Our strengths include our whole organism approach to neutrophil dynamics and our triangulation and replication of our findings via multiple experimental approaches. We have generated robust evidence for chronic stress-induced meningeal neutrophilia, and our data-led investigation of candidate mechanisms revealed IFN-I-mediated migration of neutrophils from skull BM to the meninges as a potential treatment target in stress-associated depression.
All procedures were approved by the National Institute of Mental Health Animal Care and Use Committee (protocol #LCMR06) and conducted in accordance with National Institutes of Health guidelines. Strains used were either C57BL/6J male mice purchased from Jackson Labs (Bar Harbor, ME) or male LysM offspring from C57BL/6J mice bred in our facility with LysM mice, which strongly express GFP in neutrophils (obtained from Dr. Dorian McGavern, NINDS). Upon arrival into the facility, purchased mice were randomly pair-housed in divided cages and given one week of acclimation to the facility. Mice bred in-house were weaned at 3 weeks of age into same-sex cages of 2-4 littermates. Sex was assigned based on external anatomy. Animals were housed under a reverse light cycle (lights off 8:00 AM to 8:00 PM), with food and water ad libitum. Temperature and humidity were maintained between 18-23 °C and 40-60%, respectively, by NIMH animal care staff. At 8-10 weeks of age, mice were randomly assigned to an experimental group. At sacrifice, mice were deeply anesthetized with isoflurane, confirmed by tail and leg pinch. Anesthesia was maintained until diaphragm rupture, at which point the animals were transcardially perfused before decapitation and tissue collection.
CSD stress was performed as previously reported. Briefly, the 'intruder' test mouse was introduced into the home cage of an aggressive, CD-1 (Taconic; Rensselaer, New York) retired breeder. The two were separated by a perforated barrier and given 24 h to acclimate; the barrier allowed for olfactory, visual, and auditory communication, but not tactile contact. Each day for either 1, 2, 4, 8, or 14 consecutive days, depending on the experiment, the barrier was lifted, and antagonistic encounters were allowed to occur for 5 m. Interactions were monitored by a trained individual to ensure the test mouse exhibited submissive behavior and conversely that the CD1 exhibited dominant behavior. The C57/CD1 pairs were maintained throughout the study unless the CD1 failed to show dominant aggression toward the C57. When this occurred, the C57 was paired with a different CD1 mouse until submissive behavior in the C57 was evident, and the new pairing was maintained thereafter. Efforts to minimize physical damage were taken, i.e. CD1 mice were lightly anesthetized with isoflurane and incisors were trimmed prior to starting the social defeat paradigm, and on a weekly basis thereafter. Test mice were shaved and inspected for the presence of wounds at the end of the experiment; wounds were scored on a scale from 1-10 (see Table S12). Animals with wound scores of 10 were excluded. Unless otherwise noted, animals were sacrificed exactly 2 h after exposure to the final defeat stress on day 14. We did not go beyond 14 days of defeat out of concern for animal welfare.
All behavioral testing was done by both male and female experimenters during the dark phase of the light cycle, prior to a defeat encounter on that day. On the day of testing, mice were moved to a separate behavioral room with red lighting and acclimated for one hour prior to running the behavioral assays. Tests were usually run on separate days, but when multiple tests were run on a single day, the animals were given an hour to recover in between tests. Behavioral tests were performed in the following order (least to most aversive):
As previously described, mice were placed into a novel arena (50 ×50 x 50 cm) that contained a thick sheet of paper; one corner of this paper was 'spotted' with 50 μL of urine from multiple estrus females, and testing was performed with the lights off while the experimenter was outside the room. After testing, the sheets of paper were sprayed with ninhydrin and heated to indicate the presence of proteins, allowing for visualization of urine marking. Photos were taken of the sheets by an experimenter blind to group identity, then analyzed in ImageJ. Reduced preference marking for the female scent is indicative of social anhedonia.
Exploration of the novel open field arena, 50 ×50 x 50 cm in dimension, was performed under dim white lighting (~25 lux). Mice were placed in the middle of the arena; total distance moved, crosses to center, and time in center over a 10 min testing period whilst the experimenter was out of the room were later analyzed with automated tracking software (Clever Sys TopScan Suite) to eliminate potential human bias. Fewer crosses to center, reduced time in center, and reduced movement are all indicative of increased anxiety-like behavior.
