Association of Polymorphisms in Genes Regulating the Corticotropin-Releasing Factor System With Antidepressant Treatment Response FREE

Hypothalamic-pituitary-adrenocortical (HPA) system hyperactivity has been the most prominent and well-documented neuroendocrine change associated with major depressive disorder (MDD). This HPA hyperactivity is thought to be integral to the pathogenesis, course, and treatment of this disorder.^1- 2 Hyperactivity of the corticotropin-releasing factor (CRF; also termed corticotropin-releasing hormone [CRH]) and arginine vasopressin (AVP) neuronal systems seem to play a seminal role in the etiology and symptomatology of mood and anxiety disorders.^3- 13 In addition, considerable evidence suggests that normalization of the HPA system might be the final step necessary for stable remission of MDD. Successful antidepressant treatment has been shown to attenuate and normalize HPA-axis abnormalities.^14- 18 Finally, an open-label study suggested potential clinical efficacy for a CRH receptor type 1 antagonist in depressed patients,¹⁹ but findings have not been consistently replicated.²⁰

Given these data, it is likely that genetic polymorphisms that affect the function of genes that regulate this system may modulate an individual's response to antidepressant treatment. Indeed, 3 genes within the HPA-axis system have been reported to be associated with antidepressant response. Licinio et al²¹ reported an association of a 3-SNP haplotype within CRHR1 with response to desipramine or fluoxetine, which was replicated in a Chinese sample.²² Interestingly, in both samples, the association of CRHR1 variants was most pronounced in patients with anxious depression. Binder et al²³ found an association of genetic variants within the gene encoding the glucocorticoid receptor–regulating cochaperone of hsp90, FKBP5, with response to antidepressant treatment. This has been replicated in the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) sample and a German sample.^24- 25 Finally, van Rossum et al²⁶ described carriers of a rare functional polymorphism in the glucocorticoid receptor gene (ER22/23EK) as showing a faster response to antidepressant treatment.

The aim of this study was to investigate the association of polymorphisms in genes of the CRH and AVP system with remission and response to citalopram hydrobromide treatment in the STAR*D sample. To this goal, we genotyped tagging single-nucleotide polymorphisms (SNPs) in the genes encoding the ligands (UCN, UCN2, UCN3, and CRH), receptors (CRHR1 and CRHR2), and binding protein (CRHBP [OMIM *122559]) of the CRH system as well as the ligand (AVP) and the 2 behaviorally relevant receptors (AVPR1B and AVPR1A) of the AVP system.

The rationale, methods, and design of the STAR*D study have been detailed elsewhere.²⁷ In brief, investigators at 14 regional centers across the United States implemented a standard study protocol at 41 clinical sites.

Subjects provided separate written informed consent for study participation and for the collection of blood samples for genetic studies. Outpatients aged 18 to 75 years with a baseline Hamilton Depression Rating Scale score²⁸ of 14 or higher who met DSM-IV²⁹ criteria for nonpsychotic MDD were eligible as long as the treating clinician deemed antidepressant medication use to be appropriate. Exclusion criteria are summarized in the supplementary “Methods” section (available at http://www.archgenpsychiatry.com). The 16-item Quick Inventory of Depressive Symptomatology–Clinician-Rated (QIDS-C)³⁰ score was obtained at baseline and at each treatment visit to measure symptom severity. This study only investigates treatment outcomes from the first treatment step (level 1) in STAR*D, which required an adequate dose of citalopram.³¹ Citalopram was given for up to 12 weeks, and response was assessed weekly. No concomitant medications were allowed, except for benzodiazepines and sedatives/hypnotics, if needed. Patients could move to the next treatment level if they were nonresponsive at any visit. This point is referred to as the exit visit,³¹ with patients either substantially improved or moved to a different treatment within level 2.

DNA samples were collected from 1953 participants of the 4041 patients enrolled in STAR*D. For details on DNA extraction and quality control as well as a detailed comparison of the demographic characteristics of the 1953 patients who consented to providing a blood sample for DNA vs the other 2088 patients in the whole STAR*D sample, see the study by McMahon et al.³² The selection criteria of patients with DNA for this analysis are described below.

