Transforming growth factor‐beta renders ageing microglia inhibitory to oligodendrocyte generation by CNS progenitors

Abstract It is now well‐established that the macrophage and microglial response to CNS demyelination influences remyelination by removing myelin debris and secreting a variety of signaling molecules that influence the behaviour of oligodendrocyte progenitor cells (OPCs). Previous studies have shown that changes in microglia contribute to the age‐related decline in the efficiency of remyelination. In this study, we show that microglia increase their expression of the proteoglycan NG2 with age, and that this is associated with an altered micro‐niche generated by aged, but not young, microglia that can divert the differentiation OPCs from oligodendrocytes into astrocytes in vitro. We further show that these changes in ageing microglia are generated by exposure to high levels of TGFβ. Thus, our findings suggest that the rising levels of circulating TGFβ known to occur with ageing contribute to the age‐related decline in remyelination by impairing the ability of microglia to promote oligodendrocyte differentiation from OPCs, and therefore could be a potential therapeutic target to promote remyelination.


| INTRODUCTION
Remyelination is an important regenerative process in the CNS in which new myelin sheaths are restored to demyelinated axons by oligodendrocytes which are newly generated by adult oligodendrocyte progenitor cells (OPCs) (Franklin and ffrench-Constant, 2017). This results in the restoration of saltatory conduction, which is lost during demyelination, and protection from irreversible axonal atrophy (Irvine & Blakemore, 2008;Smith, Blakemore, & Mcdonald, 1979). Remyelination is controlled by a complex interplay of multiple molecules and pathways that have either facilitatory or inhibitory effects on the proliferation, migration and differentiation of adult OPCs (Franklin and ffrench-Constant, 2017). One of the main rate limiting steps in remyelination is the efficiency of OPC differentiation. This is especially evident in multiple sclerosis (MS), a chronic demyelinating disease often of several decades' duration: chronic demyelinated lesions frequently contain oligodendrocyte lineage cells that have failed to undergo full differentiation into myelin sheath-forming cells (Chang, Tourtellotte, Rudick, & Trapp, 2002;Kuhlmann et al., 2008;Wolswijk, 1998).
The molecular mediators of remyelination are produced by a variety of cell types present within demyelinating lesions. These include microglia, the resident immune cells of the CNS, which are now recognized to play a major role in CNS remyelination (Lloyd & Miron, 2016;Rawji, Mishra, & Yong, 2016). Microglia affect OPC differentiation rate through the clearance of myelin debris (which is inhibitory to OPC differentiation) (Kotter, Li, Zhao, & Franklin, 2006;Natrajan et al., 2015) and by secreting pro-regenerative factors (Miron & Franklin, 2014;Miron et al., 2013).
Previous studies have identified the extracellular matrix (ECM) as a critical factor regulating OPC differentiation during remyelination (Lau et al., 2012). For example, chondroitin sulfate proteoglycans (CSPGs), which are deposited by different cell types in demyelination lesions, are inhibitory to OPC differentiation and remyelination (Keough et al., 2016). Since microglia cells compose a large component of the cells present within demyelinating lesions, it is conceivable that they too contribute to the ECM composition of a lesion. To test this, we have developed a protocol for capturing the ECM environment associated with microglia (the micro-niche). Moreover, since TGFβ levels increase with ageing and have been shown to negatively affect hippocampal neurogenesis in the CNS (Buckwalter et al., 2006;Carlson et al., 2009;Doyle, Cekanaviciute, Mamer, & Buckwalter, 2010), we hypothesized that TGFβ-mediated effects on the microglial micro-niche might contribute to the age-related decline in the efficiency of remyelination (Cantuti-Castelvetri et al., 2018;Miron et al., 2013;Ruckh et al., 2012;Shields, Gilson, Blakemore, & Franklin, 1999;Sim, Zhao, Penderis, & Franklin, 2002).

