Studies on the glycoprotein associated with Rh (rhesus) blood group antigen expression in the human red blood cell membrane.

The blood group Rh antigens are associated with non-glycosylated 30-kDa erythrocyte membrane proteins (the Rh30 polypeptides) and the Rh glycoprotein. We used antipeptide antibodies to study the Rh glycoprotein in human erythrocyte membranes. The Rh glycoprotein was present in Rhnull U+ve cells. However, the N-glycan chain of the Rh glycoprotein in Rhnull U+ve cells was smaller than in normal cells. In contrast, the N-glycan chain of the Rh glycoprotein was larger than normal in glycophorin B-deficient red cells. We suggest that this observation reflects a lower rate of movement of newly synthesized Rh glycoprotein through intracellular membranes to the cell surface in the absence of glycophorin B, and that in normal red cells glycophorin B facilitates the movement of the Rh protein complex to the cell surface. Our results provide evidence for the intracellular interaction of at least three proteins, the Rh glycoprotein, Rh30 polypeptides, and glycophorin B during the biosynthesis and cell surface expression of the Rh complex. These observations are likely to be important for the successful design of expression systems for the blood group Rh antigens.

Despite the importance of the human Rh blood group antigens in blood transfusion and immunohematology (Mollison et al., 19871, it is only recently that some of the erythrocyte membrane polypeptides involved in the expression of these antigens have been identified (reviewed by Agre and Cartron (1991) and Anstee and Tanner (1993)). Immunoprecipitation studies with a range of Rh-specific antisera and monoclonal antibodies have identified two unglycosylated Rh-related integral membrane proteins. Anti-RhD immunoprecipitates a protein with an apparent molecular mass of 30 kDa which lacks carbohydrate (Moore et al., 1982;Gahmberg, 1983) designated the D30 polypeptide. Anti-Rhc, anti-RhE, and mouse monoclonal antibodies of the R6A type (which react with red cells irrespective of their common Rh phenotype) immunoprecipitate slightly larger components of apparent molecular mass of 32 kDa (designated the CcEego and R6A32 polypeptides; Moore et al. (1982), Ridgwell et al. (19831, Moore and Green (19871, Avent et al. (1988a, 1988b, reviewed by Anstee and Tanner (1993)). The CcEeao and R6A32 polypeptides are probably the same molecules. These proteins are collectively described here as the Rhz0 polypeptides.
Rh-specific immunoprecipitates also contain a glycoprotein which migrates as a d i f i s e band on SDS-PAGE' due to heterogeneity in the glycan chain (Moore and Green, 1987;Avent et al., 1988a). The SDS gel mobility of the glycoprotein from surface-labeled red cells which coprecipitates with the D30 polypeptide is slightly slower than that coprecipitated with the CeEeso polypeptide or the R6A32 polypeptide (Moore and Green, 1987;Avent et al., 1988b). However, both glycoproteins have the same N-terminal amino acid sequences (Avent et al., 1988b). Although the N-terminal sequence of the Rh-related glycoproteins is different to that of the Rh30 polypeptides, the two sequences show some homology. A cDNA clone, Rh50A, representing a member of the Rh-related glycoprotein family has been isolated (Ridgwell et al., 1992). The Rh50A cDNA predicts a 409-amino acid hydrophobic, N-glycosylated integral membrane protein which is homologous with, and has a topology very similar to that predicted for the Rh30A gene product.
The coprecipitation of the Rh glycoprotein with the 30-kDa Rh polypeptides in Rh-specific immunoprecipitates suggests that the glycoprotein may be associated with the Rh30 polypeptides and be involved in Rh antigen expression. The Rh blood group locus (Bruns and Sherman, 1989;Cherif-Zahar et al., 1991) and the Rh3OA gene (MacGeoch et al., 1992) have been assigned t o chromosome l p , whereas the Rh5OA gene has been localized to chromosome 6 (Ridgwell et al., 1992). Therefore the Rh30 polypeptides rather than the Rh glycoproteins are responsible for Rh Cc, D, and Ee antigen polymorphisms. However, the glycoproteins may be required for the correct expression or presentation of Rh antigenic determinants at the erythrocyte surface.

