Distinct mouse models of Stargardt disease display differences in pharmacological targeting of ceramides and inflammatory responses

Mice were genotyped for the Abca4-mutant alleles and screened for Rpe65^Leu/Leu homozygosity to ensure consistent rates of retinoid cycling (SI Appendix, Fig. S1). They were screened additionally for Rdh8-knockout homozygosity to ensure increased synchronized, acute light-triggered retinal degeneration (14). We also assessed whether there were any structural or morphological differences among the retinas of C57BL/6 J control mice, Abca4^−/−Rdh8^−/− mice, and Abca4^PV/PVRdh8^−/− mice from 1 to 12 mo of age, raised under a standard light–dark cycle. Several methods were used for this assessment, including histological staining (SI Appendix, Fig. S2 A and B), optical coherence tomography (OCT) (SI Appendix, Fig. S2 C and D), and scanning laser ophthalmoscopy (SLO) (SI Appendix, Fig. S2E). Only minor noticeable morphological changes were observed among the sections from the age-matched groups, suggesting that most of the photoreceptors are still present in both Abca4-knock-out and Abca4-knock-in strains at 12 mo of age. Minor changes were also observed using SLO to compare the time-dependent accumulation of autofluorescent species (SI Appendix, Fig. S2E). Scotopic and photopic electroretinography (ERG) recordings showed minor deviations in flash response between age-matched groups (SI Appendix, Fig. S3). To assess baseline visual cycle activity and visual chromophore content in the three cohorts, we performed retinoid analyses on 1-mo-old dark-adapted mice (SI Appendix, Fig. S4 A–C). Retinoids relevant to the visual cycle were also within the normal range, with only slight variability. In the three cohorts, similar levels of 11-cis-retinal, which is bound in stoichiometric ratios to visual opsins, indicate that rhodopsin levels were unaffected by the alterations in Abca4. As such, no major structural differences were observed among the different groups, so we chose to focus on molecular differences among the retinas from the respective mouse strains.

To compare time-dependent changes in auto-fluorescent species within the RPE of the Abca4-knock-out and Abca4-knock-in mice relative to WT, we employed two-photon excitation fluorescence (TPEF) recordings of the RPE from the three groups of mice, all bred with the tyrosinase^C-2J albinism mutation to minimize background fluorescence. Fig. 1A shows representative TPEF images of the RPE layer after pulsed stimulation with 730-nm and 850-nm infrared (IR) light of the intact mouse eyes taken from the albino-variants of the Abca4-knockout and Abca4-mutant mice at different ages. These images demonstrate a progressive accumulation of molecular species that fluoresce after stimulation with 850-nm light and a shift in the localization of molecular species that fluoresce when stimulated with 730-nm light. Fig. 1B shows the quantification of the fluorescence ratios (850 nm/730 nm). Clearly, these data revealed that the RPE of both types of Abca4-deficient mice have 850 nm/730 nm fluorescence ratios that increase from 1-mo to 12-mo of age. For the Abca4^−/−Rdh8^−/− mice, there was an increase in the fluorescence ratio of 0.47 ± 0.08 from the 1-mo to 12-mo cohort (P < 0.0001). For the Abca4^PV/PVRdh8^−/− mice, there was a mean increase in the fluorescence ratio of 0.40 ± 0.12 from the 1-mo cohort to 12-mo cohort (P < 0.0001). No statistically significant difference was found between the two ratios at any time point, indicating a similar degree of fluorescent-species accumulation in both types of Abca4-deficient mice. At later time points, we noticed an increase in the number of RPE cells with greater than two nuclei highlighted by orange arrows (Fig. 1 A and C). Considering these observations, we assessed whether there were any differences in the percentage of RPE cells with one nucleus, two nuclei, or ≥3 nuclei when comparing 6-mo-old and 12-mo-old mice from both models of ABCA4 deficiency (knock-out vs. knock-in). We grouped RPE cells from TPEF images based on their nuclei count and determined the mean percentages of RPE cells in the image with one nucleus, two nuclei, or ≥3 nuclei at 6- and 12-mo of age (Fig. 1 C and D). When we examined age-matched RPE cells from the two Abca4 strains (binned according to nuclear multiplicity), there was only a small, but statistically significant increase in the percentage of RPE cells with ≥3 nuclei in the 12-mo-old Abca4^PV/PVRdh8^−/− cohort when compared with the 12-mo-old Abca4^−/−Rdh8^−/− cohort.

We then compared the TPEF fluorescence properties of individual species within the RPE of 1- and 6-mo-old albino mice by performing fluorescence lifetime imaging, which examines the exponential decay of individual fluorescent species within the RPE cells (Fig. 1 E–G). Comparing the phasor plots of individual species from the RPE of the two Abca4 mouse strains (plotted in Fig. 1 E–G), a pronounced increase in species with short fluorescence lifetimes is evident at 1-mo-old and 6-mo-old timepoints for both Abca4 mutants, indicative of higher retinal condensation-product content (15, 16). This is in stark contrast to the WT mice, which only display a minor accumulation of species with slightly shorter lifetimes from 1-mo to 6-mo of age. Age-matched Abca4 mutants were also found to have similar RPE phasor-plot patterns. Furthermore, examination of the emission spectra of each cohort reveals a shift to longer-wavelength fluorescence emission, indicative of higher concentrations of retinal condensation products in the 6-mo-old cohorts of both Abca4 types compared to the C57BL/6 J-Tyr^C-2J controls (Fig. 1H).

