Introduction

George Somjen was prescient in his comments about glial cells in 1988 [1]. Indeed, the crucial role of glia in psychiatric diseases has been neglected over the past two decades [2, 3]. Glial cells are involved in every major aspect of brain development, function, and disease by communicating with neurons and releasing neurotransmitters and other signals [4]. In the central nervous system (CNS), astrocytes, oligodendrocytes, and microglia are the three main types of glial cells, which play an important role in synaptic function, neuronal metabolism and migration [3]. Over the years, there has been a shift in our understanding of the relative number of glial cells, and detailed studies suggest that the proportions vary greatly by brain region [5]. Schizophrenia (SCZ), bipolar disorder (BD), and major depressive disorder (MDD) are three major psychiatric diseases, with some similarities in their occurrence and symptoms, at least in the early stages [6,7,8]. A growing body of research is attempting to clarify the pathogenesis associated with these mental disorders. Genetic studies have shown a high correlation between multiple genes involved in the regulation of the immune system and SCZ, BD, and MDD [9]. Postmortem evidence supports the role of cerebral inflammation in the etiological pathways of these mental disorders [10,11,12]. Nevertheless, the neurobiological mechanisms underlying these diseases are still not fully understood.

In the literature related to major psychiatric diseases such as SCZ, BD, and MDD, there have been a large number of reports on glial cell dysfunction, which undoubtedly provides evidence of a relationship between psychiatric disease and glial pathology [13,14,15]. However, there are problems in these studies, such as the diversity of results due to differences in the studied brain regions and research methods. It is difficult to draw straightforward conclusions based on these results. To help improve our understanding of the pathogenesis and pathophysiology of psychiatric diseases, we systematically reviewed postmortem studies on the structure of brain cells in patients with SCZ, BD, and MDD. The density, number, and size of the astrocytes, microglia, and oligodendrocytes, as well as the expression of cell type-specific markers in postmortem brain samples, were evaluated. By emphasizing the importance of glial cells in these mental diseases, we have provided a preliminary conclusion on the changes of glial cells in the postmortem brain and a guideline for future studies to improve our understanding of the mechanism of mental diseases and to identify new therapeutic targets.

Methods

We performed a literature search for records indexed within PubMed and Web of Science up to May 1, 2021. The search strategy was ‘(glia or microglia or astrocytes or oligodendrocytes) and (schizophrenia or bipolar disorder or major depressive disorder or depression) and (postmortem or brain sample)’.

Studies were screened for relevance based on their title and abstract by two researchers independently. The full text of potentially relevant articles was retrieved and screened against the following inclusion criteria: (1) studies that focused on postmortem brain samples in SCZ, BD, or MDD; (2) measured glial cells, including microglia, astrocytes, or oligodendrocytes, as well as several related markers; and (3) matched psychiatric-disease-free controls. Studies with matched samples (by age, sex, race, brain pH or postmortem interval, etc.) [16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109] or studies that statistically adjusted for these specific variables (age and sex primarily) [110,111,112,113,114,115,116] were included. Duplicates and studies that did not meet the above criteria were excluded. In addition, review articles, in vitro studies, and animal studies were excluded.

Eligible studies were assessed and the data were extracted into an Excel spreadsheet by the researcher, and any disagreements were resolved by discussion. For the eligible studies, the first author’s name, publication year, brain bank, sample size, sex, age, and death from suicide were extracted as background information. In addition to glial cell markers measured, measuring techniques and in which brain regions the measurements were made were all extracted, along with comparative results between the patient subjects and the healthy controls.

Results

Our search strategy resulted in the identification of 1070 unique studies from the initial search. After screening of the titles and abstracts for relevance, 175 articles were full-text screened against the inclusion criteria. Out of the 175 articles, 69 articles were excluded because they did not measure glial cells (50 studies); data not separable from other diagnostic groups (9 studies); did not include any psychopaths (6 studies) and healthy controls (2 studies); and incorrect measurement methods, such as single-cell sequencing (2 studies). Two more articles were found in the reference section of papers identified in the database search. Thus, a total of 108 studies were ultimately included in this review, including 31 studies on total glial cells that did not differentiate between glial cell types, 41 studies on astrocytes, 24 studies on microglia, and 30 studies on oligodendrocytes (Fig. 1).

