Sphingosine-1-Phosphate Signaling through Müller Glia Regulates Neuroprotection, Accumulation of Immune Cells, and Neuronal Regeneration in the Rodent Retina

Expression of S1P-related genes in the retina

We began by analyzing patterns of expression of S1P-related genes in scRNA-seq libraries that were established from control retinas and damaged retinas at different times after NMDA treatment; these libraries were generated and analyzed previously (Todd et al., 2019; Hoang et al., 2020; Campbell et al., 2021a, b). UMAP plots revealed discrete clusters of all major retinal cell types (Fig. 1a,b). Clusters of cells were labeled based on distinct patterns of gene expression as described in the Materials and Methods. Neuronal cells from control and damaged retinas were clustered together regardless of time after NMDA treatment (Fig. 1a,b). In contrast, resting MG, including cells from 48 and 72 h after NMDA, and activated MG from 3, 6, 12, and 24 h after NMDA were spatially separated by UMAP embedding (Fig. 1a–c). Activated MG were identified, in part, based on downregulation of genes such as Slc1a3 and upregulation of genes such as Vim or Nes (Fig. 1c). We examined patterns of expression and changes in expression levels of S1P-related genes (Fig. 1d–i). S1pr4, S1pr5, Sgpl1, and Sphk2 were either not detected at significant levels or detected in very few cells in the retina (data not shown). S1pr2 was expressed by relatively few retinal cells, but S1pr1, S1pr3, and Sphk1 were highly expressed by activated MG (Fig. 1d,e). In addition, S1pr1 was highly expressed by endothelial cells (Fig. 1d). We did not perform detailed analyses of S1P-related genes in endothelial cells because there were <350 cells captured in these libraries (Fig. 1b). S1pr1 and Sphk1 appeared to be robustly upregulated in activated MG at 3 and 6 h after NMDA treatment, but downregulated in MG thereafter (Fig. 1d,e). To perform a detailed analysis of S1P-related genes in MG, we bioinformatically isolated and re-embedded nearly 6,200 MG into a UMAP dimensional reduction (Fig. 1f,g). These MG formed a distinct trajectory of cells with resting MG from control retinas clustered to one side, ++activated MG from 3 h after NMDA-treatment clustered to the other side of the plot, and MG from later times after NMDA bridging the middle of the trajectory (Fig. 1f,g). This analysis revealed that levels of S1pr1 were relatively low in resting MG and levels were significantly upregulated in ++activated MG at 3 h after NMDA and further upregulated in +activated MG at 6 h after NMDA (Fig. 1h,i). Levels of S1pr1 were significantly downregulated in activated MG and returning to resting MG from retinas at 12–72 h after NMDA (Fig. 1i). Levels of S1pr2 were very low in MG, and levels of S1pr3 were not significantly changed, but the percentage of MG that expressed S1pr3 was greatest in MG that were returning to a resting phenotype from retinas 24–72 h after NMDA (Fig. 1h,i). By comparison, Sphk1 was low in resting MG and highly upregulated in ++activated (3 h after NMDA) and +activated MG (6 h after NMDA), but significantly downregulated thereafter (Fig. 1h,i). Collectively, these findings suggest that after acute injury, levels of Sphk1 (S1P synthesis) are very rapidly and transiently upregulated and the upregulation in S1P synthesis is followed by a rapid and transient upregulation in autocrine signaling through S1pr1 in MG.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Patterns of expression of S1P-related genes in damaged mouse retinas. scRNA-seq was used to identify patterns of expression of S1P-related factors among retinal cells with the data presented in UMAP a–h or violin plots i. Aggregate scRNA-seq libraries were generated for cells from control retinas and retinas 3, 6, 12, 24, 36, 48, and 72 h after NMDA treatment. MG were bioinformatically isolated and analyzed from the large aggregate library (f–i). UMAP-ordered cells formed distinct clusters of neuronal cells, resting MG, activated MG, and returning to resting MG based on distinct patterns of gene expression (c). UMAP heatmap plots illustrate patterns and levels of expression of S1P receptors S1pr1, S1pr2, S1pr3, and S1P metabolism gene Sphk1 (h). Violin plots illustrate relative levels of expression in clusters of MG, activated MG, and returning to resting MG (i). Significance of difference was determined by using a Wilcox rank sum with Bonferroni’s correction (supplemental Table 1). Abbreviations: MG, Müller glia; NMDA, N-methyl-D-aspartate; UMAP, Uniform Manifold Approximation and Projection.

