(Received for publication, December 27, 1996, and in revised form, February 3, 1997)
From the Department of Biochemistry and Molecular Biology, University of British Columbia, British Columbia, Vancouver V6T 1Z3, Canada
Outer segments of mammalian rod photoreceptor cells contain an abundantly expressed membrane protein that migrates with an apparent molecular mass of 220 kDa by SDS-gel electrophoresis. We have purified the bovine protein by immunoaffinity chromatography, determined its primary structure by cDNA cloning and direct peptide sequence analysis, and mapped its distribution in photoreceptors by immunocytochemical and biochemical methods. The full-length cDNA encodes a 2280-amino acid protein (calculated molecular mass of 257 kDa) consisting of two structurally related, tandem arranged halves. Each half consists of a hydrophobic domain containing six putative transmembrane segments followed by an ATP-binding cassette. A data base homology search showed that the rod outer segment 220-kDa protein is 40-50% identical in amino acid sequence to the ABC1 and ABC2 proteins cloned from a mouse macrophage cell line. Photoaffinity labeling with 8-azido-ATP and nucleotide inhibition studies confirmed that both ATP and GTP bind to this protein with similar affinities. Concanavalin A labeling and endoglycosidase H digestion indicated that the rod outer segment protein contains at least one carbohydrate chain. Immunocytochemical and biochemical studies have revealed that the 220-kDa glycoprotein is distributed along the rim region and incisures of rod outer segment disc membranes. From these studies we conclude that the 220-kDa glycoprotein of bovine rod outer segment disc membranes or Rim ABC protein is a new member of the superfamily of ABC transporters and is the mammalian homolog of the frog photoreceptor rim protein.
Outer segments of rod and cone photoreceptor cells are highly specialized subcellular structures that serve as the site for phototransduction. The rod outer segment (ROS)1 consists of a highly ordered axial array of over 500 discs surrounded by a separate plasma membrane. Each disc is made up of two closely spaced lamellar membranes circumscribed by a hairpin loop or rim region (1). The continuous disc membrane encloses a compartment called the disc lumen or intradiscal space. The perimeter of the disc is interrupted by one or more incisures that penetrate toward the center of the disc. Filamentous structures extend from the rim region of the discs to adjacent discs and to the plasma membrane (2-4).
The protein composition of ROS disc membranes differs from that of the plasma membrane (5). Although both membranes contain rhodopsin as the major membrane protein, the cGMP-gated channel (6), the Na/Ca-K exchanger (7), and the GLUT-1 glucose transporter (8) are predominantly, if not exclusively, present in the ROS plasma membrane. Guanylate cyclase (9), the peripherin/rds-rom-1 complex (10-12) and a high molecular mass (220-290 kDa) glycoprotein (5, 13) are present in disc membranes. The peripherin/rds-rom-1 complex in mammalian ROS and the 290-kDa glycoprotein or rim protein of frog ROS each constitute 3-4% of the total ROS membrane protein and are localized to the rim region and incisures of rod and cone disc membranes (10-15). The frog rim protein has been reported to undergo a light-activated phosphorylation reaction (16); however, the role of this reaction in photoreceptor cell function remains to be determined. An abundant 220-kDa membrane glycoprotein is also present in mammalian ROS disc membranes (17, 18). Photoaffinity labeling studies indicate that this protein specifically binds ATP and GTP (19). Although the distribution of the 220-kDa protein in mammalian ROS disc membranes has not yet been determined, it is generally thought that this protein may be the homolog of the frog 290- kDa rim protein.
The function of the frog rim protein is not currently known. Its large size and distribution along the rim region of ROS discs, however, has led to the suggestion that it may constitute the filament-like structures that link adjacent discs together and/or connect the discs to the plasma membrane (2, 4). As a first step in defining the role of the rim protein in ROS structure and function and its possible involvement in retinal degenerative diseases, we have cloned and sequenced the cDNA for the bovine 220-kDa protein and studied its molecular properties and cellular distribution. In this paper we report that the bovine 220-kDa glycoprotein is a novel member of the ATP-binding cassette (ABC) transporter superfamily. Like the frog rim protein, it is localized to the rim region of bovine ROS disc membranes.
