We investigated the contribution of several polymorphisms in serotonin-related candidate genes to the response of depressed patients over time treated with citalopram. This approach differs from other studies where the response criterion is dichotomized based on depression scores after few weeks of treatment. In this case, we evaluate depression scores quantitatively along the treatment period of 12 weeks. Our results in the exploratory sample do not support the previously reported antidepressant response associations with 5-HTTLPR (Arias et al 2003
; Kim et al 2000
; Rausch et al 2002
; Smeraldi et al 1998
; Yoshida et al 2002
; Yu et al 2002
; Zanardi et al 2000
_T102C (Minov et al 2001
; Peters et al 2004
(Binder et al 2004
). Most likely our small sample size may not have the power to detect an effect. In addition, there were differences in study design - not all of the previous studies used citalopram, and response criteria differ significantly among studies. Nevertheless, for the most frequently discussed and published variant, 5-HTTLPR, we find not even a trend of an effect on citalopram response.
Instead, we find association with both HTR1A
genes that encode two receptors which are downregulated after SSRI treatment (Sari 2004
). Our results although nominal and not corrected for multiple testing suggest that higher baseline level of expression of these two receptors interferes with SSRI efficacy: First, the HTR1A
variant associated with lower response, i.e. the G allele of C-1019G, prevents binding of a repressor (NUDR/DEAF-1) in the raphe (Lemonde et al 2003
). The same allele has been found associated with poorer response to antidepressants in two other studies (Hong et al 2006
; Lemonde et al 2004
). Furthermore, the G allele was associated with depression (Lemonde et al 2003
), anxiety- and depression-related personality traits (Strobel et al 2003
), agoraphobic subtype of panic disorder (Rothe et al 2004
). Our results are consistent with these findings, since we find association both with higher QIDS scores (i.e. depression severity) throughout the trial, but also independently with response at week 12 (p=0.035) while accounting for QIDS scores at baseline as a covariate. Combining the two effects, i.e. not taking initial QIDS score into account, as is done in most previous association studies on response, yields highly significant response difference (p = 0.005), which remains significant after accounting for multiple testing.
Second, after finding a nominal association between homozygosity for the G allele on HTR1B
_G861C and non-response, we tested two functional SNPs in the promoter in the exploratory sample, and found that a haplotype combination of SNPs that has previously been reported to increase transcription (Duan et al 2003
) was significantly associated with slower response to the SSRI citalopram. However, only four subjects were homozygous for haplotype 4, three of which maintained high scores over time, and one reached a QIDS score of 5, i.e. barely passed remission criterion, at week 12. Our data suggest that the higher transcriptional level may interfere with the downregulation previously found necessary for SSRI efficacy (Sari 2004
). Both genes are integral to the negative feedback control of serotonergic activity (Lifschytz et al 2004
). Interaction between these two genes could not be estimated in the exploratory sample because the combination of both the G/G in HTR1A
and the high expressed haplotype in HTR1B
was not present in this sample. However, individuals with at least one of the genotypes that increased transcription (in either HTR1A
) had higher QIDS scores over time than individuals with common variants for both genes (who showed the best response to citalopram; 106/147; 72%; see ). If these genetic variants were at least additive, one might speculate that individuals with high levels of transcription in both genes might be particularly treatment resistant, a hypothesis that can be tested in larger samples. Our findings, while clearly preliminary, would predict that genetic variants enhancing transcription may affect desensitization of both receptors by the SSRI and ultimately drug response. This hypothesis is supported by pharmacological data showing that an antagonist for both autoreceptors accelerates the onset of therapeutic effect (Starr et al 2007
). Testing of this hypothesis in a large sample is warranted.
However, there are two caveats to our results. First, only four out of 153 subjects were homozygotes for the highly expressed haplotype – and these all were not or barely responsive to citalopram. Second, haplotype phase cannot be inferred unambiguously for double heterozygote subjects for the two promoter SNPs. The PL-EM algorithm defines phase based on the maximum expectation, and double heterozygous samples thus fell into the low expression category. However, given allele frequency, about 5–6% of these double heterozygotes would be expected to have one of the rarer haplotype on one chromosome together with the high expressing haplotype on the other chromosome, and are thus misclassified as low expressing. However, the association with high expression category remains unaffected by this phase uncertainty.
