Variation in Catechol-O-Methyltransferase Is Associated with Duloxetine Response in a Clinical Trial for Major Depressive Disorder

Variation in Catechol-O-Methyltransferase Is Associated with Duloxetine Response in a Clinical Trial for Major Depressive Disorder Roy H. Perlis, Bonnie Fijal, David H. Adams, Virginia K. Sutton, Madhukar H. Trivedi, and John P. Houston Background: The study objective was to evaluate variations in genes implicated in antidepressant mechanism of action for association with response to duloxetine treatment in major depressive disorder (MDD). Methods: We assessed response over 6 weeks in 250 duloxetine-treated Caucasian patients in a randomized, double-blind study of patients with MDD. Single nucleotide polymorphisms (SNPs) were genotyped in 19 candidate genes selected based on evidence for involvement in antidepressant mechanism of action. Primary analysis examined baseline to end point reduction in the 17-item Hamilton Depression Rating Scale (HAMD17) total score, using a set-based test for association for each gene. Follow-up analyses examined individual SNPs within any significant gene for association with reduction in HAMD17 and 30-item Inventory of Depressive Symptomatology-Clinician Rated (IDS-C-30). Results: After correction for multiple comparisons, only COMT was associated with change in HAMD17 (experimentwide p ⫽ .018). Peak association was detected with rs165599 (p ⫽ .006), which accounted for approximately 3% of variance in HAMD17 change and ⬎4% of variance in IDS-C-30 change (p ⫽ .001). The least-squared mean change (SE) in HAMD17 score by rs165599 genotype was ⫺10.8 (1.2), ⫺8.7 (.6), and ⫺6.5 (.7) for patients with GG, GA, and AA genotypes, respectively. For SNPs in serotonin 2A receptor (HTR2A) previously associated with citalopram response, including rs7997012, no significant evidence of association with duloxetine response was identified. Conclusions: Single nucleotide polymorphisms in COMT were associated with symptom change in duloxetine-treated patients with MDD. If replicated, the magnitude of the COMT genotype effect is of clinical relevance. Key Words: Antidepressant, genetic association, genomics, response, single nucleotide polymorphism

M

any patients with major depressive disorder (MDD) do not experience full remission with antidepressant treatment, and patients frequently respond differentially to different antidepressants. Clinical trials provide information on how treatments work on average for a given population but are not typically informative about which individual patients will respond. The low percentage of patients that respond to a given antidepressant, e.g. 27.5% remission to a selective serotonin reuptake inhibitor (SSRI) (1-3), make such individualized prediction desirable to minimize the time patients are receiving ineffective or intolerable medications. Small studies suggest that outcome of antidepressant treatment is partially familial (4), indicating that genetic variation may be informative about antidepressant treatment outcome. In this regard, several recent association analyses of candidate gene variations with treatment response have been performed for MDD patients who were treated with the SSRI antidepressant citalopram in the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) study and were prospectively assessed

From the Department of Psychiatry and Center for Human Genetics Research (RHP), Massachusetts General Hospital, Boston, Massachusetts; Eli Lilly and Company (BF, DHA, JPH), Indianapolis, Indiana; i3 Research (VKS), Cary, North Carolina; and University of Texas Southwestern Medical Center (MHT), Dallas, Texas. Address reprint requests to Roy H. Perlis, M.D., Bipolar Clinical and Research Program and Center for Human Genetic Research, Massachusetts General Hospital and Harvard Medical School, Boston, MA 02144; E-mail: [email protected]. Received May 27, 2008; revised September 12, 2008; accepted October 1, 2008.

0006-3223/09/$36.00 doi:10.1016/j.biopsych.2008.10.002

(5-7). An initial genetic screen of variants in 68 candidate genes for response and remission to citalopram in the STAR*D cohort identified association between treatment response and a variant in the serotonin 2A receptor (HTR2A) gene (5). While antidepressants that inhibit the reuptake of both norepinephrine and serotonin (serotonin-norepinephrine reuptake inhibitors [SNRIs]) are increasingly used in clinical practice, few association studies have specifically examined response to treatments other than SSRIs (8). To evaluate variations in genes implicated in antidepressant mechanism of action for association with response to the serotonin-norepinephrine reuptake inhibitor duloxetine in MDD, we assessed response over 6 weeks in 250 duloxetine-treated Caucasian patients in a randomized, double-blind study of patients with MDD. We examined variations in genes coding for candidate monoamine receptor and other putative antidepressant sites of action using a gene-based approach.

