Retinal ganglion cell expression of cytokine enhances occupancy of NG2 cell-derived astrocytes at the nerve injury site: Implication for axon regeneration

Abstract

Following injury in the central nervous system, a population of astrocytes occupy the lesion site, form glial bridges and facilitate axon regeneration. These astrocytes originate primarily from resident astrocytes or NG2+ oligodendrocyte progenitor cells. However, the extent to which these cell types give rise to the lesion-filling astrocytes, and whether the astrocytes derived from different cell types contribute similarly to optic nerve regeneration remain unclear. Here we examine the distribution of astrocytes and NG2+ cells in an optic nerve crush model. We show that optic nerve astrocytes partially fill the injury site over time after a crush injury. Viral mediated expression of a growth-promoting factor, ciliary neurotrophic factor (CNTF), in retinal ganglion cells (RGCs) promotes axon regeneration without altering the lesion size or the degree of lesion-filling GFAP+ cells. Strikingly, using inducible NG2CreER driver mice, we found that CNTF overexpression in RGCs increases the occupancy of NG2+ cell-derived astrocytes in the optic nerve lesion. An EdU pulse-chase experiment shows that the increase in NG2 cell-derived astrocytes is not due to an increase in cell proliferation. Lastly, we performed RNA-sequencing on the injured optic nerve and reveal that CNTF overexpression in RGCs results in significant changes in the expression of distinct genes, including those that encode chemokines, growth factor receptors, and immune cell modulators. Even though CNTF-induced axon regeneration has long been recognized, this is the first evidence of this procedure affecting glial cell fate at the optic nerve crush site. We discuss possible implication of these results for axon regeneration.

1. Introduction

Following traumatic injury in the central nervous system (CNS), reactive astrocytes form a glial border around the lesion, which has been regarded as a barrier to CNS axon regrowth (Adams and Gallo, 2018; Boghdadi et al., 2020; Bradbury and Burnside, 2019; Filous and Silver, 2016; Sofroniew, 2020; Tran et al., 2018). This view has been supported by failed axon regrowth, with the presence of astrocytic scar and evidence that astrocytes produce growth inhibitory factors, including chondroitin sulfate proteoglycans (CSPGs) (Barritt et al., 2006; Bradbury et al., 2002; Cafferty et al., 2007; Duan and Giger, 2010; Jones et al., 2003a; Morgenstern et al., 2002). However, recent studies have refuted this prevailing view and proposed that astrocyte scar formation is not a primary cause of regeneration failure and that scar-forming astrocytes in fact aid CNS regeneration. For example, genetic manipulation of astrocytes leading to prevention of scar-forming astrocytes was shown to impede axon regeneration after spinal cord injury (SCI) (Anderson et al., 2016). RNA-sequencing (RNA-seq) studies have shown that astrocytes produce a repertoire of growth permissive molecules after injury (Anderson et al., 2016). Others have shown that after SCI, astrocytes not only secrete growth-promoting molecules, but they also express cell adhesion molecules (e.g., neural cadherin and fibronectin), allowing them to act as a permissive physical substrate for the elongating axons (Ferguson and Scherer, 2012; Ribeiro et al., 2020; Tom et al., 2004).

Several SCI studies have investigated the cellular sources of astrocyte populations that may be permissive to axon regeneration. Sabelström et al. demonstrated that ependymal cells give rise to astrocytes after SCI (Sabelstrom et al., 2013). However, another study has shown that ependymal-derived astrocytes are not prevalent in a clinically relevant injury model (Ren et al., 2017). Others have shown that after SCI, a small population of astrocytes penetrate the lesion core and form growth-permissive “bridges”. They further showed that ependymal-derived astrocytes are not the source of these bridges through which supraspinal axons regenerate (Zukor et al., 2013). In addition to ependymal cells, another potential source of growth-permissive astrocytes is oligodendrocyte progenitor cells (OPCs), which studies have shown to differentiate into astrocytes after contusive SCI (Hackett et al., 2016; Hackett et al., 2018). Using genetic lineage tracing, it was shown that about 20% of OPCs in the glial border region differentiate into GFAP+ astrocytes, and many of them accumulate around the astroglial-fibrotic border, with some extensions into the fibrotic region to form astrocyte bridges (Hackett et al., 2016). These observations suggest the possibility that the growth-permissive astrocytes are a subpopulation of the astrocytes derived from OPCs.

In contrast to SCI, where the lesion epicenter is largely devoid of astrocytes, the optic nerve injury site normally fills in with astrocytes over time (Hilla et al., 2017; Qu and Jakobs, 2013; Yungher et al., 2017), which raises the possibility that this astrocyte repopulation could contribute to axon regeneration. Even though optic nerve injury has been used widely to investigate the mechanisms of axon regeneration (Gokoffski et al., 2020; Luo and Park, 2012; Luo et al., 2014; Qian and Zhou, 2020; Williams et al., 2020; Yin et al., 2019), little is known about the fate of OPCs in this injury example, and whether OPCs contribute significantly to the formation of growth permissive astrocytes or to retinal ganglion cell (RGC) axon regeneration. In this study, we utilized genetic lineage tracing, cytokine-induced regeneration, and RNA-seq to investigate the cellular sources and possible mechanisms of astrocyte repopulation in the optic nerve injury site.

2. Materials and methods

2.1. Animals

All animal experimental procedures were performed in compliance with protocols approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Miami. Animals used were C57BL/6 J (The Jackson Laboratory, 000664), NG2creER^™ (Zhu et al., 2011)(The Jackson Laboratory, stock 008538), GFAPcreER^T2 (Casper et al., 2007)(a gift from Dr. Ken McCarthy, University of North Carolina), R26 loxP-STOP-loxP-tdTomato (Arenkiel et al., 2011)(a gift from Dr. Fan Wang, Massachusetts Institute of Technology), GLT1-eGFP (Regan et al., 2007)(a gift from Dr. Jeffrey Rothstein, Johns Hopkins University). All animals were housed in a viral antigen free facility and kept under standard 12-h light-dark conditions. For all surgical procedures, mice were anesthetized with isoflurane. For analgesia, buprenorphine (0.05 mg/kg) was administered post-operatively. Animals of both sexes were used.

