Microglia Transcriptional Profiling in Major Depressive Disorder Shows Inhibition of Cortical Gray Matter Microglia

,

https://doi.org/10.1016/j.biopsych.2023.04.020 Major depressive disorder (MDD) is a significant contributor to the global burden of disease and a leading cause of disability worldwide (1).Insight into disease pathophysiology and novel therapeutics are urgently needed because treatment resistance is common and occurs in up to 30% of patients with MDD (2).Among others, inflammation is a prominent hypothesis in the neurobiology of depression, based on altered levels of pro-and anti-inflammatory cytokines and chemokines in blood and postmortem brain tissue in MDD (3)(4)(5).Furthermore, a higher prevalence of MDD has been associated with chronic inflammatory diseases, such as rheumatoid arthritis, inflammatory bowel diseases, and multiple sclerosis (6,7).Peripheral and central inflammation alters microglia activity (8)(9)(10), which may alter neuronal functioning through mechanisms such as synapse stripping (11).Indeed, recent studies have shown an altered immune status of microglia in MDD (12,13).
Microglia account for 10% to 15% of the cells in the brain and influence major aspects of neuronal functioning.By surveilling the brain for debris, excessive proteins, dysfunctional synapses, and aberrant neurons, they regulate synaptic transmission and neural plasticity in health and disease (14).In psychiatric disorders, microglial changes have been demonstrated in schizophrenia, autism spectrum disorder, MDD, and bipolar disorder (13,15,16).However, it remains uncertain whether-and in which state-microglia have beneficial or detrimental effects in different neuropathological conditions.Studies of MDD have yielded contradicting results ranging from reactive microglia to microglia in a homeostatic state.Torres-Platas et al. (17) showed vascular infiltration of macrophages and reactive/primed microglia in the white matter (WM) of the dorsal anterior cingulate cortex of brains of donors with MDD who had committed suicide.In contrast, a positron emission tomography study with a radio-labeled tracer for translocator protein (TSPO)-a marker for microglial activation-showed no increase of reactive microglia and even lower TSPO levels in mild to moderate MDD (18).Another TSPO study showed increased binding in the dorsal anterior cingulate cortex exclusively in patients with moderate to severe MDD with suicidal thoughts (19).Finally, two recent postmortem studies showed a noninflammatory signature of microglia, with upregulated homeostatic molecules (TMEM119, CX3CR1, and CD195) and downregulated immune activation markers (CD163, CD14, CD68, and HLA-DR) (12,13).
These contradictory findings regarding microglia in MDD prompted us to investigate their MDD-associated properties.Here, we investigated the transcriptome of primary human microglia in the tissue of a clinically well-characterized MDD brain donor cohort and gene and protein expression of neuronal regulators of microglia in an independent cohort.

METHODS AND MATERIALS Human Postmortem Tissue
Human brain occipital cortex gray matter (GM) and corpus callosum WM was provided by the Netherlands Brain Bank (https://www.brainbank.nl).Informed consent to perform the autopsy and use tissue and clinical data for research purposes were obtained from all donors.All procedures were approved by the Medical Ethics Committee of Amsterdam University Medical Centers.For RNA sequencing, GM and WM tissue blocks were collected from the occipital cortex and corpus callosum of donors with MDD (n = 13) and age-matched nonneurological control donors (n = 10).Non-neurological control donors with cognitive problems, based on clinical data, were excluded from the analysis.Donor characteristics are summarized in Table 1 and displayed in detail in Table S1.
For immunohistochemistry and reverse transcriptasequantitative polymerase chain reaction (RT-qPCR), independently collected GM tissue blocks from the occipital cortex of donors with MDD and age-matched non-neurological control donors were explored.Detailed donor characteristics are presented in Table S2.

Microglia Isolation
Corpus callosum and occipital cortex tissue blocks of 5 to 15 g were dissected at autopsy and stored in Hibernate A medium (Invitrogen) at 4 C until further processing.Microglia isolations were performed as described before (20,21).Briefly, brain tissue was mechanically dissociated using a tissue homogenizer (VWR) followed by enzymatic digestion with trypsin for 45 minutes at 37 C in Hibernate A medium supplemented with DNAse I (Roche).Percoll (GE Healthcare) density centrifugation was performed, and glial cells were collected from the interlayer.Magnetic-activated cell sorting was performed for positive selection of microglia using anti-CD11b beads (Miltenyi Biotec).Viable cells were counted, and the cell pellet was stored in 1 mL cold TRIsure (Bioline) at 280 C for further analysis.To assess the purity of isolated cells, CD45, CD11b, and CD15 expression was analyzed by flow cytometry (Figure S1).Consistent with previous reports, WM and GM microglia expressed comparable levels of CD11b, while CD45 expression was higher in WM than in GM microglia (20,21).

