Effect of Band 3 Subunit Equilibrium on the Kinetics and Affinity of Ankyrin Binding to Erythrocyte Membrane Vesicles*

The membrane-spanning protein, band 3, anchors the spectrin-based membrane skeleton to the lipid bilayer via the bridging protein, ankyrin. To understand how band 3 subunit stoichiometry influences this membrane-skeletal junction, we have induced changes in the band 3 association equilibrium and assayed the kinetics and equilibrium properties of ankyrin binding. We observe that band 3 oligomers convert slowly to dimers and ultimately monomers following removal of ankyrin. Addition of excess ankyrin back to these membranes enriched in dissociated band 3 then shifts band 3 almost entirely to tetramers, confirming that the tetrameric form of band 3 constitutes the preferred oligomeric state of ankyrin binding. 4,4′-Diisothiocyanostilbene-2,2′-disulfonic acid (DIDS) labeling of band 3, which is shown to shift most of the band 3 population to dimers, eliminates the majority of ankyrin-binding sites on the membrane and greatly reduces retention of band 3 in detergent-extracted membrane skeletons. Furthermore, DIDS− modified membranes lack all low affinity ankyrin-binding sites and roughly half of all high affinity sites. Since labeled membranes lack the rapid kinetic phase of ankyrin binding and exhibit only half of the normal amplitude of the slow kinetic phase, it can be concluded that the rapid phase of ankyrin association involves low affinity sites and the slow phase involves high affinity sites. A model accounting for these data and most previous data on ankyrin-band 3 interactions is provided.

The membrane-spanning protein, band 3, anchors the spectrin-based membrane skeleton to the lipid bilayer via the bridging protein, ankyrin. To understand how band 3 subunit stoichiometry influences this membraneskeletal junction, we have induced changes in the band 3 association equilibrium and assayed the kinetics and equilibrium properties of ankyrin binding. We observe that band 3 oligomers convert slowly to dimers and ultimately monomers following removal of ankyrin. Addition of excess ankyrin back to these membranes enriched in dissociated band 3 then shifts band 3 almost entirely to tetramers, confirming that the tetrameric form of band 3 constitutes the preferred oligomeric state of ankyrin binding. 4,4-Diisothiocyanostilbene-2,2-disulfonic acid (DIDS) labeling of band 3, which is shown to shift most of the band 3 population to dimers, eliminates the majority of ankyrin-binding sites on the membrane and greatly reduces retention of band 3 in detergentextracted membrane skeletons. Furthermore, DIDS ؊ modified membranes lack all low affinity ankyrin-binding sites and roughly half of all high affinity sites. Since labeled membranes lack the rapid kinetic phase of ankyrin binding and exhibit only half of the normal amplitude of the slow kinetic phase, it can be concluded that the rapid phase of ankyrin association involves low affinity sites and the slow phase involves high affinity sites. A model accounting for these data and most previous data on ankyrin-band 3 interactions is provided.
Band 3 (M r ϳ101,000) is the predominant polypeptide of the human erythrocyte membrane, comprising ϳ25% of the total membrane protein. Band 3 contains two major structural domains, a membrane-spanning domain that may traverse the bilayer 14 times in ␣-helical segments (1)(2)(3), and an exposed cytoplasmic domain that exhibits an elongated segmented morphology (4,5). The membrane-spanning domain (M r ϳ 55,000) catalyzes anion transport across the phospholipid bilayer (6 -8). It also serves as the major antigen responsible for immunemediated removal of senescent and abnormal erythrocytes (9 -16). The cytoplasmic domain participates in the mechanism of senescent/abnormal cell removal (9 -15), binds and regulates glycolytic enzymes (17,18), and provides the major link of the spectrin-based membrane skeleton to the bilayer (19,20). This latter function is mediated by ankyrin, a protein that connects the ␤ subunit of spectrin to the cytoplasmic domain of band 3 in a manner that is sensitive to pH (21), the concentration of diphosphoglycerate (22), and the association state of band 3.