Fig. 1
figure 1

Flow chart of the systematic search. Asterisk indicates that various studies report on more than one cell type

Full size image

The regions of the brain studied among our reviewed papers mainly included the anterior cingulate cortex (ACC), anterior cingulate gyrus (ACG), anterior midcingulate cortex (AMC), anteroventral nucleus (AVN), corpus callosum (CC), caudate nucleus (CN), dorsolateral prefrontal cortex (DLPFC), dorsal raphe nucleus (DRN), entorhinal cortex (EC), frontal cortex (FC), hippocampus, locus coeruleus (LC), medial frontal gyrus (MFG), occipital cortex (OC), orbitofrontal cortex (OFC), prefrontal cortex (PFC), superior frontal gyrus (SFG), superior temporal gyrus (STG), temporal cortex (TC), amygdala, cerebellum, putamen, subiculum, and thalamus (Table 1; Figs. 2 and 3).

Table 1 Summary information of the included studies
Full size table
Fig. 2
figure 2

Postmortem evidence of glial abnormalities in different brain regions in SCZ, BD, and MDD

Full size image
Fig. 3
figure 3

Postmortem evidence of glial abnormalities in different brain regions in SCZ, BD, and MDD

Full size image

Total Glia

In our review, a total of 31 studies evaluated total glial cells (astrocytes, microglia, oligodendrocytes) in postmortem brain samples without the use of cell type-specific markers.

Several studies looking at total glia found that decreased cell density was associated with SCZ. Cotter et al. [37] published their first study in 2001, with glial cell density measured by cresyl violet staining, and observed a decrease in layer VI of the ACC in SCZ and MDD compared to healthy controls, whereas no change was observed in BD. In their second year of study [35], comparable effects were observed again in layer V of the DLPFC. However, in their two subsequent studies, neither Heschl’s gyrus [36] nor OFC [34] observed evidence for changes in glial cell size or density between the three groups of mental diseases and healthy controls. Beasley et al. [18] published a study on the effects on glial cells in 2009. In the white matter near the temporal plane of 15 patients in each group, they found that only schizophrenic patients, not the BD or MDD patients, had a lower glial cell density than the control subjects. Similar decreases in glial cell density were observed in the CA3 and CA4 regions of the hippocampus [41] and layer III of the primary motor cortex [19] of patients with SCZ.

In fact, changes in glial cells were also observed in postmortem brains from BD and MDD patients. For example, Rajkowska et al. [73] found that the density of glial cells decreased in sublayer IIIc of the DLPFC in patients with BD, coupled with enlargement and changes in the shape of the glial nuclei. Brauch et al. [24] also found that the area occupied by the glial cells was reduced in the TC of bipolar patients compared with that of healthy controls. In a study on the amygdala, Nissl staining showed that the density of glial cells decreased only in the MDD group, while there was no significant difference between the BD and control groups [50]. In contrast, glial cell density in the hippocampus was significantly increased in 19 major depressive patients compared with healthy controls [91].

Twenty-two studies evaluating glial cells did not detect any changes associated with SCZ, BD, and MDD. In a cohort of 18 schizophrenic patients [29], there was no difference in the numbers of glial cells in the PFC and ACC, regardless of whether the patients had superimposed mood disturbances. In another cohort of 18 schizophrenic patients, quantified analysis after hematoxylin and eosin staining showed no difference in glial cell density in the CC compared to healthy controls. However, it is worth mentioning they found gliosis in the CC in patients with late-onset SCZ compared with early-onset SCZ and controls [68]. In a study of postmortem brain samples from the Stanley Foundation Neuropathology Consortium, for SCZ, BD, and MDD, Nissl staining revealed no differences in the glial cell density or size in the amygdala [22] and no changes in glial number in the TC [24]. Similarly, whether from the EC [42], PFC [113], OC [113], dentate gyrus [23], fusiform gyrus [78], FC [82], DLPFC [82], lateral geniculate nucleus [83], and primary auditory cortex [87] of SCZ or the OC [75], basolateral amygdala [78, 114], and auditory cortex [87] of MDD, Nissl staining failed to detect any changes in the density or number of glial cells. Cresyl violet staining revealed no changes in glial cell density in the ACC [20, 30] and PFC [117] from SCZ or BD. Gallocyanin, another staining technique, also did not detect an effect of SCZ on glial cell density in the dorsal ACC [53]. Moreover, in Khundakar’s four studies of the MDD postmortem brain, no changes in glial cell density were obtained in the DLPFC [56], OFC [57], CN [58], and ACG [59] compared with healthy controls.

