Abstract
Background Microglia have been implicated in the pathophysiology of major depressive disorder (MDD), but information on biological mechanisms is limited. Therefore, we investigated the gene expression profile of microglial cells in relation to neuronal regulators of microglia activity in well-characterized MDD and control autopsy brains.
Methods Pure, intact microglia were isolated at brain autopsy from occipital cortex grey matter (GM) and corpus callosum white matter (WM) of 13 MDD and 10 age-matched control donors for RNA sequencing. Top differentially expressed genes were validated using immunohistochemistry (IHC) staining. Since gene expression changes were only detected in GM microglia, neuronal regulators of microglia were investigated in cortical tissue and synaptosomes from the cortex by RT-qPCR and Western blot.
Results Transcriptome analysis revealed 92 genes differentially expressed in GM microglia of MDD, but none in WM compared to the control Of these, 81 genes were less abundantly expressed in GM MDD, including CD163, MKI67, SPP1, CD14, FCGR1A/C, and C1QA/B/C. Accordingly, pathways related to effector mechanisms, such as the complement system and phagocytosis were differentially regulated in GM microglia in MDD. IHC staining revealed significantly lower expression of CD163 protein in MDD. Whole tissue analysis showed an increase in CD200 (p<0.001) and CD47 (p=0.068) mRNA, and CD47 protein was significantly elevated (p<0.05) in synaptic fractions of MDD cases.
Conclusions Transcriptional profiling indicates an immune-suppressed microglial phenotype in MDD, possibly caused by neuronal regulation.
Introduction
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, as treatment resistance is common and occurs in up to 30% of MDD patients(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 post-mortem brain tissue in MDD(3–5). 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(8–10), which may alter neuronal functioning through mechanisms such as synapse stripping(11). Indeed, recent studies show 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 in MDD report contradicting results, ranging from reactive microglia to microglia in a homeostatic state. Torres-Platas et al. showed vascular infiltration of macrophages and reactive/primed microglia in the white matter of the dorsal anterior cingulate cortex (dACC) of brains of donors with MDD that committed suicide(17). In contrast, a positron emission tomography (PET) 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 dACC, exclusively in moderate to severe MDD patients with suicidal thoughts(19). Finally, two recent post-mortem studies showed a non-inflammatory 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 post-mortem tissue
Human brain occipital cortex grey matter (GM) and corpus callosum white matter (WM) was provided by the Netherlands Brain Bank (Amsterdam, The Netherlands; 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 NBB procedures were approved by the Medical Ethics Committee of the Amsterdam UMC (locationVUmc, Amsterdam, The Netherlands). For RNA sequencing, GM and WM tissue blocks from the occipital cortex and corpus callosum of MDD donors (n=13) and age-matched non-neurological control donors (n=10) were collected. 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 Supplementary Table 1.
For immunohistochemistry (IHC) and quantitative RT-PCR, a second and third cohort with GM tissue blocks from the occipital cortex of MDD donors and age-matched non-neurological control donors was explored. Detailed donor characteristics are displayed in Supplementary Table 2.
Microglia isolation
Corpus callosum and occipital cortex tissue blocks of 5-15 g were dissected at autopsy and stored in Hibernate A medium (Invitrogen, Carlsbad, CA, USA) 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, Radnor, PA, USA), followed by enzymatic digestion with trypsin for 45 min at 37°C in Hibernate A medium supplemented with DNAse I (Roche, Basel, Switzerland). Percoll (GE Healthcare, Little Chalfont, UK) 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, Bergisch Gladbach, Germany). Viable cells were counted and the cell pellet was stored in 1 ml cold TRIsure (Bioline, London, UK) at -80°C for further analysis. To assess the purity of isolated cells, CD45, CD11b, and CD15 expression were analyzed by flow cytometry (Supplementary Figure 1). In line 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, Hilden, Germany) according to the manufacturer’s protocol using phase separation by addition of chloroform and centrifugation, followed by overnight precipitation in isopropanol at -20°C. RNA concentration was measured using a Nanodrop (ND-1000; NanoDrop Technologies, Rockland, DE, USA), and RNA integrity was assessed using a Bioanalyzer (2100; Agilent Technologies, Palo Alto, CA, USA).
