New Results

Vision rescue via chemically engineered endogenous retinal ganglion cells

Rana Muhammad Shoaib, Aregnazan Sandrosyan, Biraj Mahato
doi: https://doi.org/10.1101/2023.12.27.572921

Abstract

Loss of retinal ganglion cells (RGCs) is the final common end point for many optic neuropathies, ultimately leading to irreversible vision loss. Endogenous RGC regeneration from Müller cells presents a promising approach to treat these diseases, but mammalian retinas lack regenerative capacity. Here, we report a small molecule cocktail that causes endogenous Müller cell proliferation, migration, and specification to newly generated chemically induced RGCs (CiRGCs) in NMDA injured mice retina. Notably, regenerated CiRGCs extend axons towards optic nerve, and rescue vision post-NMDA treatment. Moreover, we successfully reprogrammed human primary Müller glia and fibroblasts into CiRGCs using this chemical-only approach, as evidenced by RGC-specific gene expression and chromatin signature. Additionally, we show that interaction between SOX4 and NF-kB determine CiRGC fate from Müller cells. We anticipate endogenous CiRGCs would have therapeutic potential in rescuing vision for optic nerve diseases.

Background

The regenerative capacity of the adult mammalian central nervous system, including the neurosensory retina, is limited 1–3. However, recent research has shown that Müller glia, a type of support cell in the retina, can be reprogrammed into retinal neurons in vivo 4–8. In lower vertebrates, Müller glia naturally transforms into neural progenitor cells and generate functional neurons to restore vision after injury. Researchers are currently addressing the limitation of endogenous mammalian retinal neurogenesis in two main ways: i) transplanting in vitro engineered stem cell derived retinal neurons or neural progenitor cells into injured retina9–15 or ii) overexpression of transcription factors in Müller cells to regenerate neural progenitors or retinal neurons to restore retinal function16–22. However, these methods involve the use of viruses which present challenges for its clinical implementation. An alternative method, using small molecules for chemical reprogramming, has recently garnered considerable attention 23–26. In a prior study, we demonstrated that a combination of five small molecules (referred to as 5C) could reprogram fibroblasts into photoreceptors, subsequently leading to vision restoration upon subretinal transplantation in blind mice27. By augmenting 5C with supplementary molecules, here we report a chemical combination that stimulates proliferation, migration, and trans-differentiation of resident retinal Müller cells into CiRGCs after NMDA-induced RGC damage. To validate the Müller origin of CiRGCs, we conducted lineage tracing experiments using two mouse models. Moreover, regenerated CiRGCs pass axons through the optic nerve head and rescued vision in the NMDA injured mice. Lastly, we furnish supplementary evidence by demonstrating that human primary Müller glia and fibroblasts can also be reprogrammed into CiRGCs in vitro using a similar chemically induced approach.

Results

Chemical cocktail induces endogenous Müller glia to become CiRGCs in NMDA-injured retina

Based on our previous work27, we tested if 5C [Valproic acid (V), CHIR (C), Repsox (R), Forskolin (F), and IWR1 (I)] along with additional chemicals or factors can induce endogenous Müller cell proliferation and subsequent retinal neurogenesis in NMDA mediated RGC degeneration. We found 5C plus Noggin and IGF1 (2F) along with an additional molecule S24795 (hence 6C+2F) can induce Müller cell proliferation in the inner nuclear layer (INL) and migration to ganglion cell layer (GCL) after NMDA injury (with 75%-85% RGC loss) (Fig. 1, Extended data Fig. 1a-d, Extended data Table 1, 2). Müller cell proliferation and migration was examined on day 1,2, 3 and 6 based on prior research (Fig. 1a).28, 29 On day 1, BrdU+ cells were evident along the outer edge of INL in NMDA injured retina similar to published studies (Fig. 1b) 6, 29. These BrdU+ cells were also colocalized with Müller marker glutamine synthetase (GS) (Fig. 1c, d)30, 31. By contrast no BrdU+ cells were observed at the same layer in vehicle injected eyes (Extended data Fig. 2a-d). On day 3, BrdU+ (approx. 39.13±2.7%) cells were located along the GCL that indicates cell migration (Fig.1e, f). When stained with SNCG, a RGC specific marker, we found 30±1.7% of cells in GCL layer are SNCG+ BrdU+ (Fig. 1g, h). Taken together, these data suggest that 6C+2F can induce endogenous Müller glia proliferation, migration, and differentiation into CiRGCs.

Extended data Fig. 1.
Extended data Fig. 1. Generation of NMDA induced RGC degeneration mice model.

(a) Retinal cross sections showing RBPMS+ cell death 7 days after NMDA injection in intravitreal route. (b) Approx. 72% decrease in RGC number was evident (n=3 eyes). (c) Retinal flat mount stained with RBPMS also showed decrease in RGC number. (d) Approx 80% decrease in RGC number was evident. RGCs were counted in periphery and medium regions (n=3 eyes).

Extended data Fig. 2.
Extended data Fig. 2. Generation of BrdU+ cells is 6C+2F treated retina but not in vehicle treated retina.

(a) Flat mount retina on day1 and day3 after BrdU antibody staining. (b) Magnified micrographs of flat mount retina showing BrdU staining in the DAPI+ nucleus. (c) Retinal sections stained with BrdU in PBS injected eyes on day 1 and day 3, did not show any BrdU+ cells. (d) No BrdU+ cells were evident in PBS injected flat mount retina.

Fig. 1.
Fig. 1. Small molecule cocktail (6C+2F) induces endogenous Müller cell proliferation, migration, and trans-differentiation to SNCG+ RGCs post NMDA injury.

(a) Scheme showing time point for small molecule cocktail intravitreal injections and subsequent tissue collection. (b) BrdU+ cells in the lower part of INL layer 1 day after cocktail injection. (c, d) BrdU+ cells are positive for Müller cell marker Glutamine Synthetase (GS). (e) BrdU+ cells are migrated to GCL layer on day3 after injection. (f) Number of BrdU+ cells in the GCL layer on day3. (g) On day3, BrdU+ cells in GCL layer are positive for RGC marker SNCG. (h) Number of SNCG+BrdU+ cells in the GCL layer.

Fig. 2.
Fig. 2. Lineage tracing using a AAV2-GLAST-GFP reporter confirms Müller origin of RGCs in NMDA injured retina.