MATERIALS AND METHODS
Preparation of Immunogens and Production ofdntisera-Synthetic peptides were synthesized using N=-Fmoc protected amino acids (Atherton et al., 1981), with a continuous flow solid-phase polyamide resin (Dryland and Sheppard, 1986) on a MilliGen 9050 Pepsynthesizer. The amino acids were coupled by in situ activation with benzotriazol tetramethyluroniumhydroxybenzotriazole (Knorr et al., 1989). The synthesized peptides were cleaved from the resin and deprotected with 93% (v/v) trifluoroacetic acid, 2.5% ( v h ) ethanedithiol, 2.5% (v/v) water, 2% (w/v) phenol for 2 h a t 25 "C. The 14-amino acid peptide RhGP-C corresponds to residues 3 9 6 4 0 9 at the C terminus of the predicted Rh50A glycoprotein sequence (Ridgwell et al., 1992). The 18-amino acid peptide Rh3OA-C corresponds to residues 4 0 1 4 1 7 at the C terminus of the predicted Rh30A polypeptide sequence (Avent et al., 1990;Cherif-Zahar et al., 1990) with a n additional cysteine residue added at the N terminus of the peptide. The 11-amino acid peptide Rh30A-N corresponds to residues 1-11 at the N terminus of the mature Rh,, polypeptide (Avent et al., 1988b;Bloy et al., 1988;Saboori et al., 1988) present in the red cell membrane (residues 2-12 of the protein predicted by the Rh3OA cDNA). Rh30A-N was acetylated at the N-terminal serine residue, as sequencing of the D3, polypeptide immunoprecipitated from erythrocyte membranes suggested that a proportion of the peptide may have a blocked N terminus (Avent et al., 1988b). The peptide R h 4 M corresponds to the 14 residues at the C terminus of the cDNA clones Rh4 and RhVI, which are reported to be splicing isoforms of Rh3OA (Le van Kim et al., 1992a). Peptides were coupled to keyhole limpet hemocyanin (Sigma) using the cross-linker sulfosuccinimidyl-4-(N-maleimi-domethy1)cyclohexane-1-carboxylate as described by Avent et al. (1992).
Male New Zealand White rabbits were immunized at 14-day intervals by subcutaneous injection of 100-200 pg of keyhole limpet hemocyanin-coupled peptide. The initial immunization was in Freund's complete adjuvant and subsequent boosts were in Freund's incomplete adjuvant. Immune bleeds were taken at 10-day intervals after each boost.
Rabbit antipeptide serum against residues 477-492 of the human glucose transporter (GLUT11 was a giR from Dr. S. A. Baldwin (University of Leeds).
Epitope Mapping of Antisera-Hexapeptides covering the synthetic peptide immunogen sequences of Rh30A-C, Rh3OA-N, and RhGP-C described above, were synthesized on a cellulose paper support using a SPOTS kit (Cambridge Research Biochemicals) and developed and probed with the rabbit polyclonal antisera according to the manufacturers' instructions.
Preparation and Deatment of Erythrocytes a n d Erythrocyte Membranes-Red cells were obtained from the South West Regional Blood Transfusion Centre, and were washed in phosphate-buffered saline, pH 7.4, before use. Erythrocyte membranes were prepared by lysis of cells in ice-cold 5 m sodium phosphate, pH 8, containing 2 m phenylmethanesulfonyl fluoride. In some cases red cells stored under liquid nitrogen were used and membranes were recovered from the frozen cells by direct lysis of samples in the above buffer without prior removal of cryopreservative.
To deglycosylate erythrocyte membranes, washed erythrocytes were washed once with PNGase F (peptide-N-glycanase F) buffer (112.5 m NaCI, 25 m sodium phosphate, 50 rm EDTA, 5 m glucose, 2 m phenylmethanesulfonyl fluoride, pH 7.51, resuspended in a n equal volume of PNGase F buffer containing either 200 unitdml PNGase F (Oxford Glycosystems, 4000 units/ml) or 400 pg/ml endo-p-galactosidase (prepared by the method of Kitamikado et al. (1981)) and incubated for 18 h a t 37 "C. The deglycosylated cells were washed several times with phosphate-buffered saline.
SDS-Polyacrylamide Gel Electrophoresis (SDS-PAGE) and Immunoblotting-SDS-PAGE was carried out using the glycine buffer system of Laemmli (1970) or the Tricine buffer system of Schagger and von Jagow (1987). Immunoblotting was carried out as described by Mallinson et al. (1986). Membranes were blocked with phosphate-buffered saline containing 0.2% (w/v) Tween 20, 5% (w/v) low-fat bovine milk powder.
Construction of Plasmids BSXG.Rh3OA and BSXG.Rh5OA-The polymerase chain reaction (PCR) was used to add a BamHI linker immediately 3' to the coding sequence of the Rh30A and Rh50A cDNAs, and a BamHI linker and consensus translation initiation sequence (CCACC, Kozak (1981Kozak ( , 1984) immediately 5' to the initiating methionine codon of the Rh3OA and Rh50A cDNA clones. The resulting PCR products were cut with BamHI and inserted into the compatible BglII site of the BSXG expression vector (Groves and Tanner, 1992) to give the constructs BSXG.Rh30A and BSXG.Rh50A. RNase-free preparations of BSXG.Rh30A and BSXG.Rh50A cDNA were linearized with the restriction enzyme HindIII, and Rh30A and Rh5OA cRNA synthesized using T7 RNA polymerase as described by Groves and Tanner (1992).