For phenotype definitions, we used the criteria for remission at the last visit of level 1 (exit visit), as proposed by McMahon et al. We only included patients who remained in level 1 for at least 4 weeks so that at the exit visit time, patients had at least 4 weeks of treatment and had remained an average of 11.4 weeks (SD, 3.0 weeks) in the study.

Remitters were those who achieved a QIDS-C score of 5 or less at the last treatment visit; nonremitters had a QIDS-C score of 10 or more at the last visit. Those with a final QIDS-C score in the borderline range of 6 to 9 were excluded from analysis. Thus, 38.2% of patients were nonremitters, 43.0% were remitters, and 18.8% were in the borderline range.

Responders achieved at least a 50% reduction in baseline QIDS-C score at the last treatment visit. Nonresponders did not achieve a 50% reduction in baseline QIDS-C score at the last treatment visit. According to this definition, 42.7% were nonresponders and 57.3% were responders.

As a secondary test, change in QIDS-C score from baseline to last visit was tested as a quantitative measure of change in depressive symptoms. Only 1 patient did not have both measures and was excluded from this analysis.

To test for possible interaction with anxious depression status, we used the definition described in the studies by Fava et al,^33- 34 with a baseline Hamilton Rating Scale for Depression (17-item scale) anxiety/somatization factor score of 7 or more considered to indicate anxious depression.

Similar to McMahon et al,³² we also controlled for drug tolerability to avoid misclassification of intolerant patients as nonresponders. Subjects were scored as tolerant, probably tolerant, intolerant, or probably intolerant on the basis of an algorithm that considered study exit data and the Global Rating of Side Effect Burden.³⁵ (More detail is available in the supplementary “Methods” section.)

Ten genes involved in the regulation of the AVP or CRH system were chosen. Table 1 lists the genes by chromosomal position and the number of selected SNPs. One hundred seventeen tagging SNPs were selected, as presented in the online Appendix.

Using the online tool provided by Applied Biosystems (ABlaunchSNPlex), 3 multiplexed SNP panels were designed for the SNPlex platform (Applied Biosystems) from 113 SNPs. Five SNPs failed the assay design algorithm (SNP panels could not be designed for 2 SNPs in AVP, 2 in UCN2, and 1 in CRHR1). A total of 108 SNPs were thus genotyped on the SNPlex platform using a standard protocol at the genetic core facility at University of California–Los Angeles. Four more SNPs, all in CRHR1, were genotyped using predesigned TaqMan assays (Applied Biosystems) on an ABI7900 HT Real Time PCR Instrument (Applied Biosystems). Overall, 76 (68%) SNPs reached a call rate greater than 90% in this initial experiment. Nearly 16% (15.7%) had a call rate less than 50%, and the average call rate of all the SNPs was 82%. A set of 30 DNA samples of the 1863 genotyped on SNPlex failed more than 80% of all SNPlex assays, indicating a potential problem with DNA quantity or solubility in this assay. The corrected average call rate of the SNPs reaching at least 90% is thus 94.8%. All 4 TaqMan assays were genotyped in 1897 individuals with a call rate ranging from 99.05% to 99.84%, with 100% concordance of 352 duplicate genotypes. Hardy-Weinberg equilibrium was examined separately for the 3 ethnic groups represented in STAR*D (European American, African American, and Hispanic). Three SNPs showed significant deviations from Hardy-Weinberg equilibrium in 2 or more populations and were thus also excluded from the analysis. eTable 1 lists the SNP names, gene names, positions on the chromosome, minor allele frequency (MAF), and P value for Hardy-Weinberg equilibrium for the 76 SNPs with a call rate greater than 90% that were included in the analysis. The genetic coverage of the SNP selection passing quality-control criteria (call rate ≥ 90% and in Hardy-Weinberg equilibrium, n = 73) is described in the online Appendix.

To control for the possibility of false-positive associations due to genotyping error, we regenotyped the 5 SNPs in CRHBP, with the strongest association using TaqMan-based assays. For these, we had 603 duplicate assessments for a total of 14 044 genotypes with 0 discordances. Call rates were over 99% for all 5 SNPs. For the 5 regenotyped SNPs, the discordance rate was 0.028 between SNPlex and TaqMan genotypes.