| Animal husbandry
Only Sprague Dawley rats were used for this study. Rats were bred and raised at the Animal Facility at the University of Cambridge. All animals were fed standard diet and were kept under 12 hr cycles of light and darkness. All animals used for experiments were sacrificed by schedule 1 approved methods, according to the requirements and regulations set by the United Kingdom Home Office.

| Induction of white matter demyelination
In order to induce demyelination, female Sprague-Dawley rats (Harlan Laboratories) of 18 months of age were used. The rats were anesthetized with buprenorphine (0.03 mg/kg, s.c.) and 2.5% isoflurane. Demyelination was induced by stereotaxic injection of 4 μL of 0.01% ethidium bromide (EB) into the caudal cerebellar peduncles (CCPs), as previously described (Woodruff & Franklin, 1999). EB was delivered at a rate of 1 μL/min. After EB delivery the injection needle remained in position for additional 4 min.

| Isolation and culture of aged and young microglia
Neonatal (P0-P20), young adult (P30-P90), and aged (18-24 months) rats were decapitated after lethal injection of phenobarbital. Brains were removed shortly after death (including cerebrum and cerebellum) and placed into ice cold HALF isolation medium (Hibernate A Low Fluorescence), supplemented with 2% of B27. Meninges, olfactory bulbs and remains of spinal cord (if present) were removed and brain tissue was chopped mechanically into 1 mm pieces using sterile scalps.
The minced tissue was spun down in 100 g for 2 min in RT. Brain tis- Solution, no magnesium/calcium; Gibco). The tissue was centrifuged for 5 min, in 250 g. Supernatant was completely aspirated and tissue was re-suspended in neutralizing solution (HALF+ 2% B27+ 2 mM sodium-pyruvate) for 5 min in RT. In order to create a single cell suspension, cells were first titurated using a 5 mL serological pipette and subsequently three fire polished glass pipettes in decreasing opening sizes. After each trituration step the tissue suspension was allowed to sediment and the supernatant, containing the cells, was transferred into a fresh tube, while being passed through a 70 μm cell strainers to ensure single cell suspension. After removing the supernatant each time, 2 mL of fresh neutralizing solution were added to the remaining tissue. After the whole tissue was passed though the cell strainer, a 90% pre-filtered (20 μn filters) isotonic After overnight recovery, media was completely removed and replaced with fresh OPCM supplemented with either TGFβ (Peprotech) or TGFβ inhibitor (SB-431542, Tocris). Cells were cultured for 48 hr, afterwards media was removed, filtered using 22 μm filters to remove any cell debris and was frozen immediately using dry ice. Conditioned media was stored in −80 C until use.

| Cells immunohistochemistry
At the end of each in vitro experiment, cells were fixed by adding 4% PFA to each well and incubating in RT for 10 min. Cells were then washed twice using PBS (5 min each wash, RT, shaking). Cell nuclei were labeled with 2 μg/mL Hoechst 33342 (Sigma). Before applying primary antibodies, cells were blocked using 5% normal donkey serum (NDS) for 20 min, RT. For intracellular staining, 0.1% Triton was added to the blocking solution. Primary antibodies (see Table 1) were diluted in PBS+ 5% NDS and samples were incubated overnight in 4 C. Following incubation, samples were washed twice with PBS (10 min, RT, shaking) and secondary antibodies were added. Secondary antibodies were diluted in PBS+ 5% PBS, and samples were incubated in RT for 2 hr. Following incubation, samples were washed 3 times with PBS, 10 min each time. Second wash included Hoechst 33342 for nuclear staining (2 ng/mL). Image acquisition was performed using a Leica-SP5 microscope (Leica) and LAS software (Leica) or a Zeiss Observer A1 inverted microscope (Zeiss) and Zeiss Axiovision software. For each well (in 48 well plate) 3-4 images were taken in randomly selected areas. Further image processing and analysis was performed using the ImageJ software package. Images quantification was done using Cell-Profiler software (Broad Institute, Harvard University). In total, for each condition, a minimum of 100 cells were counted.