(a) Scheme showing time points for AAV2 and small molecule cocktail injection and subsequent tissue collections. (b) Upper panel: vehicle control. Middle panel: BrdU+ cells are generated in the lower part of INL layer on day1 after cocktail injection and these BrdU+ cells are GFP+ (upper panel, inset white box). Lower panel: On day3 BrdU+ GFP+ cells (yellow) appeared in the GCL layer indicating their Müller origin. (c) GFP+ cells in GCL layer are also positive for pan RGC marker RBPMS. (d) GFP+ BrdU+ cell count in the INL layer. (e, f) GFP+ BrdU+ and GFP+RBPMS+ cell counts in the GCL layer.

Endogenous CiRGCs are of Müller origin and extend axons to the optic nerve

Next, we lineage traced CiRGCs by intravitreally injecting AAV2-GLAST-GFP into WT (C57/BL6, control) and NMDA injured mice (approx.106 vg/eye) before cocktail injection (Fig. 2a). 7 days after AAV2 injection, we observed specific GFP expression in Müller cells of WT retina, while almost no GFP was detectable in GCL (Extended data Fig. 3a-c). After administering 6C+2F into AAV2 injected mice eye, we found GFP+BrdU+ cells on the inner side of INL layer on day1 and on the GCL layer on day3 (approx. 35%±6.1 GFP+BrdU+ cells) (Fig. 2b, d, e). The expression of GFP in BrdU+ cells at INL indicated their origin from Müller cells (Fig. 2b, middle, white boxes). Furthermore, we observed GFP+RBPMS+ cells on day-3-retinal sections which further supports Müller origin of CiRGCs (Fig. 2c, f). Additionally, we used a GLAST-CreERT2;ROSA26-flox-stop-flox (Ai9) mice that specifically express TdTomato in Müller cells and allowed us to track their fate following NMDA-damage (Fig. 3a). We found RBPMS+ TdTomato+ RGCs in GCL in cocktail treated but not in PBS treated eyes on day3 (Fig. 3b, c, Extended data Fig. 4a, b). These findings confirmed the Müller cell origin of CiRGCs.

Extended data Fig. 3.
Extended data Fig. 3. AAV2-GLAST-GFP promoter reporter expression in Müller cells.

(a) GFP expression from the reporter concentrated in the INL layer of the retina. (b) No GFP expression was evident in RBPMS+ cells in GCL. (c) GFP reporter expressed in the nucleus of GS+ Müller cells.

Extended data Fig. 4.
Extended data Fig. 4. Regenerated TdTomato+ RGCs pass their axons through the optic nerve head and proximal part of the optic nerve.

(a) No RBPMS+TdTomato+ cells were present in PBS injected retina on day3. (b) percentage of TdTomato+RBPMS+ cells in GCL on day3 after cocktail administration. (c) Mean fluorescence for TUBB3 staining (n=3 eyes every fifth section was counted) (d) Number of TdTomato+ axons per section in the optic nerve. Every fifth section was counted from serial sections and 4 sections were included per eye (n=3). (e) TdTomato+ axons pass through the optic nerve head in chemical treated eyes on day 6. (f) Z-section from the indicated area from d., showing axon passing through the optic nerve head (arrowheads). Black and white images are grayscaled and inverted.

Fig. 3.
Fig. 3. CiRGC Lineage tracing and axon projection to the optic nerve in GLAST-Cre;Rosa26-TdTomato reporter mice.

(a) Outline for the lineage tracing experiment. (b) Upper panel: Representative image showing RBPMS+ TdTomato+ cells (red and green rectangle) in the GCL layer on day3 indicating their Müller origin. (c) Magnified picture from red and green rectangular box. (d) No TdTomato+ axons were evident in optic nerve of control eyes after NMDA and PBS treatment (n=3 eyes tested). TdTomato+ cells are positive for muller specific marker GS. (e) Upper panel: WT control. Middle panel: PBS injected eyes showed no TdTomato+ axon passing through the optic nerves on day6. Lower panel: In cocktail treated animals TdTomato+ axons are passed through the optic nerve head on day 6. (f) Regenerated TdTomato+TUBB3+ axons passed through proximal part of optic nerve. (g) Magnified picture from (f) showing passed axons in the proximal part of optic nerve in cocktail treated eyes. (h) Magnified image for TdTomato+TUBB3+ axons passing through the optic nerve head into the proximal part of optic nerve.

Fig. 4.
Fig. 4. Recovery of pERG & pVEP after cocktail treatment in NMDA injured mice eye.

(a) Timeline for functional measurements after cocktail treatment in NMDA injured eyes. (b) Recovery of pERG upon cocktail treatment. Upper panel: cocktail treated. Lower panel: PBS treated. Representative pERG waveforms are from day 23. (c) Pattern VEP measurements at the indicated time points. Upper panel: cocktail treated. Lower panel: PBS treated. Representative pVEP waveforms from day 23 are presented. All measurements were conducted longitudinally at indicated time points.

Next, we investigated if Tdtomato+ CiRGCs projected axons to the optic nerve, an essential step for vision rescue. Absence of TdTomato+ axons in GLAST-Cre;TdTomato optic nerve upon NMDA injury (without cocktail) indicated suitability of this model for the study (Fig. 3d). Subsequently we found TdTomato+ axons are passed through the optic nerve head in cocktail treated but not PBS treated eyes on day 6 (Fig. 3e). We observed severely damaged RGC axons and reduced intensity of TUBB3 (15.5% of WT) staining in the optic nerve following NMDA treatment which was recovered after cocktail treatment (29.6% of WT) (Fig. 3e, left panels, Extended data Fig. 4c). Further analysis of cocktail treated eyes indicated expression of TUBB3 (axonal marker) in TdTomato+ axons that are passed to the proximal part of the optic nerve via optic nerve head (Fig. 3e-h, Extended data Fig. 4d-f)32. These results suggested that endogenously regenerated CiRGCs extend their axons to the proximal part of the optic nerve through optic nerve head.

