Inhibition of GABAA-ρ receptors induces retina regeneration in zebra fish

Abstract A potential treatment for retinal diseases is to induce an endogenous Müller glia (MG)-derived regenerative response to replace damaged neurons. In contrast to mammalian MG,zebra fish MG are capable of mediating spontaneous regeneration. We seek to de fine the mechanisms that enable retina regeneration in zebra fish in order to identify therapeutic targets to induce mammalian retina regeneration. We previously used pharmacological and genetic methods to inhibit gamma aminobutyric acid A (GABAA) receptors in undamaged zebra fish retinas and showed that such inhibition could induce initiation of retina regeneration, as measured by the dedifferentiation of MG and the appearance of MG-derived proliferating progenitor cells. Here, we show that inhibition of a pharmacologically distinct subset of GABAA receptors (GABAA-ρ) can also induce retina regeneration. Dual inhibition of both GABA receptor subtypes led to enhanced retina regeneration. Gene expression analyses indicate that inhibition of GABAA-ρ receptors induces a canonical retinal regenerative response. Our results support a model in which decreased levels of GABA, such as would occur after retinal cell death or damage, induce dedifferentiation of MG and the generation of proliferating progenitor cells during zebra fish retina regeneration. Animal experiments were approved by the Vanderbilt’s Institutional Animal Care and Use Committee (Protocol M1800200) on January 29, 2019.

Key Words: gamma aminobutyric acid; morpholino; Müller glia; neurotransmitter;regeneration; retina; stem cells; zebrafish

Introduction

The health and overall economic consequences of vision loss, whether due to injury or disease, is significant with 2014 estimates of annual costs by Prevent Blindness of $145 billion, a number that will certainly increase as life expectancy increases (Wittenborn and Rein, 2014). This has led to a concerted effort to identify therapies to restore loss of vision or reduce the effects of degenerative retinal disorders. One current treatment involves intravitreal injection of either stem cells or retinal precursor cells, but these treatments are not yet capable of fully restoring vision (Barber et al., 2013; Hanus et al., 2016; Gonzalez-Cordero et al., 2017; Stern et al., 2018).Recent gene therapy approaches have been used for delivery or overexpression of factors to induce retina regeneration or to restore expression of defective genes, but in certain cases this approach is limited to those diseases where the exact defective gene is known (Jorstad et al., 2017; Russell et al.,2017; Yao et al., 2018). An alternative to these approaches is to induce damaged or diseased retinas to undergo regeneration using resident adult stem cells. Zebrafish retinas undergo spontaneous retina regeneration in response to damage (Wan and Goldman, 2016). In contrast, the mammalian retina does not naturally regenerate, more often responding to damage via reactive gliosis (Bringmann et al.,2006). Intriguingly, the adult stem cell that is responsible for regeneration in zebrafish (Müller glia; MG)(Bernardos et al.,2007) is present in the mammalian retina, but for unknown reasons, MG-derived regeneration is blocked in mammals,possibly related to the hippo pathway (Rueda et al., 2019). In fish, MG respond to damage by dedifferentiation, asymmetric division, and the generation of proliferating neuronal progenitor cells which can then migrate and differentiate into any lost retinal cell type (Wan and Goldman, 2016).

Gamma aminobutyric acid (GABA) is the main inhibitory neurotransmitter in the central nervous system (Roberts and Kuriyama, 1968; Farrant and Nusser, 2005). Beyond its role in synaptic transmission, GABA has recently been shown to regulate neural stem cell activation in the mouse hippocampus (Catavero et al., 2018). We previously showed that inhibition of GABAAreceptors can activate MG-derived stem cell proliferation during the initial stages of retina regeneration in zebra fish (Rao et al., 2017). We hypothesized that normal GABA levels maintain MG quiescence, but that disruption of the GABA levels, whether by damage,disease or pharmacological inhibition, activates MG leading to dedifferentiation, asymmetric division, production of proliferating progenitor cells, and overall induction of a regenerative response.