The individuals were recruited within a larger study to investigate the roles of genetic and environmental factors in predicting the development of posttraumatic stress disorder.^36- 37 Research participants were approached while in the waiting rooms of primary care or obstetrical-gynecological clinics of Grady Memorial Hospital in Atlanta, Georgia. Informed consent was obtained for all subjects, and all procedures in this study were approved by the institutional review boards of Emory University School of Medicine and Grady Memorial Hospital. In a subsample of 164 individuals of mostly African American descent (91.6% African American), a low-dose dexamethasone suppression test with a 0.5-mg dose was performed as described in the supplementary “Methods” section. Seven of the 164 probands had no detectable dexamethasone levels and were excluded from the analysis. Significance of the interaction effect of risk allele carrier status for rs10473984 (TT and TG vs GG) and depressive symptoms (cutoff of Beck Depression Inventory score ≥ 16, with 31.7% classified as currently depressed according to this cutoff) was determined using repeated-measures analysis of variance to examine the main effect of rs10473984 and depressive symptoms status (between-subject factor) and their interaction term on the change in serum cortisol and corticotropin from day 1 to day 2 (within-subject factor). All analyses were covaried for by age, sex, Childhood Trauma Questionnaire total score, and Posttraumatic Stress Disorder Symptom Scale total score.

We restricted our analysis to patients with at least moderate symptom severity as defined by a baseline QIDS-C score of 10 or more (30 patients had less severe baseline QIDS-C scores) and who remained in level 1 for at least 4 weeks (154 were at level 1 <4 weeks). Of the 1768 patients fulfilling these criteria, DNA was not available for 49 owing to quality-control issues, and another 47 samples were not genotyped in the SNPplex assay owing to 384-well plate layout, and thus cost reasons. This brought the number of patients entering the analysis with SNPplex genotypes to 1672 and for the TaqMan SNPs, to 1719. The initial analysis strategy was to test for allelic and additive genetic (genotypes coded as 0, 1, and 2) associations of the included 73 SNPs with remission and response at exit visit and change in QIDS-C score from baseline to exit visit. The initial analysis was performed in the combined samples, not stratified by race. Analyses were run using logistic or linear regression, adjusted for age and sex. We established the significance of main genetic effects for each SNP using permutation-based procedures that randomly assigned the outcomes (remitter-responder status and change in QIDS-C score) to subjects (sampled without replacement), while holding each subject's genotype fixed. This permutation method is robust against nonnormal distribution of the outcome variable and small cell sizes. For each analysis, the empirical P value was based on at least 10 000 permutations. We conducted these analyses using appropriate components of the SAS software system (version 9.1; SAS Institute, Cary, North Carolina) as described in previous studies.^36- 37

Secondary analyses then stratified the sample by racial category (European American, African American, and Hispanic) and used estimated proportion of African biogeographical ancestry as an additional covariate.

To assess interaction effects of SNPs with anxious depression on treatment response, we used a repeated-measures general linear model with QIDS score at the baseline and exit visits as the within-subject factor and presence of anxious depression, SNP genotype, and their interaction term as between-subject measures. Age, sex, and estimated proportion of African biogeographical ancestry were used as covariates. These analyses were carried out using SPSS, version 15.0.

For our main SNP analysis, we used Bonferroni correction for multiple testing, adjusting for the number of tested SNPs (n = 73) and phenotypes (remission, response, and change in QIDS score) as well as genetic model (additive and allelic). The adjusted α level was thus 0.05/(73 × 3 × 2) = .000114. In this article, we present the empirical P values as well as the adjusted Bonferroni-corrected P values for all analyses in the total sample.

Subjects were enrolled into the STAR*D study without regard to race or ethnicity; 65.5% described themselves as European American, 15.6% as African American, and 14.0% as Hispanic. To control for possible effects of admixture, we used 18 ancestry informative markers to estimate biogeographical ancestry from the 4 continental population groups (Appendix). Using STRUCTURE, version 2.0,³⁸ and a 4-population model, we calculated the estimated proportion of African ancestry for each individual and used this value as a covariate. Individuals missing genotype data for more than 3 of these markers were excluded from these analyses (8.8%).

