Molecular characterization of the band 3 protein from Southeast Asian ovalocytes.

Southeast Asian ovalocytosis (SAO) is a hereditary form of elliptocytosis resulting in rigid, oval-shaped erythrocytes resistant to invasion by malaria parasites. The molecular defect is due to deletion of codons 400-408, encoding a 9-amino-acid sequence located at the boundary between the cytosol and the first transmembrane segment in Band 3, the erythrocyte anion transport protein. We have carried out an extensive characterization of Band 3 isolated from SAO erythrocytes which contain about 50% mutant Band 3. A slightly higher proportion of Band 3 in SAO erythrocytes was left associated with the cytoskeleton after extraction of ghost membranes with non-ionic detergents. Size exclusion high performance liquid chromatography analysis showed that SAO Band 3 contained a higher proportion of tetramers relative to dimers (50% tetramer) than normal Band 3 (33% tetramer). The circular dichroism spectrum of Band 3 from SAO erythrocytes was very similar to the spectrum for normal Band 3. Enzymatic deglycosylation and tomato lectin binding showed that SAO Band 3 lacked the polylactosaminyl oligosaccharide found on normal Band 3. SAO Band 3 was unable to bind the anion transport inhibitor 4-benzamido-4'-aminostilbene-2,2'-disulfonate, suggesting a dramatic alteration in the inhibitor binding site. In conclusion, deletion of 9 amino acids from Band 3 on the cytosolic side of the membrane affects the properties (glycosylation and inhibitor binding) of Band 3 on the opposite side of the membrane without dramatic changes in the secondary and quaternary structure of the protein.

malarial parasites Plasmodium knowlesi and Plasmodium falciparum (Kidson et al., 1981;Hadley et al., 1983). However, individuals with SA0 are asymptomatic and are not adversely affected by the altered red cell morphology (Liu et al., 1990). The mechanism by which the SA0 mutation affords resistance to malarial invasion has not been determined. Alteration of cytoskeletal-membrane interaction is implicated since cytoskeletal rearrangements are required for erythrocyte invasion by malarial parasites (Bannister and Dluzewski, 1990). In addition to the ovalocytic shape of SA0 red cells, red cell ghosts prepared from cells of SA0 individuals undergo saltinduced shape changes which control individuals do not (Liu et al., 1990). SA0 red cells are very rigid, (Mohandas et al., 1984;Saul et al., 1984;Schofield et al., 199213) perhaps due to increased interaction of the cytoplasmic domain of Band 3 with the cytoskeleton, since the rotational (Tilley et al., 1991) and lateral (Liu et al., 1990;Mohandas et al., 1992) mobilities of Band 3 are reduced in SA0 red cells.
The molecular basis for SA0 was recently identified as a 9amino-acid deletion (residues 400-408) in Band 3 (Jarolim et al., 1991;Tanner et at., 1991;Mohandas et al., 1992;Schofield et al., 1992a). Band 3, the 95-kDa anion-exchange protein of the erythrocyte membrane, has been cloned and sequenced (Kopito and Lodish, 1985;Tanner et al., 1988). The protein is composed of two domains: an NH2-terminal 43-kDa cytoplasmic domain and a 52-kDa membrane domain, predicted to span the the bilayer 12-14 times (Kopito and Lodish, 1985). The membrane domain alone is responsible for anion-exchange activity (Grinstein et al., 1979;Lepke et al., 1992). The mutation spans the boundary between the cytoplasmic domain and the predicted first transmembrane segment of Band 3. Heterozygotes for SA0 are afforded protection against malarial infection, but the homozygous state is probably lethal since no homozygotes have been found (Jarolim et al., 1991). In all alleles coding for SA0 Band 3, an associated mutation ( L y P + Glu) known as the "Memphis mutation" (Yannoukakos et al., 1991), was found. The Memphis mutation alters the electrophoretic mobility of Band 3 (Mueller and Morrison, 1977). This allows identification of SA0 and control Band 3 by SDS-polyacrylamide gel electophoresis of proteolytic fragments of Band 3 (Palatnik et al., 1990).