As a model of membrane-skeleton junctions, the ankyrinband 3 interaction has been heavily investigated. The sites of ankyrin association on the cytoplasmic domain of band 3 have been localized to two distinct regions: one near a central proteolytically accessible hinge and a second more proximal to the anion transporter's N terminus (30 -34). Similarly, two sites of band 3 interaction with ankyrin have been identified and shown to separately reside in ankyrin repeats 7-12 and 13-24 (35). Perhaps related to this structural complexity is a similar complexity in the kinetics, affinity and magnitude of ankyrin binding to band 3. Thus, ankyrin association with the anion transporter in its membrane environment is characterized by populations of both low affinity (K D ϳ 1.5 ϫ 10 Ϫ7 M) and high affinity (K D ϳ 1.5 ϫ 10 Ϫ8 M) sites (21,36). Furthermore, the low affinity sites gradually convert to high affinity sites during prolonged incubation, and a slow accrual of new sites is also observed as the binding reaction approaches completion (21). Although only ϳ100,000 ankyrin sites are detected in freshly prepared membranes (21, 36 -38), during extended incubations at pH 6.35 with saturating ankyrin concentrations approximately 270,000 sites/cell can eventually be measured (21). Because removal of the cytoplasmic domain of band 3 eliminates virtually all ankyrin interactions with the red cell membrane (36,37), it is assumed that band 3 somehow accounts for all of the kinetic and equilibrium phases of ankyrin association. It is the goal of this paper to place these unexplained kinetic and equilibrium binding data on a more firm physical foundation by identifying the sources of the fast and slow phases of ankyrin binding as well as the causes of the high and low affinity populations of sites.

EXPERIMENTAL PROCEDURES
Materials sodium phosphate, pH 7.4. Cells were washed four times with 5 volumes of the same buffer containing 1% bovine serum albumin to remove unreacted DIDS and then three times without the serum albumin. Under these conditions, Ͼ95% of the DIDS resides on band 3 (39).
Ankyrin Purification and Binding Assay-Erythrocyte ankyrin and the 1 M KI-extracted IOV for binding studies were prepared as described by Bennett (40) with minor modifications. Briefly, ankyrin was released with 0.5 M KCl from Triton X-100 extracted red cell membrane skeletons, precipitated with 32% ammonium sulfate, and purified by gel filtration chromatography on a Sephacryl S-300 HR column (2.8 ϫ 120 cm) in 0.6 M NaBr, 0.1 mM EDTA, 20 mM sodium phosphate, 0.05% NaN 3 , 0.5 mM dithiothreitol, and 0.5% Brij 35, pH 7.5. Labeling and assay of 125 I-ankyrin binding to KI-IOV were performed as described (21,40), the latter conducted in a buffer consisting of 5% sucrose, 50 mM sodium phosphate, 50 mM boric acid, 30 mM NaCl, 1 mM EDTA, 0.2 mM dithiothreitol adjusted to pH 6.5 or 7.8 as noted. An analysis of the band 3 dimer:tetramer ratio was not conducted on the specific KI-IOV preparations used for the ankyrin binding studies, but an examination of a similar KI-IOV preparation revealed a dimer:tetramer ratio of ϳ60:40. The more extensive KI stripping and 37°C incubation procedures employed during preparation of band 3 for HPLC analyses yields a much lower fraction of band 3 tetramer (see "Discussion").
Ankyrin Fragment Purification-The 46-kDa fragment of ankyrin which contains the ankyrin-binding site for band 3 was expressed in Escherichia coli strain BL21 (DE3)/pLysS containing the ankyrin sequence for amino acids 403-827 plus an additional Met-Ala-Ser at the N terminus (expression vector a kind gift of Vann Bennett). Protein expression and purification were performed as described by Davis and Bennett (41). Briefly, plasmid protein synthesis was induced in cultures grown to an A 650 nm of 0.5 with 1 mM isopropyl-␤-thiogalactopyranoside for 2 h. The cells were harvested and washed with 10 mM sodium phosphate, 100 mM NaCl, pH 7.5, and then digested with lysozyme (1 mg/ml) and DNase I (25 g/ml) in 50 mM sodium phosphate, 1 mM sodium azide, 25% sucrose, 10 mM magnesium chloride, pH 8.7. The protein was solubilized with 10 mM sodium phosphate, 100 mM NaCl, 2 mM EDTA, 1 mM sodium azide, 1 mM dithiothreitol, 1% Triton X-100, and 1% deoxycholate, pH 7.5, then separated from unsolubilized material by spinning at 20,000 ϫ g for 30 min. The ankyrin fragment was precipitated with 32% ammonium sulfate and purified on a Sephacryl S-300 gel filtration column in 10 mM sodium phosphate, 1 M sodium bromide, 1 mM EDTA, 1 mM sodium azide, 1 mM dithiothreitol, pH 7.4. Peak fractions were collected and dialyzed against HPLC incubation buffer containing 5 mM sodium phosphate, pH 8, amended with 20 mM 2-mercaptoethanol, 1 mM dithiothreitol, 80 g/ml phenylmethanesulfonyl fluoride, 20 g/ml leupeptin, and 20 g/ml pepstatin A. Ankyrin fragment protein concentration was estimated by absorbance measurements at 280 nm using an extinction coefficient of 0.347 ml mg Ϫ1 cm Ϫ1 which we determined by the Edelhoch method (42).