Methods and Materials Gene Selection and Genotyping Methods Candidate genes were selected by the authors based on evidence for involvement in antidepressant mechanism of action. Single nucleotide polymorphisms (SNPs) were identified based on search of the Single Nucleotide Polymorphism database (dbSNP), prioritizing those with known (i.e., nonsynonymous coding SNPs) or putative (i.e., promoter SNPs) function. In general, these SNPs captured the majority of variation, with minor allele frequency of 5% or greater in most genes of interest. Genotyping was performed by Cogenics (Newton, Massachusetts) using Sequenom MassARRAY iPlex (Sequenom, San Diego, California) as previously described (9). This SNP-based approach does not allow for examination of the potential effect of other sources of genetic variation such as the serotonin transporter promoter polymorphism (5HTTLPR). All BIOL PSYCHIATRY 2009;65:785–791 © 2009 Society of Biological Psychiatry

786 BIOL PSYCHIATRY 2009;65:785–791

R.H. Perlis et al.

Table 1. Baseline Characteristics

Characteristic Age, Years, mean (SD) Age of Onset, Years, mean (SD) Number of Previous Episodes, mean (SD) HAMD17 Score, mean (SD) Gender, Male, n (%) First Mood Episode, n (%) Atypical Depression, n (%) Melancholic Features, n (%) History of Suicidality, n (%) Prior Suicide Attempt, n (%) Hospitalized for Suicidality, n (%)

Caucasian Genotyped ITT (N ⫽ 250)

Non-ITT a (N ⫽ 109)

p

44.2 (12.6) 31.2 (14.0)

42.0 (12.9) 29.7 (12.6)

.133 .390

4.7 (12.0) 18.6 (5.9) 35 (32.1) 21 (28.8) 9 (12.3) 40 (54.8) 20 (27.4) 14 (70.0) 11 (55.0)

.187 .070 .393 .451 .049 .542 .153 .707 .749

3.4 (4.7) 17.2 (5.7) 92 (36.8) 61 (24.4) 14 (5.6) 147 (58.8) 49 (19.6) 32 (65.3) 29 (59.2)

ITT, intent to treat; SD, standard deviation; HAMD17, 17-item Hamilton Depression Rating Scale. a Includes unsuccessfully genotyped patients and patients not in the intent-to-treat population (ITT) (n ⫽ 51) and genotyped non-Caucasian patients not included in analysis (n ⫽ 58).

individual samples were successfully genotyped. Of 818 SNPs for which genotyping assays were requested, 615 assays were successfully developed and validated (54 assays failed development and 149 failed validation due to the following reasons: inconsistency with Hardy-Weinberg Equilibrium [HWE] [43 assays], call rate less than 80% [49 assays], or minor allele frequency ⫽ 0 [57 assays]). Of the 615 SNPs with validated assays, 552 SNPs were successfully genotyped in at least 90% of individuals (i.e., 63 SNPs had a call rate less than 90%); 184 of these SNPs were excluded prior to analysis because of minor allele frequencies (⬍5%), resulting in 368 SNPs passing quality control.

Patient Population Patients were a subgroup of a randomized, double-blind clinical study of MDD as defined by the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text-Revised (DSM-IV-TR). This subgroup consented to donate a blood sample for genetic analysis. Genotype association analysis was limited to Caucasian patients to minimize the effects of population stratification. Additional details of the primary clinical study are available elsewhere (10). Briefly, patients were required to have a minimum total score of 15 on the 17-item Hamilton Depression Rating Scale (HAMD17) at baseline. After an initial randomization to duloxetine 30 mg (once daily [QD] or twice daily [BID]) or 60 mg (QD) for 1 week, all patients received 60 mg QD for the remaining 5 weeks of the acute treatment period used for this association analysis. All three treatment arms were also randomly assigned to one of two food groups (i.e., study drug taken or not taken within 1 hour of eating). The study design also included a 2-week placebo double-blind, variable expected-duration placebo lead-in period so that the start of active therapy was blinded (in actuality, patients received 1 week of placebo lead-in). Concomitant psychotropic medications, including benzodiazepines, were not allowed during the analysis period. For the cohort of patients who donated samples for genetic analysis, 80.8% of patients were overall compliant with their assigned treatment based on daily pill count. Study drug compliance was defined for each visit as taking between 80% and 120% of prescribed capsules. Patients were considered overall compliant for the study if they were compliant at all nonmissing visits. The study protocol was approved in accordance with the principles of the Declaration of Helsinki and with good clinical practices. For study entry, patients were required to have a level of understanding sufficient to provide informed consent and to communicate with the investigators and site personnel. All

Table 2. Gene-Based Analysis for Association with Depressive Symptom Change in Duloxetine-Treated Patients for 19 Candidate Genes Gene

pGenewise

p Experimentwise

Most Significant SNPs

# SNPs in Gene-Based Analysis

.209 .174 .001 .604 .147 .107 .895 .808 .284 .153 .022 .226 .047 .131 .205 .123 .140 .240 .055

