(Received for publication, January 15, 1997, and in revised form, May 14, 1997)
From the Medical Research Council of Canada Group in Membrane Biology, Departments of Medicine and Biochemistry, University of Toronto, Toronto, Ontario, Canada M5S 1A8
Band 3, the anion exchanger of human erythrocytes, contains up to 14 transmembrane (TM) segments and has a single endogenous site of N-glycosylation at Asn642 in extracellular (EC) loop 4. The requirements for N-glycosylation of EC loops and the topology of this polytopic membrane protein were determined by scanning N-glycosylation mutagenesis and cell-free translation in a reticulocyte lysate supplemented with microsomal membranes. The endogenous and novel acceptor sites located near the middle of the 35 residue EC loop 4 were efficiently N-glycosylated; however, no N-glycosylation occurred at sites located within sharply defined regions close to the adjacent TM segments. Acceptor sites located in the center of EC loop 3, which contains 25 residues, were poorly N-glycosylated. Expansion of this loop with a 4-residue insert containing an acceptor site increased N-glycosylation. Acceptor sites located in short (<10 residues) loops (putative EC loops 1, 2, 6, and 7) were not N-glycosylated; however, insertion of EC loop 4 into EC loops 1, 2, or 7, but not 6, resulted in efficient N-glycosylation. Acceptor sites in putative intracellular (IC) loop 5 exhibited a similar pattern of N-glycosylation as EC loop 4, indicating a lumenal disposition during biosynthesis. To be efficiently N-glycosylated, EC loops in polytopic membrane proteins must be larger than 25 residues in size, with acceptor sites located greater than 12 residues away from the preceding TM segment and greater than 14 residues away from the following TM segment. Application of this requirement allowed a significant refinement of the topology of Band 3 including a more accurate mapping of the ends of TM segments. The strict distance dependence for N-glycosylation of loops suggests that TM segments in polytopic membrane proteins are held quite precisely within the translocation machinery during the N-glycosylation process.
Band 3 (AE1) is the anion exchange
(AE)1 protein of the erythrocyte membrane
where it catalyzes the electro-neutral exchange of Cl and
HCO3
(1-6). Human Band 3 is a 911-amino acid
polypeptide that consists of two domains, an amino-terminal domain
located in the cytosol that is involved in cytoskeletal interactions
and a carboxyl-terminal membrane domain responsible for anion exchange.
Current models (see Fig. 1) indicate that the membrane domain may
contain up to 14 TM segments, but the disposition of many of the
hydrophilic loops connecting putative TM segments with respect to the
membrane has not been established (4, 5, 7, 8).
Human Band 3 is N-glycosylated at Asn642 within EC loop 4 (9, 10). The mouse, rat, and chicken proteins are also N-glycosylated on the same loop (4). Interestingly, trout erythrocyte Band 3 is N-glycosylated on an expanded EC loop 3 (Z-loop) and does not contain an acceptor site in EC loop 4 (11). AE2 and AE3, like trout Band 3, contain N-glycosylation sites in an expanded EC loop 3 but none in EC loop 4 (4). The localization of N-glycosylated sites to single loops is a common feature of native polytopic membrane proteins (12), although N-glycosylation of multiple loops is possible (13).
The placement of novel N-glycosylation acceptor sites (Asn-Xaa-Ser/Thr) into putative extracellular or intracellular domains of polytopic membrane proteins has proven very useful in defining the folding pattern of the polypeptide in the membrane (14-20). Successful N-glycosylation requires that the acceptor site is exposed to the lumen of the endoplasmic reticulum during biosynthesis and is adequately spaced from the membrane surface (15, 20-23). In this paper, we have used this biosynthetic approach to refine the topology of Band 3 and to determine the requirements for N-glycosylation of loops in polytopic membrane proteins.
