Microglial SIRPα regulates the emergence of CD11c+ microglia and demyelination damage in white matter

A characteristic subset of microglia expressing CD11c appears in response to brain damage. However, the functional role of CD11c+ microglia, as well as the mechanism of its induction, are poorly understood. Here we report that the genetic ablation of signal regulatory protein α (SIRPα), a membrane protein, induced the emergence of CD11c+ microglia in the brain white matter. Mice lacking CD47, a physiological ligand of SIRPα, and microglia-specific SIRPα-knockout mice exhibited the same phenotype, suggesting that an interaction between microglial SIRPα and CD47 on neighbouring cells suppressed the emergence of CD11c+ microglia. A lack of SIRPα did not cause detectable damage to the white matter, but resulted in the increased expression of genes whose expression is characteristic of the repair phase after demyelination. In addition, cuprizone-induced demyelination was alleviated by the microglia-specific ablation of SIRPα. Thus, microglial SIRPα suppresses the induction of CD11c+ microglia that have the potential to accelerate the repair of damaged white matter.


Introduction 42
Microglia constantly survey the microenvironment of the brain. When microglia 43 encounter tissue damage, they become activated, produce multiple humoral factors and have the potential to support the repair of damaged myelin. addition, CD11c + microglia in the aged mice expressed higher levels of CD14, Dectin-1, 148 and CD68 compared with CD11c − microglia as in SIRPα KO mice ( Figure 3F). 149 To address the effect of the increase of CD11c + microglia on inflammatory 150 responses, we examined the expression of pro-and anti-inflammatory cytokines in the 151 brain and spinal cord by quantitative PCR analysis. Among the examined cytokines 152 (TNF-α, IL-1β, IL-6, IL-10 and TGF-β), the expression of TNF-α was increased in 153 SIRPα KO as compared to WT mice (Figure 3-figure supplement 2). 154 155

Induction of CD11c + microglia in the white matter of CD47-deficient mice 156
To address the mechanism involved in the regulation of microglia activation by SIRPα, 157 we examined the effect of the genetic ablation of CD47, a membrane protein and SIRPα 158 ligand (Matozaki et al., 2009). In the brain of CD47 KO mice, CD11c + microglia were 159 increased in the white matter as observed for the brains of SIRPα KO mice ( Figure 4A). 160 The number of Iba1 + microglia was increased more than 2.5-fold in the hippocampal 161 fimbria, and about 70% of the Iba1 + microglia expressed CD11c ( Figure 4B). To address the impact of the emergence of activated CD11c + microglia on the brain Spp1, trophic factors promoting oligodendrocyte differentiation; and Cst7 (Cystatin F), 202 a cysteine proteinase inhibitor. We also found that the positive regulators of microglia 203 phagocytosis, Trem2, DAP-12, and Cx3cr1 (Lampron et al., 2015;Poliani et al., 2015), 204 were increased in the white matter of the mutant mice. 205 We next examined the gene expression in the brain mononuclear cells, in 206 which microglia were enriched. Expression of total 16,544 genes was detected in both 207 or either of CD47 KO and WT control cells. Among them, 1,323 and 2,286 genes were 208 markedly (> 2-fold) increased and decreased, respectively, in the CD47 KO cells 209 (Supplementary file 2). Genes increased in CD47 KO brain cells were significantly 210 enriched in pathways for T cell receptor signal, axon guidance, proteoglycans in cancer, 211 TNF signal, and NF-κ B signal ( Figure 5A); genes decreased in CD47 KO brain cells 212 were enriched in cancer associated pathways including Wnt, Hippo, and Rap1 signalling, 213 as well as cardiomyopathy ( Figure 5B). 214 Comparison of array data revealed 32 and 55 genes that were commonly 215 increased and decreased (> 2-fold), respectively, in both the white matter and the brain 216 mononuclear cells of CD47 KO mice (Figure 5-figure supplement 2). Shared 217 induced genes included myelin-repair related genes, such as ItgaX, Igf1, Lpl, Apoc1, 218 Ch25h, Mmp12, Spp1, and Cst7 ( Figure 5C). The expression of Clec7a, CD68, Trem2, 219 and Cx3cr1, which were markedly increased in the white matter of CD47 KO mice, 220 showed only moderate (< 2-fold) increase in the brain mononuclear cells (Figure 5C). 221 Substantially higher expression of these genes in microglia might mask the increased 222 expression of these genes in the limited CD11c + subset that was only ~5 % of the total 223 microglia when prepared from the whole brain of CD47 ΚΟ mice (data not shown). 224 225