The LD test uses an acrylic box (50 cm×25 cm with 30 cm walls) with aversive lighting (~40 lux). Approximately 1/3 of the box is enclosed and dark; an opening allows crossover between light and dark sections. Mice were placed in the light compartment and allowed to move freely for 10 min while the experimenter was out of the room. Time spent in the light compartment and number of crosses between the light and dark sides were scored from video recordings using TopScan. Low scores indicated anxiety-like behavior.
Two perforated acrylic cylinders, one containing an unfamiliar CD-1 mouse and the other empty, were placed in the OF arena with red lighting. The test mouse was placed in the middle of the arena and allowed to freely explore for 10 m with the experimenter outside of the room. TopScan was used on captured videos to track approaches to the social stimulus and time spent investigating social vs. non-social stimuli. Fewer social interactions indicated anhedonic behavior.
Unless otherwise indicated, mice were euthanized 2 h after their final exposure to the defeat stressor. Tissue collection occurred between ~8am and noon, with modifications as indicated below. Venous blood was collected into EDTA tubes via puncture of the submandibular vein and kept on ice until processing. Retro-orbital injections were administered while mice were under light isoflurane anesthesia. After, mice were deeply anesthetized with isoflurane prior to perfusion with 35 mL room temperature PBS. When indicated, hindlegs were collected for tibial BM extraction. The head was decapitated, and intact skull cleaned with a scalpel to remove muscle and connective tissue.
LysM mice received a retro-orbital intravascular injection of either DyLite 649- or DyLite 594-conjugated tomato lectin (TomL, Cat #DL-1178, Vector Labs), which was allowed to circulate for 5 min to label blood vessels before perfusion. In addition to PBS, mice were perfused with 10 mL of 4% cold paraformaldehyde. Skulls were post-fixed in 4% PFA for 24 h at 4 °C before transferring to 25% sucrose solution for cryoprotection, dehydration, and preparation for imaging.
WT mice were retro-orbitally injected with 4 μg CD45-FITC (Cat. #103108; Biolegend). Mice were then allowed to recover; after 25 min of circulation, mice were anesthetized lightly for venous blood collection. The decapitated skull was placed in cold HBSS + 0.1% BSA on ice until further processing.
We used the CUBIC tissue clearing method on whole skulls from LysM mice. Tissue was prepared as for histology except that samples were transferred to PBS instead of sucrose after 24 h fixation. Decalcification of bone prior to tissue clearing was achieved by incubation in decalcification solution (10% EDTA, 15% imidazole) for 5-7 days at 37 °C with shaking. The decalcification solution was refreshed once on day 3. Following tissue clearing, the whole skull was inverted and placed in a glass-bottom dish filled with fresh Reagent 2 and imaged using a Zeiss 780 confocal microscope fitted with 10x objective. 2-5 images were collected at random locations for each skull and analyzed using IMARIS 9.7. For each image, both the number of vascular channels (labeled with TomL) and the number of discreet LysM-GFP cells in a channel were counted. For each individual sample, the ratio of averaged LysM-GFP cells normalized to the average number of channels is presented.
~500 μL of venous blood was lysed in 8 mL ACK Lysis Buffer (Cat. # 351-029-721; Quality Biological, Inc.) for 5 min at room temperature, and the reaction stopped by diluting with 7 mL cold HBSS + 0.1% BSA. Cells were pelleted, washed, and prepared for staining.
To prepare for skull BM extraction, the dorsal calvarium was trimmed to be relatively flat, and meninges were removed under a dissecting microscope. Next, the skull was cut into small bone pieces with scissors in cold HBSS + 0.1% BSA. This entire slurry was transferred to a 70 μm cell strainer and mashed with the rubber end of a 3 mL syringe for approximately 2 min per sample. Tibia were prepared by first stripping away all tissue from the bone, then cutting the very top such that a 23 g syringe needle could be inserted to flush out the BM into cold HBSS + 0.1% BSA. This was next transferred to a 70 μm cell strainer and mashed with the rubber end of a 3 mL syringe. For both BM tissues, the resulting cell suspensions were then pelleted and prepared for flow cytometry.
