Animals
All animal procedures were approved by the Institutional Animal Care and Use Committee (covered by Animal welfare assurance No A3011-0), Renaissance School of Medicine at Stony Brook University, and conducted in accordance with the guidelines of the National Institutes of Health “Guide for the Care and Use of Laboratory Animals”. Experiments were performed using adult (2–3 months old) male mice [C57BL/6J (wt), CSPG4-EGFP (Jackson Labs, 022735 model FVB.Cg-Tg(Cspg4-EGFP*) HDbe/J), CX3CR1-GFP (Jackson Labs, 005582 model B6.129P-Cx3cr1tm1Litt/J), PDGFRα-CreERT2 (Tg(Pdgfra-cre/ERT2)1Wdr; from58), Rosa26-ΕYFP (Jackson Labs, 006148 model B6.129X1-Gt(ROSA)26Sortm1(EYFP) Cos/J)] were used wherever mentioned. All the mouse lines were backcrossed to a C57BL/6J background, bred in-house, and genotyped by PCR. CD-1 retired-breeder male mice (Charles River Laboratories, CD-1 IGS mice, strain code:022) were used as aggressors. For induction of recombination in the PDGFRα-CreERT2 :: Rosa26-EYFP mice, tamoxifen (Sigma CAS # 10540-29‐1) dissolved in ethanol : sunflower seed oil was injected intraperitoneally (i.p.) starting at P30, at a daily dose of 75 µg/g body weight (from a 10 mg/ml stock), for 5 consecutive days, as in25. All animals were housed in 12-h light/dark cycle. Food and water were provided ad libitum by the experimenters.
Repeated social defeat stress paradigm (RSDS)
Adult male mice were subjected to repeated bouts of physical aggression (‘‘defeats’’) from aggressive CD-1 mice for 10 consecutive days, as previously described8,24. The CD-1 aggressors were screened during a 3-day screening period for adequate aggressive behavior and then housed on one side of the divided mouse cage (known as the home cage), at least overnight prior to the start of defeat sessions. All defeat experiments were performed within that compartment, while the intruders were rotated across defeat days, so that the experimental animals would not habituate to a single aggressor. On day 0 (D0) of RSDS the ‘socially defeated-to be’ (SD) mice were placed into the aggressor’s space for 10 min (physical stress), and after the end of the encounter they were returned to their side overnight. The mice could see and smell the aggressor through the clear perforated divider (sensory stress). The naïve mice were exposed to wt (C57BL/6J) mice, instead, for 1 minute. About 80% of the mice that were subjected to social defeat stress displayed depressive-like behavior (SD-Sus for Susceptible), whereas the rest 20% were the nonresponding mice (SD-Res for Resilient) did not, thus modeling the heterogeneity in individual responses to stress in humans59. At the end of RSDS paradigms all animals were singly housed. Socially defeated mice were categorized based on the SI output, and further behavioral analyses for each group followed this categorization. Both SD-Sus and SD-Res exhibited anxiety-like behavior, however the SD-Sus additionally demonstrated social avoidance, reduction of reward under stress, anhedonia, despair-like behavior, and elevated plasma corticosterone levels (~ 130 ng/ml)59. In addition to this paradigm, a shorter adaptation (3 days social defeat) was also used in this study.
Behavioral analyses
For endpoints D15 and D25, mice were behaviorally tested for 4 days (1 experiment/day), while for endpoint D6, mice were tested for 3 days (1 experiment/day) prior to euthanasia. Behavioral testing and recordings were performed during the light phase. Both naïve and SD mouse groups were tested for behavioral alterations using the i) social interaction test (SI; measures social avoidance), ii) elevated plus-maze test (EPM; measures stress/anxiety), iii) forced swimming test (FST; measures despair-like behavior), iv) sucrose preference test (SPT; assesses anhedonia-lack of feeling pleasure), and v) novelty suppressed feeding test (NSF; assesses reward under stress). All experiments (except for SPT) were performed in a light-controlled and sound-isolated behavioral analysis room. The mice were acclimatized to the experimental room for 1 hour before the start of each experiment. The SI and EPM were performed under red light conditions; the remaining behavioral tests were performed with lights on. The software Noldus Ethovision XT16 was used for the automated tracking and scoring during the behavioral tests. White noise generator (70–75 dB) was used to mask intermittent disturbing sounds (from surrounding areas) that would potentially startle the animals.
