Proteolytic Shedding of the Extracellular Domain of Photoreceptor Cadherin IMPLICATIONS FOR OUTER SEGMENT ASSEMBLY*

Photoreceptor cadherin (prCAD) is a distinctive cadherin family member that is concentrated at the base of rod and cone outer segments and is required for their structural integrity. During retinal development, prCAD localizes to the site of the future outer segment before rhodopsin or other phototransduction proteins. In vivo , prCAD undergoes a single proteolytic cleavage that releases the ectodomain as a soluble fragment. The C-terminal fragment containing the transmembrane and cytosolic domains remains associated with the outer segment. In rds ( (cid:1) / (cid:1) ) retinas, in which outer segment assembly is severely disrupted because of the absence of retinal degeneration slow (RDS)/peripherin, an essential outer segment structural protein, the level of prCAD is increased, whereas the levels of other outer segment proteins are decreased relative to wild type retinas. Additionally, the ratio of intact:cleaved prCAD polypeptides is increased in rds ( (cid:1) / (cid:1) ) retinas. These data imply that prCAD ectodomain cleavage is an integral part of the outer segment assembly process, and they further suggest that outer segment assembly might be driven, at least in part, by the near irreversibility of proteolysis. In the vertebrate retina, phototransduction occurs within the outer segments (OSs) of rod and cone photoreceptors. The OS is (stock 001979) genotyped al. (7). Cloning Zebrafish and Xenopus prCAD cDNAs— Using a probe derived from bovine prCAD, multiple full-length zebrafish prCAD cDNA clones were isolated by low stringency screening of an adult zebrafish retina cDNA library (a gift of Dr. James Three independent cDNA clones were sequenced. A Xe- nopus laevis expressed sequence tag clone (za26c10; GenBank TM /EBI accession AW159202) encompassing the 3 (cid:4) 55% of the prCAD coding region was obtained from Dr. Richard McCombie (Cold Spring Harbor The complete Xenopus prCAD coding region sequence was obtained by RACE-PCR amplification using the SMART RACE cDNA kit (BD Biosciences) and total RNA from Xenopus retina. In Situ Hybridization to Zebrafish and Xenopus—In situ hybridization was performed essentially as described previously (8) using digoxygenin- labeledriboprobestranscribedfromzebrafishand Xenopus prCADcDNAs and encompassing codons 1–449 and 1–857, respectively. In Vitro Transcription and Translation— A mouse prCAD cDNA tem-plate was amplified by PCR using eight different forward primers, each of which contained (from 5 (cid:4) to 3 (cid:4) ) a T7 promoter, an initiator methionine codon within an optimal consensus (“Kozak”) sequence for translation initiation, and a segment of mouse prCAD sequence to direct the site of priming. A single reverse primer located downstream of the mouse prCAD stop codon was used for each PCR reaction. The resulting PCR products were gel purified and used for in vitro transcription/transla- tion using the T N T kit (Promega). Posttranslational Modification prCAD— retina F N oligosaccharides, endo- O ) retinas, extracts were prepared both from dissected retinas and from whole eyecups, in the latter case to eliminate the possibility that OS fragments might be lost from the dissected rds ( (cid:2) / (cid:2) ) retina. Compari-sons of these two preparations showed little difference in the recovery of OS proteins.Toisolate the Triton X-100-insoluble fraction from bovine rod OSs, purified dark adapted OSs prepared under dim red light as described previously (10) were incubated with excess 11-cis retinal to maximally regenerate rhodopsin and then fractionated as shown in Fig. 4 E . All procedures were carried out in darkness or under dim red light to minimize opsin aggregation.

In the vertebrate retina, phototransduction occurs within the outer segments (OSs) 1 of rod and cone photoreceptors. The OS is a modified cilium that projects from the apical face of the photoreceptor cell toward the overlying retinal pigment epithelium (RPE). In mammals, the typical rod OS is ϳ1 m in diameter and ϳ30 -50 m in length. Each rod OS consists of a stack of ϳ1,000 flattened membrane sacs, referred to as discs, surrounded by plasma membrane. The light-absorbing visual pigments reside within the disc and plasma membranes at millimolar concentrations; other less abundant phototransduction proteins reside within one or both of these membranes or within the cytosolic space between adjacent discs.
The OS of both rods and cones are subject to constant renewal throughout the life of the organism, with synthesis and assembly of new discs at the base of the OS and with RPEmediated phagocytosis and degradation of the oldest discs at the tip of the OS (1). In mammals, each disc requires ϳ10 days to move along the length of the OS, implying that each photoreceptor assembles ϳ100 new OS discs/day. All of the OS protein and lipid constituents originate in the cell body proper and are funneled through the thin connecting cilium to the site of assembly in the OS. The most widely accepted model for OS disc assembly, based on electron microscopic analyses of primate photoreceptors, posits that nascent discs form by evagination at the base of the OS and that as they enlarge and move distally they become progressively enclosed in plasma membrane (2).