2.2. Adeno-associated viruses (AAVs)

Animals received AAV serotype 2, expressing a secretable form of ciliary neurotrophic factor (CNTF). AAV-CNTF was produced as described previously (Yungher et al., 2015). Briefly, an AAV-compatible SIBR vector was created by PCR-amplifying the knockdown cassette of a SIBR vector. This cassette was inserted into a Stratagene AAV plasmid, replacing the CMV promoter and B-globin intron. The resulting AAV-SIBR plasmid with a ubiquitin promoter, was then modified via bridge PCR to create KpnI and BglII sites to flank the EGFP open reading frame. Plasmid DNA encoding human CNTF was purchased from Open-Biosystems (Accession: BC068030), and the open reading frame was amplified using a forward primer that incorporated both a 5′ Kpn1 restriction site and the NGF signal peptide sequence and a reverse primer that incorporated 3′ BglII site. PCR-amplified CNTF was then used to replace the EGFP ORF in the AAV-SIBR vector via standard restriction digest and ligation. All enzymes were purchased from New England Biolabs. Plasmids were then used to produce adeno-associated virus serotype-2 (AAV2) (1–9 × 10¹³ particles/ml). For making AAV2-Null, we used empty pAAV.CMV·C-HA as described in a previous publication (Bray et al., 2019). To incorporate the HA-tag at C-terminus of CNTF, the original cDNA sequence was PCR amplified using the following synthetic oligonucleotides: 5’-CCAAGGTACCATGTCCATGTTGTTCTACAC-3′ (CNTF-Acc65I) and 5’-CCGTAGATCTAGCGTAATCTGGAACATCGTATGGGTACATTTTCTTGTTGTTAGCAATAT-3′ (CNTF-BglII). The pAAV-CNTF plasmid described above was digested with Acc65I and BglII restriction enzymes. The resulting 666 bp DNA fragment was replaced with the new PCR fragment yielding pAAV-hCNTF-hemagglutinin (HA) construct that was sequenced to verify its integrity. The activity of HA-tagged hCNTF was assessed by its ability to activate the STAT3 signaling pathway in HEK293 cells and compared to untagged hCNTF using Western blotting with Rabbit anti-phospho-STAT3(Tyr705) antibody (Cell Signaling; 9145S).

2.3. Optic nerve crush (ONC)

For the ONC procedure, the left optic nerve was exposed intraorbitally by blunt dissection. The optic nerve was crushed with forceps (#5 Dumont, Fine Science Tools) for 10 s ~ 1 mm distal to the emergence from the globe (Luo et al., 2016; Park et al., 2008). Animals received optic nerve crush 3 to 14 days after AAV injection as indicated for each experiment. Sham surgery involved exposure of the optic nerve without nerve crush.

2.4. Tamoxifen injection

Injections were started at about postnatal day 35 (experiments involving GFAPcreER and NG2creER^™ mice). Procedure time-lines are outlined in the text and figure legends. The exact number of animals used for each group is indicated in the main text and figure legends. Mice received intraperitoneal injection of tamoxifen (0.124 mg/g body weight) for 5 consecutive days as previously described (Hackett et al., 2018). The mice were given a washout period of 11 days prior to receiving AAV injection.

2.5. Estimation of Cre induction efficiency in the NG2creER^™; Rosa26-tdTomato mice

NG2creER^™; Rosa26-tdTomato mice (6 weeks old) received intraperitoneal injection of tamoxifen (0.124 mg/g body weight) for 5 consecutive days. Cre induction efficiency was obtained 14 days after the last tamoxifen injection by dividing the number of tdTomato+ NG2+ cells by the total number of NG2+ cells (n = 3 mice).

2.6. Intravitreal injection

Female and male mice 5 to 7 weeks old underwent unilateral AAV injection. A fine glass micropipette was inserted into the posterior chamber, taking care to avoid damaging the lens. Using a Hamilton syringe (Hamilton 80,900) 2 μl of virus was slowly injected. Anterograde labeling of regenerating axons was performed by injecting 2 μl of 2 μg/μl Alexa 488 or Alexa-555 conjugated cholera toxin β subunit (CTB) (ThermoFisher, C22841 and C22843) 2–3 days before euthanasia.

2.7. Histology

Mice were anesthetized and then transcardially perfused with 4% paraformaldehyde in phosphate buffered saline (PBS). Optic nerves were dissected and postfixed in 4% paraformaldehyde in PBS for 2 h, and cryoprotected in 30% sucrose in PBS overnight. About 8 mm segments of the optic nerve centered at the injury site were embedded in OCT compound (Tissue-Tek) and longitudinal sections (10 μm) were cut using a cryostat. Sections were immunostained by incubating in primary antibodies in 5% Normal Goat Serum in PBS with 0.3% Triton-X overnight at 4 °C. Primary antibodies used were: RFP (Rockland 600–401-379S, 1:1000), GFAP (Invitrogen 130,300, 1:1000 or Dako Z033429, 1:500), APC/CC1 (Millipore OP80 Ab-7, 1:500), NG2 (Millipore AB5320, 1:200), GFP (Abcam ab13970, 1:2000) and Iba1 (Wako 019–19,741, 1:500). Following primary antibody incubation, sections were washed and incubated in species-appropriate Alexa Fluor IgG (H + L) secondary antibodies (Invitrogen, 1:500) at room temperature for 1 h. For CC1, goat-anti-mouse IgGγ2b secondary antibody was used (Invitrogen A-21141, 1:500). Slides were mounted using Vectashield with DAPI (Vector Laboratories H-1200). Images were obtained using a Nikon Eclipse Ti fluorescent microscope or an Olympus FluoView 1000 confocal microscope.

2.8. Quantification of lesion areas in the optic nerve

The area size of the optic nerve lesion from different animals was determined by manually drawing a contour around the relevant area approximately 1 mm away from the eye, using ImageJ software. They were presented as “YFP negative area”, “area occupied by YFP negative and GFAP positive cells”, “GFAP negative area” and “IBA+ lesion core area”. The size of the relevant lesion area was presented as μm² or as an arbitrary unit (Supplementary Fig. 1). At least three or four sections were analyzed for each animal.

2.9. Optic nerve RNA-sequencing and data analyses

Total RNA from mouse optic nerve was extracted using RNeasy Micro Kit (Qiagen #740040) and treated with DNAse I to remove genomic DNA contamination according to the manufacturer’s instructions. Total RNA-seq (ribo-depleted) libraries were produced using Truseq Stranded Total RNA library prep kit (Illumina, #20020596) with 500 ng of DNase-treated Input RNA.

The RNA-seq processing was done as described previously (Dasilva et al., 2021). Briefly, raw fastq files were trimmed with Trimmomatic v0.32 (Bolger et al., 2014) and aligned to mouse genome (mm10 version) using STAR aligner v2.5.3a (Dobin et al., 2013) with default parameters and RSEM v1.2.31 (Li and Dewey, 2011) to obtain expected gene counts against the mouse Ensembl (version 96). The differential expression analysis was determined between ONC + AAV-CNTF vs. ONC + AAV-Null animals, ONC + AAV-Null vs. sham injury + AAV-Null animals, and ONC + AAV-CNTF vs. sham injury + AAV-Null animals using DESeq2 (Love et al., 2014) and R v3.2.3. A gene was considered detected if the FPKM was >1 in three replicates of at least one sample and significantly changed if the q-value <0.05 and a fold change >1.5. Heatmap was generated using Python script. Volcano plots were generated using VolcaNoseR (Goedhart and Luijsterburg, 2020). Venn Diagram was generated using InteractiVenn. Raw RNAseq data of optic nerves were deposited to Gene Expression Omnibus (GEO: GSE196824).