To further explore the genetic variation influence of these two genes, we obtained genotype data for three SNPs from each gene, HTR1A
, from the extended sample of patients participating in the STAR*D study, which includes most of our exploratory sample (McMahon et al 2006
). As described by McMahon (McMahon et al 2006
) SNPs were selected from HapMap and none of our functional SNPs were genotyped. For HTR1A
, the three SNPs genotyped in the large sample flank the gene at about 5 kb on each side. Individuals homozygous for the G allele at rs1364043 and homozygous for a haplotype containing this allele showed better response over time to citalopram. This SNP is located 6.7 kb downstream from HTR1A
gene and no information is currently available about its functionality or affect on expression. Given that LD between rs1364043 and rs6295 (the C-1019G found associated in the exploratory study) is weak (r2 = 0.267), we cannot conclude that this association is due to this latter, functional SNP. But both data sets independently support that genetic variation at the HTR1A
locus may be involved in the response to SSRIs.
Similarly, for HTR1B we found association for SNP rs6298 in the extended sample. According to HapMap, this SNP is in complete LD (r2=1) with the coding SNP rs6296 found associated in our exploratory sample. Heterozygotes for both SNPs showed better response to citalopram over time than either type of homozygotes. For rs6298, homozygotes for the T allele show higher QIDS scores over time. Unfortunately, we cannot determine whether or not the observed association in the extended sample is due to the promoter variants, since rs6298 alone cannot tell the high expressed haplotype apart. It appears that in HTR1B there are at least two functional variants on several haplotype backgrounds. Clearly, testing more SNPs including the two promoter variants in a larger sample may help elucidate the contribution of these genes to antidepressant response.
A significant interaction between the two associated SNPs from each gene, HTR1A and HTR1B, was observed in the extended sample. Individuals homozygous for the T allele at both loci have higher mean QIDS scores over time than either one alone, suggesting a synergetic (more than additive) effect on the response to treatment. This combination was even more significantly associated with antidepressant response (p = 0.004) than any other SNP combination and the global interaction was also significant (p = 0.032).
In this same line of evidence, novel therapeutic approaches include a selective HTR1A/1B autoreceptor antagonist and 5-HT re-uptake inhibitor for a faster antidepressant action (Scott et al 2006
) and acute anxiolytic effect in rats (Starr et al 2007
). When adding a HTR1A antagonist to an SSRI, a faster improvement was observed in some studies (Artigas et al 1994
; Blier and Bergeron 1995
) but not reproduced by others (Berman et al 1999
; Berman et al 1997
) and a recent meta-analysis confirmed the faster improvement in the first two week period but not beyond that period (Ballesteros and Callado 2004
). A substantial increase in 5-HT has been observed when combining an SSRI with a HTR1A/1B antagonist compared to using HTR1A antagonists alone (Gobert et al 2000
; Sharp et al 1997
). Recently a study on the transcriptional regulation at HTR1A
C(-1019)G suggested that genetic heterogeneity may obscure the association of this functional polymorphism on mood a suggestion that is further supported by the several reports that find associations when gene-gene or gene-environment interactions are considered (Le Francois et al 2008
In summary, we investigated the natural course of response to SSRIs over time in relation to genetic variation at two SSRI target genes, an approach that allows the effect of time to randomly vary from individual to individual. Although not conclusive, we show that more than one polymorphism in a gene may contribute to the response to treatment and a more dense analysis in a large sample may help elucidate this influence. Our major finding is the interaction between the two associated SNPs in each gene, HTR1A and HTR1B where genetic variation at both loci influences the individual response to treatment, a response tested by the mixed linear model that takes into account inter-individual variability. This is consistent with preclinical studies that show that simultaneous use of antagonists for both genes accelerates and improves the effect of SSRIs.