Endogenous CiRGCs rescue vision after NMDA mediated RGC damage

To address the paramount question of whether CiRGC regeneration leads to vision rescue, we measured electrical activity from retina and visual cortex of the brain using pattern ERG (pERG) and pattern VEP (pVEP) following the scheme in Fig. 4a. We observed pERG and pVEP rescue starting from day 16, after that day 23, day 30 & day 60 in cocktail treated eyes compared to vehicle (Fig. 4b, c). pERG and pVEP amplitudes were rescued to 63% and 89% of the baseline respectively on day 60. We noted differential improvement between right and left eyes both in pERG (right 59% vs left 35.2%) and pVEP (right 81% vs left 58.7%) until day 23. This difference becomes narrower (pERG; right, 56%, left 51%, pVEP; right, 96.8%, left,73.5%) on day 30. Next, we conducted a longitudinal light avoidance test in the pERG & pVEP rescued animals following published methods to determine whether improved retinal function is sufficient to rescue visually evoked behaviour33–35. Normal mice exhibit a natural aversion to brightly illuminated areas, forming the basis for this test. All animals were placed in the apparatus for a duration of 300 seconds, and the proportion of time spent in the brightly lit (2800 lux, similar to cloudy outdoor environment) area was recorded. Wild type mice demonstrated a distinct preference for the dark compartment, while NMDA-treated mice did not display such a preference (Mean: 195 ± 32 seconds versus 120 ± 29 seconds). Intriguingly, mice treated with cocktail exhibited a clear and statistically significant preference for the dark compartment (215 ± 37 seconds) (Fig. 5a, b). Additionally, we measured latency time to first entry to the dark chamber and found significantly higher latency time in PBS treated group whereas wildtype and cocktail treated groups showed similar latency time (Fig. 5c). Subsequently, we conducted longitudinal visual cliff test, a method employed to assess depth perception and the fear of crossing the deep side of a platform36–38. Typically, normal mice exhibit an inherent tendency to avoid the deep side, favoring instead the shallow side of a visual cliff. Mice were positioned on the central platform between the deep and shallow sides, and their choices were meticulously documented. Our observations revealed that wild-type mice displayed a notable preference, with 96±3.9% choices to shallow side. In contrast, the NMDA+PBS group exhibited a significantly diminished performance, with only 32±8.9% choosing the shallow side. Conversely, the cocktail-treated group demonstrated a more favorable outcome, with 72±10.9% of choices directed towards the shallow side (Fig. 5d, e). In assessing the cumulative duration of stay on both the shallow and deep sides, it became evident that wild-type mice exhibited a distinct preference towards the shallow side. Conversely, the NMDA+PBS group showed no clear preference, indicating a compromised performance. Intriguingly, the NMDA plus cocktail-treated mice displayed a preference for the shallow side (Fig. 5f). These findings strongly indicate a recovery of visual functions in NMDA injured cocktail-treated mice.

Fig. 5.
Fig. 5. Rescue of light aversion & depth perception behavior after cocktail treatment in NMDA injured mice eye.

(a) Schematic of the experimental set up for light aversion behavior test. (b) Plot of mean time (± SEM) spent in the dark chamber by the indicated experimental groups. (c) Plot of mean latency time (± SEM) required to first entry to the dark chamber by the indicated experimental groups. (d) Schematic of the experimental set up for death perception test (visual cliff test). (d) Plot for percent shallow side chooser. Five trials were performed for each animal. (f). Plot for cumulative duration spent in deep or shallow sides by the indicated experimental groups after removal of the cliff. Student’s t-test was used to determine significance in b,c,e. 2-Way ANOVA was used to determine significance in f.

In vitro CiRGCs exhibits RGC like gene expression and chromatin signature

To reinforce our in vivo findings suggesting Müller glia is the source of endogenous RGC generation, we conducted experiment where human primary Müller and fibroblast cells were treated with similar chemical cocktail (VCRFIs+2F, without S24795) in a defined RGC differentiation medium (Extended data Table 3, 4). S24795 was omitted because in vitro Müller cells are already activated as indicated by their GFAP expression and may behave like progenitors39. Subsequently, we performed the RGC reprogramming and examined immunofluorescence with RGC markers on day 4. We observed that the morphology of CiRGCs resembled that of neurons, and these RGC-like cells exhibited the expression of RBPMS, SNCG and BRN3A (Extended data Fig. 5a-d). Müller derived RGC (Md-CiRGC) reprogramming was quick with BRN3A cells starting to appear as early as 6hrs after chemical induction (Extended data Fig. 5e). To achieve a deeper understanding of RGC gene expression we conducted SC-RNA Seq of CiRGCs collected on day 4 (Fig. 6a). A total of 9044 Md-CiRGCs were included in the analysis and resulting data were visualized by UMAP and heatmap. We identified cell populations that are divided into 12 distinct clusters (Fig. 6b). These CiRGCs express all major RGC markers exhibited in heatmap and UMAP, including BRN3A, RBPMS, ISL2, ISL1, SNCG, NEFL, NEFM, ELAVL3, ELAVL4, as well as the RGC specifying transcription factor SOX4 (Fig. 6c, Extended data Fig. 6a, b). Additionally, gene expression signature indicated the presence of mature RGC genes such as THY1, GAP43, TUBB3 (Extended data Fig. 6c). Reduced expression of Müller specific genes SLC1A3 (GLAST) and GFAP in the reprogrammed CiRGCs was noted (Extended data Fig. 6d). We analyzed cell sets for POU4F1+ and POU4F1+RBPMS+ (two best RGC markers across species) and examined expressions of major RGC genes. We found expressions of all major RGC specifying genes SNCG, RBPMS, ISL2, ISL1, SNCG, NEFL, NEFM, ELAVL3, SOX4, THY1, TUBB3, GAP43 (Fig. 6c. middle, lower). Next, we examined if human fibroblasts can be reprogrammed to CiRGCs using same chemical method by SC-RNA seq. Fibroblast derived CiRGCs (Fd-CiRGCs) on day4 was used for single cell RNA sequencing similar to Md-CiRGCs. Notably, we found Fd-CiRGCs expressing RGC marker BRN3A along with other major RGC markers such as BRN3B, RBPMS, SNCG, ISL2, NEFM, NEFL, GAP43, THY1, ISL1, ELAVL3, SOX4 etc. which is similar to Md-CiRGC gene expression signature (Extended data Figs. 7a-c and 8). Neither Md nor Fd CiRGCs do not express other retinal cell markers (Extended data Fig. 7d).

