Distinct roles for prominin-1 and photoreceptor cadherin in outer segment disc morphogenesis in CRISPR-altered X. laevis Free

Inherited retinal degenerative diseases are rare disorders that result in progressive vision loss and blindness. A common mechanism by which this occurs is disruption of the synthesis and assembly of retinal photoreceptor outer segments (OS) (Goldberg et al., 2016). These specialized organelles are highly modified primary cilia comprising membranes that contain the vision-producing pigments. There are two principal retinal photoreceptors – the rods and the cones – which differ in outer segment morphology and function. The cylindrical rod outer segments (ROS) mediate dim light (scotopic) vision and comprise organized stacks of discrete membrane discs surrounded by a plasma membrane. The tapered cone outer segments (COS) mediate daytime (photopic) vision and comprise infoldings of continuous membrane lamella. The biosynthesis and assembly of photoreceptor outer segments occurs continuously; every day new disc membrane is formed by evaginations of plasma membrane at the base of the outer segment and old membrane is shed from the tips of the outer segments (Burgoyne et al., 2015; Ding et al., 2015; Steinberg et al., 1980; Volland et al., 2015). Discarded membrane is phagocytosed by the retinal pigment epithelium (RPE). Defects in prominin-1 (PROM1) and photoreceptor cadherin (CDHR1) can result in an autosomal dominant Stargardt-like bull's eye macular dystrophy (Kniazeva et al., 1999; Michaelides et al., 2010) or autosomal recessive cone-rod dystrophies and/or retinitis pigmentosa (Gurudev et al., 2013; Rattner et al., 2001). The proteins encoded by these genes share a unique localization to the basal ROS where nascent discs are synthesized, which has led to the hypothesis that mutant PROM1 and CDHR1 cause retinal degeneration by disruption of OS morphogenesis (Rattner et al., 2001; Zacchigna et al., 2009; Han et al., 2012).

X. laevis are a common model organism for the study of the interplay between photoreceptor outer segment morphogenesis and inherited blindness. This is in part due to their large photoreceptor outer segments, large number of cones (roughly a 1:1 ratio of cones to rods in the retina), and the ease by which genetic manipulation such as transgenesis and CRISPR-Cas9 genomic editing can be performed (Feehan et al., 2019; Tam et al., 2013). They are considered to be allotetraploid animals, due to a hybridization of interspecies diploid progenitors and subsequent genomic doubling hypothesized to have occurred ∼17–18 million years ago (Kobel and Du Pasquier, 1986; Session et al., 2016). These two X. laevis subgenomes have continued to evolve independently; one chromosome set is more likely to preserve the ancestral genes (the L or long chromosomes) and one chromosome set is less stable and experiences a more frequent loss of gene conservation and expression (the S or short chromosomes). It is estimated that 50–60% of X. laevis genes are conserved and expressed in both of the L and the S chromosome sets (Session et al., 2016).

Prom1 – prominin-1, also known as prominin-like 1, CD133 or AC133 – encodes a pentaspan transmembrane glycoprotein that is evolutionarily conserved between vertebrates and invertebrates (Nie et al., 2012). Mammals, including humans, have two prominin homologs – Prom1 and Prom2. They are widely expressed in the body and are generally co-expressed, therefore they are thought to have overlapping functions. In the retina, however, Prom1 is the only homolog that is highly expressed (Corbeil et al., 2010a). X. laevis have three reported prominin homologs – prom1, prom2 and prom3 – that share ∼30% sequence identity. The prom1 and prom2 homologs are equivalent to those found in human and mouse. The prom3 homolog is found only in some non-mammalian vertebrates, such as Daniorerio, X. laevis, X. tropicalis, Gallusgallus and the platypus; it is also not significantly expressed in the retina (Han and Papermaster, 2011). Prom1 is hypothesized to be involved in ectosome formation (Marzesco et al., 2005), neurodevelopmental signaling (Singer et al., 2019), stabilization of membrane curvature (Corbeil et al., 2010b), and cytoskeletal remodeling (Thamm et al., 2019), but the only documented clinical manifestation of PROM1 mutations in humans is non-syndromic vision loss (Gurudev et al., 2013); this could be due to the overlapping expression and function of prom2 in non-retinal tissues. Prom1-null mice exhibit dysmorphic ROS and COS, mislocalized rod and cone opsins, increased photoreceptor apoptosis and impaired retinal function; retinal degeneration occurs within 3–26 weeks, depending on the genetic background of the mouse model used (Dellett et al., 2014; Zacchigna et al., 2009). ROS disorganization, an increase in lipofuscin-like deposits in the RPE, a reduced scotopic and photopic electroretinogram B wave and retinal degeneration also occur, by 4–7 weeks, in mice with overexpression of the autosomal dominant mutation, Prom1^R373C (Yang et al., 2008). It has been suggested that Prom1 and Cdhr1 proteins form a complex that performs a shared role in OS morphogenesis because Cdhr1 is mislocalized in Prom1^R373C mice, mouse Prom1 is mislocalized in Cdhr1^−/− mice, and Prom1 co-immunoprecipitates with Cdhr1 in HEK293 cells co-transfected with either wild-type or mutant Prom1 and a Cdhr1–Myc fusion construct (Yang et al., 2008).

