Characterization of retinal regeneration in adult zebrafish following multiple rounds of phototoxic lesion

Fish maintenance

Two fish lines were used for these studies: adult (9–18 months) albino (alb) and adult (9–18 months) Tg(gfap:egfp)/alb zebrafish. Unless otherwise noted, fish were maintained under a daily light cycle of 14 h light (250 lux):10 h dark at 28.5 °C (Westerfield, 1995) and fed a combination of flake food and brine shrimp. The Institutional Animal Care and Use Committee at Wayne State University approved the phototoxic lesion protocol used in this study (Protocol # 16-01-037).

Intense light exposure

A photolytic damage model was used to destroy rod and cone photoreceptors (Thomas et al., 2012). First, five adult albino fish per treatment group were dark-adapted for 10 days. For the untreated fish (0 h control group), eye tissue was harvested after the 10 day dark-adaptation. For fish that were exposed to photolytic damage to photoreceptors, following the 10 day dark-adaption, animals were first exposed to approximately 100,000 lux for 30 min using a broadband light source (Thomas et al., 2012). Next, fish were transferred to 1.8-liter tanks and exposed to approximately 10,000 lux of light from four, 250 W halogen lamps (Thomas et al., 2012). Fish were subjected to constant light exposure for up to four days, and then transferred back to standard light:dark conditions. For fish that underwent 1 round of light treatment (1× treatment group), eye tissue was collected at the time points indicated. These animals were 9–10 months of age. For fish that underwent multiple rounds of light treatment (6× treatment group), following the first light treatment, fish were allowed to recover for 28 days, dark-adapted for 10 days, and then exposed to another round of light lesion as described above. This was repeated for a total of six rounds of light treatment and 28 day recovery period. As these fish were now 18 months of age, the experiment was terminated after six rounds in order to not exceed our normal window of 9–18 months of age for light treatments.

BrdU incorporation

For the BrdU study, adult (9–18 months) albino fish were light treated for 24 h. Next, fish were transferred to a 1L solution containing 0.66 g of NaCl, 0.1 g Neutral Regulator, and 1.5 g BrdU (SeaChem Laboratories, Inc. Stone Mountain, GA, USA; Sigma). After 24 h, these fish were transferred to normal light:dark conditions for 28 days, dark-adapted for 10 days, and then exposed to another round of light lesion as described above. The timeframe for BrdU incubation was based on previous studies which indicated that all participating Müller glia re-enter the cell cycle within this window, starting at ∼31 hpL, and that the first daughter cell is visible at 48 hpL (Kassen et al., 2008; Thummel et al., 2008a; Vihtelic & Hyde, 2000). Thirty-six hours after starting the second treatment, fish were euthanized and whole eyes were enucleated and processed for immunohistochemistry. The percentage of Müller glia in each group (BrdU +, PCNA +, double-positive) were calculated within a linear distance of 300 microns. The Kruskal-Wallis nonparametric H test was used to assess whether there were significant differences between the groups at the 0.05 significance level.

Immunohistochemistry and confocal microscopy

Immunohistochemistry was performed as previously described (Thummel et al., 2008a). Briefly, five whole eyes from each experimental group were harvested from sacrificed adult (9–18 months) animals and were fixed in 9:1 ethanolic formaldehyde (100% ethanol: 37% formaldehyde) overnight at 4 °C. Eyes were cryopreserved, embedded in freezing medium, and retinal sections (16 µM) containing or immediately surrounding the optic nerve were collected on Superfrost Plus glass slides (Fisher Scientific, Pittsburgh, PA, USA). Two different immunohistochemistry protocols were used for the primary antibody incubation. Standard immunohistochemistry was performed as previously described (Thummel et al., 2008a) using the following primary antibodies: mouse Zpr-1 antibody (Zebrafish International Resource Center, Eugene, OR, USA; 1:200), rabbit anti-Rhodopsin antisera (gift from David Hyde; 1:5,000), rabbit anti-GFP (Abcam; 1:1,500), mouse anti-PCNA (Sigma; 1:1,000), rabbit anti-Blue opsin (gift from David Hyde; 1:500), rabbit anti-UV opsin (gift from David Hyde; 1:1,000), mouse anti-Glutamine Synthetase (Chemicon; 1:500), and mouse 4C4 (gift from Peter Hitchcock; 1:250). For antibodies requiring antigen retrieval, the following procedure was used. Slides were preheated to 55 °C on a slide warmer while a glass coplin jar was preheated in an Oster 6-quart Digital Food Steamer for 30 min . Sodium Citrate Buffer (10 mM Sodium Citrate/1XPBS/0.05% Tween-20, pH 6.0), heated in a microwave until boiling, was poured into the coplin jar and the slides were immediately placed in the buffer solution. After 30 min, the coplin jar was removed from the steamer and allowed to cool at room temperature for 30 min. The slides were removed and the tissue was outlined with a Pap Pen (Daido Sangyo, Tokyo, Japan). Slides were rinsed three times for 5 min with 1XPBS/.5% Triton X-100. Retinal sections were covered with a blocking solution of 1XPBS/.5% Triton X-100/20% sheep serum for 1 h. Sections were incubated overnight at 4 °C in primary antibody diluted into 1XPBS/.5% Triton X-100/2% sheep serum. The primary antibodies and antisera requiring use of this protocol were: mouse anti-HuC/D (Invitrogen; 1:50) and rabbit anti-GFAP (DakoCytomation; 1:500).