A ZAPII oligo(dT)-primed retinal cDNA
library was a generous gift of W. Baehr, and the pcDNAII
random-primed cDNA library was made from bovine retinal poly(A)
RNA. Monoclonal antibodies PMc 1D1 and PMs 4B2 against the
- and
-subunits of the bovine cGMP-gated channel, respectively, and PMe
2D9 against the Na/Ca-K exchanger have been reported (6, 20, 21).
ROS were isolated
under dim red light from freshly dissected or previously frozen retinas
by continuous sucrose gradient centrifugation; ROS membranes were
isolated by hypotonic lysis of ROS as described previously (5). The ROS
membranes resuspended in 0.01 M Tris·HCl, pH 7.4, at a
protein concentration of 8-10 mg/ml were either used immediately or
stored at 70 °C in light-tight vials. Disc membranes were
separated from ricin-gold-dextran labeled plasma membranes by mild
trypsin digestion and sucrose gradient centrifugation (5).
Monoclonal antibody Rim 3F4 against the bovine 220-kDa disc rim protein was generated from a BALB/c mouse that had been repeatedly immunized with a partially purified preparation of the 220-kDa protein by standard procedures (22). The immunogen was prepared by solubilizing ROS in 0.02 M Tris·HCl, pH 7.4, 18 mM CHAPS, and 0.15 M NaCl and passing this solution through a DEAE-Fractogel TSK column equilibrated in the same buffer. The flow-through fraction containing the 220-kDa protein, rhodopsin, and other minor proteins was devoid of the cGMP-gated channel and the highly antigenic Na/Ca-K exchanger (23). Culture fluid from Rim 3F4 hybridoma cells was routinely used at a dilution of 1:40 for Western blotting and 1:5 for immunocytochemical studies. For immunoaffinity chromatography the Rim 3F4 antibody was purified from ascites fluid by precipitation with 50% ammonium sulfate at 4 °C followed by DEAE-Sephacel ion exchange chromatography.
Polyclonal antibody PrimT1 was obtained by immunizing a New Zealand White rabbit with the 220-kDa rim protein purified on a Rim 3F4 antibody-Sepharose 2B column (see below). For use in Western blotting, immunocytochemistry, and library screening, residual contaminating antibodies to other ROS proteins were removed by treating 5 ml of the PrimT1 antiserum containing 1% (v/v) Triton X-100, with 2 ml of Sepharose 2B conjugated to 4 mg of Triton X-100-solubilized ROS proteins, in which the 220-kDa protein had been first removed by passage through a Rim 3F4 antibody-Sepharose 2B column. For electron microscopy, the PrimT1 antibody was purified by the method of Burton et al. (24).
Isolation of cDNA ClonesApproximately 100,000 plaque
forming units from an oligo(dT)-primed bovine retinal cDNA library
were screened with the preadsorbed PrimT1 antiserum diluted 1:5000
(25). Of 11 positive clones detected, 6 were subcloned and found to
contain identical sequences. A synthetic oligonucleotide from the 5
end of the longest clone (
2) was used to rescreen the library. The
5
end of the longest clone (
15) was used to screen the
random-primed retinal cDNA library. Overlapping clones were
isolated and sequenced in both directions by a combination of the
dideoxy chain termination method using [
-35S]dATP and
automated fluorescent sequencing (Licor, Inc.).
For Northern blots, 10 µg of total RNA from bovine retinas were hybridized at 65 °C for 2 h in Rapid-hyb buffer (Amersham) with 32P-labeled probes synthesized by random-priming (Pharmacia T7Quick Prime Kit). Probes were derived from an EcoRI/SacI fragment of clone p41 (nucleotides 3707-4038 of the full-length cDNA) and an EcoRI/HpaII fragment of clone p55 (nucleotides 2498-2780).
