In mammals, disease or injury of the retina leads to irreparable vision loss. Conversely, some teleost fish mount a regenerative response to retinal injury that is capable of restoring lost sight (Lindsey and Powers, 2007; Sherpa et al., 2008). Müller glial cells (MGs), the main glial cells of the retina, are key to successful regeneration. In response to a harmful stimulus they can rapidly dedifferentiate acquiring neural stem cells properties, proliferate and migrate to the damaged retinal layer and differentiate into lost neurons (Fischer and Reh, 2003; Ooto et al., 2004). In mammals, MGs generally respond to injury by reactive gliosis accompanied by hypertrophy, but they rarely re-enter the cell cycle and regenerate new neurons (Bringmann et al., 2009; Burda and Sofroniew, 2014). These data suggest that the ability of MGs to dedifferentiate in response to retinal injury would be a crucial difference between the regenerative responses of fish and mammals (Ramachandran et al., 2010). In this sense, it is not yet known how to unlock the neurogenic potential of the mammalian MGs.

As part of the central nervous system, the retina shares the characteristic high metabolism of the brain. It is currently accepted that many retinopathies like age-related macular degeneration (Zhang et al., 2020), diabetic retinopathy (Liew et al., 2017) and retinitis pigmentosa (Fu et al., 2019) could be associated with both systemic and tissular metabolic impairment. Müller cells play a critical role in retinal metabolism and are among the first cells to demonstrate metabolic changes in retinal stress or disease (Bringmann et al., 2009). Reactive gliosis of MGs is characterized by non-specific and specific responses. These responses involve changes in their morphology, biochemistry and physiology (Bringmann et al., 2009). Early after injury, gliosis plays a neuroprotective role releasing neurotrophic factors and antioxidants. However, if gliosis persists, the dedifferentiation of cells contributes to neuronal cell death, e.g., via the impairment of neurotransmitter removal promoting excitotoxicity (Newman, 2004). Generally, an impairment of the supportive functions of Müller cells may have an additive effect on dysfunction and neuron loss by increasing the susceptibility of neurons to stressful stimuli in the diseased retina. Thus, and considering that many metabolic diseases are often accompanied by neuronal dysfunction, a proper understanding of the mechanisms implicated in Müller cells under metabolic impaired conditions is essential for the development of efficient therapeutic strategies.

Somatic reprogramming and pluripotency are characterized by a transition from oxidative metabolism into anaerobic glycolysis (Folmes et al., 2011). Interestingly, the NAD⁺-dependent histone deacetylase SIRT6 has been shown to act as a key modulator of glucose homeostasis and its absence, or inactivation, favors the switch to anaerobic glycolysis by the upregulation of the levels of the glucose transporter GLUT1 and of several glycolytic genes (Zhong et al., 2010). Furthermore, SIRT6 was described as a direct regulator of the expression of several core pluripotent genes by the dependent deacetylation of acH3K56 and acH3K9 during embryonic stem cells (ESC) differentiation (Etchegaray et al., 2015), further suggesting a key role of SIRT6 as a regulator of these cells’ pluripotency capacity. Moreover, we have recently demonstrated that SIRT6 would regulate early neurodegenerative events observed during Diabetic Retinopathy development by modulating the expression of key neural factors in Müller cells (Zorrilla-Zubilete et al., 2018). In this work we aimed to study the effect of glucose variations and oxidative stress in Müller cells’ reprogramming capacity and analyze the participation of SIRT6 as a potential epigenetic modulator of this process.