I n Vitro Cell-free Danslation ofRh3OA and Rh5OA cRNA a n d Protein Sequencing of Radiolabeled Rh50A-Rh30A and Rh5OA cRNAs were expressed in the rabbit reticulocyte lysate cell-free translation system with canine pancreatic microsomes (Promega) and labeled with ~-[~~S ] m e t h i o n i n e as described by Groves and Tanner (1992). Microsomes containing the expressed proteins were purified by centrifugation in an Airfuge (Beckman Instruments) through a 100-pl sucrose cushion (250 m sucrose, 0.5 M KCl, 5 m dithiothreitol, 50 m lysine, 3 nm MgCl,, 50 rm Hepes, pH 7.5; or 330 m sucrose, 3 rm MgC1,.6Hz0, 20 m Hepes, pH 7.5).
Microsomal membranes containing labeled expressed protein were solubilized in gel sample buffer (Laemmli, 1970) and the proteins separated by SDS-PAGE using the Tricine buffer system of Schagger & von Jagow (1987). The separated proteins were transferred to Problott membrane (Applied Biosystems) and exposed to X-OMAT-AR (Kodak) film for 90 min. The required radioactive band was excised from the Problott membrane and sequenced using a Blott cartridge and modified Fastblott cycles on an Applied Biosystem protein sequencer (model 477A). The radioactivity released at each sequencing cycle was determined in liquid scintillant (Emulsifier Safe, Packard) using a 3H/'4C program (Packard 1600TR).
Immunoprecipitation of Cell-free Danslation Products-The immunoprecipitation procedure was adapted from Anderson and Blobel (1983) and Groves and Tanner (1992). Microsomes purified by Airfuge centrifugation were solubilized with an equal volume of immunoprecipitation buffer (2% (w/v) SDS, 150 nm NaCl, 50 rm Tris, pH 7.4,5 m EDTA) and then diluted with a further 4 volumes of immunoprecipitation buffer containing 2.5% (w/v) Triton X-100. The solubilized microsomes were incubated overnight at 4 "C with the rabbit polyclonal antisera. Protein A-Sepharose beads (Bioprocessing), previously washed in detergent-free immunoprecipitation buffer, were then added and incubated for 1 h at room temperature or overnight at 4 "C. The protein A-Sepharose-bound immunoprecipitates were washed twice with 1 ml of low salt buffer (0.2% (w/v) SDS, 2% (w/v) Triton X-100, 10 nm NaCl, 50 rm Tris, pH 7.4,5 rm EDTA), and three times with immunoprecipitation buffer containing 2% Triton X-100. For immunoprecipitation from intact microsomes, Airfuge-purified microsomes were diluted with 10 volumes of detergent-free immunoprecipitation buffer, and then incubated overnight at 4 "C with the rabbit polyclonal antisera. The microsomes were then repurified through a sucrose cushion as described above and solubilized in immunoprecipitation buffer containing 2% (w/v) Triton X-100 before the addition of protein A-Sepharose as described above. Immunoprecipitated material was analyzed by SDS-PAGE using the Tricine gel system described by Schiigger & von Jagow (1987). Gels were treated for fluorography with Amplify (Amersham), dried, and exposed to preflashed X-AR film (Kodak).
Northern Blottiw-Total RNA was prepared from different mouse tissues (1 g) by the single step method using guanidinium thiocyanate (Ausubel et al., 1987). Poly(A+) RNA was prepared from the total RNA samples using oligo(dT)-cellulose columns (Pharmacia) according to the manufacturers' instructions. Northern blots were prepared using formaldehyde-agarose gels and probed with Rh30A or Rh50A cDNA (radio- labeled using a random prime DNA labeling kit; Boehringer Mannheim) as described by Ausubel et al. (1987).

RESULTS
Epitope Mapping ofAntipeptide Antisera Directed against Rh Protein Sequences-Antipeptide sera were prepared in rabbits to sequences corresponding to the N and C termini of the Rh30A protein and the C terminus of the Rh glycoprotein (RhGP). The epitopes reactive with the antiserum were examined using the SPOTS technique. RhGP anti-C was found to bind specifically to residues 3 9 8 4 0 4 (DDSVYWK) of the Rh50A C-terminal protein sequence (Ridgwell et al., 1992). Rh30A anti-C bound specifically to residues 3 9 9 4 1 5 (KF-PHLAV) of the Rh3OA C-terminal sequence, and Rh30A anti-N bound to residues 1-6 (Ac-SSKYPR) at the N terminus of the Rh30 polypeptides (Avent et al., 1990).