Cdhr1 – cadherin-related family member 1, also known as protocadherin-21 (Pcdhr-21) or photoreceptor cadherin (prCAD) – encodes a single-pass transmembrane protein that is expressed exclusively in the vertebrate retina; it has no known homologs. Vertebrate Cdhr1 orthologs are expressed widely and can be identified in D. rerio, Fugurubripes, X. laevis, G. gallus, Mus musulus and Bos taurus. To date, no invertebrate orthologs have been identified (Rattner et al., 2001, 2004). Cdhr1-null mice have dysmorphic ROS, increased retinal apoptotic activity, and experience retinal degeneration by 6 months of age (Rattner et al., 2001). Mammalian Cdhr1 protein has been reported to be proteolytically cleaved, resulting in an extracellular soluble N-terminal fragment and a C-terminal fragment that remains embedded in the plasma membrane. Therefore, it has been hypothesized that this cleavage of Cdhr1 may act as an essential, irreversible step, during photoreceptor outer segment morphogenesis, such as during the sealing of discs in the ROS and subsequent fission from the surrounding plasma membrane (Rattner et al., 2004). Alternatively, Cdhr1 has been reported to act as a tether between the leading edge of nascent ROS discs and the inner segment, to guide nascent discs as they elongate (Burgoyne et al., 2015).

Despite studies demonstrating that loss of Prom1 or Cdhr1 causes photoreceptor OS defects and retinal degeneration in mice, their roles in OS morphogenesis and the mechanisms of pathogenicity for inherited retinal disease remain unknown. Here, we report the characterization of genetically modified prom1-null, cdhr1-null and prom1 plus cdhr1 double-null X. laevis, which provide new insights into the roles of these proteins in photoreceptor OS morphogenesis and retinal degenerative disease.

The prom1 gene identified on xenbase.org [xelaev18034674mg, XB-GENE-6460662, NCBI Gene ID: 100316925 (prom1.L] corresponds to the X. laevis prominin-3 sequence reported by Han and Papermaster (2011), and is therefore not expected to be significantly expressed in the retina. The prom1 sequence used in the experiments reported here – xelaev18005149m, NCBI Gene ID: 100316924 (proml-1) – was found by utilizing the xenbase.org BLAST database. It corresponds to the X. laevis prominin-1 sequence reported by Han and Papermaster (2011), and is therefore expected to have high retinal expression. There is likely no functional prom1 S chromosome gene, as only six of 27 exons were identified in mRNA, protein, and EST databases; regardless, we did design one of the sgRNAs tested to target both the L and S chromosomes. Three prom1 sgRNAs – targeting exon 1 (L), 12 (L and S), or 21 (L) – were tested and all resulted in successful editing. Sanger chromatograms for embryos with successful CRISPR/Cas9 editing were degraded near the predicted cut site, representing the occurrence of random indels due to nonhomologous end joining (NHEJ). The exon 1-targeting sgRNA was chosen for subsequent experiments because it had the highest editing efficiency (64% of eGFP-positive embryos) and the resultant phenotype from all sgRNAs was the same. Three cdhr1 guides – targeting exon 1 (L and S), 7 (L and S) or 8 (L and S) – were tested and only the exon 7-targeting sgRNA was successful in editing both the L (88% of eGFP-positive embryos) and the S (32% of eGFP-positive embryos) chromosomes. Genotypes for animals used in subsequent experiments were as follows: prom1-null F0, various random indels; prom1-null F1, L, homozygous 5 bp deletion, homozygous 6 bp deletion or heterozygous 5 bp deletion plus 6 bp deletion; cdhr1-null F2–3, L, homozygous 27 bp deletion, homozygous 10 bp deletion, or heterozygous 27 bp deletion plus 10 bp deletion and S, homozygous 4 bp deletion; prom1 plus cdhr1-null, F0 various random indels.

We found that prom1 protein is expressed in X. laevis retina in the basal ROS and in the COS opposite the disc rim protein peripherin-2 (prph-2), similar to results reported previously (Han et al., 2012). We also found prom1-positive puncta scattered throughout the plasma membrane at the surface of the ROS (Fig. S1A, black arrowheads). CRISPR-mediated prom1 gene knockdown significantly reduced prom1 immunoreactivity in the retina and in western blots, and resulted in dysmorphic photoreceptor OS (F0–F1; Fig. S1B,C). Prom1-null ROS were shortened and amorphous, with diameters changing markedly over short distances; these constrictions and bulges comprised overgrown ROS disc membranes that were bent, folded or formed circular whorls. There were also instances of overgrown loops of membranes that continued upwards alongside the ROS or downwards into the inner segment (Figs 1A and 2C). The presence of prph-2 and lack of Lucifer Yellow staining throughout the ROS indicates that, although the ROS disc membranes are overgrown, they develop rims and do not remain open to the extracellular space (F0, Fig. 1B, Fig. S2). COS in prom1-null mutants were elongated and fragmented, and were often closely associated with the ROS plasma membranes – appearing to adhere to the ROS in fragmented puncta or wrap around the ROS in cone opsin-positive tendril-like structures or sheets (F0 and F1; Fig. 1C). Mislocalization of cone opsin to the inner segment plasma membrane did not occur frequently in F0 animals, and varied greatly from a single cone inner segment to all of the cone inner segments in a section (6 of 42 specimens examined, 14–42 dpf; Fig. 1C). Cone opsin mislocalization to the inner segment was most commonly correlated with complete destruction of the COS and was not found in F1 animals with dysmorphic photoreceptors from two different genetic backgrounds (n=10). There was no observed mislocalization of cdhr1, prph-2 or rhodopsin to the inner segment (F0 and F1, Fig. S2). Between 14 dpf to 6 weeks post-fertilization, dysmorphic ROS continued to elongate but retained the dysmorphic structure of constrictions and bulges throughout the ROS and there remained no occurrence of inner segment localization of any of the retinal proteins investigated (Fig. S2). As animals aged, there was an increase in the appearance of small autofluorescent deposits that also stained heavily with Hoechst 33342; they are likely of cellular origin such as accumulations of cellular debris or possibly dying RPE cells (Fig. S2, white arrowheads). These deposits did not occur in wild-type animals.

