Tgfb3 collaborates with PP2A and Notch signaling pathways to inhibit retina regeneration.

Neuronal degeneration in the zebrafish retina stimulates Müller glia (MG) to proliferate and generate multipotent progenitors for retinal repair. Controlling this proliferation is critical to successful regeneration. Previous studies reported that retinal injury stimulates pSmad3 signaling in injury-responsive MG. Contrary to these findings, we report pSmad3 expression is restricted to quiescent MG and suppressed in injury-responsive MG. Our data indicates that Tgfb3 is the ligand responsible for regulating pSmad3 expression. Remarkably, although overexpression of either Tgfb1b or Tgfb3 can stimulate pSmad3 expression in the injured retina, only Tgfb3 inhibits injury-dependent MG proliferation; suggesting the involvement of a non-canonical Tgfb signaling pathway. Furthermore, inhibition of Alk5, PP2A or Notch signaling rescues MG proliferation in Tgfb3 overexpressing zebrafish. Finally, we report that this Tgfb3 signaling pathway is active in zebrafish MG, but not those in mice, which may contribute to the different regenerative capabilities of MG from fish and mammals.


Introduction
Blinding eye diseases like macular degeneration, glaucoma, and diabetic retinopathy result in retinal neuron death which leads to vision loss. Unlike mammals, zebrafish have a remarkable ability to regenerate neurons that were lost due to injury or disease (Goldman, 2014;Lenkowski and Raymond, 2014;Wan and Goldman, 2016). Key to this regenerative response are Mü ller glia (MG) the major glial cell type in the retina of both fish and mammals. The normal function of MG is to maintain retinal structure and homeostasis (Bringmann et al., 2009;MacDonald et al., 2015;Reichenbach and Bringmann, 2013). However, in fish, MG respond to retinal injury by dividing and generating multipotent progenitors for neuron regeneration (Bernardos et al., 2007;Fausett and Goldman, 2006;Fimbel et al., 2007;Powell et al., 2016;Ramachandran et al., 2010b;Raymond et al., 2006).
Although it is not known why MG from fish and mammals respond differently to retinal injury, it likely results from differences in their environment and intrinsic differences that are reflected in their gene expression programs. Unlike mammals, the zebrafish retina continues to grow throughout life and this growth-permissive environment may impact MG's potential to mount a regenerative response (Hitchcock and Raymond, 2004). Zebrafish MG themselves also appear to contribute to a pro-regenerative environment by releasing growth factors and cytokines after retinal injury (Calinescu et al., 2009;Kaur et al., 2018;Nelson et al., 2013;Ramachandran et al., 2011;Wan et al., 2012;Wan et al., 2014;Zhao et al., 2014). Dying neurons and immune cells may also contribute to the pro-regenerative MG niche in fish. In addition to niche factors, intrinsic differences in gene expression programs in MG from fish and mammals have been noted (Sifuentes et al., 2016). In fish, Ascl1a and Lin28a are critical factors promoting MG reprogramming and proliferation (Elsaeidi et al., 2018;Fausett et al., 2008;Gorsuch et al., 2017;Mitra et al., 2018;Mitra et al., 2019;Ramachandran et al., 2010a;Ramachandran et al., 2011). ascl1a and lin28a RNAs are highly induced in MG following injury to the fish retina, while their homologs remain undetectable in the injured mouse retina (Elsaeidi et al., 2018;Karl et al., 2008). However, forced expression of Ascl1, along with HDAC inhibition or Lin28a expression can stimulate a limited proliferative response by MG in the injured mouse retina (Elsaeidi et al., 2018;Jorstad et al., 2017).
Another intrinsic difference between MG from fish and mammals is Notch signaling. In mice, Notch signaling declines as MG differentiate and mature; while in fish, Notch signaling is maintained into adulthood (Bernardos et al., 2005;Dorsky et al., 1995;Elsaeidi et al., 2018;Furukawa et al., 2000;Nelson et al., 2011;Wan and Goldman, 2017;Wan et al., 2012). The maintenance of Notch signaling in MG of the adult zebrafish retina contributes to MG quiescence and Notch suppression is required for MG proliferation (Conner et al., 2014;Elsaeidi et al., 2018;Taylor et al., 2015;Wan and Goldman, 2017;Wan et al., 2012). Furthermore, the opposing actions of Fgf8a on MG proliferation in juvenile and adult fish is correlated with corresponding changes in Notch signaling activity (Wan and Goldman, 2017). Thus, Notch signaling is a major control point in the decision to proliferate or remain quiescent and understanding how Notch signaling is regulated in the zebrafish retina will help reveal mechanisms underlying MG's decision to mount a regenerative response.