Astroglia

Our search strategy yielded a total of 41 studies assessing astrocytes in postmortem brain samples from patients with major psychiatric diseases.

Glial fibrillary acidic protein (GFAP) is often used as an astrocyte-specific marker in postmortem studies [118]. Among 41 studies, 34 studies evaluated differences in astrocytes by measuring the expression of GFAP or the distribution of immunoreactivity.

The first study to evaluate GFAP was published in 1993. Karson et al. [54] measured the protein level of GFAP in the brains of 25 patients with SCZ by western blot and found no differences in multiple brain regions, including the FC, STG, OC, cerebellum, thalamus, and pons relative to healthy controls. Fatemi et al. [46] later published a study of lateral cerebellum samples from the Stanley Foundation Brain Collection, in which they found that GFAP protein levels were reduced to different degrees in SCZ, BD and MDD. However, only the MDD value remained statistically significant following correction. This is in agreement with several subsequent studies, where reduced GFAP protein levels were observed in the DLPFC [86], LC [119], and OFC [66] of MDD. However, different from the result observed in MDD, two studies detected increased protein levels of GFAP in the DLPFC [47, 63] in SCZ patients compared to healthy controls. Similar increases were also observed in the DLPFC [47] and FC [77] in BD patients.

Similar to the research on GFAP mentioned above, several studies have reported increased GFAP mRNA expression in the brain from SCZ and BD. For example, Farnsworth et al. [110] found that schizophrenic patients had increased GFAP mRNA levels in the PFC and SFG. Qi et al. [72] investigated the expression of GFAP mRNA in the gray matter isolated from the ACC and DLPFC samples from patients with BD and MDD and found that only BD patients had significantly higher levels of GFAP in the ACC than control subjects. The authors did note, however, a lower area fraction of GFAP immunoreactive astrocytes in the ACC of BD. Although GFAP mRNA, as measured by a riboprobe, was decreased in the white matter of the ACC in SCZ and BD [106], Barley et al. [17] again observed that the levels of GFAP mRNA were increased in several brain regions including AVN, putamen, internal capsule, and mediodorsal thalamic nuclei, in patients with SCZ and MDD. In contrast, most studies evaluating GFAP mRNA expression of various brain regions of MDD, including the LC [21], gray matter of the PFC [67], mediodorsal thalamus [97], and CN [97], detected a reduced GFAP mRNA expression in MDD. Not all studies, however, have detected differences in GFAP mRNA expression. For example, no changes were detected between SCZ patients and healthy controls in several brain regions, including the PFC [93], MFG [115], CC [85], and DLPFC [62]. Similarly, no changes in GFAP mRNA were observed in the DLPFC and ACC in patients with BD compared with healthy controls [109].

Other studies have assessed GFAP expression by immunohistochemical analysis. Rajkowska et al. [74] published a study in 2002 of a cohort of 9 schizophrenic brains and found that GFAP-positive astrocyte density in layer V of the DLPFC increased by 81%, whereas the GFAP labeling area was reduced by 32%. These changes were layer-specific, as no difference was detected between layers III and IV. This is slightly different from what two subsequent studies both observed that there were no differences in the density of GFAP cells in the DLPFC [43, 51] of patients with SCZ, although Hercher et al. [51] did find a decreased GFAP fraction area and increased cell clustering in the DLPFC in their study. Similarly, many quantitative studies found no differences in the density of GFAP cells in multiple other brain regions in patients with SCZ, including the hippocampus [69], EC [39, 69, 120, 121], amygdala [39], ACG [43], subiculum [69, 120], STG [43], primary motor cortex [120], DLPFC [28, 43], and subventricular zone [120]. Although Williams [108] and his colleagues examined the cellular structure of the substantia nigra in the postmortem brain with schizophrenia and MDD and found a decrease in astrocyte density in SCZ patients relative to healthy controls, this effect was not detected in patients with MDD.