2.10. Gene set enrichment analysis (GSEA)

Gene expression fold changes obtained from our RNA-seq data (injured relative to sham samples) were calculated with DESeq2 as described above, and the entire list of expressed genes was pre-ranked and imported into the GSEA program (Subramanian et al., 2005).

2.11. Fluorescent in situ hybridization

RNAscope fluorescent in situ hybridization (FISH) was performed in 10 μm thickness optic nerve sections using the RNAscoper^® Multiplex Fluorescent v2 Assay (ACD Biotechne, Catalog No. 323100) according to the manufacture’s protocol. Target probes used are as follows: RNAscope^® Probe - Mm-H19-O1 (Cat No. 501671), RNAscope^® Probe- Mm-Gfap-C3 - Cat No. 313211-C3), and RNAscope^® Probe - Mm-CXCL9 - Cat No. 489341). TSA-based fluorophores were from Perkin Elmer (TSA Plus Fluorescein, PN NEL741001KT; TSA Plus Cyanine 3, PN NEL744001KT; TSA Plus Cyanine 5, NEL745001KT). Images were acquired using an Olympus Confocal FV1000 microscope or an Andor Dragonfly confocal microscope.

2.12. Optic nerve Western blots for CNTF-HA

Each optic nerve was homogenized by sonication in 50 ul of 1× RIPA-SDS sample buffer supplemented with protease and phosphatase inhibitor cocktail (Roche; PhosSTOP; cOmplete Mini). Tissue lysates were separated on 4–12% gradient polyacrylamide gel (GenScript) and transferred to nirocellulose membrane, then blocked in 5% BSA-TBST for 1 h, and incubated overnight at 4 °C with rabbit-anti-HA-tag antibody (Cell Signaling; #3724; 1:1000) and mouse anti-β-actin (Sigma; A2228; 1:6000). Secondary antibodies were AzureSpectra goat-anti-mouse IR700 (1:5000) and goat-anti-rabbit IR800 (1:5000). The signal was detected using Azure C600 Gel imaging system.

2.13. EdU proliferation assay

Mice were given intraperitoneal injection of EdU (50 mg/kg; Invitrogen A10044) in 2% DMSO in PBS at 3, 5, and 7 days after ONC and sacrificed at 21 days after crush. Tissues were prepared for histology, and EdU was detected by using Click-iT EdU Alexa Fluor 488 Imaging Kit (Invitrogen C10337) according to manufacturer’s instructions. Sections were subsequently immunostained as described above. All sections were counterstained with DAPI, and only EdU⁺ cells that were also DAPI⁺ were included in the quantification.

2.14. Statistics

Statistical analyses were performed using GraphPad Prism software. Data were analyzed using Student’s t-test, one-way ANOVA, or two-way ANOVA with Tukey’s or Bonferroni’s post hoc test as indicated in the figure legends. Values of p < 0.05 were considered significant. Number of animals used for each animal group is indicated in the figure legends. All error bars represent SEM.

3. Results

3.1. GLT1-eGFP and GFAP staining show gradual occupation of the lesion site by astrocytes after optic nerve crush

Following traumatic injury in the optic nerve, macrophages and fibroblasts fill the lesion site within few days, creating growth-inhibitory fibrotic scars. However, studies have also shown that astrocytes partially close the wound weeks after optic nerve injury, possibly providing growth permissiveness (e.g., astrocytic bridge) for a small number of RGC axons (Hilla et al., 2017; Ribeiro et al., 2020). To examine the progression of astrocyte occupancy in the optic nerve lesion, we performed intraorbital ONC in adult mice and immunostained sections with a GFAP antibody. Since GFAP is not expressed uniformly in the astrocyte processes (i.e., thus, the antibody is unlikely to stain the entire astrocytes), we also used GLT1-eGFP BAC transgenic mouse line (Regan et al., 2007) in which all parts of astrocytes will be labeled, and thus allow more comprehensive examination of astrocyte distribution. At 3 days post-injury, regions devoid of GFAP⁺ cells were seen at the lesion epicenter. At 7 days, the lesion appeared larger. By 3 weeks post-injury, however, this GFAP-negative region was greatly reduced in size (Fig. 1). We observed that GFP co-localizes almost completely with GFAP, and like GFAP staining, GFP immunoreactivity showed that the lesion is markedly reduced by 3 weeks post-injury (Fig. 1). Thus, consistent with the previous studies demonstrating that lesion volume decreases over time, our results show that astrocytes fill the lesion site after ONC crush with some residual regions likely comprising a fibrotic scar.

Fig. 1. GLT1-eGFP BAC transgenic line and GFAP immunostaining show partial but substantial lesion closure weeks after optic nerve crush.

Representative optic nerve sections of GLT1-eGFP mice. A) uninjured, B) 3 days post crush (dpc) C) 7dpc and D) 21dpc. Green, GFP; Red, GFAP immunoreactivity. Adult mice received intraorbital optic nerve crush approximately 1 mm away from the eye. By 21 days after crush, the crush site is filled largely with astrocytes, leaving behind a region of what appears to be a fibrotic scar. GFAP staining and eGFP label the cells to a similar degree. Scale bars, 100 μm.