The precisely orchestrated synthesis, transport, and assembly of proteins and lipids into an almost crystalline array of OS discs raise numerous questions regarding underlying mechanisms. What controls the regular evagination of plasma membrane to form new discs? What determines the size and shape of each new disc? What guides the subsequent growth of plasma membrane around the newly forming discs, and why does this process go to completion in rods but not in cones? One insight into some of these processes has come from the observation that cytochalasin D-mediated disruption of the actin filaments within photoreceptor cilia leads to the production of aberrantly large discs (3)(4)(5). This observation implies the existence of a cytoskeletal system that controls the size of nascent discs.
The present experiments are focused on photoreceptor cadherin (prCAD), a recently discovered protein that is likely to play a role in OS assembly (6). prCAD is a single-pass transmembrane protein with six cadherin repeats in its extracellular domain. prCAD localizes to the nascent discs at the base of the OS in both rods and cones, and targeted disruption of the mouse prCAD gene leads to disorganization of photoreceptor OS (6). In this paper we show that prCAD is conserved across a wide range of vertebrates, that it is one of the first photoreceptor proteins to localize to the site of the developing OS, that the ectodomain of prCAD is released by proteolysis in the OS, and that proteolysis is coupled to OS assembly. These data suggest a model in which proteolysis of prCAD irreversibly drives assembly of the OS.

MATERIALS AND METHODS
Mice-The production and characterization of prCAD(Ϫ/Ϫ) mice have been described previously (6). rds(Ϫ/Ϫ) mice (C3A.BLiA-Pde6b ϩ .O20-Rds Rd2 /J) were obtained from the Jackson Laboratory * This work was supported by the NEI, National Institutes of Health and by the Howard Hughes Medical Institute. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AY683207 and AY684930.
Cloning of Zebrafish and Xenopus prCAD cDNAs-Using a probe derived from bovine prCAD, multiple full-length zebrafish prCAD cDNA clones were isolated by low stringency screening of an adult zebrafish retina cDNA library (a gift of Dr. James Hurley (University of Washington)). Three independent cDNA clones were sequenced. A Xenopus laevis expressed sequence tag clone (za26c10; GenBank TM /EBI accession AW159202) encompassing the 3Ј 55% of the prCAD coding region was obtained from Dr. Richard McCombie (Cold Spring Harbor Laboratory). The complete Xenopus prCAD coding region sequence was obtained by RACE-PCR amplification using the SMART RACE cDNA kit (BD Biosciences) and total RNA from Xenopus retina.
In Situ Hybridization to Zebrafish and Xenopus-In situ hybridization was performed essentially as described previously (8) using digoxygeninlabeled riboprobes transcribed from zebrafish and Xenopus prCAD cDNAs and encompassing codons 1-449 and 1-857, respectively.
In Vitro Transcription and Translation-A mouse prCAD cDNA template was amplified by PCR using eight different forward primers, each of which contained (from 5Ј to 3Ј) a T7 promoter, an initiator methionine codon within an optimal consensus ("Kozak") sequence for translation initiation, and a segment of mouse prCAD sequence to direct the site of priming. A single reverse primer located downstream of the mouse prCAD stop codon was used for each PCR reaction. The resulting PCR products were gel purified and used for in vitro transcription/translation using the TNT kit (Promega).
Enzymatic Treatments to Detect Posttranslational Modification of prCAD-Mouse retina extracts were treated with N-glycosidase F (PNGase F; New England Biolabs) to remove N-linked oligosaccharides, with endo-O-glycosidase (ProZyme) to remove O-linked disaccharides, or with bacteriophage protein phosphatase (New England Biolabs) according to the manufacturers' recommendations. The treated extracts were analyzed by immunoblotting to test for shifts in the mobility of prCAD and its proteolytic cleavage products.
Immunostaining and Immunoelectron Microscopy-Immunostaining of fresh frozen or fixed retinas and postembedding immunoelectron microscopy were performed as described by Rattner et al. (6).
Protein Extracts from Mouse Retinas and Bovine OSs-Dissected retinas were extracted in buffer containing 150 mM NaCl, 50 mM Tris, pH 7.5, 1 mM EDTA, 1% Nonidet P-40, and complete protease inhibitor mixture (Roche Applied Science). Following a 5-min incubation on ice, the lysate was cleared of nuclei and insoluble material by centrifugation at 1,600 ϫ g for 5 min at 4°C. For the comparison of WT and rds(Ϫ/Ϫ) retinas, extracts were prepared both from dissected retinas and from whole eyecups, in the latter case to eliminate the possibility that OS fragments might be lost from the dissected rds(Ϫ/Ϫ) retina. Comparisons of these two preparations showed little difference in the recovery of OS proteins.
