We identified an interaction between the TT genotype of rs1360780 and childhood physical abuse on the severity of present adult depressive symptoms. Carriers of the TT genotype who had been exposed to childhood physical abuse showed a mean BDI-II score of 17.4 compared with 10.0 for exposed subjects without the TT genotype. The effect of the primary analyses also survived Bonferroni correction for the testing of the five exposure dimensions. This result was confirmed for lifetime diagnosis of major depression according to DSM-IV. The lifetime prevalence of MDD was 23.9% in the exposed CC/CT carriers compared with 66.7% in the exposed TT carriers. In and and Supplementary Table S1a, it is clearly indicated that the interaction effect between rs1360780 and childhood physical abuse is represented by homozygote carriers of the T allele.
To explore the negative interaction effects in the other abuse and neglect dimensions, we increased the severity of childhood abuse and neglect to the most severe category. This ‘severe' category is characterized by a higher frequency of traumatic experiences and/or more reported situations of abuse and neglect compared with the ‘moderate' category. Although the number of affected subjects was further decreased, the results indicated that emotional and sexual abuse in the ‘severe' category also interacted with rs1360780 (p<0.05; Supplementary Table S3a). Comparing the group of subjects with at least one severe abuse category vs subjects with none, mild, moderate categories also demonstrated significant interaction with rs1360780 (p=0.007; Supplementary Table S4). These results indicate that all abuse dimensions with increasing severity contribute to the interaction effects. It is noteworthy that correlations between the five CTQ subdimensions were low to moderate (r=0.11–0.56; see Supplementary Table S5). The neglect dimension did not show a comparable interaction effect. Thus, the FKBP5 pathway seems to be more specific in mediating the psychobiological effects of abuse-related stress.
Our results clearly extend previous findings on the association of FKBP5 and mood disorders examined in clinical populations (Binder et al, 2004
; Lekman et al, 2008
). Especially when gene–environment interactions contribute to the pathophysiology of a distinct genetic marker, non-replications of direct gene effects could occur because of low rates of childhood abuse exposure in the investigated samples (Gawlik et al, 2006
; Papiol et al, 2007
). Likewise, in our sample, no direct association between rs1360780 and depression was observed.
In secondary interaction analyses, we additionally investigated three intronic SNPs from the FKBP5 gene with different LDs (r2) to rs1360780: 0.09 for rs7757037, 0.41 for rs4713899, and 0.83 rs9368881 (Supplementary Figure S1). All four SNPs are widely distributed over the FKBP5 gene and cover different regions of the gene (Supplementary Figure S2). The less-linked SNP rs7757037 showed no evidence of interaction whereas the other two SNPs also indicated interactions with abuse dimension (Supplementary Table S1a). However, the AA genotype of rs4713899 included only one exposed subject. Thus, none of these two SNPs was superior to rs1360780.
Our finding could be of future relevance in the identification of subjects at high risk for depression after childhood abuse. Especially, the large effect sizes in BDI-II scores and in absolute risk of MDD suggest that the TT genotype of rs1360780 could be included in prediction models for subjects at high risk for depression. Based on the effect sizes in this study, a positive predictive value of 70% for the TT genotype is found in subjects with childhood physical abuse and prevalence of 30–35% of MDD (Supplementary Figure S4). However, the true value of a TT genotype screening in exposed subjects should be assessed longitudinally.
It is noteworthy that our results are in line with previous studies, associating the high-induction T allele of rs1360780 with an altered cortisol response to experimental stress (Luijk et al, 2010
; Ising et al, 2008
), with faster response to antidepressant treatment (Binder et al, 2004
) and with the susceptibility to PTSD after childhood maltreatment (Xie et al, 2010
; Binder et al, 2008
). According to the pathophysiological model, the high-induction alleles of the FKBP5
gene are associated with a relative GR resistance (Binder et al, 2008
; Binder, 2009
), which could facilitate a long-lasting dysregulation of the HPA axis after childhood abuse and thereby increasing the risk for MDD.
As rs1360780 was associated with the diagnoses of PTSD in former studies (Binder et al, 2008
; Xie et al, 2010
), analyses were also performed excluding all subjects with PTSD (n
=29) and adjusting for PTSD (see Supplementary Table S2). All interaction effects remained largely unchanged, indicating that our findings are valid for depression. To minimize a putative effect of cumulative adult traumatization on the interaction effects, we also adjusted all analyses for the number of traumatic events from the PTSD interview section. This adjustment did not change the interaction results. However, in both cases, PTSD and the number of adult traumata could be a causal mediator and no confounder (Rothman, 2002
). In conclusion, PTSD and adult traumata did not account for the observed interaction effect between physical abuse, rs1360780, and depression in our study.
One has to discuss putative limitations of our study. The overall number of exposed TT carriers (n
=9) was small. To address this problem, statistically bootstrapping methods were applied for BDI-II scores. Moreover, to assess the effect of single observations on the statistic of interest, we performed detailed regression diagnostics for the diagnosis of MDD (Hosmer and Lemeshow, 2000
). Four subjects with the largest impact on the statistical model were excluded separately each and all together from the model. Although the 95% CI is wide, detailed model diagnostics indicate that the interaction itself was not dependent on a single observation (see Supplementary Tables S6 and S7). Moreover, the method applied is known for yielding relative large CIs, thus we have chosen a rather conservative approach (Nie et al, 2010
In order to avoid major selection effects (eg, selected treatment samples) that may bias the results, we recruited the subjects from the general population (SHIP). From the baseline study (SHIP-0) that started in 1997, only 55.7% could be included in the LEGEND study (2007–2010). Importantly, we have no evidence that lifetime depressive symptoms at baseline had influenced the later participation in LEGEND.
One strength of our study is that we analyzed two outcomes of depression: a dimensional depression score (BDI-II) and the lifetime diagnosis of major depression. There are obvious clinical differences between a self-report measure covering the last 2 weeks before the interview and an interview-based lifetime diagnosis of MDD (DSM-IV). We analyzed the statistical association between both end points. The area under the ROC curve was 0.69, which indicates a weak relationship between BDI-II scores and MDD (Hosmer and Lemeshow, 2000
). Thus, it is justified to consider MDD and BDI as two different measures.
In conclusion, our results support the role of the high-induction TT genotype of the rs1360780 polymorphism of the FKBP5 gene in the susceptibility to depressive disorders in the light of adverse, especially traumatic childhood events.