Ceramide compensation by ceramide synthases preserves retinal function and structure in a retinal dystrophy mouse model

Ceramide, a class of sphingolipids abundant in cell membranes, is a bioactive lipid that functions as an important cellular second messenger that signals for apoptosis (Hannun, 1996; Hannun and Obeid, 2002; Pettus et al., 2002). As the central molecule of the highly interconnected sphingolipid metabolic pathways, ceramide forms the backbone of all complex sphingolipids and is an essential player in key cellular functions, including the regulation of cell-stress responses (Scarlatti et al., 2004; Hannun and Obeid, 2008), the modulation of membrane dynamics (Holopainen et al., 2000; Stancevic and Kolesnick, 2010) and the activation of cellular pathways by acting as a second messenger (Chalfant et al., 1999; Ruvolo, 2003; Wang et al., 2005; Canals et al., 2018). In mammals, de novo ceramide synthesis is catalyzed by six ceramide synthases (CerSs): CerS1-6 (Mullen et al., 2012). Each CerS regulates the formation of a specific set of ceramides with variable chain lengths and mediates distinct functions (Mizutani et al., 2005; Imgrund et al., 2009; Levy and Futerman, 2010; Ginkel et al., 2012; Jennemann et al., 2012; Ebel et al., 2014). Specifically, CerS1 synthesizes mainly C18 ceramide, CerS2 synthesizes mainly C22-C24 ceramide, CerS3 synthesizes very long chain ceramides (>C26), CerS4 synthesizes C18-C20 ceramide, and CerS5 and CerS6 synthesize mostly C16 ceramide (Ben-David et al., 2011). Consequently, there is redundancy in the ceramides synthesized by each synthase, but no CerS can completely compensate for the function of another. CerSs demonstrate tissue-specific expressions, which lead to ceramide profile variations among different tissues (Levy and Futerman, 2010). Ceramides with different chain lengths have various biological functions, and ceramide imbalance may trigger stress responses or cause diseases. Cers1-deficient mice exhibited cerebellar shrinkage and Purkinje cell loss (Ginkel et al., 2012), whereas CerS2 was essential for the production of myelin by oligodendrocytes (Ben-David et al., 2011). The production of very-long-chain (>C26) ceramides by CerS3 appears to be essential in the maintenance of a functional epidermal barrier (Mizutani et al., 2008, 2009). Cers4-deficient mice, however, suffer from progressive hair loss (Ebel et al., 2014). As a result of the interplay among CerSs, the homeostasis of different ceramide species also has a profound impact. In Cers1-deficient mice, despite the overall ceramide level reduction, the dramatic decrease in C18 ceramides triggered the overproduction of C16 ceramides, which failed to compensate for the damage caused by C18 deficiency (Zhao et al., 2011). In Cers2-deficient mice, C24:0 and C24:1 ceramides declined in association with a compensatory increase in C16 and C18 ceramides. However, the total level of ceramides was unaltered, suggesting the distinct roles played by ceramide species with variable chain lengths (Ben-David et al., 2011).

Increasing evidence supports the role of ceramide as a mediator of photoreceptor death, which is integral to major inherited retinal diseases including retinitis pigmentosa, Stargardt's disease, Leber's congenital amaurosis and age-related macular degeneration (Glazer and Dryja, 2002; Allikmets, 2004; Haddad et al., 2006; Hartong et al., 2006). On the one hand, retinal cell death or retinal impairment accompanied by ceramide accumulation has been observed both in vivo (Acharya et al., 2003, 2008; Ranty et al., 2009; Strettoi et al., 2010) and in vitro (German et al., 2006; Sanvicens and Cotter, 2006; Chen et al., 2012). On the other hand, ceramide deficiency has also recently been associated with retinal dysfunction and degeneration, as evidenced by the deleterious effects on the retina owing to ceramide depletion caused by CerS deletion (Brüggen et al., 2016; Bertrand et al., 2021; Qian et al., 2022). Retinal functions were previously examined in Cers1-, Cers2- and Cers4-deficient mice (Brüggen et al., 2016). Accompanied by the C20-C24 ceramide increase and the C16 ceramide reduction, retinal dysfunction was detected by electroretinograms (ERGs), but no significant morphological defects were detected in these three mouse models (Brüggen et al., 2016).

