Overcoming the inhibitory microenvironment surrounding oligodendrocyte progenitor cells following demyelination

Chronic demyelination in the human CNS is characterized by an inhibitory microenvironment that impairs recruitment and differentiation of oligodendrocyte progenitor cells (OPCs) leading to failed remyelination and axonal atrophy. By network-based transcriptomics, we identified sulfatase 2 (Sulf2) mRNA in activated human primary OPCs. Sulf2, an extracellular endosulfatase, modulates the signaling microenvironment by editing the pattern of sulfation on heparan sulfate proteoglycans. We found that Sulf2 was increased in demyelinating lesions in multiple sclerosis and was actively secreted by human OPCs. In experimental demyelination, elevated OPC Sulf1/2 expression directly impaired progenitor recruitment and subsequent generation of oligodendrocytes thereby limiting remyelination. Sulf1/2 potentiates the inhibitory microenvironment by promoting BMP and WNT signaling in OPCs. Importantly, pharmacological sulfatase inhibition using PI-88 accelerated oligodendrocyte recruitment and remyelination by blocking OPC-expressed sulfatases. Our findings define an important inhibitory role of Sulf1/2 and highlight the potential for modulation of the heparanome in the treatment of chronic demyelinating disease.


INTRODUCTION
Endogenous myelin repair, known as remyelination, represents a regenerative process that can restore lost neurological function and prevent disease progression in degenerative conditions such as multiple sclerosis (MS) (Franklin, 2015). The failure of remyelination in MS has been attributed to a failure of oligodendrocyte progenitor cell (OPC) differentiation, in part, due to the presence of quiescent OPCs in regions of chronic demyelination (Wolswijk, 1998, Kuhlmann et al., 2008. Diverse signaling pathways have been identified that prevent timely OPC differentiation in rodent models and many small molecules have been described that improve the rate of murine remyelination by targeting individual rate-limiting steps (Franklin and Ffrench-et al., 2012). Magnetic sorting of CD140a/PDGFR positive cells was performed as described (Pol et al., 2013). Human OPCs (hOPCs) were maintained on plates coated with poly-ornithine and laminin, in human neural differentiation (ND) media supplemented with 20 ng/ml PDGF-AA (PeproTech) and 5 ng/ml NT-3 (PeproTech) (Abiraman et al., 2015). Fetal tissue was fixed in 4% PFA and cryoprotected in a sucrose gradient (7.5% sucrose overnight, followed by 15% sucrose overnight), and frozen in OCT medium (Tissue-Tek). Serial 16 µm transverse sections were cut using a Leica cryostat and stored at -80°C.

RNA extraction and RT-PCR analysis
Total RNA was isolated from hOPCs following mitogen withdrawal and real time RT-PCR performed as described in (Pol

Oligodendrocyte differentiation and immunocytochemistry
For knockdown studies, hOPCs were seeded at 5×10 4 cells/mL and transduced with the indicated lentivirus after 24 hours. To initiate differentiation, PDGF-AA and NT-3 growth factors were removed, and cells cultured in ND media supplemented with human BMP7 (PeproTech), human WNT3a (R&D Systems) and/or PI-88 (gift of Medigen Biotechnology Corp, Taiwan), as designated, with media replenished after 48 hours. Four days following growth factor removal, cells were live stained with O4 IgM hybridoma supernatant (gift of Dr. James Goldman, Columbia University), fixed in 4% paraformaldehyde, and immunostained. For assessment of SULF2 expression, hOPCs were maintained as progenitors, fixed and stained for SULF2 (Abcam ab113405, 1:500). Where indicated, Brefeldin A (Cell Signaling Technology) was added to hOPCs at 5 µg/ml for five hours prior to fixation. Alexa 488, 594, and 647-conjugated secondary antibodies (Invitrogen) were used at 1:500 dilutions. Differentiation was quantified as the proportion of stained cells in 10 random fields at 10X magnification (using an Olympus IX70 microscope), representative of over 1,000 cells total in each condition.

Assessment of hOPC HS sulfation by flow cytometry
hOPCs were cultured as previously described, trypsinized and collected in 0.02% EDTA/PBS. Cells were resuspended in HS3A8V (RB4CD12) phage display antibody (1:10, gift of Toin van Kuppevelt, Nijmegen Medical Center, Nijmegen, Netherlands).The His-tagged phage display antibody was detected by immunostaining with 6X-His Tag primary antibody (1:700, Abcam) and 1:500 Alexa-488 conjugated secondary antibody. All antibody incubations were for 30 minutes at +4°C. Flow cytometry was performed using a BD Fortessa flow cytometer. Dead cells were excluded by forward and side scatter-based gating, and doublet discrimination was performed. Fluorescence intensity (FITC-A) was quantified from ~10,000 cells in each replicate, and data was normalized to peak fluorescence to facilitate presentation.