Peptide Sequencing and Amino Acid AnalysisFor N-terminal peptide sequencing and amino acid analysis, 100-200 µg of the immunoaffinity purified 220-kDa protein or its 115- and 120-kDa tryptic fragments (see below) were subjected to SDS-polyacrylamide gel electrophoresis and electrophoretically transferred to Immobilon membranes. The immobilized protein was subjected to acid hydrolysis for amino acid analysis or Edman degradation for N-terminal microsequencing.
For internal peptide sequences, 80 mg of purified ROS membranes were resuspended in 10 ml of 0.02 M Tris·HCl, pH 7.4, and 0.02 M CaCl2. Ten ml of L-1-tosylamido-2-phenylethyl chloromethyl ketone-trypsin (0.4 µg/ml) were added and the reaction was allowed to proceed for 5 min at room temperature and then for 30 min on ice. The membranes were sedimented at 12,000 rpm in a Sorvall SS34 rotor for 10 min and subsequently washed 3 times with 20 ml each of 10 µg/ml soybean trypsin inhibitor in the same buffer. The final pellet was solubilized at 2 mg/ml in 0.02 m Tris·HCl, pH 7.4, 0.15 M NaCl, 1% (v/v) Triton X-100, and 10 µg/ml soybean trypsin inhibitor and incubated with 1 ml of Rim 3F4 antibody-Sepharose 2B for 1 h. After the matrix was washed in the same buffer containing 0.2% Triton X-100, the tryptic fragments of the 220-kDa protein were eluted with 0.6 ml of 0.1 M glycine·HCl, pH 3.5, 0.15 M NaCl, and 0.2% (v/v) Triton X-100. The eluted fraction was concentrated to 0.1 ml using a Centricon 30 (Amicon) and subjected to SDS-polyacrylamide gel electrophoresis. The 120- and 115-kDa tryptic fragments were electrophoretically transferred to Immobilon or nitrocellulose membrane and either directly subjected to N-terminal sequence analysis or digested with endoproteinase Lys-C (Boehringer Mannheim) by standard methods (26). Similarly, the immunoaffinity purified rim protein was transferred to nitrocellulose membrane and digested with sequence grade trypsin (Boehringer Mannheim). In either case, peptides were isolated by high pressure liquid chromatography using a narrow-bore C18 reverse-phase column and subjected to N-terminal microsequencing. Amino acid analysis and N-terminal peptide gas-phase sequencing were carried out by the Tripartite Microanalytical Center at the University of Victoria.
Immunoaffinity Purification of the 220-kDa Rim ProteinThe Rim 3F4 antibody was coupled to CNBr-activated Sepharose 2B at a concentration of 2 mg of protein/ml of packed beads (20). The 220-kDa rim protein was purified by incubating 1 mg of ROS membranes, solubilized in 0.5 ml of 18 mM CHAPS or 1% Triton X-100 in 0.02 M Tris·HCl, pH 7.4, 5 mM MgCl2, with 50 µl of Rim 3F4-Sepharose 2B beads at 4 °C for 1 h. The unbound fraction was isolated by low speed centrifugation through an Ultrafree filter unit (Millipore Corp.); the beads were then washed twice with 0.5 ml of solubilization buffer. Finally, the bound protein was released from the beads with 0.1 ml of solubilization buffer containing 0.1 mg/ml 3F4 peptide.
Endoglycosidase Hf Treatment and 8-Azido-[For endoglycosidase treatment, 100 µl of ROS membranes solubilized in 10 mM CHAPS, 0.02 M Tris·HCl, pH 7.4, 5 mM MgCl2 was treated with endoglycosidase Hf (New England Biolabs) according to the manufacturer's instructions and subjected to SDS-polyacrylamide gel electrophoresis and Western blotting.
The 220-kDa rim protein was photoaffinity labeled with
8-azido-[-32P]ATP (ICN; 15 Ci/mmol) using a modified
procedure of Shuster et al. (19). The reaction was performed
under dim red light in 25 µl of 0.02 M Tris·HCl, pH
7.4, 0.5 mM ZnCl2, and 0.5 mM MgCl2, 3 µM
8-azido-[
-32P]ATP, and 2 µg of ROS membranes or
approximately 20 ng of Triton X-100 solubilized immunoaffinity purified
220-kDa rim protein with or without the competing nucleotide. The
samples were irradiated for 5 min with ultraviolet light (Mineralight
UV Lamp; 254 nm) at a distance of 3 cm, and the reaction was terminated
by the addition of 25 µl of SDS buffer for SDS-gel electrophoresis.
