Citalopram-induced pathways regulation and tentative treatment-outcome-predicting biomarkers in lymphoblastoid cell lines from depression patients

Abstract

Antidepressant therapy is still associated with delays in symptomatic improvement and low response rates. Incomplete understanding of molecular mechanisms underlying antidepressant effects hampered the identification of objective biomarkers for antidepressant response. In this work, we studied transcriptome-wide expression followed by pathway analysis in lymphoblastoid cell lines (LCLs) derived from 17 patients documented for response to SSRI antidepressants from the Munich Antidepressant Response Signatures (MARS) study upon short-term incubation (24 and 48 h) with citalopram. Candidate transcripts were further validated with qPCR in MARS LCLs from responders (n = 33) vs. non-responders (n = 36) and afterward in an independent cohort of treatment-resistant patients (n = 20) vs. first-line responders (n = 24) from the STAR*D study. In MARS cohort we observed significant associations of GAD1 (glutamate decarboxylase 1; p = 0.045), TBC1D9 (TBC1 Domain Family Member 9; p = 0.014–0.021) and NFIB (nuclear factor I B; p = 0.015–0.025) expression with response status, remission status and improvement in depression scale, respectively. Pathway analysis of citalopram-altered gene expression indicated response-status-dependent transcriptional reactions. Whereas in clinical responders neural function pathways were primarily up- or downregulated after incubation with citalopram, deregulated pathways in non-responders LCLs mainly involved cell adhesion and immune response. Results from the STAR*D study showed a marginal association of treatment-resistant depression with NFIB (p = 0.068) but not with GAD1 (p = 0.23) and TBC1D9 (p = 0.27). Our results propose the existence of distinct pathway regulation mechanisms in responders vs. non-responders and suggest GAD1, TBC1D9, and NFIB as tentative predictors for clinical response, full remission, and improvement in depression scale, respectively, with only a weak overlap in predictors of different therapy outcome phenotypes.