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 (Figure 1—figure supplement 1A; Vogt et al., 2011).

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.

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–B), without any significant effect on injury-dependent cell death (Figure 6—figure supplement 1B).

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 (Figure 4—figure supplement 1B–C).

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.

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 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 (Zhao et al., 2014). 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.

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.

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 Tgfb3-dependent activation of Alk5 engaged a non-canonical Tgfb signaling pathway, like ERK, PI3K, p38, and PP2A (Derynck and Zhang, 2003; Petritsch et al., 2000; Yu, 2002; Zhang, 2017). Previous studies indicated that ERK and PI3K signaling are necessary for injury-dependent MG proliferation and therefore, are not candidates for conveying the quiescence-promoting effects of Tgfb3 (Wan et al., 2012; Wan et al., 2014). Therefore, we focused our analysis on PP2A and p38.

PP2A functions as a trimer with a dimeric core consisting of a catalytic (zebrafish genes ppp2ca, ppp2cb) and structural subunit (zebrafish genes ppp2r1a, ppp2r1b) and one of 7 regulatory subunits (zebrafish genes ppp2r2aa, ppp2r2ab, ppp2r2ba, ppp2r2bb, ppp2r2ca, ppp2r2cb, ppp2r2d) that confers subcellular targeting, substrate specificity, and regulation of holoenzyme phosphatase activity. Importantly, the PP2A Bα subunit (zebrafish Ppp2r2aa and Ppp2r2ab) can associate with and be phosphorylated by Alk5, and this association/phosphorylation is necessary to transduce the anti-proliferative effects of Tgfb receptor signaling (Griswold-Prenner et al., 1998; Petritsch et al., 2000; Wlodarchak and Xing, 2016). Interrogation of RNAseq data sets indicates that genes encoding PP2A components and p38 isoforms (mapk14a, mapk14b) are expressed in quiescent and injury-responsive MG (Figure 6—figure supplement 1A).

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 μM, OKA), or the p38 MAPK inhibitor PD169316 (3 μM, 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.

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 γ-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.

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 μl 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.

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 ΔΔCt 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.

Gene editing was performed as previously described (Hwang et al., 2013; Vejnar et al., 2016). Briefly, CRISPRscan (https://www.crisprscan.org/) was used to identify gRNA target sequences in exon 1 of the tgfb3 gene. gRNAs were transcribed using PCR products as a templates and the MEGAscript T7 transcription kit (Thermo Fisher Scientific #AM1334). Cas9-nanos mRNA was transcribed using NotI-digested Cas9-nanos pCS2 expression vector and the mMESSAGE mMACHINE SP6 kit (Invitrogen, #AM1340). Cas9-nanos mRNA and two gRNAs targeting tgfb3 exon1 were co-injected into one cell stage zebrafish embryos. Primers for making tgfb3 mutant fish: T7 universal gRNA primer: 5’-AAAAGCACCGACTCGGTGCCACTTTTTCAAGTTGATAACGGACTAGCCTTATTTTAACTTGCTATTTCTAGCTCTAAAAC-3’; gRNA one primer: 5’-TAATACGACTCACTATAGGGCACCTGACTAGGGCCCAGTTTTAGAGCTAGAA; gRNA two primer: 5’-TAATACGACTCACTATAGGCCCTCTACAACAGCACCAGTTTTAGAGCTAGAA.

The following PCR primers are 5’ to 3’. tgfb3: forward GATTGGAGGGACGGATGA, reverse GTGACAGGGGCAGTGAAC; tgfb2: forward CAGCATGAGAGCCACAGAC, reverse CTCCACAGATACGGACAGG; tgfb1a: forward GTACAAACACCACAACCCTGG, reverse GGCTTACTTATCAATCCCGAC; tgfb1b: forward ACTGGCTCTTGCTCCTAT, reverse AACTGTTCCACCTTATGC; ascl1a: forward TTGAGCGTTCGTAAA, reverse GCTGAAGGACTGGATT; fgf8a: forward CAGTGTGGATACAAACGCAGG, reverse TAGCAAAACGCAAAGAGGTGA; lepb: forward CATTGCTCGAACCACCATCAGC, reverse TCTTTATGCACCGGGGTCTCG; crlf1a: forward GGGATTCTGGGATCTAGGAAAGC, reverse TCCTTGAAGAACCTGGTTGCG; socs3a: forward CACTAACTTCTCTAAAGCAGGG, reverse GGTCTTGAAGTGGTAAAACG; il-11b: forward GCTAACAGTGTCGCCTGACTCC, reverse CTGTAGTTCAGTGAGGGCAGGG; dll4: forward GGAAATTTGACGTGCTCCAT, reverse GAGAAAGGTGAGCCAAGCTG; hbegf: forward CGATGGATGGCGAGGATGTAGA, reverse GCATTAGGGCAGGACGAAGTTG; hey1: forward GTTTGCATTTTCACGCCCCT, reverse CGCCCTCTAGTGCTCACATT; tgfb3 gRNA mutation forward GGCAAAGGACTGCTGTTTGT, reverse GAGATCCCTGGATCATGTTGA; tgfb1b MO mis-splicing forward GCACACCATAGAAGATCCAACA, reverse AGGCATCTGCAACCAGTCTT.

Lissamine-tagged, tgfb3-targeting MO: 5’TGCATGGTTAATATCTGCACACTAT; Lissamine-tagged, tgfb1b-targeting MO: 5’AAGGATAGTGCCACTCACTCATTGT; Gene Tools standard control MO: 5’ CCTCTTACCTCAGTTACAATTTATA.

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’-GCAGGATCCGGAGCCGCTTCATTCATCTG-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.

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.

Zebrafish samples were prepared for immunofluorescence as previously described (Fausett and Goldman, 2006; Ramachandran et al., 2010a; Ramachandran et al., 2010a). Primary antibodies used in this study: anti-pSmad3, Abcam Cat. # ab52903 (1/200); Zpr-1 and Zn-5, Zebrafish International Resource Center (1/500 and 1/1000, respectively); anti-HuC/D, Abcam, Cat. #ab210554 (1/500); anti-PKC_β1, Santa Cruz Biotechnology, Cat. #SC-8049 (1/200); anti-glutamine synthetase (GS), Sigma-Aldrich, Cat. #MAB302 (1/500); anti-SOX9, EMD Millipore, Cat. #AB5535 (1/500); anti-BrdU, Thermo Fisher, Cat. # MA 1–82088 (1/500) and Cat. # B35128 (1/500, clone MoBu-1 for co-staining with EdU Click-It chemistry). Secondary antibodies: Alexa Flour 555 Donkey anti Mouse-IgG (H+L), Thermo Fisher Cat. # A31570 (1:500); Alexa flour 555 Donkey anti Rabbit IgG (H+L), Thermo Fisher, Cat # A31572 (1:500); Alexa flour 555 Donkey anti Sheep IgG (H+L) Thermo Fisher Cat #A21436. Cy3, Jakson Immuno research labs, Cat #712-166-150 (1:500); Alexa Flour 488 donkey anti mouse Thermo Fisher Cat. # A21202 (1:500); Alexa Flour 488 goat anti rabbit Thermo Fisher Cat. # A11008 (1:500); Cy5 goat anti mouse, Thermo Fisher Cat. # A10524 (1:500); and Alexa Flour 647 goat anti rabbit Thermo Fisher Cat # A21244 (1:500). In situ hybridization was performed as described previously (Barthel and Raymond, 2000).

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.