.989 .975 .018 1.00 .954 .881 1.00 1.00 1.00 .977 .350 .993 .592 .926 .994 .916 .939 .996 .654

rs488323-rs526302-rs2644627 rs908867-rs10835210-rs2049045 rs165599-rs165774-rs174696 rs762551-rs2470890-rs11854147 rs5996117-rs5996119-rs28360521 rs167771-rs10934256-rs3773678 rs755698 rs1334894-rs4713916-rs3800373 rs2178865 rs878567-rs6295-rs749099 rs9534505-rs1923884-rs2760351 rs6437000-rs765458-rs4973377 rs10789980-rs11214800-rs10891613 rs8192532-rs3790756-rs4912138 rs4464147-rs10853245-rs4308014 rs258747-rs6196-rs6198 rs3785151-rs998424-rs36023 rs2020942-rs2066713-rs3813034 rs211105-rs4537731-rs684302

4 9 19 4 8 23 1 4 1 5 41 7 12 7 7 3 38 21 3

ADRA1A BDNF COMTa CYP1A2 CYP2D6 DRD3 FGFR2 FKBP5 GRIK1 HTR1A HTR2A HTR2B HTR3A HTR6 MC2R NR3C1 SLC6A2 SLC6A4 TPH1

ADRA1A, adrenergic, alpha-la-, receptor; BDNF, brain-derived neurotrophic factor; COMT, catechol-O-methyltransferase; CYP1A2, cytochrome P450, subfamily 1, polypeptide 2; CYP2D6, cytochrome P450, family 2, subfamily D, polypeptide 6; DRD3, dopamine receptor D3; FGFR2, fibroblast growth factor receptor 2; FKBP5, FK506 binding protein 5; GRIK1, glutamate receptor, ionotropic kainate 1; HTR1A, serotonin 1A receptor; HTR2A, serotonin 2A receptor; HTR2B, serotonin 2B receptor; HTR3A, serotonin 3A receptor; HTR6, serotonin receptor 6; MC2R, melanocortin 2 receptor; NR3C1, glucocorticoid receptor; SLC6A2, neurotransmitter transporter, nonadrenaline; SLC6A4, neurotransmitter transporter, serotonin; TPH1, trytophan hydroxylase 1; SNP, single nucleotide polymorphism. a COMT was the only gene with SNPs significantly associated with depressive symptom response at the experimentwise level.

www.sobp.org/journal

BIOL PSYCHIATRY 2009;65:785–791 787

R.H. Perlis et al. Table 3. Individual COMT SNPs Associated with Depressive Symptom Change Least-Squared Mean Change (SE) and n SNP (A/B)

AA

AB

BB

R2

p

Contiguous Position

rs165737 (T/C) rs165774 (A/G) rs174696 (C/T) rs174697 (A/G) rs165599 (G/A)

⫺6.09 (1.01) n ⫽ 31 ⫺6.00 (1.03) n ⫽ 28 ⫺9.76 (1.22) n ⫽ 18 ⫺12.68 (2.53) n⫽1 ⫺10.84 (1.16) n ⫽ 18

⫺7.52 (.52) n ⫽ 101 ⫺7.53 (.52) n ⫽ 105 ⫺8.57 (.61) n ⫽ 77 ⫺10.13 (1.24) n ⫽ 30 ⫺8.67 (.56) n ⫽ 105

⫺8.95 (.66) n ⫽ 107 ⫺9.06 (.66) n ⫽ 106 ⫺7.38 (.59) n ⫽ 141 ⫺7.58 (.51) n ⫽ 210 ⫺6.51 (.67) n ⫽ 107

.02428

.016

3104651

.02605

.013

3104711

.02000

.030

3105326

.01803

.038

3105982

.03254

.006

3108931

COMT, catechol-O-methyltransferase; SE, standard error of the mean; SNP, single nucleotide polymorphism.