To further examine the basis of the fluorescence changes, A2E levels were also measured by mass spectrometry of whole-eye extracts (SI Appendix, Fig. S4D). There was a significant increase in A2E concentrations in the eye homogenates from both types of Abca4 mutant mice compared to WT, with a mean increase of 15.7 ± 3.1 pmol/eye (P < 0.0001) for the Abca4^−/−Rdh8^−/− mice; and 13.6 ± 3.3 pmol/eye (P < 0.0001) for the Abca4^PV/PVRdh8^−/− mice, without a significant difference between the means for the two mouse types. A similar analysis was performed from 6-wk-old mice (n = 3) at 1, 3, and 7 d after bleaching with 10,000 lx for 30-min intense light (SI Appendix, Fig. S4E). We found a significant increase in A2E levels between 1- and 7-d post-bleach time points for the eyes from Abca4^−/−Rdh8^−/− mice, with a mean increase of 2.6 ± 0.8 pmol/eye (P = 0.005). Similarly, the analysis also indicated significant increases in A2E levels between 1- and 7-d post-bleach time points for the eyes of Abca4^PV/PVRdh8^−/− mice, with a mean increase of 4.8 ± 1.0 pmol/eye (P = 0.001). Together, these results revealed consistent elevated levels of all-trans-retinal condensation products in both types of Abca4 mutant mice and similar distribution and identity of the fluorophores in the RPE cells.

To induce degeneration, we bleached the groups of mice with 10,000 lx lamps for 30 min; then, we compared retinal morphology 24 h and 7 d after bleaching. While no significant morphological changes were observed in the retina cryosections taken 24 h after bleaching, there was pronounced thinning of the retinal ONL taken from both types of Abca4 mutant mice at 7 d post-bleach (Fig. 2A). There was a statistically significant decrease in the number of photoreceptor nuclei across all points sampled at 200-μm intervals from −2,000 μm (inferior retina) to 2,000 μm (superior retina) within the retinas of both groups of Abca4 mutant mice when compared to those from the C57BL/6 J controls (Fig. 2B).

Next, we assessed microglial accumulation in the retinas following bleaching. It is known that microglia/ macrophage accumulation occurs in bleached retinas of Abca4^−/−Rdh8^−/− mice, peaking in number at 7 d post-bleach (13, 17). We prepared retinal wholemounts from 7-d post-bleached mice and co-stained them with antibodies against Iba-1, an established microglial surface marker, and fluorophore-conjugated peanut agglutinin, which stains cone photoreceptor outer segments (SI Appendix, Fig. S5 A–C). In the bleached retinas from both types of Abca4 mice, we found an accretion of Iba1-positive microglia/macrophages as well as a morphological change from more ramified to more spherical microglial cell bodies, suggesting increased activation in response to the light-induced photoreceptor degeneration. We also assessed the production of various cytokines and chemokines in whole-eye homogenates from unbleached and 24-h post-bleach mice of different backgrounds using LegendPlex™ flow-cytometry-based immunoassays. In the whole-eye homogenates, we observed a significant increase in mean fluorescence intensity (MFI) for CCL-2 in the bleached samples from both types of Abca4 mutant mice compared to their unbleached controls and the bleached samples from WT mice (SI Appendix, Fig. S5D).

We also measured changes in galectin-3, a known modulator of microglial activation (18, 19), within neural retinas of the mice pre-bleaching and 24-h post-bleaching. While galectin-3 staining showed a diffuse pattern in paraformaldehyde-fixed retinal cryosections of each genotype of mice pre- and post-bleaching (SI Appendix, Fig. S6A), when we performed western blots on the neural retinas, we observed a significant difference in galectin-3:α-tubulin signal intensity ratios between the Abca4-knock-out and Abca4-knock-in mice. While galectin-3 levels were low in the retinas of both pre- and post-bleached WT mice, there was an elevation in galectin-3 within the retinas of both pre- and post-bleached Abca4^−/−Rdh8^−/− mice, suggesting its levels are elevated independently of light-induced retinal degeneration. In contrast to those of the Abca4^−/−Rdh8^−/− mice, the retinas of pre-bleached Abca4^PV/PVRdh8^−/− micedisplayed low levels of galectin-3 that were comparable to WT. Elevated levels of galectin-3 observed in the retinas of the Abca4^PV/PVRdh8^−/− mice post-bleaching suggests a unique, light-dependent accumulation of galectin-3 in the retinas of Abca4^PV/PVRdh8^−/− mice but not in the retinas of the Abca4^−/−Rdh8^−/− mice (Fig. 2 C and D).