The cytoplasmic domain of Band 3 links the cytoskeleton to the plasma membrane by binding to ankyrin and Band 4.2 protein (Bennett, 1990). Band 3 is associated in the membrane as a mixture of dimers and tetramers, and the tetramers are thought to interact preferentially with the cytoskeleton (Casey and Reithmeier, 1991). The SA0 mutation may increase Band 3 binding to the intermediary protein, ankyrin (Liu et al., 1990). Anion-exchange is sensitive to inhibition by stilbene disulfonates such as DIDS, BADS, and 4-acetamido-4'-isothiocyano-2,2'-disulfonate, which bind to an extracellular site on Band 3 (Cabantchik and Greger, 1992). Marked alteration of the membrane domain of SA0 Band 3 is indicated by the inability to label the protein with several anionexchange inhibitors (Moriyama et al., 1992;Schofield et al., 1992a). Moreover, the protein is not able to carry out anionexchange (Moriyama et dl., 1992;Schofield et al., 1992a). Differential scanning calorimetry of SA0 membranes has shown that the membrane domain is largely denatured (Moriyama et al., 1992). In contrast, the cytoplasmic domain retains its normal structural and functional properties (Moriyama et al., 1992).
T o understand the consequences of the SA0 deletion upon Band 3 structure, we have undertaken a characterization of Band 3 isolated from individuals with SAO.

EXPERIMENTAL PROCEDURES
Materials-Biotinylated tomato lectin was purchased from Sigma. Triton X-100, chymotrypsin, and glycosidase F mixture, containing both endoglycosidase F and N-glycosidase F, were obtained from Boehringer Mannheim, Germany. Vectastain@ was obtained from Vector Laboratories. C12E8 was purchased from Nikko Chemical Co., Tokyo. Proteins used as Stokes radii standards were supplied by Pharmacia LKB Biotechnology Inc. The SEC 4000 columns were from Beckman. Blood was collected from control individuals and individuals with SA0 into acid-citrate-dextrose as anticoagulant. All other chemicals were reagent grade or better.
Quantitating the Fraction of Mutant Band 3-Red cells (25% hematocrit in 0.1 M NaCl, 5 mM sodium phosphate, pH 7.4) were treated at 37 "C with 0.5 mg/ml of chymotrypsin, to digest Band 3 at a known exofacial site (Palatnik et al., 1990). Digestion was stopped by addition of phenylmethylsulfonyl fluoride to 1 mM and incubation for 10 min. Band 3 was then prepared from the cells as described below. Upon SDS-polyacrylamide gel electophoresis, SA0 Band 3 differed from normal Band 3 in the migration position of the 60-kDa chymotryptic fragment (Palatnik et al., 1990). The proportions of the two forms of Band 3 were quantitated by gel scanning the Coomassie Blue-stained gel, using a Hoefer GS 300 densitometer and Rainin MacIntegratorTM software.
Measurement of the Fraction of Band 3 Bound to the Cytoskeleton-The fraction of Band 3 retained by the cytoskeleton was determined in parallel for erythrocytes from both control and SA0 individuals. Red cell ghosts (4 mg protein/ml) were extracted with 5 volumes of either 1% (v/v) C12E8, 5 mM sodium phosphate, pH 8.0, or 1% (v/v) Triton X-100,5 mM sodium phosphate, pH 8.0, and incubated on ice for 15 min. The cytoskeleton and any associated Band 3 were pelleted by centrifugation at 35,000 revolutions/minute in a Beckman Ti-75 rotor. The supernatant was removed, and the tubes were each made up to their original volume with 5 mM sodium phosphate, pH 8.0. Aliquots of the sample before centrifugation and of the pellet and supernatant after centrifugation were solubilized with sample buffer (Laemmli, 1970). After electrophoresis, the amount of Band 3 in each lane was quantitated by scanning, as above.
Isolation of Band 3 and the Membrane Domain-Band 3 was prepared as previously described (Casey et al., 1989). Briefly, red cells were washed, ghosts were prepared by osmotic hemolysis, and the membranes were stripped of peripheral proteins with ice-cold 2 mM EDTA, pH 12. Membranes were solubilized with 1% CEER (v/v), applied to aminoethyl-Sepharose 4B, and eluted with a 0-0.25 M linear NaCl gradient in 0.1% C12E8 (v/v), 5 mM sodium phosphate, pH 8.0. The 52-kDa membrane domain of Band 3 was prepared by trypsin treatment of red cell ghosts, followed by alkali stripping, solubilization in CI2E8, and DEAE-Sepharose chromatography (Casey et al., 1989).