We next probed patterns of expression of S1P-related genes in a large integrated scRNA-seq database of mouse retinal cells from different postnatal ages ranging from 2 to 25 weeks; this database has been generated and described previously (Li et al., 2024). This aggregated database contained more than 330,000 cells which form numerous distinct clusters of retinal cells in UMAP plots (Fig. S1a). Glial cells formed distinct clusters based on patterns of expression of cell-distinguishing markers (see Materials and Methods; Fig. S1b). Sphk1 was detected in astrocytes, MG and a few clusters of amacrine cells (Fig. 1c). By comparison, S1pr1 and S1pr3 were detected at high levels in MG (Fig. S1d,e). Additionally, S1pr1 was detected in endothelial cells, and S1pr3 was detected at high levels in pericytes and at lower levels in endothelial cells and astrocytes (Fig. S1d,e). Sgpl1 was expressed at high levels in most microglia and in cells scattered across all different major cell types (Fig. S1f).

We next bioinformatically isolated the MG to perform more detailed analyses of S1P-related genes. We renormalized and re-established principal components for UMAP embedding, filtered clusters with abnormally high features/UMI per cell, filtered contamination from astrocytes (high Pax2 and S100b) and bipolar cells (high Lhx4, Otx2, Scgn). These processes produced distinct clusters of MG; however, these cells were segregated by platform, not developmental stage (Fig. S1g,h). Relative levels of S1pr1 and S1pr3 were significantly increased in older mice (16 and 25 weeks of age), whereas levels of Sphk1 were significantly higher in young mice (2 weeks of age; Fig. S1i,j; Supplemental Table 1). The MG from animals at 8 weeks of postnatal development had elevated levels of S1pr3; however, these cells may have been reactive with significantly elevated levels of Tgfb2 and Klf6 (Fig. S1i,j; Table S1), which is characteristic of activated MG that are returning toward a resting phenotype (Hoang et al., 2020; Palazzo et al., 2022a).

Validation of patterns of expression of S1pr1 and Sphk1

To validate findings from scRNA-seq libraries, we applied different immunolabeling strategies using different antibodies to S1pr1 and Sphk1 (Table 1). However, none of these antibodies produced plausible patterns of labeling regardless of different antigen retrieval strategies including weak fixation, methanol washes, sodium citrate washes, exclusion of detergent, or applying different types of detergents (data not shown). Thus, we applied fluorescent in situ hybridization (FISH) strategies with probes for S1pr1 and Sphk1. We did not probe for S1pr2 or S1pr3 because the scRNA-seq data indicated very low levels of expression in the retina. In undamaged retinas, S1pr1 puncta were associated with Sox2-labeled MG nuclei (Fig. 2a). S1pr1 FISH signal was also detected in putative endothelial cells associated with autofluorescent red blood cells that were scattered across retina (Fig. 2a). Consistent with findings from scRNA-seq, we found an increase in S1pr1 puncta associated with Sox2-labeled MG nuclei at 4 h after NMDA treatment (Fig. 2a).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Fluorescence in situ hybridization (FISH) for S1PR1 and SPHK1. Glast-CreER:LNL-tTA:tetO-mAscl1-ires-GFP mice received intraperitoneal injections of tamoxifen or corn oil once daily for 4 consecutive days. Eyes received intravitreal injections with NMDA or saline, and retinas were collected 4 h after NMDA treatment. Retinal sections were labeled with antibodies to Sox2 (cyan), GFP (green), or FISH probes to S1pr1 (red puncta; a, c) or Sphk1 (red puncta; b). Regions of interest (yellow boxes) are enlarged 1.8-fold and displayed in adjacent channels. Solid arrows indicate Sox2⁺/GFP^low MG nuclei, and hollow arrows indicate Sox2⁺/GFP^high MG nuclei. The area occupied by Sox2⁺ (b) or GFP⁺ cells (e), was selected to distinguish FISH puncta labeling within these regions. Small double arrows indicate endothelial cells. Scale bars: a, b, c, and e represent 50 µm. Histograms represent the mean (bar ± SD) number of FISH puncta per MG and each dot represents one biological replicate (d; n = 6 or 7). Significance of difference (p values) was determined by using a paired t test. Abbreviations: ONL, outer nuclear layer; INL, inner nuclear layer; IPL, inner plexiform layer; GCL, ganglion cell layer; NMDA, N-methyl-D-aspartate; ns, not significant.

