Adult (6–10 weeks) C57BL/6J mice were used for all lesion experiments and subsequent analysis. Animals were housed under standard conditions with access to water and food ad libitum. All procedures were performed under protocols approved by the internal animal care committee (CICUAL–CINVESTAV), and following the guidelines of the Association for Research in Vision and Ophthalmology (ARVO) Statement for the Use of Animals in Ophthalmic and Vision Research.

Mice were anesthetised with pentobarbital (0.1 mg per 10 g body weight) by an i.p. injection. Tetracaine topical anesthetic was applied to the eyes before the intravitreal injection, as well as a tropicamide/phenylephrine ophthalmic solution. The animals were then placed on a head mount, and a 30G needle was carefully inserted at the upper temporal ora serrata in the left eye, delivering 250 mM N-methyl-D-aspartate (NMDA) (Sigma) into the vitreous humor. The right eye remained intact as a control. After the injection, mice were returned to their home cages for recovery, with food and water ad libitum. Eyes were enucleated at 4, 12, 18, and 24 h post injury (hpi), and prepared for further analysis.

After enucleation, eyes were fixed with a 4% paraformaldehyde (PFA) solution for 1 h at 4°C. Following fixation, the anterior region was removed and the posterior part cryopreserved in a 30% sucrose solution overnight at 4°C. Then, eyes were embedded in OCT compound (Tissue-Tek, Sakura Finetek, USA) and frozen for cryo-sectioning with a microtome (Leica), obtaining 10 μm retinal sections placed on coated slides (Biocare Medical, USA), which were air-dried for 1 h at room temperature (RT) and then hydrated for 30 min with phosphate buffered saline (PBS). The slides were placed in a wet chamber and blocked with a 0.3% Triton X-100 solution containing 1% bovine serum albumin (BSA) and 5% goat serum for 1 h at RT. Primary antibodies against nestin (1:200; Millipore, Billerica, MA, USA, MAB353), and glutamine synthase (GS) (1:100; Abcam, AB16802), were diluted in PBS and allowed to bind to the slides overnight at 4°C. After several rounds of washing, the slides were incubated with the secondary antibodies Alexa 488 anti-mouse (1:500; Molecular Probes, Eugene, OR, USA) and/or Alexa 568 anti-rabbit (1:500; Molecular Probes, Eugene, OR, USA), diluted in PBS with 4′,6-diamidino-2-phenylindole (DAPI; 1:15000; Sigma) for 1 h and 30 min at RT. The slides were then mounted with DABCO, and viewed on an epifluorescence microscope Zeiss Axiovert 40 CFL, coupled with a digital camera Carl Zeiss Axiocam MRm, with AxioVision Rel.4.8 software.

Cell death was analyzed in slides from eyes enucleated 24 hpi, using the in situ Cell Death Detection Kit, Fluorescein (Roche Diagnostics, GmbH, Mannheim, Germany), according to manufacturer's instructions. As a positive control, some slides were incubated with 200 U/ml DNase I (Sigma) for 10 min at RT.

Total RNA was isolated from enucleated eyes using Trizol (Invitrogen, CA, USA), from which complementary DNA was synthesized using Oligo dT and Superscript-II reverse transcriptase (Invitrogen). Specific cDNAs were amplified by PCR over 30–40 cycles, using Taq polymerase (Fermentas) and gene-specific primers (Table 1), under the following conditions: denaturation at 95°C for 30 s, annealing at 50–60°C (this temperature was changed according to the primers' Tm) for 15 s, and extension at 72°C for 30 s. The PCR products were resolved on 1.5% agarose gels containing 25 ng/ml ethidium bromide, and visualized in an UV EpiChem³ Darkroom transilluminator. Images were captured using the LabWorks 4.5 software (BioImaging Systems, UVP, Upland, CA, USA). Embryonic (E10) cDNA was used as a positive control, while non-template samples were negative controls.