Reaction of Antipeptide Antibodies Directed against the C

Terminus of the Rh Glycoprotein with Human Erythrocyte
Membrane Proteins-The RhGP anti-C antibodies were tested by immunoblotting for reactivity with human erythrocyte membrane proteins separated on SDS-PAGE. The antibodies reacted with a very diffuse band with apparent molecular mass ranging from approximately 40 kDa to the high molecular mass region of the gel (Fig. 1, u-c), as expected for the Rh glycoprotein. A similar reactive band was present in membranes from cells of all common Rh phenotypes, both RhD+ve and RhD-ve ( Fig. 1, a and b), and from Rhnul) U+ve erythrocytes (Fig. Id), but not from Rhnull U-ve erythrocytes (Fig. le). The RhGP anti-C antibodies are not blood group U antigen-specific as the reactive band was also present in membranes from RhD+ve, U-ve erythrocytes (Fig. IC). The reactivity of the antibodies with membranes from all common Rh phenotypes but not from Rhnu]] U-ve red cells indicates that they are Rh-specific.
The reactivity of the RhGP anti-C antibodies with deglycosylated erythrocyte membrane proteins was studied.
After treatment of intact erythrocytes with PNGase F, which removes N-linked glycan chains from the peptide backbone, the immunoreactive band was sharpened and its apparent molecular mass was reduced, with the front of the band migrating at 30.5 kDa (Fig. If). The SDS-PAGE mobility of the deglycosylated Rh glycoprotein was similar to that of the Rh30 polypeptides, as would be expected from the similarity in predicted size and structural homology between the Rh30A polypeptides and Rh5OA glycoprotein.
Treatment of intact erythrocytes with endo-&galactosidase has been shown to reduce the apparent molecular weight of the Rh-related glycoprotein (Moore and Green, 1987). The RhGP anti-C reactive band in immunoblots of membrane proteins from erythrocytes that had been treated with endo-p-galactosidase was less diffuse and of lower apparent molecular weight than the band in untreated red cells (Fig. l , g and i). Treatment of erythrocytes with both PNGase F and endo-@-galactosidase did not reduce the mobility of the reactive band further than seen with PNGase F alone (Fig. lh). These results confirm the presence of N-glycan chains on the Rh glycoprotein, and also show that these chains are of the erythroglycan type and carry N-acetyllactosamine repeating units (Jamefelt et al., 1978). The Rh Glycoprotein in Rh, , !! Erythrocytes-Rh,,ll erythrocytes lack all antigens of the Rh blood group system and these red cells also show altered expression of a number of other blood group antigens, notably depression of the Ss antigens which are carried on glycophorin B (Dahr et al., 1978). Rhnull red cells may be positive or negative for the blood group U antigen. The RhGP anti-C antibodies reacted with the diffisely migrating Rh glycoprotein on immunoblots of membranes from all the red cell types tested except Rhnull U-ve (Fig. le). Although the RhGP anti-C antibodies reacted with the Rh glycoprotein in Rhnull U+ve cells (Fig. Id), the Rh glycoprotein band in Rhnull U+ve cells was less diffuse and had a slightly faster mobility than the Rh glycoprotein in control cells (Fig. 2, k and n). After deglycosylation of red cells using PNGase F, the deglycosylated Rh glycoprotein band from Rhnull U+ve red cells had the same mobility as the deglycosylated Rh glycoprotein band from normal red cells (Fig. 2, o andp), suggesting that the different mobility of the Rh glycoprotein of Rhnull U+ve red cells resulted from an alteration in the size of the N-glycan chain rather than the polypeptide moiety.