3.2. Investigating the cellular origin of the lesion-filling astrocytes in the optic nerve

To determine the cellular origin of the lesion-filling astrocytes in the optic nerve, we first examined whether they in fact originate from “mature” GFAP-expressing astrocytes. In our previous study, we used GFAPcreER and found that this mouse line effectively and specifically induces Cre recombination in astrocytes upon tamoxifen injection in adult mice (Ribeiro et al., 2020). Here we crossed this line to reporter mice (Rosa26-LSL-YFP) and administered tamoxifen starting about 2 weeks prior to ONC. This time scheme allowed us to label the existing astrocytes with YFP prior to injury and to track their fate after injury. Accordingly, we sought to examine the distribution of GFAP+ cells that are either YFP positive (YFP+) or YFP negative (YFP−) after ONC. Since GFAP is expressed predominantly in differentiated astrocytes in intact animals, we reasoned that cells that are both YFP+ GFAP+ should represent astrocytes derived from differentiated astrocytes, whereas YFP− GFAP+ cells are likely derived from cells other than astrocytes. In animals without any regenerative treatment (i.e., AAV-Null injected intravitreally 3 days before crush), YFP+ GFAP+ cells surrounded the crush site at 21 days post injury (Fig. 2A). To examine the lesion environment in the context of axon regeneration, we injected AAV-CNTF intravitreally 3 days before crush, a treatment well known to enforce CNTF expression in RGCs and promote RGC axon regeneration (Leaver et al., 2006; Luo et al., 2016; Yungher et al., 2015; Yungher et al., 2017). Unexpectedly, in the AAV-CNTF-treated animals at 21 days post crush, we observed that the size of the YFP− area around the crush site was markedly larger than in the AAV-Null animals (Fig. 2A and B). We also observed that the crush site was occupied largely by YFP− GFAP+ cells (Fig. 2C). On the other hand, the overall size of lesion (i.e., GFAP− regions) was not different whether the animals received AAV-CNTF or not (Fig. 2D). Upon closer look at the crush site with a high magnification in both animal groups, we found that many GFAP+ cells that have invaded the lesion core are mostly YFP− (Fig. 2E and Fig. F). Regenerating CTB+ RGC axons associated closely with these YFP− GFAP+ processes (Fig. 2E and Fig. F). These results suggest that the axon growth permissive astrocytes in the optic nerve capable of invading in the lesion site are unlikely to be derived from preexisting astrocytes.

Fig. 2. Intravitreal AAV-CNTF injection increases occupancy of YFP− GFAP+ cells in the crush site of GFAPcreER;Rosa26-YFP mice.

A) Representative optic nerve sections of AAV-Null and AAV-CNTF injected GFAPcreER; Rosa26-YFP mice. Animals received tamoxifen injection starting about 2 weeks prior to crush injury. AAVs were injected 3 days prior to crush, and animals were euthanized 3 weeks after crush. CTB555 was injected intravitreally 3 days prior to euthanasia. Blue, DAPI; Green, YFP; White, CTB555; Red, GFAP immunoreactivity. Red asterisks, core of the crush site.

B) Quantification of YFP negative (YFP−) lesion area at the crush site. N = 13 for AAV-Null and 16 for AAV-CNTF. Student t-test *, p < 0.05.

C) Quantification of GFAP positive (GFAP+) region within the YFP− region. N = 12 for AAV-Null and 16 for AAV-CNTF. Student t-test, **, p < 0.01.

D) Quantification of GFAP negative (GFAP−) lesion area at the crush site. N = 12 for AAV-Null and 16 for AAV-CNTF. Student t-test, n.s., not significant.

E and F) YFP (green) and GFAP (red) immunoreactivity at the crush site of AAV-Null and AAV-CNTF animals in A. GFAP+ cells that occupy the core of the lesion are often YFP−. Bottom images, higher magnification of the blue boxed area. Scale bars, 100 μm.

We also sought to assess the degree to which fibrotic scar is affected by CNTF. Previous studies have indicated that fibrotic scars consisting largely of macrophages and fibroblasts are strongly inhibitory to axon regeneration (Orr and Gensel, 2018; Soderblom et al., 2013; Zhu et al., 2015). The size of fibrotic scar as indicated by IBA+ lesion core was not affected by AAV-CNTF (Supplementary Fig. 1). Together, these results indicate that AAV-mediated expression of CNTF in RGCs does not reduce lesion size, but it does enhance the lesion occupancy of GFAP+ cells that are unlikely to be of astrocyte origin.

3.3. Assessment of NG2 cells and NG2 cell-derived astrocytes after optic nerve injury in the NG2creER^™ mice

Besides the local proliferating astrocytes, another source of the lesion-filling astrocytes in the CNS are ependymal cells, as demonstrated in the spinal cord (Sabelstrom et al., 2013). However, ependymal cells are unlikely to be present in the optic nerve, thus we excluded this possibility. Alternatively, these astrocytes could arise from NG2-expressing OPCs. To examine this possibility, we generated NG2creER^™; Rosa26-LSL-tdTomato double transgenic mice (hear after referred also as “NG2creER-Tdtomato”), a strategy used previously to fate map NG2-expressing OPCs (Hackett et al., 2016; Zhu et al., 2011). Previously, in-depth characterization of this NG2creER^™ line had revealed that NG2 cells in the postnatal brain generate only NG2 cells or oligodendrocytes, whereas NG2 cells in the embryonic brain generate protoplasmic astrocytes in the gray matter of the ventral forebrain in addition to oligodendrocytes and NG2 cells (Zhu et al., 2011). In this present study, we administered tamoxifen in young adult mice (postnatal day 35), and the animals were euthanized 5 weeks later. In these animals, we observed a few NG2 TdTomato+ cells in the uninjured optic nerve (Fig. 3A). Some of these TdTomato+ cells generated oligodendrocytes (i.e., CC1+) (Fig. 3A); about 25% of tdTomato+ cells were CC1+ oligodendrocytes whereas all tdTomato+ cells were negative for GFAP immunoreactivity (Fig. 3A), suggesting that NG2 cells continuously generate oligodendrocytes in the mature optic nerve. These results are similar to the previous studies that have investigated the fate of NG2 cells in the mature brain and spinal cord (Hackett et al., 2016; Zhu et al., 2011). Our previous study reported that the Cre induction efficiency in the spinal cords is about 30%. In the optic nerve however, we observe that the Cre induction efficiency is lower at about 5–10% (Fig. 3B).

Fig. 3. Intravitreal AAV-CNTF injection increases the occupancy of NG2 lineage cells at the optic nerve crush site.

A) Representative optic nerve sections of an uninjured NG2creER^™; Rosa26-TdTomato mouse showing tdTomato, CC1, and GFAP immunoreactivity. Scale bar, 20 μm.

B) A low magnification optic nerve of an uninjured NG2creER^™; Rosa26-TdTomato mouse showing tdTomato+ cells. The animal (about 6 weeks old) was given tamoxifen injection (5 consecutive days), and the optic nerve was removed 2 weeks after the first injection.

B′) Higher magnification of the boxed region in B. The section was stained with an NG2 antibody. NG2-tdTomato, yellow; NG2, green; DAPI, blue. We observe about 5–10% Cre induction efficiency in the optic nerve of this mouse line.

C) Quantification of tdTomato positive (tdTomato+) regions in the crush site. Student t-test *, p < 0.05; N = 6 for AAV-Null (or AAV-PLAP) and 13 for AAV-CNTF.

D) Representative optic nerve sections from AAV-Null and AAV-CNTF injected NG2creER^™; Rosa26-tdTomato mice. White, CTB555; Red, tdTomato fluorescence. Green asterisks, core of the crush site. Scale bar, 50 μm.

E) Higher magnification of the crush site of the AAV-CNTF injected animal in C. Green, GFAP; White, CTB555; Red, tdTomato fluorescence. Right panel, higher magnification of the blue boxed region. Black arrows indicate CTB+ axons that are closely associated with tdTomato+ cells and growing into the lesion core.