Extended data Fig. 5.
Extended data Fig. 5. In vitro generation of CiRGCs from human primary Müller (Md-CiRGCs) and fibroblasts (Fd-CiRGCs).

(a) RBPMS (green) and BRN3A (red) staining in Md-CiRGCs on day4. (b) qPCR showing increased expression of major RGC specifying genes in Md-CiRGCs on day4. (c) Left: Fd-CiRGCs express BRN3A in the nucleus (see white box). Right: SNCG-GFP promoter reporter expression in Fd-CiRGCs on day 4 (anti-GFP staining). (d) qPCR showing expression of major RGC genes in Fd-RGCs (day4). (e). Time course expression measurement of Brn3a mRNA by qPCR in Md-CiRGCs.

Extended data Fig. 6.
Extended data Fig. 6. Visualizing expression of major RGC genes in Md-CiRGCs by UMAP analysis.

(a-c) Md-CiRGCs are mostly Brn3a expressing RGCs and express other RGC marker genes [Brn3b (few), Sncg, Isl2, Isl1, NefL, NefM, Elavl3], specification genes (Sox4) and maturation genes such as (Thy1, Gap43, Tubb3). (d) Müller specific genes Glast, Gfap not expressed in most cells Md-CiRGC population indicating complete cell conversion. GFAP was expressed in 100% cells in the starting human Müller cells (immunofluorescence stydy, data not shown).

Extended data Fig. 7.
Extended data Fig. 7. Reprogramming human adult fibroblasts to Fd-CiRGCs by chemical cocktail and gene expression analysis by SC-RNASeq.

(a-c) Heatmap analysis showing expression of all major RGC specifying genes (Brn3a, Rbpms, Sncg, Isl2, Isl1, Sox4, Gap43, BIII-Tubulin, Thy1 etc.) in total CiRGC population, Pou4f1+ and Pou4f1+Rbpms+ cell sets. (d) No or very few cells express photoreceptor specifying genes in Pou4f1+ and Pou4f1+ Rbpms+ CiRGCs from both Müller cells and fibroblasts.

Extended data Fig. 8.
Extended data Fig. 8. visualizing major RGC genes in Fd-CiRGCs by UMAP analysis

(a) UMAP analysis of Fd-CiRGCs revealed 9 major clusters on day4 of reprogramming. (b, c) Fd-CiRGCs are mostly POU4F1 and POU4F2 expressing. FdCiRGCs express RGC marker genes (Rbpms, Sncg, Isl1, Isl2, NefL, NefM), specifying genes (Sox4) and maturation genes (Tubb3, Gap43, Thy1). (d) Fibroblasts specific genes NefH, Col1a2 not expressed in Fd-CiRGCs, indicating complete cell conversion.

Fig. 6.
Fig. 6. Gene expression and chromatin analysis of Md-CiRGCs by SC-RNASeq and SC-ATAC Seq.

(a) Scheme showing CiRGC generation from human primary Müller cells (HMC) and human lung fibroblasts (HFL). (b) UMAP analysis revealed 13 major clusters in Md-CiRGCs on day4 of reprogramming. (c) Heatmap analysis showing expression of all major RGC specifying genes (Brn3a, Rbpms, Sncg, Isl2, Isl1, Sox4, Gap43, BIII-Tubulin, Thy1 etc) in total cell population, POU4F1+ and POU4F1+RBPMS+ cell sets. (d) t-SNE analysis of SC-ATAC-Seq data using promoter sum function indicated presence of 8 total clusters. (e) Cells with open POU4F1 promoter peaks used for subsequent promoter analysis. (f-h) Open promoter peaks in the indicated genes.

Next, we examined the global chromatin signature of CiRGCs using SC-ATAC seq and t-SNE analysis indicated the existence of 8 major clusters based on a feature called promoter sum (Fig. 6d). Next promoter regions of all individual RGC genes were examined for presence of open chromatins using the filter function of the algorithm. BRN3A containing cells are distributed in all the clusters except one potentially immature cluster (#8) similar to our SC-RNA Seq (Fig. 6d, e). These BRN3A positive cells contain peaks or open chromatins in all the RGC promoters such as POU4F1, SNCG, NEFL, ISL1, RBPMS, ISL2, NEFM, SOX4 (Fig. 6f-h, Extended data Fig. 9). Interestingly, no peaks were available in majority of cells at promoters of, SOX11, POU4F2 and GFAP which showed expression in few cells SC-RNA Seq (Extended data Fig. 9d, e, h). These results suggest that 6C+2F cocktail can reprogram human Müller cells and fibroblasts into CiRGCs.

Extended data Fig. 9.
Extended data Fig. 9. Open promoter peaks present in the indicated RGC specifying genes in POU4F1 promoter sum+ cells.

(a-g) Chromatin opening in promoters of the indicated genes. (h) GFAP have no open promoter peak.