To isolate the Triton X-100-insoluble fraction from bovine rod OSs, purified dark adapted OSs prepared under dim red light as described previously (10) were incubated with excess 11-cis retinal to maximally regenerate rhodopsin and then fractionated as shown in Fig. 4E. All procedures were carried out in darkness or under dim red light to minimize opsin aggregation.

RESULTS
Evolutionary Conservation of prCAD Structure and Expression-In earlier work, we determined the sequences of mouse, bovine, and chicken prCAD by sequencing retina-derived cDNAs. To explore the structure of prCAD in more distant species, we determined the sequences of X. laevis and Danio rerio (zebrafish) prCAD homologues by sequencing overlapping cDNA clones and RACE-PCR products derived from Xenopus and zebrafish retinas (Fig. 1A). The Xenopus and zebrafish prCAD orthologues share ϳ50% amino acid identity with mouse prCAD, and the corresponding zebrafish gene, as well as two predicted Fugu rubripes (pufferfish) genes, shows the same intron-exon arrangement as found among mammalian prCAD genes (6). This intron-exon arrangement differs from that of classical cadherin genes; for example, the cytosolic domain is coded by a single 3Ј-exon in each prCAD gene and by four exons in classical cadherin genes.
A distinctive feature of the set of prCAD sequences is the relatively high conservation within the N-terminal extracellular cadherin repeat domains and the relatively low conservation within the C-terminal cytosolic domain, with only a few small regions of the cytosolic domain showing substantial amino acid identity across all orthologues (Fig. 1B). E-cadherin and N-cadherin sequences from the same three species show the reciprocal pattern of sequence conservation, with the highest conservation in the cytosolic domain. For these two cadherins, the interaction with cytoskeletal linker proteins such as ␤-catenin and plakoglobin could provide the selective pressure for sequence conservation within the cytosolic domain (11). For prCAD, the pattern of sequence conservation suggests a conserved function for the extracellular domain and a limited role for much of the cytosolic domain.
By in situ hybridization to Xenopus and zebrafish retinas, we observe prCAD transcripts exclusively in the photoreceptor layer (Fig. 1C). In both species, prCAD transcripts are also present in distinct regions in the brain, most prominently the pineal gland. Pineal expression of prCAD also occurs in mammals as determined by the abundance of prCAD sequences among expressed sequence tags in mammalian pineal libraries. These data indicate that prCAD orthologues are present in diverse vertebrates, and in each species thus far examined, these orthologues are expressed in photoreceptor cells.
Subcellular Localization of prCAD-In earlier work, we showed by pre-embedding immunoelectron microscopy that antibodies directed to the C terminus of prCAD preferentially stain the base of the OS in transverse sections of adult mouse retina (6). This analysis further showed that prCAD immunostaining is concentrated on the face of the OS opposite the connecting cilium/axoneme, a region corresponding to the free edges of nascent OS discs where the mature disc rim structure has not yet formed (2). As this immunostaining pattern has not been described for any other OS protein and as it has significant implications for hypotheses regarding prCAD function and OS disc biogenesis, we sought to further define the localization of prCAD by serial reconstruction of immunostained photoreceptors after sectioning parallel to the plane of the retina. Fig. 2, A-J, shows five consecutive sections encompassing part of the IS-OS junction of two photoreceptors, the locations of which are shown schematically in Fig. 2K. The silver-enhanced immunogold particles are visible as small black dots in the electron micrographs in the left-hand set of panels and as red dots in the graphical representations of one of the cells in the right-hand set of panels. As suggested by the earlier analysis of transverse retinal sections, the highest concentration of prCAD immunoreactivity is found in those regions of the OS opposite the connecting cilium and axoneme. Proceeding from the most proximal section ( Fig. 2A) to the most distal section (Fig. 2I), prCAD immunoreactivity is seen to cluster more tightly along this face of the OS. Under the mild fixation conditions employed (4% paraformaldehyde, 0.1% gluteraldehyde for 1 h at room temperature), the immunostained regions, which likely correspond to the free edges of nascent discs, show less complete preservation of ultrastructure. In regions of the OS that contain mature and fully enclosed discs, prCAD immunostaining is minimal.
