RNAi‐mediated suppression of vimentin or glial fibrillary acidic protein prevents the establishment of Müller glial cell hypertrophy in progressive retinal degeneration

Gliosis is a complex process comprising upregulation of intermediate filament (IF) proteins, particularly glial fibrillary acidic protein (GFAP) and vimentin, changes in glial cell morphology (hypertrophy) and increased deposition of inhibitory extracellular matrix molecules. Gliosis is common to numerous pathologies and can have deleterious effects on tissue function and regeneration. The role of IFs in gliosis is controversial, but a key hypothesized function is the stabilization of glial cell hypertrophy. Here, we developed RNAi approaches to examine the role of GFAP and vimentin in vivo in a murine model of inherited retinal degeneration, the Rhodopsin knockout (Rho−/−) mouse. Specifically, we sought to examine the role of these IFs in the establishment of Müller glial hypertrophy during progressive degeneration, as opposed to (more commonly assessed) acute injury. Prevention of Gfap upregulation had a significant effect on the morphology of reactive Müller glia cells in vivo and, more strikingly, the reduction of Vimentin expression almost completely prevented these cells from undergoing degeneration‐associated hypertrophy. Moreover, and in contrast to studies in knockout mice, simultaneous suppression of both GFAP and vimentin expression led to severe changes in the cytoarchitecture of the retina, in both diseased and wild‐type eyes. These data demonstrate a crucial role for Vimentin, as well as GFAP, in the establishment of glial hypertrophy and support the further exploration of RNAi‐mediated knockdown of vimentin as a potential therapeutic approach for modulating scar formation in the degenerating retina.


| INTRODUCTION
Reactive gliosis is regarded as a cellular attempt to protect the surrounding tissue from further damage, in order to promote repair and limit neuronal re-modeling. It includes morphological, biochemical, and physiological changes, each of which can vary with the type and severity of the initiating insult (Silver & Miller, 2004). In the eye, gliosis involves primarily Müller glial cells, which span the retina and provide structural and metabolic support to the surrounding neurons. In lower vertebrates, Müller cells may attempt to repair the damaged retina by de-differentiating into progenitor-like cells in response to injury (Jadhav, Roesch, & Cepko, 2009) but this capacity is largely lacking in the mammalian retina (Loffler, Schafer, Volkner, Holdt, & Karl, 2015).
Instead, reactive gliosis is characterized by the upregulation of intermediate filament proteins (IFs), particularly glial fibrillary acidic protein (GFAP) and vimentin, and glial hypertrophy, which involves the extension, thickening and stabilization of glial processes. Depending on the injury, some of these processes can extend beyond the outer limiting membrane (OLM) into the subretinal space between the neural retina and the retinal pigment epithelium (RPE; Hippert et al., 2015;Lewis & Fisher, 2000;Pearson et al., 2010;Winkler, Hagelstein, Rohde, & Laqua, 2002). In addition, there is a concomitant increase in the deposition of inhibitory extracellular matrix (ECM) molecules, such as chondroitin sulfate proteoglycans (CSPGs; Chen, FitzGibbon, He, & Yin, 2012;Hippert et al., 2015). Together, these lead to the formation of an inhibitory "scar" at the apical edge of the neural retina (Bringmann et al., 2006).
Müller glial cell reactive gliosis has been reported as a response to various retinal pathologies, including inherited retinal degeneration (Hippert et al., 2015), and geographic atrophy and wet age-related macular degeneration (Wu, Madigan, Billson, & Penfold, 2003). Prolonged gliosis, as can occur with chronic degeneration, can have deleterious effects on tissue function and is considered to be a major factor limiting endogenous regeneration throughout the central nervous system (CNS; Bringmann et al., 2006;Burns & Stevens, 2018;Fawcett & Curt, 2009). The glial scar may impede neuronal cell migration and axonal regrowth either by presenting a physical barrier in the form of hypertrophic glial processes, or a reservoir of inhibitory ECM molecules, or a combination of these (Chan, Wong, Liu, Steeves, & Tetzlaff, 2007;Cregg et al., 2014). In the eye, it has been postulated that it may prevent Müller glial cell-cycle re-entry, while the expansion of hypertrophic Müller glial processes into the subretinal space forms fibrotic tissue that has been shown to block the regeneration of outer segments and exacerbates retinal detachment (Lewis & Fisher, 2000).