The following is a list of products and (in
parentheses) their suppliers: MEGAscriptTM in
vitro transcription kit, DNase I, and the cap analog m7GpppGp (Ambion); rabbit reticulocyte translation system, canine pancreatic microsomes, and RNasin (Promega); ModulisTM in
vitro transcription/translation systems (MBI Fermentas); Single Tube Protein System 2, and T7 RNA polymerase (Novagen);
[35S]methionine (Amersham Corp.);
TransformerTM mutagenesis kit and Multi-PolTM
DNA sequencing kit (CLONTECH); octaethylene glycol
mono n-dodecyl ether (C12E8)
(Nikkol); deoxyribonucleotides (Boehringer Mannheim); restriction
enzymes, T4 DNA ligase, and Klenow DNA polymerase (New England Biolabs
and Pharmacia Biotech Inc.); and [-35S]ATP (ICN).
The cDNA of human Band 3 in Bluescript vector BSSK+ (Stratagene) was a kind gift from Drs. A. M. Garcia and H. Lodish, Whitehead Institute. The endogenous N-glycosylation site in the membrane domain of Band 3 at Asn642 was mutated to Asp in the TM1/14 construct as described (24). A series of unique N-glycosylation acceptor sites (Table I) were created in the TM1/14 N642D construct using oligonucleotide-directed mutagenesis and the double-stranded mutagenesis system from CLONTECH.
|
A "megaprimer" PCR method (25,
26) was used to expand the size and to insert a novel
N-glycosylation acceptor site into EC loop 3 in the TM1/14
N642D construct. The resulting mutant had the additional sequence
Val-Asn-Ser-Ser inserted between Tyr555 and
Asn556. The flanking and mutagenic primers were:
5-GCAGGGCATTCTCTTCGC-3
(forward primer), 5
-GTGGTGATCTGAGACTCCAGG-3
(reverse primer), 5
-ACTGCTGTTAACGTAGTTATAAGTCTTCTGTAGTG-3
(reverse mutagenic primer), and
5
-CTACGTTAACAGCAGTAACGTGTTGATGGTGCCCAAAC-3
(forward mutagenic primer). The bold lettering indicates the
insertion that codes for the sequence Val-Asn-Ser-Ser. The underline
indicates the position of a novel HpaI site that was used to
screen for mutants.
Short
EC loops were expanded by insertion of a 35-amino acid sequence derived
from EC loop 4, containing the endogenous N-glycosylation acceptor site, into the TM1/14 N642D construct. Two oligonucleotide primers were used to synthesize a DNA segment corresponding to EC loop
4 with terminal XbaI sites by PCR (35 cycles at 94 °C, 1 min; 53 °C, 2 min; 72 °C, 3 min). Primer 1 was a sense
oligonucleotide, 5-GCTTCTAGATACCTACACCCAGAAACTCTC-3
and
primer 2 was antisense oligonucleotide
5
-GAGTCTAGAAACTCGGAACGCAAGCCCAGT-3
. The underline indicates the XbaI cleavage sites. A unique XbaI
cleavage site is located in Band 3 cDNA in a position corresponding
to EC loop 2. The PCR product was ligated directly into this
XbaI site in TM1/14 N642D construct. The endogenous
XbaI site was mutated by a silent mutation
CTA(Leu484)
CTG in TM1/14 N642D construct. Single
novel XbaI restriction sites were then introduced at
positions corresponding to EC loops 1, 6, and 7 and IC loops 3, 5, and
6. The same PCR product as described above was inserted into the novel
XbaI sites. The resulting amino acid sequences at the
insertion sites of the mutants are EC loop 1, ...
Leu427-Leu428-Asp626-Thr627 ...
Glu658-Phe659-Leu428-Glu429-Lys430 ...
; EC loop 2, ...
Gly483-Leu484-Asp626-Thr627 ...
Glu658-Phe659-Leu484-Glu485-Tyr486 ...
; EC loop 6, ...
Pro784-Leu785-Asp626-Thr627 ...