Cell and secreted protein isolation and western blot
For cellular and secreted SULF2 expression analysis, hOPCs were cultured for 24 h in ND media supplemented with 20 ng/ml PDGF-AA and 5 ng/ml NT-3. For oligodendrocyte SULF2 expression, PDGF-AA and NT-3 was removed to allow oligodendrocyte differentiation for 3 days.
IR680RD and IR800CW secondary antibodies (Li-Cor) were diluted 1:5000 in Rockland blocking buffer. Blots were imaged using the Li-Cor Odyssey Infrared Imaging system.

Luciferase Assays
OPCs were seeded at a density of 2.5 x 10 4 cells/ml and maintained as progenitors in ND media supplemented with PDGF-AA and NT-3. One day post-seed, cells were infected with lentiviral luciferase reporter constructs for 24 hours, after which media was changed to fresh ND media supplemented with growth factors, BMP7, WNT and/or PI-88, as indicated. Twenty hours post-treatment, luminescence response was quantified using the Promega Bright-Glo reagent and a Bio-Tek plate reader, in accordance with manufacturer recommendations. Background luminescence was subtracted from all measurements and luminescence readings were normalized to the average value of the untreated control in each experiment. Data are presented as the mean ± SEM per individual human sample (three technical replicates).

Animals and surgery
All experiments were performed according to protocols approved by the University at

Animal tissue processing and analysis
Animals were euthanized at 3, 5, 7, or 14 days post-lesion (dpl) by transcardial perfusion of saline followed by 4% paraformaldehyde under deep anesthesia. Tissue processing and lesion identification was performed as described previously (Welliver et al., 2018). Slides immediately adjacent to the lesion centers, identified by solochrome cyanine staining, were used for all immunohistochemical procedures. Primary antibodies utilized were rabbit anti-Olig2 (1:500, Millipore), mouse anti-CC1 (1:50, Millipore), mouse anti-GFAP (1:300, Sigma), rabbit anti-Iba1 (1:300, Wako Chemicals USA), cleaved caspase-3 (Asp175) (1:250, Cell signaling), and rabbit anti-NG2 (1:200, Millipore). Alexa 488, 594, and 647-conjugated secondary antibodies (Invitrogen) were used at 1:500. In situ hybridization for Plp1 was performed as described in (Welliver et al., 2018). Higher magnification examples of CC1/Olig2 staining demonstrating positive cell criteria used for quantification are shown in Figure S9. Images of spinal cords were captured at 20X magnification by wide field epifluorescence using an Olympus IX51 with Prior stage and were stitched together (FIJI). For each marker, the average from at least two sections was quantified for each animal. Lesions with cross-sectional area < 10,000 µm 2 were excluded from the analysis, as were lesions which extended into surrounding grey matter. All quantification was performed by an investigator blinded to the identity of the samples.

Electron microscopy
For assessment of remyelination, tissue was processed as described in Welliver et al.
(2018). Briefly, mice were sacrificed at 14 day post lesion by transcardial perfusion with 2% glutaraldehyde in 0.1 M phosphate buffer and spinal cords were extracted (n = 5 per group), 1-mm-thick blocks surrounding the spinal cord lesion were processed through osmium tetroxide, dehydrated through ascending ethanol washes, and embedded in TAAB resin (TAAB Laboratories). One μm sections were cut, stained with 1% toluidine blue (Sigma-Aldrich), and examined by light microscopy to identify lesions. Selected blocks with lesions were trimmed, ultrathin sections cut, and examined by electron microscopy (Hitachi, HT7800). Images were acquired at 2500X magnification. Analyses of remyelinated axons and g-ratios were performed blinded. For remyelination counts, a minimum of 800 axons were counted for each animal from at least six different fields, with all animals per treatment group. Analysis of g-ratio was performed as described in Dillenburg et al. (2018). Briefly, axon and fiber diameter were measured using FIJI (diameter = 2 × √[area/π]). A minimum of 100 axons was analyzed per animal. The frequency distribution of axon diameter and g-ratio of remyelinated axons was calculated by binning on axon diameter and g-ratio.

Human multiple sclerosis tissue
Multiple sclerosis tissue was prepared as described in Postnatal day 7 and 28, and 24-week old mice were euthanized and perfused with 0.9% saline followed by 4% paraformaldehyde. Dissected mice brains and spinal cords and cryopreserved in sucrose gradient (10, 20, 30%) for 24hr and were snap frozen in OCT and sectioned coronally at 16 µm thickness. Fixed sections were baked at 60°C for 1 hr, washed with ethanol, followed by tissue pretreatment, probe hybridization following RNAscope fluorescence multiplex assay and sections were counterstained with DAPI to visualize nuclei. Positive signals were identified as punctate dots present in nucleus and cytoplasm. All tissues were tested using negative control probes (ACD Bio) to control for non-specific binding ( Figure S2a). For Sulf1 and Sulf2 probes, the expression patterns were verified by comparison with known sulfatase-specific patterns of neuronal expression in cortical layer V (Sulf2) (Zeisel et al., 2015) and layer VI (Sulf1)(Allen Brain Atlas) ( Figure S2b-f). For all RNAscope-based imaging, confocal Z stack images with optical thickness 0.2 µm were taken and stacked images are shown (Zeiss LSM 510 Meta confocal).