RNA sequencing
The NEBNext Ultra Directional RNA Library Prep Kit from Illumina (San Diego, CA, USA) was used to process the samples (GenomeScan, Leiden, The Netherlands). Sample preparation was performed according to the protocol of NEBNext Ultra Directional RNA Library Prep Kit from Illumina (NEB #E7420). Briefly, rRNA was depleted from total RNA using the rRNA depletion kit (NEB# E6310). After fragmentation of the rRNA-reduced RNA, cDNA synthesis was performed. This was used for ligation with the sequencing adapters and PCR amplification of the resulting product. The quality and yield after sample preparation was measured with the Fragment Analyzer (Agilent Technologies). The size of the resulting products was consistent with the expected size distribution (a broad peak between 300-500 bp). Clustering and DNA sequencing using the Illumina NextSeq 500 was performed according to manufacturer’s protocols. A concentration of 1.6 pM of DNA was used. NextSeq control software 2.0.2 was used. Image analysis, base calling, and a quality check were performed with the Illumina data analysis pipeline RTA v2.4.11 and Bcl2fastq v2.17. 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 v2.0.14) with default settings.
RNAseq data analysis
Sequencing data were normalized (FPKM), and differential expression was assessed using DESeq2 (https://bioconductor.org/packages/release/bioc/html/DESeq2.html). The read counts were loaded into the DESeq2 package v1.14.1 within the R platform v3.3.0 to find differentially expressed (DE) genes between pre-defined sample groups. The output of this analysis was used to feed all three separate gene regulatory pathway enrichment tools. Post-mortem factors (age, PMD, and pH) were plotted against the top 10 DE, which did not show any correlations.
Ingenuity pathway analyses were performed using the Qiagen IPA tool (Qiagen) and ErmineJ software (https://erminej.msl.ubc.ca/), using all DE genes with an adjusted p-value <0.05 to ensure the qualitative network analysis. Weighted gene correlation networks were identified by weighted gene co-expression network analysis (WGCNA) (version 1.64) library in R (version 3.5.1). Data was 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. The 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 (CPM) 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 then used to calculate the module–trait correlation between modules and sample traits (age, gender, and cohort). In the end, the userListEnrichment function was applied, which provides cell type-specific gene expression profiles to compare the identified modules to previously observed networks. In addition, Metascape was used to annotate gene ontology terms for each of the identified WGCNA modules(24).
Immunohistochemistry
8-µm formalin-fixed, paraffin-embedded (FFPE) sections were cut, and antigen retrieval was accomplished in citrate buffer pH 6.0 (microwave, 10 min at 700 W). Endogenous peroxidase activity was blocked with 1% H2O2 for 20 min and aspecific binding of secondary antibodies with 10% normal horse serum for 30 min. Sections were incubated with primary antibody (EDHu-1, 1:200; Novus Biologicals, Centennial, CO, USA) overnight at 4°C. The appropriate biotinylated (horse anti-mouse) secondary antibody was applied, followed by conjugation with avidin–biotin horseradish peroxidase complex (Vector Elite ABC kit; Vector Laboratories, Newark, CA, USA). Visualization was established with 3,30-diaminobenzidine chromogen (DAB), and samples were counterstained with hematoxylin.
Images of stained tissue sections were taken using an Axioscope microscope Z1 (Zeiss, Oberkochen, Germany) while using a Plan-Apochromat 20x/0.8 objective with a bright field camera (HV-F203SCL; Hitachi, Tokyo, Japan). Layer I until VI of the cortex and WM were manually outlined with Qupath 0.3.2 software. The DAB background (color deconvolved) was measured for all sections. Subsequently, the DAB threshold for a positive signal was set at four times the average background. DAB particles with size between 10 and 100 mm2, exceeding the DAB threshold, were counted as CD163+ cells .