We previously showed that injection of the GABAAantagonist gabazine led to MG-derived proliferation in undamaged fish retinas (Rao et al., 2017). However, a second class of GABA receptors, GABAA-ρ (Blarre et al., 2014; Alexander et al.,2017)(originally referred to as GABACreceptors (Drew et al.,1984)), is also expressed in the retina (Cutting et al., 1991,1992; Boue-Grabot et al., 1998). Whether GABAA-ρ receptors might also be involved in regulating retina regeneration is unknown. The GABAA-ρ receptor was first discovered as a subtype of the GABAAreceptor family that is insensitive to the GABAAreceptor antagonist bicuculline (Drew et al., 1984).GABAA-ρ receptors have a similar structure to GABAAreceptors in that they are both pentameric ionotropic ligand-gated ion channels (Connolly et al., 1996; Enz and Cutting, 1998;Ogurusu et al., 1999). However, the two types of receptors differ in the subunits that form the functional receptor. GABAAreceptors consist of two α subunits, two β subunits and a fifth subunit, most commonly a γ subunit. In contrast, the GABAA-ρ receptors consist of five ρ subunits, of which there are three main types, typically homomeric (Connolly et al., 1996). This difference in subunit composition results in GABAA-ρ receptors being insensitive to GABAAallosteric modulators such as benzodiazepines. While the ρ subunits were first discovered in mammals, they are also prevalent in zebra fish (Connaughton et al., 2008). The most abundant ρ subunit in the zebrafish retina is ρ2a, with ρ1 and ρ3a showing significantly lower expression and even lower levels of ρ2b (Cocco et al., 2017).In contrast to the GABAAreceptor, GABAA-ρ receptors display a more sustained response to activation (Bormann and Feigenspan, 1995) which could therefore better sustain retina regeneration. Here, we sought to test whether GABAA-ρ receptors are involved in regulating initiation of retina regeneration in zebra fish.

Materials and Methods

Zebra fish lines and maintenance

Animal experiments were approved by the Vanderbilt’s Institutional Animal Care and Use Committee (Protocol M1800200) on January 29, 2019. Zebra fish used in this study include Tg(gfap:GFP)mi2001(Bernardos and Raymond, 2006)which marks mature MG and Tg(tuba1a:GFP) (Fausett and Goldman, 2006) which marks dedifferentiated MG. All fish were maintained at 28.5°C in a 14:10 hour light:dark cycle.All fish used were between 5 and 8 months old, and were a random mix of males and females. A total of 296 zebrafish were used in this study: 73 Tg(gfap:GFP)mi2001zebrafish and 223 Tg(tuba1a:GFP) zebra fish.

Drug and morpholino injections

Pharmacological inhibitors and morpholinos were injected into the vitreous as described (Rao et al., 2017). Drugs included gabazine (S106, Sigma-Aldrich, St. Louis, MO,USA) and (1,2,5,6-tetrahydropyridin-4-yl)methylphosphinic acid (TPMPA) (1040, Tocris, Minneapolis, MN, USA). Briefly,zebrafish were anesthetized in 0.016% tricaine and a small incision was made in the sclera using a sapphire blade. A blunt-end 30-gauge needle was inserted into the vitreous and 0.5 μL of drug (15, 20, 25, and 35 nmol) or morpholino(0.75 nmol) were injected into the vitreous, the amounts of which are indicated in the respective figure legends. Fish were then placed in recovery tanks for the times listed in each experiment. For 5-ethynyl-2′-deoxyuridine (EdU) injections,20 μL of a 10 mM solution of EdU was administered via intraperitoneal injection as described by Kinkel et al. (2010).

Morpholino electroporation

Morpholinos (0.75 nmol) with a 3′-lissamine tag (Gene Tools,Philomath, OR, USA) were injected into the vitreous with or without drugs. 3 hours after injection, injected eyes were electroporated (75 V/pulse, two pulses, 1-second intervals between pulses). Fish were allowed to recover for the times indicated. Morpholinos used in this study wereGabrr2aMO1 (5′-AGT AGT GGC GCA GAT ATA ATG TCA T-3′),Gabrr2aMO2 (5′-TCG GCC TCA TAG TGA AGT CAT GAT C-3′),ascl1aMO1 (5′-ATC TTG GCG GTG ATG TCC ATT TCG C-3′) (Cau and Wilson, 2003),ascl1aMO2 (5′-AAG GAG TGA GTC AAA GCA CTA AAG T-3′) (Cau and Wilson, 2003), and a standard control morpholino (5′-CCT CTT ACC TCA GTT ACA ATT TAT A-3′).