Statistical analyses
All quantification and data analysis were performed by an investigator blinded to the identity of the samples. All statistical analyses were performed using GraphPad Prism (San Diego, CA

Heparan sulfate proteoglycan 6-O endosulfatases are highly expressed by OPCs
Transcriptomic network analysis of human OPC (hOPC) differentiation identified a module of highly correlated and species conserved genes associated with progenitor fate (Pol et al., 2017). Among these high connected genes, heparan sulfate endosulfatase 2 (SULF2) was identified in the OPC-associated module (M5). Real time RT-PCR analysis confirmed profound down-regulation during hOPC in vitro differentiation (Fig. 1a). Expression of SULF2 was relatively restricted to OPCs as SULF2 mRNA expression was >14-fold higher in primary human PDGFαR + OPCs than CD133 + neural progenitor cells (Wang et al., 2013). Likewise, SULF2 expression was maintained in oligodendrocyte-biased PDGFαR + O4 + progenitors (Abiraman et al., 2015) and adult human A2B5 + OPCs but not enriched in adult human astrocyte or microglial isolates (Sim et al.,

2009).
As heparan sulfate function has not been extensively studied during OPC development, we examined the expression of genes involved in HS biosynthesis, post-synthetic modifications of HS, and HSPG core proteins in mouse development (Fig. S1). We observed that several HSassociated genes were down-regulated with differentiation. Down-regulation of HS biosynthetic genes during OPC differentiation was consistent with previous studies that suggested reduced heparan/HS abundance during OPC differentiation (Stringer et al., 1999, Properzi et al., 2008. Together, these findings suggest that the heparanome is dynamically regulated in the progenitor state and becomes more static as OPCs differentiate to mature OLs. Remarkably, both mouse Sulf1 and Sulf2 mRNAs were almost entirely restricted to the oligodendrocyte lineage (Table S1).
We observed high expression of SULF2 mRNA in vivo restricted to a subset of PDGFRA + OPCs located in the human subventricular zone and white matter of 22-week fetal brain (Fig. 1d).
In 1-week old mouse brain, almost all Pdgfra + OPCs expressed Sulf2 mRNA ( Fig. 1e and Fig.   S2b) while Sulf1 mRNA was restricted to a subset of OPCs ( Fig. 1h and Fig. S2b).During early postnatal development both sulfatases were also detected in immature oligodendrocytes but expression in oligodendrocytes was not sustained in the adult ( Fig. 1f-j). Consistent with previous reports (Zeisel et al., 2015)(Allen Brain Atlas), Sulf1 and Sulf2 mRNAs were highly expressed within specific cortical layers (Fig. S2c-f). Sulf2 mRNA was also expressed by a small subset of Gfap + astrocytes and Iba1 + microglial cells (Fig. S2g-j). In the corpus callosum of aged adult mice (24-week), only Pdgfra + OPCs retained high levels of Sulf1/Sulf2 mRNA ( Fig. 1g/j). Combined with our expression data in human cell isolates, these findings suggest that OPCs expressed Sulf1/Sulf2 throughout developmental and into adulthood consistent with a functional role in OPC homeostasis.