However, many quantitative immunohistochemical studies have detected a decrease in the density of cells expressing GFAP in the brains of patients with MDD. For instance, Altshuler et al. [16] found a decrease in the density of GFAP-immunoreactive astrocytes in the amygdala of subjects with MDD compared to SCZ, BD, and healthy control postmortem samples. Similarly, in the study of Cobb et al. [33], the authors found that the density of GFAP-immunoreactive astrocytes in the dentate gyrus of MDD patients who did not take antidepressants was significantly lower than that in controls, whereas no change was found in CA1 or CA2/3. Other studies have also found that MDD patients had a lower GFAP-immunoreactive astrocyte density in the LC [119], DLPFC, dorsal CN, and MTN [70]. Nevertheless, not all immunohistochemical studies have detected this effect in MDD. No changes in GFAP cell density in several brain regions, including the ACC [38], hippocampus [48, 112], DLPFC [65], and substantia nigra [108], were observed in other studies.

Other astrocytic markers have also been measured in postmortem brain specimen from psychosis. With the increase in GFAP, aldehyde dehydrogenase (ALDH)1 mRNA in the SCZ and MDD was increased in several brain regions, including the putamen, AVN, internal capsule, and mediodorsal thalamic nucleus [17]. In contrast, two studies found no change in ALDH1 L1 mRNA in the DLPFC of patients with SCZ and BD [47, 109], whereas one other study detected a decrease in the PFC gray matter of MDD [67].

Hamidi and his colleagues examined the astrocyte marker S100b [50]. They did not observe changes in astrocyte density in the amygdala of patients with BD and MDD. Gos et al. [48], however, did find that the numerical density of S100b-immunopositive astrocytes was significantly decreased in the CA1 pyramidal layer in patients with BD and MDD compared to healthy controls. S100b has been measured by qPCR in a few other studies with mixed results. One study found a decrease in S100b mRNA in the LC of patients with MDD [21], and another found no change in the ACC and DLPFC in either BD or MDD [109].

Astrocytes have also been identified in postmortem brain by microscopic analysis with other staining techniques. In a cohort of 10 depressed suicides from the Quebec Suicide Brain Bank, Golgi-impregnated fibrous astrocytes had significantly larger cell bodies and longer, more ramified processes in depressed suicides [96]. This is consistent with another study, where examining astrocyte morphology by immunohistochemistry showed that astrocytes in both the thalamus and CN displayed larger cell bodies and extended more ramified processes across larger domains than cortical astrocytes [97]. However, two stereological counting studies on Nissl-stained astrocytes showed no differences in the number of hippocampal astrocytes, whether in SCZ, BD or MDD [63, 79, 122].

Microglia

From our search, a total of 24 studies assessed microglial marker in postmortem brain samples. Out of these 24 studies, 11 studies reported increased microglial markers in the postmortem brain, whereas 6 studies reported a decrease, and 7 studies found no change.

Cluster of differentiation (CD) is a microglial marker, and multiple studies have found lower CD gene expression in postmortem brains associated with SCZ. For example, in one study, downregulation of CD68 mRNA levels in the DLPFC was detected by qPCR in SCZ patients compared with healthy controls [62]. This is similar to the results of a recent study [85] in which CD68 mRNA levels were also decreased in the CC of SCZ. Similar decreases in the expression of CD68 and CD11b mRNA were observed in the ACC of BD patients [84], despite the expression of CD11b mRNA and protein in the FC was increased [77]. Moreover, Zhang et al. [109] reported a decrease in CD68 mRNA in the DLPFC of BD patients without suicide compared to BD patients with suicide and controls. This effect, however, was not seen in the ACC. Similarly, no difference in CD68 mRNA was detected in the dorsal ACC in MDD [95].

HLA antigen D-related (DR) is an immunohistochemical marker that specifically reacts with activated microglial cells. Bayer et al. [120] found that 3 of 14 schizophrenic patients had positive HLA-DR staining in the hippocampus and FC. This is in agreement with a study conducted the following year, where HLA-DR was increased in the DLPFC and STG [43], and although a similar increase was observed in the ACG, the results were not significant [43]. This increase in the HLA-DR marker appears to be more pronounced in the hippocampus of patients with paranoid SCZ, as HLA-DR was increased in this group compared with the residual SCZ group, but it was not significant compared to the control group [26].

In a microarray analysis, decreases in HLA-DRA and HLA-DRB4, subunits of HLA-DR, and mRNA expression were observed in the temporal lobe of SCZ patients compared to healthy controls [40]. Similar decreases in HLA-DRA1 and HLA-DPB3 mRNA expression were detected in the superior TC of SCZ [80]. The decreases, however, were not statistically significant when mRNA expression was confirmed by qPCR [80].