Deglycosylation of Band 3-Purified Band 3 (0.5-2 mg protein/ml) in 0.1% CI2E8, 100 mM NaCl, 5 mM sodium phosphate, pH 8.0, was treated with FIN-glycosidase F mixture (2 units/mg protein) at room temperature for 24 h at room temperature (Casey et al., 1992). Control samples were incubated under identical conditions, including a blank solution of the same composition as the enzyme solution. Cleavage of the carbohydrate chain from Band 3 was assessed by an increase in mobility and sharpening of the protein band in Laemmli gels (Laemmli, 1970) and by the loss of binding of tomato lectin, which detects poly-N-acetyllactosamine structures (Merkle and Cummings, 1987) on immunoblots. Tomato lectin blots were performed as described (Casey et al., 1992).
High Performance Liquid Chromatography (HPLC)-Chromatog-raphy was performed using a 0.75 X 30-cm SEC 4000 column. A Spectra-Physics SP8800 HPLC pump was used, at a flow rate of 0.5 ml/min. Typically, 5 pl of purified Band 3 (1-2 mg protein/ml) was injected onto the column using a 2O-pl injection loop. Protein elution was monitored at 215 nm, using a Spectroflow 757 Flowthrough Absorbance Detector (AB1 Analytical Co.). The elution buffer contained 0.1% (v/v) C12E8, 100 mM NaCl in 5 mM sodium phosphate, pH 7.0. The column was calibrated with suitable protein standards which do not bind detergent (LeMaire et al., 1986). Chromatograms were recorded using MacIntegratorTM software (Rainin Instruments) connected to a Macintosh computer. BADS Binding Assay-The binding of BADS to Band 3 was measured by fluorescence enhancement in a Spex Fluorolog fluorimeter (Casey et al., 1989). Excitation wavelength was 280 and 340 nm and emission was 450 nm. As described (Casey et al., 1989), 2 ml of 28.5 mM sodium citrate buffer, pH 7.4, was placed in a 3-ml fluorescence cell, and 100 pl of Band 3 (0.5 mg protein/ml) in 0.1% C12ER, 100 mM NaCl, 5 mM sodium phosphate, pH 7.4, was added. Band 3 samples were titrated with concentrated BADS to a final concentration of 12 p~. Fluorescence was corrected for dilution, selfquenching of the probe, and the background fluorescence of the sample and the probe.
Circular Dichroism-Band 3 in 0.1% C12E8 solution was diluted 10fold into distilled water and spectra were recorded at room temperature, in a quartz cell with a 0.1-cm pathlength, using a Jasco 5-720 spectropolarimeter. Spectra for buffer alone were subtracted from each spectrum. Protein determinations were performed (Lowry et al., 1951) and corrected to molar concentration (Casey et al., 1989). Mean residue ellipticities were calculated using the mean residue weight of Band 3, 111.7, based on the sequence of human Band 3 (Tanner et al., 1988). Secondary structures were estimated by best-fit over the whole spectrum, using Jasco software and reference spectra from Yang et al. (1986).
Analytical Techniques-Polyacrylamide gel electrophoresis was performed according to Laemmli (1970). Protein assays were performed using bovine serum albumin as standard (Lowry et al., 1951). R, SDS-PAGE of ghosts prepared from red cells that had been treated with chymotrypsin. Lower arrow indicates the 60-kDa fragment from the wild-type form of Band 3, and the upper arrow marks the 63-kDa SA0 form. of SA0 Band 3 (M, = 63,000) is due to a point mutation (Memphis Band 3, LysS6 + Glu) in SA0 Band 3 (Mueller and Morrison, 1977) that is present in addition to the 9-aminoacid deletion (Ala4m-Ala408). Quantitation of the proportion of 63,000 and 60,000 fragments by gel scanning showed that SA0 Band 3 made up 48% of the Band 3 in ghost membranes. The production of the 63,000-dalton fragment indicates that SA0 Band 3 retains the exofacial chymotrypsin sensitive site.

