Background
Schizophrenia is a devastating and complex brain disorder. Although many susceptibility genes have been identified through genome-wide association studies[
1‐
4], each gene exerts only a small to moderate effect on overall disease risk. Identifying endophenotypes in the brains of patients with schizophrenia is now considered the way to understand the etiology and mechanisms of the disorder. Growing evidence from postmortem[
5‐
9] and animal studies[
10‐
12] implicates abnormal neurodevelopment in the pathogenesis of schizophrenia and other psychiatric disorders. We screened >160 mutant mouse strains using a large-scale comprehensive behavioral test battery and identified several strains with behavioral traits corresponding to schizophrenia[
13]. We examined the brains of the latter group using various approaches and found abnormalities in the dentate gyrus (DG) of the hippocampus in these mutants. That is, the molecular and electrophysiological features of DG neurons in the adult hippocampus of these mouse strains showed similarities to those of immature DG neurons in normal mice, a phenomenon termed the “immature DG” (iDG). To date, identified mouse strains with an iDG phenotype include the alpha-calcium/calmodulin-dependent protein kinase II heterozygous knockout (HKO)[
12], schnurri-2 (Shn-2) KO[
11], and mutated synaptosomal-associated protein 25 (SNAP-25) knock-in mice[
10]. Importantly, postmortem analysis revealed an iDG-like signature in the brains of patients with schizophrenia or bipolar disorder[
8]. We therefore proposed that the iDG is a potential endophenotype of several psychiatric disorders, including schizophrenia and bipolar disorder. The maturation of gamma-aminobutyric acid (GABA) signaling, characterized by progressive developmental switches in expression from GAD25 to GAD67 and from NKCC1 to KCC2, is abnormal in the hippocampus of patients with schizophrenia[
6]. In addition to the hippocampus, we found abnormal development and maturation in the cortex in a mouse model of schizophrenia[
11]. Abnormalities in development and maturation have been also implicated in the cortex of patients with schizophrenia. Risk alleles for schizophrenia may directly affect PFC development[
14,
15]. Torkamani
et al. showed that the age-related decline of genes associated with developmental processes, such as neuronal differentiation, neurite outgrowth, and synaptic transmission, appeared to be slowed in the cortex of patients with schizophrenia[
7]. This suggests that the expression of a subset of genes in the schizophrenic brain has become arrested at an adolescent (up to 19 years) developmental stage[
16]. Consistent with this idea, a hyperdopaminergic state in patients with schizophrenia has been suggested to resemble the dopamine hyperactivity in the adolescent brain[
17]. In the case of the GABA
A receptor, α1 subunit expression increases during PFC development and persists into adulthood, whereas α2 subunit expression decreases[
18,
19]. A decrease in α1 subunits[
18,
20,
21] and an increase in α2 subunits[
21,
22] have been found in the schizophrenic PFC. Changes in GABA
A receptor subunits in schizophrenia may reflect a cortex held in a state of immaturity during adulthood. Furthermore, fast-spiking interneurons (FS neurons) in the cortices of patients with schizophrenia show maturational abnormalities[
5]. Several lines of evidence show that expression levels of parvalbumin (PV), a marker of FS neurons, are decreased in the PFC of these patients[
23‐
28]. PV immunoreactivity first appears in the PFC around 3–6 months of age and PV mRNA increases 20-fold from the neonatal stage to the adulthood[
28,
29], indicating that PV is a marker for mature FS neurons. The FS neurons in the cortices of patients with schizophrenia were hypothesized to be immature[
5]. It was recently shown that, in schizophrenia, this neuron type retains a pseudo-immature status with regard to gene expression profiles[
5]. Shn-2 KO mice, a mouse model of schizophrenia with iDG in the hippocampus, showed a decrease in the number of PV-positive cells in the frontal cortex, without signs of neurodegeneration in either region[
11], which suggests that the immature signature can be seen in the mutant mice not only in the hippocampus but also in the frontal cortex. In contrast, somatostatin expression, a marker for a certain interneuron type, decreases from birth in the normal PFC, but shows a significant decrease in the schizophrenic PFC[
28,
30]. Considering that there are genes with expression patterns that are inconsistent with an immature phenotype in the schizophrenic PFC, it is important to evaluate immaturity in the brain using genome-wide gene expression profiles. However, genome-wide gene expression patterns in the brains of patients with schizophrenia and those of normal infants have not been directly compared.