The Rh Glycoprotein in S-s-ErythrocytesS-s-erythrocytes lack glycophorin B and therefore do not express the Ss blood group antigens carried by this molecule (Dahr et al., 1978). It is known that glycophorin B is associated with the expression of blood group U antigens (Schmidt and Vos, 1967;Race and Sanger, 1968), but the U antigen is not carried by glycophorin B, since glycophorin B-deficient red cells (S-s-) are known that are U+ve. Membranes from both S-s-U+ve and S-s-U-ve erythrocytes reacted with the RhGP anti-C antibody on immunoblots and gave an Rh glycoprotein band with the same mobility characteristics (Fig. 2, b-e). However, the mobility of the trailing edge of the Rh glycoprotein band from both S-s-U+ve and S-s- U-ve cells was noticeably decreased compared to control red cells which contained normal glycophorin B (Fig. 2, a-g). The deglycosylated Rh glycoprotein band from PNGase F-treated S-s-U+ve and S-s-U-ve red cells had the same mobility as the deglycosylated Rh glycoprotein from control glycophorin B-containing cells (Fig. 2, h-j). These observations suggest that the N-glycan chain(s) of the Rh glycoprotein are altered in glycophorin B-deficient erythrocytes and this alteration does not depend on the U antigen type of the red cells. The mobility of band 3 (Fig. 3A, d-f) and the glucose transporter (Fig. 3A, a-c) (both of which contain the same erythroglycan type of oligosaccharide chain as the Rh glycoprotein, Fukuda et al. (1979)) was not altered in S-s-U+ve or S-s-U-ve erythrocyte membranes. Distribution of Rh50A mRNA in Different lksues-The Rh antigens are erythroid-specific in man but related proteins are present in erythrocytes of other mammalian species (Saboori et al., 1989). Rh30A mRNA transcripts have been detected by Northern blotting studies in human spleen erythroblasts and fetal liver, in the K562 and HEL erythroleukemic cell lines, and the MEGOl megakaryocytic cell line, but not in kidney or liver or nonerythroid cell lines (Cherif-Zahar et al., 1990). We used the human Rh5OA cDNA to probe Northern blots of poly(A+) RNA from several mouse tissues and from the K562 human erythroleukemic cell line. The Rh5OA probe hybridized to a major mRNA species of 2.9 kb together with a more weakly hybridizing mRNAof 2.1 kb in anemic mouse spleen, the erythropoietic tissue in anemic mice (Fig. 3B, h). It also detected a mRNA of 2.5 kb in the human K562 cell line (Fig. 3B, m ) . No hybridizing bands were detected in poly(A+) RNA from mouse kidney, small intestine, liver, testis, or lung (Fig. 3B, g and i -l ) . Probing the same Northern blot with human Rh30A showed weak hybridization to a 2.1-kb mRNA species in anemic mouse spleen mRNA only (data not shown). The 2.1-kb band that cross-hybridizes with both Rh30A and Rh50A human cDNA's may be the mouse 30-kDa Rh-related polypeptide. mRNA for members of the Rh-related protein family therefore appears to be expressed only in erythroid tissues.

Reaction of Antipeptide Antibodies Specific to the N-and C-terminal Sequences of Rh30A with Erythrocyte Membrane
Proteins-The Rh30A anti-N and Rh30A anti-C antibodies reacted with two closely spaced bands of apparent molecular mass 30 and 32 kDa on immunoblots of erythrocyte membrane proteins separated by SDS-PAGE (Fig. 4). These bands are of similar mobility to those immunoprecipitated from erythrocytes by Rh blood group-specific antibodies (the Rh3o polypeptides). The reactive bands were present in membranes from red cells of all common Rh phenotypes, both RhD+ve and RhD-ve ( Fig. 4, a, 6, d, e, fi h), but not Rhnull erythrocytes, irrespective of whether they were U+ve or U-ve (Fig. 4, c, g, andj-o). There were small differences in the mobility of the two bands in membranes from cells of different Rh phenotypes. This probably reflects the high degree of sequence heterogeneity in the Rh30 polypeptides predicted by the cDNA clones so far isolated for this group of polypeptides (Avent et al., 1990;Cherif-Zahar et al., 1990;Le van Kim et al., 1992a, 1992bKajii and Ikemoto, 1992). Treatment of erythrocytes with PNGase F or endo-agalactosidase prior to immunoblotting did not alter the mobility of the reactive bands with either the Rh3OAanti-N or Rh30A anti-C antibodies (data not shown). It has been shown that the Rh30 polypeptides do not carry carbohydrate chains (Gahmberg, 1983), and there are no N-glycan addition sites in the protein sequence predicted by the Rh30A cDNA.

Reactivity of Antipeptide Antibodies Directed against the Cterminal Sequence of Rh4 and RhVI with Erythrocyte Membrane Proteins-Le van Kim et al.
(1992a) used PCR to isolate cDNAs related to Rh3OA which appear to be derived from alternatively spliced mRNAs. These mRNAs had deletions compared to Rh3OA cDNA, which were located at intron-exon junctions in the Rh30A gene, and it was suggested that these transcripts represent Rh protein isoforms (Le van Kim et al., 1992a). However, it is not clear whether these PCR products are derived from low levels of mis-spliced mRNA transcripts, or actually represent proteins that are expressed in erythrocyte membranes. Two of the PCR products, Rh4 and RhVI, represent mRNAs that would encode shortened polypeptides (of 354 and 267 amino acids, respectively), in which a n altered reading frame produces a different C-terminal sequence to that of the Rh30A protein (Le van Kim et al., 1992a). Both Rh4 and RhVI are predicted to have the same new C-terminal sequence.
We raised rabbit antibodies against a synthetic peptide corresponding to the new amino acid sequence at the C terminus of Rh4 and RhVI and used them to investigate whether these isoforms are present in erythrocyte membranes. The antibodies reacted with the peptide immunogens on dot-blots. However, no reactive bands were obtained with these antibodies on immunoblots of erythrocyte membranes separated by SDS-PAGE, even when large amounts of membrane protein were used (data not shown), suggesting that the putative Rh isoforms encoded by Rh4 and RhVI are probably not present in the erythrocyte membrane.  Blots a d , e-h, i-k, and 1- , .