We examined NG2 cells and their progenies after ONC. Mice (postnatal day 35) received intraperitoneal injection of tamoxifen for 5 consecutive days and were given a washout period of 11 days prior to receiving AAV injection. At 3 weeks after ONC, there were only a few NG2 TdTomato+ cells in the crush site (Fig. 3C and D). On the other hand, we observed a significant increase in the presence of TdTomato+ cells in the crush site in the AAV-CNTF animals (Fig. 3C and D). Regenerating CTB+ RGC axons in the AAV-CNTF animals associated closely with tdTomato cells as they extend into the crush site (Fig. 3E). Thus, these results demonstrate that AAV-CNTF promotes accumulation of NG2 cells, which may facilitate RGC axon regeneration.

Next, we used immunohistochemistry and confocal microscopy to examine the cellular identity of tdTomato+ cells. We used a commonly used CC1 antibody to examine whether AAV-CNTF alters the proportion of TdTomato+ cells that are CC1+ at the injury site. However, as described in a previous study (Hackett et al., 2018), this mouse CC1 antibody resulted in non-specific staining with the IgGγ2B secondary antibody. This limited us in determining oligodendrocyte differentiation under the experimental conditions. Nonetheless, we did observe that many TdTomato+ cells in and around the crush site showed immunoreactivity for GFAP, indicating that NG2 cells generate astrocytes in these areas (Fig. 4A). We sought to examine whether the rate of astrocyte generation from NG2 cells differs between the AAV-CNTF and the AAV-Null treated animals. The portion of TdTomato+ cells that are GFAP+ was not different between the two animal groups (Fig. 4B), indicating that AAV-CNTF does not enhance NG2 cell differentiation into astrocytes in the optic nerve.

Fig. 4. Increase in NG2 cell-derived astrocytes is not due to increase in cell proliferation.

A) GFAP expression in tdTomato+ cells. Representative optic nerve section from an optic nerve crushed and AAV-CNTF injected NG2creER^™; Rosa26-tdTomato mouse. GFAP and tdTomato dual labeling is shown by confocal microscopy 3 weeks post-crush. The panel shows GFAP and tdTomato co-labeling in merged orthogonal views of confocal z stacks. Animals received tamoxifen injection starting 2 weeks prior to crush injury. AAVs were injected 3 days prior to crush. Green, GFAP; Red, tdTomato. Scale bar, 20 μm.

B) Quantification of the percentage of tdTomato+ regions that are GFAP+. N = 3 for AAV-Null and 4 for AAV-CNTF. Student t-test; n.s., not significant.

C) EdU pulse-chase experiment. Representative optic nerve sections from AAV-Null and AAV-CNTF injected NG2creER^™; Rosa26-tdTomato mouse. Animals received tamoxifen injection starting 2 weeks prior to crush. AAVs were injected 3 days prior to crush, and animals were euthanized 3 weeks after crush. Blue, DAPI; Green, EdU647; White, CTB555; Red, tdTomato fluorescence.

D) Quantification of the percentage of tdTomato+ cells that are EdUT. N = 4 for AAV-Null and 5 for AAV-CNTF. Student t-test; n.s., not significant.

E) Higher magnification of the crush sites. Green, EdU647; Red, tdTomato fluorescence. No visible difference in the total number of EdU+ cells are seen between the two animal groups.

To determine whether the increase in NG2 TdTomato+ cells seen in the AAV-CNTF animals is due to an increase in cell proliferation, we used pulse-chase strategy for EdU labeling. Eleven days after the first tamoxifen injection, young adult mice (postnatal day 35) received intravitreal AAV injection. Two weeks after the first tamoxifen injection, mice were subjected to ONC. At 3, 5, and 7 days after ONC, mice were injected intra-peritoneally with EdU. As shown in Fig. 4C and D, the percentage of TdTomato+ that are EdU+ was not different between the animal groups, indicating that the increase in NG2 TdTomato+ cells is not due to an increase in NG2 cell proliferation. We also found that the total number of EdU+ proliferated cells in the crush site is not obviously different between the animal groups (Fig. 4E).

3.4. Intravitreal AAV-CNTF injection results in CNTF presence at the optic nerve lesion site

Our results indicate that AAV-mediated expression of CNTF in the retina modifies the lesion environment in the optic nerve. Since intravitreal AAV2 used in this study are known to transduce predominantly RGCs with some amacrine and bipolar cells, modulation of the lesion by AAV-CNTF could be mediated by axonal transport of CNTF from RGCs to the injury site. Studies in the past have shown that ectopically expressed cytokine in neurons can be transported along the axons and to the axon terminals (Leibinger et al., 2021). To determine whether virally expressed CNTF is transported to the injured optic nerve, we generated an AAV-CNTF construct with a HA-tag. Through this AAV construct, we can readily detect HA as an indicator of CNTF presence. Animals received intravitreal injection of AAV-CNTF-HA or AAV-Null, followed by ONC 2 weeks later. We performed Western blot on the injured optic nerves (7 days post crush) and observed HA in the AAV-CNTF-HA injected animals (Fig. 5A).

Fig. 5. RNA-seq on the injured mouse optic nerve.

A) Upper panel, diagram showing the optic nerve region collected for Western blot. Optic nerve regions covering the crush site and about 3 mm distal to the crush site were collected for Western blot analysis. Animals received either intravitreal AAV-Null or AAV-CNTF-HA injection, followed by crush or sham injury 2 weeks later. Lower panel, Western blot results from the optic nerve of mice subjected to sham injury + AAV-Null (Sham + Null), ONC + AAV-Null (ONC + Null), and ONC + AAV-CNTF-HA (ONC + CNTF).

B) Venn diagram showing the numbers of unique and overlapping DEGs (fold-change >1.5 and adjusted p-value <0.05) from the three animal group comparisons; 1) ONC + Null vs. Sham + Null, 2) ONC + CNTF vs. Sham + Null and 3) ONC + CNTF vs. ONC + Null.

C) Number of upregulated (in red, 1554 genes) and downregulated (in blue, 1173 genes) genes induced by ONC + Null compared to Sham + Null with the fold-change >1.5 and adjusted p-value <0.05. Heatmap of the DEGs in ONC + Null (“NO”, 5 replicates) vs. Sham + Null (“NS”, 5 replicates) with the fold-change >1.5 and adjusted p-value <0.05.

D) Volcano plot depicting the DEGs in ONC + Null vs. Sham + Null mice with the fold-change >1.5 and adjusted p-value <0.05.

E) Top15 GSEA of KEGG Pathway Analysis terms of ONC + Null vs. Sham + Null DEGs comparison with significant Normalized Enriched Score (NES) and False-Discovery Rate (FDR) < 0.05.