SOX4 and NF-kB (RelA) collaboratively determine RGC fate from Müller cells

We investigated mechanism behind CiRGC reprogramming from Müller cells using SC-RNASeq data. In this process, NF-kB emerged as a key player in the reprogramming of CiRGCs, based upon its established role in chemically mediated photoreceptor reprogramming from our prior study27. Photoreceptor reprogramming involved the use of VCRF which triggered the activation of NF-kB. Given the presence of VCRF in the CiRGC reprogramming cocktail, we hypothesized that NF-kB, either independently or in conjunction with other transcription factors, could directly regulate RGC fate determining genes such as BRN3A, RBPMS, ISL1, SNCG and GAP43. Previous report and our bioinformatics analysis (rVista) identified NF-kB binding sites in the regulatory regions of these genes (Extended data Fig. 10)40. Heatmap analysis of SC-RNA Seq indicated existence of an intermediate unstable progenitor-like stage representing clusters 1, 11 and 12 (Extended data Fig. 11a, green box). Approx. 30%-40% cells in these clusters express RGC specifying transcription factor SOX4, a potential target of WNT signaling activator CHIR9902 used in our cocktail41. In contrast, few cells in clusters 1, 11, 12 express NF-kB, compared to 30%-50% in the remaining clusters. These results suggested that SOX4+ progenitors may have originated in clusters 1,11, 12 which subsequently express NF-kB and moved to other clusters. Further analysis revealed the existence of a few SOX4+NF-kB+ cells in clusters 1,11 and 12 and this number was increased in other clusters (Extended data Fig. 11b, green box). Trajectory analysis further confirmed that SOX4+NF-kB+ cells originated in clusters 1, 11 and 12 subsequently spread to the rest of the clusters in later time points (Extended data Fig. 11c, left). Investigating the potential induction of BRN3A expression, we examined the expression of BRN3A along with SOX4+NF-kB+ cells. Interestingly, a few BRN3A-expressing cells were present in clusters 1,11,12, suggesting a gradual generation of BRN3A+ RGCs (Extended data Fig. 11c, middle). A similar distribution was evident for SOX4+NF-kB+POU4F1+ cells (Extended data Fig. 11c, right). Trajectory analysis for all clusters along pseudotime indicated an initial origin of clusters 1 followed by clusters 5, 6, and then the remaining clusters, reinforcing the notion of a time-dependent generation of CiRGCs (Extended data Fig. 11d). Subsequently, we employed violin plots to illustrate gene expression patterns in BRN3A+ and BRN3A+RELA+SOX4+ cell sets (Extended data Fig. 12). Analysis of the BRN3A+ cell population indicated that clusters 1 and 11 exhibited a few BRN3A+ cells, likely due to the presence of a limited number of NF-kB+ or SOX4+ cells (Extended data Fig. 12a). In contrast, a higher number of BRN3A+ cells were evident in the remaining clusters, where both NF-kB and SOX4 were co-expressed. Additionally, BRN3A+NF-kB+SOX4+ population was revealed, clusters 1, 11, 12 do not possess these cells (Extended data Fig. 12b). Collectively, these results further underscore the hypothesis that the cooperation between NF-kB and SOX4 contributes to BRN3A expression in CiRGCs similar to other studies where collaboration between SOX4 and RelA was noted42 (Extended data Fig. 12c). We also found RAX+ progenitor like and ATOH7+ cell population that may generate CiRGCs through a minor POU4F2 mediated pathway (Extended data Fig. 11e).

Extended data Fig. 10.
Extended data Fig. 10. Bioinformatic analysis (r-vista) of NF-kB (RELA) binding site identification.

Putative RelA binding sites are present in POU4F1, ISL1 regulatory regions.

Extended data Fig.11.
Extended data Fig.11. SOX4 and NF-kB (RELA) mediates Md-CiRGC fate commitment through a non-canonical pathway.

(a) Md-CiRGC SC-RNASeq showed presence of an unstable progenitor population that express Sox4 but not RelA (upper panel, green and red boxes). This cell population was absent from Pou4f1+ cell set (lower panel). (b) Few RelA+Sox4+ cells generated in cluster1, 11, 12 (UMAP, green boxes). However, numbers increased in the rest of the clusters. Left UMAP: all cell populations; middle UMAP: RelA+Sox4+ cells in different clusters, right UMAP; merged. (c) RelA+Sox4+ and Pou4f1+ cells followed similar pattern of cluster wise distribution. Left UMAP: Pseudo time trajectory analysis of RelA+Sox4+ cells showing increased numbers in later time points at clusters other than 1, 11 and 12. Middle UMAP: Few Sox4+RelA+ and Pou4f1+ cells were evident in clusters 1, 11 and 12. Right UMAP: Almost no Sox4+RelA+Pou4f1+ cells were evident in clusters, 1, 11 and 12. (d) Pseudo time trajectory analysis (UMAP) for all clusters indicated clusters 1, 11, 12 generated earlier than other clusters where Pou4f1 expressed. (e) Presence of Rax+ and Atoh7+ cells in CiRGC population on day4 (UMAP).

Extended data Fig. 12.
Extended data Fig. 12. SOX4 and NF-kB (RELA) mediates Md-CiRGC fate commitment through a non-canonical pathway.

(a,b) Expression analysis of RelA, Sox4 and Pou4f1 (violin plot) in Pou4f1+ and Pou4f1+Sox4+RelA+ cell sets. (c) Schematic diagram for the mechanism of CiRGC generation.

Discussion

In our prior study we demonstrated the generation of photoreceptors from fibroblasts by 5C within 10 days. Since we observed cells morphologically similar to RGCs between reprogramming day 4 & 5 using 5C, we hypothesized that we might achieve the generation of RGCs at an earlier time point. Employing a Sncg promoter reporter our investigation indeed yielded GFP+ cells on day4, strongly indicating their RGC lineage (Extended data Figs. 13 step1 & 14a). Subsequently, we discovered PAX6+ (progenitor marker) cells on day 3, suggesting that 5C-mediated human retinal neuron reprogramming from fibroblasts may occur through a progenitor like stage (Extended data Fig. 14b)43. This was further supported by SC-RNA Seq showed presence of reminiscent PAX6+ cells on day 4 (Extended data Fig. 14c). Next, we investigated if 5C could reprogram human Müller cells into CiRGCs. Encouragingly, we observed the expression of Sncg-driven GFP in Md-CiRGCs, however when 5C coupled with 2F, we noted a higher number of GFP+ cells (Extended data Figs. 13 step2, 14d-f). These findings spurred us to investigate the potential of 5C+2F in regenerating RGCs from endogenous Müller cells. Administering this combination into NMDA-injured mice eye, we observed the proliferation of Müller cells, predominantly at the outer edge of the INL, albeit at reduced efficiency. A few were observed to progress towards GCL (Extended data Fig. 15a). At this stage, we searched for additional molecules that could potentially induce BrdU+ Müller cell proliferation and facilitate their neurogenesis. We considered S24795, a partial agonist of α7 nAChR, given its established role in neural progenitor cell survival and CNS neurogenesis44. Finally, we found S24795 along with 5C+2F (but not S24795 alone) was able to enhance Müller cell proliferation and migration. Lineage tracing with a TdTomato reporter confirmed the Müller cell origin of regenerated CiRGCs, which extended their axons through the optic nerve head into the proximal part of optic nerve— a crucial step for visual function restoration. Consistently, we found recovery of visual functions starting from day 16, which reached its peak level on day 60. We noted differential vision improvement between right and left eyes on day 16 and day 23, with right eye improved better. This may be due to variability of RGC subtypes between left and right eyes which connect differentially with newly regenerated CiRGC axons45. Encouragingly, we generated CiRGCs from human primary Müller and fibroblasts by this chemical-only approach, offering its potential applications in humans. Interestingly, Fd-CiRGCs are immature (not expressing maturation marker THY1), hence would be ideal for cell transplantation. Md-CiRGCs express both vision-forming and pan RGC markers BRN3A and RBPMS, which are reported to be the best RGC markers across species46–50. Interestingly, Md-CiRGCs exhibited axon growth-enhancing genes (Extended data Fig. 15b) and 90% of BRN3A+ Md-CiRGCs express the αRGC marker SPP1, a subtype, possess a greater propensity to regenerate their axons after axotomy51, 52. These results suggest a pro-regenerative state in CiRGCs, that support its regenerative ability in mice in vivo studies. Interestingly, αRGCs are most susceptible to injury in rat bead and ONC model and BRN3A+ Md-CiRGCs isolated from in vitro culture would be ideal for transplantation studies in these models53. Furthermore, BRN3A+ larger RGCs are preferentially susceptible to death in human glaucoma and BRN3A+ endogenously regenerated Md-CiRGCs would be appropriate for therapy54. Our mechanistic study suggests that NF-κB and SOX4 could have a central role in the formation of CiRGCs from primary Müller glia, representing a probable in vivo mechanism. We do not rule out the possibility for generation of certain CiRGCs through an ATOH7-dependent pathway, as we have identified a minor CiRGC population expressing ATOH7.