For both primary antibody procedures, after overnight incubation, sections were rinsed three times for 10 min with PBS/.5% Triton X-100 at room temperature. Sections were then incubated in the secondary antibody solution diluted in 1XPBS/.5% Triton X-100/2% sheep serum for 1 h. After secondary antibody incubation, sections were rinsed in 1XPBS/.5% Triton X-100 3 times for 10 min. Slides were covered with a coverslip using ProLong Gold (Molecular Probes, Eugene, OR). Secondary antibodies included AlexaFluor-conjugated 488, 594 and 647 goat anti-primary (1:500, Life Technologies, Grand Island, NY, USA). Nuclei were stained using TO-PRO-3 (TP3; 1:750; Life Technologies, Grand Island, NY, USA) or DAPI.

Confocal microscopy was performed with a Leica TCS SP8 confocal microscope. Z-stacked images were taken within a 4 µM thickness with 0.5 µM between slices. Quantification of retinal cell number was performed by manual count of a 300 micron linear distance on the central dorsal retina using five captured images in each group of biological replicates. The differences between groups were analyzed by a one-way ANOVA if three groups were compared as seen in all photoreceptor quantifications and the quantification of mis-localized HuC/D+ neurons. A Student’s T-test was performed using p < 0.05 as a statistical cut-off if only two groups were compared as in the comparison of PCNA+ Müller glia at 36 h post light (hpL) and qRT-PCR results.

Quantitative real-time PCR

Retinal tissue was isolated from control and experimental groups at 36 hpL, 72 hpL, and 28 days post light (dpL). Pools of five retinas were collected in biological triplicate. Tissue was collected into 1 mL of Trizol, manually homogenized, and frozen in Trizol at −80 °C. RNA was isolated using the Trizol Reagent, per manufacturer’s instructions (Invitrogen). DNase treatment was not performed. RNA was quantified and assessed for purity using a nanodrop. 1 µg RNA was used as the input for cDNA synthesis, which was performed using the manufacturer’s protocol with random oligos and Superscript II polymerase (Invitrogen, 18064). cDNA was diluted 1:25 in nuclease free H₂O and 2 µL was used in the following reaction. Quantitative Real-Time PCR (qPCR) was carried out in technical and biological triplicate in a 10 µL reaction using the SYBR green reagent (#4309155; Applied Biosystems, Foster City, CA, USA) on a CFX Connect Real-Time System (Bio-Rad). Cycling conditions were as follows: step 1. 95 °C hold for 10 min, step 2. 95 °C denature for 15 s, 60–63 ° C anneal/extend for 1 min (cycle step 2. 39 times), step 3. 65–95 °C melt-curve increasing by 0.5 °C every 5 s. Primers were designed to span introns with the exception of the gpia primer pair (BIO-RAD Assay ID:qDreCED0007433) which was used as the endogenous reference gene. Primer sequences are displayed in Table 1. PCR products were not sequenced to confirm identity; however, the melt curve was analyzed to determine single peaks for each primer set. Quantification of fold-change in expression was performed between control (1×) experimental (6×) groups using the Livak 2^−ΔΔC(t) method with a p < 0.05 for statistical cut-off (Pfaffl, 2001). Briefly, the mean CT for each gene was normalized against the gpia mean CT endogenous reference to obtain a ΔCT value. The ΔCT value of the 6× group was averaged across the biological triplicate and then normalized against the 1× group, giving the ΔΔCT value. Statistical differences between groups were determined using a student’s T-test from the normalized ΔΔCT values. Finally, ΔΔCT values were converted to log 2-fold changes in gene expression and graphed.