The prCAD immunolocalization results described above and by Rattner et al. (6) could conceivably reflect a uniform distribution of prCAD throughout the OS but a selective accessibility of prCAD to antibody probes only at the OS base. Arguing against this possibility is the observation that polyclonal antibodies directed against either the N-terminal pair of cadherin domains or the extreme C terminus give the same pattern of immunostaining predominantly at the OS base (Fig. 2, L and M). Moreover, pretreating unfixed frozen sections of mouse retina with acetone, acetic acid, or Bouin's fixative (picric acid and Formalin) did not alter the immunostaining pattern for either of these anti-prCAD antibodies. In all of these experi-ments, the specificity of the immunostaining signal was confirmed by its absence from prCAD(Ϫ/Ϫ) retinas. A similar pattern of immunostaining was seen in the macaque retina using antibodies against the N terminus of mouse prCAD. Although not evident in the images shown in The signal peptide and transmembrane domain are underlined with red and black lines, respectively. B, amino acid sequence conservation between mouse, Xenopus, and zebrafish sequences for prCAD (top) and the classical cadherins, E-cadherin (CAD1; center) and N-cadherin (CAD2; bottom). Amino acid sequence conservation was scored with a PAM 120 matrix using ClustalX software. Quality index scores (QI, range 0 -100), averaged over 10 amino acid bins, are presented above a schematic representation of the protein structure. S, signal peptide; EC, extracellular cadherin domain; PRE, prepeptide; CYTO, cytoplasmic domain. The transmembrane domain is represented by a black box. GenBank TM /EBI accession numbers for the sequences used are as follows: zCAD1, NP_571895; zCAD2, Q90275; xCAD1, S43064; xCAD2, P20310; mCAD1, NP_033994; and mCAD2, AAH22107. C, in situ hybridization with prCAD probes to sections of zebrafish and Xenopus heads (left) and eyes (right). Within the retina, hybridization is confined to the photoreceptor cell layer. prCAD hybridizes to the pineal in both species (arrow in far left image). dpf, days postfertilization; wpf, weeks postfertilization; ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer. Scale bar, 20 m. 13). In considering the function of different OS proteins, one would predict that those proteins involved in assembling and/or maintaining the structure of the OS should be the first to localize to the apical tip of the photoreceptor. OS proteins with no role in OS structure or assembly would be expected either to appear later or to be initially mislocalized; for this class of proteins, correct localization to the growing OS would only be expected after the OS transport and assembly apparatus is fully functional.
The disorganization of OS discs and the minimally perturbed electroretinogram responses seen in prCAD(Ϫ/Ϫ) retinas suggest that prCAD is involved in OS structure and/or assembly but not in phototransduction (6). To further test this idea, we immunostained developing mouse retinas at postnatal days 2, 3, 6, and 9 (P2, P3, etc.) with anti-prCAD and anti-rhodopsin antibodies (Fig. 3). At P2, when the first rods begin differentiating, rhodopsin immunoreactivity is seen throughout the cell body, but prCAD immunoreactivity is only detectable at the apical tip of the cell, the site of the developing OS (Fig. 3A). Moreover, for some developing rods there is little or no colocalization of rhodopsin and prCAD, suggesting that there may be a time window during the early development of each rod when rhodopsin (or opsin) accumulates but is not targeted to the site of the future OS. Between P2 and P9, as additional rods differentiate, the number of prCAD-stained cells increases, and progressively more rhodopsin localizes to the developing OS region. However, at each of these time points, large amounts of rhodopsin continue to accumulate in the outer nuclear layer (i.e. the rod cell bodies), whereas prCAD is only detectable at the base of the growing OS.
Analysis of arrestin immunolocalization at the same developmental times reveals a pattern and time course similar to those seen for rhodopsin. By contrast, the ␣ subunit of the rod cyclic nucleotide-gated channel (14), rod outer segment membrane protein 1 (ROM1; Ref. 15), and the axoneme protein RP1 (16) are first clearly detectable at P9, at which time they are correctly localized to the growing OS. Liu et al. (16) observed a similar pattern of localization for RP1 in developing mouse photoreceptors but were able to detect immunostaining in the OS as early as P6. Taken together, these data show that prCAD is unusual among OS proteins in its early localization to the developing OS, a temporal pattern that is consistent with a role for prCAD in OS assembly.
Models to Explain Steady-state prCAD Localization at the Base of the OS-The localization of prCAD at the base of the OS leads to an apparent paradox. During OS biogenesis, integral membrane proteins such as rhodopsin and RDS/peripherin accumulate within nascent discs. Following the complete separation of the disc and plasma membranes, such integral membrane proteins remain embedded within the disc membrane and travel from the base to the tip of the OS over a period of ϳ10 days (1). A similar time course of synthesis and turnover is presumed to apply to other integral membrane OS proteins, such as the cyclic nucleotide-gated channel, which reside in the plasma membrane and are anchored via attachment to the adjacent disc rim (17). By analogy with these other OS membrane proteins, prCAD proteins that have been transported to antibodies. In the graphical representations, silver-enhanced immuno-   representations (B, D, F, H, and J) of serial horizontal sections through a WT mouse retina immediately distal to the IS-OS junction. Prior to embedding, the tissue was immunolabeled with anti-prCAD C-terminal antibodies and 0.8-nm gold-conjugated secondary the base of the OS would be expected to move distally along the OS in association with the disc and/or plasma membranes. Therefore, at steady state, we would expect to see uniform staining of all integral membrane OS proteins, including prCAD, along the length of the OS. The low level of prCAD immunostaining beyond the base of the OS suggests that (a) the influx of newly transported prCAD to the base of the OS is matched by local degradation and/or that (b) unlike other integral membrane OS proteins, prCAD is selectively retained at the base of the OS.