Identifying precisely which aspects of reactive gliosis impede therapeutic interventions in vivo and finding ways to ameliorate them is challenging but crucial (Burns & Stevens, 2018;Hamby & Sofroniew, 2010). Recent evidence has suggested that the stiffness of the glial scar affects the regenerative properties of nervous tissue and that changes in IF expression levels correlates with changes in cell stiffness (Lu et al., 2011;Moeendarbary et al., 2017;L. Wang et al., 2018). An increase in GFAP is considered the hallmark of reactive gliosis and GFAP is thought to be critically important in the formation of the extended and thickened (hypertrophic) astrocytic processes typically seen in reactive gliosis (Lundkvist et al., 2004;Pekny, Wilhelmsson, & Pekna, 2014). Conversely, vimentin has received much less attention, perhaps because of the less dramatic changes in its expression upon injury.
Here, we use RNAi to modulate the expression of Gfap or Vimentin in a mouse model of inherited, progressive retinal degeneration. We report a crucial role for vimentin, as well as GFAP, in the establishment of retinal glial hypertrophy. This also raises the potential utility of RNAi-mediated knockdown of vimentin as a therapeutic approach for modulating glial hypertrophy as part of scar formation in the damaged retina.
Mice received food and water ad. lib. and were provided with fresh bedding and nesting. Rlbp.Gfp +/+ mice were used at postnatal (P)7-8.
Rho À/À and wild-type animals were aged P10 (±1 day) at the time of AAV injection, unless otherwise stated. Both male and female animals were used in all experiments. All experiments have been conducted in accordance with the United Kingdom Animals (Scientific Procedure) Act of 1986.

| Plasmid construction
shRNA constructs for Gfap and Vimentin were generated by subcloning the target hairpins (shGfap or shVim) into the mU6pro plasmid. The target hairpins were placed downstream of a U6 promoter to form an RNAi-expression cassette. The RNAi cassettes were excised and cloned into an AAV pD10.CBA backbone containing a Red Fluorescent Protein (RFP) reporter. The shGfap targets the Gfap sequence ShGfap and shVim hairpins were designed using a standard shRNA motif of sense-loop-antisense in which the loop consisted of nine nucleotides. shRNA is less likely to generate off-target effects compared to siRNA, but the selected sequences were further verified for low likelihood of off-target effects by using web-based RNAi design software (Dharmacon, Invitrogen) where genome-scanning algorithms assess the sequences for homology with the rest of the genome; the selected shRNA sequences were found to be highly specific and ontarget. Non-targeting hairpin was used as a control (shControl), the sequence used for this was: 5 0 -GATCGGACACTCCTCATAA-3 0 . The resulting constructs were named pshGfap, pshVim, and pshControl.
The sequences most effective for knocking down Gfap (pshGfap2) and

| Cell transfection
Human embryonic kidney (HEK) cell line 293 T cells were seeded in 24-well plates at a density of 1.5 x 10 5 cells per well in 1 ml of

| Astrocyte primary culture
Brains of P7-8 pups were dissected per previously established protocols (Schildge, Bohrer, Beck, & Schachtrup, 2013) and put into a petri dish containing cold Hank's Balanced Salt Solution (HBSS, Gibco, Thermo-Scientific, Loughborough, UK). Briefly, the meningeal layer was carefully removed from both hemispheres and the cortices were separated from the rest of the brain manually and placed in a new petri dish containing HBSS. Collected tissue was cut into small pieces using dissection scissors. The harvested tissue was transferred into a 15 mL sterile falcon tube and centrifuged at 200g for 5 min. The supernatant was aspirated, and cortices were resuspended in 1 ml/per brain of 0.05% trypsin/DNase and incubated at 37 C/5% CO 2 for 20 min. After incubation, the cell suspension was passed through a 40 μM cell strainer and centrifuged for 10 min at 200g. Trypsin solution was removed, and cell pellet was gently washed with PBS before being resuspended in DMEM/10% FBS/1% AA medium and seeded into sixwell poly-D-lysine coated plates (Invitrogen). After 2 weeks in culture, astrocytes were transfected with 5 x 10 12 particles of AAV-shControl vector and fixed for immunocytochemistry 1 week post-transduction.