patients provided written informed consent after the study procedures and possible side effects were fully explained. Patients provided a separate informed consent to provide a blood sample for genetic analysis. The full protocol was provided to the investigators’ ethical review boards as part of initial protocol review. The study was conducted at 33 sites in the United States. Analytic Methods Clinical Analyses. Caucasians in the intent-to-treat analysis cohort (n ⫽ 250) were compared with excluded subjects (n ⫽ 109) on baseline sociodemographic and clinical features using appropriate parametric tests (i.e., chi-square test or unpaired t test). Subject-level data from the entire clinical cohort (n ⫽ 647) were not available for comparison because of institutional privacy policies for genetic studies. To maximize power to detect associations, the primary analysis of efficacy pooled all randomized treatment arms, as previous analyses of the clinical cohort as a whole failed to detect significant differences in efficacy or tolerability (10). Whereas the protocol-specified primary outcome was the incidence of nausea, for the genetic association analysis, continuous change in HAMD17 from baseline to end point was used as the primary outcome measure to attempt to maximize statistical power and for consistency with most other duloxetine efficacy reports (11). Gene-Based Analyses. As suggested by Neale and Sham (12) for analysis of association studies, we considered the unit of analysis in this study to be the single gene, rather than single SNP. Therefore, primary analysis screened for association using the set-based test, implemented in PLINK 1.00 (http://pngu.mgh.harvard.edu/⬃purcell/ plink/) (13) for each gene. This approach has previously been used in an association study of bipolar disorder (14), and simulation studies show it to have greater power to detect associations if multiple SNPs within a gene are associated with a phenotype, with little cost if only one SNP is associated (S. Purcell, Ph.D., unpublished data, personal communication, February 2008). In brief, this test, analogous to that proposed by Ott and Hoh (15), computes the test statistic for each individual SNP within a gene, then calculates the average test statistic for the best single SNP per region, for the best two SNPs per region, and for the best three SNPs per region. The significance of these set statistics is then estimated by permutation, which allows a determination of genewise significance, allowing for correlation between SNPs and tests, while controlling type I error at the single-gene level. For these analyses, significance of SNP combinations, including between one and three SNPs, were estimated using 20,000 permutations. To further control type I error, permutation was used to account for all tests in all genes

(experimentwise p values). For the set-based tests, where pairs of SNPs were in linkage disequilibrium with r2 ⬎ .8, only 1 of the 2 SNPs was included in analysis, yielding 219 SNPs. Where a gene met this threshold (experimentwise p ⬍ .05), all single SNP associations in that gene with efficacy were then examined. The assessment of effect of genotype on changefrom-baseline-to-end point HAMD17 scores was evaluated using the p value for an overall difference between genotype means (AA vs. AB vs. BB) at 6 weeks from a mixed-effects repeated measures (MMRM) analysis of variance (ANOVA), with terms for visit, genotype, genotype by visit interaction, and baseline score as a covariate. For individual SNPs that were significantly associated with continuous change in HAMD17, remission rates as defined as end point, HAMD17 score ⱕ 7 and response rates as defined as ⱖ 50% reduction in HAMD17 score from baseline to end point, were compared by genotype. To attempt to distinguish effects of efficacy from overall effects on time on treatment, including tolerability, single SNP associations with time on protocol were also examined using the log-rank test. Finally, for exploratory purposes, we examined HAMD17 subfactors, including the core depressive subscale (items 1, 2, 3, 7, and 8), anxiety/somatization subscale (items 10, 11, 12, 13, 15, and 17), and the sleep subscale (items 4, 5, and 6), and the 30-item Inventory of Depressive Symptomatology-Clinician Rated (IDSC-30) total score (16) using the same MMRM model as described for the HAMD17 total score.

Results Subjects and Clinical Assessment Genotyping was completed for 250 Caucasian duloxetinetreated patients. Patient and illness characteristics are shown in

Figure 1. Least-squared mean change by visit for the 17-item Hamilton Depression Rating Scale by rs165599 genotype.

www.sobp.org/journal

788 BIOL PSYCHIATRY 2009;65:785–791

R.H. Perlis et al.

Table 1. Patient age, sex, and illness characteristics of the intent-to-treat Caucasian population used for the association analysis were similar to that of patients not included in the association analysis (genotyped non-Caucasian patients and Caucasian patients not in the intent-to-treat group).

pression and sleep factors (Table 4). The direction of genetic association was consistent for these SNPs across the HAMD17 subfactors and IDS-C-30 scale. To distinguish that the COMT SNPs associated with response were due to an effect on efficacy and not a more generalized effect, including tolerability, we examined the association of single COMT SNPs with time in the study (i.e., dropout). None of the COMT SNPs examined were associated with early study dropout (p ⱖ .30). The COMT SNPs associated with duloxetine response appeared to be in two linkage blocks (Figure 2). The most significant SNP, rs165599, is reportedly located 3= to the 3= untranslated region (UTR) of the gene and is in strong linkage disequilibrium (LD) with SNP rs174697, located in intron 5 before the last exon. The other three significant SNPs are located in intron 5. Of these, SNPs rs165774 and rs165737, which are associated with less response in contrast to the other three SNPs, are in a haplotype block that appears to have modest LD with the other 3= block. Single nucleotide polymorphism rs174696 is physically between these two linkage blocks but is not as strongly linked. The serotonin 2A receptor (HTR2A) and serotonin 3A receptor (HTR3A) were significantly associated with depression response to duloxetine at the genewise correction level but not at the experimentwise level. In addition, SNPs in HTR2A previously associated with citalopram response (rs7997012 and rs1928040) were not associated with mean symptom change in duloxetinetreated patients, though other HTR2A SNPs showed modest evidence of association (rs9534505, rs1923884, rs2760351; genewise p ⫽ .022). In linear regression models, we detected no significant evidence of epistatic effects between COMT rs165599 and the HTR2A SNPs, including rs7997012.

