The tetraspanin disc proteins, peripherin-2 and ROM1, facilitate CNG channel localization to the rod outer segment

Abstract

The light-responsive outer segment of rod photoreceptors is composed of two distinct membrane subdomains: discs and the plasma membrane. We investigate how the disk protein peripherin-2 is engaged in CNG channel delivery to the outer segment. Instead of forming outer segments, peripherin-2 knockout (Rds^−/−) photoreceptors release ciliary ectosomes, in which CNG channel levels are markedly reduced relative to other outer segment proteins. This is intriguing as downregulation of the CNG channel is not a general feature of degenerative mouse models with dysmorphic outer segments. Overexpression of the β1-subunit of CNG in Rds^−/− rods reveals that the majority is trapped in intracellular membranes, but is restored to the outer segment by co-expressing peripherin-2. We test peripherin-2 chimeras containing either the N-terminus, tetraspanin core, or C-terminus and find that the tetraspanin domain is sufficient to localize CNGβ1 to the outer segment. We further show that the membrane remodeling function of the tetraspanin domain facilitates this process by the redundant action of the tetraspanin domain from ROM1 and the reemergence of the endogenous CNG channel in aged Rds^−/− rods that have produced ciliary membrane protrusions.

INTRODUCTION

Human vision begins when photons of light initiate a biochemical signal that elicits an electrical response within the outer segment compartment of photoreceptor cells. The outer segment is a cylindrical organelle filled with hundreds of flattened membranes called discs, which house visual transduction machinery and structural proteins. This organelle has the singular function of generating the light response and therefore does not contain normal protein turnover machinery. To replenish membrane proteins, discs must be constantly renewed through the addition of new discs at the base of the outer segment and subsequent phagocytosis of old discs at the distal tip by the overlying retinal pigment epithelium (RPE; Young and Bok, 1969). The entire outer segment is renewed over an approximately 2-week time span (Young, 1967), placing a high demand for membrane trafficking of both signaling and structural proteins into the outer segment (Pearring et al., 2013; Williams, 2002). Defects in disk renewal or membrane trafficking underlie many forms of inherited retinal degeneration that cause human blindness.

In rod photoreceptors, disk membranes are physically separate from the plasma membrane surrounding them. The protein composition of these two membrane subdomains is distinct; therefore, membrane proteins must be sorted into these membrane subdomains based on their function. Disc membranes house proteins that initiate the visual response, as well as the structural proteins, peripherin-2 and ROM1, which are localized to the hairpin curve of the disk rim. Proteins that modulate the electrical properties of the cell are restricted to the plasma membrane (Molday and Molday, 1987; Skiba et al., 2013; Skiba et al., 2021). To better understand how membrane proteins populate the plasma membrane, we examined mechanisms of CNG channel delivery to the outer segment. It has been suggested that discs are stabilized laterally to the plasma membrane through interactions with disk rim proteins. The prime candidates for this interaction are peripherin-2 and the β1-subunit of the cyclic nucleotide-gated channel (CNGβ1) based on evidence that peripherin-2 has been detected following immunoprecipitation of CNG channel subunits (Poetsch et al., 2001).

The CNG channel is a heterotetramer that consists of three α1- and one β1-subunit in rod photoreceptors (Barret et al., 2022; Xue et al., 2022). While the α1-subunit (CNGα1) can form functional homotetramers in the absence of CNGβ1 (Kaupp et al., 1989), the reverse is not true (Chen et al., 1993). The 3:1 stoichiometry of the channel subunits is critical for proper localization and function of the CNG channel in the outer segment. In fact, the channel subunits have been shown to assemble early in biosynthesis and traffic as a fully formed channel to the plasma membrane of the outer segment (Pearring et al., 2021). Previous studies found that in the absence of CNGβ1, CNGα1 is severely downregulated (Hüttl et al., 2005; Zhang et al., 2009). When CNGα1 is overexpressed in CNGβ1-deficient rods, it does not target specifically to the plasma membrane of the outer segment, and instead, localizes indiscriminately throughout the plasma membrane of the cell, suggesting that CNGα1 does not contain targeting information (Pearring et al., 2021). It was shown that the outer segment targeting information for the CNG channel is contained within the β1 subunit (Pearring et al., 2021).

To get to the outer segment, membrane proteins utilize two trafficking routes that diverge shortly after biosynthesis in the ER. The conventional trafficking route utilizes vesicular transport through the Golgi before reaching the outer segment. Membrane proteins that utilize the unconventional trafficking route bypass the Golgi, relying on vesicular transport from the ER to reach the outer segment. Proteins such as rhodopsin and the CNG channel have been identified to traffic conventionally, while the disk rim proteins peripherin-2 and ABCA4 use the unconventional trafficking route to reach the outer segment (Connell and Molday, 1990; Deretic and Papermaster, 1991; Illing et al., 1997; Tian et al., 2014; Pearring et al., 2021). In mice, there is a small proportion of peripherin-2 that traffics through the conventional pathway, which was shown to require its binding partner and homolog, ROM1 (Conley et al., 2019). The biological importance of segregating disk rim proteins into each trafficking pathway remains unknown.