In addition to Notch signaling, Tgfb signaling has been implicated in regulating injury-dependent MG proliferation in the zebrafish retina (Conedera et al., 2020;Lenkowski et al., 2013;Sharma et al., 2019;Sharma et al., 2020;Tappeiner et al., 2016). However, there are inconsistencies among these reports with some suggesting it is inhibited in proliferating MG (Lenkowski et al., 2013;Sharma et al., 2019) and others suggesting it is activated in these cells (Conedera et al., 2020;Sharma et al., 2020;Tappeiner et al., 2016). Although most of the above studies suggest Tgfb signaling inhibits MG proliferation, one study suggests it is necessary for injury-dependent MG proliferation (Sharma et al., 2020). Besides these inconsistencies, the endogenous ligand responsible for stimulating Tgfb signaling and the downstream signaling components responsible for regulating MG proliferation remain unknown.
In zebrafish, Tgfb ligands are encoded by four genes: tgfb1a, tgfb1b, tgfb2, and tgfb3. Canonical Tgfb signaling occurs when a Tgfb ligand engages a type II receptor that recruits type I (Alk5) receptor to stimulate phosphorylation of Smad2 and Smad3, which stimulates their nuclear import and allows for regulation of target genes (Shi and Massagué, 2003). Non-canonical Tgfb signaling refers to receptors that engage Erk, Jnk, p38, or protein phosphatase 2A (PP2A)-dependent pathways (Derynck and Zhang, 2003;Petritsch et al., 2000). Identification of the specific Tgfb ligands mediating retina regeneration in fish and unravelling their mechanism of action are critical for understanding how the Tgfb signaling pathway regulates MG proliferation and retina regeneration.
Here we provide evidence indicating Tgfb3 controls MG quiescence via a non-canonical Tgfb signaling pathway. Of all Tgfb ligand encoding genes, we find tgfb3 expression is uniquely restricted to quiescent MG in the adult zebrafish retina. Following retinal injury, this expression is suppressed at the injury site. Using transgenic fish that allow for conditional expression of Tgfb3, we show that tgfb3 suppression is necessary for injury-dependent MG proliferation. Interestingly, our studies reveal a specificity in the actions of Tgfb ligands on MG proliferation with Tgfb3, but not Tgfb1b, stimulating MG quiescence. Our studies suggest PP2A and Notch signaling pathways act downstream of Tgfb3. Furthermore, we report that Tgfb3 stimulates pSmad3 expression in the injured retina; however, pSmad3 expression is not sufficient to drive MG quiescence. Finally, we report that the Tgfb3 expression is not detectable in mouse MG, and this may contribute to their poor regenerative potential.

Results
pSmad3 signaling is suppressed in injury-responsive MG pSmad3 immunofluorescence was used to detect canonical Tgfb signaling in the uninjured retina of gfap:GFP transgenic fish. This analysis revealed that pSmad3 expression was restricted to GFP+ MG ( Figure 1A). Importantly, this expression was suppressed when fish were immersed in water containing the Tgfb receptor 1 (Alk5) kinase inhibitors SB431542 or SB505124 (  (D) Top diagram shows time line for heat shock treatment after injury and when fish were sacrificed. Bottom panels show pSmad3 immunofluorescence in injured and heat shock-treated Wt and hsp70:ca-Alk5 transgenic fish. Asterisk marks the injury site and arrowheads point to recovery of pSmad3 expression at the injury site in heat shock-treated hsp70:ca-alk5 transgenic fish. White Figure 1 continued on next page We next investigated if pSmad3 expression was regulated by retinal injury. A needle poke was used to cause a focal injury (Fausett and Goldman, 2006). This manipulation stimulated a rapid depletion in pSmad3 expression at the injury site ( Figure 1B-C), which was rescued after heat shock of hsp70:ca-Alk5 transgenic fish that express a constitutively active Tgfb receptor 1 (ca-Alk5, T204D) under the control of a hsp70 heat shock promoter ( Figure 1D-E; Wieser et al., 1995;Zhou et al., 2011). Furthermore, forced expression of ca-Alk5 inhibited injury-dependent MG proliferation ( Figure 1F-G). However, when TUNEL stain was used to identify apoptotic cells, very few TUNEL+ cells were identified (Figure 1-figure supplement 1B). Thus, the reduced injury-dependent MG proliferation noted with ca-Alk5 overexpression may reflect increased pSmad signaling and/or reduced cell death. Regardless, these data indicate that Tgfb signaling via pSmad3 expression correlates with MG proliferation in the injured retina.
tgfb3 expression correlates with injury-dependent pSmad3 expression To identify injury-responsive Tgfb ligands that might regulate pSmad3 signaling in the retina, we interrogated RNAseq data sets from MG and MG-derived progenitors that were isolated from uninjured gfap:GFP and injured 1016 tuba1a:GFP transgenic fish retinas, respectively (Fausett and Goldman, 2006;Kassen et al., 2007). This analysis indicated constitutive and low expression of tgfb1a; injury-dependent induction of tgfb1b and tgfb2; and injury-dependent suppression of tgfb3 ( Figure 2A). qPCR was used to validate the RNAseq data (Figure 2B-C; Figure 2-figure supplement 1A). Injury-dependent reduction in tgfb3 RNA was also evident following NMDA-mediated amacrine and ganglion cell death or genetic ablation of photoreceptors using metronidazole-treated zop:nsfb-EGFP transgenic fish that harbor a zebrafish rod opsin promoter driving expression of bacterial nitroreductase that converts metronidazole into a cytotoxic product (Figure 2-figure supplement 1B; Montgomery et al., 2010;Powell et al., 2016).