Meninges samples were collected by first cutting around the lateral sutures of the skull; the dorsal skull and ventral skull were transferred to a fresh petri dish filled with cold HBSS + 0.1% BSA and kept on ice while pial and arachnoid meningeal membranes were gently picked off the entire outer surface of the brain into a second 'working dish' with Dumont #5 forceps (Cat. #RS-5058; Roboz, Gaithersburg, MD). Extra care was taken to avoid inclusion of choroid plexus from the 4 ventricle. Once finished with the brain, skull pieces were transferred as necessary into the working dish to remove attached meninges; we avoided leaving skull pieces in the'working dish' to minimize contamination with cells from skull BM. Upon completion of the meningeal dissection, samples were transferred to a fresh tube and cells were pelleted by centrifugation, then resuspended in 2 mL of BSA-free HBSS supplemented with 2.5 mg/mL Collagenase D (Cat. #11088858001; Roche) and 12.5 μL of 0.5 mg/mL DNAseI (Cat. #L5002139; Worthington) for cell dissociation. The samples were incubated at 37 °C for 30 min, diluted with cold HBSS + 0.1% BSA, and mashed through a 70 μm cell strainer into single cell suspension.
The dorsal skull was carefully removed to retain maximal attachment of the meningeal layers from tissue prepared for histology, and brains were returned to fresh 25% sucrose until sunk for further staining. Dorsal meninges were gently peeled from the skull as a single sheet, mounted ventral (brain) side up/dorsal (skull) side down onto slides and encircled with a Pap pen.
Meningeal whole mount samples from LysM mice were dried, washed with PBS, blocked for 1 h in 4% normal goat serum in 0.4% Triton-PBS, and incubated in a humidity chamber overnight at room temperature with chicken anti-GFP (1:1000, Cat #13970, Abcam), diluted in 0.2% Triton-PBS with 2% normal goat serum. Approximately 18 h later, the samples were washed 3 x 5 m with PBS and incubated for 2 h with Chicken IgY-Alexa Fluor 488 (1:500, Cat #150169, Abcam) in 0.4% Triton-PBS. Samples were washed 2 x 5 m with PBS, given a quick rinse in deionized water, and counterstained with DAPI for 5 min. They were rinsed briefly again with deionized water then cover-slipped with PVA-DABCO (made in-house).
Meningeal whole mounts were tile scanned using a 20x objective at 1024 × 1024 resolution and online stitching with a Zeiss 780 confocal microscope. Two independent investigators blind to treatment hand-counted confocal images of the meningeal whole mounts to quantify the density and location of LysM-GFP cells, which we assumed were neutrophils (characterized by high GFP expression, irregularly shaped nuclei, and a semi-round shape) within the tissue using ImageJ software. There was excellent agreement in their counts (Pearson correlation, ****p < 0.0001, r = 0.937). LysM-GFP cells were also examined in relationship to blood vessels; LysM-GFP cells > 10 μm from a blood vessel were provisionally called 'non-vascular,' whereas LysM-GFP cells ≤ 10 μm were called 'abluminal.' LysM-GFP cells were otherwise considered intravascular.
Brains were sliced into 30 μm sections on a freezing microtome and stored in an ethylene glycol solution at -20 °C until staining. Free-floating sections were washed in PBS, blocked with 4% normal goat serum, and incubated overnight at room temperature using the same staining protocol as that used for meningeal tissue. Mounted sections were tile scanned using a 20x objective at 1024×1024 resolution and online stitching with a Zeiss 780 confocal microscope. To obtain LysM myeloid cell counts, two blinded experimenters counted strongly GFP cells using ImageJ software.
A subset of samples was stained for p-Selectin as above with the addition of p-Selectin (Cat#148301, Biolegend) diluted at 1:1,000 in 0.4% Triton-PBS. An additional secondary antibody conjugated to Alexa Fluor 555 (Cat#A-31570, Invitrogen) in 0.4% Triton-PBS at a dilution of 1:500 was used. Volocity (PerkinElmer) was utilized to quantify the amount of TomL blood vessels associated with P-selectin staining, which was used to calculate the percent P-selectin coverage.