Based on immunostaining, it appeared that ceramide accumulates in the retinas of Abca4^−/−Rdh8^−/− mice (20). Indeed, anti-ceramide staining of retinal cryosections of 6-mo-old unbleached mice revealed increased ceramides in the RPE of both types of Abca4 mutant mice when compared to the WT control, as indicated by the white arrows (Fig. 3A). To examine the distribution of the elevated ceramides in the RPE, we stained the RPE wholemounts of 6-mo-old mice using a pan-ceramide antibody (Fig. 3B). We found a pronounced increase in ceramide signal within the whole RPE of the Abca4 knock-in mice; however, the peripheral RPE showed a slightly weaker signal. Moreover, when looking at the higher magnification images of RPE flatmounts taken from the equatorial regions, we found the most increased ceramides in the RPE of the Abca4^PV/PVRdh8^−/− mice (Fig. 3C).

While we observed increased pan-ceramide staining in the RPE of 6-mo-old Abca4 knock-in mice, it remained unclear which specific ceramide species are elevated. Given that different ceramide species possess unique functions, we further quantified the levels of ceramide and hexCer in the retinas and RPEs using LC–MS/MS to determine the exact molecular species that were increased (SI Appendix, Fig. S7 A and D). We found that in the retinas and RPEs of 6-mo-old Abca4-mutant mice, there is a notable increase in several ceramide and hexCer species compared to WT controls. C16 (d18:1:1/16:0) and C18 (d18:1/18:0) ceramides were the most elevated species in the retinas of both types of Abca4-mutant mice. In the retinas of Abca4^−/−Rdh8^−/− mice, C16 ceramides were increased by 248% (P < 0.0001), and C18 ceramides were increased by 162% (P < 0.0001). Within the retinas of the Abca4^PV/PVRdh8^−/− mice, C16 ceramides exhibited a 245% increase (P < 0.0001), and C18 ceramides showed a 285% increase (P < 0.0001). Notably, there was a 1.8-fold greater elevation in C18 ceramides within the retinas of the Abca4^PV/PVRdh8^−/− mice compared to Abca4^−/−Rdh8^−/− mice (P < 0.0001). HexCer d18:1/24:1 was the most notably enriched species within the RPE of the two Abca4 mutants relative to the WT control, with increases of 410% (Abca4-knock-out; P < 0.0001) and 459% (Abca4-knock-in; P < 0.0001). We also observed significant elevations in hexCers d18:1/18:0 and d18:1/24:0 within the RPEs of the Abca4-mutant mice compared to the WT controls. Their increases were 182% (P = 0.024) and 159% (P = 0.004), respectively, in Abca4^−/−Rdh8^−/− mice; and 199% (P = 0.023) and 163% (P = 0.007), respectively, in Abca4^PV/PVRdh8^−/− mice. Clearly elevated C18-ceramides in Abca4^PV/PVRdh8^−/− mice compared to Abca4^−/−Rdh8^−/−mice could potentially be linked to increased inflammatory events and added stress stemming from the dysfunctional, mutated Abca4 protein, paralleling the pathology seen in Alzheimer’s patients (21).

Additionally, we analyzed ceramide and hexCer levels in the retina and RPE taken from 6-wk-old mice in response to light damage-bleaching at 10,000 lx for 30 min. Tissues were taken before bleaching and 1, 3, and 7 d after bleaching, and ceramide and hexCer levels were assessed using LC–MS/MS (n ≥ 3) (SI Appendix, Fig. S7 B, C, E, and F and Table S1). The most dramatic increase in ceramide levels occurred between the pre-bleach and 1-d post-bleach time points before plateauing out at the 7-d post-bleach time points. At the 7-d post-bleach time point, there were statistically significant increases in bulk ceramide and hexCer content within the retinas and RPEs of each genotype when compared to genotype-matched, pre-bleached (d0) levels, with the exception of hexCer levels in the RPE of the Abca4^−/−Rdh8^−/− mice.

When comparing the time point- and genotype-matched groups of mice at 7 d after bleaching, there were no significant differences in ceramide levels in the retina, while there were significant increases in retinal hexCer levels in the Abca4^−/−Rdh8^−/− mice when compared to the time point-matched WT controls (mean increase: 58%, P = 0.042). At the 7-d post-bleach time point, there were also statistically significant increases in RPE levels of ceramide (mean increase: 54%, P = 0.047) and hexCer (mean increase: 84%, P = 0.008) in the Abca4^PV/PVRdh8^−/− mice when compared to the time point-matched WT controls. The elevated levels of hexCer, in the absence of a corresponding increase in ceramides, might indicate a cell strategy to diminish the apoptotic potential of ceramides.

Following documentation of the increases in ceramide and hexCer levels in the retinas and RPE of the Abca4-knock-out and Abca4-knock-in mice, we tested whether we could pharmacologically inhibit the accumulation of these metabolites by treating both Abca4 mutants intraperitoneally with AdipoRon, a known agonist of adiponectin receptors 1 (AdipoR1) and 2 (AdipoR2) (22). Both groups of Abca4 mutants were treated daily for 2 wk before they were killed, and retinas and RPEs were taken for mass spectrometric analysis of ceramide and hexCer levels (Fig. 4). We found that AdipoRon treatment selectively decreased ceramide and hexCer accumulation in the Abca4^PV/PVRdh8^−/− mice; retinal levels were decreased by 64% (P = 0.014) and RPE levels by 29% (P = 0.047). In contrast, AdipoRon treatment had no significant effect on ceramide and hexCer levels in the retinas of the Abca4^−/−Rdh8^−/− mice. This implies inherent genetic or biochemical differences between these two strains. It also suggests that the pathway influenced by AdipoRon in the context of ceramide and hexCer metabolism might be active in the Abca4^PV/PVRdh8^−/− strain but not in the Abca4^−/−Rdh8^−/− strain. The mouse retina has a pronounced expression of ADIPOR1, while ADIPOR2 is almost nonexistent (23). We hypothesize that ceramidase activity is not stimulated by AdipoRon in Abca4^-/-Rdh8^-/- mice, but the basis for this lack of effect warrants further investigation.