Sulfatase inhibits oligodendrocyte production following demyelination
We examined the pattern of sulfatase expression following demyelination by inducing demyelination in adult mouse spinal cord by focal injection of lysolecithin. We observed increased expression of Sulf1 and Sulf2 mRNA within the lesion at 5 days post-lesion (5 dpl) ( Fig. 2a and   b). At this time-point, OPCs are actively recruited into the lesion. We observed that many Pdgfra + OPCs within the lesion expressed Sulf1 or Sulf2 mRNAs ( Fig. 2a-i), while a subset expressed both Sulf1 and Sulf2 (Fig. 2e,j). As in normal brain, Sulf2 expression was restricted to OPCs in the unlesioned spinal cord white matter. In contrast, following demyelination Sulf2 was also expressed by a very small subset of Gfap + astrocytes and Iba1 + microglia in the vicinity of and within the lesion while only very few Sulf1 + astrocytes were observed (Fig. S3). Of note, Pdgfrb + pericytes did not express detectable Sulf1 or Sulf2 following demyelination. These observations suggest that both sulfatases are up regulated by OPCs following demyelination and that sulfatase expression was largely restricted to OPCs and oligodendrocyte lineage cells.
To determine the functional role of OPC-expressed sulfatases following demyelination, we performed tamoxifen-induced conditional ablation of Sulf1/Sulf2 1-week prior to lysolecithininduced demyelination. In order to ablate Sulf1/2 specifically from adult OPCs, we crossed NG2creER mice (Zhu et al., 2011) with floxed Sulf1/Sulf2 mice (Ai et al., 2007). Littermate controls lacking cre-expression were also injected with tamoxifen. At 5 dpl, we observed a significant increase in Olig2 + oligodendrocyte lineage cell density in Sulf1/2 cKO mice compared to wildtype (wt) (n = 3-6 animals per group, t-test p < 0.01) ( Fig. 2k-l, o). To assess OPC proliferation animals were terminally injected with EdU. Increased Olig2 + cell density was not due to increased proliferation (EdU%, p = 0.85) (Fig. 2p). This suggested that ablation of sulfatase encourages a favorable microenvironment for OPC migration and recruitment.
To more directly define progenitor recruitment following demyelination, we assessed the density of Pdgfra mRNA-defined OPCs and immature oligodendrocytes at each stage following demyelination and determined the rate of progenitor proliferation by co-labeling Pdgfra + OPCs with Ki67 (Fig. S5). As shown previously (Sim et al., 2002), in wildtype animals Pdgfra + OPC density was greatest at 5 dpl and gradually declined thereafter (Fig. S5d). Following Sulf1/2 cKO, we observed significantly increased Pdgfra + OPC density at both 5 and 7dpl compared to matched wt controls (Holm-Sidak multiple comparison test, n = 4-6 animals, p < 0.01) (Fig. S5d). By 14dpl, the Sulf1/2-dependent increase in OPC density was no longer present (n= 3-4 per group, p =0.64).
The proportion of dividing Ki67 + OPCs likewise decreased over the same time period. Consistent with the EdU analysis above (Fig. 2P), Sulf1/2 cKO did not influence the proportion of Ki67 + Pdgfra + OPCs at any stage following demyelination suggesting that proliferation of OPCs was not effected by Sulf1/2 ablation (Ki67%, p = 0.99) (Fig. S5e). Next to determine, whether differential cell death contributed to the differences observed in oligodendrocyte lineage recruitment, we analyzed apoptosis via cleaved caspase 3 within the Olig2 + cell population ( Fig.   S5f-h). The differences in proportion of apoptotic OPCs between Sulf1/2 cKO and wt mice did not reach significance (n = 3-4, p = 0.28)( Fig. S5h) and, as such, likely do not account for the large differences in Olig2 + cell density observed ( Fig. 2o and q). Thus, sulfatases Sulf1/2 act to impair recruitment and early differentiation of OPCs following demyelination.

Individual sulfatase deletion is sufficient to improve oligodendrocyte generation following demyelination
In order to define whether the effects of Sulf1/2 cKO was mediated specifically via deletion of either Sulf1 or Sulf2, we assessed the effect of individual Sulf1/2 cKO on OPC recruitment and differentiation following demyelination at 7 dpl (n = 5-6 animals per genotype) ( Fig.3a-g).
Together, these data suggest that Sulf1 and Sulf2 both contribute to delayed OPC recruitment and oligodendrocyte differentiation following demyelination. As deletion of either sulfatase resulted in substantial alterations in the density of oligodendrocytes and oligodendrocyte lineage cells, it is clear that the loss of either Sulf1 or Sulf2 was not fully compensated by the remaining sulfatase gene.

SULF2 expression inhibits hOPC differentiation and is enriched in MS lesions
Having established that Sulf2 can impair OPC recruitment and differentiation in mice following demyelination, we next studied the functional role of sulfatase expression in purified human OPCs. RNA-seq and qPCR revealed that SULF2 was the principal sulfatase expressed by hOPCs (Table S1 and  Importantly, SULF2 knockdown significantly accelerated human O4 + oligodendroglial differentiation by 46% compared to scrambled controls ( Fig. 3i-j) (n = 6 samples, p < 0.0001). To determine if hOPCs actively secreted SULF2, we performed slot blot analysis of conditioned media as well as cell lysates (Fig. S4d). Consistent with SULF2 mRNA downregulation upon differentiation ( Fig. 1a), both cellular and secreted SULF2 protein expression decreased during hOPC differentiation.
As SULF2 acted to prevent efficient hOPC differentiation in vitro, we next asked whether SULF2 was expressed by OPCs in adult human brain and in chronic active demyelinated lesions from secondary progressive MS patients (n = 2). SULF2 mRNA was expressed by PDGFRA + OPCs in normal appearing white matter and substantially increased around the demyelinated lesion border (Fig. 3l-p). We confirmed that these SULF2 mRNA transcripts were restricted to OPC and oligodendrocyte cells by their colocalization with PDGFRA ( Fig. 3m-n) and PLP1 transcripts, respectively (Fig 3o-p ). Almost all PDGFRA + and PLP1 + cells observed at the lesion border expressed SULF2 mRNA. Together, these data suggest that SULF2 is up regulated in a similar manner following demyelination in the human brain and in vitro act in a similar manner to prevent efficient OPC differentiation.