Quantitative RT-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 (qPCR) was performed using the 7300 Real Time PCR system (Applied Biosystems, Foster City, CA, USA) using the equivalent cDNA amount of 1–2 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 Supplementary Table 3. Expression of target genes was normalized to the average cycle threshold of GAPDH and EF1A. Cycle threshold values were assessed with SDS software (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 µm) 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 1,400 g for 10 min to remove nuclei. Upon pelleting mitochondria and synaptosomes at 13,800 g for 10’, the pellet was resuspended in sucrose-rich buffer (B) and layered on a gradient of 1.2 M, 1.0 M, 0.85 M sucrose buffer. After ultracentrifuge spinning at 82,500 for 2 h, the top layer between 1.0 M and 1.2 M was aspirated off and processed further for protein detection.
Western blotting
Isolated synaptosomes or brain tissue cryosections were lysed in cold RIPA buffer containing 50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1% Nonidet-P40, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate, and protease inhibitors (cOmplete EDTA-free; Sigma-Aldrich, Saint Louis, MI, USA). Protein content was determined using the Pierce BCA Protein Assay kit (Thermo Fisher Scientific, Waltham, MA, USA). For electrophoresis, proteins were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to nitrocellulose membranes (Millipore, Bedford, MA, USA) by electroblotting for 90 min (Transblot SD; BioRad, Hercules, CA, USA). Blotted membranes were incubated in blocking buffer (5% BSA in TBS with 0.05% Tween-20 (TBS-T)) for 1 h, followed by o/n incubation at 4°C with primary antibodies directed against CD47 (B6H12.2; Novus Biologicals), CD200 (AF2724; R&D Systems, Minneapolis, MN, USA), PSD95 (810401; Biolegend, San Diego, CA, USA), or β-actin (AC-74; Sigma-Aldrich) in blocking buffer. After several washes with TBS-T, membranes were incubated with DyLightℲ 649 (BioLegend) or IR800 (LI-COR, Lincoln, NE, USA)-conjugated secondary antibody for 1 h at RT (1:2,000 in blocking buffer). After washes in TBS-T, fluorescent signal was detected using the LI-COR Odyssey 9120 Imaging System (LI-COR).
Statistical analysis
Statistical analysis of RT-qPCR, Western blot and IHC experiments was performed using GraphPad Prism version 9 software (GraphPad Software Inc, La Jolla, CA, USA). Results are shown as mean with standard error of the mean. Statistical analysis was performed using either parametric or non-parametric 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 <0.05 were considered to indicate a significant difference.
Results
Microglia isolation yields pure populations of intact microglia
In this study, we isolated intact microglia using a rapid isolation protocol from human occipital cortex (GM) and corpus callosum (WM) (Figure 1A). Flow-cytometric phenotyping of intact microglia indicated no differences in CD11b expression, viability, and purity of microglia between MDD (n=13) and control (n=10) donors (Supplementary Figure 1). Membrane CD45 detection was lower in GM microglia isolated from MDD donors (p=0.0141), and for WM microglia, a trend towards decreased membrane CD45 in MDD was visible (p=0.0596) (Figure 1B). RNA sequencing of isolated microglia from control GM or WM revealed enrichment of established microglial marker genes (ADGRG1, CSF1R, P2RY12, C1QA, CX3CR1) and no notable detection of neuronal (MAP2, STMN2, SYN1, VGLUT1), astrocytic (AQP4, GFAP, ALDH1L1), oligodendrocytic (PLP, SOX10, MOG), or endothelial cell (CLDN5) marker genes (Figure 1C). As expected(20,26), hierarchical clustering of gene expression data showed that control GM and WM microglia are phenotypically distinct (Supplementary Figure 2).
Microglial samples cluster based on region and sex
Principal component analysis (PCA) of microglial gene expression data of all samples revealed segregation of the samples in four groups, showing that the largest part of the variation is explained by regional differences between GM and WM microglia (approximately 62%, Figure 1D). Interestingly, the variance in PC2 is 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 MDD donors
When comparing control GM and WM microglia using DESeq2 analysis, 2,373 DE genes were found, as expected based on previous findings(26), of which 1,691 were higher expressed in WM microglia- and 682 were higher expressed in GM microglia (data not shown). When comparing GM microglia isolated from MDD brains with control GM microglia, 92 DE genes passed multiple testing correction and fold change cutoff, of which 81 were lower and 11 were higher expressed in MDD compared to control microglia (Figure 1E/F). Notable down-regulated genes encoded the scavenger receptor CD163, the proliferation marker Ki-67, the extracellular matrix protein osteopontin, the lipopolysaccharide co-receptor CD14, the high-affinity immunoglobulin gamma Fc receptor (CD64), and complement component C1q chains (see Supplementary Table 4 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 IHC staining was performed in occipital cortex (GM) of an independent, age-matched cohort of 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=0.028). Interestingly, CD163 staining intensity was lowest in donors with a certain depressed state at decease (Figure 2A, diamonds).