Prophylactic treatment with the BMT combination is known to salvage the photoreceptors of bleached Abca4^−/−Rdh8^−/− mice (24, 25); here, we assessed whether the photoreceptors of the Abca4^PV/PVRdh8^−/− mice would also be salvaged by prophylactic BMT administration. Representative OCTs and SLO images of the eyes of the Abca4^mutants treated with vehicle or BMT demonstrate drug-dependent salvage of the ONL (increased thickness vs. vehicle-treatment) in the Abca4-mutant mice at 7 d after bleaching (Fig. 5 A–D). Representative SLOs (Fig. 5B) highlight significant reductions in punctate, auto-fluorescent spots in both groups of Abca4 mutants that were treated with BMT.

We also assessed the effects of the Cysteine-Cysteine Chemokine Receptor 5 (CCR5)-specific antagonist maraviroc to see whether we could obtain similar salvaging of the photoreceptor layer in Abca4 mutant mice by controlling microglia/macrophage-mediated immune responses to light-induced retinal degeneration (Fig. 5 E–H). Representative OCTs of the retinas of mice treated with vehicle or maraviroc demonstrate a drug-dependent rescuing of ONL thickness in both types of Abca4 mutants. Quantification of the ONL thickness of vehicle-treated vs. maraviroc-treated mice (Fig. 5 G and H) revealed statistically significant, maraviroc-dependent salvaging of ONL thickness 7 d post-bleaching. In addition, representative SLOs (Fig. 5F) demonstrate significant reductions in punctate, autofluorescent spots within the retinas of both types of Abca4 mutant mice following prophylactic administration of maraviroc.

To further investigate changes in visual processing, we performed primary visual cortex recordings (V1) on both types of Abca4 mutant mice to compare visual evoked potentials (VEPs) of unbleached mice, 7- to 12-d post-bleached and vehicle-treated mice, and 7- to 12-d post-bleached and maraviroc-treated mice. Thus, six animal groups were compared by choosing the most robust response to the flash stimulus from the cortical profile for each group (SI Appendix, Fig. S8A), and standardizing the responses using z-scores. Representative examples of the responses for Abca4^PV/PVRdh8^−/− and Abca4^−/−Rdh8^−/− mice are provided in SI Appendix, Fig. S8 B and C. The population analysis revealed that the unbleached Abca4^−/−Rdh8^−/− mice (2.74 ± 0.39 µV) and the bleached Abca4^−/−Rdh8^−/− mice (2.70 ± 0.53 µV; P = 0.660) displayed the most robust responses to the flash stimulus (SI Appendix, Fig. S8 D and E). The VEP amplitudes dropped after treatment (1.55 ± 0.27 µV) and were significantly different only from the unbleached group (P = 0.011, unbleached; and P = 0.150, bleached + DMSO). The Abca4^PV/PVRdh8^-/- unbleached animals had a lower but not significantly different response from that of the Abca4^−/−Rdh8^−/− unbleached group (1.91 ± 0.2 µV vs. 2.74 ± 0.39 µV; P = 0.140). Similarly, to the previous group, responses in Abca4^PV/PVRdh8^−/− unbleached animals were not significantly different from Abca4^PV/PVRdh8^−/− bleached + DMSO animals (1.91 ± 0.2 µV vs. 1.49 ± 0.2 µV; P = 0.076). The responses of Abca4^PV/PVRdh8^−/− mice that were bleached and maraviroc-treated were significantly lower than those of the un-bleached group (1.91 ± 0.2 µV vs. 1.32 ± 0.19 µV; P = 0.024), but there were not significant differences between the bleached + vehicle (DMSO) and the maraviroc-treated groups (1.49 ± 0.2 µV vs. 1.32 ± 0.19 µV; P = 0.076) (SI Appendix, Fig. S8E). These data suggest that even a residual number of photoreceptors could produce sufficient visual cortex responses. It should be noted that the visual pathway becomes hyperexcited during degeneration, so even residual rods and cones can produce significant VEPs (26).