k, l)
as seen at 7 dpl. Likewise, increased mature oligodendrocyte density in Sulf1/2 cKO animals was accompanied by increased total density of Olig2 + oligodendrocyte lineage cells relative to wt controls (p< 0.01) (Fig. 4l). The density of Olig2 + CC1defined cells was equivalent between groups at 14 dpl ( Fig. 4l) as was the density of Pdgfra + OPCs (Fig. S5d). The percentage of Olig2 + CC1 + oligodendrocytes within the Olig2 + population was no longer enhanced at 14 dpl (p = 0.72) (Fig. 4m). Taken together, Sulf1/2 cKO resulted in accelerated OPC recruitment and differentiation of new oligodendrocytes leading to increased density of both mature oligodendrocytes and OPCs following demyelination in the CNS.
As OPC secreted sulfatases may regulate local signaling in other cell types, we examined the effect of OPC-specific deletion of Sulf1/2 on astrocyte and microglial activation following demyelination. We did not observe differences in the gross pattern or intensity of astrocytic Gfap immunoreactivity (p = 0.66) (Fig. 4g, h). Likewise, the overall distribution and intensity Iba1 + microglia/macrophages staining was not significantly altered by Sulf1/2 cKO (p = 0.26) ( Fig.4i-j,   n). Thus, while paracrine effects of OPC-expressed Sulf1/2 cannot be excluded, there were no overt effects on the gliotic response following demyelination.
We next investigated whether decreased sulfatase activity in OPCs might accelerate remyelination ( Fig. 4o-v). At 14 dpl, Sulf1/2 cKO significantly increased the proportion of oligodendrocyte remyelinated axons relative to WT controls (n = 4 animals per group, t-test p = 0.01) (Fig. 4s). While a small proportion of axons were remyelinated by Schwann cells, we did not observe any changes in the proportion of Schwann cell myelinated axons following Sulf1/2 cKO (data not shown). Consistent with an acceleration of oligodendrocyte differentiation, myelin thickness was significantly increased in remyelinated axons following Sulf1/2 deletion. The myelin g-ratio, which represents the ratio of axon to axon plus myelin diameter, was significantly decreased in Sulf1/2 cKO (g-ratio calculated from n = 4 animals, t-test p < 0.05). Linear regression analysis of individual g-ratio vs. axon diameter also demonstrated a significant effect of Sulf1/2 cKO (Fig. 4t). Likewise, the distribution of g-ratios demonstrated a significant reduction in the frequency of very thinly remyelinated axons consistent with improved remyelination initiation following Sulf1/2 cKO (Two-way ANOVA, Sidak post hoc, 0.85-0.9 p < 0.0001 and 0.9-0.95 p<0.01) (Fig. 4u). Sulf1/2 cKO did not influence the distribution of axonal diameter within the lesion suggesting that deletion of Sulf1/2 does not induce axonal swelling (Fig. 4v). These results indicate that OPC-expressed sulfatases act as negative regulators for remyelination in a cellautonomous manner.
As modulation of these pathways could occur at multiple levels, we next asked whether sulfatase modulated WNT and BMP signaling directly in human OPCs. To this end, scrambled control or SULF2 knockdown (KD) hOPCs were transduced with viral reporters for WNT and BMP signaling (see methods). WNT-dependent TCF reporter luciferase was >11-fold induced following WNT3a treatment (n = 4 fetal samples) (Fig. 5n). Strikingly, SULF2 KD significantly attenuated WNT induction by more than 40%. Likewise, BMP response element (BRE)-dependent luciferase was increased >18-fold by BMP7 treatment and this BMP-induced BRE luciferase activity was significantly inhibited by SULF2 KD (Fig. 5o). Two-way ANOVA revealed a significant interaction effect between SULF2 KD and both WNT and BMP signaling (p < 0.05) and was consistent with a direct effect of sulfatase on these signaling pathways. To further test whether SULF2 KD could directly influence the effects of BMP signaling on hOPC fate, we treated SULF2 KD and scrambled infected hOPCs with BMP7 and assessed the effects on O4 + oligodendrocyte differentiation (n = 5-6 fetal samples) (Fig. S6). As shown previously (Sim et al., 2011), BMP7 treatment inhibited O4 + differentiation from hOPCs (Two-way ANOVA, p < 0.001) (Fig. 5p). Importantly, SULF2 KD significantly rescued the effects of BMP7 (Hold-Sidak test, p < 0.01). As such, sulfatase expressed by hOPCs directly potentiates both WNT and BMP signaling and is consistent with sulfatases exerting cell-autonomous role in OPCs.