RNA Isolation
RNA isolation was carried out by QIAzol Lysis Reagent (Qiagen) according to the manufacturer's protocol using phase separation by chloroform addition and centrifugation followed by overnight precipitation in isopropanol at 220 C. RNA concentration was measured using a NanoDrop (ND-1000; NanoDrop Technologies), and RNA integrity was assessed using a Bioanalyzer (2100; Agilent Technologies).

RNA Sequencing
RNA sequencing was performed by GenomeScan.Samples were prepared using the NEBNext Ultra Directional RNA Library Prep Kit from Illumina (NEB #E7420).Briefly, ribosomal RNA was depleted from total RNA using the rRNA depletion kit (NEB# E6310).After fragmentation of the ribosomal RNAreduced RNA, complementary DNA (cDNA) was synthesized, ligated with the sequencing adapters, and amplified by PCR.The quality and yield were measured with the Fragment Analyzer (Agilent Technologies).The size of the products was consistent with the expected distribution (a broad peak between 300 and 500 bp).Clustering and DNA sequencing using the Illumina NextSeq 500 (Illumina) was performed according to manufacturer's protocols.A concentration of 1.6 pM of DNA was used.NextSeq control software 2.0.2 (https://support.illumina.com/downloads.html)was used.Image analysis, base calling, and quality check were performed with the Illumina data analysis pipeline Real-Time Analysis version 2.4.11 and Bcl2fastq version 2.17.Reads were aligned using the human assembly GRCh37.75.The reads were mapped to the reference sequence using a short read aligner based on Burrows-Wheeler Transform (Tophat version 2.0.14) with default settings.

RNA Sequencing Data Analysis
Sequencing data were normalized by fragments per kilobase of exon per million mapped fragments, and differential expression was assessed using DESeq2 (https://bioconductor.org/packages/release/bioc/html/DESeq2.html).Read counts were loaded into the DESeq2 package version 1.14.1 within the R platform version 3.3.0 to find differentially expressed (DE) genes between predefined sample groups.The output of this analysis was used to feed all 3 separate gene regulatory pathway enrichment tools.
Pathway analyses were performed using the Qiagen Ingenuity Pathway Analysis tool (Qiagen) and ErmineJ software (https://erminej.msl.ubc.ca/), using all DE genes with an adjusted p value , .05 to ensure the qualitative network analysis without a fold change cutoff.Weighted gene correlation networks were identified by weighted gene coexpression network analysis (WGCNA) (version 1.64) library in R (version 3.5.1).Data were converted into a matrix (DESeq2, version 1.20) where transcripts with a sum lower than 10 counts (all GM samples) were considered lowly expressed and removed from the data frame.Variability among the samples was reduced by batch correction using sva (version 3.28.0)(22), where the number of the surrogate variables was predicted using the leek method (23).Genes with a mean count per million expression value in the lower-quartile range (25%) were removed to generate a signed network (beta value of 8) which was then used to perform hierarchical clustering on a topological overlap matrix.Only modules with 100 genes or more were merged with the mergeCloseModules (cutHeight = 0.25) function in the WGCNA package.The module eigengenes were recalculated to their corresponding modules and used to calculate the module-trait correlation between modules and age, sex, and cohort.The userListEnrichment function was applied, which provides cell type-specific gene expression profiles to compare the identified modules to previously observed networks.ClusterProfiler was used to annotate gene ontology terms for the identified WGCNA modules (24).