Cell-free Panslation of
the molecular mass predicted for the proteins from the cDNA sequences are 44.5 for the Rh glycoprotein and 45.5 for Rh3OA. This difference could arise because these extremely hydrophobic polytopic integral membrane proteins show very anomalous mobilities on SDS-PAGE, or because the mature polypeptide chains in the red cell membrane have undergone proteolytic cleavage. We examined this question by expressing the Rh5OA and Rh30A cRNAs using the rabbit reticulocyte lysate cell-free translation system in the presence of canine pancreatic microsomes. The radiolabeled translated proteins were compared with the red cell membrane forms detected by immunoblots of membranes separated by SDS-PAGE.
When Rh5OA cRNA was expressed in microsomes using the cell-free translation system, two radioactive bands of 30.5 and 36 kDa were observed (Fig. 5 u ) . Treatment of the detergentsolubilized microsomes with PNGase F or endo H to deglycosylate the protein reduced the 36-kDa band to 30or 30.5-kDa, respectively (Fig. 5, c and d 1. The M, of the glycosylated Rh50A cell-free translation product is lower than that of the Rh glycoprotein in erythrocyte membranes, because of the lack of processing of N-glycan chains in canine pancreatic microsomes. The slightly different mobilities of the two deglycosylated radioactive Rh50A bands (Fig. 5, c and d ) reflects the different specificities of the endoglycosidases used (Tarentino and Plummer, 1987). PNGase F removes the entire N-glycan from the polypeptide backbone, whereas endo H cleavage leaves a single N-acetyl glucosamine residue on the N-glycosylated asparagine of the polypeptide. Partial inhibition of N-glycosylation (by approximately 50%), by using the peptide acceptor Bz-NLTNHM~ in the cell-free translation mixture, decreased the amount of the 36-kDa band and increased the amount of the 30.5-kDa band (Fig. 56). The deglycosylated Rh50A cell-free translation products were separated by SDS-PAGE in parallel with membranes from red cells that had been treated with PNGase F. The RhGP anti-C immunoreactive protein from deglycosylated erythrocyte membranes had the same mobility as the deglycosylated cell-free translation product (Fig.  5). N-terminal amino acid sequencing of the 36-kDa Rh5OA translation product labeled using [3sSlmethionine resulted in the release of radioactivity at sequence cycles 1 and 8 (data not shown), as expected from the presence of methionine a t these positions in the predicted Rh5OA sequence. The radioactive Rh50A translation product could be immunoprecipitated from detergentsolubilized microsomes using the RhGP anti-C antibody (data not shown). These results show that the cell-free translation product retains the N and d terminus of the predicted Rh50A protein sequence and does not undergo proteolytic cleavage.
The Rh30A cRNA was also expressed in canine pancreatic microsomes using the rabbit reticulocyte lysate cell-free translation system. SDS-PAGE analysis of the products showed a 30-32-kDa radioactive band (Fig. 5g), which was only obtained when canine pancreatic microsomes were present, consistent with the extremely hydrophobic nature of the Rh polypeptides. On SDS-PAGE analysis, the radiolabeled Rh30A cell-free translation product aligned exactly with the 30-32 kDa band in erythrocyte membranes immunoblotted with Rh30A anti-C antibody (Fig. 5, f and g). The 30-32-kDa labeled band could be immunoprecipitated from the microsomes by the Rh3OA anti-N and Rh30A anti-C, showing that the product retained both the N and C terminus predicted from the cDNA sequence (data not shown). Sequencing of the [3sSlmethionine-labeled Rh30A cellfree translation product purified from the microsomes resulted in the release of radioactivity a t sequence cycle 1, as expected from the presence of methionine at the N terminus of the protein. The 45.5-kDa protein predicted from the Rh30A cDNA sequence clearly migrates anomalously on SDS-PAGE in a similar manner to that described for the Rh glycoproteins above, and our results show that the anomalous mobility is not the result of post-translational cleavage or modifications to the protein.