We also compared the relative expression of tgfb ligand encoding RNAs in the retinal neuron (GFP-) and MG (GFP+) cell populations that were separated by FACS using dissociated cells from uninjured gfap:GFP fish retinas. This analysis showed tgfb3 is highly enriched in MG, whereas tgfb1b and tgfb2 are more equally distributed between these different cell populations (Figure 2-figure supplement 1C). In situ hybridization assays for tgfb3 combined with glutamine synthetase (GS) immunofluorescence on retinal sections confirmed tgfb3 gene expression is restricted to GS+ MG in the uninjured retina ( Figure 2D).
We next used in situ hybridization assays to investigate when during development this MG-specific tgfb3 expression profile emerges. Consistent with a previous report (Cheah et al., 2005), we found transient expression of tgfb3 RNA in the lens at 24 hpf (hours post fertilization) ( Figure 3A). At 96 hpf when MG are differentiating (Bernardos et al., 2005), we do not detect significant levels of tgfb3 RNA in the retina, nor do we observe significant tgfb3 expression at 7 dpf (days post fertilization). By 10 dpf, tgfb3 RNA is detected in the central region of the retina where more mature MG reside ( Figure 3A). This expression continues to increase with age and the adult expression pattern is established by 3 mpf (months post fertilization) ( Figure 3A). Consistent with our RNAseq and qPCR data indicating very low tgfb1b levels in the retina ( Figure 1A-B), we were unable to detect zebrafish tgfb1b gene expression by in situ hybridization at any of the time points examined ( Figure 3A).
We next examined Tgfb3 expression in the 1 month old mouse retina. This analysis showed Tgfb3 expression was confined to cells in the GCL and INL ( Figure 3B). Co-staining retinal sections for Tgfb3 RNA and Sox9 immunofluorescence (MG marker) showed no overlap, indicating Tgfb3+ cells in the INL are not MG ( Figure 3B).   Although the lack of detectable Tgfb3 expression in the developing (1-7 dpf) fish retina suggests Tgfb3 expression is not necessary for MG differentiation, low levels of expression may go undetected by in situ hybridization. Therefore, we investigated if knocking down Tgfb3 expression with a translation-blocking morpholino-modified antisense oligonucleotide (MO) would affect MG differentiation. Control or tgfb3-MOs were delivered into single cell zebrafish embryos and at 6 dpf, fish were sacrificed and retinal sections assayed for glutamine synthetase (GS) immunofluorescence which serves as a marker of differentiated MG. This analysis revealed normal GS reactivity in the developing retina regardless of Tgfb3 knockdown (Figure 3-figure supplement 1A). We confirmed the tgfb3-MO's effectiveness by observing reduced GFP in embryos injected with a tgfb3-EGFP fusion RNA and the tgfb3-MO, but not with control MO (Figure 3-figure supplement 1B).
Because of concerns that MO-mediated Tgfb3 knockdown may not be effective at 6 dpf due to MO dilution during cell division, we generated tgfb3 -/fish using a CRISPR/Cas9 strategy (Figure 3figure supplement 1C-E). In these fish, an insertion mutation changes the tgfb3 reading frame so a premature stop codon is introduced in exon 1 (Figure 3-figure supplement 1E). We found that tgfb3 -/fish die around 2 weeks post fertilization. Therefore, we selected tgfb3 +/+ , tgfb3 +/and tgfb3 -/fish at 12 dpf to assay for retinal GS expression. Normal GS immunofluorescence was observed in both tgfb3 +/and tgfb3 -/fish retinas suggesting Tgfb3 does not impact MG differentiation (Figure 3-figure supplement 1F). However, we do note that retinas were smaller in tgfb3 -/fish.
We next used in situ hybridization assays to investigate the spatial pattern of tgfb3 expression in the injured adult retina. This analysis revealed that tgfb3 RNA is specifically suppressed at the injury site and that this suppression preceded MG proliferation that begins~2 dpi ( Figure 3C).
Together, the above results indicate that tgfb3 expression in the zebrafish retina correlates with MG maturation, pSmad3 expression, and MG quiescence.

Injury-dependent Tgfb3 suppression is required for MG proliferation
To investigate if Tgfb3 suppression was required for injury-dependent MG proliferation, we generated hsp70:tgfb3 transgenic fish that allow conditional expression of Tgfb3 with heat shock. A 1 hr heat shock at 37˚C resulted in over a 100-fold induction of tgfb3 RNA that persisted for over 3 hr and then returned to basal levels around 24 hr post heat shock (Figure 4-figure supplement 1A).