Immunohistochemistry and terminal deoxynucleotidyl transferase dUTP nick end labeling

Zebra fish were euthanized in 0.08% tricaine and treated eyes were removed and fixed in 4% paraformaldehyde overnight at 4°C. Eyes were then washed with 5% sucrose in PBS and then cryoprotected with 30% sucrose overnight at 4°C. Eyes were then transferred to a solution of two parts Tissue-Tek O.C.T. (25608-930, VWR, Radnor, PA, USA) and one part 30% sucrose for 3 hours before moving to 100% OCT for 30 minutes. Eyes were then embedded in OCT for cryosectioning.Slides were rehydrated in PBS, and then incubated in 10 mM sodium citrate at 95°C. Sections were then blocked in 3%donkey serum, 0.1% Triton X-100 in PBS. Antibodies used were mouse anti-proliferating cell nuclear antigen (PCNA)(1:500; ab29, Abcam, Cambridge, MA, USA) and rabbit antigreen fluorescent protein (GFP) (1:500; TP401, Torrey Pines Biolabs, Secaucus, NJ, USA) diluted in antibody solution (1%donkey serum, 0.05% Tween-20 in PBS) overnight at 4°C.Slides were washed with PBS, then secondary antibodies were applied: anti-mouse cy3 (1:100) and anti-rabbit Alexa fluor 488 (AF488) (Jackson Immuno Research, West Grove,PA, USA) with TO-PRO-3 (1:1000, Thermo Fisher Scientific,Waltham, MA, USA) in antibody solution described above for 2 hours at room temperature. Slides were washed, dried,and coverslipped with Vectashield (Vector Laboratories, Inc,Burlingame, CA, USA). PCNA-positive cells were counted in the inner nuclear layer across the entirety of retinal sections.Two non-consecutive sections were counted and averaged for each eye. Images for immunofluorescence staining were taken using a META Zeiss LSM 510 Meta confocal microscope.Optical slice thickness is 0.44 μm. Images shown are stacks unless otherwise noted. The number of slices per stack are indicated in the respective figure legends. Fiji ImageJ software 4.13 was used to process images (Schindelin et al., 2012).

Terminal deoxynucleotidyl transferase dUTP nick end labeling(TUNEL) labeling was done using thein situcell death detection kit, TMR Red (12156792910, Roche Life Science,Indianapolis, IN, USA) to detect apoptotic cells. EdU detection was done using the Click-iT EdU Alexa Fluor 555 Imaging Kit (C10338, Thermo Fisher Scientific) according to the manufacturer’s instructions prior to immunohistochemistry.TUNEL-positive cells were counted across the entirety of retinal sections, in all layers of the retina. Each data point is from a single eye and is an average of counts from twononconsecutive sections, with each fish providing only one eye.

Fluorescent activated cell sorting

Fluorescent activated cell sorting (FACS) was used to isolate GFP+ cells from Tg(gfap:GFP)mi2001zebrafish retinas, either undamaged or 24 hours after TPMPA injection, using BD FACSAria III (BD Biosciences, San Jose, CA, USA) in the VUMC Flow Cytometry Shared Resource. Retinas were dissociated according to Rajaram et al. (2014) with the following changes.Retinas were dissected and collected in Leibovitz L-15 media(21083-027, Thermo Fisher Scientific). Retinas were then treated with 1 mg/mL hyaluronidase (H3884, Sigma-Aldrich)rocking at room temperature for 15 minutes. Dead cells were detected via propidium iodide. A total of 24 retinas were collected and pooled from 12 adult fish per sorting.