Of the 73 tested SNPs, 7 showed nominally significant associations with remission at exit visit (Table 2). Of these, 3 SNPs were located in CRHBP, including the best SNP, rs10473984. This SNP showed P = .00044 (corrected, P > .05) for the additive test and P = .00062 (corrected, P > .05) for the allelic test. For the response at the exit visit, 6 SNPs showed nominally significant associations, with the 3 best P values observed for 3 CRHBP SNPs. The best SNP was again rs10473984, with an additive genetic value of P = .0067 (corrected, P > .05) and an allelic value of P = .0044 (corrected, P > .05). Thus, no SNP withstood correction for multiple testing with this phenotype. Five of the 6 SNPs associated with response were also within the 7 SNPs that showed nominal associations with remission (Table 2).

We performed a linear regression with the difference between baseline and exit visit QIDS scores as an outcome and the 73 SNPs as predictors, adjusting for sex and age. For this phenotype, the 4 most significant P values were seen for CRHBP SNPs. The best association was seen with rs10473984 (P = 6.56 × 10⁻⁵; corrected, P = .028), followed by rs10055255 (P = .0029; corrected, P > .05), rs28365143 (P = .0056; corrected, P > .05), and rs10062367 (P = .0145; corrected, P > .05). This included the 2 top SNPs for the association with remission (rs10473984 and rs10055255); the 2 other SNPs had not shown significant associations with remitter status.

None of the 73 tested SNPs showed a significant association with citalopram tolerability after correcting for multiple testing. For rs10473984, for example, the genotype distribution in tolerant individuals (n = 1269) was 85.74% GG, 13.08% GT, and 1.18% TT, and in untolerant individuals (n = 151), 82.78% GG, 15.23% GT, and 1.99% TT (P = .27).

We observed a clustering of SNPs associated with treatment outcome phenotypes in the CRHBP locus, with 4 SNPs at the 3′ end of the gene and rs28365143 at the 5′ end of the genes. The 4 SNPs at the 3′ end of the genes only display moderate linkage disequilibrium (Figure 1). Because the only association that withstood correction for multiple testing lies within this locus, we focused the remainder of the experiments on CRHBP.

To control for possible false-positive associations due to genotyping error, we regenotyped the top 5 SNPs in CRHBP using TaqMan-based assays. Here, the call rates were greater than 99% for all SNPs. Increasing the effective sample size in the analysis also increased the significance of the associations (with rs10473984 now associated at P = 1.0 × 10⁻⁵; corrected, P = .0043, additive model; and P = 6.0 × 10⁻⁶; corrected, P = .00043, allelic model with remission at exit visit). For response at the exit visit, the respective P values were now P = .0001; corrected, P = .043; and P = .0006; corrected, P > .05; and for the difference in QIDS score from baseline to exit visit, P = 1.0 × 10⁻⁵; corrected, P = .0043; and P = 7.0 × 10⁻⁷; corrected, P = .00031. For this SNP, the T allele is associated with poorer response to treatment, with an odds ratio (OR) of 1.88 (95% confidence interval [CI], 1.43-2.46) for remission and 1.29 (95% CI, 1.05-1.59) for response. The associations of the other CRHBP SNPs all remained nominally significant, but none passed correction for multiple testing (eTable 2).

We then repeated the analysis for association of these 5 SNPs with remission, response, and change in QIDS-C score in the split-sample design previously described,³² in which the STAR*D sample was randomly divided in a discovery (n = 587) and test (n = 1132) set. Of the 5 tested CRHBP SNPs, only rs10473984 showed associations, with P < .01 in both the discovery and test samples, with P = .0091 and P = .00006 in the additive model for remission at exit visit, P = .0003 and P = .0077 for response, and P = .0007 and P = .0002 for change in QIDS score from admission to exit visit in the discovery and test subsets, respectively. In both subsamples, the T allele was associated with poorer response (OR, 2.00; 95% CI, 1.22-3.29 in the discovery sample; and OR, 1.30; 95% CI, 1.01-1.68 in the test set for remission at exit visit) (eTable 2).