Quantitative PCR reactions were performed using KAPA SYBR FAST (Kappa BioSystems, Wilmington, MA, USA) master mix on the PikoReal Real-time PCR System (Thermo Scientific), for 35 cycles as following: denaturation at 94°C for 10 s, annealing at 50–60°C for 30 s, and extension at 72°C for 30 s. Reactions for each primer were performed in triplicate. The ΔΔCT method was used to determine mRNA levels in control and injured retinas. All data were normalized to GAPDH mRNA expression levels.

After eye enucleation and retinal extraction, genomic DNA was extracted with the EpiTect Fast DNA Bisulfite Kit (Qiagen, Hilden, Germany), and then treated for bisulphite conversion and purified according to manufacturer's instructions. MSPCR was performed using the EpiTect MSP kit (Qiagen, Hilden, Germany) master mix and previously reported specific primers for the methylated and unmethylated forms of Oct4 (Wang et al., 2013; Table 2). DNA was amplified for 40 cycles, which comprised a denaturation step at 94°C for 15 s, annealing at 50°C for 30 s, and extension at 72°C for 30 s. The resulting products were resolved on 1.5% agarose gels containing 25 ng/ml ethidium bromide, and visualized in an UV EpiChem³ Darkroom transilluminator. Images were captured using the LabWorks 4.5 software (BioImaging Systems, UVP, Upland, CA, USA). Samples treated with CpG methyltransferase M.SssI (New England BioLabs, Ipswich, MA, USA) were used as positive methylation controls.

DNA was amplified for HRM analysis on the PikoReal Real-time PCR System (Thermo Scientific), using a Luminaris Color HRM (Thermo Scientific) master mix, in a volume of 20 μl, containing 20 ng of DNA. Thermal cycling conditions were as follows: denaturation at 95°C for 10 s, annealing at 50°C for 30 s, and extension at 72°C for 30 s, for 40 cycles. Samples were then melted from 40 to 99°C, with a melting rate of 0.2°C/s. All HRM data were collected and analyzed by the dedicated PikoReal software (Thermo Scientific, version 2.2). Normalization of HRM curves was performed by the same software, as well as calculation of closest standard call (value attributed to each sample with respect to a methylation standard). All analysis were performed in triplicate. Genomic DNA from intact retinas, treated with CpG methyltransferase M.SssI (New England BioLabs, Ipswich, MA, USA) was used as methylation standard.

As previously reported (Eberle et al., 2014), isolated retinas were dissociated in DMEM containing 0.5% trypsin (Sigma) for 20 min at 37°C. The enzymatic reaction was inhibited by transferring the tissue to DMEM containing 10% fetal calf serum. Additional mechanical dissociation was performed by triturating cells with a 1 ml plastic tip and then a fire-polished glass Pasteur pipette. Cells were resuspended in MACS buffer and incubated with a rabbit anti-mouse primary antibody against GLAST (1:200; Novus Biologicals, Littleton, CO, USA, NB100-1869) for 5 min at 4°C, washed in MACS buffer, and then centrifuged. The cell pellet was resuspended in MACS buffer and incubated with 0.1 ml of goat anti-rabbit IgG magnetic beads (Miltenyi Biotec, Bergisch Gladbach, Germany) for 15 min at 4°C. Magnetic separation was performed according to manufacturer's instructions on a QuadroMACS separator, using LS columns (Miltenyi Biotec, Bergisch Gladbach, Germany). Briefly, the cells were placed on a column fixed to the separator and flow through (GLAST-negative cells) was collected. Then, GLAST-positive fraction was eluted by loading 5 ml MACS buffer and immediately applying pressure with the supplied plunger. Cells from both fractions were then prepared for further analysis.

DNA methyltransferase inhibitor SGI-1027 (Sigma) was dissolved in 0.05% DMSO, and injected intravitreally following the same procedure as NMDA injection (10 μM in 2 μl), 24 h before retinal injury.