Immunohistochemistry
Formalin-fixed, paraffin-embedded 8-mm sections were cut, and antigen retrieval was accomplished in citrate buffer pH 6.0 (microwave, 10 minutes at 700 W).Endogenous peroxidase activity was blocked with 1% H 2 O 2 for 20 minutes and nonspecific binding of secondary antibodies with 10% normal horse serum for 30 minutes.Sections were incubated with primary antibody (EDHu-1, 1:200; Novus Biologicals) overnight at 4 C.The appropriate biotinylated (horse antimouse) secondary antibody was applied followed by conjugation with avidin-biotin horseradish peroxidase complex (Vector Elite ABC kit; Vector Laboratories).Visualization was established with DAB chromogen, and samples were counterstained with hematoxylin.
Images of stained tissue sections were taken using an Axioscope microscope Z1 (Zeiss) while using a Plan-Apochromat 203/0.8objective with a bright field camera (HV-F203SCL; Hitachi).Layer I to layer VI of the cortex and WM were manually outlined with Qupath 0.3.2software (https:// qupath.github.io/).The DAB background (color deconvolved) was measured for all sections.Subsequently, the DAB threshold for a positive signal was set at 4 times the average background.DAB particles with size between 10 and 100 mm 2 , which exceeded the DAB threshold, were counted as CD163 1 cells.
Reverse Transcriptase-Quantitative PCR cDNA synthesis was performed using the QuantiTect reverse transcription kit (Qiagen) according to manufacturer's instructions, with a minimal input of 200 ng total RNA.Quantitative PCR was performed using the 7300 Real Time PCR system (Applied Biosystems) using the equivalent cDNA amount of 1 to 2 ng total RNA used in cDNA synthesis.SYBRgreen mastermix (Applied Biosystems) and a 2 pmol/mL mix of forward and reverse primer sequences were used for 40 cycles of target gene amplification.For primer sequences, see Table S3.Expression of target genes was normalized to the average cycle threshold of GAPDH and EF1A.Cycle threshold values were assessed with Sequence Detection System software (http://appliedbiosystems.com/support/software/7900/ updates.cfm;Applied Biosystems).

Synaptosome Isolation
Synaptosome fractions were isolated from frozen cortex tissue according to a public protocol from the Kelsey Martin lab that was adapted from Carlin et al. (25).In short, human cryostat sections (50 mm) were collected and stored in a disposable petri dish on dry ice.Sections were homogenized using a Dounce homogenizer in a sucrose-rich buffer with protease inhibition.Homogenates were spun down, pelleted, washed, and spun down at 1400g for 10 minutes to remove nuclei.Upon pelleting mitochondria and synaptosomes at 13,800g for 10 minutes, the pellet was resuspended in sucrose-rich buffer and layered on a gradient of 1.2M, 1.0M, and 0.85M sucrose buffer.After ultracentrifuge spinning at 82,500 for 2 hours, the top layer between 1.0M and 1.2M was aspirated off and processed further for protein detection.

Statistical Analysis
Statistical analysis of RT-qPCR, Western blot, and immunohistochemistry experiments was performed using GraphPad Prism version 9 software (https://www.graphpad.com/updates; GraphPad Software Inc.).Results are shown as mean with standard error of the mean.Statistical analysis was performed using either parametric or nonparametric testing based on the outcome of the Shapiro-Wilk normality test.The applied test for each calculated value is described in the figure legends.p Values , .05 were considered to indicate a significant difference.

Microglial Samples Cluster Based on Region and Sex
Principal component analysis of microglial gene expression data of all samples revealed segregation of the samples in 4 groups, showing that the largest part of the variation was explained by regional differences between GM and WM microglia (approximately 62%) (Figure 1D).Interestingly, the variance in principal component 2 was defined by sex (10%), with a similar effect in both GM and WM microglia.Further examination of DE genes between males and females only revealed Y chromosome genes and the X chromosome genes XIST and TSIX, for which expression from the inactive X chromosome accounts for the differential expression.Hence, no further analysis of these DE genes was performed.Cluster analysis of control samples only showed strong clustering of GM and WM samples; however, no complete segregation was seen.No apparent clustering of MDD or control samples was found in either GM or WM.

GM Microglia Show Differential Gene Expression in Donors With MDD
When comparing control GM and WM microglia using DESeq2 analysis, 2373 DE genes were found, as expected based on previous findings (26), of which 1691 were higher expressed in WM microglia and 682 were higher expressed in GM microglia.
When comparing GM microglia isolated from MDD brains with control GM microglia, 92 DE genes passed multiple testing correction, of which 81 were lower expressed and 11 were higher expressed in MDD than in control microglia (Figure 1E, F).Notable downregulated genes encoded the scavenger receptor CD163, the proliferation marker Ki-67, the extracellular matrix protein osteopontin, the lipopolysaccharide coreceptor CD14, the high-affinity immunoglobulin gamma Fc receptor (CD64), and complement component C1q chains (see Table S4 for a complete list of all DE genes) (27,28).Regarding microglia isolated from WM, no DE genes passed multiple testing corrections between MDD and control donors.
To validate the top DE gene on protein level, CD163 immunohistochemistry staining was performed in occipital cortex GM of an independent, age-matched cohort of donors with MDD (n = 12) and control (n = 12) donors.CD163 was expressed by microglia and perivascular macrophages at lower levels in MDD (Figure 2A, B) (p = .028).Interestingly, CD163 staining intensity was the lowest in donors with a certain depressed state at the time of their death (Figure 2A, diamonds).