Social Interaction (SI) One day after the final social defeat stress interaction, a SI test was performed to determine whether the animals display social avoidance, as previously established24. Mice were placed into the social interaction open-field arena [(42 cm (w) × 42 cm (d) × 42 cm (h)], and the time spent in the interaction zone (< 8 cm from the wire-mesh enclosure) was monitored for the two 2.5-min phases (without or with a novel CD-1 aggressor present in the enclosure), separated by a duration of 30 s. The socially defeated mice with SI ratio [(Time in interaction zone with aggressor) / (Time in interaction zone without aggressor)] > 1, were classified as SD-Res and mice with a ratio below 1 (social avoidance), were classified as SD-Sus8. The naïve mice were categorized as Control (C).
Elevated plus-maze test (EPM) The apparatus used for this test was cross-shaped with two open arms (30 cm x 5 cm) and two closed arms (30 cm x 5 cm x15 cm) that extended from a central platform (5 cm x 5 cm), as previously8,60. The entire maze was elevated 40 cm above the floor. The enclosed arms offered safety when the mouse was stressed; on the other hand, the open arms offered the motive of exploration. Each mouse was placed in the central square of the apparatus, facing an enclosed arm. The mice were let to roam freely for 10 min totally; an arm entry was defined when all four paws entered an arm. The total mobility, time spent in open arms and time closed arms were recorded, as an index of stress/anxiety60.
Forced swimming test (FSM) Each mouse was placed in a transparent tank [inescapable Plexiglas cylindrical tank (height: 30 cm, diameter: 22.5 cm)] that was filled with water (to a depth of 15 cm and maintained at 25°C) and their escape related mobility behavior was measured. Once the mouse was in the water, it was left to swim for 6 min in total, assessing the behavior only during the last 4 min, as previously performed61–63. The mice that acquired an immobile posture, characterized by motionless floating in the water, were termed immobile (immobility time), making only the necessary movements to keep the head above the water. Before returning the animals to their home cages, they were dried gently using paper towels to prevent hypothermia.
Sucrose preference test (SPT) The SPT was performed as previously with slight modifications8,64. The experimental groups were habituated for 72 h to 2% sucrose, in which the bottles were alternated to avoid bias for a specific cage side. The day before the testing the mice underwent an 18 h water deprivation period. The day of the testing bottles filled with sucrose 2% or water were weighed, and consumption was determined for a 24 h period. Bottles were weighed again at the end of the 24 h period. Sucrose preference was expressed as (Δweightsucrose) / Δweightsucrose + Δweightwater) x100. Sucrose preference score less than 70% was displaying anhedonia.
Novelty suppressed feeding test (NSF) The NSF test measures the time that mice take to approach and eat food in a novel environment following an extended period of food deprivation. Mice were food-deprived for 24 h and moved in freshly prepared home-cages (to avoid any pellets remaining on the bedding). The next day, each mouse was placed at the corner of an open-field arena [(42 cm (w) × 42 cm (d) × 42 cm (h)] and a pellet of chow was already positioned at the center of it. The time until the first bite of the chow pellet was recorded, with maximum trial time the 10 min. After the trial, the mice were returned to their home cage which contained a pre-weighed food pellet, and food consumption was measured for a period of 5 min, as described65,66.
Plasma corticosterone levels
Blood samples were collected from the animals before the euthanasia, unless stated otherwise. The subjects were anesthetized by isoflurane, and they were placed in a restrainer (for easier extraction). The tail was warmed with a red light and snipped by a scalpel blade. Blood samples were collected (~ 100 µl) into lithium-heparin collection tubes (BD Vacutainer® tubes, 265729). Plasma was separated from whole blood by centrifugation and was stored at -80°C, until use. For measurement of plasma corticosterone levels, the samples were diluted 1:40 in buffer and measured using the Corticosterone EIA kit (Enzo Life Sciences, ADI-900-097), according to the manufacturer’s instructions.
BrdU and EdU labeling
For the labeling of proliferating cells, 5-Bromo-2′-deoxyuridine (BrdU;Sigma-B5022) was dissolved in drinking water (1 mg/ml), and all mouse groups were given access to the water ad libitum throughout the 10 days of the RSDS paradigm8. For dual-pulse labeling of proliferating populations, 5-ethyl-2'-deoxyuridine (EdU; ThermoFisher Scientific -E10187) was administered i.p. at D17, D19 and D21 (x3 at D17, D19, D21) at a dose of 100 µg/g body weight (8 mg/ml EdU in 0.9% NaCl), as in67.