3.5. RNA-seq of the injured mouse optic nerves

Next, we sought to gain insight into the signaling pathways and cellular processes by which CNTF affects NG2 cells. To this end, we performed RNA-seq on the mouse optic nerves. As in the experiment above, animals received intravitreal injection of AAV-CNTF or AAV-Null, followed by ONC 2 weeks later. Seven days after ONC, we removed the optic nerves for RNA-seq. In total, we generated three animal groups; 1) sham injury (i.e., optic nerve exposed but not crushed) plus intravitreal AAV-Null injection, 2) ONC with AAV-Null, and 3) ONC with AAV-CNTF (five biological replicates per group). Pre-amplified poly (A)-enriched RNA was sequenced to an average depth of 83 million reads (+10 million reads) per sample. We used a series of comparisons to identify genes uniquely expressed in the injured animals. Differential gene expression was defined as expression ±1.5 fold change and adjusted (adj) p value <0.05. As shown in the Venn diagram (Fig. 5B), many genes were differentially expressed after ONC (in both the AAV-Null and AAV-CNTF animals) compared to the uninjured animals (i.e. sham injury + AAV-Null). There was a large number of overlapping differentially expressed genes (DEGs) among the two comparisons, ONC + AAV-Null vs. sham injury + AAV-Null, and ONC + AAV-CNTF vs. sham injury + AAV-Null. Although much fewer, DEGs were also detected between ONC + AAV-Null and ONC + AAV-CNTF animals. The full lists of DEGs are provided in the Extended Data, Supplementary Table 1, 2, and 3.

3.6. Comparing the transcriptional profiles between injured and uninjured optic nerves

We found hundreds of DEGs between the injured and uninjured optic nerve (i.e. ONC + AAV-Null vs. sham injury + AAV-Null); a total of 1173and 1554 genes were down-regulated and upregulated after injury, respectively (Fig. 5C). Top upregulated genes in the injured group include those involved in cell proliferation (e.g., Mki67), cathepsins (e.g., Ctsb, Ctsd, Ctsa and Ctsz), lipases (e.g. Lpl), complement components (e.g., C1qb and C1qa), myeloid cell receptors and proteins (e.g., Ly86 and Trem2), and extracellular matrix (ECM) proteins (e.g., Spp1). Top down-regulated genes include kinesin family members (e.g., Kif5a and Kif5b), those involved in the c-Jun amino-terminal kinase signaling pathway (e.g., Mapk8ip2), and Fyco which is a gene that encodes a Rab7 adapter protein implicated in the microtubule transport of autophagosomes, and the guanine nucleotide exchange factor Rapgef4 (Fig. 5C and D). We conducted GSEA using Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways as a gene set. A rank-ordered list was generated with genes identified as expressed in the injured mouse optic nerve, and this list was used as input to a pre-ranked GSEA analysis. The top upregulated pathways include cell cycle, signaling pathways involved in the immune and inflammatory response (e.g., systemic lupus erythematosus, cytokine-cytokine receptor interaction, and complement coagulation cascade), cell migration (e.g., ECM receptor interaction) and cell death (e.g., p53 signaling pathway) (Fig. 5E).

3.7. Comparing the transcriptional profiles between the injured AAV-CNTF and uninjured optic nerves

We found DEGs between the injured AAV-CNTF and uninjured groups (Extended Data Fig. 5-2), many of which overlapped with the DEGs seen from the group comparison above. These overlapping DEGs include Mki67, C1qb, Ly86, Ctsb, Ctsz, and Laptm5 (Extended Data Fig. 5-2).

3.8. Distinct genes are highly expressed in the injured optic nerve following intravitreal AAV-CNTF injection

We found genes that are differentially expressed between injured AAV-CNTF and injured AAV-Null animals. A total of 68 DEGs were found in this comparison. Of these, 13 were down-regulated and 55 were upregulated (Fig. 6A and B). GSEA analysis using the KEGG pathway as a gene set input showed that the top upregulated pathways were “cytokine-cytokine receptor interaction,” “chemokine signaling pathway”, and “cell adhesion molecules CAM”. Top upregulated genes enriched in the KEGG pathway term “cytokine-cytokine receptor interaction” were Cxcl9, Ccl22, Cxcl10, Ccr2, and Ccr5 (Fig. 6C and D). As expected, AAV-CNTF animals showed high expression of suppressor of cytokine signaling-3 (Socs3) (Extended Data Fig. 5-3), a negative regulator of the CNTF-STAT3 signaling pathway, and a gene known to be highly induced by CNTF (Park et al., 2009; Smith et al., 2009). The IL6 family of cytokines are widely known to induce expression of various chemokines (Xie et al., 2021). Consistent with this notion, we found that several chemokines were highly enriched in the AAV-CNTF animals, including Cxcl9, Cxcl10, and Ccl22. Other notable genes that were highly enriched include cell survival regulators (Fgl2 and Dtx3l), growth factors (Fgfr4), drivers of M1 macrophage polarization (Epsti1 and Gbp9), and regulators of lymphocytes (Cd3e, Slamf8, Gimap4, Slfn1 and Tgtp2) and regulators of myeloid cells (Ccr2 and CD74) (Extended Data Fig. 5-3).

Fig. 6. Intravitreal injection of AAV-CNTF alters the expression of distinct genes in the injured optic nerve.

A) Number of upregulated (in red, 55 genes) and downregulated (in blue, 13 genes) genes induced by ONC + CNTF compared to ONC + Null with the fold-change>1.5 and adjusted p-value <0.05. B) Volcano Plot depicting the DEGs in ONC + CNTF vs. ONC + Null mice with the fold-change>1.5 and adjusted p-value <0.05. C) Top 15 GSEA of the KEGG Pathway Analysis terms of ONC + CNTF vs. ONC + Null DEGs comparison with significant Normalized Enriched Score (NES) and False-Discovery Rate (FDR) < 0.05. D) Upregulated DEGs enriched in both KEGG Pathway cytokine-cytokine receptor interaction and chemokine signaling pathway. E and F) Fluorescent in situ hybridization validation of DEGs (H19 in E and Cxcl9 in F) after ONC and CNTF injection. Consistent with the RNA-seq data, H19 expression is highly induced in the optic nerve 7 days post crush (7dpc). H19 in red and Gfap in green. In E, the merged images are higher magnification of the yellow boxed areas in the single channel images. Asterisks indicate the crush site. E’) Higher magnification of the white boxed area in the merged images in E. F. Cxcl9 expression in the Null and CNTF animals. Cxcl9 expression increases in the animals with ONC and AAV-CNTF. Image on the far right, higher magnification of the boxed area. Scale bars, 100 μm in E and 40 μm in F.