For this purpose, Müller glial cells were cultured with different glucose concentrations and under oxidative stress conditions. We analyzed the levels of the glial cell marker glutamine synthetase (GS) and the reprogramming factor SOX9. Additionally, we tested MGs migration capacity under these stimuli. We studied the involvement of SIRT6 in the reprogramming capacity by analyzing SIRT6 levels and the acetylation status of one of its substrates (acH3K9). We also evaluated the levels of SOX9 in siRNA SIRT6-silenced cells. We also determined the levels of SOX9 in retinas from mice with a conditional deletion of SIRT6 in the CNS (Nes-Cre^−/−) and performed gene and transcription factor enrichment analysis of retinal Müller cells from a murine model of Diabetes. We found that the combination of high glucose (HG) concentrations and oxidative stress may induce some degree of dedifferentiation of MGs and favors their migration ability. We also observed that HG induced an increase in the levels of the well-known pluripotent factor SOX9 and a decrease in SIRT6 levels accompanied by the increase in the acetylation levels of H3K9. Notably, inhibiting SIRT6 expression in MGs by siRNA rendered an increase in SOX9 levels. We noticed a moderate, though not significant, increment in SOX9 levels in retinas from Nes-Cre^−/− mice suggesting that in vivo modulation of this transcription factor by SIRT6 would be indirect and may involve other players. Gene ontology (GO) enrichment analysis showed differentially expressed genes in categories related to metabolism, cell migration and pluripotency in MGs from diabetic retinas. Interestingly, we found that many functional categories affected in cells from these animals were directly related to SIRT6 function like glycolysis, Hif1α pathway and phototransduction. Moreover, we identified 67 predicted transcription factors, including SOX9, that could be responsible for the transcriptional changes observed in diabetic MGs. The GO enrichment analysis of these TFs showed that these genes are enriched in biological processes linked to cellular and brain development as well as neurogenesis and pluripotency. The diversity of Müller cells’ response to a metabolic defy represents a major challenge in the understanding of the regulation of reprogramming processes in vivo.

The first step for the Müller cells to mount a reprogramming process requires the dedifferentiation of their phenotype by downregulating the expression of glial specific markers such as glutamine synthetase (GS). To analyze the effect of metabolic impairment on MGs’ reprogramming capacity we determined the levels of GS in MGs cultured in low (5 mM) or high (25 mM) glucose concentrations. We did not find a significant difference in the levels of this marker between these conditions (Figure 1A, left panels). Since metabolic impairment is directly associated with the production of reactive oxygen species (ROS), we studied the effect of oxidative stress induced by the exposure to oxygen peroxide (H₂O₂) on GS levels. Different concentrations of H₂O₂ were tested in order to find the optimal dose that generates oxidative damage without affecting cell viability (Supplementary Figure S1). We observed that the combination of high glucose and H₂O₂ (100 µM) was able to induce a significant downregulation of GS levels suggesting that this experimental condition may induce some degree of dedifferentiation (Figure 1A, lower-right panel). Western blot analysis confirmed this result (Figure 1C). This effect may be transient since GS levels showed a tend to increase when measured at longer periods of H₂O₂ incubation (Supplementary Figure S2).

Concomitantly to the downregulation of cell-specific markers, MGs would have to upregulate the levels of pluripotent or reprogramming factors. SOX9 is a high mobility group box transcription factor required for the establishment and maintenance of neural stem cells in both embryonic and adult CNS (Scott et al., 2010), and has been described to be expressed in adult Müller glia and retinal pigment epithelium (RPE) cells (Poché et al., 2008). We found that HG exposure was able to induce an increase in SOX9 levels in MGs, while oxidative stress, either applied individually or in combination with HG, had no effect on the levels of this factor (Figure 2A). High glucose effect on SOX9 levels was confirmed by immunofluorescence analysis (Figure 2B).

It has been demonstrated that the NAD + dependent histone deacetylase SIRT6 is a key modulator of glucose homeostasis. Moreover, several pluripotent genes like Oct4, Sox2 and Nanog were described to be directly regulated by SIRT6 during embryonic stem cells (ESC) differentiation (Etchegaray et al., 2015). These observations prompted us to analyze SIRT6 involvement in the reprogramming capacity of MGs. Since Figure 2 showed that only glucose variations exerted a significant effect in SOX9 levels, we focused only in the effects of glucose over SIRT6 levels. We analyzed this parameter in MGs and observed that HG induced a decrease in SIRT6 expression levels (Figure 3A) accompanied by an increase in the acetylation status of one of its substrates (acH3K9) (Figure 3B). Remarkably, the silencing of SIRT6 by siRNA rendered an increase in SOX9 levels suggesting that this reprogramming factor could also be regulated by SIRT6 in these cells (Figure 3C).

