IGF1R expression by adult oligodendrocytes is not required in the steady-state but supports neuroinflammation

In the central nervous system (CNS), insulin-like growth factor 1 (IGF-1) regulates myelination by oligodendrocyte (ODC) precursor cells and shows anti-apoptotic prop-erties in neuronal cells in different in vitro and in vivo systems. Previous work also suggests that IGF-1 protects ODCs from cell death and enhances remyelination in models of toxin-induced and autoimmune demyelination. However, since evidence remains controversial, the therapeutic potential of IGF-1 in demyelinating CNS

In the current study, we assessed the role of IGF-1 in mature ODCs by using a mouse model which overcomes the limitations of IGF-1 application or developmental ablation of IGF1R. We limited genetic IGF1R ablation to mature myelin oligodendrocyte glycoprotein (MOG)-expressing ODCs and showed that signaling through the IGF1R does not regulate survival of ODCs or myelin status during aging, toxin-induced demyelination, and neuroinflammation. The absence of IGF1R from ODCs led to unexpected protection from the induction of EAE and an intrinsically ameliorated disease. Altogether, our results suggest cell-specific beneficial effects of reduced IGF-1 signaling in the context of neuroinflammation.
2 | RESULTS 2.1 | Deletion of IGF1R from mature ODCs does not affect ODC number and myelin status First, to understand whether ODCs express IGF1R in the homeostatic human brain and during MS, we analyzed mRNA levels in an existing single-nucleus RNA sequencing dataset obtained from post-mortem human brains of patients without and with MS (Jäkel et al., 2019). We observed expression of the IGF1R in all ODCs and OPCs of healthy and inflamed tissue ( Figure 1a). Further, increased expression was found in the committed oligodendrocyte (COPs) cluster of chronic inactive lesions and the oligo6 cluster of remyelinating lesions, suggesting a key function of IGF1R in ODC development (Jäkel et al., 2019). Next, to functionally study the role of IGF-1 signaling in mature ODCs, we moved to a mouse model. We crossed a strain carrying a loxP-flanked 3rd exon of the IGF1R gene (Klöting et al., 2008), coding for the ligand-binding moiety, to the MOGi-cre strain (Hovelmeyer et al., 2005) displaying Cre activity in post-mitotic mature ODCs; we hereafter call these animals oIGF1R À/À mice ( Figure 1b). In contrast to previously published models using Olig1and PLP-cre (Zeger et al., 2007), the deletion of IGF1R in oIGF1R À/À mice did not affect developmental myelination as the promoter for MOG commences expression in ODCs at the terminal stage of myelination (Solly et al., 1996), as we previously described (Locatelli et al., 2012;Locatelli et al., 2015). By using the genotyping approach reported in Klöting et al., 2008, we observed recombination of the loxP-flanked Igf1r gene specifically in CNS and not in peripheral tissue (i.e., tail) of oIGF1R À/À mice (Figure 1c). A more detailed analysis on upper and lower spinal cord, hippocampus, cerebellum, spleen, kidney, and liver confirmed tissue specificity (data not shown). To assess the degree of MOGi-cre-driven recombination in oIGF1R À/À mice and better characterize the frequency of cre-mediated genetic recombination in mature ODCs, we crossed MOGi-cre mice to a fluorescent reporter strain (ROSA26-EYFP) (Srinivas et al., 2001). We found that in these oIGF1R EYFP reporter mice, EYFP was expressed by 98% of CC1 + ODCs in the cortex and cerebellum and 67% of CC1 + ODCs in the spinal cord ( Supplementary Figure 1a,b). To obtain information about the deletion frequency on the basis of individual ODCs, we analyzed IGF1R expression by immunofluorescence analysis of CNS tissue sections. We confirmed that, compared to mature CC1 + ODCs in Crenegative littermates, most ODCs in the oIGF1R À/À brain and spinal cord did not express IGF1R (Figure 1d,e). Together, characterization of Cre-induced recombination in oIGF1R À/À mice indicated an efficient deletion of IGF1R from mature ODCs. We next assessed whether the absence of IGF-1 signaling in ODCs leads to changes in weight, basic micro-anatomy, and motor function. Analysis of the CNS of oIGF1R À/À mice by H&E staining, LFB-PAS staining, and through Iba1-specific antibodies failed to reveal structural and cellular anomalies or signs of gliosis (Supplementary Figure 1c). Furthermore, oIGF1R À/À mice did not show any difference in body weight compared to controls (Supplementary Figure 1d). We used a Rotarod test for assessing motor function and did not observe any impediments until 15 months of age (Supplementary Figure 1e).