Glu658-Phe659-Leu785-Asp786-Val787 ...;
EC loop 7, ...
Thr853Leu854-Asp626-Thr627 ...
Glu658-Phe659-Leu854-Asp855-Ser856 ...
; IC loop 3, ...
Pro598-Leu599-Asp626-Thr627 ...
Glu658-Phe659-Leu599-Glu600-Leu601 ...
; IC loop 5, ...
Ile753-Leu754-Asp626-Thr627 ...
Glu658-Phe659-Leu754-Glu755-Val756 ...
; and IC loop 6, ...
His819-Leu820-Asp626-Thr627 ...
Glu658-Phe659-Leu820-Asp821-Val822... .
The amino acids in bold represent the inserted sequence corresponding to EC loop 4, and the underlined residues indicate amino acids that were altered or duplicated due to the creation of an XbaI restriction site. All mutations and insertions were confirmed by dideoxynucleotide chain termination DNA sequencing (27).
In Vitro Transcription and TranslationPlasmid DNA was linearized downstream of the stop codon with HindIII or with SacI; in the latter case, the ends were blunted with Klenow DNA polymerase. DNA (1 µg) was transcribed with T7 RNA polymerase in the presence of m7GpppGp for 4 h at 37 °C using the Ambion in vitro transcription kit according to the manufacturer instructions. The DNA template was removed by digestion with 2 units of DNase I for 15 min at 37 °C. The RNA was extracted with 1:1 phenol:chloroform, precipitated with 1/10 volume of 5 M ammonium acetate and 1 volume isopropyl alcohol, and then resuspended in diethylpyrocarbonate-treated water containing 2 units/µl RNasin.
Translation in a rabbit reticulocyte lysate was performed according to the suppliers. One-tenth (0.5-2 µg) of the RNA transcribed in vitro was added to 50 µl of translation mixture, and translation was allowed to continue for 2 h at 30 °C. The detergent C12E8 at 0.1% (w/v) was included in the translation mixture lacking microsomes to reduce aggregation of the translated protein. Translation was also carried out in the presence of microsomal membranes (7 units/50 µl, one unit of microsomes is the amount of membranes required to cleave the signal sequence of preprolactin by 50%). [35S]Methionine at 1 µCi/µl was included in all translation mixtures.
Alternatively, a coupled transcription/translation system supplied by MBI Fermentas was used. The transcription reaction was carried out with 0.3 µg of DNA using T7 RNA polymerase in a total volume of 3 µl. The sample was divided equally into three tubes, and the translation mix was added. Translation was carried out in the presence of 1) 0.1% C12E8, 2) 1 µl (2 units) of canine pancreatic microsomes, or 3) 1 µl microsomes and 32 µM Ac-Asn-Tyr-Thr-NH2 in a total volume of 10 µl. Each tube contained 10 µCi of [35S]Met. After incubation at 30 °C for 1 h, tRNA was digested with RNase A. Membrane integration of the translation products was determined by extracting samples (10 µl) with 100 µl of ice-cold 0.1 M Na2CO3, pH 11.5, followed by recovery of the stripped microsomes by centrifugation (16,000 × g for 15 min) (24). The translation products were examined by SDS gel electrophoresis (28) followed by autoradiography. The efficiency of N-glycosylation was determined by scanning autoradiographs of SDS gels of the cell-free translation products synthesized in the presence of microsomes using a Hoefer densitometer and MacIntegrator. The percentage of N-glycosylation was calculated from the integrated areas under the two peaks corresponding to the N-glycosylated and non-glycosylated full-length products. N-Glycosylation of the mutants was always compared with the wild-type TM1/14 construct synthesized in a parallel translation reaction.