All of the 5 CRHBP-associated SNPs show significant differences in allele frequencies among the 3 ethnic groups, and poorest response to citalopram treatment overall in African American and Hispanic subjects has been reported.³¹ Our results in the combined sample could thus be due to population stratification, with more nonremitters in the 2 minority groups. However, when the analysis was repeated with ethnicity included as a covariate (European American, African American, and Hispanic), the association of rs10473984 with reduction of depressive symptoms remains significant after correction for multiple testing (P = 2.0 × 10⁻⁵; corrected, P = .0087).

eTable 2 shows the P values for the associations in the 3 subgroups. The associations with rs10473984 that withstood correction for multiple testing and are significant in both sets of the split sample are reported in more detail below (Table 3).

Figure 2 shows the changes in QIDS-C score from baseline to last visit for this SNP in all patients as well as separately for the 3 ethnic groups. None of the associations were significant in European Americans, but in both African American and Hispanic subjects, significant associations (additive model) were observed with remission (African American, P = .0076, and Hispanic, P = .012), response (P = .0073 and P = .0097, respectively), and change in QIDS-C score (P = .0065 and P = .0002, respectively). Of note, the T allele associated with poorer response is rare in European American and Hispanic subjects (MAF < 6%) but relatively common in African Americans (28.1%). While no effect is seen in European Americans (OR, 1.19; 95% CI, 0.77-1.84), there is a poorer response to citalopram treatment in T allele–carrying African American (OR, 1.82; 95% CI, 1.17-2.84) and Hispanic (OR, 4.88; 95% CI, 1.40-16.98) subjects (Figure 2 and Table 3).

Because rs10473984 is most common in populations of African descent and the degree of admixture of African ancestry within African American and Hispanic subjects could drive the association, we also controlled for estimated African ancestry using ancestry informative markers. Data describing the performance of the 18 ancestry informative markers are presented in the Appendix.

When the association of the CRHBP SNPs with treatment outcome was reanalyzed with the estimated proportion of African ancestry as a covariate, in addition to age and sex, associations with remission, response, and change in QIDS-C score remained significant for rs10473984 in the whole sample, as well for the African American and Hispanic subsamples (Table 3). For change in QIDS-C score, the corrected P value in the additive model was P = .0001 and would thus withstand correction for multiple testing (P = .043). See eTable 2 for other SNPs.

Because of the potential involvement of the CRH system in the pathophysiology of anxious depression, we investigated the interaction of the rs10473984 genotype and anxious depression on treatment response using a repeated-measures analysis on QIDS-C score at baseline and exit visits as the within-subject factor, controlling for age, sex, and estimated African ancestry in the whole sample as well as the combined African American and Hispanic subset. In the analyzed sample, 43% were categorized as having anxious depression. In the whole sample, we observed a significant main effect of rs10473984 genotype and anxious depression status. We also observed an additional significant interaction of change in QIDS-C score from baseline to exit visit with rs10473984 genotype and anxious depression (F_1421,2 = 4.82, P = .008). This interaction term was also significant when restricting the analysis to African American and Hispanic patients combined (F_397,2 = 3.67, P = .026). In both analyses, the association of the rarer T allele with poorer treatment response was much more pronounced in patients with anxious depression (Figure 3).

In an independent cohort of 157 mostly African American individuals, we tested the effects of rs10473984 T carrier status dependent on current depressive symptoms on low-dose dexamethasone suppression of serum cortisol and corticotropin using a repeated-measures analysis of variance, with day 1 and day 2 cortisol and corticotropin measures as the within-subject factors. For cortisol measures, we observed a significant change in cortisol levels from day 1 to day 2 (F_149,1 = 17.3, P < .001), a significant interaction of the change in cortisol and depression symptoms (F_149,1 = 4.07, P = .045), a significant main effect of sex (F_149,1 = 64.7, P < .001), and a significant interaction of current depressive symptoms and rs10473984 T carrier status (F_149,1 = 6.47, P = .012). In fact, depressed carriers of the rs10473984 GG genotype had higher overall cortisol levels compared with the other groups (Figure 4A). All other tested main effects and interactions were not significant.