| qRT-PCR
RNA was isolated from acutely purified OPCs and microglia or from cultured OPCs and microglia using Qiagen RNeasy Micro Kit, or Directzol RNA MicroPrep Kit (Zymo Research; cat#R2061). Isolated RNA was immediately frozen using dry ice and was further stored in −80 C. RNA quantities were measured using Nanodrop 2000 (Thermo Scientific).
cDNA was generated using Qiagen QuantiTect Reverse Transcription Kit according to the instructions of the manufacturer (Qiagen; 205310). For RT-qPCR, primers were acquired from KiCqStart SYBR Primers (SIGMA-ALDRICH) and used at a concentration of 300 μM (see Table 2). Primer efficiency was validated for all primers (90%-110%). cDNA, primers, and the Syber Green Master Mix (Qiagen; 204,141) were mixed as instructed by the manufacturer, and RT-qPCR and melting curve analysis were performed on Life Technologies Quantstudio 6 Flex Real-Time PCR System. Fold changes in gene expression were calculated using the delta delta Ct method in Microsoft Excel using TATA box binding protein (Tbp) as a control gene. Statistical significance was determined using two-tailed unpaired t tests assuming equal variances.

| Statistics
Statistical analysis was performed on GraphPad Prism 7 (GraphPad Software, Inc.). Immunohistochemical staining was identified using CellProfiler and CellProfiller Analyst (Carpenter et al., 2006) was used for machine learning cell identification and results were compared using suitable statistical tests. If >2 groups were present, a one-way analysis of variance (ANOVA) test was performed, followed by appropriate post test in order to compare individual groups (Dunnett). For all statistical analysis, differences between groups were considered significant at p value < .05. Graphs were produced using GraphPad Prism or RStudio.

| Effects of microglial ECM on OPCs
To investigate the potential effect of microglia ECM molecules on OPC differentiation in isolation of possible effects mediated by secreted factors, we developed a protocol that yielded preparations of microglia membrane molecules and ECM depositions on which OPCs could be cultured. We termed these preparations microglial "ghosts" (Figure 1a). In brief, whole brains of young adult and aged rats were lysed and single cell suspensions generated. First, MACS (Magnetic Activated Cell Sorting) for A2B5+ cells was used to remove OPCs. Microglia were then isolated using MACS for cells expressing the microglial marker CD11B+. We established primary microglia cultures (>80% CD11B+ or IBA1+ cells) from both neonatal/young rats, as well as aged rats (Figure 1b,f). For ghost preparations, microglia were cultured in PDL coated plates for 48 hr before being lysed by ddH 2 O (see Figure 1a for schematic description). This yielded substrates comprising microglia membrane bound molecules and microglia-associated ECM, but lacking active microglia (the method is similar to that used for astrocytes by Keough et al., [2016]).
Neonatal OPCs were then plated on the microglial "ghosts" generated from microglia. When OPCs were plated on ghosts from naïve young microglia there was no significant change in OL or astrocyte generation as shown by CNP ( Figure 1c) and GFAP (Figure 1d) staining when compared with PDL control wells. We next asked whether ghosts generated from activated microglia would affect OPCs properties. We treated microglia with TGFβ since (a) it is known to alter ECM production in other cell types (Ignotz & Massagué, 1986;Laping et al., 2002)