The requirements for N-glycosylation of EC loops and the topology of Band 3 were determined by scanning N-glycosylation mutagenesis. The endogenous site of N-glycosylation at Asn642 in EC loop 4 was mutated (N642D) to provide a non-glycosylated protein into which novel N-glycosylation acceptor sites (Asn-Xaa-Ser/Thr) could be introduced (Fig. 1). The loops connecting TM segments are poorly conserved (6), and therefore, single site mutations within them are not expected to have a profound effect on protein topology. A cell-free system consisting of a reticulocyte lysate supplemented with microsomal membranes was used to assay the N-glycosylation status of the mutants. N-glycosylation at a single site will produce a uniform increase of about 2,500 in the molecular weight of the protein corresponding to the cotranslational attachment of the high mannose oligosaccharide, which only occurs in the lumen of the microsomes. To confirm that the shift in mobility was due to N-glycosylation, one translation reaction in each set of translation mixes included a competitive inhibitor of N-glycosylation, the acceptor peptide Ac-Asn-Tyr-Thr-NH2. A truncated construct (TM1/14) (24) corresponding to the membrane domain of Band 3 (Fig. 1) was used in these experiments to improve the yield of full-length translation product and to better visualize the difference in molecular weight due to N-glycosylation.
EC Loops 3 and 4 MutationsThe endogenous site at
Asn642 in EC loop 4 can be N-glycosylated by
microsomal membranes in the reticulocyte translation system (Fig.
2, lanes 1 and 2). The
non-glycosylated protein has a molecular mass of 48 kDa while the
glycosylated protein is 3 kDa larger, consistent with the attachment of
a single oligosaccharide chain. The average efficiency of
N-glycosylation in the translation reaction was 65% (Table
I). Increasing the amount of microsomes in the translation reaction did not result in a significant increase in the
level of N-glycosylation. It has been reported (29, 30) that
acceptor sites containing Thr can be more efficiently
N-glycosylated than those containing Ser. To test this
with Band 3, the endogenous site was mutated from
Asn642-Ser-Ser to Asn642-Ser-Thr; however, no
increase in the level of N-glycosylation was observed in the
cell-free system (Table I). Both the N-glycosylated and
non-glycosylated products were resistant to alkaline extraction showing
that they were integrated into the lipid bilayer. The series of lower
molecular mass bands around 35-40 kDa are due to internal initiation
of translation at methionine residues located within TM 6 (24).
EC loop 3 contains 25 residues and is sensitive to cleavage by
extracellular chymotrypsin in intact red cells (31). A construct containing the mutated endogenous acceptor site and a novel site (Asn556) in the middle of EC loop 3 can be glycosylated as
expected but only poorly (Fig. 2 and Table I). Other mutants containing
acceptor sites one or two residues proximal (Asn554, or
Asn555) were also N-glycosylated but again at
low efficiency (Fig. 2 and Table I). This low efficiency may be due to
the suboptimal size of EC loop 3 (12, 13), which may
place the acceptor sites too close to the end of a TM segment to be
properly N-glycosylated (23). To test this hypothesis, EC
loop 3 was increased in size by insertional mutagenesis. This mutant
contained the additional four residue sequence Val-Asn-Ser-Ser between
Tyr555 and Asn556. The acceptor site within
this insert was readily N-glycosylated (Fig. 2). Thus, this
loop must be larger than 25 residues in size to be efficiently
N-glycosylated, which places the acceptor sites 12 residues or more from the adjacent TM segments.
Scanning N-glycosylation mutagenesis was performed on the endogenously N-glycosylated EC loop 4 to determine the spacing required for N-glycosylation of acceptor sites in loops. Acceptor sites located near the middle of this loop were N-glycosylated with an efficiency equal to the endogenous site (Table I and Fig. 3). One exception occurred with the sequon N647WS, which was poorly N-glycosylated relative to N647SS (Table I). This confirms the observation that sequons containing tryptophans are poor acceptor sites (32). Otherwise, the amino acid sequence context of acceptor sites in EC loop 4 had little effect on N-glycosylation in this cell-free assay. However, no N-glycosylation occurred at sites positioned close to either of the neighboring TM segments (Fig. 3). There was a sharp cut-off for N-glycosylation, which suggests that acceptor sites in loops must be positioned a minimum distance away from the lumenal ends of adjacent TM segments to reach the active site of the oligosaccharyl transferase. This requirement explains the poor N-glycosylation observed for the point mutations in EC loop 3.