Effector Function Pathways Are Suppressed in GM MDD Microglia
Ingenuity Pathway Analysis was used to identify the genetic regulatory pathways associated with the DE genes in MDD compared with control GM microglia (Table 2, Figure 3).The largest part of implicated pathways showed lower activity in MDD GM microglia, with the top pathways being the complement system and Fc gamma receptor-mediated phagocytosis.Overlap in regulated genes between overrepresented pathways pointed to suppressed effector functions, specifically phagocytic activity of microglia, as indicated by nitric oxide and reactive oxygen species production by macrophages, pattern recognition of bacteria and viruses (indicating complement functioning), interleukin 12 signaling and production in macrophages, phagosome formation, and ceramide degradation (indicating cell membrane remodeling).
The gene score resampling approach gave nondirectional identified pathways that match overrepresented pathways listed in the Ingenuity Pathway Analysis.Gene ontology terms include IgG binding, macrophage activation, Fc gamma receptor signaling pathway, regulation of phagocytosis, plasma membrane invagination, phagocytosis, engulfment, and regulation of complement activation (Table S5).The comparison between WM microglia derived from either MDD or control donors that yielded no DE genes through DESeq2 analysis similarly only yielded one significant hit in the ErmineJ analysis (with false discovery rate , .05):extracellular matrix binding.
WGCNA identified clusters of coexpressed genes and allowed for multifactorial association analysis.When applied to the GM microglia samples with a minimum cluster size of 100 genes, WGCNA revealed 18 distinct modules (after merging related modules), which were further investigated in their association with 3 different group parameters: age, sex, and cohort (control or MDD) (Figure 4A).Module-trait relationship analysis showed 6 modules significantly related to cohort, 4 modules particularly related to age, and 2 modules especially related to sex.Of the cohort-related modules, 3 modules (M2, M4, and M9) were unique and not associated with age or sex, showing MDD-specific effects on overall gene expression patterns.M2, having the highest significant difference related to cohort, was of particular interest and included microglia function-related gene ontology terms, such as vesicle organization, cytosolic transport, and macroautophagy (Figure 4B).

Increased Suppression of Microglia in MDD GM
To explain a lower activation status of GM microglia in MDD, we examined the expression levels of CD47 and CD200, two well-known proteins involved in neuron-to-microglia signaling that silence microglial immune responses and prevent synaptic pruning (29)(30)(31).An independent cohort selection of MDD (n = 13) and age-matched control (n = 13) occipital cortex GM samples was used to investigate CD47 and CD200 expression by RT-qPCR (Figure 5A).While CD47 showed a trend toward increased expression in MDD (p = .068),CD200 was significantly higher expressed in MDD GM than in control GM (p = .0009)(Figure 5B).Because phagocytic pathways are implicated in microglia-mediated degradation of targeted synapses, we studied CD47 and CD200 expression in the synapse fraction of frozen human cortex.This synapse fraction is highly enriched for the excitatory postsynaptic marker PSD-95, detection of which is absent in the myelin fraction (Figure 5C).CD47 abundance was higher in MDD cortical synaptosomes than in those of controls (p = .0396),whereas CD200 was unchanged (Figure 5D, E).