To validate the gene expression changes seen in the RNA-seq, we performed FISH. First, we examined the expression of H19 lncRNA, one of the most highly induced genes in both ONC groups in our RNA-seq data (mean FPKM values of 11, 117, and 182 for uninjured, ONC + AAV-Null and ONC + AAV-CNTF groups, respectively). Previous studies have shown that astrocytes and microglial cells express H19, and this gene contributes to glial cell activation and proliferation (Li et al., 2020). Nonetheless, the cell types and the extent of H19 expression in the injured optic nerve are unknown. Using two color FISH, we observed that H19 expression was undetectable or at vety low level in the uninjured optic nerve. However, ONC caused dramatic increase in H19 expression in the optic nerve. Notably, we observed that H19 expression was predominantly in the astrocytes surrounding the crush site (Fig. 6E). We also observed H19 expression in the optic nerve meninges (Fig. 6E).

One of the top upregulated DEGs in the AAV-CNTF animals was Cxcl9. We performed FISH to examine Cxcl9 expression and found that this gene is highly induced in the crush site of the AAV-CNTF animals (Fig. 6F), further validating the RNA-seq data. Together these results demonstrate that AAV-CNTF causes alteration in the expression of distinct genes in the injured optic nerve, many of which are known regulators of cell survival and migration. Whether these genes play a role in modulating the fate of NG2 cells and the pathogenesis after optic nerve injury remain unknown.

4. Discussion

After traumatic injury, reactive astrocytes form borders around the tissue lesions and separate necrotic from healthy tissue. There has been extensive effort into determining the precise roles of astrocytes and the mechanisms by which these cells influence axon regeneration. Due to the easy access to the nerve and the simplicity of eye injection, the optic nerve injury model has been used widely to interrogate the neuron-intrinsic mechanisms of axon regeneration. In this study, we use this model to examine astrocyte repopulation in the tissue lesions. Using GFAP antibody and GLT1-eGFP mice, we first confirmed that after crush injury, astrocytes partially fill the lesion site, and that regenerating RGC axons project along the astrocytes, thus supporting the growth-facilitating role played by these cells. Our study shows that intravitreal injection of AAV-CNTF increases the presence of NG2 cell-derived astrocytes in the optic nerve lesion. Even though CNTF-induced RGC axon regeneration had been described for a long time, to our knowledge, this is the first evidence of this procedure affecting glial cell fate at the ONC site. It has been described extensively that the AAV2 strategy we used in this study, induces transgene expression primarily in RGCs (Leaver et al., 2006; Luo et al., 2016; Yungher et al., 2015; Yungher et al., 2017). Our results from the AAV-CNTF-HA experiment also show that CNTF likely produced by RGCs are present in the lesioned optic nerve, raising the possibility that axonally transported CNTF acts locally in the lesion site and affects the injury environment.

Astrocytic borders and fibrotic scar have been subjects of intense investigation for SCI. Numerous SCI studies have examined the cellular sources of glial borders, lesion filling astrocytes and fibrotic scars affecting axon regeneration. As mentioned above, studies have shown that ependymal-derived astrocytes are not the source of astrocytic bridges through which supraspinal axons regenerate (Ren et al., 2017; Zukor et al., 2013). We and others have shown that OPCs can differentiate into astrocytes after contusive SCI, and that many of them accumulate around the astroglial-fibrotic border, with some extensions into the fibrotic region to form astrocyte bridges (Hackett et al., 2016; Hackett and Lee, 2016). Nonetheless, the number of OPC-derived astrocytes in the fibrotic region is usually small, and this is associated with lack of axon regeneration (Hackett et al., 2016; Hackett et al., 2018). Moreover, the fibrotic scars which astrocytes are mostly unable to fill in, persist for several months (Li et al., 2021). Thus, it is viewed that the extent to which astrocytes fill the lesion in the spinal cord after injury is less than that of the optic nerve. One can speculate that the more efficient lesion filling seen in the optic nerve could partially explain why spontaneous axon regeneration is more common in the optic nerve than in the spinal cord. Currently it is unclear what accounts for the difference in the astrocytes occupancy between the two systems. Optic nerve is strictly a white matter tissue with no residing neuronal cell bodies whereas diverse types of neurons reside in the spinal cord. It is unknown whether the absence of neurons contributes to the lesion difference seen in the optic nerve. Additionally, it is known that various cell types can affect the lesion environment, which could indirectly influence the fate of OPCs and astrocytes. For instance, infiltrating macrophages are known to recruit fibroblasts and pericytes into the lesion site (Zhu et al., 2015). It is unclear whether there are differences in the composition or types of macrophages and fibroblasts/pericytes between the two systems, and whether such differences contribute to the overall difference in the astrocyte occupancy after injury.

NG2 cells are a population of CNS cells that are distinct from mature oligodendrocytes, astrocytes, and microglia. Also referred to as OPCs, they are ubiquitously distributed throughout the CNS and are capable of differentiating into oligodendrocytes and astrocytes in the adult CNS (Nishiyama et al., 2021; Richardson et al., 2011). In this regard, their proliferation and differentiation has been an appealing target to promote remyelination after CNS injury (Hackett and Lee, 2016). Indeed, NG2 cells are found to accumulate around the lesion site (Lytle and Wrathall, 2007; McTigue et al., 2001; Tripathi and McTigue, 2007), and several studies have examined the mechanisms by which OPCs proliferate, differentiate into oligodendrocytes, and contribute to remyelination (Almad et al., 2011; Miron et al., 2011). Several factors capable of affecting NG2 cell differentiation into oligodendrocytes include fibroblast growth factor 2 (FGF2) (Tripathi and McTigue, 2008; Wolswijk and Noble, 1992), glial growth factor 2 (GGF2) (Whittaker et al., 2012; Zai et al., 2005), and Wnts (Fernandez-Martos et al., 2011). FGF2 was shown to act as a potent mitogen for NG2 cells (Wolswijk and Noble, 1992), and deletion of FGFR1 and FGFR2 in NG2 cells reduces oligodendrogenesis and remyelination chronically after cuprizone-induced demyelination (Furusho et al., 2015). Others have demonstrated that CNTF and leukemia inhibitory factor (LIF) are important for the proliferation and differentiation of NG2 cells into oligodendrocytes and astrocytes, at least in vitro (Mayer et al., 1994). Intraperitoneal administration of CNTF increases the numbers of NG2 cells and oligodendrocytes and improves outcome after experimental autoimmune encephalomyelitis (EAE) (Kuhlmann et al., 2006). Importantly, deletion of signal transducer and activator of transcription 3 (STAT3), delays oligodendrogenesis without affecting proliferation after SCI (Hackett et al., 2016), while overexpression of a constitutively active STAT3 increased oligodendrocyte differentiation in vitro (Steelman et al., 2016).