Genotyping and Genetic Analyses From 19 candidate genes, 368 SNPs met quality control criteria and had minor allele frequencies ⬎5% and were used for analysis. Response Association Catechol-O-methyltransferase (COMT) was the only gene significantly associated with depressive symptom change in duloxetine-treated patients at the experimentwise permutation level (Table 2). When individual COMT SNPs were examined for association with HAMD17 score change, 5 out of 29 genotyped COMT SNPs were associated with depressive symptom change in duloxetine-treated patients (uncorrected p ⬍ .05) (Table 3). Single nucleotide polymorphism rs165599 had the strongest statistical significance (p ⫽ .006) and least-squared mean change in HAMD17 score by genotype for this SNP at each visit as shown in Figure 1. Change in HAMD17 scores for each rs165599 genotype had clearly separated by 1 week of treatment, and this difference persisted and increased over the remaining 5 weeks, with patients with the GG (n ⫽ 18) genotype having the greatest response. In an exploratory analysis, depressive symptom remission (HAMD17 ⱕ 7) and response (ⱖ 50% reduction in HAMD17) appeared more common in the responsive COMT genotypes (Supplements 1 and 2, respectively). Remission rates by rs165599 genotype (GG, GA, AA) were 61.1% (11/18), 45.7% (48/105), and 38.3% (41/107) (p ⫽ .090) and response rates by rs165599 genotype (GG, GA, AA) were 72% (13/18), 54% (57/ 105), and 44% (47/107) (p ⫽ .024). The HAMD17 subfactors and the IDS-C-30 were analyzed for association with the three most significant COMT SNPs to assess which particular symptom domains were underlying the association with depressive response. All three COMT SNPs were significantly associated with score change on the IDS-C-30 and two of the three were significantly associated with change on the HAMD17 core de-

Discussion Catechol-O-methyltransferase was the only gene of the 19 examined genes significantly associated with depressive symptom change in duloxetine-treated patients at the experimentwise level. The association of symptom response with COMT SNPs was strongest when the IDS-C-30 was considered and was not

Table 4. COMT SNP Associations with Factor Scores for the Most Significant SNPs Associated with HAMD17 Total Score Least-Squared Mean Change (SE) SNP (A/B) IDS-C-30 rs165774 (A/G) rs174696 (C/T) rs165599 (G/A) HAMD17 Core Depression rs165774 (A/G) rs174696 (C/T) rs165599 (G/A) HAMD17 Anxiety Factor rs165774 (A/G) rs174696 (C/T) rs165599 (G/A) HAMD17 Sleep Factor rs165774 (A/G) rs174696 (C/T) rs165599 (G/A)

AA

AB

BB

R2

p

⫺13.11 (1.08) ⫺16.12 (.97) ⫺17.56 (1.08)

⫺15.95 (.85) ⫺13.84 (.99) ⫺13.50 (.91)

⫺18.80 (1.69) ⫺11.56 (2.00) ⫺9.43 (1.87)

.03501 .02741 .04381

.004 .011 .001

⫺2.74 (.31) ⫺3.23 (.28) ⫺3.78 (.31)

⫺3.29 (.24) ⫺2.98 (.29) ⫺2.73 (.26)

⫺3.84 (.48) ⫺2.73 (.58) ⫺1.69 (.54)

.01883 .007772 .03933

.034 .178 .003

⫺2.69 (.25) ⫺3.11 (.22) ⫺3.15 (.25)

⫺2.98 (.20) ⫺2.65 (.23) ⫺2.75 (.21)

⫺3.27 (.39) ⫺2.20 (.45) ⫺2.35 (.43)

.00997 .02177 .008966

.125 .024 .153

⫺1.60 (.20) ⫺2.05 (.18) ⫺2.20 (.20)

⫺2.07 (.16) ⫺1.79 (.18) ⫺1.80 (.17)

⫺2.54 (.31) ⫺1.53 (.37) ⫺1.39 (.35)

.02467 .0164 .0101

.015 .0499 .129

COMT, catechol-O-methyltransferase; HAMD17, 17-item Hamilton Depression Rating Scale; IDS-C-30, 30-item Inventory of Depressive SymptomatologyClinician Rated; SE, standard error of the mean; SNP, single nucleotide polymorphism.

www.sobp.org/journal

R.H. Perlis et al.

