Activating the regenerative potential of Müller glia cells in a regeneration-deficient retina

Regeneration responses in animals are widespread across phyla. To identify molecular players that confer regenerative capacities to non-regenerative species is of key relevance for basic research and translational approaches. Here, we report a differential response in retinal regeneration between medaka (Oryzias latipes) and zebrafish (Danio rerio). In contrast to zebrafish, medaka Müller glia (olMG) cells behave like progenitors and exhibit a restricted capacity to regenerate the retina. After injury, olMG cells proliferate but fail to self-renew and ultimately only restore photoreceptors. In our injury paradigm, we observed that in contrast to zebrafish, proliferating olMG cells do not maintain sox2 expression. Sustained sox2 expression in olMG cells confers regenerative responses similar to those of zebrafish MG (drMG) cells. We show that a single, cell-autonomous factor reprograms olMG cells and establishes a regeneration-like mode. Our results position medaka as an attractive model to delineate key regeneration factors with translational potential.


Introduction
The ability to regenerate individual cells, lost organs or even the structure of the entire body is widespread in the animal kingdom. The means by which certain species achieve remarkable feats of regeneration whereas others have restricted or no capacity to do so is poorly understood. Teleost fishes are widely used models to study development, growth and regeneration of the visual system (Centanin et al., 2011;Raymond et al., 1988Raymond et al., , 2006Rembold et al., 2006). The retina of these fish undergoes lifelong neurogenesis, and the range of retinal cell types is generated from two sources. The first are the cells of the ciliary marginal zone (CMZ), which include retinal stem cells that give rise to progenitor cells and ultimately differentiated cell types of the growing neural retina (Centanin et al., 2011(Centanin et al., , 2014aRaymond et al., 2006). The second source of new retinal cells are Müller glia (MG) cells, which generate new cell types during homeostasis and regeneration (Bernardos et al., 2007). Some teleost species, including goldfish (Carassius auratus) and zebrafish (Danio rerio) have been analyzed with respect to their ability to regenerate the retina and recover visual function after injuries (Bernardos et al., 2007;Braisted and Raymond, 1992;Raymond et al., 1988;Sherpa et al., 2008). Among these, zebrafish are the most wellstudied and have been shown to contain multipotent MG cells which can self-renew and regenerate all retinal neuronal and glial cell types after injuries. It is currently assumed that other teleost species posses the same regenerative capacities, however detailed analyses have been lacking. regeneration. Using in vivo imaging, two-photon mediated specific cell ablations and lineage tracing, we find that olMG cells react preferentially to injuries of PRCs and are only able to regenerate this cell type. We demonstrate that sox2 is expressed in MG cells in the absence of injury but, in contrast to zebrafish, is not maintained in proliferating olMG cells after injury. We show that inducing targeted expression of sox2 in olMG cells is sufficient to shift olMG cells into a regenerative mode reminiscent of zebrafish, where they selfrenew and regenerate multiple retinal cell types.

Results olMG cells reenter the cell cycle after injury but do not generate neurogenic clusters
In contrast to zebrafish MG cells, olMG cells have been shown previously to be quiescent in the juvenile retina (Lust et al., 2016). In order to address the regenerative abilities of olMG cells we used the the rx2::H2B-eGFP transgenic line that labels the CMZ, olMG cells and cone PRCs but no rods in juvenile (8dpf) and adult medaka (Martinez-Morales et al., 2009, Reinhardt et al., 2015 and Figure S1). To investigate the reaction of olMG cells and the retina upon injury, we performed needle injuries on rx2::H2B-eGFP transgenic fish.
To label cells re-entering the cell cycle we subsequently analyzed the fish either by immunohistochemistry for the mitotic marker phospho-histone H3 (PH3) at 3 dpi or incubated them in BrdU for 3 days to label cells in S-phase. We detected proliferating cells in the central retina, on the basis of both labels PH3 ( Figures 1A-1A'') and BrdU (Figures 1B-1B'') 3 days after a needle injury. These proliferating cells were also positive for rx2driven H2B-eGFP, showing that the olMG cells had re-entered the cell cycle. These results demonstrate that olMG cells in juvenile medaka are quiescent in an uninjured background (Lust et al., 2016), but begin to proliferate upon injury.
The onset of MG proliferation in zebrafish has been observed between 1 and 2 dpi (Fausett and Goldman, 2006). To understand if olMG cells show a similar mode of activation, we performed BrdU incorporation experiments and analyzed time-points after injury ranging from 1 dpi until 3 dpi. At 1 dpi, no BrdU-positive cells were detected in the retina (data not shown). At 2 dpi, the first BrdU positive cells were detected in the INL and the outer nuclear layer (ONL) of the central retina ( Figure S2A-S2B'''). Co-localization with GFP showed that these cells are olMG cells or olMG-derived cells ( Figure S2A'' and S2B'').
In response to injury olMG cells initiate DNA synthesis and divide maximally once as indicated by the appearance of single or a maximum of two BrdU-positive cells next to each other in the INL at both 2 dpi and 3 dpi ( Figures 1C and 1C').
In contrast the injury response of zebrafish MG (drMG) cells at comparable juvenile stages (4dpf) is characterized by the formation of large nuclear, neurogenic clusters in the INL ( Figures 1D and 1D'). This is consistent with the response of adult drMGs to injury in which a single asymmetric division produces a MG cell and a progenitor cell that divides rapidly to generate neurogenic clusters (Nagashima et al., 2013).
These results show that olMG cells start re-entering the cell cycle between 1 and 2 dpi but do not generate neurogenic clusters.