| TGFβ alters neonatal microglialassociated ECM
We next asked if we could identify the changes elicited in microglia by exposure to TGFβ. The effect of TGFβ was tested on young microglia in vitro by measuring mRNA levels of Cspg4, Fn1, and Cd44 (genes coding for the ECM molecules, NG2, Fibronectin, and CD44, respectively). qRT-PCR revealed that 48 hr treatment with 10-20 ng/mL (data not shown). These results are consistent with a previous study (Sugimoto et al., 2014), which showed that following stroke, rat microglia express NG2 on exposure to TGFβ.
We further verified this effect by immunostaining TGFβ1-treated young microglia for NG2 ( Figure 2b). As predicted, young microglia treated with TGFβ1 (20 ng/mL) for 48 hr in vitro showed a significant increase in the percentage of NG2+ cells (Figure 2b,c). The addition of TGFβ inhibitor SB-431542 to culture did not reduce the number of NG2+ cells, in contrast to the reduction in mRNA levels (possibly due to longer half life of the protein versus the mRNA). We therefore concluded that TGFβ induced changes in the microglia-generated ECM in a way which promoted astrocyte formation by OPCs.   control wells (Figure 4a,d,e). We then asked whether this effect could be recapitulated using aged microglia conditioned media. For this, microglia were cultured for 48 hr before retrieving the conditioned media. This conditioned media was mixed in a 1:1 ratio with fresh media. OPCs that were grown in medium conditioned by aged microglia did not show an increase in astrocyte formation (Figure 4c), supporting the inference that ECM components produced by microglia were responsible for the differentiation effects described rather than secreted factors. The lack of effect using conditioned media could also be due to technical reasons such as shorter conditioning time leading to lower concentrations of secreted molecules in the conditioned media.

| Aged microglia show ECM signature similar to TGFβ treated microglia
As BMP signaling has previously been shown to promote the generation of astrocytes by OPCs (Mabie et al., 1997), we tested whether this might be the mechanism underlying the astrocyte formation by ghosts derived from aged microglia. OPCs cultured on aged ghosts in the presence of LDN-193189, a small molecule BMP inhibitor, did not show increased astrocyte formation (Figure 4b), suggesting that the effect exerted by the microglia ghosts was via the activation of BMP signaling in OPCs.

| Aged microglia express NG2 following demyelination
We next tested the expression of NG2 by aged microglia in vivo. Focal areas of spinal cord white matter demyelination was induced in aged rats (>18 months) by the injection of ethidium bromide (EB). At 5 days post lesion-induction many cells were positive for both CD11B and NG2 within the lesion area (Figure 5b). There were also a small number of double positive cells in normal appearing white matter of aged rats (Figure 5a).

| TGFβ directly inhibits OPC differentiation
The   (Gautier et al., 2015), but also by expressing specific surface molecules that inhibit OLs from wrapping them with myelin (Redmond et al., 2016). Similarly, experiments with astrocytes show that they generate ECM that prevents OPC differentiation in vitro molecules (Keough et al., 2016). Here, we show that microglial ECM can influence cell fate decisions by OPCs, and that this, in part, depends on BMP signaling. BMP signaling relies on the phosphorylation of SMAD1/5/8 by ALK2/3 (Figure 5d). Using small molecule inhibitor (LDN-193189) which specifically targets ALK2/3, we were able to block the induction of GFAP in OPCs cultured on microglia ghosts. These results are consistent with previous studies which have shown that BMP signaling induces astrocytic fate in OPCs and other CNS progenitor cells (Gross et al., 1996;Mabie et al., 1997;Petersen et al., 2017).
Using microglia ghost preparations, we have ruled out the possible effects of microglia secreted factors, as there are no viable cells and secreted factors are removed by aspirating the cell media and replacing it with ddH 2 O. The use of conditioned media from aged microglia did not affect the behavior of the OPCs, in contrast to previous reports using conditioned media from activated microglia Yuen et al., 2013). This could be due to limited time the microglia were cultured before media was taken, as it is possible that this was not enough time for substantial concentration of factors to be released into the media in sufficient concentrations. Moreover, in previous reports the microglia were activated by potent agents (e.g., LPS) which are likely to cause significantly higher expression of secreted factors. In contrast, treatment with TGFβ, an anti-inflammatory cytokine, that we have used in this study promotes production of ECM and membrane bound molecules (Laping et al., 2002).

