We report results from the PREDICT-Gene case-control study nested in a prospective cohort designed to identify predictors of the onset of depression among adult primary-care attendees. We tested the potential gene-by-environment interaction between 5HTTLPR genotype at the serotonin transporter gene and previous exposure to threatening life events (TLEs) in depression. A total of 737 consecutively recruited participants were genotyped. Additional information was gathered on exposure to TLEs over a 6-month period, socio-demographic data and family history of psychological problems among first-degree relatives. Diagnoses of depression were ascertained using the Composite International Diagnostic Interview (CIDI) by trained interviewers. Two different depressive outcomes were used (ICD-10 depressive episode and ICD-10 severe depressive episode). Both the s/s genotype and exposure to increasing number of TLEs were significantly associated with depression. Moreover, the 5HTTLPR s/s genotype significantly modified the risk conferred by TLEs for both depressive outcomes. Thus, s/s homozygous participants required minimal exposure to TLE (1 TLE) to acquire a level of risk for depression that was only found among l/s or l/l individuals after significantly higher exposure to TLEs (two or more TLEs). The interaction was more apparent when applied to the diagnosis of ICD-10 severe depressive episode and after adjusting for gender, age and family history of psychological problems. Likelihood ratios tests for the interaction were statistically significant for both depressive outcomes (ICD-10 depressive episode: LR X 2 = 4.7, P = 0.09 (crude), LR-X 2 = 6.4, P = 0.04 (adjusted); ICD-10 severe depressive episode: LR X 2 = 6.9, P = 0.032 (crude), LR-X 2 = 8.1, P = 0.017 (adjusted)).