Gels were stained with Coomassie Blue, dried, and exposed to either x-ray film or a phosphoimaging screen. Phosphoimages were analyzed quantitatively by normalizing the intensity of the bands relative to
the sample which contained no competitive nucleotide.
A glutathione S-transferase fusion peptide
containing the C-terminal 128 amino acids of the 220-kDa rim protein
was first used to localize the epitope for the Rim 3F4 antibody. This
was carried out by subcloning a BamHI/XhoI
fragment of clone 4 (nucleotides 6458-7609) into pGEX-I (Pharmacia)
for expression in Escherichia coli (27); antibody binding
was detected by Western blotting. To more precisely map the epitope of
the Rim 3F4 antibody, peptides (9 amino acids long, 2 amino acid
overlap) spanning the C-terminal 128 amino acids were synthesized using
the Epitope Scanning Kit (Cambridge Research Biochemicals) for analysis
by an enzyme-linked immunoassay. The 3F4 peptide corresponding to amino
acids 2252-62 of the rim protein was synthesized by solid phase
peptide synthesis (28).
SDS-gel
electrophoresis was carried out as described previously (5). For
Western blotting, proteins were electrophoretically transferred onto
Immobilon membranes using the buffer system of Towbin et al.
(29). The membranes were labeled with the Rim 3F4 or PrimT1 antibody
for 1 h at room temperature followed by sheep anti-mouse or donkey
anti-rabbit Ig-F(ab)2 conjugated to horseradish
peroxidase. Alternatively, blots were labeled with concanavalin A
conjugated to horseradish peroxidase (Sigma). In either case binding
was detected using the ECL system (Amersham). Protein concentrations
were determined using the BCA assay (Pierce).
For light microscopy, cryosections of bovine retina were labeled with Rim 3F4 or PrimT1 antibodies followed by goat anti-mouse Ig or goat anti-rabbit Ig conjugated to horseradish peroxidase or Cy3 (8). In control studies Rim 3F4 antibody labeling was carried out in the presence of excess 3F4 peptide (0.1 mg/ml).
For electron microscopy, glutaraldehyde-fixed, LR-White resin embedded bovine retina sections were labeled with the purified PrimT1 antibody overnight and subsequently labeled with 10-nm diameter goat anti-rabbit Ig-gold (12).
Monoclonal antibody Rim 3F4 and polyclonal antibody
PrimT1 were used as probes to clone, characterize, and localize the
major 220-kDa protein of bovine ROS membranes. The specificity of these antibodies is shown by Western blotting in Fig. 1. Both
antibodies labeled a single protein having an apparent molecular weight
of 220,000 in both ROS membranes and the flow-through (unbound
fraction) obtained when Triton X-100 solubilized ROS membranes were
passed through a DEAE anion exchange column. The labeled protein
co-migrated with the prominent Coomassie Blue-stained 220-kDa protein
present in both samples. In contrast, monoclonal antibodies PMe 2D9
against the 230-kDa Na/Ca-K exchanger (21) and PMs 4B2 against the
240-kDa -subunit of the cGMP-gated channel (20, 30) each labeled a
high molecular mass polypeptide in ROS membranes, but not in the
unbound DEAE fraction (Fig. 1).
Cloning and Primary Structure of the 220-kDa Protein
The
220-kDa protein was cloned by initially screening an oligo(dT)-primed
bovine retinal cDNA expression library with the PrimT1 antibody. Of
the six cDNA clones isolated in the initial screen, the longest
clone (2) was 1.8 kb in length and had a polyadenylation signal and
a poly(A) tail. Rescreening of the oligo(dT)-primed library and a
random primed retinal cDNA library with oligonucleotide probes
resulted in the isolation of several overlapping clones (Fig.