Reports on studies of large-scale gene expression in various regions of the schizophrenic brain, including PFC, have been accumulating in publicly available databases. In this study, we performed bioinformatics analyses of such data (public microarray data sets) to see if the maturational state of the PFC is affected in schizophrenia. We compared genome-wide gene expression patterns of human developing PFC and adult schizophrenic PFC using different data sets reported independently.
Discussion
In this study, we show that the schizophrenic PFC resembles the juvenile PFC with respect to transcriptome-wide gene expression profiles. We compared relative gene expression in the DLFC and MFC of patients with schizophrenia (patients compared with controls) with that in corresponding regions of the human developing PFC (infants compared with adults), and showed striking similarities between them. Moreover, we revealed that transcriptional immaturity could be seen in multiple cell types in the schizophrenic PFC, including FS neurons, astrocytes, and oligodendrocytes. Our findings support the idea that immaturity in the PFC could be an endophenotype of schizophrenia.
Regarding gene expression, our bioinformatics analyses revealed highly significant similarities in the relative gene expression patterns in the PFC between patients with schizophrenia (as compared to healthy controls) and infants (as compared to adults). More specifically, the expression level of a large number of genes that normally increases during human DLFC development was decreased in the DLFC of patients with schizophrenia (Figure
2b). In other words, these genes can be considered maturation markers whose expression is disturbed in the DLFC of patients with schizophrenia. The similarity in the maturation marker expression patterns and those of schizophrenia markers in adults was extraordinary significant, and the result was replicated in 3 further analyses using other two DLFC data sets and an MFC data set, respectively (Figure
2d, f, Additional file
2: Figure S1b). Genes for immaturity markers with expression that decreases with age in the normal PFC also were affected in the PFC of patients with schizophrenia, but to a lesser extent. Thus, the transcriptional immaturity of the schizophrenic PFC could be characterized as preferential downregulation of maturation marker gene expression. When the gene expression patterns in the developing DLFC (infants vs. adults; GSE11512[
34]) were compared with those in the 2 schizophrenia cohorts (patients vs. controls; GSE53987 or GSE12649[
35]), we found statistically significant similarities in both comparisons (
P = 8.4 × 10
-22 and
P = 1.8 × 10
-5; Additional file
2: Figure S1). The relatively low overlap
P-value in Additional file
2: Figure S1c relative to other comparisons (Figure
2a, 2c, Additional file
2: Figure S1a) may be due to the small number of transcripts that were changed in the patients in the GSE12649 study[
35]. Transcript expression changes numbering 2757 (806 upregulated and 1951 downregulated), 1163 (529 upregulated and 634 downregulated) and 163 (118 upregulated and 45 downregulated) were found in the studies GSE21138[
32], GSE53987, and GSE12649[
35], respectively. The differences in the number of transcripts detected in these studies are probably due to differences in the types of microarray chip that they used (Table
1). The small number of transcripts (45) downregulated in the patients in the GSE12649 study[
35] may have caused the small number of maturation markers that downregulated in these patients (13 transcripts in Additional file
2: Figure S1d). Functionally, a deficit in working memory is common in patients with schizophrenia and has been attributed to PFC dysfunction[
46]. When the developmental changes in working memory capacity were tested in normal subjects, 8- to 12-year-olds performed more poorly than adolescents or adults (age groups 13–17 and 18–25, respectively)[
47]. Together with these previous results, our results suggest the PFC in patients with schizophrenia resembles the PFC in normal infants, both functionally and in gene expression patterns.
In the study by Maycox
et al., the MFC analysis was performed for subjects with schizophrenia and controls who had an average age of 70 years. In the present study, we compared the gene expression changes in the MFC of elderly subjects with schizophrenia (patients vs. controls) with those in the developing MFC (infants vs. relatively young adults [20–39 years]; Figure
1C), which revealed significant similarities between the two groups (
P = 2.6 × 10
-8; Figure
2e). This result might indicate that the relative change in gene expression patterns that define transcriptomic infancy can also be seen in a cohort of elderly subjects with schizophrenia. We also found that gene expression patterns in the schizophrenic MFC are similar to those in the normal infant MFC as compared to elderly adult MFC (Additional file
3: Figure S2), suggesting that both the DLFC and MFC of patients with schizophrenia represent juvenile-like gene expression profiles. In addition, we found a significant similarity in the gene expression patterns of the superior temporal cortex (STC) between normal infants and patients with schizophrenia (Additional file
6: Figure S5). These findings suggest that juvenile-like gene expression profiles also can be found in brain regions other than the PFC in patients with schizophrenia.