Reactive oxygen species (ROS) measurement
For the in-situ detection of oxidative stress in the form of ROS production, dihydroethidium (DHE; #D1148, ΤhermoFisher Scientific) was dissolved at 2 mg/ml in 100% DMSO, and then diluted 1:1 with sterile saline to 1 mg/ml. Three hours before euthanasia, the subjects were i.p. administered a DHE dose of 10 µg/g body weight, as previously described29. DHE is a redox-sensitive probe, which is oxidized by superoxide radicals to ethidium68, generating the highly superoxide-specific, red fluorescent product 2-hydroxyethidium (2-OH-E), which intercalates within the DNA. The excitation (Ex) and emission (Em) wavelengths used for the 2-OH-E were 405 nm and 580–620 nm respectively, to preferentially detect emissions from 2-OH-E and avoid nonspecific detection of ethidium (nonspecific product of DHE oxidation), as previously30,69.
Immunoblot analysis
Mouse subjects were euthanized with isoflurane overdose, and freshly extracted brains were micro-dissected using a McIlwain tissue chopper to obtain ~ 500-µm thick brain coronal sections. Medial PFC was dissected out (with the help of a dissecting microscope) from + 1.2 mm to + 2.5 mm anterior to bregma slices. Protein lysate was prepared using RIPA buffer (120–150 µl) with protease and phosphatase inhibitors (200 mM PMSF, 100 mM sodium orthovanadate, and protease inhibitor cocktail; Santa Cruz, sc-24948A). Lysates were shaken for 15 min at 4°C, cleared by centrifugation at 15,000 g for 10 min at 4°C, and protein concentration was determined using the Pierce BCA Protein Assay Kit (ThermoFisher Scientific; 23225). Samples (15–20 µg) were boiled for 5 min with 5X SDS-PAGE sample loading buffer (Thermo Fisher Scientific; 39000), separated by SDS-PAGE, and transferred to PVDF membranes (ThermoFisher Scientific; 88518) at 30 V for 16–18 h at 4°C. Membranes were blocked with 5% w/v nonfat dry milk (Cell Signal; 9999S) and incubated with primary antibodies (Supplementary Table 2a) for 16–18 h at 4°C. Washes with TBST (Cell Signal; 9997S) were followed by incubation with HRP-coupled secondary antibodies (Supplementary Table 2b). Signal was visualized by Sapphire Biomolecular Imager (Azure Biosystems) using a chemiluminescent substrate mixture (Supersignal West Pico Plus; ThermoFisher Scientific, 34580 or Immobilon Western; Millipore, WBKLS0500). Optical densities were collected with Fiji ImageJ software70. Where indicated, protein levels were expressed as fold change (versus Control) of the arbitrary units (A.U.) following normalization to the corresponding loading controls (β-Actin, GAPDH). Samples were normalized to the total protein for each sample, as determined by the BCA assay. For detailed antibody list used see Supplementary Table 2a.
Immunohistochemistry
Deeply anesthetized mice were transcardially perfused with cold 1X PBS (15 ml) followed by cold PFA fixative solution (4% paraformaldehyde in 1X PBS; 20ml) and brains were dissected and post-fixed for an additional 24 hours. Then brains were cryoprotected in 30% sucrose for 24 hours and sectioned with a sliding microtome. Free-floating brain sections (30–40 µm thick) containing the mPFC, VO, LO and fmi were blocked with 10% goat serum (ThermoFisher Scientific; 16210072) in 0.3% TritonX-100 in 1X PBS for 1–2 hour at room temperature. Tissue sections were incubated with primary antibodies overnight at 4°C at the indicated concentrations (Supplementary Table 2a). The following day, sections were washed in PBS-Triton and incubated with the appropriate cross-absorbed secondary antibodies (Supplementary Table 2c). After 4 additional washes nuclei were stained with DAPI (Sigma; D9542) and sections were mounted using MOWIOL mounting media. For BrdU IHC, sections were pretreated in 2M HCl for 40 minutes at 37°C, followed by 2-3x 5-minute washes in 0.1 M Boric acid (pH 8.5). For BrdU and EdU double staining, tissue was processed for BrdU and followed by EdU labeling with Click-iT Plus EdU Cell Proliferation Kit-Alexa Fluor 555 (ThermoFisher Scientific, C10638), according to the manufacturer’s instructions. For detailed antibody list used see Supplementary Table 2a.