To examine the effect Tgfb3 had on MG proliferation in the injured retina, hsp70:tgfb3 transgenic fish received a needle poke injury and then a 1 hr heat shock every 6 hr for 4 days. Three hours before sacrifice at 4 dpi, fish received an IP (intraperitoneal) injection of BrdU to label proliferating cells. This analysis showed that forced expression of Tgfb3 suppressed MG proliferation in the injured retina ( Figure 4A We next investigated if Tgfb3 depletion affected spontaneous or injury-dependent MG proliferation. Because tgfb3 -/fish do not survive to adults, we knocked down Tgfb3 with a tgfb3-targeting MO and determined if there was an effect on MG proliferation. Control or tgfb3-MO was delivered intravitreally and cellular uptake facilitated by electroporation as previously described (Thummel et al., 2011). Retinas were injured and at 3 dpi fish received an IP injection of EdU 3 hr prior to sacrifice. Quantification of EdU+ and TUNEL+ cells revealed Tgfb3 knockdown had no effect on MG proliferation or cell death at either the injury site or in undamaged regions of the retina ( Together, the above data indicate Tgfb3 suppression is necessary for injury-dependent MG proliferation, but this suppression is not sufficient to drive MG proliferation in the uninjured retina.

Tgfb3 regulates MG reprogramming
Reprogramming MG for retinal repair requires the regulation of gene expression programs that stimulate MG proliferation and the expansion of a MG-derived progenitor population. Essential    Figure 4 continued on next page components of these programs are genes encoding growth factors, cytokines, and transcription factors that are rapidly induced following a retinal injury (Gorsuch et al., 2017;Nagashima et al., 2020;Nelson et al., 2013;Nelson et al., 2012;Ramachandran et al., 2010a;Ramachandran et al., 2011;Ramachandran et al., 2012;Thummel et al., 2010;Wan and Goldman, 2017;Wan et al., 2012;Wan et al., 2014;Zhao et al., 2014). Because tgfb3 expression is suppressed within a few hours post injury, we suspected it might regulate these gene expression programs. To investigate this, we injured Wt or hsp70:tgfb3 fish retinas and immediately treated fish with a single 1 hr heat shock before sacrificing fish at 6 hpi and isolating retinal RNA for gene expression analysis using qPCR. Interestingly, we found that many reprogramming genes that are normally induced after retinal injury, like ascl1a, hbegfa, lepb, crlf1a, and socs3a, were suppressed by Tgfb3 expression, while genes normally repressed soon after injury, like fgf8a and dll4 were unaffected ( Figure 4C). Furthermore, forced Fgf8a expression, which we previously showed suppressed injury-dependent MG proliferation in the adult retina (Wan and Goldman, 2017), had no effect on tgfb3 expression ( Figure 4D). Thus, Tgfb3 and Fgf8a appear to act independent of each other to regulate MG proliferation.
The observation that ascl1a gene expression is suppressed in a Tgfb3-dependent manner was intriguing since it is a critical reprogramming gene that along with Lin28a and Notch inhibition is sufficient to stimulate spontaneous MG proliferation in the uninjured retina (Elsaeidi et al., 2018). We previously reported that injury-dependent induction of cytokines, like Crlf1a and Lepb, acting via gp130-coupled receptors, also act upstream of ascl1a to drive MG proliferation . Therefore, we wondered if Tgfb3 was part of this early cytokine response system. For this analysis we took advantage of previously characterized MOs to knockdown cytokine signaling in the injured retina and assayed ascl1a and tgfb3 gene expression ( Figure 4E; Zhao et al., 2014). This analysis confirmed that injury-dependent ascl1a gene induction is regulated by a cytokine signaling system, and that this signaling system also contributes to injury-dependent tgfb3 suppression ( Figure 4E). Thus, Tgfb3 and cytokine-related gene products seem to contribute to a feedback loop regulating each other's expression and also regulating MG proliferation.
Together, these data suggest Tgfb3 drives MG quiescence in the injured retina, at least in part, by inhibiting the expression of pro-regenerative gene expression programs.
Tgfb1b overexpression stimulates pSmad3 expression without affecting MG proliferation qPCR and RNAseq analysis of tgfb gene expression in the injured retina revealed that although tgfb1b and tgfb2 were induced in the injured retina, their levels remained below the suppressed levels of tgfb3 ( Figure 2A). Furthermore, injury-dependent regulation of pSmad3 expression followed that of tgfb3, but not tgfb1a, tgfb1b, or tgfb2 ( Figures 1B, 2A and 3C). These observations suggest that Tgfb3 expression largely accounts for pSmad3 expression in the uninjured and injured retina. Whether other Tgfb ligands could also inhibit MG proliferation via a pSmad3 signaling mechanism remained unknown, but seemed likely if we boosted their levels to at least that of Tgfb3 in the uninjured retina. To investigate this, we generated hsp70:tgfb1b transgenic fish.