In addition to the canonical CerSs, TLC domain-containing protein 3B (TLCD3B) is also found to have CerS activity (Yamashita-Sugahara et al., 2013). Relative to the other six canonical CerSs in the retina, TLCD3B has a much higher expression level (Bertrand et al., 2021). Patients diagnosed with cone-rod dystrophy or maculopathy were found to have TLCD3B mutations, and their phenotypes were restricted to the retina (Bertrand et al., 2021). Tlcd3b^−/− mice demonstrated both retinal dysfunction and degeneration by 7 months of age (Bertrand et al., 2021). TLCD3B has a distinct ceramide profile consisting of C16, C18 and C20 ceramides, which do not fully overlap with any of the six canonical CerSs (Yamashita-Sugahara et al., 2013; Qian et al., 2022). Although these results further support the importance of maintaining ceramide homeostasis in the retina, mechanistic explanations of the observed retinal morphological defects caused by TLCD3B disruption are still lacking. Given that previous canonical CerS-deficient mouse models failed to display retinal degeneration (Brüggen et al., 2016), an interesting question was brought up regarding the relationship between ceramide homeostasis and retinal cell death in the context of TLCD3B deficiency: is TLCD3B deficiency-associated retinal cell death caused by an absolute change in ceramide level, or does a healthy retina rely more on the profile maintenance of different ceramide species with variable chain lengths?

In a previous study (Qian et al., 2022), we performed gene therapy on Tlcd3b^−/− mice and successfully rescued the retinal degeneration phenotypes by subretinally injecting a recombinant adeno-associated virus 8 (rAAV8) vector containing the Tlcd3b gene (rAAV8-Tlcd3b). Based on targeted metabolomics analyses, Tlcd3b^−/− mice had a significant reduction of C16, C18 and C20 ceramides and the gene therapy successfully recovered the levels of these ceramide species, providing the first in vivo evidence that TLCD3B plays a role in ceramide production. To further explore the roles of TLCD3B and its interaction with canonical CerSs, we examined whether ceramide compensation in Tlcd3b^−/− mice could rescue photoreceptor degeneration by performing subretinal injections of three adeno-associated-virus (AAV)-mediated canonical CerS genes (rAAV8-CerS2, rAAV8-CerS4 and rAAV8-CerS5). To achieve high transduction efficiency, more rapid transgene expression, cell transduction specificity, and stable and long-term expression in photoreceptor cells, a tyrosine-capsid mutant AAV plasmid, AAV8 (Y733F) was used (Petrs-Silva et al., 2009; Ku et al., 2011; Pang et al., 2011; Zhong et al., 2015; Zaneveld et al., 2019). Although treatment with all three rAAV8-CerS vectors was able to restore retinal functions as indicated by improved ERG responses, only rAAV8-CerS5 treatment successfully retained retinal morphology in Tlcd3b^−/− mice. Additionally, mass spectrometry data and targeted metabolomics analyses suggested that rAAV8-CerS5 treatment best recovered key ceramide species that are largely reduced in Tlcd3b^−/− mice (C16:0, C18:0 and C20:0 ceramides) compared to the ceramide profiles of the other two viruses, indicating that different ceramide species may play different roles in retinal homeostasis. In summary, our results suggest that ceramide is a primary activator of retinal degeneration in Tlcd3b^−/− mice. Moreover, rather than acting together, different ceramide species have different impacts on retinal cells, making maintenance of the overall ceramide profile essential to retinal cell functions.

Given that the primary defect caused by loss of TLCD3B is mainly photoreceptor degeneration, we cloned mouse CerS cDNAs under the control of the human rhodopsin kinase (hGRK1) promotor and packaged the construct in an rAAV8 vector to achieve targeted expression in photoreceptor cells. During cloning, a sequence encoding an N-terminal FLAG tag was included in the vector to detect the expression of rAAV8-CerS vectors (Fig. S1). For the three treatment groups, rAAV8-CerS2, rAAV8-CerS4 and rAAV8-CerS5, subretinal injections were performed at postnatal day (P)21 to deliver rAAV8-CerS vectors to the right eyes (REs) of Tlcd3b^−/− mice, whereas the contralateral left eyes (LEs) were injected with PBS solutions to serve as internal controls.

The expression of rAAV8-CerS vectors was examined by immunofluorescence staining of the FLAG tag when mice were 7 months of age. In the treated REs of Tlcd3b^−/− mice, for all three rAAV8-CerSs, positive FLAG-tagged transgene expression was robustly detected in photoreceptor cells, as indicated by universal expression across outer segments, inner segments and outer nuclear layers (ONLs) near the injection site (Fig. 1A). In contrast, in the PBS-injected LEs, no FLAG tag expression was observed. Additionally, we also performed quantitative PCR (qPCR) to examine CerS gene expression levels, and there were significantly higher levels of Cers2, Cers4 and Cers5 mRNAs in the treated eyes (Fig. 1B). Therefore, the subretinal injections of rAAV8-CerS2, rAAV8-CerS4 and rAAV8-CerS5 vectors all resulted in stable and efficient transgene expression in photoreceptor cells.