Microscopy and histological quantification
The confocal laser-scanning microscopes Leica TCS SP8X and Leica TCS-SP5 were used for imaging of FITC, EGFP, EYFP, Alexa-488, Alexa-555, Alexa-594, Alexa-647, 2-OH-E and CY3 fluorophores. Optical sections (z = 1.0µm; total stack of 10–16µm for 1 cell layer) of confocal epifluorescence images were sequentially acquired using 10x, 20x, 40x (N.A. 1.30), 63x (N.A. 1.20) and 100x (N.A. 1.40) objectives, with LAS AF software. Fiji ImageJ software70 was used for image reconstruction, cell counting, integrated density analysis and colocalization analysis. For mPFC analysis, 6 fields (located between + 1.2mm to + 2.5mm anterior to bregma), containing the cingulate cortex (Cg1), prelimbic (PL), infralimbic (IL) and medial orbital cortex (MO) were taken from each animal, for quantitative analysis. For VO/LO analysis, 4 fields (located between + 2.0mm to + 2.5mm anterior to bregma), containing the ventral orbital cortex (VO), and lateral orbital cortex (LO), were taken from each animal for quantitative analysis. For fmi analysis, 3 fields (located between + 1.2mm to + 1.7mm anterior to bregma), containing the forceps minor (fmi), were taken from each animal for quantitative analysis. Typically, 3–4 brain slices were analyzed per animal, and at least 3 different animals for each experimental condition were evaluated. Cell counts were presented as the number of marker+ cells/section (section corresponding to 0.2mm2). Each section quantification depicts the sum of the marker+ cells from all the analyzed fields. For the integrated density analysis, the IsoData thresholding was applied to each image, and the integrated density [is defined as the product of pixel intensity (255 = the maximum pixel intensity for an 8-bit image) x area] was measured. For the colocalization analysis, the IsoData thresholding was applied to each image, images were merged, compartmentalized, and the integrated density was measured for the colocalized signal. Each section with integrated density quantification is depicted by the mean of the integrated density from all the analyzed fields. Additionally, for the analysis of weak BrdU+ cells the low threshold level was set at 120. Any cells not observable at that threshold were termed as weak-labeled BrdU+ cells, depicting the diluted labeling cells after several proliferation cycles71. All experimental group comparisons were conducted on sections stained and imaged with identical exposure and acquisition settings. All analyses were performed on raw images, prior to any image processing by Adobe Illustrator CS6 for representation purposes. The cell counting was performed in a blinded manner by 4 experimenters collectively, and tissue sections were matched across samples.
Glial morphological analysis
To quantitatively examine OPC morphology, unprocessed 30µm z-stack confocal images (100x objective; 20 individual images, step size: 1.5µm) of cells for CSPG-EGFP activity in the mPFC area were imported into Neurolucida software, as previously described3,72. Cell bodies and processes were manually traced (6 cells/mouse), and a 3D rendering was created. The 3D reconstructions were imported into Neurolucida Explorer for branched structure analysis and Sholl analysis (10µm starting radius, 80µm ending radius, and 10µm step size). Number of process intersections, cell surface area, and cell complexity were measured. OPC complexity = [Sum of the terminal orders + Number of terminals] * [Total dendritic length / Number of primary dendrites]. Terminal orders: Number of "sister" branches encountered as proceeding from the terminal to the cell body (calculated for each terminal); Terminal: Refers to process endings.
Bioinformatic Analysis of Publicly Available Dataset
snRNAseq Raw counts were downloaded from GEO (GSE144136) from Nagy et al.,9, and expression objects were created using Seurat73,74. Quality control (QC) was performed, and cells with abnormally high or low feature counts and percent mitochondrial gene expression ≤ 2% were removed. Using the WhichCells() function (Seurat) OLN cells were subsetted based on their expression of oligo-specific markers (PCDH15, DSCAM, VCAN, SOX6, PDGFRα, CSPG4, CLDN11, CNP, PLP1, PCDH9, QKI, MBP, MOG, MAG).
Oligodendroglial-Lineage Classification
To visualize discrepancies between conditions in oligodendroglial cells, the new expression object was subset by condition (MDD and Control) and the following hyperparameters were used to run UMAP on the two new objects: n.neighbors = 20, min.dist = 0.2. Four similar clusters and one MDD-specific cluster resulted from the analysis. The distinct clusters were labeled based on their expression of OPC and Committed OPC (PCDH15, DSCAM, VCAN, SOX6, PDGFRα, CSPG4, OLIG1, OLIG2), and Immature Pre-myelinating and Mature Myelinating Oligodendrocyte (CNTNAP2, CLDN11, CNP, PLP1, PCDH9, QKI, MBP, MOG, MAG) markers. The one unique MDD cluster was labeled Immune Oligodendrocytes (Im-OL), based on their expression of immune markers (P2RY12, CD74, C3, ITGAX, ITPR2, ARHGAP24, ADAM28, LPAR6).
