Androgens show sex-dependent differences in myelination in immune and non-immune murine models of CNS demyelination

Neuroprotective, anti-inflammatory, and remyelinating properties of androgens are well-characterized in demyelinated male mice and men suffering from multiple sclerosis. However, androgen effects mediated by the androgen receptor (AR), have been only poorly studied in females who make low androgen levels. Here, we show a predominant microglial AR expression in demyelinated lesions from female mice and women with multiple sclerosis, but virtually undetectable AR expression in lesions from male animals and men with multiple sclerosis. In female mice, androgens and estrogens act in a synergistic way while androgens drive microglia response towards regeneration. Transcriptomic comparisons of demyelinated mouse spinal cords indicate that, regardless of the sex, androgens up-regulate genes related to neuronal function integrity and myelin production. Depending on the sex, androgens down-regulate genes related to the immune system in females and lipid catabolism in males. Thus, androgens are required for proper myelin regeneration in females and therapeutic approaches of demyelinating diseases need to consider male-female differences.

cortical layers (grey matter to right of dashed line) and the subcortical white matter (left of dashed line) from healthy woman in (b) and man in (i) shows clear signal in the grey matter and lower signal in the white matter contrasting with virtually undetectable signal observed in sections incubated with only the secondary antibody in (a, h). An active MS lesion derived from a female patient display a strong immunostaining in (c) whereas an active lesion in (j) or a chronic active in (k) lesion from male patients displays an extremely faint signal. ARexpressing cells in the female active lesion co-express the CD68 marker of microglia/macrophages (blue and brown) in (d). e, f AR + CD68 + ramified cells in the perilesion white matter in (e) and absence of AR-expressing cells in the normal appearing white matter (NAWM) in (f). g Percentage of CD68 + AR + cells in the indicated brain areas from 3 female MS patients and 1 control female donor. The numbers are limited due to discontinuation of the antibody, but the percentage of CD68 + AR + cells appear higher in active lesions and in the perilesion in (g, values are means ± SEM). MS CA, MS chronic active lesion. Scale bars (µm) : 100 in (a, b, h, i, j, k), 50 in (c-f). l AR mRNA expression in microglia from MS and control female (F, pink) or male (M, blue) donors from a publicly available independent single-nuclei RNA sequencing database (https://malhotralab.shinyapps.io/MS_broad/) seems to show higher expression in MS and female samples 1 . AL, active lesion (samples from 8M, 13F); CAL, chronic active lesion (samples from 8M, 9F); CIL, chronic inactive lesion (samples from 6M, 7F); NAWM, normal-appearing white matter (samples from 5M, 11F); WM, white matter from control donors (samples from 6M, 7F). The median is shown with box indicating 25% quartile to 75% quartile and whiskers indicating interquartile range multiplied by 1.5. Figure 2. Aromatase is expressed in the demyelinated corpus callosum from female mice. Aromatase immunostaining at 7 days after stereotaxic injection of LPCvehicle (left) or LPC into the corpus callosum of female mice give rise to a faint or very strong labeling, respectively. These micrographs are representative of the labeling obtained in two independent experiments. The LPC-induced lesion is delineated by the dashed line. Scale bar: 50 µm. Figure 3. Blockade of the androgen receptor by its pharmacological antagonist flutamide impedes MBP expression at 10 days after LPC injection in the 129X1/SvJ strain. Visualization and quantification of MBP immunostaining at 10 dpl. The lesions are delineated by the dashed lines. Scale bar: 50 µm. Data are presented as mean values ± SEM from n=3 animals / condition examined in a single experiment. The p value was calculated by using the unpaired two-tailed t-test. ***, p=0.0003. Source data are provided as a Source Data file.  Figure 8. Data are presented as mean values ± SEM. P values were calculated by using the unpaired two-tailed t-test or Mann-Whitney test according to the value derived from the test for normality. Whelch's correction was used in (c). *, p≤0.05; **, p≤0.01; ***, p≤0.001; ****, p≤0.0001. n.s., non-significant. Source data are provided as a Source Data file.

Supplementary Figure 10. Comparison of parenchymal inflammatory cells and levels of myelin expression in EAE females and males under Vehicle (a-g) or DHT (h-n) treatment
at 14 dpi. The data are derived from spinal cord immunostaining experiments using the microglial (Iba1, Arg-1), astroglial (GFAP) and myelin (MBP) markers shown in Figure 9. Data are presented as mean values ± SEM from n=8 animals / group examined in a single experiment. P values were calculated by using the unpaired two-tailed t-test or Mann-Whitney test according to the value derived from the test for normality. Welch's correction was used for (b, e, f, j, l, m). *, p≤0.05; **, p≤0.01; ***, p≤0.001; ****, p≤0.0001. n.s., non-significant. Source data are provided as a Source Data file.