As previously, brains collected from saline-perfused mice were placed in a gentleMACS™ C tube (Cat#130-093-237, Miltenyi). 2.5 ml of 2.5 mg/ml Collagenase D plus 10 μL of Solution A and 20 μL Solution Y from a Neural Tissue Dissociation Kit (Cat#130-094-802, Miltenyi) in HBSS was added to each sample, which was minced with a gentleMACS Dissociator for 45 sec using brain protocol 1. Tubes were incubated for 30 min at 37 °C with rotation, triturated 100x with a p1000 pipette tip, and strained through a 70 μm nylon mesh cell strainer (Cat#352350, BD Biosciences). The filter was washed with 20 ml HBSS, and eluate pelleted in a swinging bucket centrifuge at 300 g for 5 min. The cell pellet was resuspended in 30% isotonic Percoll (Cat#P4937, Sigma-Aldrich) then centrifuged at 800 g for 30 min with no brake. The myelin layer was removed, and Percoll and cell pellet layers were diluted 4-fold with HBSS for washing. This was centrifuged at 300 g for 5 min; the resulting cell pellet was then stained for flow cytometry.
To exclude dead cells, samples were stained with either Fixable Viability Dye eFluor™ 780 (Cat. #65-0865; eBioscience) for 10 min at room temperature at 1:2400 dilution, or with Zombie AQUA (Cat # 423102, Biolegend) for 15 min at room temperature at 1:100 dilution. The cells were then washed and blocked with 1 μL of normal goat serum (Cat #G9023-10ML; Sigma) and 1 μL of Fc block (Cat #101302; Biolegend) for 10 m on ice. 25 μL Brilliant Violet stain buffer (Cat #563794; BD Horizon) was then added, followed by an antibody master mix (Table S13). Cells were incubated on ice for 25 min, washed with PBS, and fixed in 2% PFA at room temperature for 15 min for analysis on either a BD LSR Fortessa or a Beckman Coulter CytoFLEX flow analyzer. Compensation was performed for each session using UltraComp eBeads (eBioscience 01-2222-42) conjugated to antibodies used in the sample panels. Viability dye and GFP controls used cells instead of beads. Data were analyzed using FlowJo (BD) software (multiple versions) with manual gating. Absolute cell counts were determined using CountBrite counting beads (ThermoFisher, catalog #C36950). See Supplementary Information for gating strategies used in individual experiments.
Meningeal scRNAseq data were acquired from 8 non-stressed HC mice and 4 stress-susceptible CSD mice as described previously. In brief, live, nucleated, singlet cells (DAPIDRAQ5) were sorted on a BD FACS Aria Fusion into HBSS + 10% FBS prior to droplet encapsulation using 10x Genomics' Drop-seq platform (Chromium v2). 4 pooled samples of 4 mice each were generated (two HC pools, two CSD pools) and run on the same 10x Chromium chip, though 1 CSD sample was lost due to errors with droplet encapsulation; libraries were sequenced on Illumina NextSeq 550 and feature counts generated using Cellranger V2 pipeline. Single cell sequencing data have been deposited in GEO under accession code GSE301684.
We performed the following steps to obtain N = 6694 quality-controlled single cells: cell calling using DropletUtils::emptyDrops; exclusion of outlier cells based on mitochondrial reads (<8.3% of total) or total features per cell (range for included cells = 174 - 4548); doublet removal using scrublet, with doublet rates in the three samples estimated as 6.9%, 5.1% and 3.4%. Samples were normalized using scran deconvolution-based normalization highly-variable genes (3599 genes) were selected as genes with biological variation across samples > 0 (using scran::decomposeVar). Batch correction across the three 10X lanes was performed using batchelor::multiBatchNorm and fastMNN (default 50 components used for dimensionality reduction). Clustering of MNN-corrected PCA components across all single cells was performed using the leidenalg clustering algorithm. Marker genes were detected by two methods (a) scran findMarkers function, based on differential expression between clusters and (b) soupX quickMarkers function, which uses Term Frequency - Inverse Document Frequency (TF-IDF) to identify the genes most predictive of a cluster, based on the frequency with which a gene is expressed in a cluster. Clusters were manually annotated by comparing marker genes expressed with existing single cell datasets. differential gene expression (DGE) between CSD and HC cells was performed as follows: counts were renormalized within the cluster; genes differentially expressed in pseudobulk of empty droplets (i.e., likely representing ambient RNA) were removed as described elsewhere; genes expressed in ≤ 15% of cell in the cluster were removed; then gene expression in CSD vs. HC cells was compared using a Mann-Whitney U test, with Benjamini-Hochberg FDR correction of p-values across all tested genes. Pre-ranked gene set enrichment analysis was performed using clusterProfiler with genes ranked by -log10 (Mann-Whitney U test P-value) * sign(LFC). Cell cycle stage of each cell was estimated using scran::cyclone. Results were plotted using Seurat, bespoke code, or ktplots and pathway enrichment was performed using [R] clusterProfile GSEA function with signed -log10(p-value) as the ranking statistic and GO biological pathways accessed via msigdbr.