Proteolytic Cleavage of the prCAD Ectodomain in Vivo-To test the possibility that prCAD might be subject to proteolysis in vivo, immunoblots of mouse retina were probed with affinitypurified prCAD antibodies directed against either the N-terminal pair of cadherin domains or the extreme C terminus (Fig.  4C). Fig. 4A shows that mouse retinas contain the expected ϳ120-kDa full-length prCAD polypeptide as well as a ϳ95-kDa N-terminal fragment and a ϳ25-kDa C-terminal fragment. The specificity of each of these bands is demonstrated by their absence in protein samples prepared from prCAD(Ϫ/Ϫ) retinas (Fig. 4A). An identical cleavage into N-and C-terminal fragments was observed in the rat retina. This cleavage is unlikely to be an in vitro artifact or the result of nonspecific proteolysis because immunoblot patterns indistinguishable from the ones shown in Fig. 4A were obtained both with mouse retinas that were immediately dissected into SDS sample buffer and with retinas that were allowed to stand for 30 min at room temperature. Moreover, expression of prCAD in transfected 293 cells results in the accumulation of full-length prCAD without detectable cleavage.
Based on the apparent molecular masses of the N-and Cterminal fragments, the point of cleavage is predicted to reside on the extracellular face of prCAD, and therefore this cleavage should release the extracellular domain from the membrane. To test this idea, freshly isolated mouse retinas were incubated in hypotonic medium for 30 min at room temperature, and then the retinas and medium were analyzed separately by immunoblotting for prCAD and its fragments (Fig. 4B). Prior to immunoblotting, the medium was centrifuged at 14,000 ϫ g for 10 min to separate soluble proteins from detached OS. A comparison of the anti-N-terminal immunoblots in Fig. 4, A and B, shows that the N-terminal fragment is efficiently released into the medium during the 30-min incubation in hypotonic me-dium. A very similar result was obtained when this experiment was conducted with isotonic medium (Dulbecco's modified Eagle's medium/F12). Fig. 4B further shows that following centrifugation of the incubation medium, full-length prCAD and the C-terminal fragment are found in the OS pellet and that the N-terminal fragment is found in the supernatant. These experiments demonstrate that proteolysis of prCAD produces a soluble N-terminal fragment and a cell-associated C-terminal fragment (Fig. 4C). Although the discovery of prCAD ectodomain cleavage suggests that in situ proteolysis plays an important role in the elimination of full-length prCAD, the subsequent fates of the N-and C-terminal fragments, including their half-lives, remain to be determined.
prCAD proteolysis most likely occurs in the OS rather than the IS because both anti-N-terminal and anti-C-terminal antibodies produce essentially the same staining pattern at the base of the OS (Fig. 2, L and M), from which we infer that substantial quantities of full-length prCAD are present at the base of the OS prior to proteolytic cleavage. We note, however, that this inference rests on the assumption that the cleaved ectodomain diffuses away from the base of the OS, as suggested by its efficient release in vitro.
To more precisely define the site of proteolytic cleavage in prCAD, the electrophoretic mobility of the C-terminal fragment obtained from mouse retinas was compared with the mobilities of a series of in vitro translated polypeptides encompassing the same C-terminal sequences but differing in length at their N termini (Fig. 4D). we cannot rule out the possibility that this fragment carries modifications other than phosphorylation or glycosylation that might affect its electrophoretic mobility.
The second model noted above for the localization of prCAD at the base of the OS, selective retention, could be most easily explained if prCAD associated with OS cytoskeletal elements. Consistent with this possibility, full-length prCAD from total mouse retina or purified bovine rod OS can be pelleted by high speed centrifugation (200,000 ϫ g, 45 min) in the presence of 1% Triton X-100. Previous work has shown that this Triton X-100 high speed pellet is also enriched for axonemal and other cytoskeletal proteins (18). In purified bovine OS, the prCAD C-terminal fragment is also associated with the Triton X-100 high speed pellet and can be highly enriched by differential centrifugation (Fig. 4E). By contrast, in the mouse, only ϳ50% of the prCAD C-terminal fragment is associated with the Triton X-100 high speed pellet.
