Cell-Type-Specific Transcriptomic Analysis in the Dorsolateral Prefrontal Cortex Reveals Distinct Mitochondrial Abnormalities in Schizophrenia and Bipolar Disorder

Substantial evidence suggests that schizophrenia and bipolar disorder share common vulnerability factors, such as genetic, epigenetic, and environmental factors. In addition, both of these psychiatric illnesses show many overlapping phenotypes. Mitochondria are the major source of energy for neurons and are involved in brain development, Ca²⁺ homeostasis, neurotransmission, metabolism, synaptic plasticity, and apoptosis. Given their important role in various physiological processes, it is not surprising that mitochondria and their functions have been studied in both schizophrenia and bipolar disorder (1). In fact, one of the overlapping traits of schizophrenia and bipolar disorder is abnormal mitochondrial pathology. Interestingly, only a few studies have critically examined mitochondrial functional pathways at the diagnostic level. In this issue of the Journal, Glausier et al. (2) have elegantly addressed this topic by examining mitochondrial pathway signatures for schizophrenia and bipolar disorder at the cellular level and, more importantly, in specific cell types. It has been shown that energy needs vary not only for different brain areas but also for different cells. This energy need can even be linked directly to certain clinical symptoms and symptom severity (3, 4). Thus, studying cell-type-specific mitochondrial function is critical not just to understand its relative contribution to energy needs of specific cells but also to distinguish the phenotypic characteristics of these two disorders.

Using two independent preexisting gene expression data sets, Glausier et al. adopted a transcriptomic profiling approach to explore the cell-type-specific contribution of mitochondrial energy-producing deficits in cortical circuits in schizophrenia and bipolar disorder and established how they can be used to determine the pathological trajectories associated with these two disorders. In order to do so, they applied several strategies based on 1) analyzing differentially expressed genes between case subjects and unaffected comparison subjects, 2) applying a novel gene ontology enrichment filter to sort a set of differentially expressed genes associated with mitochondria-related functions, and 3) using weighted gene coexpression network analysis to determine differences in gene function-associated module preservation between case subjects and unaffected comparison subjects. The first set of data was chosen from RNA sequencing conducted in the gray matter of the dorsolateral prefrontal cortex (DLPFC) from 57 case subjects with schizophrenia, 35 case subjects with bipolar disorder, and 82 sex-, age-, and race-matched unaffected comparison subjects. The second data set was pooled from two independent microarray gene expression studies focused on layer 3 pyramidal neurons (L3PNs) and layer 5 pyramidal neurons (L5PNs) from the DLPFC. The first microarray data set comprised 36 pairs of well-matched schizophrenia case subjects and unaffected comparison subjects, and the second data set was based on a relatively large unique cohort of 19 triads of schizophrenia, bipolar disorder, and unaffected comparison subjects matched for sex, age, and postmortem interval. The authors initially analyzed gene expression changes and determined any differences across diagnostic groups by deducing the correlation between the test statistics for each gene analyzed. With the application of mitochondrial gene function (Gene Ontology [GO] project as “mitochondria” [GOMito]) filter, based on a curated list of 1,033 genes, Glausier et al. extracted 871 genes from gray matter DLPFC RNA sequencing data and found 356 differentially expressed genes in schizophrenia subjects in contrast to only 67 differentially expressed genes in subjects with bipolar disorder. The results show that expression changes are much stronger in schizophrenia compared with bipolar disorder. The 356 differentially expressed genes in schizophrenia demonstrated a significant functional association with mitochondrial pathways under three broad categories: mitochondrial dysfunction, oxidative phosphorylation (OXPHOS), and sirtuin signaling. Sirtuin signaling is mediated by the sirtuin protein family, which regulates the acetylation and/or ADP ribosylation of a variety of proteins and participates in physiological processes, such as cellular stress response, mitochondrial biosynthesis, lipid metabolism, and apoptosis. Interestingly, no functional association with mitochondrial pathways was found in bipolar disorder, even though a considerable number of overlapping genes were present in the constituting pathways. At the cell-type-specific level, the analysis of pyramidal neurons (L3PN and L5PN) in schizophrenia subjects detected changes in 662 genes belonging to GOMito, relative to unaffected comparison subjects. Most of the differentially expressed genes were down-regulated in both L3PN and L5PN. Pathway enrichment analysis showed the same mitochondrial pathways (e.g., mitochondrial dysfunction, OXPHOS, and sirtuin signaling) that were detected in DLPFC gray matter. A strong positive correlation between disease diagnosis and GOMito gene expression was also noted with the test statistics in schizophrenia subjects. Next, the authors studied the role of mitochondria in a smaller batch of schizophrenia subjects and examined whether comparable changes in mitochondria-related gene functions can be identified in L3PN and L5PN in bipolar disorder subjects. The GOMito-based differential gene expression analysis confirmed the perturbed mitochondria-related energy production pathways in both L3PN and L5PN in schizophrenia subjects; however, no such differences were noted in any of the mitochondrial-related gene expression abnormalities either in L3PN or L5PN in subjects with bipolar disorder. Importantly, the differential test statistics were unable to determine disease effects in L3PN- and L5PN-based differential gene expression when the schizophrenia and bipolar disorder groups were combined. On the other hand, independent analysis followed by test statistics showed moderately positive correlation with L3PN- and L5PN-specific mitochondrial-related gene expression changes in schizophrenia. Altogether, these analyses demonstrate that although mitochondrial abnormalities are postulated to be associated with both schizophrenia and bipolar disorder, at the cellular level, genes associated with mitochondrial functions are specifically associated with schizophrenia. This study also opens up an avenue to further examine other brain areas and study how abnormalities in mitochondrial gene-specific pathways play a role in different cell population, and eventually in the etiopathogenesis of these disorders.