hsp70:tgfb1b fish treated with a 1 hr heat shock exhibited over a 100-fold induction of tgfb1b RNA that returned to basal levels~24 hr later ( Figure 5-figure supplement 1A). Overexpression of Tgfb1b had no significant effect on endogenous tgfb2b expression, but did decrease tgfb3 expression by~50% at 6 hr post heat shock ( Figure 5-figure supplement 1B). Interestingly, when comparing the consequence of Tgfb1b and Tgfb3 overexpression on MG proliferation and pSmad3 expression, only Tgfb3 suppressed injury-dependent MG proliferation ( Figure 5A-B), but both Tgfb1b and Tgfb3 reestablished pSmad3 expression at the injury site ( Figure 5C). Although Tgfb1b overexpression did not regulate MG proliferation, tgfb1b RNA is induced in the injured retina (Figure 2A and C), and we wondered if this induction had any consequence on regeneration. Therefore, we knocked down Tgfb1b expression with a previously validated splice-blocking tgfb1b-targeting MO ( Figure 5-figure supplement 1C; Monteiro et al., 2016). MOs were delivered to retinas at the time of injury and cellular uptake facilitated by electroporation. Four days later, fish received an IP injection of EdU and 3 hr later some fish were sacrificed to assay MG proliferation ( Figure 5-figure supplement 1D), while others were allowed to survive until 14 dpi to lineage trace EdU+ MG ( Figure 5-figure supplement 1E). These analyses revealed that Tgfb1b knockdown had no significant effect on injury-dependent MG proliferation or the fate of MG-derived progenitors ( Figure 5-figure supplement 1D-E).
Together, the above data indicate specificity in the action of Tgfb ligands on MG proliferation and suggest that pSmad3 expression does not drive MG quiescence in the injured retina. Alk5 and PP2A inhibition rescue MG proliferation in injured retinas with Tgfb3 overexpression The above data suggested that Tgfb3-dependent activation of Alk5 may engage a non-canonical Tgfb signaling pathway to regulate MG proliferation. To investigate this, we first determined the effect Alk5 suppression had on MG proliferation in heat shock-treated hsp70:tgfb3 fish immersed in either DMSO or the Alk5 inhibitor, SB431542 ( Figure 6A). As expected, Alk5 inhibition rescued MG proliferation in injured retinas from heat shock-treated hsp70:tgfb3 fish ( Figure 6B-C).
We next investigated if PP2A or p38 signaling participated in Tgfb3-dependent regulation of MG proliferation in the injured retina. For this analysis, Wt and hsp70:tgfb3 transgenic fish retinas were injured and fish were either left untreated or received a 1 hr heat shock every 6 hr for 4 days along with daily intravitreal injections of either DMSO, the PP2A inhibitor okadaic acid (10 mM, OKA), or the p38 MAPK inhibitor PD169316 (3 mM, PD) (Bialojan and Takai, 1988;Wilson et al., 1997). Fish received an IP injection of EdU 3 hr prior to sacrifice to label proliferating cells. Quantification of EdU+ cells revealed that inhibition of PP2A, but not p38 MAPK, rescued MG proliferation in retinas with Tgfb3 overexpression ( Figure 6C). Neither forced expression of Tgfb3 or intravitreal injection of OKA affected injury-dependent apoptosis ( Figure 6-figure supplement 1B).
Together, the above data suggests Alk5 and PP2A act downstream of Tgfb3 to regulate MG quiescence.

Notch inhibition rescues MG proliferation in Tgfb3-expressing injured retinas
Previous studies revealed that Notch signaling inhibition is required for injury-dependent MG proliferation (Conner et al., 2014;Elsaeidi et al., 2018;Wan and Goldman, 2017;Wan et al., 2012). Therefore, we wondered if Tgfb3 signaling acted through Notch signaling to inhibit MG proliferation. To investigate this, we bred hsp70:tgfb3 fish with tp1:mCherry Notch reporter fish that harbor 12 RBP-Jk binding sites upstream of a minimal promoter that drives nuclear localized mCherry expression (Parsons et al., 2009). Normally, Notch signaling is suppressed in injury-responsive MG spanning the injury site ( Figure 7A, tp1:mCherry panel) (Elsaeidi et al., 2018;Wan and Goldman, 2017). However, forced Tgfb3 expression in hsp70:tgfb3;tp1:mCherry double transgenic fish prevented this injury-dependent suppression in Notch signaling ( Figure 7A-B) and this correlated with reduced MG proliferation ( Figure 4A-B). Furthermore, we found that Tgfb3-dependent inhibition of MG proliferation can be rescued by pharmacological suppression of Notch signaling using the g-secretase inhibitor RO4929097 ( Figure 7C-D). Consistent with these data, Tgfb3 overexpression in the uninjured retina increased expression of the Notch reporter gene hey1, and the ligand encoding gene, dll4. However, dll4 induction did not reach statistical significance ( Figure 7E) and further studies are needed to determine if Tgfb3 regulates Notch signaling via dll4 expression. Nonetheless, the above studies suggest Tgfb3 acts, at least in part, via the Notch signaling pathway to regulate MG quiescence.