| Intravitreal injection of viral vectors
P10 or 4.5 week old Rho À/À mice were anesthetized with an intraperi- Retinal flat mounts were prepared by excising the eye, as before, and carefully removing the sclera and retinal pigmented epithelium (RPE). The neural retina was fixed in 4% PFA for 1 h and counter-stained with Hoechst 33342 before being cut to create four conjoined "petals" (from the periphery towards the optic nerve), which were then flattened out on a slide. A drop of fluorescent mounting medium (DAKO, Ely, UK) was placed upon the tissue and the coverslip was then carefully placed on top of it. at 0.5 μm z-steps.

| Müller cell 3D reconstruction
Leica confocal xyz images (in tiff format) were directly opened in Amira 5.5.0 (FEI.com). Data series were filtered using a noise reduction median filter and then individual cells were semi-automatically segmented using a Wacom drawing tablet. The segmentation data was smoothed and "islands" with connected area of voxels containing a number of voxels less than or equal to the size value specified were removed. Three-dimensional surfaces were generated and shown using vertex normals. Three to five cells were reconstructed per file.
All data were treated in the same way.

| Müller cell apical process measurement
The area occupied by the apical terminal processes of transduced Müller glia was measured using Image J (Open Source; imagej.nih.gov/ ij/) and GNU Image Manipulation Programme (GIMP; Open Source; https://www.gimp.org/). Orthogonal views of the individual substacks were used to track individual cells to their end-foot processes.
Only those end-feet that could be clearly identified as being associated with cell bodies that occupied a position in the inner nuclear layer (INL) were used for assessment. Sub-stack xy projections encompassing the first 15-20 μm at the apical margin of the retina were generated from the flatmount confocal xyz stacks described above. The resulting image was then opened using GIMP software and converted to grayscale. Threshold Alpha tool was used to separate signal from background and the same threshold was applied to all analyzed images. Regions of interest (ROI) were selected using the free selection tool. The perimeter of each individual end-foot process was manually delineated and the histogram dialogue was used to quantify the total number of pixels within the selected area. The measured values for the total area were then converted to μm 2 . All quantifications were made by an independent assessor in a fully blinded manner. These values are presented in the text and were used for statistical analysis. A second assessor analyzed the same data sets for verification purposes. The two assessors made independent selections of end-feet to measure. These data are presented in Supplementary Information (see Figure S5).

| Statistics
All means are reported as means ± SD unless otherwise stated. n = number of independent retinae or cultures, where appropriate.
qRT-PCR results are based on at least two independent Müller cell primary cultures or cohorts of animals, as appropriate. All in vivo assessments are based on N > 3 independent retinae per condition.
Statistical significance was assessed using the Graphpad Prism 5 software and denoted as p < .05 = *, p < .01 = **, p < .001 = ***. As not all data were normally distributed, as assessed using the d'Agostino and Pearson omnibus normality test, nonparametric t-tests and oneway ANOVAs were applied as specified in the figure legends. Corrections for multiple comparisons were applied, where necessary.

| AAV-shGfap and AAV-shVim mediate effective knockdown of endogenous Gfap and Vimentin in Müller cell primary cultures
Despite intense investigation, the roles of GFAP and other IFs in gliosis, particularly hypertrophy, remain unclear. This is in part due to a reliance on knockout models, in which the protein in question is absent from the point of conception onwards, leading to the possibility of compensatory upregulation in other related proteins. The use of knockout models has also necessitated the use of acute injury models. Here, we sought to explore the role of GFAP and vimentin in prolonged gliosis, as occurs in progressive retinal degeneration. We therefore designed RNA interference (RNAi) strategy that could be introduced into models of inherited retinal degeneration at different time points, both before and after the establishment of Müller Glial hypertrophy.