Based on previous studies, at 7 months of age, Tlcd3b^−/− mice exhibited significant bipolar cell and cone photoreceptor photoresponse impairment (Bertrand et al., 2021). To determine the functional rescue by canonical CerSs, we examined electrophysiological responses to light in Tlcd3b^−/− mice by conducting ERG tests. For the three canonical CerSs that were tested, three different dosages (high, medium and low) were examined for each treatment group, and ERG results demonstrated a clear dosage-dependent pattern (Fig. S2).

For rAAV8-CerS2, the medium dosage of 1.38×10¹¹ genome copies (gc)/ml successfully preserved retinal functions of REs compared to those of LEs in Tlcd3b^−/− mice. Specifically, there was a functional improvement in cone photoreceptor cells and phototransduction/inner retinal neuron response, as indicated by significantly enhanced scotopic and photopic b-wave amplitudes of REs (Fig. 2A). However, scotopic a-waves of REs were not affected, indicating that the medium dosage of rAAV8-CerS2 does not impair rod photoreceptor function. Although the low dosage of 1.38×10¹⁰ gc/ml showed neither beneficial nor detrimental impact on retinal functions, the high dosage of 1.38×10¹² gc/ml largely impaired photoresponses, as indicated by significantly reduced scotopic a-wave, and scotopic and photopic b-wave amplitudes (Fig. S2A).

rAAV8-CerS4, however, demonstrated rescue potential at the low dosage of 1.03×10¹⁰ gc/ml (Fig. 2B). AAV-treated REs showed significantly enhanced scotopic and photopic b-waves compared to PBS-treated LEs, suggesting successful preservation of cone photoreceptor cell and bipolar cell functions or phototransduction by low-dosage rAAV8-CerS4 treatment. Although the high dosage of rAAV8-CerS4 also resulted in impaired retinal functions at a dosage of 1.03×10¹² gc/ml, the medium dosage of 1.03×10¹¹ gc/ml did not have any significant impact on the REs of Tlcd3b^−/− mice compared to the LEs (Fig. S2B).

rAAV8-CerS5 demonstrated a similar dosage-dependent pattern to that of rAAV8-CerS2. The medium dosage of 7.5×10¹¹ gc/ml significantly improved scotopic and photopic b-wave amplitudes in AAV-injected REs compared to those of PBS-injected LEs (Fig. 2C). Specifically, rAAV8-CerS5 achieved the best functional recovery of cone cells compared to that achieved by the other two viruses (Fig. S2). Scotopic a-wave responses in REs did not show any significant difference compared to those in LEs, suggesting intact rod photoreceptor functions. Similar to rAAV8-CerS2, rAAV8-CerS5 also showed a detrimental impact on retinal functions at the high dosage of 7.5×10¹² gc/ml and no impact at the low dosage of 7.5×10¹⁰ gc/ml (Fig. S2C).

Collectively, at their functional rescue dosages, rAAV8-CerS5 demonstrated the best rescue potential, followed by rAAV8-CerS2 and rAAV8-CerS4 (Fig. 2), as indicated by the 75% improvement of scotopic b-wave responses (80% of those of wild type) and 82%improvement of photopic b-wave responses (79% of those of wild type) in REs compared to LEs in rAAV8-CerS5-treated mice.

To determine whether the rAAV8-CerS treatments could restore retinal morphology in Tlcd3b^−/− mice, histological analyses on Hematoxylin and Eosin (H&E)-stained retinal sections of both the LEs and REs of Tlcd3b^−/− mice were performed at 7 months of age. As expected, untreated Tlcd3b^−/− LEs showed a reduction in overall retinal thickness as well as in the ONL thickness compared to those of the wild-type control (Fig. 3). For the AAV-treated REs, only the medium-dosage rAAV8-CerS5 treatment (7.5×10¹¹ gc/ml; later referred as ‘rescue dosage’) successfully preserved the retinal and ONL thickness of REs compared to that of the LE contralateral control (Fig. 3C). Upon the treatment of rAAV8-CerS5 at the medium dosage, Tlcd3b^−/− REs showed significantly thicker ONLs and an increased number of photoreceptor cell nuclei throughout the retina relative to those of the untreated LE. Overall, rAAV8-CerS5 treatment restored the ONL thickness to approximately 90% of that of the wild-type retina based on the quantitative morphometric analysis on retinal cross-sections.

rAAV8-CerS2 and rAAV8-CerS4 treatments, however, failed to rescue retinal morphology (Fig. 3A,B) even at their functional rescue dosages (1.38×10¹¹ gc/ml and 1.03×10¹⁰ gc/ml, respectively, as shown in Fig. S3). For all CerS treatment groups, the high dosage (later referred to as ‘toxic dosage’) failed to rescue and instead drastically reduced the retinal thickness as well as ONL thickness, suggesting toxicity of rAAV8-CerS overexpression (Fig. S3).