Extended data Fig. 13.
Extended data Fig. 13. Stepwise identification of small molecule cocktail for endogenous CiRGC regeneration.

Step1: Scheme for stepwise generation of potential CiRGCs and progenitors by modifying 5C medicated photoreceptor generation protocol. Step2: Scheme for the generation of potential CiRGCs from human Muller cells using 5C plus 2 additional factors (2F, Noggin and IGF1). Step3: Scheme for identification of novel chemical cocktail for the generation of CiRGCs from Muller cells in vivo.

Extended data Fig. 14.
Extended data Fig. 14. Generation of CiRGCs from human primary Müller cells (HPM) and foreskin fibroblasts (HFF) on day4 and identification of PAX6+ progenitor cells in earlier time points

(a) SNCG-GFP expressing CiRGCs from HADF on day4 of 5C-chemical reprogramming. (b) PAX6+ and RAX+ cells generated from HADF after day3 of 5C-reprogramming. (c) Reminiscent Pax6+ cells in SC-RNA-Seq after 6C+2F treatment in HADF on day4. (d) SNCG-GFP expressing CiRGCs from HPM on day4 of 5C-chemical reprogramming. (e) PAX6+ and RAX+ cells generated from HPM after day3 of 5C-reprogramming. (f) Reminiscent Pax6+ cells in SC-RNA-Seq after 6C+2F treatment in HPM on day4.

Extended data Fig. 15.
Extended data Fig. 15.

(a) 5C+2F induces generation of BrdU+ cells but with lower efficiency. (b) Expression of indicated axon regeneration genes in Md-CiRGCs.

In summary, this study introduces a chemical-only strategy that effectively triggers endogenous RGC regeneration from Müller cells and rescue vision after RGC degeneration. This technique has the potential to overcome challenges associated with cell transplantations and could prove valuable in treating retinal neurodegenerative disorders such as glaucoma and optic nerve hypoplasia.

Competing Interests

All other authors declare no competing interest. B.M. and R.M.S. are listed as inventors in a disclosure for patent application submitted to office of technology and commercialization at CHLA.

Author Contributions

B.M. designed experiments; B.M, wrote the manuscript; R.M.S, A, S, and BM: performed experiments.

Acknowledgements.

B.M. is supported by Knights Templar Eye Research Foundation, Glaucoma Research Foundation and Children’s Hospital Los Angeles start up fund. We thank the stem cell analytics core, cellular imaging core, and special biology and genomics core at Children’s Hospital Los Angels for pilot grant support and data collection. We thank Drs. David Cobrinik, Thomas Lee, Aaron Nagiel, Mark Borchert for critical review of the results.