In undamaged retinas, there were very few Sphk1 FISH puncta scattered randomly across the retina (Fig. 2b). At 4 h after NMDA treatment, there were numerous Sphk1 puncta scattered across the retina, and some of these puncta were associated with Sox2⁺ MG nuclei (Fig. 2b). According to scRNA-seq data, levels of Sphk1 are increased in MG at 3–6 h after NMDA, whereas the scattered distribution of Sphk1 FISH puncta did not reveal MG-specific upregulation of transcripts.

To investigate whether S1P signaling through S1pr1 is altered during the reprogramming of MG into progenitor-like cells, we applied FISH probes to retinas where the MG have been forced to express the neurogenic basic Helix-Loop-Helix (bHLH) transcription factor Ascl1 (Glast-CreER:LNL-tTA:tetO-mAscl1-ires-GFP; Jorstad et al., 2017). In adult mouse retinas, inducible forced expression of Ascl1 in MG, combined with NMDA-induced damage and HDAC inhibitor (trichostatin A), stimulates the reprogramming of MG into bipolar-like cells that integrate into circuitry and respond to visual stimuli (Jorstad et al., 2017). Ascl1-overexpressing (Ascl1-OE) MG were identified by GFP expression, which is linked to the Ascl1 transgene via an internal ribosomal entry sequence (IRES; Fig. 2). Counts of S1pr1 FISH puncta associated with Ascl1-OE MG (Sox2⁺/GFP⁺ cells) and of puncta associated with neighboring WT MG (Sox2⁺/GFP⁻ cells) within the same section were measured. In undamaged retinas, numbers of S1pr1 puncta in Ascl1-OE MG were not significantly different from WT MG (Fig. 2c–e). In damaged retinas, there were significantly more S1pr1 FISH puncta per MG in NMDA-damaged retinas compared with numbers seen in saline-treated retinas (Fig. 2c–e), consistent with scRNA-seq data. In damaged retinas, with Ascl1-OE MG, numbers of S1pr1 FISH puncta were significantly reduced in Ascl1-OE MG compared with numbers seen in WT MG (Fig. 2c–e). Collectively, these findings indicate that S1pr1 and Sphk1 transcripts are upregulated in response to NMDA damage but downregulated in response to forced expression of Ascl1 in damaged retinas.