EC Loop 4 InsertionsThe results of cell-free translation of
mutants containing acceptor sites in other putative EC loops are shown
in Fig. 4. No change in the mobility of the translation
products occurred in the presence of microsomes for any of the mutants
containing acceptor sites in EC loops 1, 2, 6, or 7. According to the
folding model presented in Fig. 1, these loops are very short (5-10
residues), which may account for their inability to be
N-glycosylated.
To determine the orientation of short loops with respect to the membrane, insertion of EC loop 4 into engineered XbaI sites was performed. A similar strategy was employed in investigations of the topology of 3-hydroxy-3-methylglutaryl (HMG)-CoA reductase (15), the glucose transporter GLUT1 (18), and the sodium-dependent glucose transporter (20). The translation products containing the EC loop 4 insert are about 3 kDa larger than the corresponding TM1-14 proteins confirming that the expressed proteins contain the additional amino acid sequence. EC loop 4 inserted into EC loop 1 was N-glycosylated (Fig. 4), in contrast to the point mutation in EC loop 1. This confirms that acceptor sites in short loops facing the lumen of the endoplasmic reticulum cannot reach the active site of the oligosaccharyl transferase. The efficiency of N-glycosylation of this loop was close to the levels found for the wild type protein (Table I). This result shows that the inserted loop is accessible to the lumen of the endoplasmic reticulum and that this region of Band 3 is extracellular. Insertion of EC loop 4 into the endogenous XbaI site in EC loop 2 or EC loop 7 also resulted in optimal N-glycosylation (Fig. 4 and Table I). In contrast, the insert in EC loop 6 was N-glycosylated at a lower level (Fig. 4 and Table I). The low level of N-glycosylation observed for the insert in EC loop 6 suggests that only partial translocation of the loop into the lumen of the microsomes occurred in the cell-free system. A series of control EC loop 4 insertions into IC loops 3, 5, and 6 were also made. Insertions in IC loops 3 or 6 resulted in barely detectable N-glycosylation (Table I). The low level of N-glycosylation of these cytosolic loops suggests that some aberrant translocation had likely occurred during cell-free translation. Insertion into IC loop 5, however, resulted in a high level of N-glycosylation (Table I), inconsistent with a cytosolic disposition for this loop (see below). The insertion mutants suggest that EC loops 1, 2, and 7 and IC loop 6 in Band 3 are lumenal during biosynthesis while the disposition of putative EC loop 6 is ambiguous.
IC Loop MutationsFigs. 5 and 6
show the results of a set of experiments with site-directed mutants
containing acceptor sites in putative IC loops 5 and 6 and the
carboxyl-terminal tail of Band 3. IC loop 5 contains a site at
Lys743 that can be partially cleaved by trypsin contained
within resealed ghosts (33). A point mutation (K743N) was made at this
particular lysine to create a novel N-glycosylation acceptor
site. Fig. 5 shows that this site, contrary to the 14 TM model, was
N-glycosylated, suggesting that this hydrophilic loop is
lumenal. This agrees with the results of the EC loop 4 insertion into
IC loop 5. To confirm this finding, scanning N-glycosylation
mutagenesis was performed on this loop. Acceptor sites located
centrally between residues 735 and 748 were N-glycosylated;
however, sites located close to the ends of the neighboring TM segments
were not (Fig. 5 and Table I). This pattern of
N-glycosylation was very similar to that found for the EC
loop 4. These results show that this part of Band 3 is exposed to the
lumen of the microsomes during biosynthesis and that acceptor sites in
this loop must also be adequately spaced from the adjacent TM segments
to be efficiently N-glycosylated. It is unlikely that all
the individual point mutations in IC loop 5 resulted in a change in the
orientation of this loop. Acceptor sites in IC loop 6 (Asn821,828) or in the carboxyl-terminal tail
(Asn880) were not N-glycosylated, consistent
with their intracellular disposition (Fig. 6). The lack of
N-glycosylation of IC 6 and the carboxyl-terminal tail also
confirms the specificity of the N-glycosylation assay.