Analysis of Band 3 Retention in Detergent-extracted Membrane Skeletons-Quantitation of intact band 3 in membrane skeletal fractions was difficult due to the diffuse nature of the anion transporter's banding pattern in sodium dodecyl sulfate-polyacrylamide gels. To circumvent this obstacle, whole erythrocytes (50% hematocrit) were digested with 1 mg/ml chymotrypsin at 37°C for 1 h to cleave band 3 into its nonglycosylated 65,000 dalton and heterogeneously glycosylated 35,000-dalton fragments. Previous studies have shown this cleavage to have little effect on anion transport, DIDS binding, or membrane skeletal interactions (39), while allowing accurate quantitation of the 65,000-dalton fragment of band 3 by SDS-PAGE.
To evaluate the band 3 content retained in the membrane skeleton, the cleaved cells (50% hematocrit) were either left unmodified or reacted with 50 M DIDS for 1 h at 37°C. The reaction was stopped by washing 3 times with phosphate-buffered saline containing 1% bovine serum albumin and 2 times with phosphate-buffered saline containing 10 mM glucose, 2 mM inosine, and 150 g/ml ampicillin. After the desired incubation period at 10% hematocrit in the latter buffer, membrane skeletons were isolated by mixing 500 l of packed cells with 500 l of phosphate-buffered saline. This suspension was then treated with 1 ml of 4% n-octyl glucoside, octyl-␤-D-thioglucopyranoside, Zwittergent 3-14, or Triton X-100 supplemented with 0.5 mM dithiothreitol, 20 g/ml leupeptin, 20 g/ml pepstatin A, 1 mM EDTA, and 80 g/ml phenylmethanesulfonyl fluoride. The membrane skeletal extract was loaded onto a 35% sucrose cushion and the sample was centrifuged at 85,500 ϫ g for 90 min to separate the solubilized membrane components from the pelleted membrane skeletons. The tubes were inverted to allow the supernatant to drain, and the pellet was solubilized in SDS buffer and examined by SDS-PAGE.
High Performance Gel Filtration Chromatography of Isolated Band 3-Red cells were either left unmodified or reacted with DIDS, and KI-IOV were prepared as described previously (43) with several modifications. Washed erythrocytes were lysed once with 40 volumes of ice-cold 5 mM sodium phosphate, 0.5 mM EDTA, 0.5 mM sodium azide, 80 g/ml phenylmethanesulfonyl fluoride, pH 8.0, and centrifuged at 23,430 ϫ g for 10 min. The red ghosts were resuspended in a 20-fold volume excess of buffer containing 0.5 mM EDTA, 0.5 mM dithiothreitol, 80 g/ml phenylmethanesulfonyl fluoride, pH 8.0, at 4°C, warmed to 37°C over 30 min, and pelleted at 23,430 ϫ g for 20 min to remove spectrin and actin. The remaining peripheral proteins were removed by addition of a 100-fold excess of 1 M potassium iodide, 25 mM sodium phosphate, 1 mM EDTA, 0.5 mM dithiothreitol, 80 g/ml phenylmethanesulfonyl fluoride, pH 7.5, at 4°C, and incubation of the pink insideout vesicles at 37°C for 30 min. The resulting white KI-IOVs were pelleted for 30 min at 24,300 ϫ g, then washed twice with 5 mM sodium phosphate, pH 8.0, supplemented with 0.5 mM dithiothreitol. The stripped membranes were then incubated for varying periods of time in 5 mM sodium phosphate, pH 8.0, containing 20 mM 2-mercaptoethanol, 1 mM dithiothreitol, 80 g/ml phenylmethanesulfonyl fluoride, 20 g/ml leupeptin, and 20 g/ml pepstatin A. After solubilizing 5 to 15 mg of membrane protein/ml suspension by vortexing in 5 volumes of 1% C 12 E 8 in 5 mM sodium phosphate, pH 8.0, the samples were centrifuged at 43,000 ϫ g for 30 min to remove any residual particulates. HPLC of solubilized band 3 (ϳ5 l of ϳ1 to 3 mg/ml) was performed at room temperature at a flow rate of 0.5 ml/min, as described by Casey and Reithmeier (25), only the column used was a 7.5 ϫ 300-mm Toso Haas TSK-4000 SW XL (Tokyo, Japan). The column buffer was 5 mM NaH 2 PO 4 , 100 mM NaCl, 0.1% C 12 E 8 , pH 7.0. Protein elution was monitored at 214 nm and sample volumes of 2-10 l were collected. The presence of band 3 in peak fractions was confirmed both by Western blotting with antibodies against the cytoplasmic domain of band 3 and by scanning for emission of DIDS fluorescence ( ex , 355 nm; em , 450 nm). Band 3 tetramer, dimer, and monomer elution peaks were assigned according to the retention times reported by Casey and Reithmeier (25) using the same detergent (0.1% C 12 E 8 ), the same molecular weight standards, and a similar gel filtration HPLC column. Since Casey and Reithmeier performed a rigorous analysis of the Stokes radii of the various oligomeric forms of band 3 in C 12 E 8 detergent solution, we have simply adopted their peak assignments.