To elucidate subtle changes in the retina on the mRNA level, single-cell transcriptomic analyses were performed on cohorts of WT mice and on both Abca4 mutants under three treatment paradigms: pre-bleached, 24-h post-bleach and vehicle-treated, and 24-h post-bleach and maraviroc-treated. A total of 69,962 barcoded single cells were successfully sequenced from 27 pairs of neural retinas (SI Appendix, Table S2). Single-cell transcriptomic datasets were compiled and submitted for public accession to the GEO database (accession number: GSE239347) (27). Following sequencing and refinement of the single-cell dataset, we generated a UMAP plot using single cells consolidated from all samples (Fig. 6A). Next, we identified 34 unique cell clusters and labeled them according to an established set of transcriptomic biomarkers for neural-retina cell types (SI Appendix, Fig. S9A). After separating single cells based on genotype and treatment group, we observed significant shifts in rod, cone, and Müller glia clustering within the retinas of the vehicle-treated and bleached Abca4 mutant mice (Fig. 6B and SI Appendix, Fig. S9B). Notably, the retinas of vehicle-treated and bleached Abca4^PV/PVRdh8^−/− mice displayed a broader transition of rod, cone, and Müller glia cells toward degeneration-associated clusters, in contrast to the retinas of the vehicle-treated and bleached Abca4^−/−Rdh8^−/− mice (Fig. 6B).

The rod, cone, and Müller glia clusters that were enriched in the retinas of vehicle-treated and bleached Abca4 mice were annotated as degeneration-associated (clusters 01, 08, and 05, respectively), and the combined degeneration-associated rod, cone, and Müller glia clusters from all samples were compared to the respective healthy clusters that were enriched in the retinas of unbleached Abca4 mutant mice and WT controls (Fig. 6 C–E). When comparing differentially expressed genes (DEGs) between the healthy (cluster 00) and degeneration-associated rods (cluster 01), 81 genes were significantly upregulated and 12 genes were significantly downregulated in the degeneration-associated cluster, while 5,627 genes did not show changes in expression between the groups (Fig. 6C). When comparing DEGs between the healthy (cluster 02) and degeneration-associated cones (cluster 08), 176 genes were significantly upregulated, and 135 genes were significantly downregulated in the degeneration-associated cluster, while 8,360 genes were not significantly altered between the groups (Fig. 6D). When comparing DEGs between the healthy (cluster 03) and degeneration-associated Müller glia (cluster 05), 228 genes were significantly upregulated, 272 genes were significantly downregulated in the degeneration-associated cluster, and 10,934 genes were not significantly changed between the groups (Fig. 6E). We defined significant DEGs as having a P < 0.05 and |Log₂ Fold Change| > 0.6.

We then interrogated differences between the transcriptional phenotypes of rods, cones, and Müller glia separated by the genotype and treatment condition. Comparing unbleached rod, cone, and Müller glia phenotypes, there were fewer significant DEGs in the cells from Abca4^−/−Rdh8^−/− mice than those from the Abca4^PV/PVRdh8^−/− mice when compared to the corresponding cells from the WT unbleached control mice. This pattern was consistent across the bleached and vehicle-treated and bleached and maraviroc-treated conditions (SI Appendix, Figs. S10 and S11 and Table S3). Notably, maraviroc treatment before bleaching resulted in a marked decrease in the number of significant DEGs in the two Abca4 mutants when compared to the corresponding vehicle-treated cohorts, with the most pronounced reduction in DEGs observed in the samples from Abca4^PV/PVRdh8^−/− mice. These findings suggest that Abca4^PV/PVRdh8^−/− mice have either a different degree or a different rate of early transcriptional responses to light-induced retinal degeneration when compared to the matched Abca4^−/−Rdh8^−/− mice. The Abca4^PV/PVRdh8^−/− mice also displayed the most prominent response to prophylactic maraviroc treatment, indicating a larger role of CCR5-based immune signaling in dictating early responses to retinal degeneration.

Not surprisingly, based on prior work, no gross differences between Abca4-knock-out and Abca4-knock-in mice were observed when examined under identical conditions. The structure of the retina was indistinguishable, and accumulation of all-trans-retinal condensation products was similarly distributed within the RPE cells. Notably, we observed a slight increase in the number of RPE cells with 3 or more nuclei in 12-mo-old Abca4^PV/PVRdh8^−/− mice when compared to the age-matched Abca4^−/−Rdh8^−/− mice; these multinucleated RPE cells have been documented previously as forming in response to instances of elevated cellular stress, aging, and retinal degeneration; they are thought to represent a component of the homeostatic response to maintain RPE integrity (28–30). The number of these multinucleated cells, however, is too few to have a large impact on the overall sizes of the RPE cells. RPE cell multinucleation has been linked previously with cell aging and disease states such as age-related macular degeneration (28, 30), suggesting that RPE cell health in the aged Abca4^PV/PVRdh8^−/− mice might be further compromised when compared to age-matched Abca4^−/−Rdh8^−/− mice.