olMG cells react preferentially to PRC injuries by apical migration
For proper regeneration to occur, the appropriate cell types must be produced. This requires not only the regulation of the proliferation of stem or progenitor cells, but also the proper control of lineage decisions in the progenitors. If and when fate decisions are made by the MG cells or proliferating progenitors during regeneration is largely unknown.
To study whether different injury sites (PRC or retinal ganglion cell (RGC) injury) result in a differential response of olMG cells, we used two-photon mediated ablations and consecutive imaging ( Figures S3A-S3D) and addressed their behavior in immediate (up to 30 hpi) and late (until 6dpi) response to injury.
We induced PRC injuries in medaka and observed that MG nuclei below the wound site started migrating apically at 17 hours post injury (hpi) (Figures 2A-2A''', see also Movie S1). These migrations were not coordinated between individual cells. Some nuclei migrated into the ONL, whereas others stayed at the apical part of the INL. Nuclei farther from the wound site did not migrate in response to the injuries. In contrast, after RGC injuries, there was no migration of MG nuclei, either apically or basally toward the wound, within the first 30 hpi (Figures 2B-2B''', see also Movie S1).
To investigate whether medaka MG nuclei migrate back at later time-points after PRC injuries or show any migratory behavior after RGC injuries, we re-imaged the injury site at two-day intervals to follow an injured retina up to 6 dpi. At 2 dpi, retinae with PRC injuries showed a gap in the INL below the injury site, at a position where MG nuclei are normally found, reflecting the migration MG nuclei towards the ONL from this location ( Figures 2C-4C''). The gap in the INL persisted until 6 dpi ( Figure 2C''). The reaction of olMG cells in retinae with RGC injuries differed. Here, we neither observed an apical nor basal migration of olMG nuclei (Figures 2D-2D'') and in fact no migration of olMG nuclei was observed at all until 6 dpi. To rule out that this is due to too little damage in the RGC layer we increased the injury size. This lead to swelling and secondary cell death in PRCs and activated olMG cells to migrate apically ( Figure S4A-S4B), indicating further that their preferential reaction is towards PRC injuries.
Taken together, these results show that olMG cell nuclei migrate towards PRC injury sites within 24 hpi and remain in this location up until 6 days, whereas they display no discernible reaction towards RGC injuries. This indicates a clear preferential reaction of olMG nuclei to refill the injured PRC layer.