DISCUSSION
Here, we showed that GM microglia, but not WM microglia, from donors with a clinical diagnosis of MDD, showed a distinct disease-associated microglia (DAM) transcriptomic profile that we have designated DepDAM.The majority of the DE genes (81 of 92) of GM microglia were downregulated, with numerous genes involved in immune responses (MK167, SPP1, C1QA/B/C) and phagocytic function (FCGR1A/C, FCGR3A, CD14, CD163).Consistent with individual genes, pathway analyses showed alterations in immune activation (complement system, inducible nitric oxide synthase signaling, interleukin 12 signaling) and phagocytic activity (regulation of phagocytosis, membrane invagination, phagocytosis, Fcg receptor signaling in phagocytosis, interleukin 12, toll-like receptor, TREM1 signaling).Compared with control donors, an immune-suppressed state of MDD GM microglia was also apparent from a lower CD45 membrane expression assessed immediately after isolation.Purified CD11b 1 cells expressed distinct microglial signature genes but lacked expression of genes defining macrophages, other glial lineages, or neurons (20).The homeostatic signature of GM microglia was unchanged in MDD, based on the unaltered expression of ADGRG1, CSF1R, P2RY12, and CX3CR1.Furthermore, there were no signs of immune The ratio presented is defined as the number of the differentially expressed genes over the total number of genes involved in each of the pathways.ARP-WASP, actin related protein-Wiskott-Aldrich syndrome protein; GM, gray matter; IL-12, interleukin 12; iNOS, inducible nitric oxide synthase; IPA, Ingenuity Pathway Analysis; MDD, major depressive disorder; RXR, retinoid X receptor; VDR, vitamin D receptor.The DepDAM phenotype reported here is consistent with the results of several recent postmortem studies of donors with MDD (12,13,(32)(33)(34) and may shed light on the underlying cause of failure of anti-inflammatory treatment trials in MDD (35,36).Snijders et al. (13) analyzed gene expression in microglia isolated from the medial frontal gyrus, superior temporal gyrus, thalamus, and subventricular zone in MDD using RT-qPCR and also identified a noninflammatory signature, which was based on reduced CD163 and CD14 expression but enhanced homeostatic gene (TMEM119 and CX3CR1) activity.In a single-cell mass cytometry study, Böttcher et al. (12) found a noninflammatory microglia phenotype in MDD characterized by reduced human leucocyte antigen and CD68 expression and an increased homeostatic phenotype.Other studies have also found lower expression of microglial inflammatory genes in MDD dorsolateral prefrontal cortex tissue (33) and decreased expression of gene sets involved in endocytosis and antigen processing in the hippocampus and orbitofrontal cortex in MDD (32).As far as we know, our study is the first to have examined the transcriptome of acutely isolated postmortem microglia of MDD cortex, emphasizing the unbiased detection of pathways specifically implicated in DepDAM.

Microglia Inhibition in
Taken together, these data are in contrast to earlier clinical findings using, for example, positron emission tomography imaging with radioligands for the TSPO receptor.This method has frequently been applied to indicate neuroinflammation in vivo.TSPO binding has been reported to be increased in several brain regions in MDD (19,37,38).However, the sensitivity and specificity of this method for detecting microglial activation in psychiatric disorders have been questioned (39,40).It should be emphasized that our autopsy cohort consists of MDD cases in a late phase, and it could be that inflammation and microglial activation do occur in earlier stages of the disease.
We only detected transcriptional differences in GM microglia and not in WM.This led us to the hypothesis that changes in microglia may be induced by neuronal interference.Indeed, here we showed increased expression of the neuronal "don't eat me" signaling molecules CD200 and CD47 in tissue (CD200) and synaptosome fractions (CD47), potentially contributing to the dampened activation status of GM microglia in MDD (41,42).While neuroimmunomodulatory roles of CD47 and CD200 have been established, their influence on depressive symptoms has not been studied before.Given that the complement system and phagocytic functions are downregulated in cortical GM microglia, this may have implications for complement-mediated pruning of synapses.CD47 protects synapses from excess microglia-mediated pruning through signaling via the SIRP1a receptor on microglia during development and knockout of either CD47 or its receptor results in reduced synapse numbers (29).In addition, binding of CD200 to its receptor on myeloid cells (CD200R) is known to control microglial activity (30), and recently, it has been shown that CD200 deficiency leads to synaptic deficits and cognitive dysfunction (43,44).Microglia play an essential role in stress-exposed mice with lipopolysaccharide or macrophage colony-stimulating factor reversed depressed-like behavior (48).One consistent finding in MDD pathology is the deficiency in functional synapses, which results in disruption of the neural circuits that underlie the regulation of mood (49)(50)(51).Chronic stress causes atrophy of neuronal processes and decreases in synaptic density, which are restored by newer rapid-acting antidepressants, such as esketamine (52,53).
Concerning limitations of the current study, almost all the MDD brain donors were treated with antidepressants, and we cannot rule out that these may have influenced the inflammatory status of microglia (54)(55)(56)(57).Based on the lack of transcriptomic changes in WM microglia in MDD cases, it seems unlikely that the changes that we have described in GM microglia are a direct anti-inflammatory effect of medication.However, a secondary effect of medication through neuronal changes is possible.Control and MDD brain donors were matched for age, postmortem delay, and pH; however, the age did differ in the RNA sequencing group (75.0 vs. 63.7,p = .067),although the difference was not significant.Furthermore, information on the clinical state of depression at the time of death was unavailable for some donors.This may be important because it has been suggested that synaptic and neuroinflammatory changes in MDD may be state and severity dependent (51).However, more than half of our cohort died in a depressive state by euthanasia for refractory depression or suicide (7 of 13) (Table S1).
In summary, we found indications for an immunesuppressed GM microglial phenotype in the occipital cortex in MDD, possibly caused by neuronal regulation.An apathic microglial status with reduced phagocytosis and complement activation may have important consequences for synapse metabolism and connectivity relating to MDD.The DepDAM phenotype further adds to the continuum of microglia states that have been identified in the human brain (34).