To assess whether rAAV8-CerS treatments successfully retain cone photoreceptor cell integrity and slow down the progression of cone cell degeneration in Tlcd3b^−/− mice at their rescue dosages, cone-arrestin staining on cross-sections and peanut agglutinin (PNA) staining on retinal whole mounts were conducted to examine cone photoreceptor cell morphology. Cone cells were best preserved by rAAV8-CerS5 treatment, as indicated by the significantly increased number of cones by 30% in REs compared to that in LEs (Fig. 4B,C). Additionally, cones in RE also showed longer outer segments as well as stronger staining in the cone synaptic terminals (Fig. 4A; Fig. S4).

For rAAV8-CerS2- and rAAV8-CerS4-treated retinas, there was also an increase in the number of cone cells in REs compared to that in LE (22% increase for rAAV8-CerS2 and 20% increase for rAAV8-CerS4), but the increase was less significant compared to what we observed in rAAV8-CerS5-treated REs (Fig. 4B,C). Although cone cell numbers were significantly improved, cone morphology was not as successfully retained in REs compared to those that received rAAV8-CerS5 treatments. As shown in Fig. 4A and Fig. S4, although cone synaptic terminals were well preserved in REs, the outer segments were much shorter and displayed weaker staining relative to those observed in rAAV8-CerS5-treated REs.

Given that Tlcd3b^−/− mice showed scotopic b-wave reductions at 7 months of age, PKC-a (encoded by Prkca) and RIBEYE (encoded by Ctbp2) co-staining was performed to assess rod bipolar cell morphology and retention of bipolar ribbon synapses. All rAAV8-CerS therapy treatment groups led to better preservation of synapses at the rescue dosages, as evident by the more robust RIBEYE staining in AAV-treated REs compared to that in the contralateral control eye (Fig. 5). Among the three rAAV8-CerSs that were tested, rAAV8-CerS5 showed the best preservation of ribbon synapses compared to rAAV8-CerS2 and rAAV8-CerS4, as indicated by the strongest RIBEYE staining.

To further study the impact of rAAV8-CerS treatments on retinal ceramide profiles, we carried out targeted metabolomics analyses to assess the change in levels of different ceramide subtypes and sphingosine in Tlcd3b^−/− mice. For each virus, we looked at two dosages: the rescue dosage and the toxic dosage.

CerS5 is reported to generate primarily C16:0 and a minimal amount of C14:0. At the rescue dosage, rAAV8-CerS5 successfully restored the levels of C14:0, C16:0 and C18:0 ceramides in REs to approximately 85-90% of wild-type levels and recovered C20:0 ceramides to approximately 70% of wild-type levels (Fig. 6A). Nevertheless, overexpressing CerS5 at the rescue dosage did not induce the overproduction of other ceramide species. Both observations gave supportive evidence backing up the best rescue effects of rAAV8-CerS5. At the toxic dosage, rAAV8-CerS5 triggered the overproduction of all three major ceramide species in the retina, C16:0, C18:0 and C20:0, which led to the toxicity and retinal degeneration of the REs.

The rescue dosage of rAAV8-CerS2, however, triggered an increase in the levels of C24:0 (Fig. 6B), which is expected as it belongs to the ceramide profile of CerS2. It also elevated the levels of C16:0, C18:0 and C20:0 in the RE compared to the LE. At the toxic dosage, rAAV8-CerS2 triggered the overproduction of its profile ceramide, C24:0 (more than two times the levels seen in wild type), as well as C24:1 (approximately 1.5 times the levels in wild type) and sphingosine, but did not cause further increase in the levels of C16:0, C18:0 and C20:0 compared to what was observed at the rescue dosage. Interestingly, C16:0 levels decreased to levels even lower than those of the contralateral left eye.

CerS4 synthesizes C18:0-C22:0 ceramides. On the one hand, as expected, the rescue dosage triggered an increase in the level of C18:0 and C20:0 in the treated RE compared to the PBS-injected LE, to approximately 80% of the wild-type levels. On the other hand, it significantly decreased the level of C24:1 (Fig. 6C). However, at the toxic dosage, rAAV8-CerS4 triggered a somewhat global change in ceramide levels. There was an overproduction of all ceramides relative to their levels in the wild-type control, except for C12:0, C24:0 and C24:1. Specifically, the levels of its profile ceramides, C18:0 and C20:0, were elevated to roughly 1.4 times the levels in wild-type control.