Evidence that the exact same population of Müller glia do not re-enter the cell cycle following each light damage event

The timeline of light-induced damage and regeneration of the zebrafish retina has previously been described (Kassen et al., 2007; Lenkowski & Raymond, 2014; Thomas et al., 2012; Thummel et al., 2008a; Vihtelic & Hyde, 2000). Briefly, photoreceptor apoptosis peaks at 24 hpL, followed by Müller glia cell cycle re-entry at 31–36 hpL. The resultant progenitors continue to proliferate over the next 3 days as they migrate from the inner nuclear layer (INL) to the outer nuclear layer (ONL), where they begin to differentiate into new photoreceptors during the following week. Previous reports showed that ∼50% of the total population of Müller glia re-enter the cell cycle during the photolesion paradigm (Thummel et al., 2008b). Given that this study used multiple rounds of phototoxic treatment, we were first interested in testing whether the exact same subset of Müller glia that acted as stem cells following one round of damage would respond in the same way a second time. To examine this, adult albino zebrafish were subjected to two rounds of light damage. During the first round, Müller glia that re-entered the cell cycle were permanently marked with BrdU incorporation during the S phase of the cell cycle. During the second round of Müller glia cell re-entry at 36 hpL, retinal sections were triple immunolabeled with antibodies against BrdU (S phase during the first round), PCNA (G1 → S phase during the second round), and GFAP (all Müller glia). We observed that 24% of GFAP-positive Müller glia were only immunolabeled with anti-BrdU (Fig. 1, arrowheads), marking cells that re-entered the cell cycle following the first round of damage, but not the second. Conversely, 40% of GFAP-positive Müller glia were only immunolabeled with anti-PCNA (Fig. 1, arrows), indicating cells that initiated cell cycle re-entry only during the second round of light damage, but not the first. Finally, 35% of GFAP-positive Müller glia were co-immunolabeled with anti-BrdU, and anti-PCNA (Fig. 1, double arrowheads), marking cells that re-entered the cell cycle during both rounds of regeneration. A Kruskal-Wallis nonparametric H-test determined that there were not significant differences between these groups (p = 0.101). These data suggest that the exact same population of Müller glia do not re-enter the cell cycle following each individual light damage event.

Multiple rounds of light damage leads to a greater loss of photoreceptors and increased Müller cell gliosis, but a normal Müller glial and progenitor cell proliferation

Next, we compared photoreceptor numbers and proliferation in untreated control animals with animals exposed to one round (1×) and six rounds (6×) of light treatment at two intermediate time points, 36 and 72 hpL (Thummel et al., 2008a). At 36 hpL, a one-way ANOVA analysis was performed among all photoreceptors quantified and revealed that there were statistically significant differences among the groups (p < 0.0001). Notably, a post-hoc test also revealed statistical differences between the 1× and 6× groups. Specifically, 6× light treated retinas exhibited decreased Rhodopsin immunolocalization (Figs. 2B–2D), rod photoreceptor nuclei in the outer nuclear layer (Fig. 2Q; p < 0.05) and significantly fewer numbers of long-single (blue) cones (Figs. 2H–2J, S; p < 0.02) and short-single (UV) cones (Figs. 2K–2M, T; p < 0.006) when compared with the 1× treated retinas. Zpr-1 positive double cones were nearly completely destroyed in both groups (Figs. 2E–2G, 2R; p > 0.05). Additionally, 6× retinas showed a significantly higher number of 4C4-positive microglia/macrophages compared with 1× retinas (Figs. 2O–2P; Average number of 13.8 vs. 46.2 per 300 µm, respectively; p < 0.01), which is consistent with the visually more pronounced cellular debris observed in the 6× retinas (Figs. 2G, 2J, 2M; arrows). These data suggest that retinas that have undergone 6× rounds of damage and regeneration exhibit increased photoreceptor loss to a subsequent light insult and exhibited an increased inflammatory response.

Increased photoreceptor loss would predict a higher percentage of Müller glia to re-enter the cell cycle at 36 hpL (Thomas et al., 2012). Contrary to this prediction, we observed that the percentage of Müller cells that re-entered the cell cycle was not significantly different between the 1× and 6× groups (Figs. 3B–3C, 3J). In addition, at 72 hpL, both 1× and 6× retinas show Müller glia surrounded by clusters of PCNA-positive progenitors migrating to the outer nuclear layer (ONL; Figs. 4E–4F), suggesting that progenitor amplification and migration was not severely altered in 6× retinas. However, retinas from the 6× groups did show signs of increased Müller cell gliosis, including unorganized, hypertrophied Müller glia with a significant upregulation of glial fibrillary acidic protein (gfap) mRNA expression (Fig. 3K), and thickened GFAP-positive appendages in the photoreceptor layer at 36 hpL (Figs. 3E–3F, 3H–3I). Expression of signal transducer and activator of transcription 3 (stat3), one of the earliest molecular markers of retinal stress, was also significantly upregulated (Fig. 3K) (Gorsuch & Hyde, 2014; Nicolas et al., 2013). At 72 hpL, retinas from the 6× groups showed continued signs of Müller cell disorganization. GFAP immunohistochemistry revealed obvious disorganization of GFAP within the Müller glia with evidence of redistribution of the GFAP fibers away from the endfeet and towards the cell body, perhaps indicative of a hypertrophy phenotype (Figs. 4A–4C). These data suggest that 6× rounds of light damage resulted in increased signs of Müller glia stress and gliosis, but did not affect their ability to re-enter the cell cycle and produce retinal progenitors.