All experiments were performed by triplicate and all data are expressed as mean ± SEM. Statistical significance of the differences was assessed by a one-way analysis of variance (ANOVA), followed by a Tukey test for post-hoc comparisons, or by Student's t-test.

We demonstrated the excitotoxic effect of the NMDA injection by a TUNEL assay, comparing injured retinas (at 24 hpi) with positive controls. NMDA produced massive cellular death throughout all retinal layers (Figure 1A), as well as a general disruption of the laminar structure of the retina. MG response to damage was assessed by immunofluorescence against GS and nestin, a filamentous protein usually found in progenitor cells. Intact retinas (Figure 1B) lack nestin-immunopositive cells. After 24 hpi, co-labeling of nestin and GS is observable across the inner nuclear layer (INL), reflecting MG early dedifferentiation (Figure 1C).

To characterize the early molecular events during MG response to NMDA-induced injury in rodent retina in vivo, we analyzed the changes in expression of pluripotency-associated genes of retinas extracted at 4, 12, 18, and 24 hpi (Figure 2A) from a group of mice (n = 12, 3 per group). We found that Oct4 and Nanog, both regarded as essential for pluripotency acquisition and maintenance, rapidly increase their expression after 4 and 12 hpi. However, this change is transient, and both genes become undetectable at 24 hpi. Klf4 and Pax6 also appear to have a transient expression, since both are substantially reduced after 18 hpi. In contrast, Lin28 and Sox2 maintain their expression throughout all time points. To validate these data, we performed qPCR analysis and determined the relative expression level of each gene. A one way ANOVA revealed that Oct4 and Nanog reach an expression peak at 4 hpi (p < 0.001, when compared to controls) and then reduce their levels at 18 and 24 hpi (Figure 2B). In a similar fashion, Klf4 exhibits a significant increase at 4, 12, and 18 hpi (p < 0.01) and then a reduction at 24 hpi. Pax6 actually undergoes a slight decrease at 4 hpi (p < 0.05), before increasing its expression levels at 12 and 18 hpi (Figure 2C). In order to achieve a better visualization of the changes in mRNA levels for each gene, we compiled the data and performed a logarithmic transformation (Figure 2D). The resulting graph shows the expression dynamics of the evaluated pluripotency-associated genes.

The transient expression of pluripotency-associated genes suggests the presence of a silencing mechanism acting on MG at 24 hpi. Since DNA methylation is considered the main device of epigenetic memory (Vierbuchen and Wernig, 2012), we chose to analyse the expression of genes encoding DNA methyltransferases and Gadd proteins, by RT-PCR (Figures 3A,B). As expected, we did not observe any change in the expression levels of the maintenance methyltransferase Dnmt1, but the expression of Dnmt3a and 3b, which are involved in de novo methylation of DNA, is diminished at 4 and 12 hpi. Gadd proteins, associated with DNA demethylation, also exhibited changes in their expression profiles. Particularly, we found a significant increase of Gadd45b at 4 and 12 hpi. These mRNA is then drastically reduced at 18 and 24 hpi. Quantification of gene expression changes by qPCR demonstrated that Dnmt3b exhibits a biphasic, time-dependent decrease, which becomes statistically significant (by an ANOVA test) at 12 hpi (p < 0.01), before reverting to its basal levels. Gadd45b also shows a significant increase at 4 hpi (p < 0.001), and then returns to normal levels (Figure 3B). We also analyzed the expression of the genes that encode active DNA demethylation-associated Tet proteins (Figure 3C), and observed that, unexpectedly, the expression of Tet1 is decreased at 12 hpi (p < 0.05); Tet2 is decreased at 4 and 12 hpi (p < 0.01 and 0.05, respectively), and Tet3 at 4, 12, and 24 hpi (p < 0.001, 0.001, and p < 0.01, respectively). We then performed a logarithmic transformation of data in order to achieve a better visualization, and compared the changes in the expression of Dnmt3b and Gadd45b with those of Oct4 and Nanog (Figure 3D). Noteworthy, the expression peak of pluripotency genes correlates with a major decrease in Dnmt3b levels (at 4 and 12 hpi), and with a significant increase in Gadd45b.

