5- HTTLPR on internalizing symptoms as for example showed by Hoefgren et al. in a case–control study on depression [
31].
Third, traumatic life events showed an impact on child behavior in bivariate models, but not in the multivariate analyses, indicating that in this material, life events did not explain as much of the variance as other factors. We used the mothers’ reports on traumatic life events experienced by the children. Previous studies have shown differences between parents’ reports and children’s self-reports on exposure to violence [
59]. Considering this, self-reports might have contributed to a more fair estimation of traumatic life events experienced by the child. Given the known relationship between life events and depressive symptoms, one would have expected an effect on the internalizing scales [
12,
15,
17].
We included measure of externalizing problems since previous research has shown high comorbidity between depressive and externalizing disorders in adolescents [
60] and since boys more often express their psychological problems or stress by showing externalizing problems [
2]. Furthermore, Laucht
et al.[
61] reported that adolescents with depressive disorders were more likely to have a history of externalizing problems. As expected, the analysis showed that boys expressed significantly more externalizing problems than girls, but there were no differences relating to sex for internalizing problems.
Despite the dropout level, the number of its participants and the longitudinal approach strengthens the study. There are some limitations that need to be mentioned. Although dropout rates of the same magnitude in longitudinal, multi wave studies are common [
7], the skewed dropout in this study is a limitation - psychosocially disadvantaged families and families of foreign origin were less likely to participate in the follow-up. Considering the known connection between psychosocial strain and behavior problems, it is plausible to think that a more representative participating population would have strengthened the results.
In general, both child and parent-reports indicated low frequencies of behavior problems. Reports on good mental health in 12-year old children have been published previously. Costello
et al.[
2] examined mental health in 9–16 year old children using parent’s reports on child behavior (CBCL) and clinical interviews, and found the lowest prevalence of depression in 12-year olds (0.4%). Conducting the follow-up at ages eight-ten or during adolescence, where previous reports have shown more behavior problems, might have increased the prevalence rate and thereby rendering a larger group for different analyses.
Since there are studies indicating that depressed mothers are prone to overestimate behavior problems in their children [
62], another limitation of the present study is that mothers’ reports, not self-reports or teachers' reports, on child behavior and traumatic life events were used. We decided to use mothers as informants since they have known their children for 12 years around the clock compared to the current teachers that have known the children for about just over a year 5–6 hours a day, five days a week. Another support for this strategy was that although the association between maternal depressive symptoms and a positive mother–child reporting discrepancy has been demonstrated, in the general population this association is small and does not bias research, using maternal reports on child problems [
62]. The chosen methodology used in the study also allowed following data from the same informants from baseline to the 12-year follow-up.
Feder, Nestler and Charney’s report of 2009 called for an increased understanding of the psycho-biological factors which are involved in risk and resiliency for psychiatric disorders [
63], and Duncan and Keller‘s review called for well-powered direct replications of cG×E findings [
58]. Thus, many studies are still to be carried out to shed light on the psychogenetic underpinning of depression.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
SA was responsible for the data collection, and for writing the manuscript. EC was responsible for the genetic analyses, and made substantial contribution to the writing of the manuscript. MB was responsible for the statistical analyses, and took part in writing the manuscript. LD took part in data collection and writing the manuscript. CGS and GS planned and supervised the research project. LO took part in planning the project and supervised the genetic analyses. All authors took part in reviewing draft versions of the manuscript, and approved of the final version.