Our present results also show that juvenile-like gene expression in the PFC of patients with schizophrenia could be due to immaturity in multiple cell types, including FS neurons, astrocytes, and oligodendrocytes. In the case of immaturity in FS neurons and oligodendrocytes, our results are consistent with those of previous studies analyzing molecular expression or structures in these cell types in schizophrenia[
5,
48]. Dysregulated gene expression in GABAergic neurons is one of the most robust findings in schizophrenia neuropathology. It is also a well-replicated finding that expression of PV, a marker for FS neurons, is decreased in the PFC of patients with schizophrenia[
49]. Considering that PV expression appears in the PFC only postnatally and drastically increases with age[
28,
29], PV is thought to be a marker for mature FS neurons. Therefore, FS cells were hypothesized to be immature in schizophrenia[
5]. In this context, Gandal
et al. developed an FS cell maturation index to evaluate maturity of FS neurons in the cortices of patients with psychiatric disorders including schizophrenia, bipolar disorder, and autism[
5]. Using time-course gene expression data from developing FS cells that were positively correlated with PV expression levels, Gandal
et al. showed a reduction of the index in cortices of patients with schizophrenia, bipolar disorder, and autism. In addition, a decrease in perineuronal nets (PNNs) has been reported in the schizophrenic PFC[
50]. Considering that PNNs are extracellular matrices predominantly enriched around mature FS neurons[
51], their decrease may imply the presence of pseudo-immature FS neurons in schizophrenic cortices. These results suggest that FS neurons stay at partially immature state in the cortices of patients with schizophrenia, which is consistent with our results.
Possible contributions of myelin and oligodendrocyte dysfunction to schizophrenia also have been suggested by many postmortem studies of the human brain at molecular[
48] and ultrastructural levels[
52]. This well-documented evidence for dysmyelination seems consistent with our finding of immaturity in oligodendrocytes. Specific markers identify the different stages of oligodendrocytes maturation: PDGFRα in oligodendrocyte progenitor cells, GalC in premyelinating oligodendrocytes, and MOG in myelinating oligodendrocytes. We used the data set for gene expression changes between premyelinating and myelinating mouse oligodendrocytes, which were purified based on maturation marker expression, as an index for oligodendrocyte immaturity in the present study[
38]. The gene expression changes during oligodendrocyte maturation showed a significant positive correlation with those shared by the human developing and schizophrenic DLFC, suggesting immaturity of oligodendrocytes in the DLFC of patients with schizophrenia.
In addition to the previous studies suggesting abnormal maturational status of FS neurons and oligodendrocytes in the schizophrenic brain, we showed that immaturity of astrocytes could also be seen in schizophrenic brains. Possible observations of changes in astrocyte densities in the cortices of schizophrenia patients are controversial; however, previous studies have shown the expression of several astrocyte-related genes is abnormal in these patients[
52]. For example, increased expression of GLT-1, a major glutamate transporter, has been reported in the PFC of patients with schizophrenia[
53]. Considering the expression of GLT-1 in rat astrocytes declines as they mature[
54], the increased GLT-1 expression in the schizophrenic PFC may imply the existence of astrocytes that partially resemble those in the immature brain. Such immaturity would support the idea that astrocyte function is linked to the pathophysiology of schizophrenia.