In this paper, we have defined the spacing requirements for the N-glycosylation of loops in the polytopic membrane protein Band 3. EC loop 3 in Band 3 contains about 25 residues, and three different point mutations near the center of this loop were barely N-glycosylated. Increasing the size of this loop by only four residues resulted in efficient N-glycosylation. The N-glycosylated EC loop 3 in trout Band 3 and in AE2 and AE3 contain an insert (Z-loop) doubling the size of the loop to about 50 residues. One reason for the larger size of EC loop 3 in these isoforms may be to allow efficient N-glycosylation. EC loop 4 in Band 3 contains about 35 residues, and sites located within a sharply defined threshold region near the middle of this loop were efficiently N-glycosylated. Point mutations in small EC loops in Band 3 were not N-glycosylated. However, insertion of EC loop 4 into these loops permitted efficient N-glycosylation. These findings are in agreement with observations made with other multispan membrane proteins in which short exofacial loops were also found to be poorly N-glycosylated (15, 18, 20-22).
A survey of polytopic membrane proteins revealed that EC loops containing N-linked oligosaccharides have a minimum size of about 30 residues, with the N-glycosylation site at least 10 residues away from a transmembrane segment (12). The distance between the oligosaccharyl transferase active site and the endoplasmic reticulum membrane has been determined using the Escherichia coli leader peptidase as a model membrane protein (23). This protein contains two TM segments with both the amino and carboxyl termini located in the periplasmic space. It was found that acceptor sites must be spaced greater than 12 residues COOH terminal from the second TM segment or greater than 14 residues NH2 terminal from the first TM segment to reach the active site of the lumenal oligosaccharyl transferase. The experiments described in this paper define similar limits for N-glycosylation of loops in the polytopic membrane protein Band 3. The distance requirements found for the leader peptidase can be applied to loops in polytopic membrane proteins. Thus, the acceptor site asparagine in loops must be spaced greater than 12 residues from the proximal and 14 residues from the distal TM segment to be efficiently N-glycosylated. For simplicity, this will be referred to as the "12 + 14 rule".
Mapping the Ends of TM SegmentsThe lumenal ends of TM
segments can be mapped within a residue or two using the 12 + 14 rule.
Acceptor sites at Asn554, Tyr555, or
Asn556 in EC loop 3 are all poorly
N-glycosylated, suggesting that they are closer than 12 or
14 residues from the ends of TM segments 5 and 6, respectively. The
acceptor site in the four amino acid insertion (VNSS) between residues
555 and 556 in EC loop 3 can be N-glycosylated. This
information maps the end of TM segment 5 to Phe544 and the
beginning of TM segment 6 to Pro568, in agreement with the
hydropathy analysis (Fig. 7).
TM segment 7 contains a well defined 21-residue hydrophobic sequence that extends from Val604 to Ile624 (Fig. 3). This segment is bordered by hydrophilic sequences including a pair of arginines on the cytosolic side of the membrane and a poorly-conserved Gln625 on the lumenal side (6). The closest acceptor site to TM segment 7 in EC loop 4 that can be N-glycosylated is located at Gly637. The 12 + 14 rule would place the end of TM7 at Ile624, 13 residues away from Gly637, in perfect agreement with the hydropathy profile prediction. Maximal N-glycosylation occurred at Lys639, two additional residues away from Gly637. Hydropathy profiles predict that a 21-residue segment corresponding to TM 8 begins at Pro660 and ends at Leu680. An acceptor site at Trp648 can be N-glycosylated, with maximal N-glycosylation one residue proximal at Gly647. The 12 + 14 rule defines the beginning of TM 8 at Met663, spaced 15 residues away from Trp648.