Other immunohistochemical studies have assessed microglial activation by measuring ionized calcium-binding adapter molecule (IBA)1. In a cohort of 24 patients with MDD, a significant increase in the density of IBA1-immunoreactive amoeboid-like cells was found in the surrounding blood vessels in depression suicide, accompanied by an increase in IBA1 gene expression, although the total microglial densities in the dorsal ACC did not change [95]. This was confirmed in a subsequent study with a larger cohort, in which amoeboid microglial cells were increased in the ventrolateral PFC in depressed patients compared to healthy controls, but the overall density was very low [32]. However, IBA1 as measured by immunohistochemistry, showed no differences in microglial density in several brain regions of SCZ, such as the DLPFC [51], AMC [71], ACC [84], MFG [115], CC [115], PFC [98], and STG [88], as well as the DLPFC [51, 109], ACC [84, 109], and MFG [116] in BD.

Several anatomical studies have also examined microglial density in postmortem brain samples. In the two studies by Steiner et al. [89, 90], there were no changes in microglial density in various brain regions between the psychiatric group and the control group, but it was noted that suicide was accompanied by higher microgliosis. This effect is similar to a recently published study, where there was no significant difference in microglial density among individuals with SCZ, BD and control subjects. However, there was significantly higher microglial density in suicidal BD individuals than in nonsuicidal BD individuals [71]. Similarly, Brisch et al. [25] evaluated HLA-DR-positive microglial cell density in the DRN and found that nonsuicidal depressed patients revealed significantly lower microglial reactions than controls. In addition, two other studies reported no changes in microglial density in the hippocampus [48] of SCZ and the amygdala [50] of BD and MDD.

Changes in microglial cells have also been investigated by electron microscopy. In a cohort of 21 schizophrenic patients, patients with positive and negative symptoms of SCZ both showed significant microglial activation and dystrophic alterations in layer V of the gray matter in the PFC, although the microglial density did not differ from the control group [101].

Oligodendrocytes

Oligodendrocytes have been measured in postmortem brain samples from patients with major psychiatric diseases in 30 studies. Of the 30 studies, 20 studies reported decreased oligodendrocyte-related markers in the postmortem brain, whereas 2 studies reported an increase, and 8 studies found no change.

Several studies assessed differences in the expression of oligodendrocyte-related or myelin-related genes in postmortem brain samples. Tkachev et al. [93] reported a reduction in key oligodendrocyte- and myelin- related genes, such as oligodendrocyte lineage genes (Olig)2, myelin oligodendrocyte glycoprotein (MOG), and coding region of proteolipid protein (PLP)1, in SCZ and BD, and these gene expression changes in the PFC for both disorders showed a high degree of overlap. Wesseling et al. [107] used a labeled multiplexed selected reaction monitoring assay and found that in the PFC of patients with BD, four oligodendrocyte-specific proteins, 2′,3′-cyclic nucleotide 3′-phosphodiesterase (CNPase), MOG, myelin basic protein (MBP), and myelin proteolipid protein (MYPR), showed a decrease, whereas this decrease was less pronounced in SCZ and MDD. However, in other brain regions, such as the AVN, putamen, internal capsule, and mediodorsal thalamic nuclei, the mRNA expression levels of oligodendrocyte-associated genes, such as CNPase, MOG, galactosyl ceramidase (GALC), and myelin-associated glycoprotein (MAG), were only low in SCZ, whereas no significant difference in BD and MDD [17]. The downregulation of MOG [111] and MBP [63] expression seems to be closely related to SCZ. In a cohort of 10 long-term schizophrenic brains, in the middle layer of the STG and stratum lucidum of CA3 in the hippocampus, the thickness of the MOG-positive fiber-like structures decreased significantly along with reduced MOG expression [111]. In a recent study [94], the thickness of MOG-positive fibrous structures also decreased in the middle layer of the STG, regardless of the presence of 22q11DS in patients with SCZ compared to controls. MBP protein expression was decreased in the DLPFC of patients with SCZ in a proteomic analysis [63]. Furthermore, two studies on the expression of oligodendrocyte-related genes in the brains of patients with MDD have yielded conflicting results [76, 92]. Rajkowska et al. [76] detected oligodendrocyte-related mRNA expression of CNPase, PLP1, MBP, MOG, MOBP, oligodendrocyte transcription factor (OLIG) 1, and OLIG2 in the PFC by qPCR and found that PLP1 mRNA levels were significantly reduced in MDD, whereas there was increased mRNA expression of CNPase, OLIG1, and MOG. No changes in protein expression for CNPase, PLP1, MOG, and MAG were detected in the ventromedial prefrontal white matter of suicidal patients with MDD, but there was decreased Olig2 and MBP protein expression [92].