In the peripherin-2 knockout mouse (aka retinal degeneration slow, Rds), photoreceptor outer segment proteins and membrane material are not incorporated into discs, but instead are released in extracellular vesicles called ectosomes (Salinas et al., 2017). Curiously, it was reported that the CNGβ1 subunit was virtually undetectable in peripherin-2 knockout retinas, suggesting that the CNG channel may require peripherin-2 for delivery to the outer segment (Chakraborty et al., 2015, 2016). Supporting this possibility, a previous study using bimolecular fluorescence complementation assays in transgenic Xenopus rods showed that overexpressed peripherin-2 and CNGβ1 successfully complement in the inner segment where protein synthesis occurs (Ritter et al., 2011). In this study, we investigated whether peripherin-2 facilitates CNG channel localization to the outer segment by performing a comprehensive analysis of channel expression and localization in Rds^−/− retinas. We found that CNGβ1 is detectable in ectosomes released from the photoreceptor cilia of young Rds^−/− mice, though at much lower levels than other disk-specific proteins. Interestingly, we find progressive accumulation of CNGβ1 in the outer segment region of Rds^−/− retinas at a timepoint when photoreceptors begin to elaborate membrane whorls from their cilia. Since CNGβ1 contains the outer segment targeting signal for the channel, we went on to investigate peripherin-2′s role in CNGβ1 and, therefore, channel delivery. We find that trafficking of the CNG channel does not require peripherin-2/ROM1 complexes that utilize the conventional pathways through the Golgi. Further, we found that the membrane-spanning tetraspanin region of peripherin-2 is sufficient to rescue outer segment localization of the channel. We also discovered that the tetraspanin core of ROM1 can facilitate CNGβ1 localization, congruent with these tetraspanin proteins having redundant functions.

RESULTS

In aged Rds^−/− rods, CNG channel localization is restored to lamellar membrane outgrowths

We characterized CNG channel subunit expression and localization in a side-by-side comparison between WT, Rds^+/−, and Rds^−/− retinas at P21 before the onset of photoreceptor degeneration. We performed immunofluorescence staining on WT and Rds^−/− retinal sections (Rds^+/− retinal sections found in Supplemental Figure S1A) and found CNGα1 and ROM1 localized to the outer segment region (Figure 1A). To detect CNGβ1, we used an anti-GARP antibody that recognizes the N-terminal GARP domain of CNGβ1. This antibody also recognizes two soluble GARP1 and GARP2 proteins that are expressed from the Cngb1 gene as a result of alternative mRNA splicing (Ardell et al., 2000). GARP proteins have normal outer segment localization in WT and Rds^+/− retinas, while no fluorescence was observed in the outer segment region of Rds^−/− retinal sections at P21 (Figure 1A). We corroborated the apparent absence of CNGβ1 in Rds^−/− retinal sections using a CNGβ1-specific antibody (Supplemental Figure S2). qRT-PCR analysis from retinal lysates revealed that Cnga1, Cngb1 Rom1, and Abca4 mRNA expression were reduced by ~50% in Rds^−/− compared with WT or Rds^+/− retinas (Figure 1B). This suggests a global downregulation of outer segment components in Rds^−/− retinas. We then quantified protein levels of CNGβ1, CNGα1, ROM1, and ABCA4 from serial dilutions of retinal membrane fractions by Western blotting. We found that protein levels of CNGα1, CNGβ1 and ROM1 were drastically reduced compared with ABCA4, a protein previously shown to be present in Rds^−/− ectosomes (Figure 1, C and D; Spencer et al., 2019). We were unable to detect either α1 or β1 subunits of the CNG channel by Western blot analysis of Rds^−/− retina membrane fractions. To determine if there is even a small amount of CNG channel delivered to the cilium (and therefore in ectosomes), we purified ectosomes from multiple P21 Rds^−/− retinas and analyzed this highly concentrated preparation by Western blotting. By this method, we were able to detect both CNGα1 and CNGβ1 in ectosomes (Figure 1E). Likewise, a previous study conducted mass spec analysis of purified Rds^−/− ectosomes and detected CNGα1 along with virtually all other outer segment proteins. However, this previous study did not include an analysis of CNGβ1 (Spencer et al., 2019). We revisited this proteomic data and confirmed the presence of CNGβ1 in the Rds^−/− ectosome proteome (Supplemental Table S1). Together, our data show that compared with other outer segment membrane proteins, such as ABCA4, the CNG channel is severely downregulated in Rds^−/− retinas. We hypothesized that the inability to form an outer segment in the absence of peripherin-2 specifically influences either CNG channel stability or delivery.

FIGURE 1:

Characterization of CNG channel subunit localization in Rds^−/− retinas. A. Retinal sections from WT and Rds^−/− mice at P21 immunostained for ROM1, CNGα1, GARP proteins, and peripherin-2. Nuclei counterstained in blue. Scale Bar, 10 μm. Here and throughout the manuscript: outer segment (OS), inner segment (IS), and nucleus (N). B. Cnga1, Cngb1, Rom1, and Abca4 mRNA levels in P21 WT, Rds^+/−, and Rds^−/− retinal lysates were determined by qRT-PCR analysis. One-way ANOVA was performed. pValues for Cnga1 ****, p < 0.0001; ns, p = 0.152; Cngb1 **, p = 0.0021; ns, p = 0.923; Rom1 **, p = 0.0092; ns, p = 0.246; Abca4 **, p = 0.0087, ns = 0.718. C. Representative Western blots show serial dilutions of WT, Rds^+/−, and Rds^−/− membrane fractions for ABCA4, ROM1, Prph2, CNGα1, and CNGβ1. D. Relative protein expression is plotted in the bar graph. One-way ANOVA was performed. p values for ABCA4 ns, p = 0.0525; **, p = 0.0059 and ROM1 ****, p < 0.0001. Welch’s t test performed. p values for CNGα1 **, p = 0.0016; and CNGβ1 ***, p = 0.0006. E. Representative Western blots show serial dilutions of an Rds^−/− ectosome fraction for ABCA4, CNGβ1, CNGα1, and ROM1 compared with a WT membrane fraction at a single concentration. F. SEM images of the apical surface of Rds^−/− retinas at P14 and P45. Scale Bar, 2 μm. Closed yellow arrowhead indicates photoreceptor cilia in young Rds^−/−. Open yellow arrowhead indicates membrane mass protruding from photoreceptor cilia in aged Rds^−/−. G. TEM images of Rds^−/− retinal cross-sections at P14 and P45. Scale bar 0.5 μm. H. Retinal sections from WT and Rds^−/− mice at P45 immunostained for ROM1, CNGα1, GARP proteins, and peripherin-2. Nuclei counterstained in blue. Scale Bar, 10 μm.