As IGF-1 signaling is known to regulate myelination in ODC progenitors (Wrigley et al., 2017), we next assessed the overall myelin content of the CNS of oIGF1R À/À and control mice at 3, 5, and 9 months of age by immunoblotting. Western blot analysis did not reveal any significant differences in protein levels of MOG and myelin basic protein (MBP) . To exclude the possibility of regional changes of myelin content through IGF1R ablation, we performed a detailed histological analysis of PLP expression. Again we did not find any difference in the myelination of white matter tracts of oIGF1R À/À mice compared to controls, with myelin sheaths appearing uncompromised up to at least 15 months of age   Figure 2d). Accordingly, and in contrast to reported developmental ablation of IGF1R along ODC maturation (Zeger et al., 2007), ODC density was unchanged in the CNS of oIGF1R À/À mice compared to controls (Figure 2e and Supplementary Figure 2e).
Next, we investigated whether the density of NG2 + OPCs would also remain unaltered in oIGF1R À/À mice. Interestingly, while the number of NG2 + OPCs in tissue sections was comparable to littermate controls in most CNS regions, we detected an increased number of NG2 + cells specifically in the brain stem of 9 months old oIGF1R À/À mice (Figure 2f,g), thus indicating that prolonged absence of IGF1R can have local indirect effects on the OPC lineage during aging. Analysis of axonal structures by NF staining and electron microscopy did not show differences between oIGF1R À/À and controls (Supplementary Figure 2f,g and data not shown). Taken together, absence of IGF1R signaling in mature ODCs did not lead to major CNS abnormalities in young and aged mice.

| IGF1R signaling does not affect ODC survival during toxic demyelination
The copper chelator cuprizone is a widely used agent for investigating non-inflammatory sterile demyelination and subsequent remyelination. It leads to mitochondrial impairment in ODC (Benetti et al., 2010;Werner et al., 2010) and subsequent transient depletion of ODCs mainly from the corpus callosum (Zatta et al., 2005). Given the described myelo-protective role of IGF-1 in the cuprizone model Mason et al., 2000), we tested the sensitivity of IGF1R-negative ODC toward toxin-induced degeneration. Feeding of oIGF1R À/À and control mice with 0.2% cuprizone led to both comparable ODC depletion and subsequent repopulation in the two groups ( Figure 3a). We were also unable to find differences in animal analyzed the density of OPCs in the corpus callosum and observed similar accumulation of NG2 + cells in the mutant animals following cuprizone intoxication (Figure 3e,f). Also, analysis of the axonal status by NF-and Smi32-specific staining did not reveal differences between the genotypes (Figure 3g,h). In summary, the absence of IGF1R from mature ODCs during cuprizone-induced demyelination, hence mitochondrial stress, did not have an impact on the survival of ODCs nor affect the timing of remyelination following toxin removal.