When prom1-null retinas were examined by transmission electron microscopy (TEM), ROS membranes were found to be convoluted but contained by the plasma membrane (F0, Fig. 2A–C). Local disc structure, such as the hairpins and spacing between lamellae, was well-preserved, but higher-order disc organization was dramatically disrupted. Frequently observed features were overgrown membranes that folded in on themselves, thin tracts of disc membranes and whorls of disc membrane. There were a few instances of penetration of disc membrane into the inner segment (n=2/4 samples observed). In some ROS, the basal discs did not form hairpins; instead, they made a 90° turn from horizontal to the vertical direction, and then continued upwards along the ROS. We found that COS disc membranes were very difficult to visualize with TEM due to their tortuous morphology and fragmentation of the membranes, as observed using light microscopy. Almost no recognizable COS structures were observed by TEM other than occasional loops of disc membrane or masses of completely disordered thin membranes in close proximity to the cone inner segment (F0, Fig. 2E).

The impact of prom1-null mutations on the scotopic electroretinogram (ERG) in 6-week-old tadpoles was minimal. At the lowest light intensities, the scotopic A-wave amplitude was reduced compared to wild-type controls [25% of wild-type response amplitude at 2.5–25 candela (cd) s/m²; simple main genotype effect, F(1,78)=13.13, P=0.0005], but the difference between prom1-null and wild-type A-wave amplitudes decreased as light intensities increased (47–56% of the wild-type response at 250–750 cd s/m² and no difference at 1250–2500 cd s/m²). Prom1-null scotopic B-wave amplitudes were not significantly different from wild-type controls, and no change in the shape of the scotopic ERG waveform was observed (F0, Fig. 3A).

The impact of prom1-null mutation on the photopic ERG of 6-week-old tadpoles was significant. Prom1-null photopic A- and B-wave amplitudes were significantly reduced at higher light intensities relative to wild-type controls (F0, Fig. 3B). For the photopic A-wave [interaction effect, F(5,72)=3.435, P=0.0077], there was no difference in response amplitude at 0.25–7.5 cd s/m², but the prom1-null mutant response decreased to 70% and then 50% of the wild-type response amplitude at 25 and 75 cd s/m² [simple main genotype effect, F(1,72)=4.686, P=0.0337]. For the photopic B-wave [interaction effect F(5,72)=6.290, P<0.0001], there was no difference from 0.25–2.5 cd s/m², 54% of wild-type response at 7.5 cd s/m², and 34–36% of wild-type response at 25–75 cd s/m² [simple main genotype effect, F(1,72)=34.29, P<0.0001]. The Prom1-null cone response to 5 Hz photopic flicker was also reduced relative to wild-type at higher light intensities [interaction effect, F(5,72)=4.179, P=0.0022]. There was no difference from 0.25–0.75 cd s/m², 42% of wild-type response at 2.5–7.5 cd s/m², and 46–50% of wild-type response at 25–75 cd s/m² [simple main genotype effect, F(1,72)=34.84, P<0.001] (F0, Fig. 3C).

We determined that X. laevis cdhr1 protein is expressed at the basal ROS and in the ROS plasma membrane. Cdhr1-null animals lost cdhr1 immunoreactivity in the ROS, but some signal remained in the COS and possible RPE microvilli (F2–3, Fig. S3A,B). This immunoreactivity is likely not cdhr1-specific because western blots demonstrated loss of one strong band of the expected size in cdhr1-null animals (∼99 kDa), but also showed the presence of several non-specific bands (F3, Fig. S3C). Cdhr1-null mutants did not have grossly dysmorphic photoreceptor OS. The expression and localization of rhodopsin, cone opsin, prom1 and prph-2 were all normal (Fig. S4). Using super-resolution microscopy, we observed that some rod photoreceptors (∼20%) had areas of membrane overgrowth that extended from the basal ROS upwards, alongside the outside of the regularly ordered disc membranes and incisures. There was also visible ‘pock-marking’ or holes in some of the basal ROS (Fig. 4A–D). There was no penetration of Lucifer Yellow into the ROS, however, indicating that although there is abnormal membrane growth, the disc membranes are sealed off from the extracellular space (F2–3, Fig. 4E).

Ultrastructural analysis by TEM confirmed disc membrane orientation and growth defects in the ROS (F0–F1, 14–49 dpf, Fig. 2F). The principal defect observed was that some disc membranes (∼30%) were oriented vertically within the ROS plasma membrane; these defects occurred both as long, thin sections of vertically oriented disc membranes and shorter ‘stacked’ thicker sections of vertical membranes comprising short pieces of disc membrane and many rim structures. Rim structures and ‘bubbles’ of membrane – which likely correspond to the pock marking seen in the super-resolution images – were usually present at the point at which disc orientation was altered. Horizontally and vertically oriented ROS membrane discs had normal lamination, and the discs remained tightly packed into the ROS plasma membrane. COS discs appeared relatively free of defects, although some COS (∼10%) had a ‘frayed’ appearance in which disc lamella were not uniformly registered, suggesting overgrowth of the disc membranes at somewhat regularly spaced intervals (Fig. 2G).