Gfap and Vimentin silencing hairpins and non-targeting control (shControl) were designed (Table S2)  3.2 | AAV-shGfap and AAV-shVim mediate effective knockdown of endogenous Gfap and Vimentin in vivo in Rho À/À mice We next assessed AAV-shGfap and AAV-shVim in vivo. We chose the Rho À/À model of inherited retinal degeneration because it exhibits a moderate rate of photoreceptors loss (compared to very fast models like PDEβ (rd1/rd1) and much slower models like Prph2 (rd2/rd2) ) and we have previously found it to exhibit a strong upregulation of Gfap and vimentin and clear evidence of Müller glia hypertrophy (Hippert et al., 2015;Pearson et al., 2010). In the first instance, vector was administered at P10 since Müller glia become post-mitotic in the central retina by P6 (Jadhav et al., 2009) and retinogenesis is complete by P10 in the rodent retina (Young, 1985).
Degeneration and the concomitant reactive gliosis begins shortly after this in the Rho À/À mouse (Hippert et al., 2015;Humphries et al., 1997). Recipient P10 Rho À/À mice received AAV-shControl, AAV-shGfap or AAV-shVim intravitreally and were assessed 3 weeks later in the first instance. In the untreated Rho À/À mouse (Figure 2ai, bi), Gfap expression is very low at P10 and largely restricted to the outer edge of the ganglion cell layer (GCL) and the outer plexiform layer (OPL; Figure 2ai). It is markedly upregulated during the course of degeneration and by 4.5 weeks of age, GFAP is widespread and demarcates the full extent of the Müller glial processes, which appear thickened (Figure 2bii).
Close inspection of the apical margin of the neural retina reveals the lateral extension of terminal GFAP +ve processes along the outer edge of the outer nuclear layer (ONL; Figure 2biii).
Administration of AAV-shControl led to a small, but statistically significant, reduction in Gfap mRNA (Figure 2ai) compared to uninjected controls, although this did not result in a noticeable reduction in GFAP as assessed by IHC (Figure 2ci,ii). However, given this result, and because the assessment of Müller Glial morphology (see later) necessitates the use of the RFP reporter within both control and RNAi-containing vectors, the effects of AAV-shGfap and AAV-shVim were compared against AAV-shControl, as opposed to untreated controls. As such, any effect seen compared to shControls is likely to rep- In the untreated P10 Rho À/À mouse, Vimentin is expressed throughout the retina (Figure 3ai,bii-iv and Hippert et al., 2015).
However, this was not reflected in a reduction in vimentin levels, as assessed by IHC (Figure 3ciii,iv). Given that vimentin is stably expressed within the normal P10 retina (Figure 3bi), we reasoned that it might take longer for the knockdown in Vimentin RNA to be reflected in a change in vimentin protein levels (Desclaux et al., 2015).
Vimentin levels were largely unchanged or increased in the untreated ( Figure 3biv) and AAV-shControl treated (Figure 3di,ii) 7.5-week-old Rho À/À retina. In contrast, vimentin was markedly reduced in AAV-shVim expressing cells (Figure 3diii-iv). ShGfap did not result in any indirect effects on Vimentin expression (Figure 3aii). Taken together, these data show that administration of AAV-shGfap and AAV-shVim prevent the up-regulation of GFAP and vimentin, respectively, which is associated with progressive retinal degeneration in the mouse retina. As before, there was no indication of compensatory changes in the expression of the other IF (Figures 2aii and 3aii).