In terms of total ceramide levels, rAAV8-CerS5 treatment also achieved the best restoration of all tested ceramide species and the sum of TLCD3B profile ceramides, C16:0, C18:0, and C20:0, to more than 80% of wild-type levels at the rescue dosage (Fig. 6D). However, despite rAAV8-CerS2 treatment having slightly better functional and morphological retention in the retina, the relative increase in the overall ceramide levels as well as that of TLCD3B profile ceramides were lower than those of rAAV8-CerS4-treated retinas (Fig. 6D), suggesting that the changes in the overall ceramide levels are not a major contributor to retinal degeneration in the context of TLCD3B loss of function. Interestingly, the C16 ceramide levels of rAAV8-CerS4-treated retinas remained unchanged compared to its levels in PBS-treated retinas, but rAAV8-CerS2-treated retinas demonstrated much higher C16 ceramide levels. At the toxic dosages, rAAV8-CerS5 treatment caused the highest level of ceramide overproduction (approximately 23% higher for overall ceramide levels and 27% higher for TLCD3B profile ceramide levels compared to those of wild type), followed by rAAV8-CerS4 treatment, which was approximately 20% higher for overall ceramide levels and 19% higher for TLCD3B profile ceramide level compared to those of wild type (Fig. 6E). Interestingly, although the toxic dosage of rAAV8-CerS2 did not drive further changes in overall ceramide levels, it led to a 10% decrease in C16 ceramide levels compared to its levels in rescue dosage-treated REs (Fig. 6E).

We report that the compensation of ceramides through the delivery of different rAAV8-CerSs successfully alleviated both impaired retinal function and photoreceptor degeneration of Tlcd3b^−/− mice. Specifically, rAAV8-CerS5 had the best rescue potential, given that it demonstrated the best ERG response restoration, ONL thickness retention and preservation of cone photoreceptor cells as well as synaptic ribbons. In addition, the targeted metabolomics analyses showed that at the rescue dosage, rAAV8-CerS5 best recovered both the TLCD3B profile ceramides and overall ceramide levels, which can be explained by CerS5 having the largest profile overlap with TLCD3B relative to that of CerS2 and CerS4.

Previous studies have mainly focused on the link between overall ceramide level and activation of apoptosis in photoreceptors. However, there is a gap in the understanding of roles each ceramide subtype plays. It is known that different CerSs are highly interactive with each other, and the overexpression of one CerS may enhance the ceramide subtypes produced by the other CerSs (Zhao et al., 2011). Interestingly, all three overexpressed CerSs induced an increase in the levels of ceramide subtypes that were not under their respective profiles. For CerS4 and CerS5, their impact appeared to be on subtypes with a similar chain length. CerS2 overexpression, however, had a global impact both on ceramides with different chain lengths and on sphingosine, which is the backbone and the intermediate between ceramide and sphingosine-1-phosphate conversion, suggesting that CerS2 may have a more universal role in ceramide synthesis relative to the roles of CerS4 and CerS5.

Our previous rAAV8-Tlcd3b gene therapy study suggested a key role played by ceramides and the association between TLCD3B regulation and ceramide profile change (Qian et al., 2022). In this study, we further demonstrated that in the context of Tlcd3b-associated retinal degeneration, the disruption of overall ceramide level is not the leading cause of photoreceptor cell death. Instead, different ceramide species may play distinct roles and the balance among them may be the key to retinal health in Tlcd3b^−/− mice. At the rescue dosages, rAAV8-CerS2 restored overall ceramide levels and TLCD3B profile ceramides to a lesser extent compared to rAAV8-CerS4 treatment, yet rAAV8-CerS2 treatment had better functional rescue and significantly improved cone photoreceptor cell preservation. There are three possible explanations for the improved rescue potential of rAAV8-CerS2. First, rAAV8-CerS4 treatment failed to restore C16:0, which is the major TLCD3B profile ceramide and the most dominant ceramide subtype in the retina. Given the abundance of C16:0 ceramide in the normal retina and the severe C16:0 reduction owing to TLCD3B loss of function in Tlcd3b^−/− mice, C16:0 ceramide might be the leading player causing retinal degeneration, and a higher compensation by C18:0 and C20:0 ceramides does not rescue retinal degeneration as effectively as C16:0 ceramide restoration. Another possibility is that when increasing the level of different ceramide species, these species need to be kept at a certain ratio to maintain homeostasis. Although rAAV8-CerS5-treated REs clearly demonstrated a descending level of the C16:0, C18:0 and C20:0 ceramides detected and rAAV8-CerS2-treated REs followed a similar pattern, rAAV8-CerS4-treated REs broke the pattern by failing to elevate C16:0 production. A third possibility would be that ceramides with different chain lengths may have different roles in terms of activating or preventing photoreceptor cell apoptosis. Some studies reported that ceramides with longer chain lengths (>C20) may have a beneficial impact on cell survival (Mesicek et al., 2010; Hartmann et al., 2012). It is possible that an elevated level of C24:0 in rAAV8-CerS2-treated REs mitigated the detrimental impact caused by the loss of TLCD3B profile ceramides. Collectively, all these possibilities point to the delicacy of retinal ceramide subtype homeostasis in the context of Tlcd3b^−/− mice, highlighting the importance to look not only at the overall ceramide level changes, but also at ceramides at the subtype level for retinopathies. As fluctuations of each ceramide subtype may have a complicated impact on the ceramide metabolic ecosystem in the retina, this further implicates the complexity of the sphingolipid pathways. Therefore, the exact relationships between different ceramide subtypes and their exact role in retinal degeneration still remains unanswered and requires further study.