BIOL PSYCHIATRY 2009;65:785–791 789

Figure 2. Linkage disequilibrium between 29 catechol-O-methyltransferase (COMT) single nucleotide polymorphisms (SNPs) for which assays were developed and validated. Linkage determined by r2 and calculated with Haploview (http://www.broad.mit.edu/node/443) (30). The five SNPs in boxes were associated with 17-item Hamilton Depression Rating Scale (HAMD17) score change in duloxetine-treated patients (uncorrected p ⬍ .05) (graphic generated by Haploview).

accounted for solely by HAMD17 anxiety or sleep factors, suggesting that COMT genotypes were predominately associated with core depression response. Depressive symptom remission (HAMD17 ⱕ 7) also appeared more common in the responsive COMT genotypes. Peak association was detected with rs165599, which accounted for approximately 3% of variance in HAMD17 change and ⬎4% of variance in IDS-C-30 change, a relatively large effect for a single genetic variation involving a complex phenotype. The least-squared mean change in HAMD17 score for the most responsive rs165599 genotype (GG) was ⫺10.8 compared with ⫺6.5 for patients with the least responsive AA genotype. This magnitude of difference appears similar to that of the difference between duloxetine and placebo in other studies of patients with mild baseline depression (11). For example, in the studies examined by Shelton et al. (11), mean change from baseline to 8 weeks in HAMD17 in patients with mild depression was ⫺6.13 for 40 mg to 120 mg duloxetine-treated patients compared with ⫺4.7 for placebo. Thus, the magnitude of the genetic effect may be of clinical relevance if replicated, ideally in a placebocontrolled study. However, the clinical utility of a single SNP to predict response is likely limited.

Catechol-O-methyltransferase is located at chromosome 22q11 and catalyzes the transfer of a methyl group from S-adenosylmethionine to catecholamines, including the neurotransmitters dopamine, epinephrine, and norepinephrine, resulting in one of the major degradative pathways of the catecholamine transmitters. The most studied COMT SNP is commonly referred to as val158met (rs4680). This G-to-A transition in codon 158 of (membrane-bound [MB]-) COMT converts a valine (val) high-activity allele to a methionine (met) low-activity allele, resulting in a threefold to fourfold reduction in COMT activity (17). The met allele is also associated significantly with decreased brain enzyme levels and lower enzymatic activity (18). The met allele is associated with improved (compared with the val allele) cognitive performance in prefrontal cortex tasks, such as various working memory tests (19-21) and potentially schizophrenia (19,22). Yoshida et al. (23) did not find a significant difference in response rates by COMT val/met for the SNRI Milnacipran in 81 Japanese patients, although carriers of the met allele did respond significantly faster. In addition, the val158met variation was associated with response to mirtazapine but not paroxetine in 102 patients in a German study (24). In our study, although other SNPs within the COMT gene were associated with www.sobp.org/journal

790 BIOL PSYCHIATRY 2009;65:785–791 response, val158met was not statistically associated with response to duloxetine (p ⫽ .141). In addition, the significant SNPs in our study appeared to be in separate haplotype blocks from the val158met SNP, suggesting independent association from that functional allele. Other COMT SNPs have not been studied as extensively. The SNP with the highest degree of statistical significance in our study was rs165599. This SNP, which is located 3= to the end of the 3= UTR in the COMT gene, has been studied in several previous reports. This SNP had the strongest individual significance of a 3-SNP haplotype, which was associated with schizophrenia (25) and associated with smoking cessation with bupropion (26). Although this SNP has been described as located 3= to the 3= UTR, Bray et al. (27) reported that it is found in transcripts expressed in human brain with significant differences in allelic expression, suggesting an additional alternative transcript(s) in brain. A recent computation study of microRNA binding sites also suggested that the A-allele significantly lowered expression of the longer 3= UTR COMT transcript in human prefrontal cortex (28). In the largest antidepressant pharmacogenetic study to date, 768 variants in 68 candidate genes were examined for response and remission to citalopram in the STAR*D cohort (5). That study implicated rs7997012 in HTR2A in citalopram response, while finding no strong associations with SNPs in COMT. In our study, rs7997012 was not associated with duloxetine response nor did the HTR2A gene meet experimentwise significance. The three HTR2A SNPs that showed modest evidence of association in our study (rs9526240, rs6561335, rs9534505) appeared to be in completely separate LD blocks from rs7997012. It should be noted that although we were unable to detect a significant association with rs799012, we do not have adequate power to definitively conclude that the HTR2A rs799012 SNP associated with citalopram response in STAR*D is not associated with duloxetine response. However, if this discordance does not arise from type II error, it suggests that duloxetine response does not appear to be influenced by the same variations as citalopram. Thus, it is possible that COMT and HTR2A may be useful in distinguishing individuals more likely to benefit from an SNRI versus an SSRI. Key limitations of this study include the absence of a placebo or active comparator group, which prevents our consideration of treatment specificity, as well as biases inherent in a candidatebased design. We also recognize the risk of spurious association due to population stratification, despite the exclusion of nonCaucasian subjects, and conversely the risk of type II error because of limited sample size. As such, replication is necessary in independent datasets. Although it would be optimal to use a similar patient population and phenotype, this may not be necessary. Duloxetine has also been studied in generalized anxiety disorder, fibromyalgia, and pain, and it would be worthwhile to determine whether the same SNPs that impact response in patients with MDD also impact response in patients with these disorders. For example, Houston et al. (29) showed that the same dopamine receptor D3 (DRD3) SNPs impacted response to olanzapine treatment in both patients with schizophrenia and bipolar I depression.