| Simultaneous knockdown of Gfap and
Vimentin has deleterious effects on both degenerating and normal retina to either anophthalmia or severe microphthalmia. In Rho À/À mice, 68% exhibited anophthalmia and 32% presented with microphthalmia (N = 9 eyes; Figure 4a). In the microphthalmic eyes, transduction was patchy, but both GFAP and vimentin were reduced in regions of good DsRed expression by 6 weeks post-injection (Figure 4b). This was associated with a small, but significant reduction in ONL thickness

F I G U R E 3 RNAi targeting vectors mediate knockdown of vimentin in Müller glia in vivo in the degenerating retina. (ai) Histogram show that
Vimentin mRNA levels, as assessed by RT-qPCR, are already abundant at the time of injection (P10) and increase significantly with degeneration. Administration of AAV-shControl vector had no effect on this increase. (ii) Administration of AAV-shVim led to a significant decrease in Vimentin mRNA at 3 weeks post-transduction (4.5 weeks of age), compared to AAV-shControl while no off-target effects were seen with AAV-shGfap. In contrast to the results in vitro, AAV-lh did not significantly reduce Vimentin mRNA in vivo. p < .05 = *, p < .01 = **, p < .001 = *** with a one-way ANOVA with Tukey's for multiple comparisons; Error bars: SD; n ≥ 5 independent retinae per group. (b) Immunohistochemistry shows that vimentin (green) expression is widespread in Müller glia in both P10 (i) wild-type and (ii-iv) Rho À/À animals. (ci-iv) Examination of the retinae injected with RNAi targeting vectors (red) showed that, at 3 weeks post-injection, there were no notable changes in Vimentin expression in any of the three conditions, compared to untreated age-matched eyes. (di-iv) Examination 6 weeks post-injection showed that vimentin expression remained robust in (i,ii) AAV-shControl treated eyes but was significantly reduced in transduced cells (red) in eyes treated with (iii,iv) AAV-shVim. and number of photoreceptor rows (see Figure 6ai). Importantly, even in the absence of degeneration, we observed a similarly detrimental effect following co-injection of shGfap and shVim into wild-type eyes; 100% eyes (N = 8; Figure 4a) were microphthalmic and presented very disturbed cytoarchitecture, including extensive whorl formation (Figure 4c,e). Qualitatively, the presence of whorls was typically associated with higher levels of transduction, as measured by DsRed expression in the combined knockdown ( Figure S3a). To control for the increased viral load, we injected wild-type and Rho À/À mice with  Figure 5a,bi,ii). Müller glial lateral spread in Rho À/À retina was already significantly larger at 3 weeks of age (1.2-fold increase; 90.1 μm 2 ± 20.5; p < .01) and increased further by 6 weeks (1.5-fold increase; 125.6 μm 2 ± 36.8; p < .0001) compared to agematched AAV-shControl treated WT (Figure 5a,ci,ii). These results indicate that marked and rapid changes in Müller glial cell hypertrophy occur at the apical margin of the retina between 3 and 6 weeks of age in the Rho À/À model of retinal degeneration.
We then examined the impact of Gfap or Vimentin knockdown on Müller glial hypertrophy apical terminal process spread. As expected, no significant changes in lateral spread were observed at 3 weeks post-transduction of AAV-shGfap (88.5 μm 2 ± 23.3, compared to 90.1 μm 2 ± 20.5 in AAV-shControl-treated Rho À/À cells; N.S.). However, by 6 weeks post-injection, AAV-shGfap transduced Müller cells presented a lateral spread (86.2 μm 2 ± 24.2) that was very similar to age-matched wild-type retinae, and significantly reduced compared to shControl-treated Rho À/À retinae at the same time point (p < .001; Figure 5aii). We observed an even larger difference in AAV-shVim treated cells; here, lateral spread was significantly reduced (À1.9-fold reduction; 67.8 μm 2 ± 17.7; p < .001) compared to age-matched AAV-shControl treated Rho À/À , and the cells retained a gross morphology similar to age-matched wild-type AAV-shControl treated Müller glia We next sought to determine whether it is possible to modify hypertrophy once it is established. Rho À/À or wild-type mice were injected at 4.5 weeks of age, when GFAP and vimentin expression is F I G U R E 5 Vimentin is required for degeneration-associated hypertrophy of Müller glial apical processes. (a) Scatter plots show the area occupied by the apical terminal processes of Müller glia in Rho À/À and wild-type animals transduced with AAV-shControl. At 4.5 weeks of age (i.e., 3 weeks post-transduction) there were no significant differences between the two strains, but at 7.5 weeks old (6 weeks post-transduction) Rho À/À Müller glial apical terminal processes were hypertrophic and occupied a significantly larger area, compared to Müller glia in both 4.5 week old Rho À/À and 7.5 week old wild-type controls . Comparison of the three RNAi vectors at 3 and 6 weeks post-transduction of Rho À/À retinae showed that AAV-shVim or AAV-shGfap each prevent the increase in area occupied by Müller glial apical terminal processes (red significance bars). p < .001 = *** and p < .0001 = ****, Statistical test applied, nonparametric t-test ( already upregulated, and examined 6 weeks later ( Figure S4b,d).