Apical Accumulation of Uncleaved prCAD in rds(Ϫ/Ϫ) Photoreceptors-If proteolytic shedding of the prCAD ectodomain normally occurs in the OS, then proteolysis could be compromised in photoreceptors with defective OS biogenesis. To test this idea, we examined prCAD proteolysis and subcellular localization in mouse retinas lacking RDS/peripherin (referred to hereafter as RDS). RDS is an integral membrane  (␣N, right). Full-length (FL) prCAD was detected with both antibodies, whereas the N-terminal (NT) and C-terminal (CT) fragments were detected only with the anti-N-terminal and anti-C-terminal antibodies, respectively. All three of these prCAD polypeptides are absent from prCAD(Ϫ/Ϫ) retinas. In the prCAD(Ϫ/Ϫ) sample, the anti-N-terminal antibody detects a low abundance prCAD fragment of ϳ45 kDa produced from the partially deleted prCAD gene (see Fig. 5 in Ref. 6). A nonspecific band of ϳ60 kDa is also detected by the anti-N-terminal antibody in both WT and prCAD(Ϫ/Ϫ) retinas. For all of the immunoblots and protein gels in this figure, molecular size markers are (starting from the top) 180, 110, 80, 60, 48, 36, 24.5, and 19 kDa. B, the prCAD ectodomain is released from the photoreceptor cell membrane. Isolated mouse retinas were incubated in hypotonic medium (2 mM CaCl 2 ) at room temperature for 30 min. The retina was then removed from the medium, and the medium (containing soluble proteins and some detached OS) was centrifuged at 14,000 ϫ g for 10 min. The recovered retina (R), the pellet (P), and supernatant (S) fractions were analyzed by immunoblotting with anti-prCAD C-terminal (left) or N-terminal (right) antibodies. Most of the N-terminal fragment was washed away from the retina and was detected in the soluble fraction (right, compare R and S); none was detected in the particulate fraction (right, P). The C-terminal fragment was associated either with the retina (left, R) or with the particulate fraction (left, P); none was detected in the soluble fraction (left, S). C, schematic depiction of prCAD structure and processing. prCAD is a single-pass membrane protein with six extracellular cadherin repeats. The cleavage site, as predicted from the size of N-and C-terminal polypeptides, is within the sixth extracellular cadherin domain. The immunogens used for production of anti-N-and anti-C-terminal antibodies are indicated at the left. TM, transmembrane domain; N, N terminus; C, C terminus. D, mapping of the cleavage site in prCAD. Top, the sequence of mouse prCAD between amino acids 657 and 677 is shown with the approximate location of the cleavage site indicated above. Eight proteins, forming a nested deletion series encompassing the mouse prCAD C terminus and including an additional initiator methionine (red), were synthesized in vitro using a rabbit reticulocyte lysate. For each polypeptide, the first prCAD-derived amino acid and its position in the prCAD sequence are indicated on the right. Bottom, the in vitro translated polypeptides were mixed with a mouse retina protein extract, resolved by SDS-PAGE, and immunoblotted. The retina-derived C-terminal fragment (green dots and arrows) and the in vitro translated polypeptides (red dots) were detected with anti-mouse prCAD C-terminal antibody. E, the C-terminal fragment of bovine prCAD is associated with the Triton X-100-insoluble fraction of bovine rod OS. Left, schematic representation of OS fractionation into Triton X-100-soluble (S1 and S2) and -insoluble (P1 and P2) fractions; all manipulations were performed under dim red light to minimize bleaching and aggregation of rhodopsin. Right, SDS-PAGE analysis of the four fractions. Identical sets of samples were either stained with Coomassie Blue (CBB, left), showing mainly the distribution of rhodopsin (rho, arrow on the left), or immunoblotted (right) with an anti-bovine prCAD C-terminal antibody (␣bC), which recognizes the bovine prCAD C-terminal fragment (arrow on the right). The band at ϳ80 kDa is presumed to represent a cross-reacting protein; a faint band at ϳ125 kDa is likely to represent full-length prCAD. In bovine OS, the ratio of C-terminal fragment:full-length prCAD is higher than in mouse retina. ROS, rod OS. structural protein that resides at the OS disc rim and plays an essential role in disc morphogenesis (19,20). In the absence of RDS, the OS fails to form, and instead, large numbers of photoreceptor-derived membrane vesicles accumulate in the space between the photoreceptors and the RPE (21). Progressive photoreceptor degeneration occurs over the first several months of life.