Proportion of Mutant Band 3 in
Glycosylation-The broad nature of Band 3 in SDS gels is due in part to the heterogeneous nature of the oligosaccharide chain attached to A d 4 ' (Tsuji et al., 1980;Fukuda et al., 1984). Purified samples of normal and SA0 Band 3 were treated with N-glycosidase F to remove the oligosaccharide chain. The mobility of Band 3 on SDS gels increased upon deglycosylation, with a considerable sharpening of the protein band for Band 3 isolated from both control and S A 0 erythrocytes (Fig. 2). Deglycosylated Band 3 from SA0 individuals ran as a broader band than Band 3 from normal erythrocytes with some overlap between the leading edge of S A 0 Band 3 (panel A, lane 3) and the trailing edge (panel A, lane 4 ) after deglycosylation. The slower migrating deglycosylated protein may represent the SA0 form, as is the case for the aminoterminal chymotryptic fragment shown in Fig. 1. Band 3 contains a polylactosaminyl oligosaccharide structure that binds tomato lectin (Fukuda et al., 1984;Casey et al., 1992).

Blotting of the control and SA0 Band 3 with tomato lectin
showed that the staining of the SA0 sample was reduced by half relative to the control sample (Fig. 2, panel B ) . Since SA0 Band 3 makes up 48% of the sample, the dramatic decrease in tomato lectin binding suggests that SA0 Band 3 does not contain the polylactosaminyl oligosaccharide. Association of Band 3 with the Cytoskeleton-SA0 Band 3 has been reported to have an enhanced interaction with the cytoskeleton (Liu et al., 1990;Tilley et al., 1991;Mohandas et al., 1992). Ghost membranes prepared from control and chymotrypsin-treated cells were therefore selectively extracted with Triton X-100 or CI2E8 to determine the proportion of Band 3 that was bound to the detergent-insoluble cytoskeleton. Triton X-100 (1%) solubilized 77% of Band 3 from control ghost membranes and 66% from S A 0 ghost membranes, the remainder associating with the cytoskeleton. Similarly, (1%) solubilized less Band 3 from SA0 ghosts (50%) than from control ghosts (66%). The results indicate that slightly more (11-16%) SA0 Band 3 was bound to the cytoskeleton, but certainly not equal to the amount of mutant protein (48%). Oligomeric Structure-Tetramers of Band 3 are the oligomeric form of the protein immobilized on the cytoskeleton (Casey and Reithmeier, 1991). The oligomeric structure of Band 3 in control and SA0 samples was compared by sizeexclusion HPLC (Fig. 3). Previous studies have shown that the major peak represented Band 3 dimers (67%) and the leading shoulder (33%) was Band 3 tetramers and higher oligomers (Casey and Reithmeier, 1991). The HPLC analysis (Fig. 3) showed that there was only a slight increase in the proportion of tetrameric Band 3 in the SA0 sample (50%). This increase in tetramer may account for the slightly enhanced binding of Band 3 to the cytoskeleton. The increase in tetramer was not, however, equivalent to the amount of mutant protein.
Circular Dichroism Spectroscopy-The deletion of amino acids predicted to be in the first transmembrane segment suggests that the conformation of the transmembrane segment must change to span the membrane (e.g. helix to extended) or that adjacent segments of Band 3 normally exposed to water must be pulled into the membrane. The cytosolic sequence that precedes the deleted residues (Ala4w-Ala40R), if plunged into a hydrophobic environment, would assume a helical conformation. The deletion of a portion of the first transmembrane segment may also result in a more global change in Band 3 structure due to disruption of helix-helix interactions within the membrane domain. Conformational changes of this type may be detected by circular dichroism spectroscopy. The circular dichroism spectra of Band 3 purified from normal and SA0 erythrocytes are presented in Fig.  4. No significant difference in the spectra were detected, suggesting that any conformational differences that may exist between normal and SA0 Band 3 are too small to be detected by this method. The membrane domains of normal and SA0 Band 3 were also purified and characterized by circular dichroism spectroscopy; however, no differences were found in the circular dichroism spectra of these samples.