Investigating complex gene expression patterns in polygenic psychiatric disorders is a challenging task. Furthermore, analyzing large-scale gene expression data may often result in batch effect, which can substantially mask true expression signals. As a countermeasure, a paired design format can be adopted for high throughput gene expression analysis. Weighted gene coexpression network analysis has emerged as a powerful tool to effectively characterize correlation patterns among genes by constructing a gene coexpression network. This type of analysis helps in creating densely connected subnetworks from gene modules, which are related to biological functions. A gene coexpression network is an undirected graph, where each node corresponds to a gene and each edge connects a pair of genes that are significantly correlated (5). The Glausier et al. study used a similar approach to identify modules based on the hierarchical clustering of genes related to mitochondrial functions in schizophrenia and bipolar disorder. Preservation analysis in DLPFC gray matter found that the module structure present in the unaffected comparison group was highly to moderately preserved in both schizophrenia and bipolar disorder subjects, suggesting that modules containing mitochondrial genes were not affected in these disorders. On the other hand, comparison between the schizophrenia and bipolar disorder groups showed that although the two modules containing differentially expressed genes found in these two groups were not enriched for any specific functional pathways, the schizophrenia group had an additional module that contained differentially expressed genes enriched for mitochondrial dysfunction, OXPHOS, and sirtuin signaling. This analysis further distinguishes schizophrenia from bipolar disorder. When determined individually in L3PNs and L5PNs, module preservations were noted between schizophrenia subjects and unaffected comparison subjects, confirming the findings of DLPFC gray matter. Lastly, in vivo studies in monkeys were used to test whether the observed effects on GOMito genes in schizophrenia subjects were related to antipsychotic drugs (haloperidol, clozapine, and olanzapine). No effects of both first- and second-generation antipsychotic drugs on the expression of mitochondrial-related genes were noted either in DLPFC gray matter or pyramidal cells, thus suggesting that these effects were not related to the antipsychotic treatment of schizophrenia subjects.

This study has several implications. The energy-consuming process in the cortical brain area is supported by the high energy production abilities of mitochondria through an activated OXPHOS system (6). As a prerequisite, mitochondria supply nearly all the energy needs central to dendritic spine formation, synaptic transmission, maintenance of ionic homeostasis in synaptic terminals, and synaptic vesicle recycling in the cortical neurons (7). The abnormalities in energy production pathways, due to mitochondrial deficiencies, is a hallmark of dendritic spine pathology and have been implicated in the reduced gray matter volume of the DLPFC in schizophrenia (8). Compelling evidence also shows reduced dendritic spines in layer 3 and, to a lesser extent, in layer 5 pyramidal neurons of the neocortex in schizophrenia (9). These changes in dendritic spine morphology, accompanied by disrupted mitochondrial energy production, have been associated with cognitive impairment in schizophrenia (10). Mitochondrial pathway abnormalities in L3PNs and L5PNs suggest that certain phenotypic characteristics of schizophrenia could be ascribed to these abnormalities. As mentioned earlier, perturbations in mitochondrial functions have been linked to specific behavior, and mitochondrial pathology could be an important factor in the manifestation of clinical symptoms. It has been suggested that in the anterior cingulate cortex, there is a reduced number of mitochondria per neuronal somata in schizophrenia due to selective loss in layers 5 and 6, although the density of mitochondria between neuropil was not altered (11). It will be interesting to examine whether gene expression changes in the DLPFC associated with mitochondrial pathways are the function of mitochondrial number and/or other neuroanatomical abnormalities in the DLPFC observed in schizophrenia subjects.

As is well known, both mitochondria-encoded and nuclear-encoded genes contribute to mitochondrial genes and mitochondrial functions (12). Also, crosstalk between mitochondria and the nucleus is critical in regulating the metabolic needs of cells. In fact, several genes associated with OXPHOS, which are a part of the mitochondrial genome, are nuclear encoded. A previous study noted that mitochondrial and neuronal gene coexpression modules are shared between bipolar disorder and schizophrenia cortical regions (13). On the other hand, an epistatic interaction study between nuclear-encoded and mitochondria-encoded genes identified potential interactions between mitochondrial DNA and the nuclear genome in bipolar disorder (12). In this context, it will be of great interest to dissect how the mitochondrial and nuclear genomes, individually or in coordination, contribute to modules in a cell-specific manner and how they interact with each other to influence schizophrenia pathophysiology.