Effector function pathways are suppressed in GM MDD microglia
Ingenuity pathway analysis (IPA) was used to identify the genetic regulatory pathways associated with the DE genes in MDD compared to control GM microglia (Table 2 and 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 (NO) and reactive oxygen species (ROS) production by macrophages’, ‘pattern-recognition of bacteria and viruses’ (indicating complement functioning), ‘IL-12 signaling and production in macrophages’, ‘phagosome formation’, and ‘ceramide degradation’ (indicating cell membrane remodeling).
The gene score resampling (GSR) approach gave non-directional identified pathways that match overrepresented pathways listed in the IPA analysis. GO-terms include ‘IgG binding’, ‘macrophage activation’, ‘Fc-gamma receptor signaling pathway’, ‘regulation of phagocytosis’, ‘plasma membrane invagination’, ‘phagocytosis, engulfment’, and ‘regulation of complement activation’ (Supplementary Table 5). The comparison between WM microglia derived from either MDD or control that yielded no DE genes through DESeq2 analysis similarly only yields one significant hit in the ErmineJ analysis (with false discovery rate <0.05): ‘extracellular matrix binding’.
WGCNA identified clusters of co-expressed genes and allowed for multi-factorial 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 three 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. Module 2, having the highest significant difference relating to cohort, was of particular interest and included microglia function-related GO terms, such as ‘vesicle organization’ and ‘cytosolic transport’ (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(29–31). An independent cohort selection of MDD (n=13) and age-matched control (n=13) occipital cortex (GM) samples were used to investigate CD47 and CD200 expression by RT-qPCR (Figure 5A). While CD47 showed a trend towards increased expression in MDD (p=0.068), CD200 was significantly higher expressed in MDD GM compared to control (p=0.0009) (Figure 5B). Since phagocytic pathways are implicated in microglial-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 post-synaptic marker PSD-95, detection of which is absent in the myelin fraction (Figure 5C). CD47 abundance was higher in MDD cortical synaptosomes than controls (p=0.0396), whereas CD200 was unchanged (Figure 5D/E).
Discussion
Here, we show that GM microglia, but not WM microglia, from brain donors with the clinical diagnosis MDD, show a distinct disease associated transcriptomic profile. The majority of the DE genes (81 out 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). In line with individual genes, pathway analyses showed alterations in immune activation (complement system, iNOS signaling, IL-12 signaling) and phagocytic activity (regulation of phagocytosis, membrane invagination, phagocytosis, FcγR signaling in phagocytosis, IL-12, TLR, TREM1 signaling). An immune-suppressed state of MDD GM microglia is also apparent from a lower CD45 membrane expression compared to control donors, assessed immediately after isolation.
Purified CD11b+ 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 activation, such as increased expression of pro-inflammatory cytokines (IL1, IL6, TNF).
The MDD associated microglia (MDD-DAM) with an immune-suppressed phenotype reported here are in line with the results of several recent post-mortem studies of MDD donors(12,13,32–34) and may shed light on the underlying cause of failure of anti-inflammatory treatment trials in MDD(35,36). Snijders et al. (2020) 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 non-inflammatory signature, based on reduced CD163 and CD14 expression but enhanced homeostatic gene (TMEM119 and CX3CR1) activity(13). Böttcher et al. (2020) found in a single cell mass cytometry study a non-inflammatory microglia phenotype in MDD characterized by reduced HLA and CD68 expression and an increased homeostatic phenotype(12). Other studies also reported 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 examine the transcriptome of acutely isolated post-mortem microglia of MDD cortex, emphasizing the unbiased detection of pathways specifically implicated inMDD-DAM.