olMG nuclei but not their cell bodies are depleted after PRC injuries
Long-term in vivo imaging of fish that were injured in the ONL made it apparent that olMG nuclei migrate apically into the wound site but remain there which might indicate a complete remodeling of the soma of these neuroepithelial cells. To understand whether cell bodies of the olMG cells remain intact during this nuclear migration, we observed nuclear movements (transgenic line rx2::H2B-eGFP) in the context of the olMG cell body (transgenic line rx2::lifeact-eGFP). We imaged the animals at two-day intervals following ONL injuries. As previously observed, olMG nuclei migrated out of the INL into the wound site ( Figures 3A-3A''). Cell bodies of the olMG cells spanning the entire apico-basal distance remained intact until 6 dpi in the absence of a nucleus in the INL ( Figure 3A'').
The previous site of the nucleus could still be detected by a small enlargement of the soma in this region.
Additionally, to extend the time range of analysis, we performed immunohistochemistry on fish injured in the ONL. After incubation in BrdU for 3 days and fixation at 10 dpi, we observed similar results (Figures 3B-3B''). In this procedure, we used immunohistochemistry to detect GS to label the MG processes ( Figure 3B). BrdU-positive cells in the ONL mark the location of the injury ( Figure 3B''). In the region directly below, the majority of MG nuclei, which had been labelled by rx2::H2B-eGFP, were absent from the INL ( Figure 3B''). GS-positive cell bodies remained spanning the apico-basal width, but without the apparent presence of nuclei. In contrast, unaffected GS-positive olMG cells located on either side of the wound site still contained their nuclei, as could easily be detected by the large size of the soma. This data shows that the cell bodies of injuryactivated olMG cells are still intact despite the migration of their nuclei into the ONL.

olMG cells divide in the INL with an apico-basal distribution
Since the injury response of olMG cells apparently does not involve self-renewal of olMG cells we wondered about the position and orientation of the cell division plane, a factor which has been associated with cell fate in various systems. others were located more basally (data not shown). This is in contrast to findings in zebrafish where, in a light injury paradigm, PH3-positive MG cells can be found in the ONL 2 days after injury (Nagashima et al., 2013) To address the cleavage plane of dividing olMG cells we employed in vivo imaging of

olMG cells are lineage restricted
In zebrafish drMG cells are able to regenerate all neuronal cell types and self-renew after injury (Nagashima et al., 2013;Powell et al., 2016). We followed a BrdU-based lineage tracing approach successfully applied in zebrafish (Fausett and Goldman, 2006;Powell et al., 2016) to address the potency of olMG cells. Transgenic rx2::H2B-eGFP fish retinae were injured either by two-photon laser ablation of PRCs or RGCs specifically or using a needle ablating all cell types. The injured fish were incubated in BrdU for 3 days to label proliferating cells. This allows to efficiently detect all injury triggered S-phase entry of olMG cells ( Figure  that they were not MG cells anymore ( Figure 5D), but were positive for Recoverin, a PRC marker ( Figure 5E). These results demonstrate that olMG cells do not self-renew and rather function as mono-potent repair system restricted to the generation of PRCs, most of which belong to the rod lineage.

Sox2 expression is not maintained in proliferating olMG cells after injury
The previous results show that olMG cells re-enter the cell cycle after injuries introduced by needle to the complete retina or by 2-photon ablation to the PRC layer. They regenerate PRC but do not undergo self-renewal. This suggests that olMG cells lack intrinsic factors that trigger self-renewal and multi-potency upon injury. One transcription factor which is well known for its involvement in the self-renewal of stem cells -particularly neural stem cells -is Sox2 (Sarkar and Hochedlinger, 2013). It has been shown that cells expressing sox2 are capable of both self-renewal and the production of a range of differentiated neuronal cell types (Sarkar and Hochedlinger, 2013). Data from zebrafish have shown that a ubiquitous gain of Sox2 expression triggers a regenerative response of the drMGs in the absence of injury (Gorsuch et al., 2017).
To investigate the expression of sox2 in MG cells, we performed immunohistochemistry on uninjured retinae in medaka and zebrafish. In the medaka retina, Sox2 protein is detected the in the CMZ (data not shown) as well as in amacrine cells (ACs) and olMG cells ( Figure 6A-6A'''). In zebrafish, the pattern was similar: Sox2 protein was present in the CMZ (data not shown), in ACs and drMG cells ( Figure 6B-6B'''). This data is consistent with data from other vertebrates including human, whose MG cells also maintain sox2 expression (Gallina et al., 2014).
To investigate the expression of sox2 after injuries in olMG and drMG cells respectively, we performed needle injuries, incubated the fish in BrdU and fixed them between 1 and 4 dpi. Proliferating olMG cells exhibited a downregulation in the expression of sox2, which was mostly absent by 3 dpi ( Figures 6C-6D, 6% of all BrdU positive cells were Sox2positive). Conversely, in zebrafish, sox2 expression could still be detected after 3 days in drMG cells that strongly proliferated in response to injury ( Figures 6E-6F, 84% of all BrdU positive cells were Sox2-positive). These findings strongly argue that the downregulation of Sox2 expression restricts the regenerative properties of olMG cells.