The shape and scale of the scotopic ERG from 6-week-old cdhr1-null tadpoles was not significantly different from wild-type animals in regards to the A- or B-wave amplitude, but the B-wave tended to return to baseline more quickly in cdhr1-null animals (F3, Fig. 5A). There was no statistically significant effect on photopic ERG amplitudes, although there was a trend towards a slightly larger B-wave response and a small increase in latency for B-wave onset; this latency was not statistically significant for any condition other than 5 Hz flicker at 25 cd s/m² [one-way ANOVA, F(5,24)=6.625, P=0.0005] (F3, Fig. 5B,C).

Photoreceptors with prom1 plus cdhr1 double-null mutations were not significantly more dysmorphic or prone to degeneration than prom1-null knockout alone. The effects on OS structure and protein localization were not distinguishable from prom1-null retinas for both TEM and light microscopy (F0, Fig. 2D,H; Fig. 6). There was no mislocalization of any of the outer segment proteins assayed, with the exception of cone opsin at about the same frequency as the prom1-null mutants. Older double-null animals also had the small Hoechst 33342-stained autofluorescent deposits in the OS layer (white arrowhead, Fig. S5).

The impact of prom1 plus cdhr1 double-null mutations on photoreceptor function was similar to that of prom1-null mutations. Rod function was not impaired; there was no difference in scotopic A-wave or B-wave amplitude or change in the shape of the scotopic ERG waveform compared to wild-type controls (Fig. 7A). Cone function of prom1 plus cdhr1 double-null mutants was impaired compared to wild-type animals, and the effects were similar to those of prom1-null mutants. The photopic A-wave amplitude in response to increasing white light intensities was reduced [no difference from 0.25–7.5 cd s/m², 58% of wild-type response at 25 cd s/m², and 23% of wild-type response amplitude at 75 cd s/m²; simple main genotype effect, F(1,42)=7.861, P=0.0076]. The B-wave response amplitudes were also flattened [interaction effect, F(5,42)=12.93, P<0.0001; no difference from 0.25–2.5 cd s/m², 56% of wild-type response at 7.5 cd s/m², and 44–46% of wild-type response at 25 and 75 cd s/m²; simple main genotype effect, F(1,42)=62.69, P<0.0001]. The response to photopic flicker (5 Hz) of prom1 plus cdhr1 double-null animals was also reduced compared to wild-type controls (no difference from 0.25–2.5 cd s/m² and 50–67% of wild-type response amplitude at 7.5–75 cd s/m² [interaction effect, F(5,42)=3.641, P=0.0080; simple main genotype effect, F(1, 42)=27.67, P<0.0001] (F0, Fig. 7B,C).

The central finding of this study is that neither prom1 nor cdhr1 are necessary for photoreceptor OS disc membrane evagination, disc fusion, or the maintenance of spacing between membrane discs and lamellae. The observation that the prom1-null mutant phenotype is significantly more severe than the cdhr1-null mutant phenotype indicates that prom1 and cdhr1 have distinct roles in OS disc morphogenesis. Our data suggest that prom1 may regulate disc size and provide structural support to the OS by aligning and reinforcing interactions between the leading edges of nascent discs. Cdhr1 may regulate disc membrane organization in the ROS by helping to keep the discs horizontal before fusion occurs, and it also has a role in regulating the growth of new disc membranes in the ROS and lamellae in the COS. A secondary important finding of this study is that the retinas of prom1-null, cdhr1-null and double-null X. laevis do not degenerate quickly, giving us the opportunity to study the mutant photoreceptors in the relative absence of cell death. The maintenance and growth of OS, the lack of any OS protein mislocalization to the inner segment, the preservation of photoreceptor function in 6-week-old animals, and the increasing appearance of autofluorescent deposits in the OS layer of older prom1- and double-null animals suggests that retinal degeneration caused by prom1-null mutations may be due to secondary toxic retinal effects – e.g. RPE toxicity or accumulation of cellular waste products – and not due to direct effects of these mutations on OS morphogenesis or the improper trafficking of OS proteins. This is an important finding, as it may not be critical to prevent the morphological defects in photoreceptors caused by prom1-null mutations to preserve vision; therapies could instead be targeted to prevent the secondary events that ultimately cause cell death.

The effects of prom1-null mutations on photoreceptor OS structure are more severe in COS than ROS. The presence of convoluted overgrown membranes indicates that prom1 plays a role in regulating the size of OS membrane discs, possibly by aligning and reinforcing interactions between the leading edges of the discs as they elongate or by controlling the amount of membrane that is added before disc fusion occurs. Evidence for this is that the loss of prom1 results in severe membrane overgrowth in the ROS and elongation and fragmentation of the COS; nascent disc evagination and fusion still occur in the ROS, as evidenced by the presence of hairpins and properly spaced disc membranes within the ROS and the lack of Lucifer Yellow dye penetration into the ROS. Recent studies have suggested that prom1 may be involved in cytoskeletal remodeling, specifically by interacting with phosphoinositide 3-kinase and the Arp2/3 complex (Thamm et al., 2019). Interestingly, a conditional knockdown of the Arp2/3 complex in mice ROS results in abnormal OS that form knob-like protrusions made up of whorls of overgrown disc membranes (Spencer et al., 2019). These mutants differ from prom1-null mutants, however, in that they lose the ability to form membrane evaginations for new OS discs. Instead, membrane is continuously added to a single ‘disc’ outgrowth. At the molecular level, prom1 has been reported to associate with actin and to regulate membrane localization and retention of cholesterol (Röper et al., 2000; Yang et al., 2008). Overgrowth of disc membranes also occurs when eyecups are treated with cytochalasin D, a mycotoxin that inhibits actin polymerization (Williams et al., 1988), and cholesterol is known to reinforce positive membrane curvature, such as the leading edges of evaginating discs, and provide rigidity and structure to the plasma membrane. The loss of membrane rigidity due to dysregulation of cholesterol content in the leading edges of the COS lamella could explain the elongated and fragmented appearance of prom1-null X. laevis cones; unlike ROS, the COS do not have the extra structural support of full disc fusion or a surrounding plasma membrane.