Transduction efficiency is lower following intravitreal injection into adult mice ( Figure S4c), compared with P10 recipients, reducing the total number of cells that could be analyzed. However, we saw no reduction in apical spread in either shGfap-(1.02-fold change; N.S.) or shVim-treated (À1.03-fold change; N.S.) compared to shControltreated Rho À/À retinas (N > 3 retinae for all conditions). Together, this suggests that reducing the expression of IFs alone may not be sufficient to reverse hypertrophy, once it has established.
3.5 | Preventing Müller glial hypertrophy in the Rho À/À retina does not alter photoreceptor cell death or gliosis-related changes in the extracellular matrix

| DISCUSSION
Glial scars, the cellular changes that constitute a physical barrier between the region of damage and the surrounding tissue, can be beneficial and protect the healthy tissue from the damaged environment, but they are also recognized as a major obstacle for neuronal repair , (Burns & Stevens, 2018;Escartin, Guillemaud, & Carrillo-de Sauvage, 2019). In the context of regenerative medicine, they are hypothesized to impede endogenous repair mechanisms (Burns & Stevens, 2018) and represent a barrier to efficient connectivity between tissue grafts and implant devices and the host tissue, as well as reducing the efficiency of cell and gene therapy approaches (Hippert, Graca, & Pearson, 2016;Pearson, 2014 GFAP has received most attention due to its striking upregulation following injury (Messing & Brenner, 2020). This upregulation has been postulated to help form and stabilize the hypertrophic cellular processes exhibited by reactive astrocytes and glia, which represent the most striking morphological change associated with glial scarring (Lundkvist et al., 2004;Verardo et al., 2008). However, this notion remains an area of significant debate, due to numerous conflicting reports of attempts to manipulate glial IF expression and results from in vitro studies have often differed from those obtained in vivo (reviewed in Pekny, Wilhelmsson, Tatlisumak, & Pekna, 2019;Wilhelmsson et al., 2017;Zeisel et al., 2018), underlining the importance of testing gliosis in context. Indeed, in our study, knockdown of vimentin led to little or no changes in the morphology of Müller glia in vitro, but almost completely prevented degeneration-induced Müller glial hypertrophy in vivo. Such differences exemplify not only the regional differences in the role of IFs, but also the importance of the cytoarchitectural context when assessing the role of a given IF.
F I G U R E 6 Suppression of gliosis-related upregulation of GFAP and vimentin does not affect photoreceptor degeneration or CSPG deposition. (a) Administration of AAV RNAi vectors by intravitreal injection resulted in little or no changes to ONL thickness, compared to noninjected age-matched controls, at (i) 3, or (ii) 6 weeks post-injection. (b) Administration of AAV RNAi vectors by intravitreal injection resulted in little or no changes to number of photoreceptor rows, compared to non-injected age-matched controls, at (i) 3 or (ii) 6 weeks post-injection, with the exception of a small, but significant, decrease in those eyes co-injected with AAV-shGfap and AAV-shVim at 3 weeks post transduction, compared to AAV-shControl treated eyes. p < .05 = *, p < .01 = **, p < .001 = ***, with a one-way ANOVA with correction for multiple comparisons. Error bars: SD; n ≥ 5 independent retinae per group. N.B. a small but significant increase in ONL thickness was seen at 3, but not 6, weeks post-injection of AAV-shControl treated Rho À/À retinae, compared to age-matched untreated eyes; this likely reflects an injectionrelated disturbance, since no difference was observed in the number of photoreceptor rows. (c) CSPG expression (as shown by CS-56 staining; grey) increases between 4.5 and 7.5 weeks of age in the degenerating Rho À/À retina. (d,e) Administration of AAV RNAi vectors (red) by intravitreal injection at P10 resulted in little or no changes to ONL thickness, compared to non-injected age-matched controls, at (d) 3 weeks or (e) 6 weeks post-injection. Scale bar: 25 μm [Color figure can be viewed at wileyonlinelibrary.com] A further limitation of preceding investigations has been the reliance on knockout models, in which the protein in question is absent from the point of conception onwards and the development of that animal may be subject to multiple compensatory mechanisms. Given that GFAP has been known of for more the 50 years, it is perhaps surprising that relatively few studies have sought to manipulate its expression (or that of vimentin) after development is complete (Messing & Brenner, 2020). Moreover, the majority of studies have focused on non-degenerative models, meaning that assessments of the role played by different IFs invariably reflect their role in the glial response to acute trauma, rather than to progressive degenerative injuries (Burda & Sofroniew, 2014).