Sustained Sox2 expression restores olMG driven regeneration
The results presented above indicate that after injury, olMG cells and MG-derived progenitors do not maintain an expression of sox2, in contrast to the situation in zebrafish.
We hypothesize that the prolonged sox2 expression facilitates drMG cells to undergo selfrenewal and to generate neurogenic clusters and ultimately all cell-types necessary to regenerate a functional retina.
To test this hypothesis, we chose the inducible LexPR system (Emelyanov and Parinov, 2008) targeted to olMG cells to sustain sox2 expression. In those retinae, we ablated all retinal cell types by a needle injury and performed BrdU mediated lineage tracing as described above. To induce the persistent expression of sox2 in olMG cells after injuries, we employed the LexPR transactivation system (rx2::LexPR OP::sox2, OP::H2B-eGFP) allowing to follow individual sox2 expressing cells by the nuclear eGFP expression. We induced sox2 expression for two days and provided BrdU in parallel, performed a needle injury to ablate all cell types and maintained the expression of sox2 until 3 dpi. After a chase until 14 dpi the retinae and regenerated cell types were analyzed ( Figures 7A and   7B). In needle injured wildtype fish, olMG cells did not self-renew and only gave rise to These data indicate that a targeted maintenance of sox2 expression after injury is sufficient to induce self-renewal and increase potency in MG cells in the medaka retina turning a mono-potent repair system into a regeneration system with increased potency.

Discussion
Here, we have characterized a differential regenerative response between two teleost fish and used it as a framework to address the molecular determinants of regeneration during evolution. By using a combination of in vivo imaging, targeted cell type ablation and lineage tracing, we investigated the dynamics of the injury response in the medaka retina.
We focused on MG cells, which play a prominent role in zebrafish retinal regeneration.
While upon injury olMG cells re-enter the cell cycle, they fail to undergo self-renewal.
Furthermore, olMG cells do not generate the neurogenic clusters which arise in zebrafish, nor do they produce all neuronal cell types in the retina. We traced this effect prominently to Sox2, the expression of which is maintained in proliferating drMG cells after injury, but not in olMG cells. We demonstrated that the sustained expression of sox2 is sufficient to convert an olMG into a dr-like MG. The fact that this response is acquired cell autonomously and in the context of a non-regenerative retina can be relevant for putative translational approaches. Since olMG cells did not self-renew after injuries and only had the capacity to regenerate PRC, olMG cells are not true multipotent retinal stem cells. Instead, olMG cells should be considered lineage-restricted progenitors. They re-entered the cell cycle between 1 and 2 dpi, similar to the re-entry observed in zebrafish. This indicates that the signals that are essential for cell-cycle re-entry are present in medaka and are activated in a window of time similar to that of zebrafish. However, since olMG cells did not self-renew, the first cell division is presumably symmetric. Given extensive data that the orientation of the In the uninjured retina, olMG cells express sox2, as is the case for many other vertebrates, including humans. However, sox2 expression in olMG cells is downregulated in response to injury, in contrast to the injury response of drMG cells, which upregulate sox2 (Gorsuch et al., 2017). We speculate that this upregulation is due to epigenetic modifications of the sox2 locus. A recent study in the mouse retina showed that the expression of oct4 is upregulated shortly after injury and then downregulated at 24 hpi (Reyes-Aguirre and Lamas, 2016). This correlates with a decrease in the expression of DNA methyltransferase 3b and its subsequent upregulation at 24 hpi, triggering a decrease in methylation and subsequent re-methylation of oct4. Furthermore, a recent study on zebrafish regeneration discovered the existence of so-called tissue regeneration enhancer elements (TREEs) (Kang et al., 2016). One TREE was associated with leptin b, which is expressed in response to injuries of the fin and heart. This TREE acquires open chromatin marks after injury, can be divided into tissue-specific modules and can drive injury-dependent expression in mouse tissue. This raises the possibility that the sox2 locus in olMG cells experiences epigenetic modifications after injury which differ from modifications in zebrafish. The fact that sox2 expression is detected in all vertebrate MG cells analyzed to date in the absence of injury raises the question whether a decrease in sox2 expression after injury might be a common feature of non-regenerative species, like chicken, mouse and even humans. Data from a conditional sox2 knockout in mouse shows that Sox2 is necessary for maintenance of MG morphology and quiescence (Surzenko et al., 2013). The results shown here may provoke an evolutionary question: is retinal regeneration an ancestral or derived feature within the infraclass of teleosts? The question might be resolved by investigations of this capacity in other fish species more closely related to medaka, such as Xiphophorus maculatus, whose last common ancestor with medaka lived around 120 million years ago (Schartl et al., 2013). Additionally, species like the spotted gar, whose lineage diverged from teleosts before teleost genome duplication (Braasch et al., 2016), might provide insights about the ancestral mode of retinal regeneration.
Recently, the retinal architecture of the spotted gar has been analyzed (Sukeena et al., 2016). Here, proliferative cells have been detected in the central retina likely representing proliferating MG cells, which generate rod PRCs during homeostasis as seen in zebrafish, suggesting that regeneration is indeed an ancestral feature in the ray-finned fish lineage.
With a potential translational perspective, regenerating and non-regenerating systems can now be systematically compared to delineate the underlying factors and mechanisms.
To date, our cumulative results show that the regenerative potential of olMG cells in the context of homeostasis and injury in medaka resemble that of mammals and birds more than zebrafish. We propose that this provides an added value to medaka as a model species for regeneration studies that bridge the differences between zebrafish and mammals. Studies of heart regeneration that have compared zebrafish and medaka lend