Analysis of Ankyrin's Effect on the Oligomeric Equilibrium of Band 3-It has been noted previously that band 3 exists in a slow subunit association equilibrium in the membrane and that shifts in this equilibrium are halted upon solubilization of the membrane in nondenaturing detergent (25). Therefore, to evaluate the effect of ankyrin on this equilibrium, KI-IOVs from either unmodified or DIDS-labeled erythrocytes were allowed to equilibrate for approximately 150 h at 37°C. By this time, Ͼ90% of unmodified band 3 and ϳ50% of DIDS-labeled band 3 had shifted to an elution position corresponding to the monomer; the remaining approximately 50% of the DIDS-modified band 3 eluted as the dimer (see below). The 46.5-kDa ankyrin fragment was then added at the molar ratio of ankyrin to band 3 used in the binding assays (1:4), and following the desired re-equilibration time at 37°C, the ankyrin fragment was removed by 10 min extraction at 37°C in 1 M potassium iodide, 25 mM sodium phosphate, 1 mM EDTA, 1 mM dithiothreitol, pH 7.5. After washing the membranes in 5 mM sodium phosphate, pH 8, and solubilizing in 5 volumes of the same buffer containing 1% C 12 E 8 , the sample was clarified by centrifuging at 43,000 ϫ g for 30 min and immediately analyzed by HPLC, as described above. Membrane concentrations ranged from 5 to 15 mg/ml and volumes injected onto the column ranged from 1 to 10 l. SDS-PAGE was also performed on each fraction to ensure that the band 3 remained intact during the various equilibration and chromatography steps.

Effect of DIDS and Ankyrin on the Oligomeric State of Band
3-There are conflicting data regarding the influence of DIDS on the subunit dissociation equilibrium of band 3 (25,44). Therefore, we have re-examined the issue by HPLC gel filtration chromatography using several different membrane preparations varying in their lengths of equilibration at 37°C after removal of ankyrin and prior to detergent extraction of band 3. Other laboratories have shown that when freshly prepared IOV are stripped of peripheral proteins, solubilized in 1% C 12 E 8 , and separated by HPLC, the solubilized band 3 elutes as a mixture of predominately dimers with some tetramers and higher oligomers (25,44). We have reproduced these observa-tions, measuring a distribution of ϳ40% tetramer and 60% dimer, but we have also noted that the ratio between the dimeric and tetrameric forms of band 3 can be significantly affected by the membrane preparation procedure employed (data not shown). In order to keep band 3 native and to avoid unwanted aggregation/denaturation, the ghosts we employed for our HPLC analyses were not extensively washed prior to KI-IOV preparation. Furthermore, since we and others have found that NaOH stripping of peripheral proteins denatures the cytoplasmic domain of band 3 (37,(45)(46)(47), peripheral proteins were more gently (but thoroughly) removed by 1 M KI incubation. We additionally did not perform an anion exchange chromatography step before size analysis of band 3 by gel filtration chromatography, since anion exchange chromatography has been shown to remove essential lipids surrounding the solubilized band 3 and cause its aggregation (29,48). In our hands, the band 3 was predominantly dimeric in both the unlabeled and DIDS-labeled preparations by the time the KI-IOV were solubilized in detergent solution (Fig. 1). Presumably, the KI stripping procedures at 37°C allowed dissociation of most band 3 tetramers to dimers. Furthermore, as the KI-IOV were further incubated at 37°C prior to solubilization, a gradual decrease in abundance of the dimeric band 3 was observed with a concomitant rise in a lower molecular weight species which we presume to be the monomer, but alternatively could be a hydrodynamically smaller dimer (Fig. 1A). Importantly, prelabeling of the cells with DIDS retarded this further dissociation of band 3 to the monomer (Fig. 1B).