In the original publications characterizing the ultrastructure of Rds^−/− photoreceptors at P21, it was reported that a laminar membrane protrusion was occasionally seen emanating from the photoreceptor cilium (Sanyal and Jansen, 1981; Cohen, 1983; Jansen et al., 1987). We performed scanning electron microscopy (SEM) from Rds^−/− retinas at P14 and P45 to assess the global appearance of photoreceptor cilia over time. By en-face imaging of the apical surface of photoreceptor cells in young Rds^−/− retinas, we observed photoreceptor cilia surrounded by extracellular vesicles (Figure 1F). In contrast, many photoreceptor cilia of aged Rds^−/− retinas develop an enlarged morphology (Figure 1F). We performed transmission electrony microscopy (TEM) of Rds^−/− retinas to assess the ultrastructure of these membrane bulges and found that they contain layered membrane structures attached to photoreceptor cilia (Figure 1G). To investigate whether these membrane layers retained at the cilium were sufficient to drive CNGβ1 trafficking to the outer segment, we immunostained WT and Rds^−/− retinas for GARP, CNGα1, and ROM1 at P45, when these structures are abundant. Represented by GARP and CNGα1 staining, we observed a dramatic increase in the ciliary localization of the CNG channel in the aged Rds^−/− retinas (Figure 1H; Supplemental Figure S2). Additionally, ROM1 staining is also present in these structures. These findings suggest that in the absence of peripherin-2, the α1- and β1-subunits can form a stable tetramer, but that the channel delivery is enhanced by the accumulation of laminated membranes at the ciliary tip.

Ciliary delivery of the β1-subunit of the CNG channel is reduced in Rds^−/− mice

While we observed both α1 and β1 subunits in ciliary-derived ectosomes from Rds^−/− mice, it remains unclear why CNG channel levels are specifically reduced. We hypothesized that overexpressed CNGβ1 could drive enhanced ciliary localization of the CNG channel in Rds^−/− rods. MYC-tagged CNGβ1 was overexpressed using a rhodopsin promoter in Rds^−/− rod photoreceptors by in vivo electroporation. Overexpression resulted in an abundance of MYC-CNGβ1 expression, but localization was restricted to areas surrounding the nucleus and within internal structures of the inner segment (Figure 2A). This localization pattern suggests that overexpressed CNGβ1 is almost completely trapped in the endomembranes and not in ciliary ectosomes. Importantly, overexpression of other outer segment-resident proteins, such as FLAG-tagged rhodopsin, can traffic to outer segment ectosomes (Figure 2B; Salinas et al., 2017). These data suggest the downregulation of CNGβ1 in Rds^−/− rods is a consequence of improper trafficking. We demonstrate this further by co-expressing full-length, FLAG-tagged peripherin-2 with MYC-CNGβ1 in Rds^−/− rods and observe that re-expression of peripherin-2 is sufficient to restore localization of CNGβ1 to the outer segment (Figure 2C). This can also be seen with endogenous CNGβ1 when peripherin-2 is overexpressed and retinal sections are stained for CNGβ1 (Supplementary Figure S3A). CNG channel subunit stoichiometry is important for trafficking, so we confirmed that endogenous CNGα1 could accommodate overexpressed MYC-CNGβ1 by staining with anti-CNGα1 antibodies (Supplementary Figure S3B). We show that this phenomenon is specific to peripherin-2 and not an artifact of overexpression by co-expressing MYC-CNGβ1 and FLAG-rhodopsin in Rds^−/− rods (Figure 2D). FLAG-rhodopsin localizes to ectosomes, but MYC-CNGβ1 remains almost entirely excluded, similar to Figure 2A.

FIGURE 2:

Ciliary delivery of the β1-subunit of the CNG channel is reduced in Rds^−/− mice. Rds^−/− mouse retinal cross-sections expressing A. MYC-CNGβ1 (n = 3), B. FLAG-rhodopsin (n = 3), C. MYC-CNGβ1 and FLAG-Prph2 (n = 5), or D. MYC-CNGβ1 and FLAG-Rhodopsin (n = 6). Retinal sections were stained for MYC and/or FLAG. Nuclei counterstained in blue. Scale Bar, 10 μm.

The peripherin-2/ROM1 complex is not required for CNG Channel trafficking to the outer segment

We then wanted to determine how peripherin-2 facilitates trafficking of the CNG channel to the outer segment. One possible scenario could be that peripherin-2 and the CNG channel traffic together to the outer segment. We investigated whether the small proportion of peripherin-2/ROM1 that utilizes the same conventional trafficking route as the CNG channel plays a role in channel delivery. We performed a deglycosylation assay on Rom1^−/− retinal lysates in which all of peripherin-2 traffics unconventionally (Conley et al., 2019). WT and Rom1^−/− retinal lysates were treated with PNGase F or EndoH to assess the glycosylation state of glycoproteins. PNGase F cleaves all N-linked glycoproteins and is used to monitor the molecular weight shift following deglycosylation. EndoH cannot cleave hybrid oligosaccharide chains that are produced in the Golgi, so proteins that traffic along the conventional pathway, like rhodopsin and CNGα1, are EndoH-resistant. Proteins that rely on the unconventional pathway, like peripherin-2, remain EndoH sensitive (Figure 3A). The CNGβ1 subunit is not glycosylated, so we analyzed the glycosylation state of the CNGα1 subunit because the subunits have been shown to traffic together (Pearring et al., 2021). We confirmed that the fraction of peripherin-2 that traffics conventionally through the Golgi is indeed absent in Rom1^−/− retinas and found that trafficking of the CNG channel via the conventional pathway is unaffected (Figure 3A).