Another step in the reprogramming process includes the migration of the cells to the layer of the damaged neurons. We tested the migration capacity of MGs by using the wound healing assay. While glucose variations did not induce a significant effect in the migration ability, the combination of HG and oxidative stress induced a decrease in the gap size between t0 and t18 indicating that these experimental conditions would favor this process (Figure 4). Oxidative stress had no significant effects under low glucose conditions (not shown).

In order to validate our in vitro results, we determined SOX9 levels in retinas from Nes-Cre^−/− mice. These animals bear a specific deletion of SIRT6 in the CNS and, hence, in their retinas (Zorrilla-Zubilete et al., 2018). Immunofluorescence studies showed that SOX9 is specifically expressed in Muller cells as it co-localizes with GS and in the retinal pigment epithelium (RPE) (Figures 5A,B). It was observed a moderate, though non-significant, increase in SOX9 levels in retinas from Nes-Cre^−/− mice compared to Nes-Cre^+/+ (Figures 5C,D). This result suggests that SIRT6 regulation of SOX9 levels in vivo may not be direct and probably involves other factors.

To deepen the in vivo regulatory pathways altered in Müller cells under a metabolic impaired environment, we evaluated the gene expression changes of a published dataset generated by using retinal Müller cells from diabetic rats compared to cells from healthy controls (Gerhardinger et al., 2005). Since in a diabetic state retinas are subjected to glucose variations and oxidative stress damage, among other stressors, we considered that this model would suit our experimental system. The differential expression analysis showed that under diabetic conditions, 163 genes changed their expression in Müller cells. While 146 increased their transcript levels, 17 decreased their expression levels (Supplementary Table S1). To determine the biological relevance of these transcriptional changes, we performed a Gene Ontology enrichment analysis. As expected, the transcriptomic profile showed that the enriched genes were associated with the glucose and monosacharyde metabolism, and ketone bodies synthesis and usage (Figures 6A,B; Supplementary Table S2). Interestingly, and in accordance with our in vitro observations, the enrichment analysis showed that metabolic impairment positively regulates several genes involved in glial cells migration (Figure 6C). This positive impact on cellular migration may mainly be due to changes in metalloproteases, gelatinases, and growth factors such as FGF (Figure 6D). Additionally, we observed a consistent enrichment in pathways associated with organogenesis, embryonic and brain development, and neurogenesis (Figure 6E). The genes mapped into developmental and neurogenesis ontologies include transcription factors such as FOXG1 and HES1, and growth factors like FGFR and NeuroD1 (Figure 6F). To validate these results, the GO enrichment was corroborated with the Reactome database obtaining similar results (Supplementary Table S3). Upregulated genes were referenced to the KEGG pathways database and remarkably, we found that many functional categories affected in cells from diabetic animals were directly related to SIRT6 function like glycolysis, Hif1α pathway and phototransduction (Supplementary Table S4). Lastly, to determine which transcription factors (TF) could be responsible for the transcriptional changes observed in MGs from diabetic rats, we used two approaches to perform TF enrichment analysis. We used the chromatin enrichment (ChEA) algorithm that includes results of ENCODE chromatin immunoprecipitation (ChIP), and the Opossum 3.0 algorithm that predicts TF binding sites. Figure 6G shows that 67 TFs are shared between ChEA and Opossum enrichment analysis. Interestingly, we found SOX9 among the top 10 transcription factors in terms of ChEA score, confirming that this TF may have a key role in regulating the differential expression program observed in Müller cells under metabolic impaired conditions (Figure 6G; Supplementary Tables S5–S7). Moreover, the GO enrichment analysis of these predicted 67 TFs showed that these factors are enriched in biological processes like cell differentiation and development (including brain and eye development) as well as neurogenesis (Figure 6H; Supplementary Table S7). Likewise, we performed an enrichment analysis using the list of TFs generated by the ChEA algorithm and validated with Reactome, and observed several TFs that may be involved in regulating the acquisition of a pluripotent phenotype like KLF4, NANOG, SOX2, POU5F1 and LIN28 (Supplementary Figure S3), which are also regulated by SIRT6 (Etchegaray et al., 2015; Kugel et al., 2016). Further studies are needed in order to elucidate the role of other pathways that may have opposite effects in vivo, thus preventing MGs to adopt a fully reprogramming phenotype in physiological conditions.