The influence of DIDS labeling on the band 3 subunit equilibrium was also manifested when the 46,500-dalton band 3-binding fragment of ankyrin (49) was added to a 4 molar excess of band 3 in the KI-IOV membranes to drive the band 3 subunit equilibrium back toward the tetramer, i.e. the preferred oligomeric state for ankyrin association (21,27,28). As seen in Fig. 2B (dotted line), unlabeled membranes allowed to incubate until band 3 dissociated to predominantly its monomeric state were shifted almost quantitatively to the tetrameric state upon 24 h incubation with ankyrin (solid line). This reassociation reaction occurred very slowly, with an approximate half-time of 12 h (data not shown). DIDS-labeled membranes, in contrast, re-equilibrated to approximately equal amounts of tetrameric and dimeric band 3 over the same time span (Fig. 2A, solid line). We did not extend this reassociation reaction in the DIDS-labeled membranes beyond 24 h, since studies described below argue that quantitative conversion to the tetramer would never have been achieved. These data thus suggest that DIDS stabilizes the dimeric state of band 3, resisting both an ankyrin-induced association to the tetramer and an unfacilitated slow dissociation to the monomer. The data further argue that the oligomeric form of unmodified band 3 preferred by ankyrin is the tetramer, since ankyrin shifts the subunit equilibrium to this association state.
Effect of DIDS on the Kinetics of 125 I-Ankyrin Binding to KI-IOV-In view of the impact of DIDS on the band 3 subunit dissociation equilibrium, it was of interest to further explore the influence of DIDS labeling on ankyrin binding. Ankyrin has been shown previously to associate with band 3 in two kinetic phases, a fast phase requiring approximately 90 min to reach completion and a slow phase extending for more than 8 h after mixing (21). To learn whether DIDS derivatization might influence ankyrin association, the kinetics of 125 I-ankyrin binding to DIDS-labeled and control KI-IOV was compared. As confirmed in Fig. 3 (top curve), 125 I-ankyrin associates with unmodified red cell membranes in both a rapid and slow binding phase. In contrast, interaction with DIDS-labeled KI-IOV is monophasic, and the number of available sites is significantly diminished (middle curve). To learn which kinetic phase of ankyrin binding is affected by DIDS labeling, the two binding curves were subtracted from each other point by point. The resulting difference curve (Fig. 3, open circles) indicates that the fast phase is eliminated by DIDS derivatization. Curiously, although Thevenin and Low (21) interpreted their earlier binding data to suggest completion of the fast phase only after 1-2 h of binding, and even though the top curve in Fig. 3 would seem to corroborate this interpretation, the difference analysis suggests the rapid phase is actually complete within 20 min. Repeat analysis of this difference curve using a second preparation of ankyrin and KI-IOV also revealed a 20 -30 min completion time for phase 1 of ankyrin association.

Effect of DIDS Modification of Band 3 on Its Association with 125 I-Ankyrin at True and Quasi-equilibrium-Previous studies
have demonstrated that band 3 offers both high and low affinity binding sites to ankyrin (21,36,37) and that these sites are generated during the slow and fast phases of ankyrin association, respectively (21). Since DIDS reaction with the membrane-spanning domain of band 3 appears to diminish the fast phase of binding (Fig. 3), it would be predicted that DIDS might selectively remove the population of low affinity sites. As shown in the Scatchard analysis of Fig. 4A, DIDS not only eliminate all low affinity ankyrin sites, as anticipated, but it also reduces the quantity of high affinity sites to approximately half their normal number (Table I).