To compare our neutrophil transcriptional data to neutrophils from other tissues, we reprocessed two public mouse RNAseq datasets that each included neutrophils acquired from multiple tissues. For the public Kolabas dataset (scRNAseq), we summed raw cell counts per tissue and cell type to create pseudobulk profiles for neutrophils and preneutrophils across multiple tissues, retaining pseudobulk profiles including at least 25 cells, and excluding two outlier samples. Variance stabilizing transformation (VST) implemented in the DESeq2 package was applied to the pseudobulked counts. Samples included bone marrow from femur, humerus, pelvis, scapula, skull, and vertebra. Our meningeal neutrophil and pre-neutrophil data were pseudobulked and transformed in the same way. For the public Evrard dataset (bulk sorted cell RNAseq), we selected control samples representing mature, immature and preneutrophils from femur, plus mature neutrophils from blood, then applied DESeq2 VST. PCA was performed on the Kolabas dataset using the intersect between the 20% most highly variable genes in the Kolabas dataset and the genes present in all three datasets. VST matrices from the stress and Evrard datasets were quantile-normalized and projected into the Kolabas PCA space to enable direct comparison (Fig. S14).
We generated cluster centroids and a minimum spanning tree for the neutrophil subclusters using scran createClusterMST. We obtained a pseudotime ordering by mapping cells to this tree using TSCAN mapCellsToEdges. Scran testPseudotime was used to find genes significantly associated with pseudotime.
Meningeal tissue was dissected as described above, but with the following modifications: no digestion was used, intact tissue was passed through a 70 μm nylon mesh cell strainer using a 1 ml syringe plunger. Cells were then centrifuged and stored in Trizol Reagent (Cat #15596026, ThermoFisher) at -80 °C. For RNA extraction, samples were thawed and triturated using syringe needles before using a Qiagen miRNeasy Mini kit (Cat#217004). Labelled probes were run on the Affymetrix GeneChip™ Mouse Gene 2.0 ST Array (Cat#902118) according to the manufacturer's guidelines. Data were RMA-normalized and used to test for DEGs in CSD vs HC conditions, as described previously. Microarray data have been deposited in GEO under accession code GSE275966.
Single cell suspensions were generated as above, and relative ROS production was analyzed using the CellROX™ assay (Cat #C10492, ThermoFisher) following the manufacturer's instructions. After ROS labeling, cells were stained for flow cytometry and promptly run unfixed on a flow analyzer.
Single cell suspensions of blood and meningeal cells were generated and stained for flow cytometry as described above. Cortical actin was stained by adapting previously published methods. Specifically, prepared cells were washed in PBS, then fixed in a modified "superfix" solution: 7.75 ml custom buffer (100 mM KCL, 3 mM MgCL, 10 mM HEPES, 150 mM sucrose, pH 7.4), 1 ml freshly made 37% formaldehyde solution (from powder, made in PBS), 1 ml DMSO, 200 μl of 100 mM EGTA, and 10 μl of 50% glutaraldehyde (in PBS). To this, Alexa Fluor™488 phalloidin (Cat #A12379, ThermoFisher) was added at a final concentration of 2 U/ml for filamentous (F)-actin visualization, and Hoechst-33342 (Cat #H3570, ThermoFisher) was added at a dilution of 1:10,000 to label nuclei. Imaging flow cytometry analyses was performed on an ImageStream MarkII (Cytek Bio) 2-laser, 12-channel instrument. Spectral compensation was performed with single color beads and cells as described above. Samples were acquired at 60x magnification and low speed. Data were analyzed using IDEAS 6.02 software (Amnis Corp.) and FlowJo.