FIGURE 3:

The peripherin-2/ROM1 complex is not required for CNG channel conventional trafficking or localization. A. Representative Western blot of Rom1^+/+ and Rom1^−/− retinal lysates deglycosylated with either PNGase F or EndoH enzymes and immunoblotted for Rho, CNGα1, or Prph2. Red arrowhead marks EndoH-resistant peripherin-2, which is absent in Rom1^−/− B. Representative images from Rom1^+/+ or Rom1^−/− retinal sections immunostained for ROM1, rhodopsin, peripherin-2, CNGα1, and CNGβ1. Scale Bar, 10 μm. n = 3 for all.

Rom1^−/− retinal sections were stained for the channel components, CNGα1 and CNGβ1. We found that both subunits of the CNG channel are properly localized to the plasma membrane of the outer segment in Rom1^−/− retinas (Figure 3B). These data show that CNGβ1 does not require peripherin-2/ROM1 complexes to traffic through the conventional pathway or localize to the outer segment plasma membrane.

The peripherin tetraspanin membrane domain is sufficient to direct CNGβ1 to the outer segment

Our finding that CNGβ1 transport through the Golgi does not rely on peripherin-2 led us to investigate the role peripherin-2 plays in facilitating outer segment localization of the CNG channel. Specific domains of peripherin-2 were examined in isolation for their ability to restore CNGβ1 localization to the outer segment. We started with two previously characterized fusion proteins that have been shown to localize to the outer segment and build a rudimentary membrane structure (Salinas et al., 2017). The Prph2-Rho_CT construct fuses the first 290 amino acids of peripherin-2—including the four transmembrane domains—to the C-terminal tail of rhodopsin that contains its outer segment targeting motif. Alternatively, the Rho-Prph2_CT construct takes the C-terminal, outer-segment targeting region of peripherin-2 and fuses it to the seven transmembrane domains of rhodopsin. We co-expressed these constructs with MYC-CNGβ1 and found that Prph2-Rho_CT can rescue MYC-CNGβ1 localization to the outer segment, while Rho-Prph2_CT cannot (Figure 4, A–C). The percentage of outer segments in Rds^−/− rods with co-localization between MYC-CNGβ1 and FLAG-Prph2 constructs was quantified across animals and displayed as a bar graph in Figure 4F. We confirmed these data by staining for endogenous CNGβ1 in Rds^−/− rods electroporated with each construct (Supplemental Figure S3, C and D).

FIGURE 4:

The tetraspanin membrane domain of peripherin-2 is required for CNGβ1 localization to the outer segment. Rds^−/− mouse retinal cross-sections co-expressing MYC-CNGβ1 and FLAG-tagged A. Prph2 (n = 5), B. Prph2-Rho_CT (n = 4), C. Rho-Prph2_CT (n = 7), D. ROM1_NT-Prph2 (n = 3), or E. Prph2_ΔNT (n = 4). Sections are immunostained for MYC and FLAG. Nuclei counterstained in blue. Scale Bar, 10 μm. F. Bar graph shows percentage of outer segment structures containing both FLAG (peripherin chimeras) and MYC-CNGβ1. One-way ANOVA was performed with post hoc t tests. p values: ****, p < 0.0001; ns, p>0.05.

Rhodopsin was previously shown to facilitate the trafficking of some outer segment membrane proteins (Pearring et al., 2015; Spencer et al., 2016). To ensure that the trafficking of CNGβ1 is specific to peripherin-2 features, we recreated these chimeras by replacing rhodopsin with a complementary GPCR, the serotonin-6 receptor (Htr6). We fused the C-terminal tail of peripherin-2 to the seven transmembrane domains of Htr6 (Htr6-Prph2_CT) as well as fused the four transmembrane domains of peripherin-2 to the targeting region of Htr6 (amino acids 392–440, Prph2-Htr6_CT; Barbeito et al., 2021). We confirmed that these molecular chimeras can localize to the outer segment in WT retinal sections (Supplemental Figure S4; Salinas et al., 2017). Expressing these chimeras in Rds^−/− rods replicates that only the chimera containing the first 290 amino acids of peripherin-2 can rescue CNGβ1 localization to outer segment membranes (Supplemental Figure S5, A and B). From these data, we find that the tetraspanin core of peripherin-2, and not its C-terminus, restores CNGβ1 localization in Rds^−/− rods.