For corticotropin measures, we observed a significant change in corticotropin serum concentration from day 1 to day 2, and significant interaction of the change in corticotropin with sex, age, and total Childhood Trauma Questionnaire score (P < .05 for all). We also observed a significant main effect of rs10473984 T carrier status (F_148,1 = 5.36, P = .022), a significant interaction of the change in corticotropin with rs10473984 T carrier status (F_148,1 = 4.77, P = .031), and a significant triple interaction of the change in corticotropin with rs10473984, T carrier, and depressive symptom status (F_148,1 = 4.27, P = .040). T allele carriers had overall higher corticotropin levels and showed more pronounced suppression following treatment with dexamethasone as a group. The enhanced dexamethasone suppression associated with this group is most pronounced when comparing individuals with a current Beck Depression Inventory total score greater or equal to 16 (Figure 4B).

In this study, we describe associations of SNPs within genes regulating the CRH system with remission and response to antidepressant treatment in the STAR*D sample. The association of 1 SNP located at the 3′ end of CRHBP (rs10473984), withstanding correction for multiple testing, is significantly associated in the discovery as well as the test set of the STAR*D sample and remains significant after correction for ethnicity and estimated proportion of African ancestry. This association is restricted to individuals with African American or Hispanic ethnicity, and this result seems to be predominantly carried by individuals with anxious depression. Furthermore, this variant is associated with glucocorticoid-receptor resistance and higher HPA-axis hormone levels. A hyperactive HPA axis at inpatient admission has been reported to correlate with better response to antidepressant therapy in male patients.³⁹

Although these results need independent replication, the association of rs10473984 is seen in the discovery as well as the test sets of the STAR*D sample, using a previously described random split approach.³² Furthermore, the T allele of this SNP was consistently associated with poorer response to treatment in 2 ethnic subsamples and showed nominally significant associations in the African American and Hispanic subsets of the sample, despite differences in MAF and linkage disequilibrium structure. While not significant in European Americans, the genotype distribution is in the same direction as for the 2 other groups. The T allele is associated with worse response and this is likely responsible for the strong association in the total sample. A much smaller effect size of this association in Europeans owing to differences in linkage disequilibrium structure and thus correlation with the putative functional variant might necessitate much larger samples to detect statistically significant associations in this subgroup.

Because of large differences in MAF between the ethnic subgroups, with the SNP being more common in African Americans (MAF = 28%) than the 2 other ethnic groups (MAF < 6%), and because poorer treatment response for minorities has been documented in this sample,³¹ we were concerned that population stratification could be a confounder of this association. This is not very likely the case, because in addition to independent significant associations in African American and Hispanic patients, the associations remained significant when controlling for ethnicity as well as for a genetic estimate of the proportion of African ancestry. Although we are relatively confident that our results are not a spurious finding owing to the increased proportion of African Americans or individuals with African admixture in the nonremitter group, our analyses cannot exclude more subtle effects of population stratification within the larger continental groups.

Another interesting aspect of the association of the CRHBP SNP rs10473984 with treatment response is its interaction with anxious depression status. In the STAR*D sample, the qualifier anxious depression is associated with a poorer treatment response overall,³¹ and data presented by Fava et al³⁴ suggest that it may be a valid diagnostic subtype of MDD, with distinct psychiatric comorbidities and clinical and sociodemographic features. Although rs10473984 was not associated with the presence of anxious depression itself (data not shown), we observed that for this SNP, the association with treatment response was mostly carried by individuals with anxious depression (43% of all STAR*D subjects³⁴). Two previous articles have shown that associations of variants in CRHR1 were also most pronounced in patients with anxious depression.^21- 22 A series of previous findings^1,40- 42 suggest that the CRH system is critically involved in both features of anxiety and depression, and it is thus possible that patients with anxious depression show higher activity of this system than patients with nonanxious depression. Genetic variants in genes regulating the CRH system may have a greater impact on treatment response in anxious depression.