Inhibitor Binding-The binding of stilbene disulfonates such as BADS to Band 3 provides a sensitive assay for the native state of the protein (Casey et al., 1989). Fig. 5 compares the BADS titration of normal and SA0 Band 3. Both samples bound the inhibitor with the same affinity (1 p~) ; however, the maximum fluorescence signal of the SA0 sample was half that of the control sample at equal protein concentrations. Similar results were obtained by exciting the probe directly at 340 nm and with the purified membrane domains. These results show that SA0 Band 3 is completely incapable of binding stilbene disulfonates.

DISCUSSION
In this paper we have noted two dramatic alterations in SA0 Band 3, both of which involve the external aspect of the protein. The first change is in the carbohydrate structure of the protein. Normal Band 3 receives the addition of a polylactosaminyl moiety, a process that occurs in the trans-Golgi (Tsuji et al., 1980;Fukuda et al., 1984). SA0 Band 3 escapes this modification. This may be due to an altered structure of the protein such that the recognition elements for this modification are lost, or that SA0 Band 3 does not reside for a sufficient length of time in the Golgi compartment responsible for the modification. The change in oligosaccharide structure would not, however, compromise the functioning of the protein as a transporter since we have shown (Casey et al., 1992) that the deglycosylated protein is fully capable of carrying out anion-exchange. It is possible that SA0 Band 3 is not glycosylated. Attempts to separate normal and SA0 Band 3 by lectin affinity chromatography have not been successful. This The second alteration is more serious. The stilbene disulfonate-binding site in SA0 Band 3 has been destroyed by the S A 0 mutation. This site faces the cell exterior. Since the deletion is in the first transmembrane helix, this suggests that this helix may form a part of the inhibitor binding site. Alternatively, the structure of the protein may be changed to such an extent that the stilbene site is altered indirectly by the change in the first transmembrane helix. It is significant that the deletion encompasses a portion of the cytosolic domain and the first quarter of the first transmembrane segment, yet the effects are felt on the cell exterior. The deletion must cause a transmembrane effect and suggests that portions of the protein other than the first helix are affected by the mutation. The loss of the stilbene disulfonate-binding site accounts for the inability to covalently label Band 3 with tritiated DIDS (Schofield et al., 1992a).
The mutation does not, however, affect the secondary struc--I 0 2 4 6 8 10 12 14 Elution Volume (ml) ture of Band 3 as determined by circular dichroism measurements of the purified protein. The same conclusion was reached by Moriyama et al. (1992) by comparison of the circular dichroism spectra of normal and SA0 inside-out vesicles which are enriched in Band 3. In addition, the SA0 Band 3 forms dimers and higher oligomers, and it binds to the cytoskeleton. The regions of Band 3 involved in dimer and tetramer formation and cytoskeleton binding are not altered greatly by the SA0 mutation. The membrane domain of Band 3 is dimeric and, due to its high helical content, the interactions responsible for dimer formation likely involve helix-helix interactions. Since S A 0 Band 3 forms oligomers, it is likely that the first transmembrane helix is not involved in dimer formation. S A 0 Band 3 is present in about equal amounts as normal Band 3 in SA0 ovalocytes. This indicates that the biosynthesis, movement from the endoplasmic reticulum through the Golgi to the plasma membrane, and the stability of the mutant protein are quite normal. One issue of importance is the possibility that SA0 Band 3 forms heterodimers or higher oligomers with normal Band 3. Attempts to The excitation wavelength was 280 nm and the emission wavelength was 450 nm. Purified Band 3 (100 pl) in 0.1% ClzEa (v/v), 0.1 M NaC1, 5 mM sodium phosphate, pH 8.0, was diluted into 2 ml of 28.5 mM sodium citrate, pH 7.0, at 28 "C, to a final protein concentration of 50 rg/ml. separate normal and SA0 Band 3 by lectin or inhibitor affinity chromatography have not been successful, suggesting that hetero-oligomers may exist. The fact that SA0 Band 3 affects the thermal transition of normal Band 3 supports this view (Moriyama et al., 1992). If heterodimers form, then the mutant protein may "piggyback" its way to the cell surface using the normal protein. The stilbene binding results suggest that if heterodimers exist, then the ability of normal Band 3 to bind the inhibitor or transport is not seriously compromised when bound to a totally defective protein.