Quantitative reverse transcription polymerase chain reaction

RNA was collected from sorted cells using TRIzol-LS (10296028,Thermo Fisher Scienti fic). Taqman small RNA assays (Thermo Fisher Scientific) were used to perform quantitative reverse transcription polymerase chain reaction (RT-qPCR) for let-7a.Briefly, 5 μL RNA was used per RT, which was then diluted 1:2. The diluted cDNA (1.33 μL) was used in a 10 μL qPCR reaction in technical triplicates. qPCR reactions were done in 384-well plates using the Bio-Rad CFX384 Real-time System(Bio-Rad Laboratories). The reactions were normalized to a U6 snRNA control. qPCR of mRNAs was done by first treating RNA with DNase (TURBO DNA free kit, AM1907, Thermo Fisher Scientific) and converted to cDNA using AccuScript High-Fidelity 1stStrand cDNA Synthesis Kit (Cat# 200820;Agilent Technologies, Stratagene, La Jolla, CA, USA). qPCR was performed using SYBR Green (Bio-Rad Laboratories).All qPCR primers spanned exon-exon junctions (Integrated DNA Technologies, Inc, Coralville, IO, USA). The reactions were normalized to 18S rRNA levels. qPCR reactions were done in 384-well plates using the Bio-Rad CFX384 Real-time System. Analysis was done using the ΔΔCt method (Livak and Schmittgen, 2001). The primers used for qRT-PCR are shown inTable 1.

In situ hydbridization

In situhybridization of Gabrr2a and Gabrg2 was performed using RNAScope (Advanced Cell Diagnostics, Newark, CA, USA).Hybridizations were performed according to manufacturer’s instructions for fixed frozen tissues on cryosections from Tg(gfap:GFP)mi2001retinas, with the following changes. After creating a hydrophobic barrier around the tissue sections,RNAScope Protease III was applied for only 20 minutes rather than 30 minutes. After applying the HRP blocker solution,immunohistochemistry was performed according to ACDBio’s instructions, using rabbit anti-GFP, and anti-rabbit AF488 and TOPRO-3 from above.

Statistical analysis

Two tailed Student’st-tests were used when comparing two sample means, a one-way analysis of variance was used to compare multiple means, and Fisher’s LSD test was used for qRT-PCR analysis with ΔCt values. Each data point represents an average of two separate counts per eye. For each count,only the inner nuclear layer was counted. Eyes that were damaged were not used for analysis. Damage was indicated by disrupted morphology or high amounts of proliferation. Forin situhybridization analysis, a Costes’ image randomization and evaluation of Pearson’s coefficient was performed using the JACoP plugin for ImageJ across all optical slices per z-stack,each comprising between 68 and 70 optical slices (Costes et al., 2004; Bolte and Cordelieres, 2006).

Table 1 ｜The primers used for quantitative reverse transcription polymerase chain reaction

Results

Inhibition of GABAA-ρ receptors induces proliferation in the undamaged retina

Our lab has previously shown that inhibition of GABAAreceptor signaling induces spontaneous proliferation in an undamaged zebra fish retina (Rao et al., 2017). However, the ρ2a subunit of the GABAA-ρ receptor is also expressed in the whole retina(Cocco et al., 2017) and in puri fied MG by RNAseq (Additional Table 1). Thus, we tested whether pharmacological inhibition of GABAA-ρ receptors would also induce proliferation. For this, we injected the GABAA-ρ receptor antagonist TPMPA(Murata et al., 1996; Ragozzino et al., 1996) into undamaged eyes from Tg(1016tuba1a:gfp) transgenic fish and assessed proliferation using immunostaining against PCNA or by direct incorporation of EdU. The Tg(1016tuba1a:gfp) transgenic line specifically marks dedifferentiated MG and MG-derived neural progenitors in actively regenerating retinas (Fausett and Goldman, 2006). TPMPA is a commercially available competitive antagonist of the GABAA-ρ receptor with only minimal effects on GABAAor GABAB receptors (Ragozzino et al., 1996). Intravitreal injection of TPMPA induced a signi ficant increase in PCNA+cells compared to control PBS injections(Figure 1). Similar increases were detected using incorporation of EdU (Additional Figure 1). Induction of DNA replication was dose dependent up to 25 nmol, but we noticed a decrease in the number of PCNA positive cells at higher concentrations(Additional Figure 2). This is consistent with a speci fic effect of TPMPA, especially because it has been reported that TPMPA can act as an agonist at higher concentrations (Ragozzino et al., 1996). Importantly, alltuba1a-GFP+cells co-localized with PCNA+cells indicating that the proliferative cells were derived from MG (Figure 1). To ensure that the proliferation we observed was not an indirect consequence of cell death due to application of TPMPA, we utilized TUNEL staining which showed no difference in the number of apoptotic cells between the PBS-injected eyes and the TPMPA-injected eyes (Additional Figure 3). Combined with earlier work, the data indicate that impaired GABA signaling can activate MG and induce a regenerative response in undamaged zebra fish retinas.