We then examined the cytosolic N-terminus of peripherin-2 to further narrow down the region of peripherin-2 required for CNGβ1 delivery. Peripherin-2 and ROM1 are structurally homologous, but the amino acid sequences differ within the cytosolic N-terminal region, making ROM1 a good target for a chimeric protein. We started by replacing the N-terminus of peripherin-2 with the N-terminal region of ROM1. When electroporated into Rds^−/− rods, we found that MYC-CNGβ1 was localized to the outer segment membranes emanating from the rod cilia (Figure 4, D and F). However, we cannot exclude that the N-terminus of ROM1 is redundant to that of peripherin-2, so we followed up by removing peripherin-2′s N-terminus and replacing it with a FLAG-tag (FLAG-Prph2_ΔNT). After electroporation in WT rods, we found that this construct did not express in the majority of transfected rods and was unable to build rudimentary outer segments in Rds^−/− rods (Supplemental Figure S5, C and D). To improve expression, we extended the sequence to a 3XFLAG-tag (3XFLAG-Prph2_ΔNT). We found this construct had improved expression in WT rods (Supplemental Figure S4E) and was able to build small, rudimentary outer segments in Rds^−/− rods. When co-expressed with MYC-CNGβ1 in Rds^−/− rods, the 3XFLAG-Prph2_ΔNT was sufficient to rescue CNGβ1 localization to these outer segment structures (Figure 4, E and F). Overall, these data show that the cytosolic N- and C-termini of peripherin-2 are dispensable for CNGβ1 outer segment localization and reiterate that the tetraspanin region of peripherin-2 is required for CNGβ1 delivery.

Overexpression of ROM1 constructs in Rds^−/− rods restores CNGβ1 localization to rudimentary outer segments

In a recent study, it was shown that the peripherin-2 homolog, ROM1, can function redundantly with peripherin-2 in outer segment disk formation (Lewis et al., 2023a). We were curious if ROM1 is sufficient to facilitate CNGβ1 localization to the outer segment. We started by overexpressing FLAG-ROM1 in Rds^−/− rods and found that FLAG-ROM1 was primarily localized to internal membranes and unable to build ciliary membrane structures (Supplemental Figure S5E), even though it was properly localized to the outer segment in WT rods (Supplemental Figure S3F). Previous studies have improved targeting and expression of ROM1 in Rds^−/− rods by replacing its C-terminus with that of peripherin-2 (Conley et al., 2019). We co-electroporated MYC-CNGβ1 with a FLAG-tagged ROM1-Prph2_CT construct into Rds^−/− rods and found small FLAG-positive structures in the outer segment region that were colocalizing with MYC-CNGβ1 (Figure 5A). We then wanted to determine whether the ROM1 tetraspanin region was sufficient to facilitate CNGβ1 delivery by further replacing ROM1’s cytoplasmic N-terminus with that of peripherin-2 (Prph2_NT/CT-ROM1). When we co-expressed this construct with MYC-CNGβ1 in Rds^−/− rods, we found that Prph2_NT/CT-ROM1 can rescue MYC-CNGβ1 localization to the outer segment (Figure 5B). Together, this shows that ROM1 is redundant to peripherin-2 in its ability to restore CNGβ1 to the outer segment of Rds^−/− rods and that this function is facilitated through the tetraspanin domain.

FIGURE 5:

ROM1’s tetraspanin domain restores localization of CNGβ1 to the outer segment. Rds^−/− mouse retinal cross-sections co-expressing MYC-CNGβ1 and FLAG-tagged A. ROM1-Prph2_CT (n = 4) or B. Prph2_NT/CT-ROM1 (n = 5). Sections are immunostained for FLAG and MYC. Nuclei counterstained in blue. Scale Bar, 10 μm.

DISCUSSION

The visual response in photoreceptors relies upon proper outer segment organization, which includes the accurate localization of membrane proteins to this organelle. It has been suggested that the structural disk protein, peripherin-2, could guide CNG channel localization to the outer segment. Interactions between these two proteins have been identified, and it was reported that CNGβ1′s localization was compromised in the absence of peripherin-2. The mechanism behind the CNG channel’s dependence on peripherin-2 to make it to the outer segment was perplexing, as peripherin-2 and CNGβ1 mostly utilize separate trafficking pathways to reach the outer segment and reside in two distinct membrane domains within the outer segment. We set out to investigate how peripherin-2 may be involved in the localization of the CNG channel to the photoreceptor outer segment.

We found in Rds^−/− retinas that mRNA levels of several outer segment genes are reduced. We posit that the loss of the outer segment compartment drives general downregulation of outer segment genes. We noted that the amount of CNG channel in the outer segment region in Rds^−/− retina sections from young mice was severely reduced, which is consistent with previous reports (Chakraborty et al., 2015; Chakraborty et al., 2016). When we examined protein levels, we found that peripherin-2′s binding partner, ROM1, and the CNG channel subunits were severely downregulated when compared with the disk protein ABCA4. In fact, only when we examined a concentrated preparation of ciliary-derived ectosomes from these mice did we observe that low levels of the CNG channel are actually present. Our finding that there is a decrease in the CNG channel at the protein level suggests that not all outer segment proteins behave similarly when the outer segment is not formed. Supporting this notion, previous reports have shown the reverse effect: PCHD21 protein is upregulated in Rds^−/− retinas (Rattner et al., 2004). The selective downregulation of the CNG channel in the absence of peripherin-2 is curious, as other models of outer segment dysmorphia (e.g., Rho^−/−, PCDH21^−/−, or WASF3^−/−) localize the CNG channel properly (Rattner et al., 2001; Pearring et al., 2015; Spencer et al., 2023). Recent studies have investigated mechanisms driving G protein-coupled receptor sorting into extracellular vesicles released from the ciliary membrane and inner segment plasma membrane (Nager et al., 2017; Ropelewski and Imanishi, 2020; Lewis et al., 2024). It is possible that these or similar mechanisms play a role in excluding the CNG channel from ciliary ectosomes, which could result in its selective downregulation.