Immune-cell crosstalk, morphology, and cytokine/chemokine levels in a tissue change in response to retinal degeneration and provide a means of obtaining information about the contributions of inflammation and immune-cell-derived phagocytosis to a retinal-degenerative phenotype (31, 32). At the terminal, 7 d post-bleach time points, we found that the microglia of both mouse models exhibited significant changes in morphology in response to light, becoming more elliptical and less ramified, an established hallmark of microglial activation in the central nervous system (33, 34). Looking at earlier, 24-h post-bleach time points, there were also elevations in Ccl2 and a trend toward an increase in Ccl4 in both Abca4 mutant mice, with Ccl4 MFIs being significantly elevated in the Abca4^−/−Rdh8^−/− mice. This is in accordance with the previous quantifications of cytokine mRNA levels in the Abca4^−/−Rdh8^−/− mice, as presented in Kohno et al. (17). Though cytokine profiling and wholemount immunofluorescence staining would imply highly similar immune responses to light-induced retinal degeneration, we observed a notable difference in galectin-3 levels pre-bleaching between the two models, with the Abca4^−/−Rdh8^−/− mice seeming to possess a consistent, high level of baseline galectin-3 expression in their retinas. This observation differs considerably from that for the Abca4^PV/PVRdh8^−/− mice, where all but one of the pre-bleached samples had much less galectin-3 signal than those from the Abca4^−/−Rdh8^−/−mice. Galectin-3 was previously characterized as a ligand of toll-like receptor-4 (TLR-4), which, upon activation, initiates pro-inflammatory gene transcription and promotes phagocytosis in microglia (18, 35, 36). Conditional deletion of galectin-3 in microglia led to photoreceptor death, RPE damage, and vision loss, suggestive of its protective role (37). TLR-4 ablation before bleaching was previously shown to protect photoreceptors in Abca4^−/−Rdh8^−/− mice (17); in contrast, another study found a paradoxical effect of photoreceptor protection from apoptosis when TLR-4 was activated pre- and post-exposure to oxidative stress (38). This study found that TLR4 pre-activation prior to exposure to oxidative stress primed photoreceptors, making them more resilient to oxidative stress-induced apoptosis. This finding could relate to the higher level of galectin-3 protein in the retinal homogenates from the unbleached Abca4^−/−Rdh8^−/−mice, suggesting that their photoreceptors might be specifically primed for resilience to light-induced degeneration. In contrast, the lack of consistent elevations in galectin-3 expression within the retinas of the Abca4^PV/PVRdh8^−/− mice suggests a lack of tissue priming for resilience to light-induced degeneration.

Previous studies have highlighted the role of ceramides as mediators of inflammation and apoptotic programs within the retina, and studies of specific blockade of ceramide production in murine models of retinitis pigmentosa and ADIPOR1 deficiency have demonstrated retinal-protective effects (23, 39, 40). Broader disruptions in sphingolipid metabolism and ceramide production can have widespread effects on pro-inflammatory gene expression, plasma and organelle membrane integrity, and lipid metabolism (41, 42). There have also been recent observations that Abca4 deficiency leads to widespread changes in lipid profiles within the retina and RPE (43). We initially thought that if there were indeed greater underlying metabolic stress within the Abca4^PV/PVRdh8^−/− mutant photoreceptors, there would be a significantly higher accumulation of ceramides within the Abca4^PV/PVRdh8^−/− retinas and RPE at baseline without light-induced degeneration or pharmacological interventions. There was a significantly higher amount of the most abundant retina-derived ceramide (d18:1/18:0) in retinal homogenates from the Abca4^PV/PVRdh8^−/− mice when compared to those from the Abca4^−/−Rdh8^−/− mice. This specific perturbation in a single ceramide species could imply a shift in cellular fatty acid content and access to metabolites used to specifically synthesize ceramide (d18:1/18:0). Other than significant differences in ceramide (d18:1/18:0) content, there were no other notable differences in ceramide or hexCer species between the two types of Abca4-mutant mice in the unbleached condition. There were, however, significant elevations in several ceramide and hexCer species within both types of Abca4-mutant mice compared to the matched C57BL/6 J controls. These findings suggested little difference in overall ceramide content between the two types of Abca4-mutant mice. There were no consistent differences in bulk ceramide or hexCer content in the retinas and RPE of the Abca4-knock-in vs. the Abca4-knock-out mice at 1, 3, or 7 d after bleaching.

Despite this overall lack of differences in ceramide and hexCer content, we tested whether both types of Abca4 mutant mice were equally responsive to 2 wk of treatment with AdipoRon and would exhibit reduced ceramide and hexCer levels following treatment. Notably, the Abca4^PV/PVRdh8^−/− knock-in mice demonstrated a marked reduction of ceramide levels with AdipoRon, a specific ADIPOR1 and ADIPOR2 agonist, while the Abca4^−/−Rdh8^−/− knock-out mice did not demonstrate any reduction in ceramide levels. This trend was consistent across many individual ceramide and hexCer species in both the retinas and the RPE and across bulk ceramide and hexCer levels. These findings suggest that the Abca4^PV/PVRdh8^−/− mice are uniquely sensitized to respond to systemic adiponectin receptor activation and subsequently display a reduction in bulk ceramide and hexCer levels within the retina and RPE. Furthermore, light-induced stress might intensify the pre-existing stress in the RPE of Abca4^PV/PVRdh8^−/− mice, as observed on the RPE flatmount, which displays pathologically distorted RPE cells (Fig. 3 C, Left; F-actin staining). We later investigated transcriptomic differences between the Abca4 knock-in and knock-out models and identified differential gene expression between the two models that could explain the difference in responsiveness to AdipoRon treatment.