Together, these data are in contrast to earlier clinical findings using for example positron emission tomography (PET) imaging with radioligands for the TSPO receptor. This method is frequently applied to indicate neuroinflammation in vivo. TSPO binding was 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 well be that inflammation and microglia activation does occur in earlier stages of the disease.
We only detected transcriptional differences in GM microglia in MDD versus controls and not in subcortical WM. This led us to the hypothesis that changes in microglia might be induced by neuronal interference in MDD. Indeed, we here show 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 neuroimmuno-modulatory roles of CD47 and CD200 are 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 in MDD, this may have implications for complement-mediated pruning of synapses. CD47 protects synapses from excess microglia-mediated pruning through signaling via the SIRP1α receptor on microglia during development, and knockout of either CD47 or its receptor results in reduced synapse numbers(29). Also, 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 synaptic circuit remodeling and phagocytosis of synaptic material, and defects may contribute to synaptic abnormalities seen in both neurodevelopmental and neurodegenerative diseases(45,46). Pre-clinical studies in depressed-like rodents showed microglial depletion with attenuated antidepressant and neurogenesis-enhancing treatment effects with imipramine or electroconvulsive stimulation(47). Also, microglial stimulation in chronic unpredictable stress-exposed mice with lipopolysaccharide or macrophage colony-stimulating factor reversed depressed-like behavior(48). One of the most consistent findings in MDD pathology is the deficiency in functional synapses, resulting in disruption of the neural circuits that underlie the regulation of mood(49–51). Chronic stress causes atrophy of neuronal processes and decreases in synaptic density, which is restored by administering newer rapid-acting antidepressants, such as esketamine(52,53).
Concerning limitations, almost all of the MDD brain donors were treated with anti-depressants, and we cannot rule out that these may have influenced the inflammatory status of microglia(54–57). Based on the lack of transcriptomic changes in WM microglia in MDD cases, it seems unlikely that the changes we describe in GM microglia are a direct anti-inflammatory effect of medication. However, a secondary effect of medication through neuronal changes is possible. Controls and MDD brain donors were matched for age, PMD, and pH; however, the age did differ in the RNAseq group (75.0 vs. 63.7, p=0.067), albeit not significant. Furthermore, information on the clinical state of depression at the time of death was unavailable for some donors. This may be important, as it is suggested that synaptic and neuroinflammatory changes in MDD may be state- and severity-dependent(49). However, more than half of our cohort deceased in a depressive state by euthanasia for refractory depression or suicide (7 out of 13; Supplementary Table 1).
In summary, we found indications for an immune-suppressed phenotype of MDD-DAM in the occipital cortex, 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 immune-suppressed MDD-DAM further adds to the continuum of microglia states identified in the human brain.
Data Availability
All data produced in the present study are available upon reasonable request to the authors
Funding
The NBB-PSY (https://www.brainbank.nl/nbb-psy/) is funded by grant 240-921200 from the Netherlands Organization for Scientific Research (NWO).
Disclosures
Dr. Scheepstra reported no biomedical financial interests or potential conflicts of interest. Dr. Mizee reported no biomedical financial interests or potential conflicts of interest. Dr. van Scheppingen reported no biomedical financial interests or potential conflicts of interest. A. Adelia reported no biomedical financial interests or potential conflicts of interest D. Wever reported no biomedical financial interests or potential conflicts of interest. Dr. Mason reported no biomedical financial interests or potential conflicts of interest. Dr. Dubbelaar reported no biomedical financial interests or potential conflicts of interest. Dr. Hsiao reported no biomedical financial interests or potential conflicts of interest. Prof. dr. Eggen reported no biomedical financial interests or potential conflicts of interest. Dr. Hamann reported no biomedical financial interests or potential conflicts of interest. Prof. dr. Huitinga reported no biomedical financial interests or potential conflicts of interest.
Data availability
After acceptance, the RNA-sequencing dataset is available online in the Gene Expression Omnibus (GEO) database.
Acknowledgments
We thank the Psychiatric Donor Program of the Netherlands Brain Bank (NBB-PSY) team for their excellent support and the donors for their valuable donation. A preprint of this manuscript was posted on medRxiv.