The evaluation of such molecules in schizophrenia might also reveal affected molecular pathways in PV and/or SST neurons, which could be used for developing therapeutic strategies targeting selectively these neurons. In order to identify such molecules, we first used published gene expression data for mouse cortical neuron subsets and selected 70 genes found to be either developmentally upregulated or preferentially enriched in PV and/or SST neurons. We then evaluated the expression patterns of these 70 genes in the online atlases of gene expression in the mouse or human cortex and excluded genes that were detected in pyramidal-like neurons with an apical dendrite, or that exhibited an apparently different laminar expression pattern from those of PV and/or SST mRNAs. We found that KCNS3, LHX6, KCNAB1 and BAY 73-4506 PPP1R2 had cortical mRNA expression patterns similar to those of PV and/or SST mRNAs. KCNS3 encodes voltage-gated K+ channel Kv9.3 modulatory a-subunit that coassembles with Kv2.1 a-subunits and leads to an enhanced conductance and modified gating properties of the heteromeric channels. LHX6 encodes LIM homeobox protein 6, a transcription factor suggested to be involved in the development of PV and SST neurons in the mouse cortex. KCNAB1 encodes K+ channel Kvb1 accessory subunit that confers fast N-type inactivation to Kv1.1 channels. PPP1R2 gene encodes protein phosphatase 1 regulatory subunit 2, which inhibits PP1 and controls signal transduction and synaptic plasticity. In this study, we determined whether KCNS3, LHX6, KCNAB1 and PPP1R2 mRNAs are selectively expressed in PV and/or SST neurons in the human prefrontal cortex. For assessing coexpression of two molecules in single cells, ISH detection of mRNAs has several advantages over the detection of proteins by immunohistochemistry. First, because of predominant somal expression of most mRNAs, ISH can demonstrate directly overlapping signals for two mRNAs in the soma. On the other hand, differences in subcellular distribution of certain proteins and polypeptides among neuronal subtypes may mask the detection of their expression by IHC in certain cell types. For example, the majority of cholecystokinin immunoreactive somata in the cortex were shown to be GABA neurons, but not pyramidal neurons. However, CCK mRNA could be easily detected in the pyramidal neuron somata with ISH. This discrepancy was explained by the predominant accumulation of CCK protein in the axon terminals, but not in the somata, of pyramidal neurons. Second, ISH with 35S-labeled riboprobes has a high sensitivity that enables the detection of as few as 10 mRNA copies per cell, whereas the sensitivity of IHC, which depends on the specificity of antibodies, might not be high enough to identify cells containing low levels of target proteins. Third, in ISH with emulsion autoradiography, specifically labeled cells are identified using quantitatively determined cut-offs based on the distribution of grain numbers per cell, whereas in IHC, identification of positive cells tends to depend on visual impression which is less reliable for cells with weak immunopositive signals. These methodological factors might explain why PPP1R2 was demonstrated to be selective to PV neurons by IHC in the mouse cortex, but in our ISH study, PPP1R2 mRNA was detected in a larger neuronal population than PV neurons. Although the types of cortical neurons that express PPP1R2 might differ between mouse and human, the absence of protein signals and the presence of mRNA for PPP1R2 in many non-PV neurons might reflect a subcellular localization or the low expression levels of PPP1R2 protein in non-PV neurons.