One might expect that, in patients with schizophrenia, a decreased expression of a large number of maturation marker genes might reflect the loss of cells expressing those genes (i.e., PV in FS neurons and MOG in oligodendrocytes). In one study, there was no significant change in the number of PV-expressing cells in the PFC of patients with schizophrenia, whereas the PV expression per cell decreased when compared with controls[
25]. The number of astrocytes between the schizophrenic and control PFC also were not significantly different[
55‐
57], even though expression of astrocyte-related genes is altered in the cortex of patients with schizophrenia[
52,
58]. Analyses of oligodendrocytes suggest that subtle oligodendrocyte or myelin anomalies, such as myelin sheath damage and a decreased mitochondria density[
59], may be more important than the changes in cell density associated with the pathophysiology of schizophrenia[
60]. In the frontal cortex of Shn-2 KO mice, no obvious hallmarks of neurodegeneration, including cell death, cell swelling, protein deposition, or nuclear condensation, were observed with immunohistological or electron microscopic analyses. The number of PV-positive cells was significantly reduced in the frontal cortex of Shn-2 KO mice when compared with controls[
11]. This suggests that the PV expression per cell decreased to undetectable levels in the mutant mice, which is similar to the pathology of schizophrenia in humans. Thus, in the cortex of patients with schizophrenia, a lower gene expression may reflect changes in gene expression in these cell types, rather than cell death.
Recent studies suggest that certain cell types in several brain regions of patients with schizophrenia may exhibit maturation abnormalities. By assessing the expression levels of maturational markers, Walton
et al. showed that the dentate granule cells of the patients may be persistently in pseudo-immature state[
8]. Expression of PV is decreased not only in the PFC but also in the hippocampus of patients with schizophrenia[
61]. Furthermore, a decrease in PNNs has been reported in the entorhinal cortex and amygdala of patients[
51,
62]. The decreases in PV and PNNs imply that FS neurons in those brain regions stay at pseudo-immature state. Expression of KCC2, a K
+-Cl
- cotransporter that plays a role in GABAergic neurotransmission, is decreased in the PFC and hippocampus of the patients[
6,
63]. Considering that KCC2 expression rises as brain development progresses[
64], the decreases indicate a pseudo-immature state in a certain type of neuron in the PFC and the hippocampus. In addition, we observed transcriptomic immaturity in the superior temporal cortex (STC) of the patients (Additional file
6: Figure S5). These findings suggest that maturational abnormalities can be seen in certain cell types in several regions across the brain of patients with schizophrenia.
Although we showed that about half of the genes representing juvenile-like expression patterns in the schizophrenic PFC were developmentally regulated in three cell types (Figure
3d, h, l), there is also a possibility that these altered expression signals are partly due to maturational abnormalities in other cell types. Gene expression patterns representing transcriptomic immaturity in the schizophrenic PFC (Bioset 1, 2, and 3) were similar to those in entire frontal cortex of developing mice (Additional file
7: Figure S6). Considering that pyramidal neurons are the major population in the cortex[
65], maturational abnormality of this cell type would contribute to transcriptomic immaturity in the schizophrenic PFC. Glutaminase is specifically expressed in pyramidal neurons[
66] and expression increases during development in human PFC (Additional file
1: Table S8), suggesting that this gene may be a maker for mature pyramidal neurons. The gene was found in all Biosets (Additional file
1: Table S8). This result raises the possibility that pyramidal neurons are also in the immature-like state in the schizophrenic PFC. To examine this question, data on developmental gene expression changes of specific cell types, such as pyramidal neurons, GABAergic neurons other than FS neurons, and microglia, are needed.
Important issues remain to be resolved: how and when the maturational abnormality phenomena emerge in the brains of patients with schizophrenia. Our results indicating the immaturity of multiple cell types in the schizophrenic PFC indicate three possibilities for how such immaturity might occur: failure in cell maturation, reversal of a once-established cell maturation, or recruitment of immature cells. The development and maturation of the brain has long been believed to be a one-way process. However, growing evidence suggests the maturational state is regulated bidirectionally for some cell types in adult brains. Established maturation of granule cells in hippocampal DG can be reversed by chronic treatment with the selective serotonin reuptake inhibitor fluoxetine, which is widely used as an antidepressant[
67,
68], and by the induction of spontaneous recurrent seizures[
69]. Mutation of Shn-2[
11] and SNAP-25[
10] may also reverse the maturational state of DG granule cells postnatally. In the DG of Shn-2 KO mice, 2-week-old animals showed no significant differences between genotypes in the expression of calbindin, a marker of mature granule cells, or calretinin, a marker of immature granule cells. Calbindin expression was decreased and calretinin expression was increased in the DG of 4-week-old Shn-2 KO mice compared with that of wild-type mice, thus indicating that an iDG phenotype emerged during postnatal development[
11]. In SNAP-25 KI mice, it was suggested that an iDG phenotype is caused by epileptic seizures that occur after P21–25[
10]. As for FS neurons, fluoxetine treatment induces an immature-like state in the visual cortex[
70], basolateral amygdala, hippocampus[
71], and the PFC[
72] in adulthood. Recently, it was shown that experience could reverse the differentiation state of FS neurons in the adult hippocampus, as monitored by PV expression levels, and the consequent plasticity may influence learning ability[
73]. Considering several environmental and genetic factors can induce a reversal in maturational status in these neuron types, it is possible that the immaturity of FS neurons in the schizophrenic PFC represents reversal of the maturational state. Demyelination induced by cuprizone, which could possibly be a partial dematuration of oligodendrocytes, was shown to cause schizophrenia-like symptoms in adult rodents[
74], suggesting that reversal of oligodendrocyte maturation may participate in the pathophysiology of schizophrenia.