Thevenin and Low (21) also observed that the low affinity sites seen during short incubation times with ankyrin were gradually converted to high affinity sites during the more extended slow phase of ankyrin binding. Since the data in Fig. 4A were obtained after more extended (8.5 h) incubation, we investigated whether low affinity sites might have been present earlier in the binding reaction, e.g. after only 1 h incubation. As displayed in Fig. 4B, under conditions where low affinity binding normally represents Ͼ95% of total binding (21), only the high affinity interaction is still detectable in DIDS-modified KI-IOV. We conclude that DIDS conjugation to band 3 eliminates low affinity sites at all time points. DIDS derivatization also initially reduces high affinity binding which nevertheless recovers partially during more extended incubation, i.e. similar to the slow accrual of high affinity sites seen with unmodified band 3 (21).

Effect of DIDS on Band 3 Retention in Detergent-extracted Membrane Skeletons-
The DIDS-induced loss of all low affinity and substantial high affinity ankyrin-binding sites predicts that band 3 retention in membrane skeletons prepared directly from DIDS-labeled whole cells should be reduced. Because band 3 migrates as a highly diffuse polypeptide in SDS-PAGE gels of normal membrane skeletal preparations, its quantitation in the detergent-extracted membrane skeletons is difficult. This quantitation, however, can be facilitated by proteolytically releasing the carbohydrate-containing fragment from band 3 prior to detergent extraction to yield the sharply defined 65,000-dalton skeleton-anchored fragment of band 3 (39). Because this treatment has no known effect on anion transport, DIDS binding, or ankyrin association (39), it can be exploited to improve densitometric analysis of band 3 retention in the membrane skeletons. Fig. 5 shows an analysis of the retention of the 65,000-dalton band 3 fragment in n-octyl glucoside-insoluble membrane skeletons prepared from unmodified and DIDS-

FIG. 2. Effect of ankyrin on the band 3 subunit equilibrium in preincubated control or DIDS-labeled KI-IOV.
KI-IOV (2 mg/ml) were prepared from DIDS-labeled or unlabeled erythrocytes as described under "Experimental Procedures." The KI-IOV were allowed to equilibrate at 37°C for 150 h until the DIDS-labeled band 3 had dissociated to half-dimer and half-monomer (panel A, dashed line) and ϳ90% of the unmodified band 3 (panel B, dashed line) was shifted to the monomer. The 46.5-kDa ankyrin fragment (0.23 mg/ml) was then added, allowed to equilibrate for 24 h at 37°C, and re-extracted with buffered 1 M KI to again remove the ankyrin fragment. The ankyrindepleted membranes were then washed with 5 mM sodium phosphate, pH 8.0, solubilized in detergent (1% C 12 E 8 ), and analyzed by gel filtration HPLC (solid lines), as described under "Experimental Procedures" and the legend to Fig. 1.   FIG. 3. Effect of DIDS on the kinetics of 125 I-ankyrin binding to KI-IOV at pH 6.35. Red cells were either left unlabeled or DIDSlabeled as described under "Experimental Procedures." KI-IOV were then immediately prepared for use in binding studies and kept on ice to prevent any further shift in the band 3 subunit equilibrium. 125 I-Ankyrin (final 7 g/ml) was then incubated at 24°C with the DIDSlabeled (OE) and unmodified (q) KI-IOV (final 35 g/ml) in 50 mM NaH 2 PO 4 , 50 mM H 3 BO 3 , 20 mM NaC1, 1 mM EDTA, 0.2 mM dithiothreitol, 0.25 mg/ml bovine serum albumin, 5% sucrose, pH 6.35. At the times indicated, bound 125 I-ankyrin was separated from free 125 Iankyrin by pelleting the KI-IOV at 43,000 ϫ g for 40 min through a 20% sucrose cushion, and the bound 125 I-ankyrin was quantified by ␥-counting. The difference between ankyrin bound to control and DIDS KI-IOV was then calculated by direct subtraction of the amount of ankyrin bound at each time point (E). This experiment was done in triplicate and also repeated on an independent set of membrane preparations. Since the error bars were generally less than twice the width of the data point symbols, they are not shown on the graph. treated whole cells. As anticipated, DIDS reduces retention of band 3 in the skeletal extract, exerting an increasingly greater impact as the time between DIDS labeling and detergent extraction proceeds. In contrast, the amount of band 3 retained in membrane skeletons from unmodified cells remains unchanged. By 1 h incubation, a reduction in band 3 content of ϳ55% was seen in four separate skeleton preparations from DIDS modified cells using n-octyl glucoside as the extracting detergent (Fig. 5). While not all detergents yielded quantitatively the same result in this study, all detergents evaluated (which included Zwittergent 3-14, octyl-␤-D-thioglucopyranoside, Triton X-100, and n-octyl glucoside) revealed a consistent loss of band 3 from the membrane skeletons of DIDS-modified cells with no loss of band 3 from control skeletons (Table II). We interpret the differences in the amount of band 3 associated with membrane skeletons to simply reflect the differing solubilizing capacities of the several detergents. For example, octyl glucoside-extracted cells retain roughly four times as much band 3 in their membrane skeletons as Triton X-100 extracted cells (50). Apparently, the band 3-ankyrin complexes most sensitive to DIDS modification constitute at least part of the population most readily extracted by Triton X-100. DISCUSSION We have provided evidence that DIDS binding to the membrane-spanning domain of band 3 significantly reduces 125 Iankyrin binding to the cytoplasmic domain. Not only were all low affinity (rapid phase) ankyrin-binding sites eliminated, but roughly half of all high affinity (slow phase) sites were also lost. Loss of sites was also observed in intact cells where the DIDSmediated decrease in ankyrin binding resulted in elimination of more than half the normal linkages to the membrane skeleton ( Fig. 5 and Table II). Presumably, the presence of high intracellular ankyrin concentrations prevented a more quantitative disappearance of binding sites in the whole erythrocyte.