Given the apparent correlation between Oct4, Dnmt3b, and Gadd45b, we reasoned that Oct4 DNA methylation profile might exhibit changes at different times post injury. To demonstrate this, we extracted genomic DNA from another group of mice (n = 12, 3 per group) and performed a methylation-specific PCR analysis, using previously tested primers, which recognize the methylated form of Oct4 (Wang et al., 2013). We observed a decrease in methylated Oct4 at 4 and 12 hpi, with an apparent return to basal levels at 24 hpi (Figure 4A). Assuming the limitations of MS-PCR, we also performed an HRM analysis, using several dilutions of DNA treated with CpG methyltransferase M.SssI as a methylation standard. By using an ANOVA test, we observed that NMDA injury induced a significant decrease in Oct4 methylation after 4 and 12 hpi (down to 20% when compared to the fully methylated standard; p < 0.001; Figure 4B). Afterwards, DNA methylation tends to increase, up to 60% when compared with the standard (p < 0.001) at 18 hpi, and around 88% at 24 hpi (p < 0.01). These results indicate a tendency of Oct4 to return to its fully methylated state as damage proceeds.

To evaluate whether the expression changes we observed are happening exclusively in MG, and not in other cell types in the retina, we separated Müller cells by magnetic associated cell sorting (MACS) after extracting retinas from 12 mice (3 per condition). This procedure depends on a reliable surface marker, and we chose the glutamate transporter GLAST, mainly expressed in radial glia in the central nervous system and specific to MG in the retina (Namekata et al., 2009). First, we confirmed that Slc1a3 (which encodes GLAST) maintains its expression levels in vivo after retinal injury (Figure 5A), making GLAST a suitable marker for MACS in our injury model. As a validation to our cell sorting results, we found that the GLAST-positive fraction shares the expression of both Slc1a3 (GLAST) and Glul (GS) with cultured MG. In contrast, the GLAST-negative fraction expresses the photoreceptor marker Nrl, also found in a whole, intact retina sample (Figure 5B). We extracted total RNA from both GLAST-positive and negative fractions of injured retina for qPCR analysis, and found that changes in Oct4 expression are restricted to the positive fraction, i.e., MG, closely resembling the pattern previously found in whole retina (Figure 5C). We obtained similar results after quantifying Nanog (Figure 5D), Lin28 (Figure 5E), and Dnmt3b (Figure 5F). These genes are expressed exclusively in MG after retinal injury, and show no significant changes in GLAST-negative cells, after statistical analysis by a Student's t-test.

To demonstrate a causal relationship between DNA methylation and Oct4 silencing, we intravitreally administered SGI-1027, a DNA-methyltransferase inhibitor which has been shown to block and degrade DNMT1, DNMT3a, and DNMT3b (Yoo et al., 2013; Gros et al., 2015), to a group of mice (n = 10, 5 per condition). We evaluated the expression of Oct4 in GLAST-positive and negative fractions of retinas injured in the presence and absence of SGI-1027. The retinas were extracted 24 h after NMDA injection, since the aforementioned pluripotency-associated marker was silenced at this time in previous experiments. Our results show that SGI-1027 allows the sustained expression of Oct4 after retinal injury, only in the GLAST-positive fraction of retinas (Figure 6A). This increase was revealed to be statistically significant (Student's t-test) by qPCR analysis (p < 0.001, when compared to control; p < 0.001 when compared to damaged retinas at 24 hpi without SGI-1027 treatment; Figure 6B). These results suggest that DNA methylation could be involved in Oct4 silencing at 24 hpi in vivo, and restrict this response to MG.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.