Our study went on to examine factors that contribute to CNG channel ciliary localization. One hypothesis was that peripherin-2 facilitates CNG channel trafficking to the outer segment—possibly through the small fraction of peripherin-2 that traffics through the Golgi in a complex with ROM1. We examined ROM1^−/− retinas, in which no peripherin-2 goes through the Golgi, and found CNG channel trafficking to the outer segment was unaffected. These data show that peripherin-2 and the CNG channel traffic separately, and are corroborated by previous work from our lab that found that CNGβ1 outer segment targeting occurs in the absence of peripherin-2 binding (Pearring et al., 2021). While our data do not rule out that CNGβ1 and peripherin-2 could interact before trafficking pathways diverge, we find this interaction does not play a role in CNG channel delivery to the cilia. We went on to determine the minimal region of peripherin-2 that can restore CNGβ1 delivery to the outer segment. We originally hypothesized that the C-terminus of peripherin-2 would be critical for CNGβ1 delivery because this region contains both the outer segment targeting sequence for peripherin-2 and a signal to retain ectosomes at the ciliary tip (Tam et al., 2004; Salinas et al., 2013; Salinas et al., 2017). However, we found that both the cytoplasmic N- and C-terminal regions were dispensable, and rather the membrane-spanning domain is necessary and sufficient for CNGβ1 delivery.

The major role of the tetraspanin domain of peripherin-2 is to create curvature in the membrane through its assembly into homotetramers or heterotetramers with its homolog, ROM1. These tetramers are fundamental units that create the hairpin structure at the edges of outer segment discs known as the rim (Milstein et al., 2020; Pöge et al., 2021). The tetraspanin proteins peripherin-2 and ROM1 are also required in proper ratios for successful disk enclosure (Lewis et al., 2023a; Lewis et al., 2023b). The essential role that these tetraspanins play in forming disk membranes suggests that the delivery of the CNG channel may be driven by the delineation of a lamellar membrane within the outer segment. In the P45 Rds^−/− retinas, we found that ciliary protrusions also contain ROM1, which has been shown to have redundant functions with peripherin-2 (Lewis et al., 2023a). We hypothesize that lamellar protrusions in Rds^−/− at P45 are built by ROM1; however, we were unable to increase their prevalence through overexpression of FLAG-tagged full-length ROM1. We were able to circumvent this by appending the C-terminus of peripherin-2 to the tetraspanin domain of ROM1, which we found was sufficient to restore CNGβ1.

In the photoreceptor outer segment, discs are formed as an evagination of the ciliary membrane through a lamellipodia-like mechanism that relies on branched actin filaments. It has been shown that inhibiting actin polymerization prevents new disk formation (Spencer et al., 2019) and also prevents CNG channel entry into the plasma membrane (Nemet et al., 2014). These data suggest that these two events are linked. Interestingly, when actin polymerization is inhibited, the newly synthesized CNG channel is held at the distal end of the connecting cilium, primed for entry (Nemet et al., 2014). This is also the site where tetraspanin proteins are found during new disk synthesis, as they are restricted to the axonemal side of nascent discs (Nemet et al., 2014; Burgoyne et al., 2015; Ding et al., 2015; Klementieva et al., 2025). Together, these data suggest that the partitioning of disk membranes from the plasma membrane by tetraspanin proteins is a key step for localizing the CNG channel and could play a role in segregating other proteins into the plasma membrane. Our results support this conclusion, and additional evidence comes from analysis of rhodopsin knockout outer segments, which have internal disk-like structures and localize the CNG channel properly (Pearring et al., 2015). Therefore, peripherin-2′s ability to delineate a disk is sufficient to drive CNG channel localization even in the absence of a fully elaborated outer segment.

MATERIALS AND METHODS

Request a protocol through Bio-protocol.

Animals

Rds^−/− described by (Sanyal and Jansen, 1981) and confirmed as a peripherin-2 knockout by (Ma et al., 1995). WT CD1 mice used in electroporation experiments were obtained from Charles River; WT C57BL/6J and Rds^−/− from The Jackson Laboratory. Rds^−/− mice maintained in our colony were backcrossed to C57BL/6J to produce Rds^± and Rds^+/+ littermates, included in qRT-PCR, immunofluorescence, and Western blot experiments as heterozygous and wild-type mice, respectively. All animals were analyzed at P21 unless otherwise noted. Mice were handled following protocols approved by the Institutional Animal Care and Use Committee at the University of Michigan (registry number A3114–01). As there are no known gender-specific differences in vertebrate photoreceptor development and/or function, male and female mice were randomly assigned to experimental groups. All mice were housed in a 12/12-h light/dark cycle with free access to food and water.

DNA constructs

DNA constructs were generated using overhang extension PCR methods. Generation of Prph2-Rho_CT, Rho-Per_CT, and Htr6-Per_CT was previously described (Salinas et al., 2017). Primers used to generate the remaining molecular fusion proteins are detailed in Supplemental Table S2. All forward and reverse primers were designed to introduce AgeI and NotI restriction digest sites, respectively. For rod-specific expression, cDNA sequences were cloned between 5 AgeI and 3 NotI cloning sites into the pRho vector to drive expression by a 2.2 kb bovine rhodopsin promoter (Addgene plasmid #11156; http://n2t.net/addgene:11156).

In vivo electroporation

DNA constructs were electroporated into neonatal mouse retinas using the protocol described in (Matsuda and Cepko, 2007) with modifications detailed in (Salinas et al., 2017). Briefly, neonatal (P0–P2) mice were anesthetized on ice, followed by puncture of the eyelid and cornea with a 30-gauge needle at the periphery of the eye. A blunt-ended, 32-gauge Hamilton syringe was used to inject 0.3–0.5 μl of concentrated plasmid DNA subretinally (~2 μg/μl of the construct of interest mixed with ~1 μg/μl of mCherry transfection marker). The positive side of a tweezer-type electrode was placed over the injected eye (BTX, Holliston, MA), and five 50 ms pulses were applied at 100–110 V using an ECM830 square pulse generator (BTX). After neonates recover, they are returned to their mother to develop until collection at P20–P22.