SIRPα-deficient mice 227
To examine the cell type involved in the suppression of CD11c + microglia by SIRPα, 228 we analysed microglia-specific SIRPα conditional KO (cKO) mice that were generated 229 by crossing SIRPα-flox mice (Washio et al., 2015) and Cx3cr1-CreER T2 mice to 230 achieve microglia-specific gene targeting (Goldmann et al., 2013;Safaiyan et al., 2016;231 Wolf, Yona, Kim, & Jung, 2013). Flow cytometric analysis revealed that greater than 232 98% of microglia were SIRPα negative in the brain of tamoxifen-treated 233 SIRPα fl/fl :Cx3cr1-CreER T2 mice (Figure 6A), and the numbers of Iba1 + and 234 Iba1 + /CD11c + microglia were increased in the hippocampal fimbria of these cKO mice 235 ( Figures 6B and 6C). These data suggest that a lack of the interaction between SIRPα 236 on microglia and CD47 on neighbouring cells is the primary cause for the induction of 237 CD11c + microglia in CD47-SIRPα signal-deficient mice. We also analyzed CD11c + 238 cell-specific SIRPα cKO mice (Washio et al., 2015). However, CD11c + microglia were 239 not increased in these mutant mice (Figure 6-figure supplement 1), suggesting that a 240 lack of SIRPα in the resident CD11c + microglia, a small subset of microglia in normal 241 mouse brain (Bulloch et al., 2008), did not cause the expansion of these cells. 242 243

microglia-specific SIRPα-deficient mice 245
Although myelin damage induces CD11c + microglia (Remington et al., 2007), 246 demyelination was not observed in SIRPα total KO mice by light microscopy ( Figures  247   2C and 2D). We examined the myelin structure again in microglia-specific cKO mice 248 by electron microscopy (Figures 6D and 6E). In the cross section of the anterior 249 commissure, the frequency of myelinated axons and the g-ratio (the ratio of the inner 250 axonal diameter to the total outer diameter) were comparable between SIRPα cKO and 251 control mice [frequency of myelinated axons: 34.3 ± 4.37% for control (n = 3), 36.3 ± 252 4.18% for cKO (n = 3), P = 0.757, Student's t-test: g-ratio: 0.63 ± 0.025 for control (n = 3), 0.657 ± 0.009 for cKO (n = 3), P = 0.374, Student's t-test], suggesting the myelin 254 structure was normal in the mutant mice. microgliosis, CD11c + /Iba1 + microglia were markedly increased, even in control mice 263 ( Figure 7A). In SIRPα cKO mice, demyelination as well as microgliosis was 264 significantly reduced after the same treatment (Figures 7A-7C). At an early stage of 265 Cpz treatment (after 3 wks feeding with Cpz), demyelination was not observed in 266 control or SIRPα cKO mice ( Figure 7A), but abnormally strong immunoreactivity of 267 MBP was observed ( Figure 7A). This was probably related to myelin damage, because 268 the epitope of the anti-MBP antibody we used (DENPVV) is similar to that of a myelin 269 damage-detectable antibody (QDENPVV) (Matsuo et al., 1998). Consistently, 270 significant microgliosis as well as an increase in CD11c + cells were observed in both 271 genotypes after 3 wks feeding with Cpz ( Figures 7A and 7C). At this stage, the area of 272 microgliosis was significantly larger in SIRPα cKO mice compared with control mice 273 ( Figures 7A and 7C). The repair of demyelination was observed after feeding with 274 normal chow for 2 wks ( Figure 7A). Microgliosis was still present in both genotypes at 275 this time point but was significantly smaller in SIRPα cKO mice compared with control 276 mice (Figures 7A and 7C). In the white matter, including the corpus callosum and 277 alveus, the number of cells expressing Olig2 + , an oligodendrocyte progenitor and 278 lineage undergoing terminal differentiation into mature olidodendrocytes (Nishiyama, suggesting demyelination was preceded by the loss of oligodendrocytes ( Figures 7A  281   and 7D). The number of Olig2 + cells was increased at 5 wks of Cpz treatment and 282 recovered to normal levels after another 2 wks feeding with normal chow (Figures 7A  283   and 7D). Throughout the de-and re-myelination process, no significant differences in 284 the number of Olig2 + cells were noted between control and SIRPα cKO mice ( Figures  285   7A and 7D). 286