2). Upstream clone (p114) contained an ATG start codon
with an adjacent upstream Kozak consensus sequence (31). The nucleotide
sequence immediately downstream from this ATG coded for an amino acid
sequence that was identical to that determined by direct N-terminal
sequencing of the isolated 220-kDa protein (see below).
The overlapping clones contained an open reading frame of 6843 base
pairs that encodes a protein of 2280 amino acids having a calculated
molecular mass of 257,000 daltons (Fig. 3). The
predicted amino acid sequence consists of two structurally related,
tandemly arranged halves: each half of the protein has a cluster of
hydrophobic segments followed by a conserved ATP-binding cassette (32). Hydropathy profiles (33) predict at least 12 transmembrane segments, six in each half of the protein. Fourteen consensus sequences for
N-linked glycosylation are present (Fig. 3).
Data base homology searches indicate that the ROS 220-kDa protein is a member of the ABC transporter superfamily. Overall, it is 40-50% identical in amino acid sequence to the ABC1 and ABC2 proteins from a mouse macrophage cell line (34) and 37% identical to ABC-C, a putative transporter recently cloned from a human medullary thyroid carcinoma cell line (35). The ABC-binding domains of these proteins display an even higher degree of identity (>60%), particularly within the two short stretches of the Walker motifs and the active transport signature located upstream of the Walker B pattern (32). The ROS 220-kDa protein is more distantly related to other well studied members of the ABC transporter superfamily such as the cystic fibrosis transregulator and the multi-drug resistance protein (MDR1 P-glycoprotein), showing a sequence identity of 20-22% overall and 20-26% for the ABC domains. A 43% sequence identity is observed between the N- and C-terminal ABC domains of the bovine ROS 220-kDa protein, a value similar to that between the two cassettes within ABC1 and MDR1 P-glycoprotein. However, a high degree of conservation is found for the C-terminal ABC domain from different mammalian ROS 220-kDa proteins; for example, the bovine and human proteins are 94% identical in this region.2
Northern Blot Analysis of Retinal RNAThe size of the
mRNA transcripts for the 220-kDa protein was estimated by Northern
blotting. As shown in Fig. 4, an
EcoRI/SacI fragment from clone p41 (see Fig. 2)
hybridized to transcripts of ~9.5 and 6.8 kb in length. An upstream
fragment from clone p55 gave a similar labeling pattern (data not
shown).
Peptide Sequencing and Amino Acid Composition of the 220-kDa ROS Protein
Peptide sequences of the 220-kDa ROS protein were
obtained to confirm that the cDNA clones code for the protein
sequence of the major 220-kDa protein of bovine ROS membranes. The
N-terminal 18 amino acids of the immunoaffinity purified 220-kDa
protein was found to be identical to a sequence derived from clone p114 and immediately downstream from a methionine (Fig. 3). Mild trypsin digestion of the 220-kDa protein in ROS membranes followed by purification on a Rim 3F4 antibody affinity column resulted in two
large peptide fragments having apparent molecular masses of 120 and 115 kDa (see Fig. 8). The N-terminal sequence of the 120-kDa fragment was
identical to the N-terminal sequence of the full-length 220-kDa,
whereas the sequence of the 115-kDa fragment was identical to a
sequence starting at position 1311, close to the middle of the 220-kDa
protein (Fig. 3). These results indicate that mild trypsin digestion
cuts the 220-kDa protein essentially in half, producing tryptic
fragments corresponding to two homologous domains. Trypsin digestion
has also been reported to split MDR1 P-glycoprotein into two homologous
halves (36).
In addition, nine smaller peptide fragments were derived from trypsin and/or endoproteinase Lys-C digestion of the purified, detergent-solubilized 220-kDa protein. The N-terminal sequences of these peptides were identical to sequences found within the 220-kDa protein sequence deduced from cDNA cloning (Fig. 3). The sequence of peptide number 21 was of particular interest. Although 36 of the 37 amino acids of this peptide correspond to positions 1564-1600 of the full-length 220-kDa protein sequence, the amino acid at position 23 of the peptide sequence (position 1586 of the 220-kDa sequence) could not be identified. Since this residue is an Asn that is part of a N-linked glycosylation consensus sequence, it is likely that this residue is covalently modified with an oligosaccharide chain, and therefore, not identifiable by sequencing.