| TGFβ changes the microglial micro-niche
Since naïve young microglia ghosts did not significantly affect OPC differentiation (Figure 1b,c), unlike the effects shown previously by naïve astrocytes (Keough et al., 2016), we considered whether activated microglia more similar to a state likely to be present in demyelinated lesions would have a different effect. We used TGFβ to activate the microglia since it has been shown previously to be present in elevated levels in MS lesions (De Groot et al., 1999) and circulating blood of MS patients (Nicoletti et al., 1998) molecules in different cell types (Ignotz & Massagué, 1986;Laping et al., 2002;Roberts et al., 1988). Following stroke, TGFβ is linked to upregulated NG2 expression in microglia (Sugimoto et al., 2014).
Expression of NG2 in microglia has been associated previously with lower phagocytic capacities (Zhu et al., 2012). Moreover, microglia in NG2 knockout mice show increased expression of neurotrophic factors and decreased expression of pro-inflammatory markers following facial nerve axotomy (Zhu et al., 2016).
These findings, in combination with rising levels of TGFβ with ageing (Carlson et al., 2009), position TGFβ as a leading candidate for explaining the underlying mechanisms that alter microglia membrane bound proteins. Consistent with these studies, we also show that TGFβ can induce the expression of NG2 in microglia in vitro, and furthermore can induce an "aged like" membrane bound proteins signature. TGFβ exerts its effects in cells through phosphorylation of SMAD2/3 by ALK4/5/7.
Using ALK4/5/7 specific small molecule inhibitor (SB-431542) (Inman et al., 2002;Laping et al., 2002), we show that these effects are abolished and thus the increased expression of NG2 by microglia is prevented. We therefore conclude that TGFβ alters microglia in such a way that it activates BMP signaling in OPCs, driving them toward astrocyte generation. Although our primary focus has been on NG2 expression, we do not exclude the involvement of other ECM molecules in the effects described. RNA sequencing databases of aged and young microglia show increase in multiple ECM molecules in aged microglia, including fibronectin and VCAM (Hickman et al., 2013). Our identification of NG2 expression in microglia is supported by previous reports that which show that NG2 expression plays a role in microglia activation in pathological conditions (Sugimoto et al., 2014;Zhu et al., 2016), even though NG2 is generally regarded as reliable marker for OPCs (Nishiyama, Chang, & Trapp, 1999). Our data and those of others thus confirm that NG2 does not exclusively identify OPCs.

| TGFβ directly inhibits OPC differentiation
Since TGFβ was shown to alter OPC differentiation via its effects on microglia, we further tested whether TGFβ could affect OPCs directly.
Previous reports claimed that TGFβ promotes OPC differentiation (McKinnon et al., 1993;Palazuelos et al., 2014). In contrast, we find evi-  stage of development (already expressing CNP) than the cells used in our study. Thus, it is possible that TGFβ has different effects in different stages of OPC differentiation. This is supported by previous studies in which higher levels of TGFβ were associated with lower levels of MBP following focal demyelination (Kotter, Zhao, Van Rooijen, & Franklin, 2005).
Our findings reveal TGFβ as a potential therapeutic target for enhancement of remyelination in aged patients. These results support previous studies implicating the harmful effects of TGFβ in ageing, and specifically in regeneration (Buckwalter et al., 2006). Future experiments targeting TGFβ by, for example, the administration of TGFβ pathway inhibitors such as the FDA approved compound Losartan, could be undertaken to assess their effects on remyelination in aged animals Holm et al., 2011).
In summary, we report a novel role for microglial ECM in determining OPC differentiation fate. As microglia constitutes the majority of the cells in most demyelination lesions (especially during the early stages of remyelination), these findings could prove important in the design of remyelination therapies. Moreover, we identify TGFβ as a key cytokine which can exert inhibitory effects on the differentiation of OPCs, an effect that increases with ageing.