S1P signaling and reprogramming of Ascl1-over expressing MG

To investigate how S1P signaling influences Ascl1-mediated neurogenesis, we applied small molecule inhibitors to Sphk1 (PF543) and S1pr1/3 (VPC23019) to the retinas of Ascl1-OE mice. All pharmacological compounds used in this study are described in Table 3. Retinas were treated with NMDA and inhibitors, inhibitors at 1 day post-injury (DPI), TSA at 2 DPI, and harvested at 16 DPI (Fig. 3). Application of PF543 or VPC23019 alone suppressed glial differentiation but did not significantly influence numbers of differentiating neurons (data not shown). Combined application of Sphk1 + S1pr1/3 inhibitors significantly decreased the percent of GFP⁺/Sox2⁺ cells (Fig. 3a–c), indicating that the S1P pathway promotes glial dedifferentiation in Ascl1-OE MG. By comparison, application of Sphk1 + S1pr1/3 inhibitors significantly increased the percent of GFP⁺/Otx2⁺ and GFP⁺/Lhx4⁺ cells (Fig. 3d–f,g,j,l), indicating that S1P pathway suppresses neuronal differentiation in Ascl1-OE MGPCs. Sphk1 + S1pr1/3 inhibitors had no significant effect upon total numbers of proliferating EdU-positive cells or numbers of EdU⁺/GFP⁺ cells (Fig. 3g,h). By comparison, there was no significant effects of S1P pathway inhibition on numbers of GFP⁺ cells colabeled with the cone bipolar cell marker Scgn (Fig. 3i,k), suggesting that the S1P pathway does not promote differentiation toward this type of bipolar cell. Collectively, these findings indicate that inhibition of S1pr1/3 and Sphk1 in Ascl1-OE MG enhances the differentiation of bipolar-like cells, while glial phenotype was suppressed.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Inhibition of S1P signaling enhances Ascl1-mediated reprogramming of MG into neurogenic progenitor cells. Schematic summary of S1P signaling and sites of action for different small molecule agonists and antagonists a, Glast-CreER:LNL-tTA:tetO-mAscl1-ires-GFP mice were intraperitoneally injected with tamoxifen or corn oil once daily for 4 d. Eyes were intravitreally injected NMDA and either a vehicle control or inhibitors to Sphk1 (PF543) and S1pr1/3 (VPC23019). At 24HPI, eyes were treated with vehicle or inhibitors. Eyes were treated with TSA at 48HPI. Retinas were collected 24 d from the first tamoxifen injection. Retinal sections were labeled with DAPI (blue, b, d) or EdU (blue, e, h) and antibodies to GFP (green) and Sox2 (red, b, d), Otx2 (red, e, h), Scgn (red, i), or Lhx4 (red, j). Regions of interest (yellow boxes) are enlarged 1.5 (b, d) or 2-fold (e, h, i, j) and displayed in separate panels. Solid arrows indicate GFP⁺/Sox2^high cells in b and d, and hollow arrowheads indicate GFP⁺/Sox2^low cells. In e and h, solid arrows indicate GFP⁺/Otx2⁺/EdU⁺ cells. Scale bars: b and e represent 50 µm. Histograms represent the mean (bar ± SD) where each dot represents one biological replicate (n = 8), and blue lines connect control and treated eyes from the same individual (c, g, f). Significance of difference (p values) was determined by using a paired t test. Abbreviations: ONL, outer nuclear layer; INL, inner nuclear layer; IPL, inner plexiform layer; GCL, ganglion cell layer; NMDA, N-methyl-D-aspartate; ns, not significant.

Cross-regulation of S1P/S1pr1 and NFκB signaling in MG

S1P receptors differ by their affinity to S1P and by their coupling to different G-proteins, thereby resulting in distinct cellular responses and physiological outcomes. S1pr1 has been shown to predominantly couple to Gi, activating Phospholipase C (PLC), MAPK/Erk, and PI3K/AKT pathways (Pyne and Pyne, 2017), and secondarily impact NFκB signaling (Sun et al., 2024). We have recently reported that NFκB signaling is rapidly, transiently, and selectively activated in MG in damaged mouse retinas, and this activation is mediated by proinflammatory cytokines produced by reactive microglia (Palazzo et al., 2022a, 2023). Additionally, NFκB signaling promotes glial fate in MGPCs of chicks and Ascl1-OE mice (Palazzo et al., 2020, 2022a). S1P signaling is known to activate inflammatory responses via NFκB in many cell types, including endothelial cells, cancer cells, and macrophages (Liang et al., 2013; Campos et al., 2016; Xiao et al., 2018; Zheng et al., 2019; Del Gaudio et al., 2020). Thus, we investigated whether pharmacological activation or inhibition of S1P-related receptors and enzymes influenced NFκB activation and expression of other cell signaling biomarkers in normal and damaged retinas.