DISCUSSION
The antipeptide antisera used in the present study have been shown to be Rh blood group specific since they react with all erythrocytes except Rhnull. The Rh30A anti-C and Rh30A anti-N antisera react with two closely spaced bands of molecular mass 30-32 kDa on immunoblots of erythrocyte membrane proteins from red cells of all Rh phenotype except Rh,,,ll. This is probably because there is a high degree of homology between the terminal sequences of the Rh30 family of polypeptides which results in cross-reactivity of the Rh30A-specific antisera. Similar results have been obtained by Avent et al. (1992) and Suyama and Goldstein (1992). The 30-kDa Rh polypeptides have been reported not to carry glycan chains, and consistent with this finding, deglycosylation of erythrocytes with PNGase F or endo-P-galactosidase does not affect the mobility of the Rh30A anti-N and Rh30A anti-C immunoreactive bands on The Rh3o polypeptides have an apparent molecular mass on SDS-PAGE of 30-32 kDa (Moore et al., 1982;Gahmberg, 1983;Ridgwell et al., 1983) but the Rh30A cDNA clone predicts a molecular mass of 45.5 kDa (Avent et al., 1990;Cherif-Zahar et al., 1990). It has been suggested that the Rh polypeptides may behave anomalously on SDS-PAGE (Gahmberg, 19831, because of their hydrophobicity, or because they undergo post-translational cleavage. Suyama et al. (1991) suggest that the disparity between the predicted and observed M , of the Rh30 polypeptides was due to the unusual SDS binding characteristics of the Rh proteins. The SDS-PAGE mobility of the Rh30A polypeptide expressed in microsomes using the reticulocyte lysate cell-free translation system is the same as that of the Rh30 polypeptides in erythrocyte membranes (Fig. 5). This shows that both the 30-kDa Rh proteins in erythrocyte membranes and the Rh3OA cell-free translation product behave in a similar anomalous manner on SDS-PAGE, and confirms that the Rh30 polypeptides are not subject to post-translation cleavage.
The RhGP anti-C antisera react with the diffusely migrating Rh glycoprotein from all red cells except Rh,,ll U-ve. Since Rh,,ll U+ve red cells do not carry Rh blood group antigens but contain the Rh glycoprotein, it is clear that the Rh glycoprotein alone does not express Rh blood group antigens. The Rh5OA cDNA clone predicts a polypeptide of molecular mass of 44.5 kDa but the deglycosylated protein produced in cell-free translation has a molecular mass of 30.5 kDa, and this has the same SDS-PAGE mobility as the deglycosylated Rh glycoprotein detected in immunoblots of membranes from PNGase F-treated erythrocytes (Fig. 5). The Rh glycoprotein clearly shows the same anomalous mobility on SDS-PAGE as the Rh30 polypeptides. The difference in the apparent molecular mass of the Rhso polypeptides and the deglycosylated Rh glycoprotein immunoprecipitated by BRIC 69 from PNGase F-treated ccDEE erythrocytes is approximately 1 kDa. This is similar to the calculated molecular mass difference between the predicted sequences of the Rh3OA and Rh50A cDNA clones (approximately 0.9 kDa).
The expression of Rh5OA cRNA by cell-free translation with microsomes gives rise to two bands of molecular mass 30.5 and 36 kDa (Fig. 5). Deglycosylation of the translation products with PNGase F or endo H, or inhibition of glycosylation with Bz-NLTNHMe, reduces the amount of the 36-kDa component without changing the mobility of the 30.5-kDa band, which suggests that the 30.5-kDa band is unglycosylated. The presence of only one glycosylated band in the Rh5OA cell-free translation product is consistent with other data which show that, of the three potential N-glycosylation sites present in the Rh5OA cDNA sequence, only Asn-37 is glycosylated in the Rh glycoprotein isolated from erythrocyte membranes (Eyers et al., 1994).
The monoclonal antibody MB2D10, produced by immunization of mice with isolated Rh immunoprecipitates which included the Rh glycoproteins (von dem Borne et al., 1990), reacts with a diffusely migrating band of M , similar to the Rh glycoprotein on immunoblots of all erythrocyte membranes, except Rh,,ll U-ve. This reactivity was abolished by prior removal of N-linked carbohydrates from the erythrocyte membranes (von dem Borne et al., 1990;Mallinson et al., 1990). MB2DlO also reacted with the isolated glycoprotein component of Rh immunoprecipitates (Mallinson et al., 1990). The RhGP anti-c terminal specific rabbit antibodies described in this paper show similar specificity to MB2D10, in that they react with the Rh SDS-PAGE. glycoprotein on immunoblots of membrane proteins from all red cells, including Rhnull U+ve, but not with Rhnull U-ve red cells. However, the MB2D10 antibody epitope is dependent on the presence of the N-glycan chain (Mallinson et al., 19901, whereas the rabbit RhGP anti-C antibodies are peptide specific. The Rh glycoprotein in S-serythrocytes (which lack glycophorin B) shows a different mobility on SDS-PAGE to that in control cells, irrespective of whether the S-sred cells are U+ve or U-ve. Although the red cell Ss antigens are carried by glycophorin B, and glycophorin B is associated with U antigen expression (Dahr and Moulds, 19871, the U antigen is not carried by glycophorin B since erythrocytes lacking glycophorin B may be U+ve. All erythrocyte types, except the Rhnull U-ve type, contain the Rh glycoprotein. Our results are consistent with earlier suggestions that the expression of the U antigen requires both the Rh glycoprotein and glycophorin B to be expressed in the erythrocyte membrane (Mallinson et al., 1990;von dem Borne et al., 1990). The absence of the Rh glycoprotein in Rhnull U-ve cells is associated with the lack of U antigen expression, and may also account for the reduced expression of Ss antigens (glycophorin B) in these cells (Dahr et al., 1987a). This is consistent with suggestions that glycophorin B may form part of a multimolecular complex containing the Rh antigens in the erythrocyte membrane (reviewed by Agre and Cartron (1991) and Anstee and Tanner (1993)).