Discussion 287
Cell-cell interactions between SIRPα on microglia and CD47 on neurons were 288 proposed to suppress the activation of microglia (Ransohoff & Cardona, 2010). 289 However, direct evidence for the function of this module in physiological context has 290 not been provided. In this study, we demonstrated that SIRPα is a key molecule for the 291 suppression of microglia activation in vivo. Molecules other than CD47, such as 292 surfactant protein-A and -D (SP-A, -D), have also been reported as SIRPα ligands 293 (Matozaki et al., 2009). However, we found that CD47 KO mice exhibited the same 294 phenotype as SIRPα KO mice suggesting that the CD47-SIRPα interaction is indeed 295 important for the suppression of microglia activation in the brain. In addition, 296 microglia-specific SIRPα cKO mice also exhibited the same phenotype. Therefore, 297 SIRPα expressed on microglia directly control microglia activation and the induction of 298 CD11c + microglia in vivo. The emergence of CD11c + microglia in the brains of 299 microglia-specific SIRPα cKO mice establishes that CD11c + microglia are derived from 300 resident microglia in the brain, and not recruited monocytes that are spared in the  The characteristics of CD11c + microglia in our mutant mice were similar to that of 303 "primed microglia" that have been observed in aging and neurodegenerative diseases 304 (Holtman et al., 2015); thus, SIRPα is a possible key component regulating microglia 305 priming in vivo. 306 We found that the cell-surface expression of SIRPα was significantly 307 increased in CD47 KO mice. It is likely that the binding of CD47 destabilises SIRPα 308 molecules on microglia in the brain of WT mice, because our previous study suggested 309 that the interaction between CD47 and SIRPα induced endocytosis of the CD47-SIRPα 310 complex in CHO cells (Kusakari et al., 2008). It remains unclear which cell type 311 expressed CD47 that contributed to the suppression of CD11c + microglia. It was through interactions with SIRPα (Ransohoff & Cardona, 2010). An in vitro study 314 suggested that CD47 on myelin sheets interacted with SIRPα and suppressed microglial 315 phagocytosis (Gitik et al., 2011). Another study reported that direct cell-cell contact 316 between astrocytes and microglia suppressed the expression of CD11c on microglia 317 has not been reported in the brain. In addition, our data suggested that the cell surface 332 expression of SIRPα was not decreased in microglia from aged mice. Thus, the 333 age-dependent dysfunction of CD47-SIRPα signals was unlikely. The increase of 334 CD11c + microglia in aged brains may be explained by the age-dependent accumulation 335 of tissue damage that stimulates microglia activation beyond the suppressive ability of 336