2.3 | Diminished neuroinflammation in oIGF1R À/À mice compared to controls The cellular and molecular mechanisms leading to demyelination in neuroinflammation differ substantially from cuprizone-intoxication and include oxidative stress, excitotoxicity, CD8 + T cell cytotoxicity, and stimulation of death receptors (McTigue & Tripathi, 2008). However, the role of IGF-1-mediated signaling in protecting ODCs during neuroinflammation has remained controversial (Bilbao et al., 2014;Cannella et al., 2000;DiToro et al., 2020;Genoud et al., 2005;Lovett-Racke et al., 1998). We hence addressed whether IGF1R signaling was involved in the response of mature ODCs to autoimmune CNS inflammation and induced EAE by immunizing oIGF1R À/À mice and oIGF1R +/+ controls against the MOG 35-55 peptide. Surprisingly, we consistently observed less severe development of clinical neuroinflammation by ablation of IGF1R. Disease incidence (91.3% in control oIGF1R +/+ mice vs. 69.2% in oIGF1R À/À ) and clinical severity both significantly decreased (Figure 4a To investigate whether priming of the immune response against the MOG peptide differed between oIGF1R À/À and control animals, we assessed the number of antigen-specific T cells in mice immunized with MOG 35-55 . We performed an in vitro restimulation with the peptide and found a comparable response for all genotypes (Supplementary Figure 4c). Since a decreased number of regulatory Foxp3 + T (Treg) cells was reported to lead to enhanced EAE severity (Koutrolos et al., 2014) we compared Treg cell numbers in both genotypes. Accordingly, the numbers of Treg cells were unaffected by absence of IGF1R on ODCs in the induction phase of EAE in lymph nodes (Supplementary Figure 4d). To investigate the underlying transcriptomic differences as a result of the absence of IGF1R from ODCs, we performed microarray analysis of spinal cord tissue of immunized mice 9 days p.i. (pre-clinical phase). However, this analysis revealed no relevant and significant differences between oIGF1R À/À and control mice (data not shown).
In conclusion, the absence of IGF1R from mature ODCs decreased the clinical severity and incidence of EAE without directly affecting the T cell response.

| IGF1R does not control ODC number or apoptosis during EAE
To unmask the distinct role of IGF1R from overlying effects of the observed severity of EAE in individual mice, we performed a detailed histological analysis in time-and clinical score-matched oIGF1R À/À and control animals. We did not observe any relevant histopathological differences after H&E and LFB-PAS stainings and by ultrastructural analysis (data not shown). Notably, ODC density in the different experimental groups remained unchanged at peak EAE and after 1 month of disease ( Figure 4c). Similarly, the level of myelin proteins such as MBP and MOG was comparable in oIGF1R À/À mice compared to control mice ( Figure 4d). To investigate whether IGF1R plays an anti-apoptotic role in ODCs during EAE, we then analyzed the number of apoptotic CC1 + cells in the acutely inflamed spinal cord through staining for activated Caspase 3 (Casp3) (Supplementary Figure 5a). Both, the densities of Casp3 + CC1 + cells and the total number of Casp3 + cells appeared comparable in oIGF1R À/À mice compared to controls (Figure 4e,f), thus showing that the absence of IGF1R from ODCs did not have a general influence on ODC apoptosis during EAE. We next analyzed whether recruitment of progenitors of ODCs was affected in oIGF1R À/À and found a comparable accumulation of NG2 + OPCs in the inflamed F I G U R E 1 Expression of IGF1R in human and mouse ODCs, and receptor deletion in oIGF1R À/À mice. (a) Expression of IGF1R by the indicated cell types from MS patients and healthy individuals. Single cell RNAseq data as published by Jäkel et al. (Jäkel et al., 2019). Ctrl, nuclei from healthy individuals; from MS patients: NAWM, normal-appearing white matter; A, active; CA, chronic active; CI, chronic inactive; RM, remyelinated. (b) Mice carrying an IGF1R gene with loxP-flanked 3rd exon (IGF1R f/f , left) were crossed to mice expressing the Cre recombinase specifically in ODCs (MOGi-cre, center). The resulting offspring (MOGi-cre Tg/WT /IGF1R f/f = oIGF1R À/À ) features recombination of the IGF1R locus specifically in MOG-expressing, mature ODCs. (c) DNA was isolated from CNS and tails of oIGF1R À/À and control animals. PCR with primers spanning exon 3 of the IGF1R gene reveals bands of 1100 bp (WT gene) in IGF1R +/+ control mice and bands of 1350 bp (IGF1R f allele) and 570 bp (IGF1R Δ allele/recombined locus) in oIGF1R À/À mice. Recombination of the IGF1R f allele was specific for the CNS. (d) Representative picture of spinal cord white matter sections from