Next, we analyzed the expression of genes that have been associated with proliferation in other cellular events, such as cyclin dependent kinase 1 (cdk1), and c-jun (Malumbres, 2014; Munshi & Ramesh, 2013), and genes shown to be required for amplification of the progenitor population during retinal regeneration, such as proliferating cell nuclear antigen (pcna), paired box protein 6 (pax6a and pax6b), sonic hedgehog (shha and shhb), sinus oculis homeobox 3 (six3b), and transforming growth factor-beta-induced factor 1 (tgif1) (Fuccillo, Joyner & Fishell, 2006; Hatton et al., 2006; Kaur et al., 2018; Lenkowski et al., 2013; Sherpa et al., 2014; Thomas et al., 2018; Thummel et al., 2010; Thummel et al., 2008b; Todd & Fischer, 2015). Compared with 1× control retinas, retinas that underwent 6× rounds of light damage exhibited a significant downregulation of many pro-proliferative genes, including cdk1, pcna, shhb, six3b, and tgif1 (Fig. 4J). However, no change was observed for c-jun and pax6a, and one pro-proliferative gene, shha, was significantly upregulated (Fig. 4J). In addition, we analyzed a group of genes associated with Müller glial function of retinal homeostasis and support, such as glutamine synthetase (glulb), potassium inward-rectifying channel (kcnj10a), and retinaldehyde binding protein 1 (rlbp1b) (Bringmann & Wiedemann, 2012). Compared with 1× control retinas, 6× retinas exhibited a significant downregulation of glulB and rlbp1b (Fig. 4J), suggesting that the Müller glia have begun to lose their ability to recycle extracellular glutamate and participate in the regeneration of the cone visual pigment. Notably, expression of rlbp1a, which is expressed in the retinal pigmented epithelium but not in Müller glia (Collery et al., 2008), was not significantly different between groups (Fig. 4J). Finally, we analyzed the expression of two intermediate filaments of the Müller glia, gfap and vimentin (vim), and found no differences in gfap expression and a significant downregulation of vim expression between groups at 72 hpL (Fig. 4J). Together, these data suggested a complex set of consequences to 6× rounds of light treatment in regards to maintaining proliferation of the progenitor pool and Müller glia function.

Multiple rounds of light damage results in aberrant long-term regeneration of the inner retina, a reduction of regenerated long single cones, and full regeneration of all other photoreceptor subtypes

Finally, we analyzed 1× and 6× groups at 28 dpL, an established time-point for phenotypically defined full regeneration as previously described (Thomas et al., 2016; Thummel et al., 2008a; Thummel et al., 2008b). Immunolocalization of GFAP showed an apparent resolution in Müller cell gliosis and organization at 28 dpL in both 1× and 6× retinas (Figs. 5A–5C). Interestingly, immunolocalization of HuC/D, a marker of amacrine and ganglion cells, demonstrated aberrant localization of HuC/D-positive cells in both the 1× and 6× retinas compared with non-light treated control retinas (Figs. 5D–5F). In addition, we observed a significantly greater number of HuC/D-positive cells in the inner plexiform layer (Fig. 5F,G arrowhead; p < 0.05) and outer retinal layer (Figs. 5F, 5G; arrow; p < 0.05) in the 6× retinas compared with the 1× retinas. This is consistent with reports demonstrating that Müller glial cells can produce excess neurons even after one round of regeneration (Hitchcock & Raymond, 1992; Powell et al., 2016; Sherpa et al., 2008; Sherpa et al., 2014).

Lastly, we compared photoreceptor numbers in untreated control retinas with 1× and 6× retinas at 28 dpL by immunolabeling the rod and cone photoreceptor populations with cell-specific antibodies. Immunolocalization of rhodopsin (rods), Zpr-1 (red-green double cones), blue opsin (long single cones), or UV opsin (short single cones) showed that retinas subjected to 1× and 6× rounds of light treatment regenerated both rods and cones to levels observed in undamaged control retinas (Fig. 6). In contrast, blue cones exhibited reduced regeneration in the 6× retinas, but not the 1× (Fig. 6O, p < 0.02) and UV cones regenerated in elevated numbers in the 1×, but not the 6× retinas (Fig. 6P, p < 0.03). These data indicated that although six rounds of light damage results in a significant increase in the acute gliotic response to damage, the Müller glia largely retained their innate ability to regenerate photoreceptors following multiple rounds of phototoxic lesion.