SIRPα. 337
We demonstrated that CD11c + microglia were specifically increased in the 338 white matter of mutant mice. Thus, the white matter may contain an endogenous factor 339 that promotes the induction of CD11c + microglia in mutant mice. One potential candidate is myelin. As reported (Poliani et al., 2015), CD11c + microglia were increased 341 in control SIRPα +/+ mice after Cpz treatment. Thus, myelin damage effectively induces 342 CD11c + microglia even in the control mice. This suggests that myelin degradation 343 products during homeostatic turnover of the myelin structure might stimulate the our present data support the model whereby the lack of CD47-SIRPα signal upregulates destruction of myelin structures during the normal turnover process, and thereby 367 triggers the induction of CD11c + microglia without damage. 368 Transcriptome analyses showed that the expression of TNF-α, a proinflammatory 369 cytokine, was increased 2-3-fold in the brain monocytes isolated from CD47 KO mice. 370 In addition, pathway analysis showed other genes related to TNF and NF-κ B signalling 371 pathway, including RelA and IKKB (Ikbkb), were specifically increased in the mutant 372 brain cells. Thus, it is likely that proinflammatory TNF axis is strengthened in microglia 373 by the lack of CD47-SIRPα signal. In contrast, Wnt signal pathway, which contributes 374 to the proinflammatory transformation of microglia (Halleskog et al., 2011), was 375 attenuated in the CD47 KO brain cells. The transcriptome analysis also showed that 376 expression of TGF-β, an anti-inflammatory cytokine, and brain-derived neurotrophic 377 factor (BDNF), a neuroprotective factor, were increased in the CD47 KO brain cells. For myelin staining, frozen brain sections were mounted on a glass slide first, and 512 then stained with a Black-Gold II myelin staining kit (Merck Millipore) according to the 513 manufacturer's protocol. Images were acquired with a light microscope DM IRBE 514 (Leica) equipped with a cooled CCD camera (Penguin 600CL; Pixera Corp., Santa 515 Clara, CA). 516

Preparation of microglia and flow cytometry 517
Mononuclear cells including microglia were isolated by the method described by Sierra Cells were first gated on their forward (FSC) and side (SSC) scatter properties to 541 discriminate putative monocytes from other events, and then were gated on CD45 and 542 CD11b. The CD11b + /CD45 dim/lo fraction obtained was analysed as microglia. All data 543 were analysed with FlowJo 8.8.4 software (Tree Star Inc., Ashland, OR). 544

Quantitative PCR analysis 545
Total RNA was extracted from the whole brain, spinal cord, or dissected optic nerve

Microarray analysis 572
Total RNAs were prepared from the white matter (optic nerve and optic tract) or brain 573 mononuclear cells of WT or CD47 KO mice as described above. For the white matter, the brain mononuclear cells, RNAs from seven (WT; 10-15 wks of age) and six (KO; 577 12-16 wks of age) male animals were pooled and analysed. Microarray analyses were 578 performed by the Dragon Genomics Center of Takara Bio (Otsu, Japan). The quality of 579 Data are presented as the means ± SEM and were analysed by the Student's t-test. A P 631 value < 0.05 was considered statistically significant. 632

Data availability 633
The microarray data have been deposited to the Gene Expression Omnibus database 634 (https://www.ncbi.nlm.nih.gov/geo/) (accession numbers: GSE118804 and GSE118805 635 for the white matter and the brain mononuclear cells, respectively). 636

Acknowledgments 637
We thank E. Urano and T. Maegawa for technical assistance. This work was supported 638 by a Grant-in-Aid for Scientific Research on Innovative Areas ("Brain Environment"), a 639 Immunofluorescence staining of coronal brain sections prepared from control (WT) or 907 CD47 KO mice at 19 wks of age with antibodies to Iba1 (red) and CD11c (green). 908 cKO (SIRPα fl/fl :Cx3cr1-CreER T2 ) mice at 11 wks of age were stained as in Figure 3A. 950 The percentage of CD11b + /CD45 dim/lo /SIRPα + or /SIRPαmicroglia among total 951      Genes (probe sets) commonly changed in both of the white matter and the brain mononuclear cells of CD47 KO mice ( > 2-fold: |Log2 ratio| > 1) were listed. Expression of total 139 genes were changed in both of the white matter and the brain mononuclear cells of CD47 KO mice. 32 genes (34 probe sets) and 55 genes (58 probe sets) were markedly (> 2-fold) increased and decreased, respectively, in the CD47 KO mice.