In conclusion, our results showed that the subretinal injections of all three rAAV8-CerSs successfully restored retinal functions, but only rAAV8-CerS5 was able to preserve retinal morphology, indicating that CerSs and TLCD3B play partially redundant roles (Qian et al., 2022). However, Tlcd3b^−/− mice demonstrated a greater sensitivity to dosage variations in the case of rAAV8-Tlcd3b treatments than rAAV8-CerSs treatments. Additionally, the sensitivity of different phenotypes of Tlcd3b^−/− mice also varied under treatments by the three rAAV8-CerSs tested, suggesting the delicacy of regulating and maintaining the level of each ceramide species. Findings from this research suggested that ceramide is an effective target for molecular therapy development for patients who have retinal pathologies and for those with TLCD3B mutations. Future studies will explore more direct ways of ceramide delivery into the retina and test the best combination and dosage of different ceramide subtypes for treatments on Tlcd3b-associated retinopathies.

cDNAs encoding full-length CerS2, CerS4 and CerS5 (GenScript OMu18716, OMu21307 and OMu18680) were sequence verified and amplified by PCR. For detection of exogenous CerS transgene expression, a sequence encoding an N-terminus FLAG tag was present in the original GenScript vectors, and PCR amplification was performed using primer sets that included the sequences of the FLAG tag and the AgeI and EcoRI restriction enzyme sites at the ends (Fig. S1A). Both the pTR-hGRK1 AAV vector (Addgene, plasmid #60957; see vector map in Fig. S1B) and the PCR-amplified cDNA were digested with AgeI and EcoRI. The digested vector was gel-purified and then ligated with the digested insert. For AAV packaging, rAAV8 (Gene Vector Core, Baylor College of Medicine) was used for packaging hGrk1-CerS genes to achieve robust transduction efficiency and expression in retinal photoreceptors (Gene Vector Core, Baylor College of Medicine).

Tlcd3b^−/− mice were generated as previously described (Bertrand et al., 2021) and maintained on a C57BL/6J genetic background. All mice in this study were maintained in a 14-h light/10-h dark cyclic environment. All animal operations were approved by the Institutional Animal Care and Use Committee at Baylor College of Medicine. For all treatment groups, both sexes of mice were examined.

P21 mice were anesthetized with Rodent III (Center for Comparative Medicine, Baylor College of Medicine), a combination drug consisting of ketamine (22 mg/kg), xylazine (4.4 mg/kg) and acepromazine (0.37 mg/kg), which was injected intraperitoneally. Subretinal injections were performed as described previously (Zhong et al., 2015). A shallow incision was made through the sclera with a beveled 30-gauge needle. A 32-gauge blunt needle was presented inside the vitreous cavity and pushed forward until the tip of the needle had moved past the retina. The viral solution was injected into the subretinal space using an Ultra-Micro-Pump II and Micron-4 Controller (World Precision Instruments, Sarasota, FL, USA). All mice were treated once in the subretinal space with 1 µl rAAV8-hGRK1-CerS vectors (rAAV8-CerS2, rAAV8-hGRK1-CerS4 or rAAV8-hGRK1-CerS) in the RE, whereas the contralateral LE served as an internal control and was injected subretinally with 1 µl PBS. Age-matched wild-type mice were used as the external controls.