To begin to address these issues, we designed an in vivo gene ander's Disease . These studies contrast with early reports of the glial response to an acute needle stab injury in either the cortex or spinal cord in Gfap À/À knockout models, which was indistinguishable between Gfap À/À and wild-type mice (Pekny et al., 1995;Pekny et al., 1999 (Linberg, Sakai, Lewis, & Fisher, 2002;Merriman, Sajdak, Li, & Jones, 2016). In these retinae, despite the lack of GFAP upregulation, vimentin +ve Müller glial processes apparently still filled the empty spaces left by dead photoreceptors, preserving gross retinal structure.
In the absence of vimentin, it will be interesting to explore the composition of the intercellular space in hypertrophic and AAV-shVimtreated diseased retina using serial section electron microscopy.
Vimentin's role in hypertrophy may explain the surprisingly mild effects of knocking out Gfap observed in earlier reports. Vimentin filaments are absent in astrocytes in the uninjured corpus callosum, spinal cord and hippocampus of Gfap À/À mice but Vimentin expression is markedly increased following injury or culturing of Gfap À/À astrocytes, suggesting an attempt to compensate for the lack of GFAP . In uninjured Gfap À/À mice, vimentin filaments are seen in the anterior column of the cervical spinal cord (X. Wang, Messing, & David, 1997), in cerebellar Bergmann glia (Galou et al., 1996) and in Müller cells in the retina (Gomi et al., 1995). In Vim À/À mice, reactive brain astrocytes can still form IFs but they are comprised of abnormally tightly packed GFAP bundles , in keeping with the notion that vimentin acts as a facilitator of GFAP filament assembly (Galou et al., 1996). Vimentin itself also requires a partner for polymerization, but in the absence of GFAP, this can be other IF proteins, such as nestin or synemin. Surprisingly, however, other studies of Vim À/À mice have shown that in the injured brain, astrocytes are able to form some abnormal IFs, but these are made only of GFAP, and that loss of vimentin was associated with a decrease in nestin and synemin protein expression . It would, therefore, appear that potential compensatory changes associated with knockout of a given IF ( If prolonged, glial scarring is understood to exacerbate neuronal degeneration. It is not clear, however, if the converse is true, that is, whether prevention or reduction in glial scarring has any beneficial effect on neuronal degeneration. Miller and colleagues (Nakazawa et al., 2007) reported that chemical or detachment-induced retinal degeneration was less extensive in Gfap À/À /Vim À/À mice compared to wild-type controls, while Sivak and colleagues (Livne- Bar et al., 2016) reported that blockade of Type III IF dynamics (which includes GFAP and vimentin) using Withaferin A led to a reduction in metabolic injury-induced inner retinal apoptosis. By contrast, Perez and colleagues (Wunderlich et al., 2015) noted that cell death was accelerated in the inner retina in Gfap À/À /Vim À/À mice after ischemia/ reperfusion, but that photoreceptor cell loss was comparable to that in wild-type mice. Here, we found that the degree of photoreceptor loss associated with progressive retinal degeneration was largely unaffected by RNAi-mediated suppression of GFAP or vimentin alone.