Increasing evidence suggests that metabolic alterations strongly influence the initiation and progression of neurodegenerative disorders including retinopathies (Lin and Beal, 2006; de la Montea and Tongd, 2014; Kim et al., 2017; Harder et al., 2020). Cells of the CNS that do not cope properly with the metabolic, morphological and neurophysiological changes needed to maintain the homeostasis of the tissue would eventually contribute to the development of neurodegenerative processes. As a highly metabolic tissue, the retina needs to adapt as well. Müller cells, the major glial component of the retina, play a critical role in retinal metabolism and are among the first cells to demonstrate metabolic changes in retinal stress or disease. They are under the focus of current investigations due to their potential regeneration capacity. Adult Müller glial cells retain a progenitor cell-like gene expression profile that makes them a suitable source of regenerative cells (Das et al., 2006). Since the availability of glucose and oxygen are important regulatory elements that may help directing stem cell fate, in this work we aimed to study the effect of glucose variations and oxidative stress in the reprogramming capacity of Müller cells and tried to identify the epigenetic mechanisms that regulate this process by studying the role of the histone deacetylase SIRT6.

In order for MGs to initiate a reprogramming process they have to adopt a dedifferentiated phenotype by downregulating their specific glial markers such as glutamine synthetase (GS). We analyzed GS levels in MGs cultured under different metabolic settings and found that the combination of high glucose and oxidative stress was able to induce a significant decrease in GS levels suggesting that this experimental condition may cause some degree of dedifferentiation. The balance between pluripotent embryonic and multipotent adult stem cells may be related to metabolism variations which, in turn, influence the cellular redox state. In this sense, changes in the levels of reactive oxygen species (ROS) could mediate the interplay between metabolism and stem cell fate (Ryu et al., 2015). Our observations are in accordance with the fact that optimal ROS levels are required for the onset of a reprogramming process (Zhou et al., 2016). Moreover, GS levels variations may explain the reduced dependence on exogenous glutamine, which is a feature of pluripotent stem cells (Vardhana et al., 2019). However, since GS downregulation may reduce the protection against glutamate excitotoxicity, this decrease should be temporary to avoid further damage of the tissue under a harmful stimulus.

In parallel with the downregulation of specific markers, MGs would have to upregulate the expression of pluripotent or reprogramming factors. The high mobility group box transcription factor (TF) SOX9 was demonstrated to be indispensable for the establishment and maintenance of neural stem cells in both embryonic and adult CNS (Scott et al., 2010). It was described to be vital in triggering the switch from the neurogenic to the gliogenic program at the germinal zones of different neural tissues. During retinogenesis, Sox9 is expressed in multipotent progenitor cells, although in the adult retina it is expressed only in Müller glia and RPE cells (Poché et al., 2008). We determined that high glucose induced an increase in SOX9 levels in MGs; while oxidative stress, applied individually or in combination with HG, had no effect on the levels of this factor. Zhou et al. showed that the addition of exogenous hydrogen peroxide induced a biphasic pattern of mouse embryonic fibroblasts (MEFs) reprogramming capacity. They observed a tendency towards an increase in the yield of induced pluripotent stem cells (iPSC) derived from MEFs treated with a low dose of H₂O₂, as well as an impaired reprogramming efficiency with higher doses (Zhou et al., 2016). Further studies using different incubation times with H₂O₂ will help us define whether a transient acquisition of a pluripotent phenotype could be achieved in vitro.