The third possibility that might explain the transcriptional immaturity of the schizophrenic PFC is adult cortical neurogenesis. New neurons are generated in the cortex of adult rodents and primates under pathological conditions, including ischemia and chemical neurodegeneration[
75‐
78]. Considering that elevated inflammatory conditions have been reported in the cortex of patients with schizophrenia[
79], adult neurogenesis might be increased in the schizophrenic cortex[
80,
81]. A recent study demonstrated that the density of GABAergic interneurons increased in the white matter of the schizophrenic PFC[
82], which suggests that new neurons might be generated or recruited in the white matter of patients with schizophrenia. However, neurogenesis is upregulated by less than 1% of the total neuron count in the adult cortex under pathological conditions, such as focal or global ischemia, cortical tissue aspiration, or a laser-induced lesion[
83], suggesting that newly generated neurons can hardly account for the transcriptional immaturity of the schizophrenic PFC, even under pathological conditions. Although cortical adult neurogenesis cannot be excluded as a possible contributory factor, it might not be a major factor.
Some studies have addressed the issue of when the maturational abnormality phenomena emerge in the schizophrenic PFC. If the phenomena reflect the consequences of illness chronicity, the magnitude of the alterations would be expected to correlate with illness duration. However, it has been shown that neither illness duration nor age explain the expression levels of some genes related to GABAergic neurons, which are decreased in the schizophrenic PFC, suggesting that the gene expression changes in schizophrenia are not a consequence of illness chronicity[
84]. It has also been shown that normal age-related decreases in expression of genes related to central nervous system developmental systems do not occur in patients with schizophrenia during the aging process[
7], suggesting that disturbances in gene regulatory mechanisms appear before clinical onset or at an early stage of clinical illness. In the present study, we compared data sets from the DLFC of patients with schizophrenia, who were grouped according to age (defined by Narayan
et al.[
32]), to the two data sets from the developing DLFC (developmental data sets in Figure
1a and b). We found the highest overlap in
P-values in comparisons with the short-DOI schizophrenia group (Additional file
1: Table S2). Gene expression patterns representing a juvenile-like phenotype are more likely to be associated with younger patients than with older patients. Thus, the altered gene expression that causes abnormalities in neural maturation in schizophrenia seems to emerge during postnatal developmental stages prodromally or concomitantly with clinical onset.
Shn-2 KO mice were previously described as showing multiple schizophrenia-like phenotypes at molecular, anatomical, electrophysiological, and behavioral levels[
11]. The transcriptome pattern in the MFC of the Shn-2 KO mice, which we found to be similar to that of human infant MFC in the present study, is also highly similar to that from postmortem patients with schizophrenia[
11]. In the present study, we found that genes related to FS neuron development, rather than oligodendrocyte or astrocytes development, showed significant overlap with genes that are commonly up- or downregulated in the MFC of Shn-2 KO mice and human infants (Figure
4). Although the altered gene-expression signals linked to the development of FS neurons may have been derived from maturational abnormalities in other cell types, pseudo-immaturity of FS neurons specifically has been suggested by findings in the MFC of Shn-2 KO mice. We previously reported a decrease in the number of PV-expressing neurons in the MFC of Shn-2 KO mice[
11]. Expression of PV in FS neurons in the immature brain was quite low (only one-sixtieth of the adult level)[
37], suggesting that the expression of PV could be undetectable in immature FS neurons. Thus, the decrease in the number of PV-expressing neurons in the MFC of Shn-2 KO mice can be at least partly explained by the immaturity of FS neurons.