However, simultaneous knockdown of both Gfap and Vimentin in both the degenerating Rho À/À and in the normal wild-type retina resulted in severe abnormalities, with many injected eyes failing to develop normally. This may be explained by the fact that when both are removed, functional IFs cannot form since other IFs, such as nestin  or synemin (Jing et al., 2007), are able to form IFs alone. As discussed earlier, differences between knockout and RNAi approaches may be explained in part by compensatory mechanisms that come into play in the knockout, which would otherwise not be present in a normal developing/diseased retina. In those retinae that did go on to develop following GFAP/vimentin knockdown using RNAi, a small exacerbation of photoreceptor loss was observed, indicating a dose-dependent effect. Taken together, these findings indicate that reducing either GFAP or vimentin alone does not ameliorate (nor exacerbate) photoreceptor loss during progressive degeneration. They also suggest that the impact of IF deficiency on retinal cell survival is both disease-and cell type-specific.
There is a growing body of evidence demonstrating the negative consequences of reactive gliosis, particularly when it is not resolved within the post-acute phase after injury. While the acute response may ensure the survival of cells through the post-traumatic phase, prolonged gliosis and scarring have negative effects on regeneration and plasticity, as shown in a range of different experimental models (reviewed in . The stiffness of glial cells, which correlates with IF expression, is proposed to affect the regenerative properties of nervous tissue (Lu et al., 2011;Moeendarbary et al., 2017;L. Wang et al., 2018). Reducing hypertrophy, as one aspect of gliosis, may therefore provide a more permissive environment for regeneration. Indeed, Gfap À/À /Vim À/À mice showed reduced hypertrophy of astrocytes and a partial restoration of synaptic connectivity after entorhinal cortex lesion (Wilhelmsson et al., 2004). Similarly, extensive sprouting was observed in Gfap À/À /Vim À/À mice after spinal cord hemi-section and these animals also performed better in behavioral tasks (Menet, Prieto, Privat, & Gimenez y Ribotta, 2003).
Interestingly, Gfap À/À animals performed similar to wild type and did not show any improvement regarding neurite sprouting and outgrowth or function, again underlining the importance of vimentin in this process. There is much excitement around the potential for Müller glial cell cycle re-entry and Müller glial-mediated endogenous repair of the retina, a phenomenon common in lower vertebrates, but rarely seen in the mammalian retina (see (Langhe & Pearson, 2020;Pearson & Ali, 2018) and it will be of significant interest to explore the role of Müller glial-stiffness in modulating this response.
Novel therapeutic approaches include the subretinal and intravitreal transplantation of various donor cell populations (Aghaizu, Kruczek, Gonzalez-Cordero, Ali, & Pearson, 2017;Barber et al., 2013;Barnea-Cramer et al., 2016;Kruczek et al., 2017;Pearson et al., 2012;Tassoni et al., 2015;Waldron et al., 2018) and neonatal (Seiler & Aramant, 2012) and stem cell-derived (Mandai et al., 2017) retinal sheet grafts and the introduction of subretinal and epiretinal electronic implants (Pardue et al., 2001;Zrenner et al., 2011). Each of these requires close physical interaction between the graft and the remaining host inner retina, where extensive hypertrophy at the apical surface of the neural retina may impede (Hippert et al., 2016). Further investigations are required to determine the relative contributions of glial hypertrophy and gliosis-related changes in ECM composition on the efficacy of these different therapeutic approaches. Although we did not see a beneficial effect of attenuating glial hypertrophy on photoreceptor loss directly, it may still be of indirect benefit to patients with progressive retinal degeneration. For example, gliosis is common in patients with diabetic retinopathy (Lechner, O'Leary, & Stitt, 2017;Mizutani, Gerhardinger, & Lorenzi, 1998) and may contribute to pathogenesis; for example, diabetic retinopathy is frequently exacerbated by glial hypertrophy at the vitreal and apical surfaces of the retina leading to retinal detachment. In the current study, we were able to almost completely prevent Müller Glial hypertrophy, at least at the apical surface (we did not examine hypertrophy at the vitreal surface), raising the possibility of designing therapeutic approaches that might at least partially ameliorate this condition. Notably, however, we could not reverse hypertrophy by RNAi once it was established, underlining the importance of correct, and most likely early, timing for any such interventions.
Together, our findings deepen our understanding of the roles