Discussion
Here we report that Tgfb3 signaling regulates MG quiescence in the zebrafish retina. Although both Tgfb1b and Tgfb3 can induce pSmad3 expression, only Tgfb3 stimulates MG quiescence, suggesting the involvement of a non-canonical Tgfb signaling pathway. We found that PP2A or Notch inhibition partially rescued MG proliferation in injured retinas overexpressing Tgfb3 and that Tgfb3 acts, at least in part, by stimulating Notch signaling. We also found that Tgfb3-driven MG quiescence is associated with suppression of regeneration-associated genes. Finally, our study reveals that tgfb3 is highly expressed by pro-regenerative MG of the zebrafish retina, but remains undetectable in nonregenerative MG of the mouse retina.
Tgfb ligands are expressed as latent pre-pro-polypeptides that must be released from latency for their action (Shi et al., 2011). Prior to secretion, the immature polypeptide is cleaved between the prodomain and mature peptide domain, but they remain associated. The prodomain is required for proper folding, dimerization, and binding to integrin in the extracellular matrix. In order for Tgfb ligands to signal through their receptors they must be released from the prodomain, which is accomplished via integrin interaction. Among the four different Tgfb ligand encoding genes expressed in the uninjured retina, our data suggests that Tgfb3 predominates. Tgfb3 expression is restricted to quiescent MG and its suppression is required for MG proliferation. Our data suggest that Tgfb3 is constitutively released from integrin in the extracellular matrix and that tgfb3 gene expression is a major control point in regulating Tgfb signaling in MG.
The mechanism by which tgfb3 expression is suppressed in the injured retina remains unknown, but diffusible factors emanating from either dying neurons and/or immune cells that accumulate at the injury site are good candidates. The observation that both the ligand (Tgfb3) and the response (pSmad3) are restricted to MG suggests that Tgfb3 acts in an autocrine or paracrine fashion; however, we cannot rule out the involvement of an intervening neuronal or immune cell. An autocrine/ paracrine type of regulation may be a common theme in the injured zebrafish retina since most of the reported secreted factors regulating MG proliferation emanate from MG themselves (Nelson et al., 2013;Wan and Goldman, 2017;Wan et al., 2012;Zhao et al., 2014). In the adult mammalian retina, Tgfb signaling is low and Tgfb expression does not correlate with MG quiescence or proliferation (Anderson et al., 1995;Kugler et al., 2015;Lutty et al., 1993;Tosi et al., 2018). However, similar to what we observed in adult fish retina, Tgfb signaling has been associated with reduced MG proliferation in young chicks and postnatal rats (Close et al., 2005;Todd et al., 2017), and with reduced stem cell proliferation in certain regions of the mouse brain (Falk et al., 2008). The high basal level of tgfb3 in mature MG from zebrafish, but not mice, is intriguing and one cannot help but speculate that it may contribute to their stemness. Interestingly, Notch signaling has been reported to contribute to neural stem cell stemness in the adult zebrafish brain (Than-Trong et al., 2018), and in the retina, our data indicates Tgfb3 stimulates Notch signaling. Thus, Tgfb3 expression in the zebrafish retina provides a mechanism for maintaining Notch signaling into adulthood, which is absent in mammals.
Our conclusion that Tgfb3-pSmad3 signaling is active in quiescent MG and suppressed following retinal injury differs from previous reports (Conedera et al., 2020;Lenkowski et al., 2013;Sharma et al., 2020;Tappeiner et al., 2016). However, even among these previous studies, inconsistencies emerge, reinforcing the importance of carefully controlled experiments. Lenkowski et al., 2013 assayed the expression of putative Tgfb responsive genes, like tgif1 and tgfbi to concluded Tgfb signaling was transiently induced prior to injury-dependent MG proliferation and then repressed when MG proliferate. In contrast, Tappeiner et al., 2016 used pSmad3 immunofluorescence to conclude Tgfb signaling was increased in proliferating MG. Further disparities emerge when examining the effect of the Tgfb signaling inhibitor SB431542 on MG proliferation in the INL where Tappeiner et al., 2016 reported no effect, and Sharma et al., 2020 reported inhibition. Remarkably, even in work coming from the same group, inconsistencies emerge. For example, when assaying tgfb gene expression, one report suggests increased injury-dependent tgfb1a expression, while another indicated reduced tgfb1a expression (Conedera et al., 2020;Tappeiner et al., 2016); and another group reported Tgfb signaling is inhibited in an Oct4-dependent fashion in injuryresponsive MG, while a later report indicates enhanced Tgfb signaling in these reprogrammed MG Sharma et al., 2020). The reason for these disparities is not known, but reinforces the idea that manipulating and assaying Tgfb signaling in the zebrafish retina is not trivial. We also note that many of these studies relied solely on SB431542 to inhibit Tgfb signaling; however, this drug is a more potent inhibitor of CK1 and RIPK2 (Vogt et al., 2011), which further clouds the interpretation of results.