The long-term survival of dysmorphic photoreceptors in X. laevis indicates that prom1 is not required for biosynthesis of OS discs or trafficking of opsins to the OS. There is no mislocalization of key OS proteins, such as rod opsin, cone opsins or prph-2 to the inner segment, and prom1-null X. laevis lack the severe retinal degeneration frequently associated with defects in ciliary or trafficking components (Goldberg et al., 2016). Cone opsin mislocalization occurred only in a small subset of prom1-null X. laevis cones, unlike previous reports in Prom1^−/− and Prom1^R373C mice (Yang et al., 2008; Zacchigna et al., 2009), and our results indicate that this mislocalization is usually the result of complete destruction of the COS. Conservation of the scotopic and photopic ERG response is an interesting result, considering the significant disruption of OS morphology in prom1-null X. laevis photoreceptors. The lack of complete impairment is not likely to be due to mosaicism in F0 animals; ERG analysis was repeated in F1 prom1-null X. laevis, and the results were similar, although with a greater depression in the scotopic ERG B-wave response amplitude (Fig. S6). Histology shows that the majority of rod and cone photoreceptors in these animals still have some sort of an intact, albeit highly dysmorphic, OS. Compared to rods, cones have less sensitivity and greater dynamic range; the photopic ERG measures their ability to detect changes in light above background rather than their absolute sensitivity. Our data suggests that, although dysmorphic, the remaining COS membranes contain all the components required for phototransduction, and that a highly organized outer segment is not essential for maintaining basic cone signaling function.

The lack of early and severe retinal degeneration in prom1-null X. laevis supports the hypothesis that disruption of OS disc morphogenesis is not the cause of photoreceptor death, but that an indirect secondary effect could be responsible instead. In support of this, there are increasing numbers of small, heavily Hoechst-stained and autofluorescent deposits in the OS layer of 6-week-old prom1-null X. laevis and there have been reports of an increase in lipofuscin-like deposits in Prom1^R373C mice (Yang et al., 2008). Raising mice in constant darkness has also been reported to be protective against retinal degeneration caused by Prom1-null mutations (Dellett et al., 2014), which indicates a role of secondary toxicity in the pathogenesis of Prom1-null mutations; lipofuscins or toxic retinal metabolic byproducts should build up less quickly when photoreceptors are less active. Clinically, some PROM1 mutant retinal diseases resemble Stargardt disease (Kniazeva et al., 1999; Lee et al., 2019; Michaelides et al., 2010), which is also caused by secondary toxic effects – that is, the build-up of the bisretinoid A2PE (lipofuscins) due to the lack of ABCA4 kills the RPE, leading to subsequent photoreceptor death (Sparrow et al., 2012; Tanna et al., 2017). It should be noted, however, that PROM1-null and PROM1^R373C patients lack the retinal hyperfluorescence characteristic of ABCA4-associated Stargardt disease.

The phenotype of cdhr1-null X. laevis is less severe than reported in mouse (Rattner et al., 2001), and is limited to changes in disc membrane orientation, poor disc stacking, occasional overgrown ROS membranes and oversized membrane lamellae in the COS. There is no mislocalization of rhodopsin, cone opsin, prph-2 or prom1, and no Lucifer Yellow dye penetration into the overgrown ROS membranes. The subtle changes in ERG response suggest that all components required for phototransduction are present, and functional. X. laevis photoreceptors appear to be largely unaffected by the loss of cdhr1 and do not exhibit early or severe retinal degeneration, even though mice have been reported to experience up to 50% loss of outer nuclear layer density by 6 months of age (Rattner et al., 2001). This difference in reported photoreceptor death between species could be the result of differences in lifespan of mice and X. laevis (1.5–2 years vs 15–30 years, respectively) or differences in cdhr1 protein cleavage and localization in the ROS. In mice, Cdhr1 is localized only to the basal ROS, where it is cleaved into a soluble N-terminus and a membrane-embedded C-terminus (Rattner et al., 2004). It was hypothesized that this cleavage represents an irreversible step in ROS morphogenesis, such as during the membrane fusion process when nascent open discs are sealed and then enclosed within the ROS plasma membrane. However, X. laevis appear to require neither cdhr1 nor cdhr1 cleavage for membrane fusion; N-terminal immunoreactivity is present in the basal ROS as well as throughout the ROS plasma membrane in wild-type animals and hairpins are detectable using immunohistochemistry for prph-2 and TEM in cdhr1-null mutants. This suggests that cdhr1 and cdhr1 cleavage are not integral to the fusion event that controls disc sealing in the ROS. It has also been suggested that cdhr1 may act as a tether between the leading edge of nascent ROS discs and the inner segment, and that it could guide OS disc growth until the disc has reached the correct size, after which the tether is severed (Burgoyne et al., 2015). Our data support this hypothesis over the cleavage hypothesis because of the presence of overgrown and vertically oriented disc membranes within the ROS plasma membrane and overgrown COS lamellae. X. laevis have calyceal processes made up of F-actin fibers that form a cage-like structure around the base of the rod and cone OS, which mice lack (Sahly et al., 2012). The presence of calyceal processes could lessen the impact of the loss of cdhr1 if its function is to tether nascent discs in alignment, as they could provide additional structural support and guidance for nascent disc and lamellar membranes as they elongate. A previous study hypothesized that new COS membrane is added at somewhat regular intervals along the non-ciliary side of the COS in areas termed distal invaginations (Eckmiller, 1987). Our data may support this theory, as the regularly spaced membrane overgrowth that gives the cdhr1-null COS a ‘frayed’ appearance could represent these distal invaginations and areas of new membrane growth in the COS or failure of these distal invaginations to develop. A significant difference between this previous report and ours, however, is that we observed areas of overgrown membrane at both the ciliary and non-ciliary sides of the COS in cdhr1-null X. laevis.