Further evidence that the immunoaffinity purified 220-kDa protein is the same as the protein encoded by the cloned cDNA was obtained from amino acid analysis. In general, the experimentally determined amino acid composition of the purified 220-kDa protein is within 15% of the amino acid composition determined from the deduced protein sequence (data not shown).
Immunoaffinity Purification of the 220-kDa Protein from Bovine Rod Outer SegmentsThe 220-kDa protein could be efficiently purified
from detergent-solubilized ROS by immunoaffinity chromatography. As
shown in Fig. 5A, the 220-kDa protein was
quantitatively adsorbed to a Rim 3F4-Sepharose column and eluted with
an 11 amino acid peptide corresponding to the Rim 3F4 epitope localized
near the C terminus of the protein (see Fig. 3). No other prominent
proteins in the immunoaffinity purified sample were detected on
SDS-polyacrylamide gels stained with Coomassie Blue.
The 220-kDa Protein in Bovine ROS Is Glycosylated
Concanavalin A labeling and endoglycosidase H digestion were carried out to determine if the ROS 220-kDa protein is glycosylated. As shown in Fig. 5A, concanavalin A labeled the 220-kDa protein in both ROS membranes (lane a) and immunoaffinity purified preparations (lane c). Monomeric and oligomeric forms of rhodopsin were also labeled with concanavalin A in ROS membranes (lane a) and in the unbound fraction from the affinity column (lane b). The immunoaffinity purified 220-kDa protein fraction, however, appears to be largely free of rhodopsin.
Endoglycosidase H digestion was carried out to further evaluate the extent to which the 220-kDa protein is glycosylated. As shown in Fig. 5B, only a slight increase in mobility of the 220-kDa protein was observed after treatment with endoglycosidase H. These results indicate that the 220-kDa protein contains at least one N-linked oligosaccharide chain, but it is unlikely to be highly glycosylated.
Photoaffinity Labeling of the 220-kDa Protein with 8-Azido-[Photoaffinity labeling was
carried out to determine if the putative ATP-binding domains of the
220-kDa protein bind nucleotides. As shown in Fig.
6A, 8-azido-[-32P]ATP
covalently labeled the 220-kDa protein in both ROS membranes and
immunoaffinity purified preparations. Labeling could be largely inhibited by the addition of 1 mM cold ATP or 1 mM GTP (Fig. 6, A and B). These
studies are in agreement with the earlier studies of Shuster et
al. (19) indicating that a high molecular weight polypeptide in
ROS is covalently labeled by this reagent.
Inhibition studies were carried out to determine the binding
specificity of the ROS 220-kDa protein for ATP and GTP. As shown in
Fig. 6B both ATP and GTP inhibited
8-azido-[-32P]ATP labeling. Half-maximum inhibition
occurred at a slightly lower GTP concentration (~70 µM)
than ATP concentration (~95 µM), in general agreement
with an earlier study (19). These results further support the view that
the ROS 220-kDa protein is an ABC transporter capable of binding both
ATP and GTP.
The distribution of the 220-kDa protein in retina tissue
was examined using immunocytochemical labeling methods. The Rim 3F4 antibody labeled the outer segments of rod, but not cone, photoreceptor cells in bovine retinal cryosections (Fig.
7A). Weak staining was observed in the inner
segment layer, but no staining was detected in other retinal cells. A
similar pattern of labeling was observed for rat retina and when the
PrimT1 polyclonal antibody was used as primary antibody (data not
shown). Labeling was specific since addition of excess 3F4
peptide inhibited Rim 3F4 antibody labeling (Fig. 7B).
The subcellular distribution of the 220-kDa protein in ROS was studied using immunogold labeling methods for electron microscopy. When LR White resin embedded retinal sections were labeled with the PrimT1 antibody, immunogold particles were found to be distributed along the periphery and incisures of the ROS where the rim region of the discs are in close proximity to the plasma membrane (Fig. 7, C and D). This labeling pattern is similar to that described by Papermaster et al. (13) for a 290-kDa rim protein in frog outer segments. The Rim 3F4 monoclonal antibody did not label these sections presumably due to inaccessibility of the 3F4 epitope in resin embedded samples.