BrdU incorporation For BrdU incorporation, fish were incubated in 2.5 mM BrdU diluted in 1x Embryo
Rearing Medium (ERM) or 1x Zebrafish Medium for respective amounts of time.

Induction of the Lex PR system
For induction of the Lex PR system, fish were induced by bathing them in a 5 µM to 10 µM Mifepristone solution in Embryo Rearing Medium (ERM) for respective times.

In vivo imaging and laser ablations
For in vivo imaging fish in a Cab background were kept in 5x 1-phenyl-2-thiourea (PTU, Sigma) in 1x ERM from 1 dpf until imaging to block pigmentation. Fish in a QuiH background could be imaged without any treatment. Fish were anesthetized in 1x Tricaine diluted in 1xERM and mounted in glass bottomed Petri dishes (MaTek) in 1% Low Melting Agarose. The specimens were oriented lateral, facing down, so that the right eye was touching the cover-slip at the bottom of the dish. Imaging and laser ablations were performed on a Leica SP5 equipped with a Spectra Physics Mai Tai ® HP DeepSee Ti:Sapphire laser, tunable from 690-1040nm and Leica Hybrid Detectors. A wound was introduced using the bleach point function or the region of interest function, together with the high energy 2-photon laser tuned to 880nm. Follow-up imaging was performed using same laser at 880nm and a 40x objective.

Retinal needle injuries
Larvae (zebrafish 5dpf, medaka 8dpf) were anesthetized in 1x Tricaine (A5040, Sigma-Aldrich) in 1x ERM and placed on a wet tissue. Under microscopic visualization, the right retina was stabbed multiple times in the dorsal part with a glass needle (0.1 mm diameter). Left retinae were used as controls.

Immunohistochemistry on cryosections
Fish were euthanized using Tricaine and fixed over night in 4% PFA, 1xPTW at 4°C. After fixation samples were washed with 1x PTW and cryoprotected in 30% sucrose in 1xPTW.
To improve section quality, the sections were incubated in a half/half mixture of 30% Washes were performed with 1x PTW instead of PBS.

Immunohistochemistry imaging
All immunohistochemistry images were acquired by confocal microscopy at a Leica TCS SPE with either a 20x water objective or a 40x oil objective.

Image processing and statistical analysis
Images were processed via Fiji image processing software. Statistical analysis and graphical representation of the data were performed using the Prism software package (GraphPad).   (A-A'') In vivo imaging of a juvenile rx2::H2B-eGFP, rx2::lifeact-GFP medaka retina which was injured in the ONL (asterisk) and imaged every second day after injury. Close to the injury site an olMG process without a nucleus can be detected at 2 dpi (A, arrowhead).
The empty process remains until 6 dpi (A") (n=3 fish, data obtained from three independent experiments). Scale bar is 10 μm.                Figure S6