Finally, our study does not support the existence of a prom1–cdhr1 protein complex that performs a single, shared role in OS disc morphogenesis as previously hypothesized (Yang et al., 2008). If this were true, then it would be expected that prom1-null and cdhr1-null mutations should affect OS morphology similarly. Instead, the prom1-null phenotype is significantly more severe than the cdhr1-null phenotype, and there is no mislocalization of prom1 in cdhr1-null X. laevis or cdhr1 in prom1-null X. laevis. These proteins may still interact, and each does appear to have a role in regulating ROS disc and COS lamellae membrane size, but their function is not dependent on the presence of the other protein. Double-null mutants do not have a significantly different phenotype to prom1-null mutants, which provides further evidence against a genetic or protein interaction. If a relationship existed, we would expect that double-null animals would have either a significantly more severe phenotype (synergy) or possibly even a mitigated phenotype, which may occur when gene products operate in series within the same pathway (Mani et al., 2008).

In summary, the results reported in this study have provided significant new insights into the function of prom1 and cdhr1 proteins in photoreceptor OS morphogenesis and the pathogenesis of PROM1-null mutations in human disease. Our data support a role for prom1 in the regulation of nascent disc size and structural support for the OS and a role for cdhr1 in the regulation of disc membrane elongation, tethering and organization. Our study shows definitively that these proteins are not required for OS disc evagination, the formation of hairpins or ROS disc fusion. Furthermore, we are the first to report that prom1-null mutations may cause retinal degeneration by secondary effects instead of direct effects on photoreceptor OS morphogenesis. This new insight could lead to a paradigm shift in the development of therapies for human patients, as it may not be necessary to rebuild the photoreceptors to preserve vision; therapies could instead be targeted to preventing the secondary events that ultimately cause cell death. This is an exciting subject of future investigation.

Animal use protocols were approved by the University of British Columbia Animal Care Committee and carried out in accordance with the standards set by the Canadian Council on Animal Care. Xenopus laevis tadpoles were housed at 18°C under a 12-h cyclic light schedule (08:00–20:00; 900–1200 lux).

Single-guide RNAs (sgRNAs) were synthesized on the basis of X. laevis prom1 and cdhr1 sequences identified by xenbase.org [prom1: xelaev18005149m, NCBI Gene ID: 100316924 (proml-1); cdhr1: xelaev18035010mg, XB-GENE-865231, NCBI Gene ID: 100337587 (cdhr1.L) and xelaev18000599mg, XB-GENE-17339736, NCBI Gene ID: 108703385 (cdhr1.S)]. sgRNAs, corresponding oligonucleotides and primer sequences were designed in silico using the ChopChop (http://chopchop.cbu.uib.no/), ZiFiT (http://zifit.partners.org), and Integrated DNA Technologies OligoAnalyzer (idtdna.com/calc/analyzer) online tools. The sgRNA target sequences, corresponding complementary oligonucleotides for cloning into the pDR274 vector, and PCR primer sequences used for subsequent genotyping are listed in Table S1.

To synthesize sgRNAs, oligonucleotides encoding targeting sequences were cloned into pDR274 (Addgene plasmid #42250, deposited by Keith Joung) and the resulting derivatives were linearized and used as templates for in vitro RNA transcription with the HiScribe™ T7 High Yield RNA Synthesis Kit (NEB, Ipswich, MA); 1.5 µg template was incubated for 4 h following kit protocols. Cas9 mRNA was transcribed in vitro from linearized pMLM3613 (Addgene plasmid #42251; deposited by Keith Joung) using the HiScribe T7 ARCA mRNA kit with poly-A tailing (NEB). A combination of three separate reactions was used to achieve high concentrations (≥45 µg) of Cas9 mRNA. eGFP mRNA was transcribed in vitro from a linearized pBluescript II SK+ construct using the T7 mMessage Ultra kit (Ambion/Thermo Fisher Scientific, Waltham, MA, USA). All synthesized RNA was treated with DNase I (NEB) to remove template contamination, and then purified using a Qiagen RNeasy kit (Hilden, Germany). Final products were quantified by absorbance at 260/280 nm with a NanoDrop 2000c spectrophotometer (Thermo Fisher Scientific), evaluated for size and quality by agarose gel electrophoresis (1.5%), and stored at −80°C.