Since postembedding immunogold labeling studies could not unambiguously
establish whether the 220-kDa protein was localized to the rim region
of disc membranes or the plasma membrane, disc membranes were isolated
from trypsin-treated ROS (5) for Western blot analysis. As shown in
Fig. 8, the Rim 3F4 monoclonal antibody intensely
labeled the 220-kDa protein in ROS membranes and its 115-kDa tryptic
fragment in highly purified disc membranes. A less intensely labeled
120-kDa fragment was also observed in the disc membrane fraction; this
fragment is most likely derived from trypsin cleavage at a secondary
site within the 220-kDa. Absence of plasma membrane contamination in
the disc fraction was verified in labeling studies employing a plasma
membrane specific cGMP-gated channel antibody (6). As shown in Fig. 8,
the anti-channel antibody PMc 1D1 labeled the 63-kDa -subunit of the
cGMP-gated channel in ROS membranes, but no labeling of the trypsinized
channel subunit (6) was observed confirming its absence in the disc fraction. The immunocytochemical and biochemical studies taken together
indicate that the bovine 220-kDa protein is preferentially, if not
exclusively, localized to the rim and incisures of ROS discs.
Although the high molecular weight rim protein had been shown to be a major membrane protein of ROS almost 20 years ago (13), no detailed studies have been reported on its structural properties since this time. In this study we have purified the bovine 220-kDa rim protein, mapped its distribution in ROS, and determined its primary structure by cDNA cloning and direct peptide sequencing.
The bovine ROS 220-kDa glycoprotein has been localized to the rim region and incisures of bovine ROS disc membranes by a combination of immunogold labeling and biochemical techniques. On this basis we conclude that this protein is the mammalian homolog of the 290-kDa frog rim protein first reported by Papermaster et al. (13). However, a notable difference in labeling of photoreceptors is apparent. Whereas a polyclonal antibody to the frog rim protein labels both frog rod and cone outer segments (15), antibodies against the bovine rod protein used in the present study labels only outer segments of rod photoreceptors. It appears that either mammalian cone photoreceptors do not express a rim protein, or more likely, the cone rim protein is immunochemically different from the rod rim protein and may be encoded by a separate gene. Alternatively, the rim protein in cones may be less abundantly expressed or less accessible such that cone labeling is not detected using the reagents employed in this study.
The Rim 3F4 monoclonal antibody has been used to purify the rim protein from Triton X-100 and CHAPS solubilized ROS in a single step. Analysis of the purified protein by SDS-gel electrophoresis indicates that no other intensely stained proteins are present in this preparation. Furthermore, the amino acid composition of the purified rim protein is in close agreement with the composition calculated from its sequence. On the basis of these results, it is likely that the rim protein is composed of a single type of subunit and, at least in detergent solution, is not tightly associated with other proteins in stoichiometric amounts. It remains to be determined if the rim protein exists as a single polypeptide or as a protein complex consisting of several identical subunits.
Primary structural analysis indicates that the rim protein is a new member of the superfamily of ABC transporters. Like other eukaryotic ABC transporters (37), the rod rim protein consists of two structurally related halves, each half consisting of a hydrophobic domain with multiple predicted membrane spanning segments followed by a highly conserved ATP-binding cassette. The presence of nucleotide binding folds is consistent with earlier photoaffinity labeling studies (19) and studies carried out here, indicating that the rim protein specifically binds both ATP and GTP. The large size of the rim protein (2264 amino acids; calculated Mr ~257,000) makes this protein one of the largest ABC transporters reported to date.