Eyes were treated with a single intraocular injection of S1P or S1pr1 agonists (CYM5422 or SEW2871) and retinas were harvested 24 h later. Alternatively, eyes were injected with an S1pr1 agonist (SEW2871 or CYM5442), S1pr1 antagonist (MT1303 or NIBR0213), S1pr3 antagonist (TY52156), S1pr modulator (FTY720), Sphk1 antagonist (PF543), or S1P lyase antagonist (S1PL-in-31) before and with NMDA, and retinas were harvested 24 h after the last injection. FTY720, or Fingolimod, is a clinically approved S1pr1 modulator which initially acts as an agonist to S1pr1 but later induces receptor internalization and degradation (Chun et al., 2021a). We used NFκB reporter mice wherein eGFP expression is driven by a chimeric promoter containing three HIV-NFκB cis-regulatory elements (Magness et al., 2004). In the NFκB-reporter mice, there is MG-specific NFκB activation in retinas treated with NMDA, cytokines, and different growth factors (Palazzo et al., 2023; Fig. 4a). In undamaged retinas, we found that S1pr1 activation, via S1P or different agonists, significantly increased numbers of Sox2-positive MG that express NFκB reporter in undamaged retinas (Fig. 4b,c). Additionally, S1pr1 agonist treatments increased numbers of NFκB-expressing MG in damaged retinas (Fig. 4b). Consistent with findings from studies using S1pr1 agonists, numbers of NFκB-GFP-expressing MG were increased in retinas treated with S1P lyase inhibitor, decreased in retinas treated with S1pr1 antagonists, and decreased in retinas treated with Sphk1 inhibitor (Fig. 4d,e). S1pr3 antagonist (Fig. 4d) and S1pr1 modulator (data not shown) had no significant effects on numbers of MG that expressed NFκB reporter in damaged retinas. Collectively, these findings suggest that autocrine S1P signaling via Sphk1-mediated synthesis of S1P and activation of S1pr1 activates NFκB signaling in MG in normal and damaged retinas.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Cross-regulation of S1P/S1pr1 and NFκB signaling in MG. Retinas were obtained from undamaged or NMDA-injured eyes treated with a vehicle control or a small molecule agonist/antagonist a–g, Retinal sections were labeled with DAPI (e) and antibodies to Sox2 (red, a, c, e; pseudocolored cyan, f) and NFκB-GFP (green, a, c, e) or FISH probes to S1pr1 (red puncta; f). Arrows indicate the nuclei of MG. Scale bars: a, c, and f represent 50 µm. Regions of interest (yellow boxes) are enlarged 1.8-fold and displayed in separate panels (f). Histograms illustrate the mean (bar ± SD) number of NFκB-eGFP in MG (b, d) or number of FISH puncta per MG (g), where each dot represents one biological replicate (n = 5–8), and blue lines connect replicates from control and treated retinas from one individual. Significance of difference (p values) was determined by using a paired t test (b, d, g). Abbreviations: ONL, outer nuclear layer; INL, inner nuclear layer; IPL, inner plexiform layer; EdU, 5-ethynyl-2′-deoxyuridine; NMDA, N-methyl-D-aspartate; ns, not significant; UMAP, Uniform Manifold Approximation and Projection.

S1P signaling is known to activate NFκB (Alvarez et al., 2010a), and NFκB is known to reciprocally regulate S1P signaling (O’Sullivan et al., 2014; Hutami et al., 2017). Accordingly, we sought to identify changes in patterns of S1pr1 expression in retinas where NFκB signaling was inhibited. We applied a small molecule NFκB inhibitor, PGJ2, which has been shown to potently block NFκB reporter activation in NMDA-damaged mouse retinas (Palazzo et al., 2022a), followed by FISH for S1pr1. We found that the number of S1pr1 FISH puncta per MG were significantly increased in NMDA-damaged retinas treated with PGJ2 (Fig. 4f,g). We next probed for changes in levels of S1P-related genes in scRNA-seq of retinas where NFκB signaling has been selectively deleted from MG. We analyzed libraries that we generated and described in a prior study (Palazzo et al., 2022a). Rlbp1-CreERT mice were crossed to mice carrying floxed alleles of Ikkb to prevent NFκB signaling in MG. Activation of canonical NFκB signaling involves the IKK complex (Ikka and Ikkb) which is activated to phosphorylate IkBα/IkBβ, thereby releasing NFκB transcription factors (p65 and p50) and allowing them to translocate into the nucleus to regulate transcription of target genes (Zhang et al., 2017). Deletion of Ikkb blocks signaling through the NFκB pathway by preventing IKK-mediated phosphorylation and degradation of IkBa/IkBb, thereby causing sequestration of NFκB transcription factors in the cytoplasm (Li et al., 2003; Zhang et al., 2017, p. 30). Consistent with patterns of expression seen in scRNA-seq libraries in WT retinas (Fig. 1), at 8 h after NMDA treatment, S1pr1, S1pr3, and Sphk1 were highly expressed by MG (Fig. S2a–f). In addition, Sgpl1 was detected in immune cells and MG (Fig.S2g), consistent with patterns of expression seen in WT retinas. We bioinformatically isolated the nearly 1,400 MG to permit a more detailed analysis. Consistent with the effects of the small molecule inhibitor, in MG where Ikkb was deleted, levels of S1pr1 were significantly upregulated, whereas levels of S1pr3 and Sphk1 were significantly reduced, and levels of Sgpl1 were unchanged (Fig. S2h). Collectively, these findings suggest that NFκB regulation of S1pr1/3 expression and downstream NFκB signaling through S1pr1/3 reciprocally influence expression levels in MG in damaged retinas.