In vitro fertilization and microinjections were performed at 18°C as previously described (Feehan et al., 2017). Eggs and sperm were incubated in a Petri dish for 20 min, the follicle cell sheath was removed from fertilized embryos using 2% cysteine and gentle shaking, and then the embryos were tightly packed into a monolayer in 2% agarose injection plates flooded with 0.4× Marc's modified Ringer's solution (MMR; 1× is 100 mM NaCl, 2 mM KCl, 1 mM MgCl₂•6H₂O, 2 mM CaCl₂, 5 mM HEPES) plus 6% Ficoll. Cas9 mRNA (5 ng), eGFP mRNA (750 pg), and sgRNAs (1.25–2 ng) were combined and loaded into a pulled glass micropipette with a 20–25 μm bore. Micropipettes were mounted in a micromanipulator and connected to 25 µl Hamilton gastight syringes mounted in a Hamilton syringe pump set to deliver 36 μl/h. Embryos were injected with the RNA solution for 1 s (equal to ∼10 nl); time was kept with a metronome. Injected embryos that exhibited symmetrical division at the 4-cell stage were transferred to 6% Ficoll plus 10 µg/ml gentamycin in 0.1× MMR at 18°C. At 1–3 dpf, surviving embryos were screened for eGFP fluorescence using an epifluorescence-equipped Leica MZ16F dissecting microscope; uniformly fluorescent embryos were selected and transferred to 1× tadpole Ringer's solution (10 mM NaCl, 0.15 mM KCl, 0.24 mM CaCl₂•H₂O, 0.1 mM MgCl₂•H₂O) and raised at 18°C in a 12-h cyclic light incubator.

Whole embryos (1–3 dpf) or tail snips (14 dpf and older) were used for genomic DNA extraction. Tissues were placed in 75 µl of genomic prep buffer (50 µM Tris-HCl pH 8.5, 1 µM EDTA, 0.5% Tween 20 and 0.2 mg/ml proteinase K) and incubated at 55°C for 2 h and 95°C for 10 min. Extracted genomic DNA was used as template for PCR amplification of target exon sequences using the primers listed in Table S1. Primers and free nucleotides were removed by ExoSAP enzyme treatment (37°C for 30 min and then 95°C for 5 min) and the products were analyzed by Sanger sequencing (Genewiz, Seattle, WA).

For fluorescence immunohistochemistry, whole eyes were fixed in 4% paraformaldehyde plus 3% sucrose in 0.1 M sodium phosphate buffer (PB; pH 7.4) overnight at 4°C. Fixed eyes were then cryoprotected in 0.1 M PB plus 20% sucrose (pH 7.4) for 3 h at 22°C with gentle shaking or 18–48 h at 4°C without shaking. Cryoprotected eyes were embedded in Optimal Cutting Temperature medium (OCT; Thermo Fisher Scientific) and then quick-frozen to −80°C. Sagittal or coronal cryosections of the central retina were cut at 12 µm, thaw-mounted onto Fisherbrand Superfrost plus slides and stored at −20°C.

For immunolabeling, sections were washed (3×8 min) in 1× phosphate-buffered saline (PBS) and then incubated in blocking buffer (1% goat serum plus 0.1% Triton X-100 in 1× PBS) for 30–45 min. After blocking, sections were washed and then incubated overnight in primary antibody in dilution buffer (0.1% goat serum plus 0.1% Triton X-100 in 1× PBS). After incubation, sections were washed and then secondary antibodies and counterstains were applied in dilution buffer and incubated for 4–6 h at 22°C; all tissues were counterstained with Hoechst 33342 and fluorophore-conjugated wheat germ agglutinin (WGA) to visualize nuclei and photoreceptor OS membranes. Sections were cover-slipped using Mowiol mounting medium (Millipore Sigma, St. Louis, MO) and imaged using a Zeiss 510 or 800 confocal microscope equipped with a 40× N.A. 1.2 water immersion objective or a Zeiss LSM 880 with Airyscan equipped with a 63×1.4 N.A. oil immersion objective. Sections were double-labeled in pairs – rhodopsin and cone opsin, cdhr1 and prph-2, and prom1 and WGA – to maximize tissue use from small tadpole eyes. Antibody sources, counterstains, and concentrations used are listed in Table S2. Micrographs represent maximum intensity projections of whole retinal sections (z=0.28 µm/step) unless specified otherwise.

To label photoreceptor disc membranes with Lucifer Yellow, a ‘peeled grape preparation’ of whole eyes was made. This is suitable for tadpoles and froglets, but works best in animals whose scleras have not been hardened by cartilaginous growth. Using a 30 gauge needle and extra fine forceps, the sclera was split open and then gently peeled away, along with the RPE, leaving a globe comprising the retina and intact lens. The retina-lens globes were then incubated in 20 µl of 0.4% Lucifer Yellow-VS lithium salt in 60% L15 culture medium for 45 min at room temperature. Excess dye was removed by using a transfer pipette to drop in and remove the globes from three different Eppendorf tubes filled with 1000 µl of clean 60% L15. Rinsed globes were then fixed and prepared for imaging the same way as whole eyes for fluorescence immunohistochemistry (described above).