CRHBP encodes the CRH-binding protein, which is present in the circulation and the interstitial spaces as a soluble 37-kDa glycoprotein that binds CRH and all urocortins with high affinity, reduces their bioavailability, and prevents their binding to CRH receptors.^43- 45CRHBP is highly expressed in a series of tissues, including the brain,^46- 48 thereby permitting its role in modulation of CRH neurotransmission.⁴⁷ De Luca et al⁴⁹ have already reported on an association of a variant within this gene with suicidal behavior in patients with schizophrenia.

rs10473984 is located at the 3′ end of CRHBP. At least in populations of African descent, the linkage disequilibrium structure would allow for this SNP to tag potential causal variants in the last intron and exon of the gene, which could influence splicing, change the coding sequence, or lead to differences in gene expression.⁵⁰ Using the HapMap gene expression and SNP genotype data,⁵¹ we could not find any significant effect of this SNP on CRHBP messenger RNA expression in lymphoblastoid cell lines (data not shown). Regulatory effects are known to be tissue specific, so this SNP might well be associated with CRHBP levels in other tissues. The fact that this SNP was associated with corticotropin levels (Figure 4B) could support a functional effect of this variant. The T allele, associated with poorer response to citalopram treatment, was also associated with higher corticotropin serum concentrations in depressed and nondepressed individuals. This might suggest that this allele is associated with reduced CRHBP expression and thus higher levels of free CRH, thereby increasing corticotropin secretion. In addition, individuals with clinically significant depressive symptoms carrying the GG genotype (associated with best treatment outcome) of this SNP showed the least degree of dexamethasone suppression of corticotropin. Previous studies have shown that depressed patients with dexamethasone nonsuppression of HPA-axis activation at treatment initiation have a beneficial treatment-response profile.³⁹

In conclusion, despite some limitations, such as the lack of a formal replication and incomplete genetic coverage, which was most pronounced in UCN and UCN2, our results support a role of the CRF system in treatment response to citalopram in patients with MDD and expand upon previous preclinical and clinical studies that demonstrated a central role of this system in the pathophysiology of depression and mechanism of action of antidepressants. Our results also support the notion that genetic variants in this system might be most relevant in predicting treatment response in anxious depression.

Correspondence: Elisabeth B. Binder, MD, PhD, Max-Planck Institute of Psychiatry, Kraepelinstr 2-10, 80804 Munich, Germany (ebinder@emory.edu).

Submitted for Publication: December 18, 2008; final revision received June 28, 2009; accepted August 12, 2009.

Author Contributions: Dr Binder had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Financial Disclosure: In the past, Dr Binder has received grant support from Pfizer and GlaxoSmithKline. Currently, Dr Binder receives grant support from the National Institute of Mental Health and the Doris Duke Foundation.

In the past 3 years, Dr Owens has had research grants from National Institutes of Health, Eli Lilly and Company, Pfizer, GlaxoSmithKline, Lundbeck, Cyberonics, Ortho-McNeil Janssen, and AstraZeneca. He has served as a consultant to H. Lundbeck A/S and received speaker's honoraria from Eli Lilly and Company. Dr Owens has a patent entitled “A method to estimate transporter occupancy.”

Dr Rush has received research support from the National Institute of Mental Health, the Robert Wood Johnson Foundation, and the Stanley Medical Research Institute; has been on the advisory boards and/or a consultant for Advanced Neuromodulation Systems, AstraZeneca, Best Practice Project Management, Bristol-Myers Squibb, Cyberonics, Eli Lilly and Company, Gerson Lehman Group, GlaxoSmithKline, Jazz Pharmaceuticals, Magellan Health Services, Merck, Neuronetics, Ono Pharmaceutical, Organon, Pam Lab, Personality Disorder Research Corp, Pfizer, The Urban Institute, and Wyeth-Ayerst Laboratories; has been on speakers' bureaus for Cyberonics, Forest Pharmaceuticals, and GlaxoSmithKline; has equity holdings (excluding mutual funds/blinded trusts) in Pfizer; and has royalty income affiliations with Guilford Publications and Healthcare Technology Systems.