References

  1. 1.↵
    Tanaka, E.M. & Ferretti, P. Considering the evolution of regeneration in the central nervous system. Nat Rev Neurosci 10, 713–723 (2009).
    OpenUrlCrossRefPubMed
  2. 2.
    Li, H. & Chen, G. In Vivo Reprogramming for CNS Repair: Regenerating Neurons from Endogenous Glial Cells. Neuron 91, 728–738 (2016).
    OpenUrlCrossRef
  3. 3.↵
    Ming, G.L. & Song, H. Adult neurogenesis in the mammalian brain: significant answers and significant questions. Neuron 70, 687–702 (2011).
    OpenUrlCrossRefPubMedWeb of Science
  4. 4.↵
    Goldman, D. Müller glia cell reprogramming and retina regeneration Nature Review Neuroscience 15, 431–442 (2014).
    OpenUrlCrossRefPubMed
  5. 5.
    Jorstad, N.L. et al. Stimulation of functional neuronal regeneration from Muller glia in adult mice. Nature (2017).
  6. 6.↵
    Ueki, Y. et al. Transgenic expression of the proneural transcription factor Ascl1 in Muller glia stimulates retinal regeneration in young mice. Proc Natl Acad Sci U S A 112, 13717–13722 (2015).
    OpenUrlAbstract/FREE Full Text
  7. 7.
    Cope, E.C. & Gould, E. Adult Neurogenesis, Glia, and the Extracellular Matrix. Cell Stem Cell 24, 690–705 (2019).
    OpenUrlCrossRefPubMed
  8. 8.↵
    He, G.Y.a.Z. Glial inhibition of CNS axon regeneration. Nature Reviews Neuroscience 7, 617 (2006).
  9. 9.↵
    Goldman, S.A. Stem and Progenitor Cell-Based Therapy of the Central Nervous System: Hopes, Hype, and Wishful Thinking. Cell Stem Cell 18, 174–188 (2016).
    OpenUrlCrossRef
  10. 10.
    Trounson, A. & McDonald, C. Stem Cell Therapies in Clinical Trials: Progress and Challenges. Cell Stem Cell 17, 11–22 (2015).
    OpenUrlCrossRefPubMed
  11. 11.
    Kikuchi, T. et al. Human iPS cell-derived dopaminergic neurons function in a primate Parkinson’s disease model. Nature 548, 592–596 (2017).
    OpenUrlCrossRefPubMed
  12. 12.
    Vierbuchen, T. et al. Direct conversion of fibroblasts to functional neurons by defined factors. Nature 463, 1035–1041 (2010).
    OpenUrlCrossRefPubMedWeb of Science
  13. 13.
    Victor, M.B. et al. Generation of human striatal neurons by microRNA-dependent direct conversion of fibroblasts. Neuron 84, 311–323 (2014).
    OpenUrlCrossRefPubMed
  14. 14.
    Xue, Y. et al. Direct conversion of fibroblasts to neurons by reprogramming PTB-regulated microRNA circuits. Cell 152, 82–96 (2013).
    OpenUrlCrossRefPubMedWeb of Science
  15. 15.↵
    Yoo, A.S. et al. MicroRNA-mediated conversion of human fibroblasts to neurons. Nature 476, 228–231 (2011).
    OpenUrlCrossRefPubMedWeb of Science
  16. 16.↵
    Liu, Y. et al. Ascl1 Converts Dorsal Midbrain Astrocytes into Functional Neurons In Vivo. J Neurosci 35, 9336–9355 (2015).
    OpenUrlAbstract/FREE Full Text
  17. 17.
    Weinberg, M.S., Criswell, H.E., Powell, S.K., Bhatt, A.P. & McCown, T.J. Viral Vector Reprogramming of Adult Resident Striatal Oligodendrocytes into Functional Neurons. Mol Ther 25, 928–934 (2017).
    OpenUrlCrossRef
  18. 18.
    Niu, W. et al. SOX2 reprograms resident astrocytes into neural progenitors in the adult brain. Stem Cell Reports 4, 780–794 (2015).
    OpenUrl
  19. 19.
    Gaia Colasante, G.L., Alicia Rubio, Lucian Medrihan, Latefa Yekhlef, Alessandro Sessa, Luca Massimino, Serena G. Giannelli, Silvio Sacchetti, Massimiliano Caiazzo, Damiana Leo, Dimitra Alexopoulou, Maria Teresa Dell’Anno, Ernesto Ciabatti, Marta Orlando, Michele Studer, Andreas Dahl, & Raul R. Gainetdinov, S.T., Fabio Benfenati, and Vania Broccoli Rapid Conversion of Fibroblasts into Functional Forebrain GABAergic Interneurons by Direct Genetic Reprogramming. Cell Stem cell 17, 719–134 (2015).
    OpenUrlCrossRefPubMed
  20. 20.
    ia Rivetti di Val Cervo, R.A.R., Giada Spigolon, Débora Masini, Elisa Martín-Montañe, Enrique M Toledo, Gioele La Manno, Michael Feyder, Christian Pifl2, Yi-Han Ng, Sara Padrell Sánchez, Sten Linnarsson, Marius Wernig, Tibor Harkany, Gilberto Fisone & Ernest Arenas induction of functional dopamine neurons from human astrocytes in vitro and mouse astrocytes in a Parkinson’s disease model. Nature Biotechnology 35 (2017).
  21. 21.
    Torper, O. et al. In Vivo Reprogramming of Striatal NG2 Glia into Functional Neurons that Integrate into Local Host Circuitry. Cell Rep 12, 474–481 (2015).
    OpenUrlCrossRefPubMed
  22. 22.↵
    Guo, Z. et al. In vivo direct reprogramming of reactive glial cells into functional neurons after brain injury and in an Alzheimer’s disease model. Cell Stem Cell 14, 188–202 (2014).
    OpenUrlCrossRefPubMed
  23. 23.↵
    Hu, W. et al. Direct Conversion of Normal and Alzheimer’s Disease Human Fibroblasts into Neuronal Cells by Small Molecules. Cell Stem Cell 17, 204–212 (2015).
    OpenUrlCrossRefPubMed
  24. 24.
    Li, X. et al. Small-Molecule-Driven Direct Reprogramming of Mouse Fibroblasts into Functional Neurons. Cell Stem Cell 17, 195–203 (2015).
    OpenUrlCrossRefPubMed
  25. 25.
    Babos, K. & Ichida, J.K. Small Molecules Take a Big Step by Converting Fibroblasts into Neurons. Cell Stem Cell 17, 127–129 (2015).
    OpenUrl
  26. 26.↵
    Zhang, L. et al. Small Molecules Efficiently Reprogram Human Astroglial Cells into Functional Neurons. Cell Stem Cell 17, 735–747 (2015).
    OpenUrlCrossRefPubMed
  27. 27.↵
    Mahato, B. et al. Pharmacologic fibroblast reprogramming into photoreceptors restores vision. Nature 581, 83–88 (2020).
    OpenUrl
  28. 28.↵
    Takeda, M. et al. alpha-Aminoadipate induces progenitor cell properties of Muller glia in adult mice. Invest Ophthalmol Vis Sci 49, 1142–1150 (2008).
    OpenUrlAbstract/FREE Full Text
  29. 29.↵
    Mike O. Karl, S.H., Branden R. Nelson, Kristine Tan, Brian Buckingham, and Thomas A. Reh Stimulation of neural regeneration in the mouse retina. PNAS 105, 19508–19513 (2008).