Our study demonstrates significant, differential responses to pharmacological targeting of baseline ceramide accumulation and prophylactic targeting of immune cell crosstalk within the two mutationally distinct models of ABCA4 deficiency. We hypothesized that the catalytically inactive Abca4^PV/PVRdh8^−/− knock-in mice would demonstrate a more severe retinal-degenerative phenotype when compared to the knock-out Abca4^−/−Rdh8^−/− mice, as a result of the increased metabolic burden from producing and subsequently degrading mutated Abca4 transcripts. This proved to be true in our transcriptomic comparisons of the two mice strains pre- and post-bleach and their susceptibility to pharmacological interventions that target CCR5-specific immunological signaling and ceramide accumulation. Though initial histological, biochemical, and electrophysiological phenotyping yielded few noticeable phenotypic and morphological differences between the two models pre- and post-light-induced degeneration, subtle phenotypic differences began to emerge in the comparison of RPE multinucleation and two-photon fluorescence lifetime imaging. At 12 mo of age, Abca4^PV/PVRdh8^−/− mice raised under a standard light–dark cycle demonstrated a significantly higher percentage of RPE cells with 3 or more nuclei when compared to age-matched Abca4^−/−Rdh8^−/− mice.

Because we observed differences in baseline galectin-3 expression, we tested whether the two Abca4 mouse models would respond differently to another immunomodulatory drug, maraviroc, an established antagonist of CCR5 shown to promote recovery from ischemic stroke and hemorrhaging in rodent models, via modulation of microglial and astrocyte activation and inflammatory gene transcription (44). Additionally, we investigated the potential retina-protective effects of prophylactic administration of maraviroc prior to bleaching, relative to the effects of a retinal-protective, monoaminergic drug cocktail (BMT) previously characterized by Leinonen et al. (45). Indeed, we saw a robust recovery of ONL thickness in both cohorts of bleached and maraviroc-treated Abca4 mice. Interestingly, there seemed to be a more consistent, significant retinal-protective effect of CCR5 blockade between the two groups in comparison to treatment with the BMT cocktail, suggesting another potential difference in the signaling pathways that dictate apoptotic/necrotic responses to light-induced retinal degeneration between the Abca4-knock-out and Abca4-knock-in mice. Encouraged by these observations, we proceeded to make a comprehensive, single-cell RNA transcriptomic comparison between the Abca4 mutant mice and WT controls under three independent treatment conditions to elucidate whether there are underlying transcriptional differences between the two Abca4 types that could inform differential responses to therapeutics.

In our single-cell RNA sequencing, which analyzed transcriptomic changes in retinal cell types pre-bleach, 24-h post-bleach and vehicle-treated, and 24-h post-bleach and maraviroc-treated, notable differences emerged when we compared the two Abca4 models to each other and to the WT controls in all three treatment conditions. There were also more DEGs between the WT and Abca4^PV/PVRdh8^−/− mice across the rods, cones, and Müller glia compared to the corresponding number of DEGs between the WT and Abca4^−/−Rdh8^−/− mice. This hints at a higher degree of overall transcriptional deviation from the WT cohort in the Abca4^PV/PVRdh8^−/− cohort, suggesting a more pronounced difference in baseline cellular function within the Abca4^PV/PVRdh8^−/− cells.

Specifically, cones from both Abca4-mouse types displayed increased expression of genes associated with cytoskeletal remodeling, cyclic-AMP-based signaling, cell–cell adhesion, Egr1 signaling, and Fos signaling (46–49). In the cones of the unbleached Abca4^−/−Rdh8^−/−mice, several genes were uniquely enriched that have been linked with Stat1 signaling, modulation of serine/threonine kinase, regulation of apoptosis, and GTPase-signaling pathways when compared to the cones from unbleached Abca4^PV/PVRdh8^−/− mice (50–52). Notably, in the cones from the Abca4^−/−Rdh8^−/− mice relative to those from the Abca4^PV/PVRdh8^−/− mice, there was a significant upregulation of Lpcat2, which is known to co-localize with TLR-4 and promote pro-inflammatory gene expression in macrophages (53). In the cones from unbleached Abca4^PV/PVRdh8^−/− mice compared to those from the unbleached Abca4^−/−Rdh8^−/− mice, there were increases in the expression of one of the interphotoreceptor matrix proteoglycans (Impg2); Abca4; Ppp1r14c, a regulator of C/EPB homologous protein transcription factor activity; and a modulator of TLR activation and response to inflammatory stimuli (54–56). Comparing the rods of the two unbleached Abca4 strains, there looked to be a unique upregulation of an intracellular lipid receptor and a mitochondrial gene within the rods from the Abca4^−/−Rdh8^−/− mice (57). The rods from the unbleached Abca4^PV/PVRdh8^−/− mice displayed enriched Galnt13 transcript levels relative to those from both the WT and Abca4^−/−Rdh8^−/− mice, suggesting increased levels of O-linked glycosylation on serine/ threonine residues within the rods of unbleached Abca4^PV/PVRdh8^−/− mice (58).