Dr Trivedi has been a consultant for Abbott Laboratories, Akzo (Organon), AstraZeneca, Bayer, Bristol-Myers Squibb, Cephalon, Cyberonics, Eli Lilly and Company, Fabre-Kramer Pharmaceuticals, Forest Pharmaceuticals, GlaxoSmithKline, Janssen Pharmaceutica, Johnson & Johnson, Eli Lilly and Company, Meade Johnson, Neuronetics, Parke-Davis Pharmaceuticals, Pfizer, Pharmacia & Upjohn, Sepracor, Solvay Pharmaceuticals, VantagePoint, and Wyeth-Ayerst Laboratories; has served on speakers' bureaus for Abdi Brahim, Akzo (Organon), Bristol-Myers Squibb, Cephalon, Cyberonics, Eli Lilly and Company, Forest Pharmaceuticals, GlaxoSmithKline, Janssen Pharmaceutica, Pharmacia & Upjohn, Solvay Pharmaceuticals, and Wyeth-Ayerst Laboratories; and has also received grant support from Bristol-Myers Squibb, Cephalon, Corcept Therapeutics, Cyberonics, Eli Lilly and Company, Forest Pharmaceuticals, GlaxoSmithKline, Janssen Pharmaceutica, Merck, the National Institute of Mental Health, the National Alliance for Research in Schizophrenia and Depression, Novartis, Pfizer, Pharmacia & Upjohn, Predix Pharmaceuticals, Solvay Pharmaceuticals, and Wyeth-Ayerst Laboratories.

Dr Fava has received research support from Abbott Laboratories, Alkermes, Aspect Medical Systems, Astra-Zeneca, Bristol-Myers Squibb, Cephalon, Eli Lilly and Company, Forest Pharmaceuticals, GlaxoSmithKline, J & J Pharmaceuticals, Lichtwer Pharma GmbH, Lorex Pharmaceuticals, Novartis, Organon, PamLab, Pfizer, Pharmavite, Roche, Sanofi/Synthelabo, Solvay Pharmaceuticals, and Wyeth-Ayerst Laboratories. He has served on advisory boards and done consulting for Aspect Medical Systems, Astra-Zeneca, Auspex Pharmaceuticals, Bayer AG, Best Practice Project Management, Biovail Pharmaceuticals, BrainCells, Bristol-Myers Squibb, Cephalon, Compellis, CNS Response, Cypress Pharmaceuticals, Dov Pharmaceuticals, Eli Lilly and Company, EPIX Pharmaceuticals, Fabre-Kramer Pharmaceuticals, Forest Pharmaceuticals, GlaxoSmithKline, Grunenthal GmBH, Janssen Pharmaceutica, Jazz Pharmaceuticals, J & J Pharmaceuticals, Knoll Pharmaceutical Company, Lundbeck, MedAvante, Merck, Neuronetics, Novartis, Nutrition 21, Organon, PamLab, Pfizer, PharmaStar, Pharmavite, Precision Human Biolaboratory, Roche, Sanofi/Synthelabo, Sepracor, Solvay Pharmaceuticals, Somaxon, Somerset Pharmaceuticals, Takeda, TetraGenex, Transcept Pharmaceuticals, and Wyeth-Ayerst Laboratories. Dr Fava has served on speakers' bureaus for Astra-Zeneca, Boehringer-Ingelheim, Bristol-Myers Squibb, Cephalon, Eli Lilly and Company, Forest Pharmaceuticals, GlaxoSmithKline, Novartis, Organon, Pfizer, PharmaStar, and Wyeth-Ayerst Laboratories. He has equity in Compellis and MedAvante.

Dr Ressler reports receiving awards and/or funding support related to other studies from Howard Hughes Medical Institute, Lundbeck, Burroughs Wellcome Foundation, National Alliance for Research on Schizophrenia and Depression, National Institute of Mental Health, and National Institute of Drug Abuse, and has a consulting agreement with Tikvah Therapeutics for N-methyl-D-aspartate–based therapeutics.

Currently, Dr Nemeroff serves on the scientific advisory boards for Astra-Zeneca, Johnson & Johnson, Pharma Neuroboost, Quintiles, and National Alliance for Research on Schizophrenia and Depression. He previously served on the Forest Laboratories scientific advisory board. He has received grant money from the National Institutes of Health. He serves on the Board of Directors of AFSP, NovaDel Pharmaceuticals, Mt Cook Pharma, and the George West Mental Health Foundation. He owns equity in CeNeRx and Reevax. He owns stock or stock options in Corcept, Mt Cook Pharma, and NovaDel.

Funding/Support: This work was supported by GlaxoSmithKline; grants MH-90003, MH-077083-01, and MH-58922 from the National Institutes of Health; and the Doris Duke Clinical Scientist Development Award.