In the work reported here, we not only suppressed Tgfb signaling with SB431542 and the more specific Alk5 inhibitor, SB505124 (Vogt et al., 2011), but also stimulated Tgfb signaling with conditional expression of Tgfb1b, Tgfb3, and ca-Alk5 using transgenic fish. All these manipulations resulted in the expected change in pSmad3 immunofluorescence in MG, supporting our contention that pSmad3 immunofluorescence reflects Tgfb signaling. Importantly, we found that Tgfb signaling is active in quiescent MG and rapidly suppressed in injury-responsive MG, and that this suppression correlates with tgfb3 gene expression that was visualized by in situ hybridization and quantified by RNAseq and qPCR. Furthermore, unlike previous studies, we determined the relative proportion of RNAs encoding different tgfb isoforms, revealing tgfb3 is expressed at least 80-fold higher than the other isoforms in quiescent MG, and the only isoform to be repressed after retinal injury. These data suggest that Tgfb3 is largely responsible Tgfb signaling in quiescent MG. Remarkably, Conedera et al., 2020 reported a~2 fold increase in tgfb3 expression at 1-14 dpi, which is inconsistent with their previous study (Tappeiner et al., 2016), and hard to reconcile with our data.
Using transgenic fish, we found that forced expression of Tgfb3, but not Tgfb1b, suppressed injury-dependent MG proliferation. This result contrasts with Sharma et al., 2020 who reported intravitreal injection of recombinant human Tgfb1 enhanced MG reprogramming and proliferation in the injured. Zebrafish Tgfb1b shares 42% amino acid identity with human TGFb1, and it is not clear if the human ligand will engage zebrafish Tgfb receptors and stimulate pSmad3 expression in the zebrafish retina. Although transgenic approaches used to conditionally express zebrafish ligands are generally preferable to intravitreal injection of human factors, these transgenic approaches do have their limitations since heat shock causes cell stress, and the hsp70 promoter is induced in all retinal cell types. Our studies suggest that heat shock-induced cell stress had no effect on MG proliferation in wild type and hsp70:tgfb1b fish; however, we were unable to achieve conditional, cell-type specific expression of Tgfb3, so it remains possible that some of its actions are indirectly related to its expression in retinal neurons.
One of the more remarkable observations we made during this study is that bothTgfb3 and Tgfb1b can stimulate pSmad3 expression, but only Tgfb3 can drive MG quiescence. This suggested that Tgfb3 must be mediating if effects on MG proliferation via a non-canonical Tgfb signaling pathway. Although the mechanism coupling Tgfb3-dependent activation of Alk5 to MG quiescence remains unknown, our data indicates that it impacts Notch signaling as indicated by increased tp1: mcherry transgene and endogenous hey1 expression.
Experiments designed to rescue MG proliferation in Tgfb3 overexpressing retinas revealed a role for Alk5 and PP2A in regulating MG quiescence. Although it is not surprising that inhibition of Tgfb3's Alk5 receptor would relieve its effects on MG proliferation, the observation that this phenotype is recapitulated by PP2A inhibition was unexpected. Interestingly, PP2A can be recruited to Alk5 to stimulate G1 arrest (Griswold-Prenner et al., 1998;Petritsch et al., 2000;Wlodarchak and Xing, 2016), and PP2A is a major regulator of cell cycle check points and signaling pathways that impinge on the cell cycle, like Wnt, MAPK, PI3K, and mTor (Wlodarchak and Xing, 2016). Importantly, these PP2A-regulated signaling pathways have been previously shown to contribute to injurydependent MG proliferation (Ramachandran et al., 2011;Wan et al., 2012;Wan et al., 2014;Zelinka et al., 2016). However, additional studies are needed to determine the mechanism of action of PP2A in regulating MG quiescence and in particular, whether it directly interacts with Alk5. Animals, injury models, and cell proliferation assays Animal studies were approved by the University of Michigan's Institutional Animal Care and Use Committee. Zebrafish were kept at 26-28˚C with a 10/14 hr light/dark cycle. Adult male and female fish from 6 to 12 months of age were used in these studies. 1016 tuba1a:GFP, gfap:GFP, tp1: mCherry, zop:nsfb-EGFP, and hsp70:ca-Alk5 fish were previously described (Fausett and Goldman, 2006;Kassen et al., 2007;Montgomery et al., 2010;Parsons et al., 2009;Zhou et al., 2011). We generated hsp70:tgfb1b and hsp70:tgfb3 transgenic fish using standard recombinant DNA techniques using Tol2 vector backbone. Expression constructs were injected into single cell zebrafish embryos as previously described (Fausett and Goldman, 2006). Fish were anesthetized in tricaine and retinas were injured with a needle poke injury (2-4 injuries/retina for analysis of proliferation and protein expression on retinal sections and 8-10 injuries/retina when harvesting total RNA for qPCR), NMDA, or genetically as previously described (Fausett and Goldman, 2006;Montgomery et al., 2010;Powell et al., 2016). To investigate cell proliferation, fish received an IP injection of BrdU or EdU (10 ml of 10 mg/ml stock) as indicated in the text and detected by immunofluorescence or Click-It chemistry as previously described (Wan and Goldman, 2017). Wild-type FVB/N mice were obtained from our breeding colony.