The majority of studies have reported that changes in the number or density of oligodendrocytes are associated with SCZ. Byne et al. [27] reported a decrease in oligodendrocyte number in the anterior principal thalamic nucleus in men with SCZ; although women had a similar trend, the differences were not significant. In one of his subsequent studies [31], however, there were no changes in oligodendrocyte-specific genes in the thalamus of schizophrenic patients, including MAG, CNPase, and MBP. In fact, CNPase is a marker of oligodendrocyte progenitor cells in addition to differentiating and myelinating oligodendrocytes. Hof et al. [52] reported a decrease in oligodendrocyte total number in layer III of the SFG in 7 schizophrenic brains compared with healthy controls, which was measured by CNPase staining and confirmed by Nissl counts. Another study evaluated oligodendrocyte numbers in the entire hippocampus of 10 schizophrenic patients and found a significant decrease in CA4 in the hippocampus [44]. No effect of SCZ was observed in the dentate gyrus, CA1,2/3, and subiculum [44]. Mauney et al. [64] quantified cells that were immunoreactive for neural/glial antigen (NG)2, a selective marker for oligodendrocyte progenitor cells (OPCs), and those that were immunoreactive for OLIG2, an oligodendrocyte lineage marker of mature oligodendrocytes, and found that there was no significant change in the density of NG2-immunoreactive cells in the PFC of patients with SCZ, but the OLIG2-immunoreactive cell density was decreased significantly. The authors suggest that impaired OPC differentiation and myelin sheath lesions appear to be involved in the pathogenesis of SCZ [64]. A disintegrin and metalloprotease (ADAM)12, a member of the family of multidomain metalloprotease-disintegrins that possess cell-binding, cell-signaling and proteolytic properties [45], might also be involved in the pathophysiology of SCZ since the author found that the density of ADAM12-immunoreactive oligodendrocytes in the white matter of the ACC was significantly decreased in schizophrenic patients [45]. Kolomeets et al. [60] found that the density of oligodendrocytes was reduced in layer V of the PFC in patients with SCZ. In their next study with a large sample size, they also found that the numerical density of oligodendrocytes and oligodendrocyte clusters was reduced in the anterior putamen in SCZ [61]. Although this effect was not observed in BD and MDD, the density of oligodendrocyte clusters was significantly reduced in all male clinical cases compared to male controls [61]. Reductions in oligodendrocyte density were also observed in other brain regions of schizophrenic patients, such as the thalamus and [55] layer VI of the PF [102]. This effect, however, was not seen in the anterior cingulum bundle [81] or substantia nigra [81].

A few studies have also detected differences in the density of oligodendrocytes in the postmortem brains of patients with BD and MDD. For example, Gos et al. [48] reported a decreased density of S100B-immunopositive oligodendrocytes in the left alveus of the hippocampus in a cohort of 6 BD patients compared to the MDD and control groups. In contrast, in another stereological study of the posterior hippocampus [122], Nissl staining revealed that BD patients had significantly more oligodendrocytes in the CA1 region of the hippocampus than healthy controls. Similarly, in patients with MDD, the density of oligodendrocytes was higher in the CA2/3 region, CA4 region, and subiculum of the hippocampus. The authors also noted that antidepressant doses correlated with the density and number of oligodendrocytes in CA2/3 [122]. Although one study reported an increase in oligodendrocyte density in the DLPFC of patients with BD [123], in the other two studies, varying degrees of decrease in oligodendrocyte density were observed in the amygdala [50] and layer VI of the PFC [102] in MDD.

The morphological changes in oligodendrocytes have also been investigated by electron microscopy. Vostrikov et al. [104] reported that SCZ patients had a significantly lower number of pericapillary oligodendrocytes than controls. The author also noted obvious ultrastructural dystrophic and degenerative alterations of pericapillary oligodendrocytes in the PFC of schizophrenic brains. In two studies by Uranova et al. [99, 100], significant ultrastructural changes in oligodendrocytes were observed in the PFC [99, 100] and CN [99] of SCZ and BD, despite no changes in the density of oligodendrocytes. This is consistent with what Vikhreva et al. [103] observed, where the ultrastructure of the oligodendrocytes was changed, which was observed in layer IV of the PFC white matter in SCZ. These pathological changes in oligodendrocytes are aggravated when antipsychotic drugs are administered [105].

