Pharmacogenetics of Antidepressant Response

Major depressive disorder is one of the most common and debilitating psychiatric disorders. While psychopharmacological treatments exist, they are not universally effective and can produce significant side effects in some patients. The most common psychopharmacological agents used to treat major depression are the selective serotonin reuptake inhibitors, or SSRIs. Often, these drugs take several weeks to relieve depressive symptoms. If the initial therapy fails, other antidepressants are often prescribed. This "trial and error" process creates a delay in remission which can be frustrate the patient and lead to further decreased well-being. Individualized therapy would have great clinical utility by identifying patients that are likely to respond positively to SSRI therapy. The goal of this thesis is to investigate the use of genetic markers for guiding treatment with SSRIs. We utilized several complementary pharmacogenetic approaches and two depressed populations treated with SSRIs. The first was a small (N=96) population given fluoxetine, and the second was a large (N=1,953) population taking citalopram. We used the fluoxetine population and a linkage disequilibrium mapping approach to investigate variants in seven pharmacodynamic candidate genes for association to response and specificity of response. Several variants in HTR2A and TPH1 were associated with fluoxetine outcome. We then resequenced the coding region and 5' conserved non-coding regions of these genes in the fluoxetine population in order to uncover novel variation and additional tagging SNPs. These tagging SNPs were genotyped in our citalopram population, and none of the SNPs were associated with clinical outcome. We then genotyped known, functional polymorphisms in relevant pharmacokinetic genes for association to citalopram response and tolerance. Using a two-stage study design, none of the variants were significantly associated with outcome following citalopram treatment. We also utilized a whole genome association study using over 590,000 SNPs from across the genome. Using a two-stage study design, none of the most highly associated markers in the discovery sample were also associated in the validation sample. Similar non-significant results were obtained using multi-SNP decision trees. However, further genotyping is necessary in the validation sample, as the most highly associated SNPs may not be the most consistently associated