RNA isolation, PCR, and RNAseq
Total RNA was isolated using Trizol (Invitrogen). cDNA synthesis and PCR reactions were performed as previously described (Fausett et al., 2008;Ramachandran et al., 2010a). Real-time qPCR reactions were carried out in triplicate with ABsolute SYBR Green Fluorescein Master Mix (Thermo Scientific) on an iCycler real-time PCR detection system (BioRad). The DDCt method was used to determine relative expression of mRNAs in control and injured retinas and normalized to either gapdh or gapdh s mRNA levels. Individual comparisons were done using unpaired 2-tailed Student t-test. ANOVA with Fisher's PLSD post hoc analysis was used for multiple parameter comparison. Error bars are standard deviation (SD). For RNAseq, retinas from 1016 tuba1a:GFP and gfap:GFP fish were dissociated and GFP+ MG were purified using FACS using the University of Michigan's Cell Sorting Core as previously described (Powell et al., 2013;Ramachandran et al., 2010a). RNA from GFP+ cells was used to generate libraries and DNA was sequenced on an Illumina HiSeq2000 instrument. Sequencing reads were analyzed by the University of Michigan's Bioinformatics Core. The number of reads for each expressed gene was determined and differentially expressed genes were restricted to those exhibiting at least a 2-fold difference in expression with threshold abundance greater than 5 Fragments Per Kilobase of transcript per Million mapped reads to eliminate very low abundant transcripts whose estimates of fold-change are unreliable. GEO accession for RNAseq data is GSE145330.

Morpholino (MO) functional assays
For testing tgfb3-targeting MO, we generated pCS2+tgfb3 EGFP construct that contained tgfb3 cDNA MO target site upstream and in-frame with the EGFP initiator AUG. Primers for generating tgfb3 MO targeting sequence are: forward primer (BamHI): 5'-GCAGGATCCGGAGCCGCTTCA TTCATCTG-3' and reverse primer (Nco1): 5'-TCACCATGGTGGACAGAGACAAGCTCATG-3'. The pCS2+tgfb3 EGFP plasmid was linearized with Not1 restriction enzyme and capped sense tgfb3-EGFP RNA was synthesized using SP6 RNA polymerase using Invitrogen's mMESSAGE mMACHINE SP6 Transcription Kit (Invitrogen, #AM1340) according to manufactures directions. Following purification, the capped tgfb3-EGFP RNA was dissolved in nuclease free water containing 0.2% phenol red and injected with experimental or control MO into single cell zebrafish embryos. Each embryo received approximately 50 pg of RNA and 250 pg of control or experimental MO. For testing splice blocking tgfb1b-targeting MO, we injected either control MO (2.4 ng) or tgfb1b-targeting MO (0.3 ng and 2.4 ng) into single cell zebrafish embryos. RNA was extracted from embryos at 24-48 hr post injection and assayed for tgfb1b mRNA by PCR.

Heat shock and pharmacological inhibitors
For heat shock, fish were immersed in a water bath at 37˚C for 1 hr before returning to system water at 28˚C. For extended periods of heat shock, this was repeated every 6 hr. For inhibition of Tgfb signaling we used two different Tgfb receptor 1 (Alk 5) inhibitors, SB431542 and SB505124 (Fisher Scientific) and to inhibit Notch signaling we used RO4929097 (Cayman). PP2A was inhibited with okadaic acid (Cell Signaling Technology) and p38 MAPK was inhibited with PD169316 (Cayman Chemical). Pharmacological reagents were prepared in DMSO as a 10 mM stock and diluted 1/200 in fish water for immersion or PBS for intravitreal injections. Control fish were treated with vehicle.

Microscopy, TUNEL, cell quantification, and statistical analysis
BrdU and EdU labelling were used to identify and quantify proliferating cells in retinal sections as previously described (Fausett and Goldman, 2006;Ramachandran et al., 2010a;Wan and Goldman, 2017;Wan et al., 2012;Wan et al., 2014). TUNEL assays were performed on retinal sections using the in situ cell death, fluorescein kit (Sigma Aldrich, Cat # 11684795910). Images were captured by a Zeiss Axiophot fluorescence microscope or a Leica DM2500 microscope. Autofluorescence is defined as variable, background fluorescence that can be detected in multiple channels. All experiments were done in triplicate (three trials) with three animals per trial unless otherwise indicated. We routinely quantify the total number of proliferating cells around the injury site in all nuclear layers and also the number of proliferating cells restricted to the INL in order to be sure proliferative changes represent MG proliferation and are not solely due to rod progenitor proliferation. Unless specifically stated, quantification shown in graphs is proliferation in all nuclear layers at the injury site. Quantification of pSmad3 immunofluorescence was restricted to the area surrounding the injury site and represents the field of view. Error bars are standard deviation (SD). ANOVA with Fisher's PLSD post hoc analysis was used for multiple parameter comparison; two-tailed Student's t test was used for single parameter comparison.