Next, we investigated whether pharmacological activation or inhibition of S1P-related receptors and enzymes influenced ERK phosphorylation, mTor activation (pS6), and cFos expression in normal and damaged retinas. In undamaged retinas treated with exogenous S1P or S1PR1 activator, we applied antibodies to different second messengers for cell signaling pathways in the retina (Fig. S3a–d). In retinas treated with S1pr1 agonist (SEW2871), pERK1/2 levels were significantly elevated in Sox2-expressing MG (Fig. S3a,b). However, intravitreal delivery of S1P did not result in the accumulation of pERK1/2 in MG (Fig. S3b), possibly due to rapid degradation by Sgpl1. Levels of immediate early gene cFos or mTor target pS6 were not significantly different between treatment groups (Fig. S3c,d), indicating that these pathways are not affected by S1P signaling.

The accumulation of immune cells in damaged retinas is influenced by S1pr1, but not Sphk1

To further investigate how signaling through S1pr1 and synthesis of S1P via Sphk1 in MG influences the retina, we generated conditional knock-out (cKO) mice. We crossed Rlbp1-CreERT mice with mice carrying floxed alleles of S1pr1 or Sphk1. Mice were treated with vehicle or tamoxifen for 4 consecutive days, followed by 3 d to allow recombination and tamoxifen to clear. Relative to wild-type animals, S1pr1 activator had no significant effect on pERK1/2 levels in Sox2-expressing MG in mice with S1pr1 cKO (Fig. S3e,f). To evaluate gene expression, we applied FISH probes for S1pr1 and Sphk1 to retinas harvested 4 h after NMDA (Fig. S4a,b). As expected, FISH signals for S1pr1 and Sphk1 were prominent in the INL and were nearly absent in damaged retinas with MG-specific cKO of S1pr1 or Sphk1 (Fig. S4a,b).

Next, we probed for the accumulation of resident microglia and peripheral immune cells that are known to migrate into the retina after injury (White et al., 2017; Yu et al., 2020). In response to NMDA damage, peripheral monocytes infiltrate the neural retina to amplify inflammatory resident microglia signaling and inhibit Ascl1-mediated retinal regeneration (Todd et al., 2020; Blasdel et al., 2024). Eyes were injected with saline or NMDA, and retinas were harvested 2 d after NMDA (Fig. 5). Resident retinal microglia and macrophages typically express low levels of CD45 and high levels of Iba1, whereas peripheral monocyte-derived macrophages maintain steady high levels of CD45 and relatively lower levels of Iba1 in the homeostatic and injured retina (Sedgwick et al., 1991; O’Koren et al., 2016). Accordingly, we identified CD45^high/Iba1^low cells as putative infiltrating monocyte-derived macrophages (CD45⁺/Iba1⁻) and CD45^high/Iba^high or CD45^low/Iba1^high cells as microglia (CD45⁺/Iba1⁺ or CD45⁻/Iba1⁺, respectively). We found that cKO of S1pr1 or Sphk1 had no effect on numbers of immune cells in undamaged retinas (Fig. 5a–d). In WT retinas, 2 d after NMDA treatment, we found significant increases in total numbers of Iba1⁺ cells, but not CD45⁺/Iba1⁻ cells (Fig. 5a–d). Numbers of Iba1⁺ and CD45⁺/Iba1⁻ immune cells in NMDA-damaged retinas with S1pr1^−/− MG were significantly increased when compared with numbers seen in undamaged retinas with S1pr1^−/− MG and NMDA-damaged retinas with WT MG (Fig. 5a,b). By comparison, there was no significant difference in the number of CD45⁺/Iba1⁻ immune cells in NMDA-damaged retinas with Sphk1^−/− MG compared with numbers seen in NMDA-damaged retinas with WT MG (Fig. 5b–c). However, numbers of microglia in damaged retinas with Sphk1^−/− MG were significantly increased compared with numbers seen in undamaged retinas with Sphk1^−/− MG (Fig. 5b,d). In summary, the deletion of S1pr1, but not Sphk1, from MG causes the increased accumulation of Iba1-expressing microglia and CD45-expressing immune cells in acutely damaged retinas.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Immune cell accumulation in S1pr1 KO and Sphk1 KO retinas. Rlbp1-CreERT:S1pr1^fl/fl and Rlbp1-CreERT:Sphk1^fl/fl mice were intraperitoneally injected with tamoxifen or corn oil once daily for 4 d. Eyes were intravitreally injected with NMDA or saline control, and retinas were collected 48HPI. Retinal sections were labeled with antibodies to Iba1 (blue) and CD45 (green; a, b). Solid arrows indicate Iba1⁺/CD45⁺ cells, and hollow arrowheads indicate Iba1⁺/CD45⁻ cells. Scale bars: a and b, represent 50 µm. Histograms represent the mean (bar ± SD) where each dot represents one biological replicate (c, d; n = 5–18). Significance of difference (p values) was determined by using ANOVA with Sidak’s correction. Abbreviations: ONL, outer nuclear layer; INL, inner nuclear layer; IPL, inner plexiform layer; GCL, ganglion cell layer; NMDA, N-methyl-D-aspartate; ns, not significant.