Figure 1 .
Figure 1.Differential gene expression in microglia isolated from MDD and control brain tissues determined by RNA sequencing.(A) Acute microglia isolation from WM (CC) and GM (OC) of 13 donors with MDD and 10 age-matched control donors by sorting CD11b 1 magnetic beads.Flow cytometric analysis was performed, and RNA was isolated for RNA sequencing.(B) CD45 expression on GM and WM microglia detected by flow cytometry directly after isolation.Data passed the Kolmogorov-Smirnov normality test, and differences between MDD and control microglia were assessed using unpaired t tests.(C) Normalized expression of genes associated with microglia, neurons, astrocytes, oligodendrocytes, or endothelial cells in control WM or GM microglia.(D) PC analysis of gene expression demonstrates apparent differences between microglia from GM and WM and minor differences between microglia derived from female or male donors.(E) Number of DE genes for MDD vs. control microglia and for both WM and GM.(F) Volcano plot of DE genes between microglia obtained from donors with MDD or control donors.Threshold indicates DE genes with fold change .1.5, and purple circles indicate the notable downregulated genes of particular interest.Differential expression was determined using an adjusted p value (false discovery rate) , .05.CC, corpus callosum; DE, differentially expressed; Endo, endothelial cells; FPKM, fragments per kilobase of exon per million mapped fragments; geoMFI, geometric mean fluorescent intensity; GM, gray matter; MDD, major depressive disorder; OC, occipital cortex; oligo, oligodendrocytes; PC, principal component; RBC, red blood cell; WM, white matter.

Figure 2 .
Figure 2. Immunohistochemical staining of CD163.An independent validation cohort of 12 donors with MDD and 12 age-matched control donors was selected, and formalin-fixed, paraffinembedded tissue was extracted from the occipital cortex gray matter.(A) Immunoreactivity of CD163, presented as relative expression to total surface.Diamond-shaped symbols indicate donors with a definite depressed state when they died.(B) Representative images of CD163 staining of a control donor (40x magnification, scale bar = 250 mm), showing CD163 1 microglia (arrowhead) and perivascular macrophages (arrow).Mann-Whitney U test: *p , .05.MDD, major depressive disorder.

Figure 5 .
Figure 5. Increased suppression of microglia in MDD cortex.(A) An independent validation cohort of 13 donors with MDD and 13 age-matched control donors was selected.Frozen tissue was extracted from the OC for whole tissue RT-qPCR and Western blot analysis of synaptosomes.(B) Expression of CD200 and CD47 in MDD whole tissue determined by RT-qPCR.(C) Western blot analysis of PSD-95 expression in whole tissue homogenate (input), isolated synaptic fractions (synapse), and myelin fractions.(D, E) Western blot analysis of CD47 and CD200 expression in isolated synaptic fractions.MDD, major depressive disorder; OC, occipital cortex; RT-qPCR, reverse transcriptase-quantitative polymerase chain reaction.

Table 1 .
Characteristics of Donors With MDD and Control Donors Used for RNA Sequencing of Isolated Microglia

Table 2 .
Reveals Differential Expressed Signaling Pathways Between GM Microglia From Donors With MDD and Control Donors Major Depressive Disorder 624Biological Psychiatry October 15, 2023; 94:619-629 www.sobp.org/journal Coexpression networks for gray matter microglia from donors with major depressive disorder or control donors were determined by weighted gene correlation network analysis.(A) Eighteen modules identified are provided with their module-trait relationships and the number of genes that belong to each module displayed between brackets.Numbers in the heat map indicate the p value for the module-trait correlation.Minimum cluster size was set at 100 genes.(B) Gene ontology annotation of the genes in modules M2, M4, and M9, corresponding to overrepresented pathways in Ingenuity Pathway Analysis and ErmineJ analysis, visualized in a dot plot.The size of the dot represents the gene count.