Grouped Violin and Dotplots To visualize the gene expression of early and late oligodendroglia as a whole, we grouped the OPC and Committed OPC clusters (termed OPCs), and the Immature Pre-Myelinating OL and Mature Myelinating OL clusters (termed OLs). Using the WhichCells() function, we defined which cells belonged to the four initial clusters and then re-labeled them using the SetIdent function. The Violin plots and Dotplots were then created using the standard Seurat functions. Dotplots displaying a hierarchical clustering were also created using the standard Seurat function. To perform this hierarchical clustering, after grouping the cells (as mentioned above) the object underwent a pseudobulking technique in which we averaged the counts for each of the labeled cell types. This new matrix was then cleared of missing values and fed into the hclust() function which performed the hierarchical clustering and visualized with as dendrogram() cluster.
Integration with published dataset To verify the cluster labeling, and the credibility of our immune oligodendrocytes cluster, our oligodendrocyte object was integrated with those published by Jäkel et al, 201936 using Seurat’s integration and label transfer method. Raw counts for the Jäkel data were downloaded from GEO (GSE118257)36, and expression objects were created using Seurat. In preparation for the integration, we subset the oligodendroglia based on the Jäkel labels. The two objects were then integrated and a pseudobulking method was applied, in which we averaged the counts across each of our and Jäkel’s labeled cell types. We then computed the correlation coefficient between samples (clusters) using the Cor() function and plotted a heatmap using the pheatmap() function75.
Comparison with Microglia Given that the Im-OL cluster displays several immune components, we sought to compare it to the innate immune cells, microglia. To do so, we subsetted OL lineage cells and microglia from the originally read in object and again pseudobulked the cell clusters. The correlation coefficient between samples (clusters) was then computed using the Cor() function and a heatmap plotted using the pheatmap() function.
Pseudotime Analysis The two separated Seurat objects (MDD and Control) were used for the Pseudotime analysis. Separately, the UMAP coordinates and feature loadings for each object were passed to Monocle3 and pseudotime graph was learned using default parameters37,76.
Gene Scoring A Gene Scoring technique was used to calculate a Gene Signature “Score” for each of the OLN clusters. First, a signature containing the genes of interest was created. The raw counts were then extracted from the Seurat object and log normalization was performed. We then averaged the log normalized counts for each of the genes in the signature. For the HLA gene complex (HLA-A, HLA-DPA1, HLA-DRB1, HLA-DQB1, HLA-C, HLA-DPB1, HLA-DOB, HLA-DRB5, HLA-DQB2, HLA-B, CD74), signature “score” was added back into the Seurat object metadata and plotted using the VlnPlot() function. For the myelin gene scoring (CNP, PLP1, PCDH9, QKI, MBP, MOG, MAG), column statistics were performed.
Gene Set Enrichment Analysis
To conduct the gene set enrichment analysis (GSEA) for the Im-OL cluster, the Seurat FindMarkers() function was utilized to attain the genes that were differentially expressed (DE) in the Im-OL cluster versus the other OL clusters. The DE genes were ranked using the following function: sign(average- log2Fold-Change) x -log10(p-value). The ranked gene list was then fed into the gseGO() function of the clusterProfiler package77, using “Biological Processes” as the ontology. The plot was made using the standard clusterProfiler functions.
Colocalization plots To visualize the colocalization of OLN and Im-OL characteristic genes within the ImOL cluster, the cluster was first subsetted from the rest of the object. After re-running standard clustering parameters, the FeaturePlot function from Seurat was used with the following hyperparameters: Blend = TRUE.
Statistical analysis
No statistical methods were used to predetermine sample size a priori, but the sample sizes used were similar to those reported in8. Furthermore, representative data from each experiment were examined by Shapiro-Wilk’s test (p > 0.05)78 and a visual inspection of their histograms to confirm normal data distribution79,80. The majority of the experimental analyses was performed by parametric ordinary one-way analysis of variance (ANOVA), followed by Tukey’s multiple comparison post-hoc tests. For the glial morphological analyses two-way ANOVA was performed, followed by Tukey’s multiple comparison post-hoc tests. For the myelin DEG scoring analyses, one-way Kruskal-Wallis analysis was performed, followed by Dunn’s multiple comparisons test. The GraphPad Prism 8 and Excel (Microsoft) were used for all statistical analyses, except for the Bioinformatic Analyses (R-Studio). The data were reported as mean ± S.D. with symbols indicating the following P value ranges: *P ≤ 0.05, ** P ≤ 0.01, *** P ≤ 0.001 and **** P ≤ 0.0001).