PI-88 modulates OPCs sulfation and regulates BMP and WNT signaling in hOPCs
As sulfatases act in the extracellular space, they represent a pharmacologically relevant target for manipulation in demyelinating disease. PI-88, a highly sulfated structural mimetic of heparan sulfate acts as a non-cleavable substrate and inhibitor for sulfatases (Parish et al., 1999, Yu et al., 2002 and is currently in clinical trials for cancer therapy. We first assessed basal HS sulfation on rat CG4 and human primary OPCs by flow cytometry using RB4CD12 antibody that recognizes highly sulfated HS GAG motifs (Jenniskens et al., 2000) ( Fig. 6a, b). Treatment with PI-88 caused a progressive and persistent enrichment of HS sulfation at 30-min and 24 hours (Fig. 6b).
We next examined whether PI-88 via induced increased HS sulfation could inhibit WNT and BMP signaling in OPCs. In rat CG4 cells, activated β-catenin, a downstream effector of canonical WNT signaling was substantially induced following WNT3a treatment (Fig. 6c).
Importantly, PI-88 treatment effectively blocked the effect of WNT3a. Likewise, PI-88 attenuated BMP7-induced Smad 1/5 phosphorylation (pSmad) (Fig. 6d). Using luciferase-expressing lentiviral reporters of WNT and BMP signaling, we found PI-88 exhibited a clear dose-dependent effect and antagonized WNT and BMP signaling with IC50s of 0.38 μg/ml and 0.89 μg/ml for WNT3a and BMP7 respectively ( Fig. 6e-f). Interestingly, when we compared the effect of PI-88 treatment on low and high dose WNT/BMP-induced luciferase we observed a downward shift of the dose-response curve following PI-88 treatment (data not shown). This suggests that the effect of PI-88 is noncompetitive and is consistent with inhibition of sulfatase rather than direct agonist binding or another receptor-mediated mechanism.
Next, we explored the effects of PI-88 on human primary OPC signaling. Lentiviral reporter infected hOPCs were treated with BMP/WNT ligands and/or PI-88. BMP and WNT ligand treatment (50 ng/ml) induced robust >15-fold increase in reporter activity (two-way ANOVA, p < 0.05) that was significantly attenuated by PI-88 ( Fig. 6g-h). PI-88 treatment alone did not alter BMP or WNT-dependent luciferase activity. Thus, we concluded that PI-88 was capable of inhibiting BMP and WNT signaling in both rat and hOPCs.
Linear regression analysis of axon diameter vs. g-ratios demonstrated a significant reduction of g-ratio following PI-88 treatment (Fig. 7u). Likewise, the distribution of g-ratios binned by axon diameter was left shifted and indicated increased myelin sheath thickness in PI-88 treated mice (Fig 7v). Similar to Sulf1/2 cKO, PI-88 treatment did not induce axonal swelling confirmed by axon diameter frequency distribution (Fig 7w). Together, these data demonstrate that PI-88 treatment promoted both OPC differentiation and accelerated remyelination following demyelination.
Finally, to determine if PI-88 might similarly regulate WNT and BMP signaling and exert a pro-differentiative effect via WNT/BMP antagonism, we examined the expression of WNT and BMP target genes following Sulf1/2 cKO and/or PI-88 treatment using in situ hybridization ( Fig.   8e-l). Similar to our results using pharmacological manipulation of WNT and BMP pathways, we observed significantly decreased expression of Apcdd1 (WNT target) and Id4 (BMP target) in Pdgfra + OPCs in Sulf1/2 cKO mice (two-way ANOVA, Sulf1/2 cKO main effect p < 0.01, n = 4-6 animals per group) (Fig 8p-q). Likewise, PI-88 treatment alone significantly inhibited expression of both genes in Pdgfra + OPCs (two-way ANOVA, PI-88 effect p < 0.05). Combined treatment of PI-88 in Sulf1/2 cKO mice did not significantly decrease WNT or BMP target gene expression any further compared to Sulf1/2 cKO alone (Holm-Sidak, p > 0.5). These observations suggest that PI-88 acts via OPC-expressed Sulf1/2 to influence BMP and WNT signaling in OPCs and thereby promote myelin repair.