    OpenUrlAbstract/FREE Full Text
  30. 30.↵
    Natalia Surzenko, T.C., Amelia Bachleda, Lee Langer and Larysa Pevny SOX2 maintains the quiescent progenitor cell state of postnatal retinal Müller glia. Development 140, 1445–1455 (2013).
    OpenUrlAbstract/FREE Full Text
  31. 31.↵
    Shen, F. et al. Glutamate-induced glutamine synthetase expression in retinal Muller cells after short-term ocular hypertension in the rat. Invest Ophthalmol Vis Sci 45, 3107–3112 (2004).
    OpenUrlAbstract/FREE Full Text
  32. 32.↵
    Miesfeld, J.B. et al. The Atoh7 remote enhancer provides transcriptional robustness during retinal ganglion cell development. Proc Natl Acad Sci U S A 117, 21690–21700 (2020).
    OpenUrlAbstract/FREE Full Text
  33. 33.↵
    Bin Lin, A.K., Nobushige Tanaka, Satchidananda Panda, and Richard H. Masland Restoration of visual function in retinal degeneration mice by ectopic expression of melanopsin. PNAS 105, 16009–16014 (2008).
    OpenUrlAbstract/FREE Full Text
  34. 34.
    Katada, Y. et al. Highly sensitive visual restoration and protection via ectopic expression of chimeric rhodopsin in mice. iScience 26, 107716 (2023).
  35. 35.↵
    Divya, M.S. et al. Intraocular Injection of ES Cell-Derived Neural Progenitors Improve Visual Function in Retinal Ganglion Cell-Depleted Mouse Models. Front Cell Neurosci 11, 295 (2017).
  36. 36.↵
    Gu, L., Kawaguchi, R., Caprioli, J. & Piri, N. The effect of Rbfox2 modulation on retinal transcriptome and visual function. Sci Rep 10, 19683 (2020).
  37. 37.
    Guo, X. et al. Preservation of vision after CaMKII-mediated protection of retinal ganglion cells. Cell 184, 4299–4314 e4212 (2021).
    OpenUrlCrossRefPubMed
  38. 38.↵
    Su, J. et al. In vivo base editing rescues photoreceptors in a mouse model of retinitis pigmentosa. Mol Ther Nucleic Acids 31, 596–609 (2023).
    OpenUrl
  39. 39.↵
    de Melo, J. et al. Injury-independent induction of reactive gliosis in retina by loss of function of the LIM homeodomain transcription factor Lhx2. Proc Natl Acad Sci U S A 109, 4657–4662 (2012).
    OpenUrlAbstract/FREE Full Text
  40. 40.↵
    Takae Kiyama, H.Y.A., Tudor C. Badea, and Chai-An Mao. Pou4f1-Tbr1 transcriptional cascade controls the formation of Jam2-expressing retinal ganglion cells. Frontiers in Ophthalmology (2023).
  41. 41.↵
    Moreno, C.S. SOX4: The unappreciated oncogene. Semin Cancer Biol 67, 57–64 (2020).
    OpenUrlCrossRef
  42. 42.↵
    Jones, K. et al. SOX4 and RELA Function as Transcriptional Partners to Regulate the Expression of TNF-Responsive Genes in Fibroblast-Like Synoviocytes. Front Immunol 13, 789349 (2022).
  43. 43.↵
    Till Marquardt, R.A.-P., Nicole Andrejewski, Raffaella Scardigli, Francois Guillemot, and Peter Gruss Pax6 Is Required for the Multipotent State of Retinal Progenitor Cells. Cell 105, 43–55 (2005).
    OpenUrl
  44. 44.↵
    Kita, Y. et al. Galantamine promotes adult hippocampal neurogenesis via M(1) muscarinic and alpha7 nicotinic receptors in mice. Int J Neuropsychopharmacol 17, 1957–1968 (2014).
    OpenUrlCrossRefPubMed
  45. 45.↵
    Rheaume, B.A. et al. Single cell transcriptome profiling of retinal ganglion cells identifies cellular subtypes. Nat Commun 9, 2759 (2018).
    OpenUrlCrossRefPubMed
  46. 46.↵
    Francisco M. Nadal-Nicolás, M.J.n.-L.p., Paloma Sobrado-Calvo, Leticia Nieto-López, Isabel Cánovas-Martínez, Manuel Salinas-Navarro, Manuel Vidal-Sanz, and Marta Agudo Brn3a as a Marker of Retinal Ganglion Cells: Qualitative and Quantitative Time Course Studies in Naïve and Optic Nerve–Injured Retinas. Investigative Ophthalmology & Visual Science 50, 3860–3838-3868 (2009).
    OpenUrlAbstract/FREE Full Text
  47. 47.
    Iain M Dykes, J.L., S Raisa Eng, Eric E Turner Brn3a regulates neuronal subtype specification in the trigeminal ganglion by promoting Runx expression during sensory differentiation. Neural Development 5 (2010).
  48. 48.
    Quina, L.A. et al. Brn3a-expressing retinal ganglion cells project specifically to thalamocortical and collicular visual pathways. J Neurosci 25, 11595–11604 (2005).
    OpenUrlAbstract/FREE Full Text
  49. 49.
    Vladimir Vladimirovich Muzyka, M.B.a.T.C.B. Postnatal developmental dynamics of cell type specification genes in Brn3a/Pou4f1 Retinal Ganglion Cells. Neural Development 13:15 (2018).
  50. 50.↵
    Nadal-Nicolas, F.M. et al. Pan-retinal ganglion cell markers in mice, rats, and rhesus macaques. Zool Res 44, 226–248 (2023).
    OpenUrl
  51. 51.↵
    Gallego-Ortega, A. et al. Alpha retinal ganglion cells in pigmented mice retina: number and distribution. Frontiers in Neuroanatomy.
  52. 52.↵
    Duan, X. et al. Subtype-specific regeneration of retinal ganglion cells following axotomy: effects of osteopontin and mTOR signaling. Neuron 85, 1244–1256 (2015).
    OpenUrlCrossRefPubMed
  53. 53.↵
    VanderWall, K.B. et al. Differential susceptibility of retinal ganglion cell subtypes in acute and chronic models of injury and disease. Sci Rep 10, 17359 (2020).
  54. 54.↵
    Lis​a A. Kerrigan–Baumrind, H.A.Q., Mary E. Pease, Danielle F. Kerrigan, and Rebecca S. Mitchell Number of Ganglion Cells in Glaucoma Eyes Compared with Threshold Visual Field Tests in the Same Persons. Invest Ophthalmol Vis Sci. 41, 741–748 (2000).
    OpenUrlAbstract/FREE Full Text
Back to top
Posted December 27, 2023.
Download PDF

Supplementary Material

Email
Vision rescue via chemically engineered endogenous retinal ganglion cells
Rana Muhammad Shoaib, Aregnazan Sandrosyan, Biraj Mahato
bioRxiv 2023.12.27.572921; doi: https://doi.org/10.1101/2023.12.27.572921
Twitter logo Facebook logo LinkedIn logo Mendeley logo
Citation Tools

Subject Area