Conclusion Catechol-O-methyltransferase but not HTR2A variation was significantly associated with symptom change in duloxetinetreated patients with MDD. If replicated, these preliminary results would suggest that predictors of duloxetine response may differ from those previously reported for SSRI response. www.sobp.org/journal

R.H. Perlis et al. This work was sponsored by Eli Lilly and Company. We would like to thank Stacia Mellinger at Eli Lilly and Company for editorial review of the manuscript. RHP has received honoraria or consulting fees from Eli Lilly and Company, AstraZeneca, Bristol-Myers Squibb, Glaxo Smith Kline, Pfizer, and Proteus and is a stockholder in Concordant Rater Systems, LLC. BF, JPH, and DHA are employees and stockholders of Eli Lilly and Company. VKS was an employee of Eli Lilly and Company during a portion of this work. MHT has received consulting fees from Eli Lilly and Company. Supplementary material cited in this article is available online. 1. Trivedi MH, Rush AJ, Wisniewski SR, Nierenberg AA, Warden D, Ritz L, et al. (2006): Evaluation of outcomes with citalopram for depression using measurement-based care in STAR*D: Implications for clinical practice. Am J Psychiatry 163:28 – 40. 2. Rush AJ, Trivedi MH, Wisniewski SR, Stewart JW, Nierenberg AA, Thase ME, et al. (2006): Bupropion-SR, sertraline, or venlafaxine-XR after failure of SSRIs for depression. N Engl J Med 354:1231–1242. 3. Trivedi MH, Fava M, Wisniewski SR, Thase ME, Quitkin F, Warden D, et al. (2006): Medication augmentation after the failure of SSRIs for depression. N Engl J Med 354:1243–1252. 4. Franchini L, Serretti A, Gasperini M, Smeraldi E (1998): Familial concordance of fluvoxamine response as a tool for differentiating mood disorder pedigrees. J Psychiatr Res 32:255–259. 5. McMahon FJ, Buervenich S, Charney D, Lipsky R, Rush AJ, Wilson AF, et al. (2006): Variation in the gene encoding the serotonin 2A receptor is associated with outcome of antidepressant treatment. Am J Hum Genet 78:804 – 814. 6. Paddock S, Laje G, Charney D, Rush AJ, Wilson AF, Sorant AJ, et al. (2007): Association of GRIK4 with outcome of antidepressant treatment in the STAR*D cohort. Am J Psychiatry 164:1181–1188. 7. Kraft JB, Peters EJ, Slager SL, Jenkins GD, Reinalda MS, McGrath PJ, Hamilton SP (2007): Analysis of association between the serotonin transporter and antidepressant response in a large clinical sample. Biol Psychiatry 61:734 –742. 8. Kim H, Lim SW, Kim S, Kim JW, Chang YH, Carroll BJ, Kim DK (2006): Monoamine transporter gene polymorphisms and antidepressant response in Koreans with late-life depression. JAMA 296:1609 –1618. 9. JurinkeC,vandenBoomD,CantorCR,KosterH(2002):Automatedgenotyping using the DNA MassArray technology. Methods Mol Biol 187:179–192. 10. Whitmyer VG, Dunner DL, Kornstein SG, Meyers AL, Mallinckrodt CH, Wohlreich MM, et al. (2007): A comparison of initial duloxetine dosing strategies in patients with major depressive disorder. J Clin Psychiatry 68:1921–1930. 11. Shelton RC, Andorn AC, Mallinckrodt CH, Wohlreich MM, Raskin J, Watkin JG, Detke MJ (2007): Evidence for the efficacy of duloxetine in treating mild, moderate, and severe depression. Int Clin Psychopharmacol 22:348 –355. 12. Neale BM, Sham PC (2004): The future of association studies: Genebased analysis and replication. Am J Hum Genet 75:353–362. 13. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, Bender D, et al. (2007): PLINK: A tool set for whole-genome association and populationbased linkage analyses. Am J Hum Genet 81:559 –575. 14. Perlis RH, Purcell S, Fagerness J, Kirby A, Petryshen TL, Fan J, Sklar P (2008): Family-based association study of lithium-related and other candidate genes in bipolar disorder. Arch Gen Psychiatry 65:53– 61. 15. Ott J, Hoh J (2003): Set association analysis of SNP case-control and microarray data. J Comput Biol 10:569 –574. 16. Rush AJ, Trivedi MH, Ibrahim HM, Carmody TJ, Arnow B, Klein DN, et al. (2003): The 16-Item Quick Inventory of Depressive Symptomatology (QIDS), clinician rating (QIDS-C), and self-report (QIDS-SR): A psychometric evaluation in patients with chronic major depression. Biol Psychiatry 54:573–583. 17. Shield AJ, Thomae BA, Eckloff BW, Wieben ED, Weinshilboum RM (2004): Human catechol O-methyltransferase genetic variation: Gene resequencing and functional characterization of variant allozymes. Mol Psychiatry 9:151–160.