There were even larger numbers of DEGs when the Müller glia sampled from both Abca4 cohorts were compared with Abca4^−/−Rdh8^−/−-specific upregulation of genes involved in phosphatidylinositol metabolism/signaling, receptor tyrosine kinase signaling, and maintenance of cell–cell adhesions/cell junctions, as well as a significant elevation in Stat3 transcript (59, 60). Upregulated genes were also involved in Egr2 and Egr3 signaling, mTOR signal inhibition, and NF-κB-dependent signaling pathways (61). Knockdown of Stat3 was previously shown to inhibit Wnt3a-dependent protection from oxidative stress in RPE cells, and increases in Stat3 expression were shown to improve mutant photoreceptor survival in models of retinal degeneration (60, 62). There was a statistically significant upregulation of Nkd1, a negative regulator of Wnt signaling upon dysregulation in the Müller glia of the unbleached Abca4^PV/PVRdh8^−/− mice, suggesting that Stat3 and Wnt-β-catenin signaling axes may have a role in ensuring Müller glia cell resilience in models of ABCA4 deficiency (63, 64). There were also significant enrichments in other cadherins and synaptic maintenance-associated genes in the Müller glia from the Abca4^PV/PVRdh8^−/− mice when compared to the matched Müller glia from the Abca4^−/−Rdh8^−/− mice (65). This enrichment implies that there is a baseline prioritization of modulating synaptic connections and neurotransmitter release within the Müller glia of the Abca4^PV/PVRdh8^−/− mice, rather than a prioritization of priming the microenvironment for resistance to oxidative stress in the case of the Müller glia from the Abca4^−/−Rdh8^−/− mice.

When we transitioned to comparing the rods, cones, and Müller glia from the 24-h post-bleach vehicle (DMSO)-treated cohorts, there were even larger differences in the transcriptional landscapes of the Abca4 cohorts. When comparing DEGs in the cones from the bleached Abca4 mice of both strains, there was significantly higher expression of many visual cycle and phototransduction-associated genes within those from the Abca4^−/−Rdh8^−/− mice compared to those from the Abca4^PV/PVRdh8^−/− mice. The cones from the Abca4^−/−Rdh8^−/− mice also displayed unique enrichments in many solute carrier family protein transcripts and those associated with synaptic maintenance (66, 67). There were also some upregulated genes involved in the maintenance of cell–cell adhesion and positive regulators of Wnt-β-catenin signaling, consistent with the apparent enrichment in Wnt-β-catenin signaling in the cones from the unbleached Abca4^−/−Rdh8^−/−mice (68–70). The cones from Abca4^PV/PVRdh8^−/− mice exhibited higher expression of specific receptor-tyrosine-kinase and tyrosine-phosphatase genes, a GABA receptor subunit, chloride channels, ubiquitin E3 ligases, and Rock2 signaling genes (71–74). These transcriptional changes suggest homeostatic responses to suppress degeneration-associated hyperexcitability of the cones from the Abca4^PV/PVRdh8^−/− mice. This pattern of visual cycle- and phototransduction-gene downregulation was carried over when the transcriptional levels of the rods from the bleached and vehicle-treated Abca4^PV/PVRdh8^−/− mice were compared to those from the like-treated Abca4^−/−Rdh8^−/−mice. Additionally, there was an upregulation in genes linked to GABA/ glutamate homeostasis (75). There was a significant increase in ceramide kinase-like (Cerkl) gene expression within the rods from the Abca4^PV/PVRdh8^−/− mice, consistent with a previously characterized increase in ceramide-associated autophagy via stabilization of SIRT1 (76). There were also noteworthy increases in Stat1 and Stat2 expression within the rods from the Abca4^PV/PVRdh8^−/−mice, which could be linked to differential sensitivity to interferon-based signaling pathways, which can act in concert with PI3K and MAPK signaling pathways to drive interferon-stimulated gene expression (77).

Comparison of the Müller glia from the bleached and vehicle-treated cohorts showed that those from the Abca4^−/−Rdh8^−/− mice had significant upregulations in genes involved in lipid transport, glutamate metabolism, synaptic maintenance, axonal projection, and neurogenesis (78). In contrast, the Müller glia from Abca4^PV/PVRdh8^−/− mice displayed higher expression of some adhesion molecules such as vimentin (Vim), endothelin signaling receptor Ednrb, as well as several EIF-complex genes, and ERK signaling cascade genes (79). These observations suggest that the Müller glia of Abca4^PV/PVRdh8^−/− mice display higher levels of cellular stress and participate in extracellular matrix and synaptic remodeling, while mounting responses to suppress angiogenesis and facilitate repair of the inner retinal vasculature.

All mice were housed in the vivarium at the University of California, Irvine, and were maintained on a normal mouse chow diet and a 12-h/12-h light (<150 lx)/dark cycle. Male and female 1- to 12-mo-old mice were used in this study. The age of animals for each set of data is indicated in the individual figures. All animal procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of the University of California, Irvine, and were conducted in accordance with the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research.

Z.J.E., D.L., A.T.F., and K.P. designed research; Z.J.E., D.L., Z.D., G.P., J.Z., K.K., J.P., D.P., A.T.F., and M.T. performed research; Z.D. contributed new reagents/analytic tools; Z.J.E., D.L., Z.D., G.P., J.Z., J.P., D.P., A.T.F., and M.T. analyzed data; and Z.J.E., D.L., and K.P. wrote the paper.

K.P. is a consultant for Polgenix, Inc. and serves on the Scientific Advisory Board at Hyperion Eye, Ltd. All other authors have declared that no conflict of interest exists.