Figure 1 ｜ Inhibition of the GABAA-ρ receptor in the absence of retina damage results in a Müller glia-derived proliferative response.

While TPMPA is ＞ 100-fold more potent against GABAA-ρ receptors compared to GABAAreceptors (Murata et al.,1996), it was possible that under the conditions of intravitreal injection, the effect of TPMPA could have been due to unexpected inhibition of GABAAreceptors. To complement the TPMPA experiments, we tested whether knocking down the ρ2a subunit (GABRR2a), the most abundant ρ subunit in the undamaged retina (Additional Table 1), would also be sufficient to induce a proliferative response. For this,we independently electroporated two different antisense morpholinos targeting the ρ2a subunit into undamaged retinas (Figure 2). Compared to control morpholino injections,injection of the two different morpholinos targeting the ρ2a subunit induced significantly higher proliferation (Figure 2CandD). Again, alltuba1a-GFP+cells co-localized with PCNA indicating that antisense inhibition of the GABAA-ρ receptor induces a MG-based regenerative response.

Inhibition of GABAA-ρ signaling induces a regenerative response

To further test whether inhibition of GABAA-ρ receptors induces a bona fide regenerative response, we combined TPMPA injections with antisense inhibition ofascl1a(Figure 3).Ascl1a is a transcription factor that is required for MG-derived retina regeneration (Fausett et al., 2008; Ramachandran et al., 2010; Brzezinski et al., 2011; Ramachandran et al., 2011;Pollak et al., 2013; Ueki et al., 2015; Wohl and Reh, 2016). The prediction is that if TPMPA is inducing a bona fide regenerative response, loss of Ascl1a should reduce the proliferation observed after injection of TPMPA. As shown inFigure 3, after knockdown ofascl1a, we observed a significant decrease in the amount of proliferation compared to co-injection of TPMPA and a control morpholino (Figure 3EandF).

Lastly, to confirm that TPMPA-induced proliferation is indeed activating the canonical retina regeneration pathway,we used qRT-PCR on purified MG to determine if specific regeneration-associated genes are differentially expressed after treatment with TPMPA. Previous work has shown that retina regeneration in zebrafish results in an increase in the expression ofascl1a,insm1a, andsox2, and decreased expression ofdkk1bandlet-7a. Like Ascl1a, Insm1a and Sox2 are transcription factors that are required for retina regeneration (Ramachandran et al., 2012; Gorsuch et al.,2017). Downregulation of Dkk1b, a negative regulator of Wnt signaling, is also required for regeneration (Ramachandran et al., 2011) as is downregulation ofLet-7a, a miRNA that represses several mRNAs encoding factors required for retina regeneration, including Ascl1a (Ramachandran et al.,2010). For these experiments, we used the Tg(GFAP:GFP)mi2001zebra fish line in which MG expression of GFP is controlled by the GFAP (glial fibrillary acid protein) promoter (Bernardos and Raymond, 2006; Nagashima et al., 2013). After damage,expression of the GFP reporter and/or GFAP is increased in the zebrafish retina (Vihtelic et al., 2006; Bernardos et al.,2007; Lenkowski et al., 2013; Lenkowski and Raymond, 2014;Sifuentes et al., 2016). Thus, retinas from either undamaged fish or fish injected with TPMPA were dissociated and sorted to enrich for GFP+cells (Figure 4). RNA was isolated and qRTPCR was performed to determine fold changes in expression between undamaged and TPMPA injected GFP+cells. We observed signi ficantly increased levels of expression ofascl1aandinsm1aand slightly upregulated expression ofsox2(Figure 4). We also detected signi ficantly decreased expression levels ofdkk1bandlet-7a. These results support the hypothesis that inhibition of GABAA-ρ receptors by TPMPA induces expression of factors consistent with a canonical regenerative response in zebra fish.

Figure 2 ｜ Inhibition of GABAA-ρ receptors via antisense morpholino injections in the absence of damage results in a proliferative response.

Figure 3 ｜ Inhibition of ascl1a blocks TPMPA induced-proliferation.