BIOL PSYCHIATRY 2009;65:785–791 791

R.H. Perlis et al. 18. Chen J, Lipska BK, Halim N, Ma QD, Matsumoto M, Melhem S, et al. (2004): Functional analysis of genetic variation in catechol-O-methyltransferase (COMT): Effects on mRNA, protein, and enzyme activity in postmortem human brain. Am J Hum Genet 75:807– 821. 19. Egan MF, Goldberg TE, Kolachana BS, Callicott JH, Mazzanti CM, Straub RE, et al. (2001): Effect of COMT Val108/158 Met genotype on frontal lobe function and risk for schizophrenia. Proc Natl Acad Sci U S A 98:6917–6922. 20. Goldberg TE, Egan MF, Gscheidle T, Coppola R, Weickert T, Kolachana BS, et al. (2003): Executive subprocesses in working memory: Relationship to catechol-O-methyltransferase Val158Met genotype and schizophrenia. Arch Gen Psychiatry 60:889 – 896. 21. Malhotra AK, Kestler LJ, Mazzanti C, Bates JA, Goldberg T, Goldman D (2002): A functional polymorphism in the COMT gene and performance on a test of prefrontal cognition. Am J Psychiatry 159:652– 654. 22. Tunbridge EM, Harrison PJ, Weinberger DR (2006): Catechol-o-methyltransferase, cognition, and psychosis: Val158Met and beyond. Biol Psychiatry 60:141–151. 23. Yoshida K, Higuchi H, Takahashi H, Kamata M, Sato K, Inoue K, et al. (2008): Influence of the tyrosine hydroxylase val81met polymorphism and catechol-O-methyltransferase val158met polymorphism on the antidepressant effect of milnacipran. Hum Psychopharmacol 23: 121–128. 24. Szegedi A, Rujescu D, Tadic A, Muller MJ, Kohnen R, Stassen HH, Dahmen N (2005): The catechol-O-methyltransferase Val108/158Met polymor-

25.

26.

27.

28.

29.

30.

phism affects short-term treatment response to mirtazapine, but not to paroxetine in major depression. Pharmacogenomics J 5:49 –53. Shifman S, Bronstein M, Sternfeld M, Pisanté-Shalom A, Lev-Lehman E, Weizman A, et al. (2002): A highly significant association between a COMT haplotype and schizophrenia. Am J Hum Genet 71:1296 –1302. Berrettini WH, Wileyto EP, Epstein L, Restine S, Hawk L, Shields P, et al. (2007): Catechol-O-methyltransferase (COMT) gene variants predict response to bupropion therapy for tobacco dependence. Biol Psychiatry 61:111–118. Bray NJ, Buckland PR, Williams NM, Williams HJ, Norton N, Owen MJ, O’Donovan MC (2003): A haplotype implicated in schizophrenia susceptibility is associated with reduced COMT expression in human brain. Am J Hum Genet 73:152–161. Barenboim M, Hyde TM, Kleinman JE, Lipska BK, Weinberger DR (2008): Computational prediction of 3’UTR SNP effect on expression of COMT emphasizes the role of its rare transcript. Biol Psychiatry 63:152S. Houston JP, Perlis RH, Sutton V, Adams DH, Farmen M, Breier A (2007): Genetic association study of treatment response with olanzapine/fluoxetine combination or lamotrigine in bipolar I depression. Biol Psychiatry 61:206S. Barrett JC, Fry B, Maller J, Daly MJ (2005): Haploview: Analysis and visualization of LD and haplotype maps. Bioinformatics 21:263–265.

www.sobp.org/journal