DISCUSSION
The inhibitory tissue environment of chronic demyelinated lesions acts to prevent efficient myelin repair and remyelination in MS (Lau et al., 2013). The cellular environment and local extracellular matrix determine the availability and signaling of multiple growth factor and cytokine pathways. Here, we show that sulfation of heparan sulfate proteoglycans (HSPG) critically influences the paracrine signaling environment surrounding OPCs and can be modulated to improve recruitment and formation of new oligodendrocytes leading to accelerated myelin repair.
These data demonstrate that extracellular sulfatases expressed by human and mouse OPCs promote inhibitory WNT and BMP signaling following demyelination and that ablation of sulfatase function either by conditional genetic deletion or pharmacological inhibition can enhance remyelination.
The mammalian genome contains two sulfatase genes which share substrate specificity but differ in expression pattern in a cell-and tissue-specific manner (Ai et al., 2006, Morimoto-Tomita et al., 2002. We found that OPCs express high levels of sulfatases that are eliminated during oligodendrocyte differentiation. During mouse CNS development, both sulfatases were expressed by OPCs and immature oligodendrocytes. In contrast, human OPCs express SULF2 more than 10 4 -fold higher than SULF1. The significance of this species-difference in expression is unclear but has significant implications for the design of sulfatase-specific small molecules. In contrast, the similarities in terms of resultant phenotype following sulfatase deletion suggest an important conservation of function between human and mouse signaling. In the adult mouse CNS, Sulf1/2 expression is largely restricted to OPCs and a subpopulation of cortical layer V neurons, as previously described (Zeisel et al., 2015). Neuronal Sulf2 expression suggests that the heparanome within cortical demyelinated lesions differs from subcortical white matter and may potentiate inhibitory signals that contribute to failed remyelination. Intriguingly, sulfatase expression remained predominantly restricted to OPCs following demyelination in both mouse and human lesions. While the specific HSPG core proteins that undergo sulfatase-mediated editing are unknown, syndecan-3 (SDC3) is noteworthy due to its high expression in fetal and adult human OPCs (Sim et al., 2009, Sim et al., 2011 and in cultured murine cells (Winkler et al., 2002).
The role of heparanome sulfation has not been previously studied in the context of demyelination. Constitutive sulfatase expression by OPCs suggests these enzymes prevent complete 6-O sulfation and our functional studies indicate that increased HSPG sulfation is associated with inhibition of various extracellular signaling cascades. The resting state of OPC heparanome is predominately composed of highly sulfated HexA2S-GlcNS6S trisulfated disaccharide units which are specific among glial subtypes (Stringer et al., 1999). As such, sulfatases are ideally situated to modulate OPC signaling. The generation of highly sulfated HSPGs is catalyzed by sulfotransferases. These enzymes, expressed at high levels by OPCs, are localized to the Golgi apparatus and do not act once HSPGs are presented on the cell surface (Habuchi et al., 2004). As such, the increased prevalence of surface HS sulfation following sulfatase inhibition is most likely due to de novo HSPG presentation and suggests a rapid turnover of OPC-expressed HSPGs.
In the adult CNS, sulfatase deletion in OPCs results in a substantial increase in recruitment and differentiation of oligodendrocyte lineage cells following demyelination leading to improved remyelination. Interestingly, antisense Sulf1 treatment during embryonic development results in reduced migration (Kakinuma et al., 2004). Similarly, we observe that SULF KD in hOPCs results in perturbed migration in response to PDGF-AA (data not shown). As HSPG sulfation is expected to influence multiple pathways, we observed distinct effects on OPC dynamics following demyelination. In addition to accelerated differentiation (assessed by density and proportion of post-mitotic oligodendrocytes), we noted that sulfatase deletion resulted in progressive recruitment of Olig2 + cells such that the density of oligodendrocyte lineage cells progressively increased from 5-14 days post lesion. This contrasts with wild-type lesions which typically reach maximal density at 5 days. The mechanisms by which overall recruitment of OPC and subsequent generation of oligodendrocytes remain unclear and are likely complex due to the number of signaling pathways that may be influenced by sulfatase activity. Regardless, the effect of dual sulfatase deletion on oligodendrocyte number was striking as we observed a more 2-fold increase in oligodendrocyte number. Consistent with shared substrate specificity, the effect of individual sulfatase deletion was consistent with a model in which sulfatases act in a dose-dependent manner. As such, once a lower limit of sulfatase activity was reached, oligodendrocyte recruitment and differentiation were increased. The dose-dependent effects of sulfatase inhibition in vitro is consistent with this model and supports the potential for future therapeutic intervention.
Importantly, the effect of conditional deletion of Sulf1/2 in NG2-expressing cells was restricted to OPCs as we did not observe pericyte Sulf1/2 expression either in normal CNS or following demyelination with similar results observed in purified human OPCs. As pericytes did not express Sulf1/2, conditional deletion of sulfatase in NG2-expressing cells resulted in specific deletion of OPC-expressed sulfatase activity.
HSPG sulfation is known to regulate several signaling pathways (Rosen and Lemjabbar-Alaoui, 2010). Here, we show both WNT and BMP pathways are inhibited by sulfatase inhibition in OPCs. Canonical WNT signaling prevents efficient oligodendrocyte differentiation following demyelination (Fancy et al., 2009, Fancy et al., 2011. WNT target genes are activated in OPCs following human white matter injury and correlate with failed differentiation and repair (Fancy et al., 2014). We found that endogenous sulfatases promote WNT signaling and sulfatase inhibition decreases WNT target gene expression in OPCs. This likely occurs in a ligand-receptor dependent manner as TCF/LEF transcriptional activity was dependent on SULF2 expression in purified human OPCs.