Figure 4 ｜ Inhibition of the GABAA-ρ receptor induces gene expression changes consistent with canonical retina regeneration.

GABAA-ρ receptors are localized to the inner and outer nuclear layers

If detection of reduced GABA levels is mediated by MG,GABA receptors should be expressed in MG. By RNAseq of undamaged retinas using the Tg(GFAP:GFP)mi2001transgenic line, we found that GABAAand GABAA-ρ receptor subunits are indeed expressed in MG (Additional Table 1).Besides transcriptomic analysis, we also tested whether GABAA-ρ receptors co-localize with MG. We previously used immunostaining to show close association between GABAAreceptors on MG processes flanking horizontal cell processes (Rao et al. 2017). Because antibodies against the zebra fish GABAA-ρ receptors are not available, we used in situ hybridization to localize RNAs encoding both the ρ2a and γ2 subunits of the GABAA-ρ and GABAAreceptors, respectively.RNA transcripts encoding ρ2a subunits (gabrr2a) were detected in the outer plexiform layer, the inner nuclear layer,and cell bodies of the outer nuclear layer (Figure 5). GABAA-ρ receptors are known to be expressed in both horizontal and bipolar cells, (Qian and Dowling, 1993; Fletcher et al.,1998; Lopez-Chavez et al., 2005), but we also detected ρ2a transcripts associated with MG processes (Figure 5). To better determine whether ρ2a subunits co-localize with MG processes, we used the ImageJ plug-in JACoP (Bolte and Cordelieres, 2006) to evaluate the extent of co-localization and also applied Costes’ image randomization and evaluation of Pearson’s coefficient on three sets of optical slices (single slice shown inFigure 5B) from one of which the Z stack inFigure 5was generated. This analysis resulted in an average Pearson’s coefficient of 0.133667. For these analyses,Pearson’s coefficients can range from 1 (perfect correlation) to-1 (no correlation). Thus, there is a positive correlation for colocalization between RNA transcripts encoding ρ2a subunits with MG processes. The correlation is weak, but the resultingP-value is 1.0, meaning high confidence (＞ 95%) that the colocalization is not due to random chance.

Figure 5 ｜ Localization of GABAA-ρ transcripts via in situ hybridization.

Additionally, we used in situ hybridization to localize the γ2 subunit (gabrg2), which showed that RNA transcripts encoding the γ2 subunit are broadly expressed across the retina(Additional Figure 4). Closer examination of the merged image revealed that puncta corresponding to γ2 subunits are detectable in retinal layers containing both MG cell bodies and processes, consistent with previous immunostaining.Combined, immunostaining, RNA localization, and RNAseq support the hypothesis that MG processes are in position to sense GABA levels in the retina.

Synergistic activation of regeneration by simultaneous inhibition of GABAA-ρ and GABAA receptors

Individually, inhibition of GABAA-ρ (Figures 1and2) or GABAAreceptors (Rao et al., 2017) is sufficient to induce proliferation as part of a canonical retina regenerative response. However,since both receptors are associated with MG processes,overall detection of GABA levels by MG could be mediated by both receptors. If true, combined inhibition of both receptors should synergize to activate MG during regeneration. To test this, we co-injected gabazine, a GABAAreceptor antagonist,and TPMPA. We observed a signi ficant increase in the number of proliferating PCNA+cells compared to either treatment alone or the PBS control treatment (Figure 6).

Discussion

We previously showed that inhibition of GABAAreceptors can induce retina regeneration in adult undamaged retinas (Rao et al., 2017). Here, we extend that work to show that inhibition of GABAA-ρ receptors can induce a similar regenerative response. Several lines of evidence support that the induction of proliferation that we observe is mediated by MG in a canonical regenerative pathway. First, the transcription factor Ascl1 must be activated during retina regeneration in both fish and mice (Fausett et al., 2008; Brzezinski et al., 2011;Jorstad et al., 2017) and knockdown of ascl1 blocks the effects of inhibition of GABAA-ρ receptors. Second, we observed the expected activation of ascl1 and insm1a (Ramachandran et al.,2012) following inhibition of GABAA-ρ receptors, and we also observed reduced levels of dkk1b and let-7a (Ramachandran et al., 2010, 2011). We did not observe signi ficant activation of sox2, but this seems to be more related to timing. Sox2 is normally activated during regeneration, reaching its peak by 31 hours of light damage (Gorsuch et al., 2017) whereas we examined sox2 levels only 24 hours after inhibition of GABAA-ρ receptors. Given the significant differential expression of the other factors, it is likely that sox2 would show higher expression if tested at a later timepoint. Altogether,immunostaining (Rao et al., 2017),in situhybridization,and RNAseq support the idea that MG are positioned to respond to reduced levels of GABA and induce activation of MG (Additional Figure 5). Normally, this cascade would be induced after retinal injury, but our experiments show that simply mimicking the loss of GABA in an undamaged retina can induce regeneration.