While more than one mechanism can account for these observations, we hypothesize Scheme I to explain our current and previous data on ankyrin-band 3 interactions. As proposed by others (23)(24)(25)(26)(27)(28) and ourselves (21), ankyrin appears to associate predominantly if not exclusively with the tetramer of band 3 in situ. Consistent with this observation is the fact that addition of ankyrin to KI-IOVs drives the band 3 subunit equilibrium nearly quantitatively to the tetramer (Fig. 2B). According to the above mechanism, the fast (low affinity) phase of ankyrin binding arises from its association with pre-existing band 3 tetramers, i.e. the receptive population at the time of ankyrin addition. By the same argument, the slow phase of ankyrin binding must derive from generation of new band 3 tetramers from band 3 dimers as ankyrin addition and complex formation shifts the subunit equilibrium to the right (Fig. 2). DIDS elimination of the fast phase of ankyrin binding may, therefore, result from depletion of pre-existing tetramers in the KI-IOV membranes, as noted by native gel electrophoresis (44). Retention of a slow phase of binding to DIDS-modified membranes (Fig. 3) would then simply represent the slow, but only partial recruitment of band 3 dimers back to tetramers, as seen in Fig.  2A. The structural basis for conversion of low to high affinity sites is the only unexplained feature of this model, but FIG. 4. A, effect of DIDS on the equilibrium binding of 125 I-ankyrin to KI-IOV at pH 7.8. Increasing amounts of 125 I-ankyrin were incubated with KI-IOV (final 35 g/ml) with (OE) or without (q) DIDS modification (labeling was done on whole cells, and KI-IOV were prepared and stored on ice until used) for 510 min at 24°C in 50 mM NaH 2 PO 4 , 50 mM H 3 BO 3 , 30 mM NaC1, l mM EDTA, 0.2 mM dithiothreitol, 0.25 mg/ml bovine serum albumin, 5% sucrose, pH 7.8. Ankyrin binding was quantitated, as described in the legend to Fig. 3 and under "Experimental Procedures." The data are presented in a Scatchard plot. This experiment was conducted in triplicate and also repeated on a separate set of membrane preparations. Error bars were generally less than twice the width of the data point symbols, and similar results were obtained on an independent preparation of DIDS-labeled and control KI-IOV. B, Scatchard plot of the binding of 125 I-ankyrin to DIDS-labeled KI-IOV following a short (1 h) incubation at pH 7.8. Red cells were left unmodified or labeled with DIDS; then KI-IOVS were prepared before use. Various amounts of 125 I-ankyrin were incubated with DIDS modified KI-IOV at 24°C for 1 h at pH 7.8 (ࡗ), and the binding was quantitated as described in the legend to Fig. 3. The line marked with solid triangles (OE) is from the Scatchard plot of 125 I-ankyrin binding to DIDS membranes following the 510-min incubation shown in Fig. 4A. The use of a Scatchard plot to present data at quasi-equilibrium has been justified elsewhere (20). This experiment was done in triplicate and repeated once on a separate KI-IOV preparation. Error bars generally fell within the width of the data point symbols, and similar results were obtained on an independent preparation of DIDS-labeled and control KI-IOV.  Fig. 4A. b The curvilinear Scatchard plots were assumed to result from the presence of two classes of independent sites characterized by binding capacities, N 1 and N 2 , and dissociation constants, K D1 and K D2 . Derivation of these four parameters was conducted by iteration using the equation, where B and F represent the bound and free concentrations of ankyrin, respectively. c ND, not detected. d The monophasic Scatchard plots were analyzed for K D and N according to the equation, previous observations place significant constraints on any physical interpretation of this transition. For example, the two forms must be interconvertible and they must be generated in an ordered chronology, where low affinity sites are occupied before high affinity complexes can arise (21). Furthermore, the maximum number of high affinity complexes should approximate the potential