The results of postmortem analyses often can be confounded by the effects of medications. However, in the genetically engineered rodent model, we can exclude the effect of drugs and control other environmental factors that are not study variables. Therefore, results showing a significant similarity in gene expression patterns between the human developing MFC and the Shn-2 KO mouse MFC suggest that the transcriptional immaturity observed in the schizophrenic PFC may not be due to medication. We also showed that the PFC of rodents treated with antipsychotics exhibited no apparent similarities with the human developing PFC, except for a comparison between olanzapine-treated rodent PFC and the human developing MFC (Additional file
5: Figure S4). These results suggest that medication effect may not be a major contributing factor to the transcriptional immaturity found in the schizophrenic PFC. Other studies also suggest that a decreased expression of GABAergic neuron-related genes in the schizophrenic PFC, including PV, seems unrelated to confounding factors such as medication or substance use[
25,
84,
85].
To date, many hypotheses have been developed to explain the mechanisms of schizophrenia. Immaturity of the PFC seems to be consistent with several major hypotheses, such as the neurodevelopmental[
9,
86], oligodendrocyte[
48,
87], and inflammation hypotheses[
79,
88]. The immaturity of FS neurons literally reflects a neurodevelopmental problem. Dysfunction of oligodendrocytes characterized by demyelination could be accounted for by immaturity of oligodendrocytes, as discussed above. Pathway enrichment analyses showed that enrichment of inflammation-related pathways is likely to be accompanied by a juvenile-like PFC phenotype (Additional file
1: Table S6, S7). Activation of the NADPH-oxidase/interleukin-6 (IL-6) pathway, which is known to play an important role in inflammatory processes, could increase superoxide production in the brain and induce a reversible loss of PV-positive cells in adulthood[
89]. In a previous study that used Next Generation Sequencing Expression, IL-6 mRNA was increased in the PFC of patients with schizophrenia, suggesting an increase in inflammation in the schizophrenic PFC[
90]. Similarly, oxidative stress increased significantly in the schizophrenic PFC when compared with controls[
91,
92]. These findings suggest that neuronal pseudo-immaturity in the PFC could be induced by brain inflammation, followed by PFC dysfunction, such as deficits in attention, working memory, and executive functions, which are symptoms of schizophrenia. Together, these findings could be a link between inflammatory conditions and evidence of neuronal immaturity. Enrichments in energy metabolism- and mitochondria-related pathways also were found for genes representing a juvenile-like PFC phenotype (Additional file
1: Table S6, S7, S9). Energy metabolism mediated by mitochondria plays an important role in the development and maintenance of mammalian brains[
93]. Because FS neurons have high metabolic demands and show dramatic upregulation of energy related genes during development, it has been hypothesized that defects in energy metabolism genes impair FS neuron development[
5]. Thus, enrichments in energy metabolism-related pathways are likely related to abnormalities in cell maturational status, especially in the FS neurons examined in this study. Taken together, these data show that a transcriptional immaturity in the PFC can be considered an endophenotype of schizophrenia, which is consistent with several etiological hypotheses.
Collectively, our results demonstrate that the genome-wide expression profile of the schizophrenic PFC resembles that of the juvenile PFC, which could be due to immaturity in multiple cell types, including FS neurons, astrocytes, and oligodendrocytes. Given that the pseudo-immature cells, especially FS neurons, are not actually lost or absent from the schizophrenic PFC, attempts to restart the normal maturation process could be a potential therapeutic strategy. Considering that adult Shn-2 KO mice also have a juvenile-like PFC, treatments that induce PFC maturation in the mouse model might be candidate therapies for schizophrenia. For example, chronic administration of anti-inflammatory drugs lessened the immaturity of the DG granule cells and some behavioral abnormalities in Shn-2 KO mice[
11]. Further investigation of transcriptional immaturity in the PFC as a factor in the precipitation of, as well as recovery from, episodes of schizophrenia would facilitate study of the disorder.