Blood plasma was analyzed using ELISA kits for murine CXCL1 (Cat #900-K127, PreproTech), CXCL2 (Cat #900-K152, PreproTech), and CXCL12 (Cat #DY460, R&D Systems) according to manufacturer's instructions. To compensate for sample plasma effects, assay standards were diluted in the CXCL12 RD6Q mouse standard diluent. In all assays, EIA/RIA high binding polystyrene flat bottom well plates (Cat # 2580, Costar) were used for the attachment of capture antibody. Sample absorbance was detected at 450 nm using a Victor 3 plate reader. Data analysis was performed using a four-parameter fit of resultant absorbances in Prism 9.0.
Samples used for RT-qPCR were resuspended and stored in Trizol at -80 °C until further purification with the Qiagen RNeasy kit (Cat #74104, Qiagen) according to the manufacturer's instructions. Total RNA was quantified and 1 μg was converted into cDNA using Superscript II reverse transcriptase (Cat #18064014, ThermoFisher) and Oligo dT primers (Cat #18418020, ThermoFisher). RT-qPCR was performed using the below listed primer sets and 2xSYBR Green Master Mix (Cat #172570, Bio-Rad) in a Bio-Rad MyIQ iCycler. The cycling program used an initial denaturation for 3 m at 95 °C, followed by 40 cycles of the following: 95 °C for 15 s, 58 °C for 30 s, 72 °C for 30 s. Levels of the targeted gene expression were determined by comparison of the sample Ct values to standard curves of Ct vs. dose of the amplicon of interest, with normalization to TATA binding protein (TBP). The amplicons of each primer set for any target gene were validated by sequencing prior to these experiments. Primer sequences and GenBank accession numbers are: CXCL1 (NM_008176. Fwd: GCTGGGATTCACCTCAAGAA, Rev: TGGGGACACCTTTTAGCATC), CXCL2 (NM_009140. Fwd: AAGTCATAGCCATCTCAAGGG, Rev: CTTCCGTTGAGGGACAGCAG), CXCL12 (NM_021704. Fwd: CTGCATCAGTGACGGTAAAC, Rev: TCCACTTTAATTTCGGGTCA), TBP (NM_013684. Fwd: GACCCACCAGCAGTTCAGTA, Rev: AAACACGTGGATAGGGAAGG).
Skulls were obtained and prepared for single-cell suspension. The cells were then fixed, pelleted onto thin coverslips using funnel centrifugation (Cat #10-354, Fisher HealthCare), and imaged using a Zeiss 780 confocal microscope fitted with 10x objective.
Anti-IFNAR (clone: MAR1-5A3, Cat # BE0241) and non-specific, IgG isotype control (clone: MOPC-21, Cat # BE0083) antibodies for repeated in vivo injections were purchased from BioXCell and diluted in InVivoPure pH 7.0 Dilution Buffer (Cat # IP0070) to a concentration of 5 mg/mL. Mice were i.p. injected with 1 mg of antibody on day zero (d0), before the start of the defeat paradigm, and received 0.5 mg of antibody every third day thereafter, for a total of 5 injections per mouse. Defeats were done approximately 1-3 h after injections. LysM mice were ear-tagged and randomly assigned to one of three groups: HC+IgG, CSD+IgG, or CSD + IFNAR. WT mice were ear-tagged and randomly assigned to one of four groups: HC+IgG, HC + IFNAR, CSD+IgG, or CSD + IFNAR.
To further explore the mechanistic role of neutrophils in the response to chronic social defeat (CSD) stress, we first performed neutrophil depletion experiments using methods optimized for prolonged treatment. Briefly, 8 mice were randomly assigned to either the HC or CSD group. 2 mice from each group were then randomly assigned again to either treatment (Ly6G antibody depletion) or control (IgG antibody) conditions, comprising 4 groups in total: HC+IgG, HC+Ly6G, CSD+IgG, CSD+Ly6G (n = 2 per group). Primary (1) rat anti-mouse Ly6G antibody (clone: 1A8, Cat # BE0075, Bio X Cell) and 50 μg secondary (2) anti-rat antibody (clone: MAR18.5, Cat # BE0122) were administered on alternating days for the treatment group, and an equivalent amount of IgG antibody (clone: 2A3, Cat # BE0089) was administered to control condition mice, according to the timeline shown in Fig. S26A. In this pilot we noted (but did not quantify) ear tag holes in both HC and CSD groups that became enlarged over time, indicating possible wound healing deficits. Using our rubric for assessing fight-related wounds in CSD animals (Table S12), we determined that neutrophil depletion significantly impaired wound healing of the typically minor injuries sustained during the CSD protocol (Fig. S26B). We also noted massive splenomegaly (>300 mg) in two Ly6G-treated mice (one HC, one CSD). Due to concerns about animal welfare, we ceased using this method for CSD experiments.