Discussion

Although the pathogenesis of SCZ, BD and MDD is largely unclear, accumulating postmortem brain sample studies allow for the study of the pathological mechanism of these psychiatric diseases at the cellular level. Therefore, we systematically reviewed the literature reporting glial cell number, density and cell type-specific markers in the postmortem brains of patients with SCZ, BD and MDD.

We tried to elucidate a trend for each cell type in the different diseases, although most of the study designs and results were heterogeneous. Astrocyte dysfunction appears to be a unique pathology in MDD, as multiple autopsy studies have reported a decrease in the number/density of GFAP-reactive astrocytes in several brain regions of patients with MDD, including the LC [119], OFC [66], amygdala [16], and hippocampus [33]. Similarly, decreases in GFAP gene and protein levels were also observed in other brain regions, including the cerebellum [46], DLPFC [86], PFC [67], thalamus, and CN [97] in patients with MDD. However, not all MDD studies have detected this effect. One plausible explanation is that GFAP immunoreactivity varies with the duration of depression because the GFAP protein levels [86] and the packing density of GFAP-immunoreactive astrocytes [65] are positively correlated with the age of death, which may reflect a compensatory response to neuronal injury in elderly patients with MDD [75]. This may also partly explain why Davis and his colleagues [38] observed an increase in the density of GFAP-immunoreactive astrocytes in MDD brains with an average age of 75. Thus, the pattern of astrocyte pathology in the cerebral cortex in younger patients with MDD appears to be different from that in older patients.

Antidepressants are also most likely a confounding factor for this heterogeneous result, as antidepressants have been found to alter the expression pattern of glia-specific genes, including genes encoding GFAP, vimentin, and aquaporin, and affect glial cell numbers [124]. In this respect, it is important to know that a study using the chronic social defeat stress paradigm as an animal model for depression demonstrated that chronic exposure to antidepressant medications (fluoxetine) can counteract the significant reduction in the number of astrocytes induced by stress, resulting in hippocampal astrocyte numbers in the control range [125]. In our systematic review, supporting evidence from Cobb and his colleagues indicated that depressed people who were taking antidepressants (including fluoxetine) had more GFAP-immunoreactive astrocytes in the dentate gyrus [33]. One possible explanation for such findings is that antidepressant treatment may influence glial cell numbers by affecting glial cell proliferation, although the evidence thus far has shown only that fluoxetine-treated experimental animals have an increase in gliogenesis in the prefrontal cortex [126, 127]. Together, although controversial and somewhat inconsistent, the current evidence from postmortem MDD brain studies tends to support the hypothesis that astrocytic pathology represents a prominent feature of MDD and participates in the pathogenesis of MDD.

Astrogliosis was previously considered to be the basis of the changes in the morphology and number of astrocytes in SCZ. However, consistent evidence for astrocytic pathology in postmortem brains from patients with SCZ has proven elusive, in part due to some conflicting findings in the included studies. Investigations of astrogliosis include studies measuring not only GFAP mRNA and GFAP protein but also glial morphology and size. From our search, multiple studies on GFAP expression of genes/proteins have reported increased levels of GFAP gene and protein in different brain regions of SCZ [47, 63, 74, 110]. However, few immunochemical studies have detected an increased number/density of GFAP astrocytes [39, 43, 48, 51, 66, 120, 121]. Antipsychotic medications appear to be an important factor, which was supported by studies in which increased GFAP mRNA levels [17], GFAP protein levels [28], and GFAP immunoreactivity [128] were significantly correlated with lifetime antipsychotic treatment. Therefore, whether astrogliosis is a pathological feature of SCZ and the influence of antipsychotic treatment on glial pathology of SCZ needs further study.