Current models of most eukaryotic ATP transporters have 12 transmembrane segments, six in each hydrophobic domain, and both the
ATP-binding cassettes and the N and C terminus localized on the
cytoplasmic side of the membrane (37). At least 12 membrane spanning
segments (see Fig. 3) are predicted for the rim protein from hydropathy
plots (33). The rim protein is glycosylated like other well
characterized ABC transporters. However, the number of
N-linked oligosaccharide chains is likely to be small since digestion with endoglycosidase H has only a minor effect of the mobility of the rim protein on SDS gels. Although the site(s) of
glycosylation has yet to be determined directly, the inability to
detect Asn-1586 in peptide number 21 by direct peptide sequencing strongly suggests that this residue, which is part of a consensus sequence for N-linked glycosylation, is covalently bound to
an oligosaccharide chain. Therefore, one predicts that this segment resides on the intradiscal or lumen side of the disc membrane. On the
basis of these studies, we propose a working model for the membrane
organization of the Rim ABC protein as shown in Fig. 9.
A novel aspect of this model is the presence of a large intradiscal loop between each of the first two membrane spanning segments for each
half of the protein. Such segments are not predicted in models for
other ABC transporters such as P-glycoprotein and cystic fibrosis
transregulator and contribute to the unusually large size of the Rim
ABC protein relative to most other ABC transporters. Relatively large
intradiscal segments are also found in the peripherin/rds and rom-1
subunits and have been suggested to be involved in the formation and
stabilization of the highly curved rim region of the disc membranes
(11, 38).
The rod rim protein is most closely related to ABC1 and ABC2
transporters of macrophages (34) as shown in the dendrogram comparing
the C-terminal nucleotide-binding domains of representative ABC
transporters (Fig. 10). Other related proteins include
an ABC-C protein recently cloned from a human medullary thyroid
carcinoma cell line (35), the NodI protein involved in nodulation of
plant roots by Rhizobium (39), and the YLH4 putative
transporter from Caenohabditis elegans (40). The rim protein
is more distantly related to MDR1 P-glycoprotein, involved in the
active extrusion of drugs from cells (41), and cystic fibrosis
transregulator, which functions as a chloride channel and has been
linked to cystic fibrosis (42).
The function of the rim protein is not currently known. On the basis of its large size and location along the disc rims, it had been suggested that this protein may constitute the filamentous structures that link adjacent discs together and/or connects discs to the plasma membrane (2, 4). The finding that the Rim ABC protein is a member of the ABC transporter superfamily makes this less likely. However, it could play a role in anchoring such filaments to the disc rim membrane.
In a number of cases, ABC transporters have been shown to be involved in the active transport of a variety of hydrophobic molecules across membranes including drugs (43), lipids (44), metabolites (45), and peptides (46, 47). The mouse MDR2 P-glycoprotein and its human counterpart (MDR3) have also been reported to function as a flippase promoting the transfer of phosphatidylcholine from the inner to the outer leaflet of the plasma membranes of hepatocytes (48). It is possible that the Rim ABC protein may also function in the transport of specific hydrophobic substrates between the cytoplasm and intradiscal side of the disc membrane. Retinal derivatives which are important in phototransduction are hydrophobic and might be considered as possible substrates for translocation into or out of the disc lumen. However, there is no evidence that retinal derivatives accumulate inside the discs or require an active process for transport across cell membranes. Other possible substrates include lipids or peptides which after translocation to the intradiscal space may interact with regions of proteins exposed on the lumen side of the disc membrane.
Some ABC transporters have been reported to be involved in morphological processes. The ABC1 protein of mouse macrophage and the ced-7 protein of C. elegans have been implicated in the process of engulfment of cell corpses generated by apoptotic cell death (49, 50), and the NodI protein has been linked to the process of Rhizobium nodulation in roots (39). The 220-kDa protein, being more similar in structure to these proteins, may function in the morphogenesis of the rim region of disc membranes. Clearly, more studies are required to determine the role of the rim protein in outer segment structure and function.
In summary we have shown that the major 220-kDa glycoprotein of mammalian ROS is a member of the superfamily of ABC transporters, and on the basis of its size and location, it appears to be the mammalian homolog of the frog rim protein. To our knowledge, the Rim ABC protein is the first ABC transporter found in vertebrate retina.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBankTM/EMBL Data Bank with accession number(s) U90126[GenBank].
We thank Theresa Hii for expert technical assistance in the generation of the Rim 3F4 monoclonal antibody.