The NAD⁺-dependent histone deacetylase SIRT6 has been described as a critical modulator of metabolism by repressing the expression of several glycolytic genes (Zhong et al., 2010) and as a key regulator of the retinal function (Silberman et al., 2014). We determined that HG was able to induce a decrease in SIRT6 levels and an increase in acH3K9 levels in Müller cells, as seen previously in primary cell cultures (Zorrilla-Zubilete et al., 2018). Since glycolytic enzymes represent candidate targets for enhancing somatic reprogramming, inducing the expression and/or activity of glycolytic enzymes in Müller cells would improve their reprogramming capacity. Remarkably, SIRT6 was described to regulate embryonic stem cell differentiation by modulating the acetylation status of its substrates at the promoter level of several pluripotent genes like sox2, nanog and oct4 (Etchegaray et al., 2015). In this sense, we found that the downregulation of SIRT6 expression by siRNA rendered an increase in SOX9 levels as seen in HG-treated cells. Considering that stem cells have been described to prefer aerobic glycolysis rather than oxidative phosphorylation (Bigarella et al., 2014; Zheng et al., 2016), and that an increment in ROS was shown to be required for the onset of a reprogramming process, the modulation of the shift in the metabolic imbalance between these two processes may play a critical role in defining whether cells are in quiescent, pluripotent, proliferative, or differentiating states (Shyh-Chang et al., 2013). The downregulation of SIRT6 levels by the exposure to high glucose might induce the expression and activity of glycolytic enzymes and the expression of pluripotent genes enhancing the reprogramming capacity of MGs thus facilitating a regenerative response. This observation is in accordance with recent works that show that a histone deacetylase inhibitor promotes chromatin accessibility at key gene loci in MGs allowing a more effective reprogramming process (Jorstad et al., 2017), and that several pluripotent transcriptional factors are increased in SIRT6-knockout MEFs (Xu et al., 2019). Additional studies will help us define the role of SIRT6 in MG’s reprogramming capacity. It is worth noting that metabolic impairment may induce an increased burden of DNA damage causing cells to accumulate genomic instability. The concentration of H₂O₂ used in our in vitro experiments was able to induce DNA damage as seen by the increase in ɣH2AX levels depicted in Supplementary Figure S1. SIRT6 promotes DNA double-strand break (DSB) repair contributing to the maintenance of genome stability (Toiber et al., 2013; Onn et al., 2020). Thus, if DNA repair becomes a priority under oxidative stress conditions, a redistribution of SIRT6 to sites of breaks, and the consequent reduction of its levels or activity in other genome locations, may lead to transcriptional misregulation favoring metabolic imbalance and neurodegeneration. Moreover, DNA damage can activate the poly ADP-ribose polymerase (PARP1) that catabolizes NAD+, and the extent of NAD+ depletion may likely correlate with the extent of the damage. Several studies have shown that the effect of H₂O₂ on NAD+ levels depends on the cell type and the concentration of peroxide used (Zhu et al., 2016; Ye et al., 2019; Martín-Guerrero et al., 2020). Additional studies are needed in order to determine if NAD+ levels are affected in our system, and if so, whether some of the peroxide effects could be reversed by using a NAD+ precursor.

Mice with a conditional deletion of SIRT6 in the CNS (Nes-Cre^−/−) have been shown to exhibit a neurodegenerative phenotype, including behavior impairment, learning and memory deficit. They also showed neurodegenerative markers like increased DNA damage and apoptotic cells, and augmented Tau phosphorylation in the brain (Kaluski et al., 2017; Portillo et al., 2021). Moreover, retinas from these mice presented structural and molecular evidence of neurodegeneration (Zorrilla-Zubilete et al., 2018). We found that retinas from Nes-Cre^−/− mice showed a moderate, yet non-significant, increase in SOX9 levels compared to Nes-Cre^+/+ suggesting that in vivo regulation of this factor by SIRT6 may be more variable and may involve multiple players. It is worth mentioning that not all Müller cells of a retina may respond to a pathogenic stimulus in the same way (Bringmann et al., 2009). Such a heterogeneity between neighboring Müller cells in the same region of the retina would explain the lack of significance of SOX9 levels observed in the tissue and reveal how challenging it could be to identify in vivo the subpopulation/s of these cells with reprogramming capacity.