Total RNA used in the study was extracted from the retinal tissues of 7-month-old mice using the TRIzol reagent (Invitrogen, 15596018). For each rAAV8-CerS treatment, three mice were used (three AAV-treated REs and three PBS-treated LEs). cDNA synthesis was performed with 1 µg of total RNA according to the manufacturer's protocol (Invitrogen, 18091050). Gapdh (forward primer, 5′-GAGTGGGAGTTGCTGTTGAAGT-3′, and reverse primer, 5′-AGGATAGTCATTTTGGGGTTTGT-3′) was used as an internal control in SYBR Green PCR Master Mix kits (Thermo Fisher Scientific, 4367659). The relative RNA levels of Cers2, Cers4 and Cers5 in AAV-treated eyes (test groups) and PBS-treated eyes (control groups) were determined using the CFX Opus 96 Real-Time PCR System (Bio-Sad, #12016658) and the following primers: Cers2 (forward, 5′-GGACCGGTGCCACCATGCTCCAGACCTTGTATGA-3′, and reverse, 5′-GGGGAATTCTCAGTCATTCTTAGGATGATT-3′), Cers4 (forward, 5′-GGACCGGTGCCACCATGTCGTTCAGCTTGAGTGAG-3′, and reverse, 5′-GGGAATTCTGTCTATGTGGCCCGGGTGT-3′) and Cers5 (forward, 5′-GGACCGGTGCCACCATGGCGACTGCAGCAGCG-3′, and reverse, 5′-GGGAATTCCTAGTCACAGGAGTGTAGAT-3′). All experiments were carried out in triplicates. Ct values were extracted using the Bio-Rad CFX Maestro Software to calculate the relative quantitative values, which were shown by 2^−ΔΔCt. Two-tailed unpaired Student’s t-tests were performed for statistical analyses.

Mice were dark adapted overnight before ERG acquisition, and all procedures were performed under dim red light. Mice were anesthetized and topical tropicamide (1.0%; Generic, 1168129) and phenylephrine hydrochloride (2.5%; Bausch & Lomb Americas, 42702010210) eye drops were used to dilate the pupils. The cornea was anesthetized with proparacaine hydrochloride (0.5%) eye drops (Generic, 2963726). Goniosoft (2.5%; McKesson, 1122510) was applied to the cornea to keep the eyes hydrated. A ground electrode was placed on mice subcutaneously. ERGs were recorded using electrodes (LKC Technologies, Gaithersburg, MD, USA; 95-033 M) placed at the center of each cornea. Data acquisition and storage were conducted using the LKC UTAS Visual Diagnostic System and EMWIN software (LKC Technologies). The oscillatory potential was isolated and removed using the EMWIN software prior to wave analysis. A two-way repeated measures ANOVA with Sidak post hoc test for multiple comparisons was performed between AAV-treated REs and PBS-treated LEs of Tlcd3b^−/− mice to determine group-to-group statistical significance, whereas two-tailed unpaired Student’s t-tests were conducted for each luminescence to determine statistical significance.

Mice were euthanized with isoflurane, followed by cervical dislocation. Eyes were enucleated and fixed in fresh Davidson's fixative [(20% formaldehyde (Sigma-Aldrich, 252549), 35% ethanol, 10% acetic acid (Sigma-Aldrich, 695092), and 53% ddH₂O)] overnight at 4°C. After fixation, eyes were dehydrated in ethanol series and embedded in paraffin wax according to standard protocol (Qian et al., 2022). The paraffin-embedded tissue was then cut into 7-μm-thick sections (Leica, RM2255).

H&E staining was performed on retinal cross-sections, and stained sections were visualized using light microscopy (Zeiss Apotome). ONL thickness was measured using the Zen software (Zeiss). A two-way repeated measures ANOVA with Sidak post hoc test for multiple comparisons was performed between AAV-treated REs and PBS-treated LEs of Tlcd3b^−/− mice to determine group-to-group statistical significance, whereas two-tailed unpaired Student’s t-tests were conducted for each position to determine statistical significance.

For immunofluorescence staining, paraffin-embedded retinal sections were prepared as described above. Following deparaffinization and rehydration, antigen retrieval was performed by boiling retinal sections in citrate buffer [for 1l citrate buffer, it is composed of 800 ml ddH₂O, 24.268g sodium citrate dihydrate (Sigma-Aldrich, 567446), 3.358g citric acid (Sigma-Aldrich, 791725), adjusting pH to 7.4 using 0.1 N hydrochloric acid] for 30 min. Retinal sections were washed in PBS five times for 5 min each. Retinal sections were incubated in hybridization buffer NGST (10% normal goat serum with 0.1% Triton-X 100 in PBS) for 1 h at room temperature. Next, sections were incubated in primary antibodies (anti-FLAG antibody, Sigma-Aldrich, F3165, 1:500; anti-cone arrestin, Sigma-Aldrich, AB15282, 1:500; Anti-Pkca, Sigma-Aldrich, P4334, 1:5000; or anti-RIBEYE, BD Biosciences, 612044,1:1000) or Alexa Fluor 488-conjugated PNA (Invitrogen, L21409, 1:400) diluted in NGST overnight at 4°C. The following steps were conducted in the dark. Retinal sections were incubated in secondary antibodies [anti-Cy5-rabbit (Invitrogen, A21242) or anti-Cy3-mouse (Invitrogen, A32728)] diluted to 1:400 in NGST for 2 h at room temperature and then washed in PBS five times for 5 min each. All sections were then counterstained with DAPI diluted 1:1000 in PBS for 15 min at room temperature and then washed in PBS four times for 2 min each. Sections were air dried for 5 min at room temperature before ProLong Gold mounting medium (Life Technologies, P36934) was applied to each section, and slides were coverslipped. Stained sections were visualized using fluorescence microscopy (Zeiss Axio Observer Z1).