The absence of glycophorin B in S-serythrocytes is associated with a decrease in the SDS-PAGE mobility of the Rh glycoprotein (Fig. 2, a-g), which probably reflects a n increase in the average number of N-acetyllactosamine repeating units present in the N-glycan chain on the Rh glycoprotein when an interaction between the Rh glycoprotein and glycophorin B cannot take place. This suggests that interaction occurs between glycophorin B and the Rh glycoprotein during their biosynthesis and movement through intracellular membrane compartments to the cell surface, so that in the absence of glycophorin B the processing of the N-glycan chain on the Rh glycoprotein is altered in these intracellular compartments.
A parallel situation occurs in the case of glycophorin A and red cell band 3, which also has an erythroglycan type N-glycan chain containing repeating N-acetyllactosamine units. The band 3 in human red cells which lack glycophorin A (the En(a-) or Mk/Mk phenotypes) has a reduced mobility on SDS-PAGE, which results from an increase in the average number of Nacetyllactosamine units on the protein (Gahmberg et ul., 1976;Tanner et al., 1976). In red cells which effectively express more glycophorin A than normal because of the presence of additional glycophorin A hybrid proteins (the Dantu and St(a+) types), the average size of the N-glycan chain on band 3 is reduced (Dahr et al., 1987b). This effect occurs because the two proteins are associated during their biosynthesis. The presence of glycophorin A accelerates the movement of newly synthesized band 3 through intracellular membranes to the cell surface (Groves and Tanner, 1992).' The effects on the size of the band 3 glycan chain appear to result because band 3 moves more slowly through internal membranes in the absence of glycophorin A, allowing more time for extension of the carbohydrate chains by the glycosyltransferases in these internal compartments.
By analogy with the case of glycophorin A and band 3, it is likely that the increased size of the N-glycan on the Rh glycoprotein in cells lacking glycophorin B occurs because the Rh glycoprotein remains in the intracellular membrane system longer when glycophorin B is not present during its biosynthe-L. J. Bruce, D. J. Groves sis. From this we conclude that, in the normal red cell, glycophorin B accelerates the movement of the newly synthesized Rh complex through intracellular membranes to the cell surface. Although the amount of band 3 found on the surface of red cells which lack glycophorin A is normal, the protein appears to be only partially competent for anion transport.2 Similarly, the expression of Rh antigens in glycophorin B-deficient red cells is not abnormal, but it is possible that the functional activity (which is presently unknown) of the complex is altered in glycophorin B-deficient cells.
In contrast to the increase in size found when glycophorin B is absent, the Rh glycoprotein N-glycan chains are reduced in size when the Rh30 polypeptides are absent (as in Rhnull U+ve red cells), suggesting that the Rh glycoprotein moves faster through intracellular membranes to the cell surface in this situation. Thus the R30 polypeptides interact with the Rh glycoprotein during the biosynthesis of the Rh complex and appear to slow the movement of the Rh glycoprotein through intracellular membranes to the cell surface. No red cells have been found that express the Rha0 polypeptides in the absence of the Rh glycoprotein. This suggests that the Rh glycoprotein may be required for the expression of the Rh30 polypeptide at the red cell surface. Since Rhnull U+ve erythrocyte membranes contain the Rh glycoprotein but not the Rh30 polypeptides, it is clear that the expression of the Rh glycoprotein in red cells does not require the co-expression of Rh,, polypeptides.
In the case of the proteins of the Rh complex, it is clear that both glycophorin B and the Rh30 polypeptides interact with the Rh glycoprotein during the biosynthesis of the complex, so that the intracellular association of all three proteins is likely to be necessary for the normal biosynthesis and expression of the Rh complex at the red cell surface. This is likely to be important in designing expression systems for the Rh antigens. Co-expression of both the Rh glycoprotein and glycophorin B (and perhaps other Rh-associated proteins such as the CD47 glycoprotein) with the Rh30 polypeptides will probably be necessary for the successful expression of the Rh complex at the cell surface. Alterations in the size of proteins carrying N-glycan chains of the erythroglycan type may prove to be a generally useful diagnostic tool for assessing changes in the rates at which these proteins move through intracellular membrane compartments to the cell surface under different circumstances. When used in conjunction with the study of naturally occurring deletion variants of different components associated with the Rh blood group system, this approach has given a useful insight into the intracellular associations which occur between the different components during the biosynthesis and translocation of the Rh complex to the surface of the normal red cell.