We observed a highly significant interaction between sulfatase function and pharmacological manipulation of WNT signaling. CHIR-99021 acts as a GSK-3β inhibitor to potentiate canonical WNT signaling. As Sulf1/2 cKO was observed to completely block the effect of the GSK-3β antagonist, these data suggest that canonical WNT signaling acts to inhibit oligodendrocyte differentiation and remyelination via secondary pathways which are themselves dependent on and modulated by Sulf1/2. In purified rat OPCs, the inhibitory effect of WNT3a treatment on differentiation occurs in a BMP-dependent manner (Feigenson et al., 2011). In contrast, recombinant human WNT3a treatment of purified hOPCs while effectively activating TCF/LEF transcription did not alter basal oligodendrocyte differentiation (data not shown) suggesting the need for engagement of additional pathways to elicit the inhibitory WNT effect.
Together with our data, these observations suggest that WNT may be acting via and dependent on BMP and other extracellular signaling following demyelination. In addition to these paracrine mechanisms, it is likely that sulfatases also directly modulate WNT signaling. Supporting a direct role of sulfatase manipulation on WNT signaling in demyelination, we observed the lack of an additive effect of XAV939 treatment and sulfatase deletion. In addition, we found that WNT target gene expression is reduced following Sulf1/2 cKO and that WNT-dependent transcriptional activity in hOPC was similarly reduced following SULF2 KD. While WNT pathways is therefore likely a key target of sulfatases following demyelination, WNT antagonist treatment did not entirely phenocopy Sulf1/2 cKO, as treatment with XAV939 did not increase overall Olig2 density whereas Sulf1/2 cKO resulted in a significantly increased density.
Our data support a clear role for sulfatase mediated potentiation of BMP signaling following demyelination. Genetic sulfatase deletion in OPCs resulted in impaired BMP-response element-dependent transcription and reduced Id4 expression by OPCs in vivo. BMP signaling is up regulated following demyelination (Ara et al., 2008, Sabo et al., 2011 and is active in MS lesions (Deininger et al., 1995). We found that BMP receptor antagonist increased both Olig2 recruitment and oligodendrocyte differentiation, as shown (Govier-Cole et al., 2019). Sulfatase deletion in the context of BMP blockade did not further increase differentiation suggesting this process is dependent on BMP signaling. In contrast, intracellular activation of BMP signaling by treatment with A01, a Smurf1 E3 ligase specific antagonist, could not be effectively compensated for by sulfatase deletion. Given the intracellular mode of action for A01, this is entirely consistent with a model in which HSPG sulfation regulations BMP ligand/receptor accessibility.
In addition to WNT and BMP, modulation of HSPG sulfation by sulfatases regulates several additional signaling cascades that influence the demyelinated lesion microenvironment.
Previously, Sulf activity has been to shown to potentiate PDGFαR signaling in glioblastoma (Phillips et al., 2012). PDGF is the principle OPC mitogen during development (Sim et al., 2011, Sim et al., 2006, Noble et al., 1988, Raff et al., 1988 but does not appear to be rate-limiting following murine spinal cord demyelination (Calver et al., 1998, Woodruff et al., 2004. In addition to PDGF, previous studies suggest that sulfatase activity inhibits FGF2 signaling (Otsuki et al., 2010, Wang et al., 2004, Lamanna et al., 2006, Holst et al., 2007, Seffouh et al., 2013 and pharmacologically reduced sulfation blocks FGF responsiveness in OPCs (Bansal et al., 1996, Fortin et al., 2005. As FGFR ablation in oligodendrocyte lineage cells results in hypomyelination (Furusho et al., 2012), inhibition of sulfatase leading to increased FGF signaling is consistent with our observed acceleration of remyelination and increased myelin sheath thickness following Sulf1/2 cKO. In addition to PDGF and FGF2, heparan sulfatase regulation of inflammatory signaling cascades such as IFN-γ have been described (Lortat-Jacob et al., 1991). As such, HSPG 6-O sulfation via regulation by OPC-expressed sulfatases provide the basis for the coordinated regulation of the lesion microenvironment.
Given that sulfatases are secreted into the extracellular milieu, it is likely that paracrine effects of sulfatase inhibition may influence other glial and inflammatory cell signaling following demyelination. Although a broad assessment of microglial number via Iba1 immunoreactivity and astrogliosis via Gfap did not indicate a major effect following sulfatase deletion, we cannot exclude the possibility that paracrine effects may influence the infiltration, proliferation, or activation of microglia and other immune cells within the lesion environment. Importantly, while lysolecithininduced demyelination provides the ideal model to assess the effects of sulfatase deletion on the cellular dynamics of remyelination, paracrine influences of sulfatase inhibition cannot be ruled out due to the relative absence of adaptive immune system related signaling in this model.
PI-88 is a heparin mimetic acting as a competitive antagonist to sulfatases (Parish et al., 1999). The effects of PI-88 treatment on OPC dynamics following lysolecithin demyelination essentially phenocopied that of OPC-specific sulfatase deletion. Importantly, when PI-88 treatment was combined with Sulf1/2 cKO, there were no additive effects and the PI-88 did not further influence OPC dynamics or differentiation. This is consistent with the effects of PI-88 being Taken together, OPC-expressed sulfatases by regulating their local heparanome, potentiate inhibitory signals present in demyelinating lesions that prevent efficient myelin repair.