Detailed methods for TEM tissue preparation are published elsewhere (Tam et al., 2015). Briefly, whole eyes were fixed in 4% paraformaldehyde+1% glutaraldehyde in 0.1 M PB at 4°C for ≥24 h. Fixed eyes were then infiltrated with 2.3 M sucrose in 0.1 M PB for 1–4 h at 22°C with gentle shaking, embedded in OCT (Thermo Fisher Scientific), cryosectioned at 20 µm, and then thaw-mounted (one section per slide) onto gelatin-coated Fisherbrand Superfrost plus slides. Optimally oriented sections were washed with 0.1 M sodium cacodylate and then stained for 30 min with 1% osmium tetroxide. After staining, sections were dehydrated in increasing concentrations of anhydrous ethanol and then infiltrated with increasing concentrations of Eponate 12 resin (Ted Pella Inc., Redding, CA) diluted with anhydrous ethanol. Once tissues were infiltrated with 100% resin, Beem^® capsules with the ends trimmed off were placed over the section on the slide and then the capsule was filled with resin and allowed to polymerize (16–24 h at 65°C). Ultrathin sections (silver-gray; 50–70 nm) were cut with a diamond knife and collected on 0.5% Formvar-coated nickel slot grids. Sections were stained with saturated aqueous uranyl acetate (12 min) and Venable and Coggeshall's lead citrate (0.25%, 5 min). Imaging was performed with a Hitachi 7600 TEM at 80 kV.

Single retinas from tadpoles aged >45 days post-fertilization (minimum equatorial diameter ∼1.6 mm) were isolated from the eye cup and solubilized in 50 µl of solubilizing buffer (1× PBS, 2.5% SDS, 5 mM Tris-HCl pH 6.8, 20% sucrose, Bromophenol Blue, 2 mM EDTA, 1 mM PMSF and 4% β-mercaptoethanol). Protein samples (12 µl) were separated with a 10% SDS-PAGE resolving gel using the Laemmli discontinuous buffer system and then transferred to a PVDF transfer membrane (Immobilon-FL, Merck KGaA, Darmstadt, Germany) using a Bio-Rad wet transfer apparatus. Blots were blocked for 30 min (1% skim milk in 1× PBS) and then probed with anti-N-cdhr1 (∼99 kDa) or anti-N-prom1 (∼95 kDa) overnight. Blots were then incubated with IRDye800CW- or IRDye700CW-conjugated secondary antibodies (1:10,000; Rockland, Gilbertsville, PA) for 3 h and analyzed on a LI-COR Odyssey imager (Li-Cor, Lincoln, NE, USA).

Electroretinograms (ERGs) were recorded as previously described (Vent-Schmidt et al., 2017). Electrodes were connected to a model 1800 AC amplifier and head stage (AM Systems, Sequim, WA) and an Espion Ganzfeld stimulator (ColorDome) and recording unit (Diagnosys LLC, Lowell, MA) were used. Corneal recordings were made from a silver wire electrode set in a glass micropipette filled with 0.1× MMR and mounted into a micromanipulator. The combined reference and ground were a modified gold EEG electrode glued into a 60 mm Petri dish. Tadpoles at Niewkoop and Faber stages 52–54 were anesthetized by exposure to 0.03% tricaine for 2 min and then mounted on their right side on the reference electrode using 2% low-melting point agarose infused with 0.01% tricaine. Scotopic ERGs were recorded in animals that had been dark-adapted overnight and prepared under dim red light; recorded stimuli were the average of five trials. Photopic ERGs were recorded in animals that had been exposed to a normal light cycle and prepared under regular lab lighting (350–500 lux); recorded stimuli were the average of 10 trials. The A-wave amplitude was measured from baseline, which was determined by the voltage at 15 ms before the light flash to A-wave trough, and the B-wave amplitude was determined from A-wave trough to the B-wave peak, including oscillatory potentials. All ERGs were recorded from the left eye, which was then fixed and processed for histology. Genotyping was performed after recording ERGs and animals were removed from the final data analysis if Sanger sequencing of tail snips determined they did not have successful editing in the desired target genes.

Animals used for analysis were selected randomly from tanks that contained either wild-type or CRISPR-edited mutants; gene editing was confirmed after data collection was complete. If animals did not have successful gene editing, as determined by Sanger sequencing of tail snips, those data were not included in the final analysis. Tadpole sex was not determined, so data from both sexes were pooled. The generation and number of animals used for each set of experiments are indicated in the Results and in figure captions. Statistical analysis and graphing was performed using GraphPad Prism (V.6-8; San Diego, CA, USA). Western blot band density was quantified using FIJI (V 1.52p, Bethesda, Maryland, USA) and then unpaired, two-tailed t-tests were used to analyze differences between wild-type and mutant eyes. ERG waveforms were visualized in Excel (Microsoft, Redmond, WA) and analyzed by measuring the A-wave and B-wave peak amplitudes, plotting them, and then fitting the resulting curves using non-linear regression analysis. Genotype effects, light intensity effects and interaction effects were analyzed by two-way ANOVA with Sidak's post hoc test. Photopic latency in cdhr1-null animals was analyzed by measuring the difference between peaks (cdhr1-null mutant minus wild-type) and then comparing the mean differences of all groups using a one-way ANOVA with Tukey post hoc test (data not shown). Micrographs were processed using Adobe Photoshop (Creative Cloud 2019) and FIJI; any nonlinear adjustments in signal intensity are reported in the figure captions. Osmium peppering artifacts that obscured the underlying OS structure were digitally removed from some of the TEM micrographs. A preliminary report of some of our findings was presented previously in abstract form (Carr et al., 2019).

Transmission electron microscopy imaging supported by the University of British Columbia Bioimaging Facility and staff. Purchase of the Zeiss LSM 800 and LSM 880 with Airyscan funded by an infrastructure grant from the Canadian Foundation for Innovation (O.L.M.).

The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.253906.reviewer-comments.pdf