Similarly, our data show that conflicting findings have also been found in studies evaluating microglia in the brains of SCZ patients postmortem. In support of the fact that most studies have not observed a difference in microglial density, several studies have observed changes in gene expression between controls and SCZ patients. One possible explanation is that microglia are known to be highly reactive and can show changes in a host of genes or proteins, which may or may not reflect changes in their morphology or quantity. Our results show that a reduction in CD68 mRNA levels was found in the DLPFC [62] and CC [85] of SCZ patients, which partly reflects that microglial activation was decreased in these brain regions of SCZ patients, as CD68 is generally thought to be highly expressed in round/activated microglia [129]. In two studies, however, more activated microglia were observed in the cortical areas of postmortem brains of several individuals with SCZ [51, 120]. The confounding factor may be that psychiatric patients include some suicide victims, since studies have shown that the suicidal brain may have high levels of pro-inflammatory cytokines [130]. This is consistent with the findings of two studies by Steiner et al. [89, 90] where patients with any psychiatric condition who committed suicide had the highest number of HLA-DR-positive cells. However, when suicide victims were considered, there were no differences in the same groups between diagnostic groups. The same effect was also observed for BD and MDD, and patients who committed suicide tended to have a higher microglial density [25, 71]. Furthermore, the cohort of healthy controls also included suicide victims, which may potentially confound the results.

In postmortem brain studies of oligodendrocytes, oligodendrocyte density and the expression levels of myelin-related genes, such as MAG, Olig2, and CNPase, were decreased in SCZ. A genetic study has shown that variations in oligodendrocyte-related genes affect the microstructural integrity of white matter bundles and cognitive performance in SCZ [131]. Repeated findings of the downregulation of these genes in the brains of postmortem patients with SCZ provides supporting evidence, although the hypothetical function of oligodendrocyte-related genes and white matter structural integrity is currently unclear. Oligodendrocyte dysfunction leads to changes in synaptic formation and function, which in turn leads to cognitive dysfunction, which is considered to be one of the core symptoms of SCZ [132, 133].

However, not all studies have detected this effect, and a plausible explanation is that there may be sex differences in gene expression, since Byne et al. [31] found higher expression of CNPase and MAG in females compared to males in the thalamic regions of SCZ. In addition, multiple studies have detected an effect of a decreased number/density of oligodendrocytes in several brain regions in SCZ patients by immunohistochemistry and Nissl staining, and it is very likely that this effect is due to the loss of mature oligodendrocytes rather than oligodendrocyte precursor cells. A quantitative study detected a significant loss of Olig2-immunoreactive cells in the PFC of patients with SCZ but not NG2-immunoreactive cells [64]. This suggests that in future studies, the use of stage-specific markers to describe the specific stages of differentiation and maturation of oligodendrocytes in patients with SCZ may be more helpful to understand the pathology of oligodendrocytes related to SCZ.

However, there were still several limitations to this study that merit emphasis. Firstly, considering the relatively small number of BD studies that met the inclusion criteria for this review, we cannot draw a definite conclusion, which limited our ability to explain similar and different trends at the cellular level across diseases. Second, anxiety disorders are highly comorbid with MDD. However, the reviewed studies did not report anything with regard to comorbidities with anxiety; therefore, the presence of an anxiety disorder in MDD patients could be an important confounder for the observed between-study heterogeneities [134]. A third limitation is the high level of heterogeneity across studies derived from the research design, measurement methods, and sample selection. For example, some studies have found layer-specific effects in areas of the brain that evaluate glial cells [20, 35, 37, 73]. However, not all studies have measured the cortical layer of glial cells. Although a few studies have considered the impact of suicide on their measurements, many studies have not included it in their statistical analysis, which makes this another limitation. In addition, postmortem histopathological studies often include psychiatric patients with a history of use of various antidepressants and/or antipsychotic medications, so medication therapy itself may also lead to changes in glial morphology and number. Finally, since we did not consider non-English articles and unpublished data, there is a possibility of publication bias. Thus, our findings should be interpreted with caution.

In conclusion, although most studies evaluating total glial cells in postmortem brain samples from patients with psychiatric disease reported no difference, multiple other studies evaluating glial cell subtypes found evidence of glial cell abnormalities (see Supplementary Table S1). For example, although 19 studies did not find any effect of psychiatric disease on astrocyte markers, 22 studies presented both increased and decreased astrocyte markers. In particular, 11 studies found that astrocyte number/density and GFAP expression were decreased in various brain regions in MDD. Similarly, 15 studies reported a decrease in the density and related gene expression of oligodendrocytes in multiple brain regions of SCZ. These findings suggest that glial subtypes seem to have specific patterns of change in each disease, which is of great significance for us to understand the pathophysiology of glial cells in major psychiatric diseases and to provide new directions for disease treatment.