Figure 6 ｜ Dual inhibition of GABAA and GABAA-ρ receptors enhances proliferation.

Direct vs. indirect effects of GABA inhibition

Because multiple cell types express GABA receptors in the retina, it remains possible that the effects we observe are not directly due to sensing of decreased GABA levels by MG. This could explain the increase in PCNA+cells in the ganglion cell layer, as GABA receptors have been found in retinal ganglion cells (Popova, 2015). Inhibition of GABAAand GABAA-ρ receptors on bipolar cells (Connaughton et al.,2008) could result in excess glutamate leading to excitotoxic damage (Olney, 1982) and MG-derived proliferation. Excess glutamate could also activate AMPA receptors on MG, leading to an influx of Ca2+(Zhang et al., 2019) and subsequent proliferation (Pinto et al., 2015). The Ca2+activated protein CAPN5 is upregulated in MG after damage (Coomer and Morris, 2018); excess glutamate activating AMPA receptors on MG could be responsible for upregulation of CAPN5 and other Ca2+activated proteins during regeneration. Lastly, it is also possible that damage induced by injection could be causing cell death leading to induction of a regenerative response, as opposed to regeneration as a consequence of inhibition of GABA signaling. We did not observe increased levels of apoptosis as measured by TUNEL assays, but it remains formally possible that damage leading to necrosis or autophagy could also result in initiation of regeneration.Definitive testing of the model that GABA levels are directly sensed by MG awaits generation of transgenic zebra fish lines with inducible, MG-speci fic knockouts of these receptors.

There also remains the possibility that the effects of TPMPA are not entirely due to inhibition of GABAA-ρ receptors. While TPMPA is a highly selective antagonist of GABAA-ρ receptors,it is also a weak antagonist of GABAAreceptors and an even weaker agonist of GABAB receptors (Ragozzino et al., 1996).It is therefore formally possible that the effects of TPMPA we observe could be mostly due to inhibition of GABAAreceptors.However, the morpholino knockdown of GABAA-ρ subunit ρ2a argues against this possibility and provides further support that the effect of TPMPA is through inhibition of GABAA-ρ receptors. In addition, the dual inhibition of GABAAand GABAA-ρ receptors provide further support that the effects are synergistic and not solely due to effects on GABAAreceptors.This synergistic effect is likely due to more GABA receptors being inhibited. As shown, both here and in our previous work(Rao et al., 2017), higher doses of the inhibitors increase the resulting number of PCNA+cells.

GABA receptors and adult neurogenesis

Neural stem cell activity in the mouse hippocampus has been proposed to be regulated by sensing of non-synaptic GABA levels (Chell and Frisen, 2012; Song et al., 2012; Catavero et al., 2018). Our proposed activation of MG (Additional Figure 5)is very similar to that proposed by Song et al. (2012) although we do not have evidence of long range GABAergic inputs (Bao et al., 2017) which would not seem to be necessary in the retina. Interestingly, GABABreceptors have been proposed to play a role in adult neurogenesis in the mouse hippocampus(Giachino et al., 2014). We have no evidence thus far for an involvement in G-protein coupled GABABreceptors in regulating activation of MG in the zebra fish retina. Supporting our work in the retina, inactivation of GABAAreceptors was shown to inhibit proliferation of cultured progenitor cells from adult mouse retina (Wang et al., 2019). Thus, despite some differences, there appears to be a an evolutionarily conserved mechanism involving stem cell activity and the sensing of GABA levels. A major question, then, is how loss of GABA signaling mechanistically induces regeneration in the zebra fish retina. GABAAand GABAA-ρ are both ionotropic receptors that selectively transport Cl-ions either into or out of the cell depending on membrane potential. Recently, reduced levels of intracellular Cl-were found to induce tumor necrosis factor α in endothelial cells (Yang et al., 2012). Tumor necrosis factor α has been shown to be involved in the early stages of retina regeneration (Nelson et al., 2013). It is therefore possible that loss of GABA signaling after retina damage results in reduced intracellular Cl-, which then leads to an upregulation of tumor necrosis factor α.