Studies have shown that NG2 cells can differentiate into astrocytes under certain conditions (Hackett and Lee, 2016; Mayer et al., 1994). The NG2-Cre mouse line showed that some NG2 cells could differentiate into protoplasmic astrocytes in the ventrolateral forebrain gray matter (Zhu et al., 2008a) and spinal cord (Zhu et al., 2008b) during development. Interestingly, in NG2-CreER mice, only 8% of NG2 cells expressed GFAP at 10 days post cortical stab injury, indicating minimal differentiation into astrocytes under this condition. On the other hand, after contusive SCI, 25% of reporter-labeled cells in the NG2-CreER mice expressed GFAP at 1 week post-injury (Hackett et al., 2016), indicating that the degree of astrogliogenesis from NG2 cells likely differs depending on the type of injury (Hackett and Lee, 2016; Hackett et al., 2018). The molecular factors that regulate astrogliogenesis from NG2 cells are yet unclear. Possible pathways include bone morphogenetic protein (BMP) (Rajan and McKay, 1998), and/or Olig2 signaling pathways (Fukuda et al., 2004; Setoguchi and Kondo, 2004). BMP2 and BMP4 were shown to promote astrogliogenesis from NG2 cells in vitro (Mabie et al., 1997). Both BMP2 and BMP4 are upregulated after SCI (Hampton et al., 2007), and intraspinal injection of BMP4 leads to increased differentiation of transplanted NG2 progenitors into GFAP⁺ astrocytes (Sellers et al., 2009).

How does CNTF increase the presence of NG2 cells and NG2 cell-derived astrocytes in the optic nerve injury site? There are at least three possible explanations; CNTF might be increasing NG2 cell migration, proliferation, or survival. In previous studies, two-photon live imaging has revealed that NG2 cells react to laser injury by migrating only short distances toward the lesion (Hughes et al., 2013), suggesting that the large number of NG2 cells at the injury site is likely due to local proliferation rather than migration. In support of this, the percentage of proliferating NG2 cells was found to increase several folds (McTigue et al., 2001) and NG2 cells comprise nearly one half of bromodeoxyuridine (BrdU)-labeled cells after SCI (Zai and Wrathall, 2005). These results indicate that NG2 cells have a significant capacity to proliferate after SCI. However, the EdU pulse chase experiment in our study does not show an obvious increase in proliferation capacity, arguing against this scenario. Thus, one can speculate that CNTF instead promoted survival and maintenance of NG2 cells in the lesioned environment. Indeed, CNTF and other IL6 family members are well known to promote survival of various cell types, including glial cells in various pathological conditions (Murakami et al., 2019), further supporting the cell survival scenario. Examination of cell death specifically (e.g., using TUNEL staining) will help determine whether an increase in the survival of NG2 cells was the cause of NG2 cell and NG2 cell-derived astrocyte increases seen in the injured optic nerve.

To examine the signaling pathways and cellular processes that might be involved in the underlying AAV-CNTF effects, we performed bulk RNA-seq on the optic nerve. As expected, we found hundreds of genes that are differently expressed after injury, many of which are known to play roles in modulating glial cell activation, immune responses, and inflammation. Furthermore, we also observed numerous genes that are uniquely induced after AAV-CNTF injection. It is possible that some of these genes may act downstream of CNTF and influence NG2 cells. For example, CXCL9 and CXCL10 are known to regulate immune migration, differentiation, and activation of immune cells, such as cytotoxic lymphocytes (CTLs), natural killer (NK) cells, NKT cells, and macrophages. Others have shown that CXCL9 regulates cell apoptosis via the AKT pathway (Song et al., 2019).

Given the prominent induction after CNTF, we have examined the expression of Cxcl9, but it is unclear which cells highly express this gene in the optic nerve. However, previous studies have shown that CXCL9 is expressed in multiple cell types including macrophages and microglial cells in response to pathological conditions (Ellis et al., 2010). Others have shown that these chemokines are expressed in neurons in response to CNS injury and can activate microglia, directing them to the lesion site. CXCR3, a receptor for CXCL9 is expressed in microglia and that this receptor system controls microglial migration (Rappert et al., 2004). Recent study has shown that macrophages are the predominant source of CXCL9 and their depletion abrogates CD8+ T-cell infiltration (House et al., 2020). Importantly, studies have also shown that CXCR3 is expressed in OPCs as well as in mature oligodendrocytes (Omari et al., 2005). Based on these observations, it is possible that CXCL9 is expressed in the infiltrating macrophages (or microglial cells) and regulate OPCs directly or indirectly in the optic nerve. However, the exact cell types in the optic nerve that express CXCL9 and CXCR3 remain to be determined. It is also possible that growth factors (Fgfr4) and the drivers of macrophage polarization (Epsti1 and Gbp9) that were highly induced after AAV-CNTF affect the survival of NG2 cells in the lesion site. However, whether these and other differentially expressed genes in the optic nerve in fact mediate the CNTF’s effects warrants further investigation.

What roles do NG2-derived astrocytes play for CNTF-induced axon regeneration? Since the regenerating RGC axons were in close contact with NG2-Tdtomato cells, it is likely that these cells were growth permissive for these axons. This does not align with the notion that NG2 cells and reactive astrocytes, which are considered major sources of CSPGs, are major inhibitors of axon regeneration (Dou and Levine, 1994). However, the inhibitory properties of CSPGs do not suggest that NG2 cells and astrocytes themselves are inhibitory. Similar to our study, several studies have shown that NG2 cells and astrocytes are often associated with regenerating axons (Anderson et al., 2016; Zukor et al., 2013). Regenerating axons are observed frequently in areas of the spinal cord that are NG2⁺ after SCI (Jones et al., 2003b; McTigue et al., 2006). These data suggest that NG2 cell-derived astrocytes themselves may be permissive to axon growth (Yang et al., 2006), and AAV-CNTF promotes regeneration of RGC axons into the lesion site at least partly by promoting the presence of NG2 cell-derived astrocytes. Further supporting this, our previous study using sparse cell labeling demonstrated that most of the regenerating RGC axons after CNTF treatment project along the paths of astrocyte processes within the injured optic nerve (Bray et al., 2017). Others have noted that while axons may be able to use NG2 cells as a growth-permissive substrate, they can form terminal synaptic contacts with NG2 cells, causing prevention of axon growth (Filous et al., 2014; Han et al., 2012). In this regard, it is possible that CNTF switches on the regenerative program in the RGCs and promotes the accumulation of NG cell-derived astrocytes that are favorable for regeneration, but at the same time, these cells create a “trap” in the lesion site, preventing further growth of axons to the distal targets. Delineating the precise role of NG2 cell-derived astrocytes in AAV-CNTF animals and the underlying molecular mechanisms could help devise better strategies for enhancing enhance axon regeneration beyond the site of injury.