oIGF1R À/À and control animals immunostained with the ODC-specific CC1 and with IGF1Rspecific antibodies. CC1 is shown in red, IGF1R in green. Dashed white lines indicate CC1 + ODCs highlighting the IGF1R negativity of ODCs in oIGF1R À/À mice. Scale bar, 30 μm. (e) Quantification of IGF1R-expressing CC1 positive cells in brain and spinal cord isolated from oIGF1R À/À and control mice, following manual blinded quantification of immunofluorescent staining for pIGF1R and CC1, for brain n = 3, for spinal cord n = 6. Two tailed student's T test, *** = p < .001. A, active; CA, chronic active; CI, chronic inactive; CNS, central nervous system; MOG, myelin oligodendrocyte glycoprotein; MS, multiple sclerosis; NAWM, normal-appearing white matter; ODC, oligodendrocyte; RM, remyelinated spinal cords of oIGF1R À/À and control mice (Supplementary any obvious difference between groups (data not shown). Taken F I G U R E 2 Absence of IGF1R from mature ODC does not lead to structural or cellular abnormalities in the CNS. (a) CNS sections from oIGF1R À/À and control mice of 1 month of age were stained with a PLP-specific antibody (displayed in red). Scale bar, 200 μm. (b) Relative intensity of PLP-specific staining from sections treated as in (a). Shown is mean ± s.e.m., n = 5. (c) Representative electron micrographs of ODCs and myelinated axons from the cerebellum of 1 month old and (d) 12 months old oIGF1R À/À and control mice. (e) Sections from indicated CNS areas oIGF1R À/À and control mice at 3 or 12 months of age were immunostained with an ODC-detecting, ASPA-specific antibody. Cells were manually counted in a blinded fashion; shown is the relative ASPA + cell density in oIGF1R À/À compared to control mice. The bar graph depicts the mean density ± s.e.m., n = 5. (f) Representative sections of the brain stem of 9 months old oIGF1R À/À and control mice were immunostained with the OPC-detecting NG2-specific antibody. Scale bar, 50 μm. (g) CNS sections from 9 months old oIGF1R À/À and control mice were immunostained with NG2-specific antibodies and cells manually counted (field size: 0.25 mm [Floyd et al., 2007]). The bar graph shows relative cell density ± s.e.m., n = 4. *p < .05, Student's T test. ASPA, aspartoacylase; CNS, central nervous system; IGF1R, insulin-like growth factor receptor 1; ODC, oligodendrocyte; PLP, proteolipid protein F I G U R E 3 ODC loss and demyelination following cuprizone intoxication. (a) oIGF1R À/À and control mice were fed with 0.2% cuprizone and sacrificed at days 4, 9, 21, and 35 after treatment start. Sagittal sections of the corpus callosum were immunostained for ASPA and cells counted manually and blinded. The values are relative to control mice which did not receive cuprizone. The bar graph shows mean density ± s.e.m., n = 5. (b and c) Mice were fed with 0.2% cuprizone and sacrificed after 3 or 5 weeks. Brain lysates were immunoblotted and analyzed using MOG-(b) and PLPspecific (c) antibodies. Values are relative to control mice, which did not receive cuprizone in the diet. The bar graph shows mean intensity ± s.e.m., n = 4. (d) oIGF1R À/À and control mice were sacrificed 3 and 5 weeks after cuprizone feeding and sagittal sections of the corpus callosum were stained with MOG-specific antibodies. As controls served mice which did not receive cuprizone.
Scale bar, 200 μm. (e) Mice were treated as in (a). Sagittal sections of the corpus callosum were immunostained for NG2 and cells counted manually in a blinded fashion (field size: 0.12 mm [Floyd et al., 2007]). Bar graph shows mean ± s.e.m., n = 5. *p < .05, two tailed Student's T test. (f) Representative sections of the corpus callosum of oIGF1R À/À and control mice 3 and 5 weeks after beginning of cuprizone treatment, immunostained with NG2-specific antibody. Scale bar, 200 μm. (g) Mice were treated as in (a). Sagittal sections of the corpus callosum were immunostained with the neuronal cytoskeleton marker NF200-and non-phosphorylated neurofilament Smi32-specific antibodies and the staining intensity quantified. The bar graph shows mean ± s.e.m., n = 4.

| Absence of IGF1R from ODCs does not impact inflammation upon EAE development
To dissect the underlying mechanisms of amelioration of EAE development in the absence of IGF1R signaling, we analyzed recruitment of peripheral immune cells into the CNS and the local glia response to inflammation. Histological analysis revealed comparable lesion size between oIGF1R À/À and control mice (Supplementary Figure 5d). By flow cytometry, we observed that the total number of invading lymphocytes in the CNS was not different between clinical score-matched oIGF1R À/À and controls, both at the peak of the disease and during the chronic phase ( Figure 5a).  Figure 5f).
