Our finding that the TAT haplotype in CRHR1
reduced the risk of major depression among AA women who experienced one or more ACEs extends findings from some prior studies in non-substance-dependent populations to a predominantly substance-dependent one. Our findings are consistent with those originally reported by Bradley et al. 
and may be limited to AAs by virtue of the different haplotype structure among AAs and EAs. Notably, not all studies have replicated the effects reported by Bradley et al. 
, possibly due to differences in the study populations. The populations studied have been of African or European ancestry, in varying proportions. Heim et al. 
, extended the AA sample initially studied by Bradley et al. and found a moderating effect of a “tag SNP” for the TAT haplotype (rs110402) only among men. In an initial sample of women of European ancestry, Polancyzk et al.  found a significant moderating effect of the TAT haplotype, but failed to replicate the finding in a second sample of European ancestry comprised of both sexes. Grabe et al. 
, in a German population sample, found that the presence of the TAT haplotype increased
the risk of depression. After correction for multiple comparisons, only one SNP (rs17689882) of the 16 examined by Grabe et al. 
significantly moderated the effect of physical neglect on depressive symptoms. This SNP, as well as the haplotype block in which it was located (i.e., Block 3), were found to be protective. Of note, these Block 3 SNPs all effectively tag the H2 haplotype and would be expected to be in complete LD. It is thus surprising that the level of significance reported for rs17689882 exceeded those found for other H2-tagging SNPs by an order of magnitude. The findings reported by Grabe et al. 
, which run counter to those of most other reports, may reflect in part the confounding effects of different haplotype blocks in CRHR1
. Although the Block 3 (H2) haplotype and the TAT haplotype (located in Block 1 and not present in H2) showed modest linkage disequilibrium (R2
=0.21), moderator analyses did not take the relative contributions of these two haplotypes into account.
Studies in this area have also used different instruments to measure ACEs and have examined different kinds of ACEs. The Childhood Trauma Questionnaire [Bernstein et al. 1994
], a validated measure of ACEs, was used by Bradley et al. 
, Grabe et al. 
, and in one of the two samples studied by Polanczyk et al. 
. We used a measure of self-reported ACEs for which reliability data are available [Douglas et al. 2010
], but the validity is not established. An important limitation of this measure of ACEs is that it does not assess differences in the timing and severity of the events, which could have important effects on the risk associated with them. Although Grabe et al.  found that genetic moderation was limited to physical neglect, our most robust finding was for exposure to (i.e., having witnessed or experienced) violent crime. We obtained similar effects for both physical and sexual abuse, although, perhaps due to the fact that these ACEs were not endorsed as frequently as violent crime, the effects did not reach statistical significance. Although the four subgroups examined by us differed in the proportion of individuals endorsing the different kinds of ACEs, these differences do not appear to explain why we observed genetic moderation of the risk of MDE only among AA women.
Another source of variability in the findings reported in the literature on the moderating effect of variation in CRHR1
on depression risk is the specific measure of depression used. We used a dichotomous lifetime MDE as the dependent measure in our analyses, which is most similar to the measure used by Polanczyk et al. 
, who defined depression in terms of a DSM-IV major depressive disorder, assessed over the preceding year. These dichotomous diagnoses contrast with the studies by Bradley et al. 
(and the extension of that study by Heim et al. 
and Ressler et al. 
) and Grabe et al. 
, which used the Beck Depression Inventory, a dimensional approach, to measure current depressive symptoms.
Differences in socioeconomic status among the four groups in the present report do not appear to explain the observed effects on risk of depression, since the inclusion of age, education level, and income level as covariates in the logistic regression analyses did not substantially alter the findings. However, because only among AA women was there a correlation between having experienced an ACE and the number of TAT haplotypes, it is possible that the observed interaction in this group between the number of TAT haplotypes and ACE on risk of depression could reflect a statistical, rather than a biological effect.
Findings reported by Tyrka et al. 
support a biological basis for the observed moderating effect of CRHR1
variation on risk of MDE. These investigators found a moderating effect of the CRHR1
SNPs rs110402 and rs242924, which together with rs7209436 constitute the 3-SNP Block 1 haplotype, on a history of childhood maltreatment, using the dexamethasone/corticotropin-releasing hormone (DEX/CRH) test. In this study, both SNPs (which were in near-complete linkage disequilibrium) moderated the effect of maltreatment on cortisol response to the DEX/CRH test. The authors suggested that excessive HPA axis activation could represent a mechanism of interaction of risk genes with stress in the development of mood disorders. Heim et al. (2009)
found that the moderating effect of rs110402 on the cortisol response to the DEX/CRH test was present only in men, consistent with their findings on the risk of depressive symptoms. Recent evidence points to multiple brain regions being involved in the effects of stress, including not only HPA-mediated effects, but also CRH neurotransmission in, for example, the amygdala [Gallagher et al. 2008
, Wang et al. 2010
]. Further research on the specific mechanisms by which stress and variation in CRHR1
interact to produce depression could help to refine the pathophysiology of the disorder.
The findings reported here may also have implications for the treatment of depression. Liu et al. 
examined the moderating effect of three CRHR1
SNPs (rs1876828, rs242939 and rs242941) on the response to 6 weeks of treatment with fluoxetine in 127 Han Chinese patients with MDD. Individuals who were homozygous for the G allele at rs242941 and those with two copies of the GAG haplotype comprised of these three SNPs showed a more robust fluoxetine therapeutic response, an effect that was limited to individuals with high levels of anxiety. If replicated, these findings may help to identify individuals for whom SSRI therapy may be most efficacious.
In summary, we found evidence that the TAT haplotype, variants limited to the H1 inversion haplotype moderated the risk of MDE in AA women exposed to ACEs and extend the findings of a protective effect of this haplotype on risk of depression to a largely substance-dependent population. We found no moderating effects of the TAT haplotype on the risk of AD. These findings add to a growing, but complex and inconsistent, literature on the moderating role of this variation in depression and are consistent with prior negative findings in relation to AD [Treutlein et al. 2006
, Blomeyer et al. 2008
, Nelson et al. 2010
]. Differences in the measurement of adverse childhood experiences and depression, as well population differences could contribute to the variable results. Nevertheless, taken together, these findings suggest a differential impact of variation in the two large non-recombinant haplotypes containing CRHR1
among individuals who have experienced ACEs: namely, that variation moderating the risk of depression may be localized to the H1 haplotype and the one moderating risk of AD may be localized to the H2 haplotype. A model of differential vulnerability associated with the H1 and H2 haplotypes is supported by the lack of recombination across the two haplotypes (Stefansson et al. 2005
). Further, other differential effects have been seen with the two haplotypes, including ongoing positive selection of the H2 haplotype (Stefansson et al. 2005
) and a propensity for microdeletions, which have been associated with a risk of developmental delay and learning disability (Shaw-Smith et al. 2006
). Further research to test the hypothesis that the H1 and H2 haplotypes are differentially associated with risk of depression and AD is warranted.