Another key step during reprogramming involves the migration of undifferentiated cells to the site of damage. We tested the migration capacity of MGs and found that, while glucose variations did not induce a significant effect, the combination of HG and oxidative stress stimulates this process. This result differs from previous observations in retinal astrocytes (Shin et al., 2014) and other cell types (Lamers et al., 2011). However, activated microglia results in increased proliferation and migration as described in diabetic retinopathy, a disease that is a direct consequence of sustained hyperglycemia and oxidative stress induced during Diabetes (Altmann and Schmidt, 2018). Thus, further studies will be necessary to define the signals that modulate the migration capacity of retinal Müller cells during a reprogramming process which may differ from the ones involved in a pathological disorder.

In an attempt to better understand the regulatory pathways altered in Müller cells under a metabolic impaired environment like the one in a diabetic state, we evaluated the gene expression profile of Müller cells from a published dataset generated by using murine a model of Diabetes compared to cells from healthy controls (Gerhardinger et al., 2005). The differential expression analysis showed 146 upregulated transcripts and 17 downregulated transcripts under diabetic conditions. To determine the biological relevance of these transcriptional changes, we did a Gene Onthology enrichment analysis that revealed, as expected, that the transcriptomic change was enriched in genes associated with glucose and monosacharyde metabolism, as well as ketone bodies synthesis and usage. Interestingly, some of the observed changes were also implicated in lipid metabolism, mainly due to the change in the expression of the transcription factors PPARα and NF-κβ, and the expression of glycolytic regulatory enzymes HK2 and PFKFB2. Moreover, in accordance with our in vitro observations, the enrichment analysis also showed that a diabetic state positively regulates genes related to migration processes. This positive impact on cellular migration may be due to different genes that code for metalloproteases, gelatinases, and growth factors such as FGF. Furthermore, we observed a consistent enrichment in pathways associated with organogenesis, embryonic and brain development, and neurogenesis. Interestingly, when upregulated genes were referenced to the KEGG pathways database we found that many functional categories affected in MGs from diabetic animals were directly related to SIRT6 function like glycolysis, Hif1α pathway and phototransduction. These observations support previous findings from our group and others, further endorsing the role of SIRT6 as a key modulator of glucose homeostasis and retinal function (Zhong et al., 2010; Silberman et al., 2014).

To determine which transcription factors could be responsible for the transcriptional changes observed in diabetic Müller glial cells, we performed two TF enrichment analysis. We used a ChEA algorithm that includes the ENCODE ChIP results in its analysis and the Opossum 3.0 algorithm that predicts putative TF binding sites. When we crossed the results from both analyses, we obtained a list of 67 shared transcription factors, which includes SOX9 among the top 10 factors, confirming that this TF may play a strategic role in regulating the differential expression program observed in Müller cells under metabolic impaired conditions. The GO enrichment analysis of these predicted 67 TFs showed that these factors are enriched in biological processes like cell differentiation and development (including brain, visual systems and eye development) as well as neurogenesis and stem cell differentiation. Additionally, we used the list of TFs generated by the ChEA algorithm to perform an enrichment analysis and found several factors that could be involved in the acquisition of a pluripotent phenotype like KLF4, LIN28, SOX2, NANOG and POU5F1 (Supplementary Figure S3). It is worth noting that these TFs were described to be regulated by SIRT6 (Etchegaray et al., 2015; Kugel et al., 2016) further validating the role of this enzyme in regulating reprogramming processes and modulating stem cells’ fate during development.

Our results suggest that metabolic impaired Müller glial cells could alter their genetic landscape favoring a phenotype associated with a developmental profile that shares some characteristics with a reprogramming state. Müller cells response to a metabolic defy in vivo may involve other regulatory pathways that exert opposite inhibitory effects preventing MGs to adopt a fully reprogramming phenotype in physiological conditions. This complexity exposes the need of additional studies that will help us understand the mechanisms that regulate MGs response under metabolic impaired conditions. This, in turn, would help develop new therapeutic strategies to treat neurodegenerative diseases.