F I G U R E 4 oIGF1R À/À mice show ameliorated disease but no change in ODC density. (a) oIGF1R À/À and control animals were induced with EAE and scored daily for clinical signs of disease. Shown is clinical score of disease and a table reporting incidence, day of clinical onset and average maximum clinical score. The data was pooled from 6 independent EAE experiments (total of 52 oIGF1R À/À and 46 control mice). (b) Area under the curve (AUC) calculation of disease severity in the mice shown in (a), shown is average ± s.e.m. **p < .001, Mann Whitney U test.
(c) oIGF1R À/À and control mice were induced with EAE and sacrificed at peak disease and 1 month (Chronic EAE) after induction of EAE. CNS sections were immunostained with ASPA-specific antibody and cells manually counted in a blinded fashion (field size 0.22 mm [Floyd et al., 2007]). The data is representative of two independent experiments. The bar graph shows mean ± s.e.m., n = 9. (d) Brain lysates of oIGF1R À/À and control mice 1 month after induction of EAE were immunoblotted and analyzed using MOG-and MBP-specific antibodies. The bar graph shows mean staining intensity ± s.e.m., n = 5. (e) Spinal cord sections of oIGF1R À/À and control mice 2 days after clinical onset of EAE were stained with DAPI, CC1 and activated caspase 3-specific antibodies. The number of CC1 + Casp3 + cells or. (f) Casp3 + cells, manually counted in a blinded fashion (field size 0.25 mm [Floyd et al., 2007]), is shown. The data is representative of two independent experiments. The bar graph shows the mean number ± s.e.m., n = 7. (g) NeuN + cells manually counted (field size 0.25 mm [Floyd et al., 2007]  ( Moore et al., 2020). Furthermore, experimental ODC ablation (Locatelli et al., 2012), ODC-specific peroxisome impairment, and PLP overexpression (Ip et al., 2006;Kassmann et al., 2007) were invariably followed by neuronal degeneration. Notably, in the two latter genetically-modified models, ODC impairment also led to the recruitment of lymphocytes to the CNS (Ip et al., 2006;Kassmann et al., 2007), thus indicating that ODCs participate in the complex interplay between the CNS and the immune system (Kerschensteiner et al., 2009;Zeis & Schaeren-Wiemers, 2008).
The key role of ODCs in CNS homeostasis becomes particularly relevant in MS and is well-represented in demyelinating animal models such as EAE (Locatelli et al., 2012;Lucchinetti et al., 1999;McTigue & Tripathi, 2008) and cuprizone intoxication (Benetti et al., 2010;Zatta et al., 2005). In EAE, ODC death and demyelination within CNS inflammatory lesions are central to the clinical progression of disease (Lucchinetti et al., 1999). ODCs in MS patients also show altered heterogeneity that may suggest susceptibility to demyelination and epigenetic differences in genes such as BCL2L2 and NDRG1 potentially decreasing anti-apoptotic mechanisms even in non-affected CNS regions (Huyhn et al., 2014;Jäkel et al., 2019). However, finding suitable therapeutic strategies protecting ODCs and myelin during neuroinflammation and other CNS diseases is a relatively untapped research area. The IGF-1 pathway has been among those protective strategies assessed in recent years in MS and its model EAE, but results remained sadly inconsistent among studies (Bilbao et al., 2014;Cannella et al., 2000;DiToro et al., 2020;Genoud et al., 2005;Lanzillo et al., 2011;Li et al., 1998;Liu et al., 1995;Lovett-Racke et al., 1998;Shahbazi et al., 2017;Wilczak et al., 2008;Yao et al., 1995). To finally settle the specific question about the role of the IGF-1 pathway during CNS inflammation we decided for genetically ablating its receptor from mature ODC. In this genetic model, we observed that the presence of IGF1R appeared dispensable for physiological maintenance of myelin and cell survival of mature ODCs in adult animals up to 15 months of age, even though IGF-1 is known to deliver a robust anti-apoptotic signal to OPCs during development , While at first sight this result appears to differ from observation in an IGF1R ablation model using PLP-cre, in which the authors