The N-glycosylation pattern for IC loop 5 is similar to EC loop 4, suggesting a lumenal disposition in the endoplasmic reticulum. The acceptor site at Ala735 is N-glycosylated, placing it greater than 12 residues away from the end of the preceding TM segment. This positions the end of TM segment 9 at Pro722. Similarly, the N-glycosylated acceptor site at Ala749 must be greater than 14 residues away from the beginning of TM 10. The first residue of TM 10 can be localized to Leu764, which defines the clear beginning of a very hydrophobic sequence (Fig. 7). This places the EC loop 4 insertion into EC loop 7 at Leu785 in the middle of the TM 10 (Fig. 7), which may account for its partial translocation.
Biosynthetic ConsiderationsThe finding that acceptor sites in loops must be located a minimum distance away from adjacent TM segments to be N-glycosylated shows that TM segments are held quite precisely within the endoplasmic reticulum membrane during the N-glycosylation process (23). This precision is likely due to the interaction of TM segments with elements of the translocation machinery including the Sec61p complex and TRAM (34-37). These segments assume a helical conformation before entering the hydrophobic phase of the lipid bilayer (38, 39). The positioning of these helical TM segments within the translocon may be a prerequisite to their lateral movement into the membrane.
Polytopic membrane proteins usually contain a single N-glycosylated loop, and this loop is usually the largest EC loop in the protein (12). The results obtained in this study show that acceptor sites located in small loops or those in large loops located too close to a TM segment cannot reach the active site of the lumenal oligosaccharyl transferase. This would account for the lack of N-linked carbohydrate on acceptor sites found in short loops in some polytopic membrane proteins (12). The ability to N-glycosylate loops placed in various positions within the sequence of Band 3 and other polytopic membrane proteins (13, 15, 18, 20) shows that all lumenal loops of sufficient size are scanned by the active site of the oligosaccharyl transferase. This efficiency is due to intimate association of this lumenal enzyme with the translocation machinery in the endoplasmic reticulum (40-42).
Band 3 TopologyFig. 7 presents a refined model for the folding of human erythrocyte Band 3 based on the results of scanning N-glycosylation mutagenesis. At the beginning of the membrane domain, there are five TM segments between the amino-terminal cytosolic domain and EC loop 3. Insertion of EC loop 4 into either EC loop 1 or 2 resulted in efficient N-glycosylation. The extracellular disposition of EC loop 1 is in agreement with labeling studies that show Lys430 is accessible to eosin maleimide (43) and reductive methylation (44) from the extracellular media. Also consistent with exposure of EC loop 1 is the finding that a peptide encompassing Leu427 to Gly436 can be released as a water-soluble fragment by pepsin digestion of ghost membranes (45). A recent study of the membrane insertion of truncated Band 3 molecules has shown that EC loop 2 is lumenal (46), in agreement with the results obtained in this paper. The conclusion that this loop is extracellular is not, however, supported by the examination of the orientation of Tyr486 (47). In that study, it was observed that Tyr486 could be radioiodinated by lactoperoxidase in inside-out vesicle but not in intact cells. Unfortunately, a labeled peptide corresponding to the predicted sequence Y486IVGR could not be unambiguously identified, and the yield of purified peptide was low, such that the total dpm in the final peptide fraction were only slightly above background. An expanded EC loop 3 can be N-glycosylated, and EC loop 4 contains the endogenous N-glycosylation site, consistent with the model in Fig. 1.