Supplementary Figure 11. Normalization of female and male RNA-Seq analysis. (a)
Integration of female and male datasets by using normalization via housekeeping genes present in the two datasets aimed at reducing unwanted variation from RNA-Seq data (RUVSeq). (b) Differential analysis performed before (top) or after (bottom) housekeeping gene normalization indicating comparable numbers of DEGs in particular between females (n=3 / group) treated (DHTF) or not (CTF) with DHT, and between males (n=3 / group) treated (DHTM) or not (CTM) with DHT examined in two independent experiments. CT corresponds to non-treated male and female animals. In order to control the false discovery rate (FDR, p-adjust), multiple testing correction was performed by using the Benjamini-Hochberg method.  Tables and  barplots showing the contribution of oligodendroglial curated DEGs genes to promote (positive,  blue) or inhibit (negative, orange) each process of oligodendrogenesis in females (a-e) and in males (f-g). In all the bar plots, the title of the x-axis is indicated just above the axis. The title of the y-axis corresponds to the oligodendrogenesis processes labeled in each bar. Note that DHT mainly promotes myelination. (k, l) Venn-diagram comparing the number of oligodendroglial genes involved in the myelination process shared (25 genes) between DEGs in females (94 genes) and males (50 genes).  Table 12) according to their deregulation in DTH-treatment comparisons, representing the top 10 gene sets activated or suppressed in (a) females (DTH-treated compared to controls) and (b) males (DTH-treated compared to controls). Note that only in females there is an enrichment in suppressed gene sets related to immune and inflammatory processes, with no enrichment of suppressed gene sets in males. c GSEA plots visualization illustrating immune/inflammatory gene sets suppressed (i.e., genes being downregulated) in DTH-treated females. In order to control the false discovery rate (FDR, p-adjust), multiple testing correction was performed by using the Benjamini-Hochberg method.  AL, active lesion; CAL, chronic active lesion; CIL, chronic inactive lesion; NAWM, normalappearing white matter; PMI, post-mortem interval; other metadata are available on request.

Supplementary Table 2. Characteristics of human post-mortem cases included in immunostaining.
AL, active lesion; CAL, chronic active lesion; PMI, post-mortem interval; other metadata available on request.

Supplementary Methods
Chromogenic immunohistochemistry on human cryosections. Human tissue cryosections (10 µm) were fixed in 4% paraformaldehyde in 1xPBS for 1 hour at room temperature. Endogenous peroxidase and alkaline phosphatase activity was blocked by a 10 min incubation with Bloxall solution (SP-6000, Vector Laboratories Bulk RNA-Seq deconvolution. We used CIBERSORTx tool 2 on the docker module Cibertsortx/fractions, with 100 permutations as input parameter, in order to deconvolute our bulk RNA-Seq datasets obtained from EAE spinal cord samples. The signature of scRNA-Seq matrix was generated according to the book methods described by Steen et al. 3 with the GSE113973 public scRNA-Seq dataset from mouse EAE model. The deconvolution analysis was performed on two mixture files corresponding to the RNA-Seq count matrices generated as described before, containing females and males' comparisons, DHTFvsCTF and DHTMvsCTM, respectively. The results obtained are estimated as the proportions of each cell types in each RNA-Seq sample inferred from the prior knowledge of the scRNA-Seq sample. R script has been deposited in https://github.com/ParrasLab/Androgen-signaling-andremyelination-Nat-Commun-paper. Statistical analysis. Statistical analysis of mouse histological staining was performed with GraphPad Prism 7.0 software (La Jolla, CA). The significance of differences between means was evaluated by the two-tailed Student's unpaired t test for two independent group comparisons when distribution was normal. In case of absence of normal distribution (analyzed D'Agostino & Pearson normality test and Shapiro-Wilk normality test), the non-parametric two-tailed Mann-Whitney test was used. Appropriate corrections were done according to the determination of the variance of each sample. The values are the means ± SEM from the number of animals indicated in each plotted graph or as indicated in the corresponding legends. Significance of p<0.05 was used for all analyses. *, p0.05; **, p0.01; ***, p0.001; ****, p<0.0001. For transcriptomic analyses, multiple testing correction aimed at controlling the false discovery rate (FDR) was performed using the Benjamini-Hochberg method in supplementary tables 8 and 12. Cutoff used for FDR was 5%. In supplementary tables 3, 9, 10, 11, the workflow used edgeR's quasi-likelihood (QL) pipeline (edgeR-quasi) for differential expression. This statistical methodology uses negative binomial generalized linear models but with F-tests instead of likelihood ratio tests. This method provides stricter error rate control than other negative binomial based pipelines, including the traditional edgeR pipelines or DESeq2 4 .