Analyses suggested that gene expression changes in autism, schizophrenia, and bipolar disorder were consistent with an immature developmental gene expression program of parvalbumin-positive fast-spiking interneurons (FS cells). This conclusion was indicated by a robust relationship between a FS cell maturation index and parvalbumin expression levels across multiple studies and a decrease in both measures in association with autism, schizophrenia, and bipolar disorder.
The association of parvalbumin expression levels with the FS cell index was surprising because the FS cell index was broad-based and used equally weighted expression levels of all genes except for parvalbumin (see methods), whereas, the level of parvalbumin expression represented a very restricted developmentally regulated marker of FS cells. Though decreased parvalbumin levels had been reported in association with neuropsychiatric disorders 
, it was unclear whether this decrease was gene specific or related to transciptome-wide changes in the FS cell developmental program. That a broad-based measure robustly correlated with the expression level of a single cell-type specific developmentally regulated gene is strongly suggestive of transriptome-wide co-regulation of a large number of genes and is consistent with a shift towards immaturity in the FS cell developmental program. The fact that genes were segregated in the FS index based solely on direction of gene expression change during FS cell maturation in purified cortical FS cells further suggested that the FS cell index reflected the maturity of the FS cell developmental program in cortex. This conclusion was also supported by the finding that the association between the FS cell index and parvalbumin remained significant after controlling for covariation with other cell type indices and with other interneuron specific markers. Finally, chi-square analyses provided an additional level of support for FS cell developmental program-associated gene expression changes in disease that were opposite in direction from changes that occurred during FS cell development.
An approach where averages of equally weighted genes are used to create a cell-type specific index is well suited to the detection of transcriptome-wide changes in the maturity of gene expression programs. Since all genes are weighted equally, only non-random gene expression differences with respect to program maturation can contribute to a change in the index, whereas, random changes will cancel out. Thus, an index measurement can only vary when a large percentage of genes move in a coordinated manner that is consistent in direction with expression level changes during development. Comparing variability in an index measure that relies on transcriptome-wide changes in gene expression levels to expression levels of an established cell-type specific marker, such as parvalbumin, can support specificity of the transcriptome-wide approach. Specificity can be further confirmed when the relationship between the index and marker withstands correction for covariation with other cell type specific indices and markers. The use of equally weighted gene expression levels to create the index also likely contributes to the robustness of the approach. In all cases, the FS cell index was closely related to parvalbumin levels regardless of array platform or species. Thus, the equal weighting and averaging of genes, it is hypothesized, makes it possible to capture the essence of redundant but noisy information about the state of developmental programs.
The robust correlation of the FS cell index with parvalbumin levels and their coordinated decrease in disease provides additional support for the hypothesis of FS cell program immaturity in disease that compliments previous experimental data consistent with this possibility. Numerous studies reported decreased parvalbumin levels in neuropsychiatric disease 
. While this could be interpreted to suggest decreased numbers of FS cells, evidence suggested FS cell immaturity. For instance, in situ
hybridization in post-mortem brains of individuals with schizophrenia showed that parvalbumin-positive interneurons were still present in the same numbers but that levels of parvalbumin mRNA per cell were markedly reduced 
. Also, neurophysiological signatures of FS cells were immature in a mouse model of schizophrenia 
. Thus, both the current data analyses and previous studies were consistent with the interpretation that the FS cell developmental gene expression program is immature in autism, schizophrenia, and bipolar disorder.
Current results also provide an interesting context to traditional studies of gene expression changes. For instance, a significant overlap of gene expression changes was found in autism, bipolar disorder, and schizophrenia with 137 genes significantly changing in all 3 diseases. Importantly, 99.3% of these overlapping genes moved in the same direction in all diseases and 69.9% of these moved in the opposite direction during FS cell development. Further, these genes were biased (63.2%) toward those that were upregulated during FS cell development. The unified direction of gene expression changes across disease states is suggestive of a similarly coordinated regulation of this set of genes across disease states. That this set of genes change in the opposite direction during FS cell development suggests abnormal regulation of FS cell maturation may be the causal mechanism behind these disease-related gene expression changes. Further supporting this idea was enrichment for developmentally upregulated genes that would be expected in a heterogeneous tissue such as brain where downregulated genes might be obscured by higher expression levels of those genes in other cell types. Suggesting, that abnormal FS maturation might be important to the interpretation of published studies, pathway analysis identified energy metabolism, oxidative phoshporylation, and mitochrondial function as overrepresented in the subset of genes that were downregulated in all diseases and upregulated during FS cell maturation. Mitochondrial dysfunction and abnormal measures of oxidative phosphorylation and energy metabolism have been widely reported in autism, bipolar disorder, and schizophrenia 
. FS cells have high metabolic demands and dramatically upregulated energy related genes during development (table s7
). An interesting interpretation of these results might be that an underlying defect in FS cell maturation accounts for the widely reported disease-related abnormalities in oxidative phosphorylation and energy metabolism. It is also possible that primary defects in oxidative phosphorylation or energy metabolism genes might impair FS cell development.
Variability in transcriptional program maturity could have important implications for brain function. This is because the function of FS cells varies dramatically across development. FS cells have significant changes across development in firing frequency, firing type, firing rates, and resting membrane potential 
. Thus, a change in the developmental age of FS cells would be expected to have functional consequences. A role for transcriptional immaturity in disease could also inform the study of risk factors. For instance, might prenatal viral infections associated with increased risk of neuropsychiatric disease 
systematically interfere with the progression of developmental gene expression programs? Finally, the current study adds to the growing body of literature suggesting links between autism, schizophrenia, and bipolar disorder that were found to share genetic 
and environmental risk factors 
as well as endophenotypes 
To conclude, the current study describes a novel method for assessing the maturity of the FS cell developmental program using gene expression data from mixed cell-types in brain homogenate. Use of this method suggested a new interpretation of changes in gene expression levels in neuropsychiatric disease to be at least partially related to an immature gene expression program in FS cells. Understanding mechanisms underlying impaired developmental program progression could improve understanding of disease pathogenesis.