We pursued a modified neutrophil depletion protocol with shortened timeline for defeat exposure (Fig. S26C) and reduced dosage of antibody compared to that typically used -- i.e., 100 μg compared to 200 μg. 14 mice were randomly assigned to either CSD+saline (n = 6) or CSD+Ly6G conditions (n = 8). After 10 days of defeat, wound scores were similar between groups, but massive splenomegaly was again observed in the neutrophil-depleted animals (Fig. S26D, E). Discussion of these data with animal care staff led us to cease pursuit of these studies due to a combination of animal welfare concerns and uncertainty about experimental confounding, as any immunological or behavioral differences observed could have been driven by gross immune abnormalities related to acute splenomegaly rather than effects on meningeal neutrophils in neutrophil-depleted CSD mice.
Prism 9.0.2 and 10.0 (GraphPad Software, LLC) was used for statistical testing and graphing of univariate analyses. Normality and equal variance were assessed, and appropriate tests were conducted thereafter. For simple, two-group comparisons, two-tailed tests were used. Samples with non-equal variance were analyzed using a Mann-Whitney U test, whereas samples with equal variance were analyzed using a Student's t test. In almost all instances, CSD-stressed mice failed to mark in the USM test; the behavioral output was thus better modeled statistically as a binary response (did or did not mark). We therefore used χ tests for univariate analyses of group in the USM test; posthoc testing was performed with the chisq.posthoc.test package in R, using the Benjamini-Hochberg method for multiple comparisons corrections. For multiple comparisons of normally distributed data, ordinary one- or two-way ANOVAs were used for analysis. When analyzing complete data from multiple tissues within the same mice, repeated measures with an interaction term were selected for the model; for data with missing values, Restricted Maximum Likelihood (REML) mixed-effects models were used. When a main effect was present, post-hoc analyses were conducted (Dunnett's, Tukey's, or Śídák, respectively). For multiple comparisons of non-parametric data, a Kruskal-Wallis test was used, and Dunn's test for multiple corrections was run for post-hoc analysis. Univariate comparisons are summarized as the mean ± SEM and considered statistically significant at p < 0.05. No power calculations were performed at the outset of the experiment.
Prior to debatching and for univariate analyses of meningeal histology data, predictor variables were assessed for outliers (using Grubb's test for outliers), skew, kurtosis, normality (using the Shapiro-Wilks test), and equal variance (using Bartlett's test) in R version 4.3.0. Increasingly strict transformations were applied to individual variables until normal distributions and assumptions of equal variance were met (in order from least stringent to most stringent transformations: square root, log natural log). Ordinary least squares (OLS) regression was used to assess relationships between predictor variables (meningeal neutrophils, circulating white blood cell populations) and behavioral outcomes. Linear models were constructed in R with the following formula: [dependent variable] ~ [continuous predictor] + batch. Model fit was assessed via quantile-quantile plots and by examination of residual distributions. For multivariate analysis of USM, a generalized linear model was constructed with binomial family selection according to the formula USM_binary ~ [continuous predictor] + batch. Model fit was assessed via quantile-quantile plot. Coefficient estimates of the models were plotted using dotwhisker::dwplot in R. The mixed linear regressions were treated as exploratory; thus, unadjusted p values are presented in the Figures. However, For mixed linear regressions, Benjamini-Hochberg (FDR) corrections were performed to control the type I error rate; unadjusted p values are presented in Tables S1-S4.
Linear regression was used to assess wounding as a predictor for immune cell populations according to the formula [cell population] ~ [wound score]. To assess relationships between immune cells in different tissues, a Pearson correlation matrix was first generated with the function Hmisc::rcorr. Unsupervised hierarchical clustering was then performed using the Lance-Williams dissimilarity formula ("hclust" in R base stats) and the gplots::heatmap.2 function for plotting. The unsupervised clustering order for neutrophils was used to force clustering of monocytes and lymphocytes. Bonferroni corrections were applied to control the type I error rate; significant surviving associations were added to the heatmap manually, indicated with an asterisk.
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.