S1P signaling in MG regulates neuronal survival in damaged retinas

NMDA treatment is known to selectively destroy bipolar cells, amacrine cells, and retinal ganglion cells, but not MG in the mouse retina (Todd et al., 2019). Since neuroinflammation is known to impact both neuronal survival and regeneration in NMDA-damaged retinas (Todd et al., 2019), and S1P signaling influences the NFκB pathway and the accumulation of microglia/macrophage (current study), we investigated whether S1P influences cell death and neuronal survival. In undamaged retinas with WT or S1pr1 cKO MG, we found no evidence for dying TUNEL-positive cells (Fig. 6a–c). We found significant increases in numbers of TUNEL-positive cells in the INL/ONL or GCL in NMDA retinas at 2 d after treatment, and these numbers were significantly reduced in the GCL with MG-specific cKO of S1pr1 (Fig. 6a–c). Consistent with these findings, we found significantly more surviving retinal ganglion cells (RGCs) labeled for Brn3a or HuC/D in the GCL at 2 weeks after NMDA treatment in S1pr1 cKO retinas compared with numbers seen in WT retinas (Fig. 6d–g). However, we did not find a significant difference in numbers of HuC/D+ or Pax6+ amacrine cells in the INL at 2 weeks after NMDA treatment in retinas with WT MG and retinas with S1pr1^−/− MG (Fig. 6f,h–j).

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

cKO of S1pr1 from MG is neuroprotective to retinal ganglion cells. Sections of the retina were labeled for fragmented DNA (TUNEL; green; a) and DAPI (blue), or Brn3a (green; d), tdTomato (red) and DAPI (blue), HuC/D (green; f), tdTomato (red) and DAPI (blue), or Pax6 (green, i), tdTomato (red), and DAPI (blue). Arrows indicate TUNEL⁺ nuclei (a) or Brn3a⁺ nuclei (d) or HuC/D⁺ cells (f) and small double arrows indicate GFP⁺ endothelial cells. Scale bars: a, d, and f, represent 50 µm. Histograms represent the mean (bar ± SD) where each dot represents one biological replicate (b, c, e, g, h, j; n = 5–10). Significance of difference (p values) was determined by using ANOVA with Sidak’s correction. Abbreviations: ONL, outer nuclear layer; INL, inner nuclear layer; IPL, inner plexiform layer; GCL, ganglion cell layer; NMDA, N-methyl-D-aspartate; ns, not significant; TUNEL, terminal deoxynucleotidyl transferase dUTP nick end labeling.