Contrary to the predictions of the model in Fig. 1, acceptor sites or the EC loop 4 insertion in the middle of IC loop 5 were N-glycosylated. These results suggest that during biosynthesis this loop is exposed to the lumen of the endoplasmic reticulum. This conclusion is consistent with the analysis of hydropathy profiles (6) that predicts that there is only one TM segment between residues Lys698 to Arg730 (Fig. 7). A previous study (33), however, showed that trypsin within resealed ghosts can partially cleave Band 3 at Lys743, producing a carboxyl terminal 20-kDa fragment at low yield. No such fragment could be produced by extensive treatment of intact cells with trypsin, suggesting that this site is exposed to the cytosolic side of the membrane. Soluble peptides derived from this loop can be produced by trypsin digestion of ghost membranes but only if Band 3 is first alkaline-denatured (45, 48). It appears that Lys743 is partially accessible from the cytosolic side of the membrane in the mature, folded protein but is lumenal during biosynthesis (Fig. 7). The lumenal exposure of this loop may be a feature of an intermediate in the folding pathway of Band 3. It was noted that the maximal level of N-glycosylation of sites within this loop was uniformly lower than that observed for the endogenous site (Table I). This hydrophilic loop is about 40 residues in length, which is large enough to be efficiently N-glycosylated. It is known that competition occurs between the processes of N-glycosylation and protein folding in the lumen of the ER (49). If this loop folds rapidly during biosynthesis, this may be responsible for limiting the extent of N-glycosylation. During the cell-free translation, this loop, in a fraction of the molecules, became trapped in the lumen by N-glycosylation at the novel acceptor site. Normally this loop folds into the protein to expose Lys743 to the cytosolic side of the membrane (Fig. 7). It is also possible that this part of Band 3 exhibits alternative topologies as has been reported for P-glycoprotein (50, 51) and has a dynamic aspect.
N-glycosylation acceptor sites located in the IC loop 6 were not N-glycosylated, consistent with their cytosolic location. This is in agreement with previous studies using monoclonal antibodies that localized the epitope encompassing Phe813 to Tyr824 on the cytosolic side of the membrane (52). The loop containing Lys743 and the cytosolic IC 6 are separated by a long hydrophobic sequence with a short hydrophilic and poorly conserved region (Glu777-Arg782), corresponding to putative EC loop 6 near the middle (Fig. 1). Insertion of EC loop 4 at the position of EC loop 6 resulted in poor N-glycosylation, inconsistent with a lumenal disposition for this region. We suspect that partial translocation of the EC loop 4 insertion occurred in the cell-free system due to disruption of the hydrophobic domain and its stop-transfer function. It is known that the amino-terminal halves of TM segments are not sufficient to act as a stop-transfer element in lactose permease-alkaline phosphatase fusions (53). Partial translocation and N-glycosylation of a large insert has also been observed with the glucose transporter (18).
Insertion of EC loop 4 into EC loop 7 resulted in efficient
N-glycosylation of the mutant, placing EC loop 7 on the
extracytosolic side of the membrane (Fig. 7). This agrees with the
finding that the mutation P864L is associated with the Diego blood
group antigen (54). In addition, Lys851, also located
within EC loop 7, can be labeled with H2DIDS from the cell
exterior, and a small peptide (Val849-Ala855)
can be released from the membrane by extensive proteolysis (45). An
N-glycosylation acceptor site within the carboxyl-terminal tail of Band 3 was not N-glycosylated. These results agree
with previous studies (52, 55, 56) using antibodies to map the carboxyl
terminus of Band 3 to the cytosolic side of the membrane. Cys843 is palmitoylated (57), which positions this residue
on the cytosolic side of the membrane with the fatty acid extending
into the inner leaflet of the lipid bilayer. The number of amino acids (~10) between the palmitoylated Cys843 and the last
hydrophilic EC loop (K851STP) is not sufficient to cross a
30 Å bilayer as an -helix. This short sequence is not likely to be
exposed to the lipid bilayer but would be protected within a bundle of
outer helices formed by other TM segments (58).
We thank Dr. David Andrews (McMaster University) and MBI Fermentas (Flamborough, Canada) for the generous donation of high quality microsomal membranes and a coupled transcription/translation system.