Immunoblot analysis of 3-4-week-old WT and rds(Ϫ/Ϫ) retinas using the anti-C-terminal prCAD antibody shows a ϳ4-fold higher level of full-length prCAD in the rds(Ϫ/Ϫ) sample and a ϳ2-fold increase in the ratio of intact to cleaved prCAD polypeptides (Fig. 5, A and B). A similar analysis with the anti-Nterminal antibody shows a ϳ2-fold increase in both the absolute level and the proportion of full-length prCAD in the rds(Ϫ/Ϫ) sample. By contrast, immunoblot and/or immunohistochemical analysis of six representative rod OS proteins (rhodopsin, ABCR, the ␣ subunit of the cGMP-gated channel, ROM1, arrestin, and the ␣ subunit of transducin) shows decreased protein levels in the rds(Ϫ/Ϫ) retina in each case. By immunoblotting, rhodopsin, transducin, and ROM1 each show at least a 10-fold decrease at this age, consistent with an earlier analysis of rhodopsin levels (22). (The decrease in steady-state levels of ROM1 may be a special case, as ROM1 normally associates with RDS (23).) ABCR and the ␣ subunit of the cGMP-gated channel show at least a severalfold decrease as judged by immunoblotting and immunocytochemistry, respectively. Fig. 5A shows WT versus rds(Ϫ/Ϫ) immunoblots for arrestin, which shows the smallest decrease (ϳ2-fold) among the six OS proteins tested, and for the ␣ subunit of rod transducin.
Double labeling of 3-week-old WT retinas with anti-C-terminal prCAD antibodies and antibodies to the Na,K-ATPase, an IS marker, RP1, an axoneme marker, or CNGC, an OS marker (16), shows intense prCAD immunostaining confined to a small region distal to the Na,K-ATPase and proximal to RP1 and CNGC (Fig. 5, C, E, and G). (We note that the resolution of the light microscope does not allow an unambiguous determination of the extent to which these double immunolabeling signals might partially overlap.) By contrast, in the rds(Ϫ/Ϫ) retina, the zone of intense prCAD immunostaining occupies a considerably larger region distal to the IS, consistent with the increased level of prCAD observed by immunoblotting (Fig. 5, D,  F, and H). A nearly identical immunostaining pattern is seen with the anti-N-terminal prCAD antibody. The pattern of prCAD localization corresponds to the extracellular vesicles that accumulate in rds(Ϫ/Ϫ) retinas (21). Interestingly, in rds(Ϫ/Ϫ) photoreceptors, RP1 remains localized to discrete structures; these presumably correspond to the truncated axoneme or the connecting cilium (Fig. 5F).
The comparison of WT and rds(Ϫ/Ϫ) retinas shows that efficient proteolytic cleavage and turnover of prCAD require correct OS assembly. It is particularly striking that prCAD levels are elevated in the absence of RDS because RDS and prCAD show reciprocal patterns of OS localization; RDS occupies the mature disc rim region, from which prCAD is largely excluded, and prCAD is most concentrated at the free edges of nascent discs, where RDS is largely absent (24). The data presented here are consistent with a model in which the vesicles extruded from the apical face of rds(Ϫ/Ϫ) photoreceptors resemble nascent discs, and they support the idea that prCAD proteolysis is coupled to normal OS disc morphogenesis. DISCUSSION A Conserved Role for prCAD in OS Assembly-The present work provides several lines of evidence that implicate prCAD in the assembly of rod and cone outer segments. First, prCAD orthologues have now been identified in widely divergent ver-tebrates (mammals, birds, fish, and amphibians), but they have not been found in invertebrates (Caenorhabditis elegans and Drosophila melanogaster), a distribution that matches the distribution of eyes with ciliary photoreceptors. Second, prCAD is concentrated at the base of the OS, as determined by immunostaining following tissue treatments with a variety of denaturing agents and using antibodies to two different regions of the FIG. 5. Accumulation of full-length prCAD in the rds(؊/؊) retina. A, immunoblot analysis of total eyecup proteins from wild type and rds(Ϫ/Ϫ) mouse retinas. Equal quantities (30 g) of total eyecup proteins from WT or rds(Ϫ/Ϫ) retinas were probed with antibodies to the prCAD C terminus (␣C, top), rod arrestin (Arr, middle), or rod transducin ␣ (T␣, bottom). FL, full-length; CT, C-terminal fragment. B, the levels of full-length and C-terminal prCAD polypeptides were compared using 30 g of WT and either 15 or 7.5 g of rds(Ϫ/Ϫ) total eyecup proteins. Full-length prCAD is present at a ϳ4-fold greater level, and the prCAD C-terminal fragment is present at a ϳ2-fold greater level in rds(Ϫ/Ϫ) retinas. C-H, immunohistochemical localization of prCAD in rds(Ϫ/Ϫ) retinas. 10-m sections of fixed WT (C, E, and G) or rds(Ϫ/Ϫ) (D, F, and H) retinas were double-labeled with anti-prCAD C-terminal antibody (green) and anti-Na,K-ATPase antibody (red) to mark the IS, anti-RP1 antibody (red) to label the axoneme, or anti-CNGC antibody (red) to label OS membranes. In the rds(Ϫ/Ϫ) retina, prCAD accumulates distal to the shortened inner segments (D) and the deformed axoneme (F, arrowheads), presumably in OS-related vesicles released from the photoreceptors into the interphotoreceptor matrix (IPM), as indicated by its colocalization with CNGC (H, yellow indicates overlapping red and green signals). ONL, outer nuclear layer; DAPI, 4Ј,6diamidino-2-phenylindole. Scale bar, 5 m.