showed up to 25% decrease in ODC numbers in the corpus callosum at 25 weeks of age (Zeger et al., 2007), this work also presented IGF1R deletion in a fraction of NG2 + OPCs. It hence appears possible that PLP-cre induced recombination partially affected CNS cell-types other than mature ODCs (Michalski et al., 2011 F I G U R E 5 Inflammatory cell phenotype during anti-CNS autoimmunity in absence of oligodendroglial IGF1R. (a) Cells from the CNS of oIGF1R À/À and control animals 2 days (acute) or 12 days (chronic) after induction onset of EAE were analyzed by flow cytometry. They were stained with CD4-, CD8-, CD11b-, CD45-specific antibodies. Shown is the relative number of CD4 + and CD8 + T cells in the CD11b negative , CD45 + population in oIGF1R À/À compared to control mice. The data is representative of two independent experiments, shown is the mean ± s.e. m. (n = 8). (b) Bar graph showing percentage of pro-inflammatory cytokine producing T cells in the CD4 + , CD11b negative , CD45 + population isolated from the CNS of oIGF1R À/À and control mice induced with EAE and analyzed at peak disease. The data is representative of two independent experiments; shown is the mean ± s.e.m, n = 8. (c) Cells were isolated from the spinal cord and cerebellum of oIGF1R À/À and control animals 2 days after disease onset (day 12 post immunization) and analyzed by flow cytometry after staining with CD4-, CD11b-, CD45-, Foxp3-specific antibodies. Shown is the total number of Foxp3 + , CD4 + , CD11b negative , CD45 + cells per CNS (n = 4). (d) EAE was induced in oIGF1R À/À and control animals and animals sacrificed at clinical peak of disease. Spinal cord sections were stained with DAPI, MHC-II, and IB4-specific antibody and positive cells manually and blindly counted. Shown is the mean ± s.e.m., n = 4. (e) oIGF1R À/À and control animals were induced with EAE. CNS cells isolated at peak disease and immunostained for CD45, CD11b were analyzed by flow cytometry. Shown are relative percentages per CNS of CD45 intermediate -CD11b + microglia and of CD45 high -CD11b + activated microglia/macrophages in oIGF1R À/À compared to oIGF1R +/+ controls. The data is representative of three independent experiments. The bar graph shows the mean ± s.e.m., n = 10. (f) Cells were isolated at peak disease from the CNS of MOG-immunized oIGF1R À/À and control animals. The cells were immunostained for CD45, CD11b, CD11c, CD95l, MHC-II, CD95, and CD44 and analyzed by flow cytometry. Shown are relative percentages. The data is representative of two independent experiments. The bar graph shows mean ± s.e.m., n = 8. *p < .05, Mann Whitney U test. (g) Spinal cord sections from EAE-induced oIGF1R À/À animals 2 days after disease onset were immunostained with DAPI, Isolectin B4, and with MHC-II-, activated IGF1R-specific antibodies. Scale bar, 60 μm. (h) Microglial cells were seeded in 96-well E-plates at a density of 15,000 cells/well and monitored with a real-time impedance-based xCELLigence system. After 24 h, the cells were treated with IGF-1 (500 ng/ml) in the absence of serum in the medium (as indicated with the arrow [Fernandez & Torres-Aleman, 2012]). A short time after each treatment, the cell index drastically drops for 30-60 min due to the media change and temperature difference. The time point of treatment or media change is marked as "0 h" in the graph (n = 3). Subsequently, microglia was activated with recombinant mouse (rm)TNF-α 100 ng/ml and rmIFN-γ 50 U/ml in the presence or absence of IGF-1500 ng/ml (as indicated with the arrow [Floyd et al., 2007]). Results are given as mean values ± S.D. (***p < .001 vs. non-treated microglia, ANOVA, Scheffé's test, n = 3). ASPA, aspartoacylase; CNS, central nervous system; EAE, experimental autoimmune encephalomyelitis; IGF1R, insulin-like growth factor receptor 1; ODC, oligodendrocyte; MOG, myelin oligodendrocyte glycoprotein (Benarroch, 2009), anoxia (Khwaja & Volpe, 2008), oxidative stress (French et al., 2009), and IFN-γ signaling (Lin et al., 2007) show stagespecific survival potential for the ODC lineage.