As above, excess glutamate in the retina can be excitotoxic(Olney, 1982) which could be due to TPMPA acting on bipolar cells. While our previous work showed that inhibiting glutamate receptors in the retina leads to MG-derived proliferation (Rao et al., 2017), inhibiting AMPA receptors in an injury model involving injection of CoCl2led to reduced proliferation, an apparent neuroprotective effect (Medrano et al., 2018). While other explanations are possible, these seemingly contradictory results could simply be due to a CoCl2-mediated injury response, whereas we blocked AMPA receptors in undamaged retinas.

GABA and β-cell regeneration

Beyond the mouse hippocampus and the zebra fish retina, GABA levels can drive pancreatic α cells to a β cell fate (Ben-Othman et al., 2017; Li et al., 2017). Intriguingly, the role of GABA in this case, both in adult mice and cultured cells, is the opposite of what we observe in the retina in that increased GABA levels or administration of indirect agonists of the GABAAstimulated increased numbers of α cells derived from glucagon secreting β cells. Even though the mechanism of action of GABA is the opposite, the common finding is that altered GABA signaling can activate regeneration.

Conclusion

We have shown that inhibiting GABAA-ρ receptors is sufficient to induce a regenerative response in the zebrafish retina in the absence of damage and that inhibiting both GABAAGABAA-ρ receptors simultaneously produces a synergistic effect. It will be important to determine if this effect is directly mediated through MG, but, together, our results suggest a novel approach to induce a regenerative response in the mammalian retina.

Acknowledgments:We would like to thank members of the Patton lab for constructive discussions. Zebra fish were maintained by the Stevenson Zebra fish facility with help from Qiang Guan. Zebra fish lines were shared by Pamela Raymond (Tg(gfap:GFP)mi2001) and Daniel Goldman(Tg(tuba1a:GFP).

Author contributions:MRK designed the study, collected and analyzed data, and drafted and edited the manuscript. NK collected and analyzed data and helped with drafting the manuscript. JGP helped design the study, helped with analyzing data, drafting and editing the manuscript.All authors approved the final version of the paper.

Con flicts of interest:The authors have no con flict of interest.

Financial support:This work was supported by grants from the NIH R01EY024354-S1 and UO1 EY027265 to JGP, and T32 EY021453 to Milam Brantley, as well as additional support from the Stevenson family and Gisela Mosig endowments to Vanderbilt University.

Institutional review board statement:Animal experiments were approved by the Vanderbilt’s Institutional Animal Care and Use Committee (Protocol M1800200) on January 29, 2019.

Copyright license agreement:The Copyright License Agreement has been signed by all authors before publication.

Data sharing statement:Datasets analyzed during the current study are available from the corresponding author on reasonable request.

Plagiarism check:Checked twice by iThenticate.

Peer review:Externally peer reviewed.

Open access statement:This is an open access journal, and articles are distributed under the terms of the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 License, which allows others to remix,tweak, and build upon the work non-commercially, as long as appropriate credit is given and the new creations are licensed under the identical terms.

Open peer reviewers:Salvatore L. Stella, Penn State College of Medicine,USA; JiaJie Teoh, Columbia University Irving Medical Center, USA.

Additional files:

Additional file 1:Open peer review reports 1 and 2.

Additional Figure 1:Inhibition of the GABAA-ρ receptor in the absence of damage results in a proliferative response.

Additional Figure 2:Dose dependent inhibition of the GABAA-ρ receptor.

Additional Figure 3:Inhibition of the GABAA-ρ receptor does not result in increased apoptosis.

Additional Figure 4:Localization of transcripts encoding the γ2 subunit of GABAA receptors.

Additional Figure 5:Model of GABA receptor inhibition-induced proliferation.

Additional Table 1:Average RPKM values of GABA subunits from sequencing of sorted Müller glia.