Immunofluorescence

Mouse eyes were collected and drop-fixed with 4% paraformaldehyde in 1X PBS (Thermo Fisher Scientific, AAJ75889K8) for 1–2 h at room temperature. Eyes were washed and stored in 1X PBS before they were dissected into posterior eyecups. Eyecups were embedded in 4% agarose (Thermo Fisher Scientific, BP160–500) and cut into 100 μm sections through the central retina using a Leica VT1200s vibratome. Floating sections were blocked in 5% donkey serum and 0.5% Triton X-100 in 1X PBS for 1 h at room temperature before incubation with primary antibody overnight at 4°C. The following day, sections were washed with 1X PBS and then incubated with fluorophore-conjugated secondary antibodies and 10 μg/ml DAPI (Sigma-Aldrich, 102362760) for 1–2 h at room temperature. Sections were then rinsed three times in 1X PBS and mounted with Prolong Glass (Thermo Fisher Scientific, P36980) under High Precision #1.5 coverslips (Electron Microscopy Sciences, 72204–10). Images were acquired using a Zeiss Observer 7 inverted microscope equipped with a 63X oil-immersion objective (1.40 NA) and LSM 800 confocal scanhead controlled by Zen 5.0 software (Zeiss).

Enriched ectosome preparation

Ectosomes were purified from Rds^−/− mice using a protocol adapted from (Spencer et al., 2019). In brief, at least 10 Rds^−/− retinas were detached in ice-cold Ringer’s solution containing 130 mM NaCl, 3.6 mM KCl, 2.4 mM MgCl2, 1.2 mM CaCl2, and 10 mM HEPES (pH 7.4) solution containing Complete Protease Inhibitor (CPI; Millipore Sigma, 11836170001). Detached retinas were transferred to a 1.5 ml low-retention microcentrifuge tube and agitated at 4°C for 10 min. Large debris was removed by centrifugation at 300 g for 5 min at 4°C. The supernatant was further cleared by centrifugation at 2000 g for 5 min at 4°C. The ectosome fraction was then pelleted by ultracentrifugation at 150,000 g for 20 min at 4°C. The membrane pellet was resuspended in 20 μl of 2% SDS in PBS+CPI. Multiple volumes (10, 5, 2.5, and 1.25 μl) of the sample were then loaded on an SDS–PAGE gel, followed by Western blot.

Western blot and deglycosylation

Whole retinas were sonicated in 200 μl of 1X PBS with CPI and centrifuged at 60,000 rpm for 20 min at 4°C. The supernatant was collected as the soluble fraction, and the pellet was resuspended in 250 μl of 2% SDS in PBS with CPI and collected as the membrane fraction. Membrane fractions were subsequently spun at 14,000 rpm to pellet any remaining insoluble material. A DC Assay was performed on the cleared membrane fractions and compared with a BSA standard curve to determine protein concentration. Samples were diluted to a concentration 0.8–1 μg/μl in Laemelli sample buffer with 5% β-mercaptoethanol and loaded on an AnyKd Mini-PROTEAN TGX Precast Protein Gels (Bio-Rad, 4569033) in increasing μl amounts to achieve the desired μg amount. SDS–PAGE gels were run at 50 V for 30 min before increasing to 120 V until the dye front ran to the bottom (~1 h).

The SDS–PAGE gel was transferred to Immun-Blot Low Fluorescence PVDF Membrane (Bio-Rad, 1620264) at 90 mV for 90 min. Membranes were blocked for ~1 h with Intercept Blocking Buffer (LI-COR Biosciences, 927–70003) at room temperature before incubation with primary antibodies diluted in 1:1 ratio of Blocking Buffer and PBS-T overnight at 4°C. The next day, membranes were washed with 1X PBS-T and incubated with IRDye secondary antibodies diluted in 1:1 ratio of Blocking Buffer and PBS-T with 0.2% SDS for 1–2 h at room temperature. Bands were visualized and quantified using the Odyssey CLx infrared imaging system (LiCor Bioscience). Images of the uncropped Western blots can be found in Supplemental Figure S6. For CNGα1 immunoblots, membranes were blocked for ~1 h with 5% non-fat dry milk, 5% donkey serum, and 1% TritonX-100 in PBS at room temperature before incubation with primary antibodies in blocking solution overnight at 4°C. The next day, membranes were washed with 1X PBS-T and incubated with an HRP-conjugated secondary antibody diluted in blocking solution for 1–2 h at room temperature. Bands were visualized and quantified using the Azure imaging system at multiple exposure times ranging from 30 s to 20 min.

To perform a deglycosylation assay, whole retinal lysates were prepared by sonicating one retina in 250 μl of 2% SDS in 1X PBS with CPI. Lysates were cleared by centrifugation at 14,000 rpm and 23°C. Protein concentration was then determined by DC Assay, and three 20 μl reactions containing 10 μg of retinal lysate were prepared: control, PNGase F, and EndoH (NEB, P0708L and P0702L). Control and PNGase F reactions contained 2 μl GlycoBuffer 2, 2 μl of 10% NP-40, and 2 μl of PNGase F or water, respectively. The EndoH reaction contained 2 μl GlycoBuffer 3 and 2 μl EndoH. Reactions were incubated for 1 h at 37°C before performing the Western Blot procedure as described above.