Even though the absence of IGF1R did not affect ODC survival in our model, it consistently lowered disease presentation upon induction of anti-myelin autoimmunity, reducing the incidence of the disease and the severity of affected animals. We show that this protection was not mediated by reduction of the encephalitogenic T cell response. It rather seems to be an inherent property of the IGF1R-deficient ODC compartment. However, when we analyzed mRNA expression in the spinal cord of pre-clinical animals, we did not find any significant difference that could provide an explanation or support further hypothesis-driven exploration. Therefore, the actual mechanism behind EAE amelioration in oIGF1R À/À mice remains at this point unclear.
We  (Funfschilling et al., 2012;Lee et al., 2012). This hypothesis is supported by previous observations: IGF1R signaling is a regulator of insulin-like anabolism in the adult CNS in a context-dependent manner (Cheng et al., 1998;Fernandez & Torres-Aleman, 2012). Lower metabolic rates in IGF1R KO ODCs might protect these cells, as was found for human ODCs in vitro (Rone et al., 2016). Interestingly, this counterintuitive detrimental role of IGF1R signaling might also explain observations in Alzheimer's disease (Cohen et al., 2009) and ischemic models (Endres et al., 2007). In both studies, reduced IGF-1 signaling improves survival of animals and decreases neuronal loss and brain infarct size, respectively.
In summary, the described protection against EAE in absence of IGF1R signaling by mature ODCs highlights the importance of this cell type in CNS homeostasis and in their dynamic interplay with the immune system. Our results also stress the complexity of IGF-1 mediated pathways and indicate a detrimental role of IGF1R signaling in mature ODCs during neuroinflammation.

| Microglia cultures
Cortex was removed from 1-3 days-old C57Bl/6 N pups, dissociated for 15 min in 1 mg/ml trypsin and 2 min in 1 mg/mL trypsin inhibitor. difference (Diemert et al., 2012), followed by a total recovery to values before media change or drug treatment.

| Histology
Mice were euthanized with CO 2 and perfused with PBS. For cryostat sections the tissue was fixed overnight with 4% paraformaldehyde (PFA), cryoprotected in 30% sucrose and frozen at À80 C. Frozen tissue was cut sagittally in 40 μm thick sections and stained as described before (Locatelli et al., 2012) with antibodies or antisera against fol- 4.11 | RNA array oIGF1R À/À and control female animals were induced with EAE and sacrificed after 9 days. RNA was isolated from spinal cord tissue and assessed by microarray analysis (n = 5). Total RNA was labeled and hybridized to Affymetrix GeneChip ® Mouse Gene 1.0 ST arrays according to the manufacturer's instructions. We used the oligo package to import and process the scanned arrays into R (Carvalho & Irizarry, 2010). Further, we assessed the quality of scanned arrays with array Quality Metrics (Kauffmann et al., 2009). "Robust multichip average" was then used for background correction, normalization and to control for technical variation between arrays within the study (Irizarry et al., 2003). For annotation, we used both the original Affymetrix probe set and an updated probe grouping provided by Brainarray (Dai et al., 2005). Only probe sets that map to a unique gene were considered for further analysis. Then, a linear model was fitted to each of the remaining probe set's expression data, and the estimated coefficients given the set of contrasts were computed using Limma (Smyth, 2004).
Due to high variability in disease progression and small sample size, the list of differentially expressed genes was reported without adjusting for multiple testing and should be interpreted with caution.

| Electron microscopy
Electron microscopy analyses were performed as previously described (Locatelli et al., 2012). Briefly, animals were sacrificed and transcardially perfused with saline followed by a fixative containing 4% paraformaldehyde (Serva, Heidelberg, Germany) and 2% glutaraldehyde (Serva) in PBS. After post-fixation for 24 hours in the same fixative, the tissue was rinsed and cut into sections of 50 to 60 μm using a vibrating micro-

| Code availability
The code to reproduce the single-cell RNA-Sequencing figure can be found under: https://github.com/rsankowski/locatelli_ms_celltype_ comparison.

CONFLICT OF INTEREST
The authors declare no competing interests.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.