For retinal whole-mount staining, eyes were then fixed in 4% formaldehyde in PBS for 1 h at 4°C. Following fixation, retinal cups were dissected and incubated in NGST overnight at 4°C. The following steps took place in the dark. Retinal cups were incubated in Alexa Fluor 488-conjugated PNA (Invitrogen, L21409, 1:400) in NGST for 48 h at 4°C. Retinal cups were washed in PBS three times for 5 min each. Retinal cups were then transferred to a slide, cut into four petals, and flattened with the scleral side up. Samples were visualized using fluorescence microscopy (Zeiss Axio Observer.Z1). A two-way repeated measures ANOVA with Sidak post hoc test for multiple comparisons was performed between AAV-treated REs and PBS-treated LEs of Tlcd3b^−/− mice for each region (dorsonasal, dorsotemporal, ventronasal and ventrotemporal) to determine statistical significance.

Mouse retinas were collected at 3 months of age. Simultaneously, retinas of age-matched wild-type mice were collected as the control. For tissue preparation, each retina sample (∼20 mg) was mixed with 200 µl 10× diluted PBS (4°C) and 80 µl internal standard solution [phosphatidylcholine (17:0/17:0) and phosphatidylglycerol (17:0/17:0) in methanol; 50 µM; 4°C] in an Eppendorf tube (1.5 ml). Homogenization (2 min) was performed using a Bullet Blender homogenizer (Next Advance, Averill Park, NY, USA). After homogenization, 400 µl methyl tert-butyl ether (MTBE; Sigma-Aldrich) was added to the sample. Following this, the sample was vortexed for 30 s, stored under −20°C for 30 min, and sonicated in an ice bath for 10 min. After centrifugation (1600 g, 10 min), 500 µl of the upper MTBE layer was collected into a new Eppendorf tube. The MTBE layer was then dried in a Vacufuge Plus Evaporator (Eppendorf, 022820168), and samples were reconstituted with 100 μl 1:1 CHCl₃:methanol. A quality-control sample was prepared by pooling all study samples, and it was measured every ten samples to ensure consistent output by a liquid chromatography tandem mass spectrometry (LC-MS/MS) system. All mass spectrometry experiments were done on an Agilent 1290 LC-6490 QQQ-MS (Santa Clara, CA, USA). Standard sphingolipid compounds were purchased from Thermo Fisher Scientific, Sigma-Aldrich and Avanti Polar Lipids (Alabaster, AL, USA) in order to confirm detected lipids. Reverse-phase chromatography was used with a Waters XSelect HSS T3 column (150×2.1 mm, 2.5 µm particle size; Waters Corporation, Milford, MA, USA). The flow was maintained at 0.3 ml/min. The mobile phase solvent A consisted of 10 mM ammonium acetate in 60% H₂O/40% acetonitrile (ACN). Solvent B consisted of 10 mM ammonium acetate in 90% isopropyl alcohol/10% ACN. Isocratic elution was used with 50% solvent B for 3 min, before its percentage was gradually increased to 100% over the next 12 min. Following 10 min of continued 100% solvent B, at t=25 min, the percentage of solvent B was decreased gradually back to 50% to prepare for the next sample injection. MS data were extracted using Agilent's MassHunter Quantitative Analysis for QQQ (Santa Clara, CA, USA). Finally, the levels of sphingolipids were normalized to the protein amount of each retina sample (in micrograms), which was measured using the BCA protein assay kit (Thermo Fisher Scientific, PA198579). Statistical significance was determined by two-tailed unpaired Student's t-test for each lipid species.

The authors thank the Gene Vector Core at the Baylor College of Medicine for providing the AAV8 serotype and helping us generate AAV vectors.

All relevant data can be found within the article and its supplementary information.