In retinas with cKO of Sphk1 in MG, we found significantly fewer TUNEL+ cells in the INL and ONL compared with WT retinas (Fig. 7a–c). We did not find a significant difference in numbers of RGCs labeled for Brn3a or HuC/D at 2 weeks after NMDA treatment in retinas with Sphk1^−/− MG compared with numbers seen in retinas with WT MG (Fig. 7d,e). However, we found a significant increase in numbers of HuC/D+ or Pax6+ amacrine cells in the INL at 2 weeks after NMDA treatment in retinas with Sphk1^−/− MG (Fig. 7f–i). Collectively, these findings indicate that cKO or S1pr1 or Sphk1 in MG is neuroprotective to inner retinal neurons that are damaged by NMDA.

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

cKO of Sphk1 from MG is neuroprotective to inner retinal neurons. Sections of the retina were labeled for fragmented DNA (TUNEL; green; a) and DAPI (blue), or Brn3a (green; d), tdTomato (red) and DAPI (blue), or HuC/D (green; f), tdTomato (red) and DAPI (blue), or Pax6 (green; h), tdTomato (red) and DAPI (blue). Single arrows indicate TUNEL⁺ nuclei (a) or Brn3a⁺ nuclei (d) or HuC/D⁺ cells (f), and small double arrows indicate GFP⁺ endothelial cells. Scale bars: a, d, f, and h represent 50 µm. Histograms represent the mean (bar ± SD) where each dot represents one biological replicate (b, c, e, g, i; n = 6–8). Significance of difference (p values) was determined by using ANOVA with Sidak’s correction. ONL, outer nuclear layer; INL, inner nuclear layer; IPL, inner plexiform layer; GCL, ganglion cell layer; NMDA, N-methyl-D-aspartate; ns, not significant; TUNEL, terminal deoxynucleotidyl transferase dUTP nick end labeling.

Todd and colleagues reported that microglia suppress Ascl1-mediated neurogenesis without influencing cell survival in mouse retinas (Todd et al., 2020). Additionally, we have previously reported that NFκB signaling in MG is induced by Il1α, Il1β, and TNF from reactive microglia and regulates genes responsible for neuron survival (Palazzo et al., 2022a, 2023). Accordingly, we investigated whether the neuroprotective effects of S1pr1 inhibition are dependent on immune cell proliferation and recruitment. We ablated microglia in S1pr1 KO mice using the Colony Stimulating Factor 1 Receptor (CSF1R) inhibitor PLX5622 prior to NMDA treatment and collected retinas at 2 DPI (Fig. S4). Cell death was not detected in saline-injected retinas of PLX-treated animals, but we observed significant increases in numbers of TUNEL-positive cells in the INL and GCL in NMDA-treated retinas at 2 d after treatment (Fig. S5a–c). In PLX-treated animals, numbers of TUNEL+ cells were significantly reduced in both the GCL and the INL with MG-specific cKO of S1pr1 (Fig. S5a–c). Collectively, these findings indicate that S1pr1 signaling regulates cell survival following NMDA damage, and this process operates independently of signals from immune cells.

Finally, we investigated whether pharmacological inhibition of S1pr1, Sphk1, or Sgpl1 influences cell death. Eyes were injected with vehicle, MT1303 (S1pr1 inhibitor), NIBR0213 (S1pr1 inhibitor), PF543 (Sphk1 inhibitor), FTY720 (S1pr1 modulator), or S1PLin31 (SGPL1 inhibitor) before and with NMDA, and retinas harvested 1 d after NMDA treatment. Although MT1303 had no significant effect, the NIBR0213 significantly decreased numbers of dying cells across all layers of the retina, particularly the INL (Fig. S6a–d). The solubility and pharmacodynamics of the MT1303 may have limited the efficacy of this drug (Shimano et al., 2019). In damaged retinas treated with S1pr1 modulator FTY720, cell death was reduced overall but reached significance only in the GCL (Fig. S6a–d). This was similar to S1pr1 inhibitor NIBR0213 and consistent with the notion that FTY720 functions mostly to decrease signaling through S1pr1. In retinas treated with Sphk1 inhibitor, numbers of dying cells were reduced in the INL/ONL and GCL (Fig. S6a–d). Complementary to these findings, treatment of retinas with S1PLin31 resulted in significant increases in numbers of TUNEL + cells in the INL (Fig. S6a–d). Collectively, these findings suggest inhibition of S1P synthesis or S1pr1 signaling suppresses and inhibition of S1P degradation increases cell death in acutely damaged retinas (Table 3).