Antibodies

The following commercial primary antibodies were used: mouse monoclonal M2 anti-FLAG (Sigma-Aldrich, F1804); rabbit polyclonal anti-FLAG (Sigma-Aldrich, F7425); goat polyclonal anti-FLAG (Abcam, ab1257); rabbit monoclonal 71D10 anti-MYC (Cell Signaling, 2278S); mouse monoclonal 9B11 anti-MYC (Cell Signaling, 2276S); mouse monoclonal 1D4 anti-rhodopsin (Abcam, ab5417); goat polyclonal M-18 anti-ABCA4 (Santa Cruz, SC-21460); mouse monoclonal PMc 1D1 anti-CNGα1 (Abcam, ab255766); goat polyclonal anti-ABCA4 (Everest Biotech, EB08615). The following antibodies were generously provided by: Vadim Arshavsky, Duke University (sheep polyclonal anti-ROM-1); Robert Molday, University of British Columbia (mouse monoclonal 1D1 PMC, anti-CNGα1; mouse monoclonal 4B1, anti-GARP); Steven Pittler, University of Alabama at Birmingham (rabbit polyclonal anti-CNGβ1).

The following commercial secondary antibodies were used: Donkey anti-mouse AF488 (A21202, Thermo Fisher Scientific); Donkey anti-rabbit AF488 (A21206, Thermo Fisher Scientific); Donkey anti-sheep (A11015, Thermo Fisher Scientific); Donkey anti-sheep AF568 (A21099, Thermo Fisher Scientific); Donkey anti-mouse AF647 (A31571, Thermo Fisher Scientific); Donkey anti-Sheep AF647 (A21448, Thermo Fisher Scientific); Donkey anti-mouse HRP (Jackson ImmunoResearch 715–035-150).

Scanning electron microscopy

Mouse retinas were dissected in ice-cold Ringer’s solution adjusted to 314 mOsm. The retinas were fixed with 2% paraformaldehyde, 2% glutaraldehyde, and 0.05% calcium chloride in 50 mM MOPS (pH 7.4) for 1 h with gentle agitation. Retinas were then washed with PBS and post-fixed with 1% Osmium Tetroxide in PBS (Electron Microscopy Sciences [EMS]) for 1 h. Next, the retinas were washed with PBS, subsequently dehydrated through graded ethanol to 100%, and finally dried via the critical point method utilizing a Tousimis Samdri PVT-3D. Dried retinas were then mounted onto 12 mm SEM stubs with carbon conductive paint (EMS) and immediately sputter coated with ~15 nm of Au-Pd alloy utilizing a Denton Desk-V coater. Lastly, retinas were analyzed with a JEOL JSM-IT100 Scanning Electron Microscope operated at 10kV, and digital micrographs were collected with an Everhart-Thornley secondary electron detector.

Transmission electron microscopy

Fixation and processing of mouse eyes for thin plastic sections were performed as described previously (Ding et al., 2015). Mice were deeply anesthetized and transcardially perfused with 2% paraformaldehyde, 2% glutaraldehyde, and 0.05% calcium chloride in 50 mM MOPS (pH 7.4), resulting in exsanguination. The eyes were enucleated and fixed for an additional 2 h in the same buffer at 22°C. The fixed eyes were washed in PBS, eyecups were dissected and embedded in PBS containing 2.5% agarose (KU VF-AGT-VM; Precisionary), and cut into 200 μm thick slices on a vibratome (VT1200S; Leica). The vibratome sections were stained with 1% tannic acid (Electron Microscopy Sciences) and 1% uranyl acetate (Electron Microscopy Sciences), gradually dehydrated with ethanol and embedded in Spurr’s resin (Electron Microscopy Sciences). Around 70 nm sections were cut, placed on copper grids, and counterstained with 2% uranyl acetate and 3.5% lead citrate (19314; Ted Pella). The samples were imaged on a JEM-1400 electron microscope (JEOL) at 60 kV with a digital camera (Orius SC200D; Gatan).

SIGNIFICANCE STATEMENT.

• Human vision begins when light is captured and transduced into an electrical response within the outer segment of photoreceptors. The outer segment contains hundreds of membranous discs that, in rods, are separated from the plasma membrane surrounding them. Proteins must be functionally sorted into these membrane subdomains, but little is known about the mechanisms guiding this process. We investigate how the CNG channel is delivered to the outer segment plasma membrane and find that it requires the structural proteins, peripherin-2/ROM1, to delineate disc membranes for outer segment enrichment. Determining how membrane subdomains are defined is foundational to understanding photoreceptor pathobiology.

Abbreviations used:

DAPI

4’,6-diamidino-2-phenylindole

AF

alexa fluor

ABCA4

ATP-binding cassette transporter subfamily A member 4

CPI

complete protease inhibitor

CT

C-terminus

CNG

cyclic nucleotide-gated

EndoH

endoglycosidase H

GARP

glutamic-acid-rich protein

GPCR

G protein-coupled receptor

IS

inner segment

NT

N-terminus

N

nucleus

OS

outer segment

PNGase F

peptide:N-glycosidase F

Prph2

peripherin-2

Rds

peripherin-2 knockout

PBS

phosphate-buffered saline

PCDH21

protocadherin-21

ROM1

retinal outer segment membrane protein 1

RPE

retinal pigment epithelium

Rho

rhodopsin

SEM

scanning electron microscopy

Htr6

serotonin-6 receptor

S

synapse

TEM

transmission electron microscopy

WT

wild type

WASF3

Wiskott-Aldrich Syndrome Protein Family Member 3