BACKGROUND: Reelin (RELN) is a glycoprotein secreted preferentially by cortical gamma-aminobutyric acid-ergic (GABAergic) interneurons (layers I and II) that binds to integrin receptors located on dendritic spines of pyramidal neurons or on GABAergic interneurons of layers III through V expressing the disabled-1 gene product (DAB1), a cytosolic adaptor protein that mediates RELN action. To replicate earlier findings that RELN and glutamic acid decarboxylase (GAD)(67), but not DAB1 expression, are down-regulated in schizophrenic brains, and to verify whether other psychiatric disorders express similar deficits, we analyzed, blind, an entirely new cohort of 60 postmortem brains, including equal numbers of patients matched for schizophrenia, unipolar depression, and bipolar disorder with nonpsychiatric subjects. METHODS: Reelin, GAD(65), GAD(67), DAB1, and neuron-specific-enolase messenger RNAs (mRNAs) and respective proteins were measured with quantitative reverse transcriptase-polymerase chain reaction (RT-PCR) or Western blot analyses. Reelin-positive neurons were identified by immunohistochemistry using a monoclonal antibody. RESULTS: Prefrontal cortex and cerebellar expression of RELN mRNA, GAD(67) protein and mRNA, and prefrontal cortex RELN-positive cells was significantly decreased by 30% to 50% in patients with schizophrenia or bipolar disorder with psychosis, but not in those with unipolar depression without psychosis when compared with nonpsychiatric subjects. Group differences were absent for DAB1,GAD(65) and neuron-specific-enolase expression implying that RELN and GAD(67) down-regulations were unrelated to neuronal damage. Reelin and GAD(67) were also unrelated to postmortem intervals, dose, duration, or presence of antipsychotic medication. CONCLUSIONS: The selective down-regulation of RELN and GAD(67) in prefrontal cortex of patients with schizophrenia and bipolar disorder who have psychosis is consistent with the hypothesis that these parameters are vulnerability factors in psychosis; this plus the loss of the correlation between these 2 parameters that exists in nonpsychotic subjects support the hypothesis that these changes may be liability factors underlying psychosis.