protein. Third, at the earliest stages of rod differentiation prCAD localizes to the site of the future OS before any of a variety of other OS proteins, including rhodopsin, arrestin, the ␣ subunit of the cyclic nucleotide gated channel, ROM1, and RP1. Fourth, prCAD undergoes a highly specific proteolytic cleavage, most likely at the base of the OS, which releases the ectodomain as a soluble fragment. Fifth, prCAD cleavage is partially inhibited, and full-length prCAD accumulates in the absence of orderly OS assembly in the rds(Ϫ/Ϫ) retina. In earlier work, prCAD(Ϫ/Ϫ) mice were observed to have disorganized OS but no defects in OS targeting of phototransduction proteins and only minimally perturbed electroretinogram responses (6). Taken together, these data imply that prCAD functions in OS assembly, although its exact role remains to be determined.
prCAD Ectodomain Shedding-Members of the cadherin superfamily interact with a variety of extracellular targets. Ecadherin mediates cell-cell adhesion through homophilic interactions (25). Protocadherins of the CNR class interact with Reelin (26), and cadherin-23 assembles its large extracellular domain into cables that constitute the tip links between adjacent stereociliary bundles in auditory hair cells (27). If the ectodomain of prCAD also interacts with binding partners on photoreceptors, the RPE, or the extracellular matrix, then its proteolytic release suggests that there is a specific stage in OS biogenesis when (a) the cell membrane disengages (via proteolysis) from an adhesive interaction and/or when (b) the prCAD ectodomain is released to diffuse to more distant binding targets.
Precedents exist in other systems for processes analogous to each of these two possibilities. The loss of cell-cell and cellmatrix attachment that accompanies apoptosis is associated with shedding of the ectodomains of E-cadherin and VE-cadherin (28,29). In Drosophila, both the Notch receptor and its ligand Delta are subject to juxtamembrane cleavage, which has emerged as an important regulatory mechanism in the Notch pathway (30 -32). In humans, the extracellular ligand binding domain of the growth hormone receptor is released from the cell surface via proteolysis, which decreases the responsiveness of target cells. Additionally, the presence of the released growth hormone receptor ectodomain in serum leads to a decrease in the concentration of free growth hormone (33). Analogous ectodomain shedding is observed with other single-span transmembrane receptors, including the tumor necrosis factor and the interleukin-6 receptors. Several growth factors and cytokines, including tumor necrosis factor-␣ and colony stimulating factor-1, which are initially synthesized as membrane-anchored precursors, are similarly released by proteolysis (34,35). For these membrane-anchored ligands, proteolytic release is presumed to extend their spatial range of action.
Cadherin family members also interact with a variety of intracellular targets. The cytosolic domains of classical cadherins bind to ␤-catenin and organize cytoskeletal attachments to regions of plasma membrane adjacent to cell-cell contacts (11). The CNR protocadherins are presumably involved in signal transduction via their interaction with the cytosolic Fyn tyrosine kinase (36). The relative resistance of the prCAD C-terminal fragment to Triton X-100 extraction suggests that the cytosolic domain of prCAD may interact with cytoskeletal proteins, but the nature of these interactions and their response to ectodomain proteolysis remain to be determined.
A major goal for future work will be to identify the protease responsible for cleaving prCAD and to determine its subcellular localization and regulation. Work on ligand and receptor shedding in other systems suggests that this enzyme will be a member of the large family of metalloproteinases, and more specifically a member of either the ADAM (a disintegrin and metalloproteinase) or matrix metalloproteinase subfamilies (34,35).
Implications for the Molecular Mechanism of OS Assembly-Although the exact role of prCAD in OS assembly remains uncertain, the discovery of prCAD ectodomain shedding suggests a model in which proteolysis of prCAD contributes to the thermodynamic driving force for OS assembly. In particular, if the correct stacking of nascent OS discs or the "zippering up" of the plasma membrane around them were coupled to proteolysis of prCAD, then these reactions would be rendered effectively irreversible. A number of precedents exist in which proteolysis ensures that a biochemical or cell biological process proceeds unidirectionally. These include destruction of cyclins to drive the eukaryotic cell cycle (37), cleavage of the cohesin protein to initiate chromosome segregation at mitosis (38), conversion of fibrinogen to fibrin to initiate blood clotting (39), and the activation of transcription factors by proteolysis for sporulation in Bacillus subtilis (40) or for the transcription of cholesterol biosynthetic genes in vertebrates (41). It seems reasonable to speculate that OS assembly, like many other cell biological processes, might be driven, at least in part, by the near irreversibility of proteolysis.