number of band 3 tetramers in the membrane. While other explanations may be possible, the sequential interaction of two or more sites on band 3 with two or more sites on ankyrin represents an attractive mechanism to explain this interconversion. Thus, as noted in the Introduction, ankyrin is known to occupy two noncontiguous sites on band 3 (30 -34), and recently two distinct domains of ankyrin have been shown to contribute to its high affinity association with band 3 (35). If completion of a single site interaction were required prior to isomerization to a multisite association, then sequential conversion of low affinity to high affinity sites would be expected. Furthermore, if most or all ankyrin-band 3 complexes eventually isomerize to their high affinity state (21), then the absence of low affinity sites in DIDS-labeled membranes would be required, since the rate-determining (kinetically visible) step in ankyrin binding would be the slow association of band 3 dimers to tetramers. Regardless of the interpretation, DIDS binding clearly shifts a subunit dissociation equilibrium in band 3 slowly toward the dimer and simultaneously eliminates all of the fast phase and most of the slow phase of ankyrin binding. It is our contention that these two perturbations are mechanistically linked.
We have attempted to analyze our HPLC, polyacrylamide gel electrophoresis, and ankyrin binding data as quantitatively as possible. While such quantitation can be useful in identifying conditions that shift the band 3 subunit equilibrium, caution must also be exercised to avoid overinterpretation of these data. Thus, the ratio of band 3 tetramers, dimers, and monomers in any sample has been shown to depend on: (i) the amount of ankyrin remaining in the sample (e.g. Fig. 2), (ii) the duration of KI-IOV incubation at 37°C (Fig. 1), (iii) the temperature of the incubation (changes in subunit equilibrium are slow or nonexistent at 0°C, Ref. 25), (iv) the extraction detergent employed (C 12 E 8 and octyl glucoside are less disruptive of band 3 oligomers than Triton X-100 (50)), and (v) pH (25). Consequently, the oligomeric ratios observed in KI-IOV do not likely correspond to the ratios present in situ. Nevertheless, the qualitative shifts in this ratio induced by DIDS are undoubtedly real, since membrane skeletons extracted directly from whole cells display the impact of these shifts on band 3 retention.
Another laboratory has reported that DIDS derivatization causes little or no change in the size of the band 3 oligomer (25). In contrast, Salhany et al. (51), Tomida et al. (44), and ourselves observe that DIDS can substantially alter the subunit interactions of band 3. We suspect that this discrepancy does not arise from inaccuracies in the measurements, but rather from differences in the delay between DIDS labeling and band 3 analysis, or alternatively, in the temperature and method of preparation of peripheral protein-stripped membranes. In our FIG. 5. Effect of DIDS on the retention of band 3 in the detergent-insoluble membrane skeletons of human erythrocytes. Red cells were cleaved with chymotrypsin to allow sharper resolution of band 3 in the membrane skeletons. After proteolysis, the cells were either left unlabeled (C) or labeled with DIDS (D), incubated for the indicated times at 37°C, and then directly extracted with 2% n-octyl glucoside, as described under "Experimental Procedures." The resulting membrane skeletons were then analyzed by SDS-PAGE to allow direct visualization of the quantity of band 3 retained in the skeletons. The fragment of band 3 generated by the proteolysis lacks the heterogeneous carbohydrate tree, and therefore, migrates as a sharp band at M r ϳ 65,000. hands, the subunit dissociation of band 3 was relatively slow, as was the reassociation induced by excess ankyrin (Fig. 2B). If different groups were to sample the band 3 population at different times after DIDS labeling, then the observed discrepancy would be expected. Additionally, we have observed that stripping of peripheral proteins from erythrocyte membranes with ice-cold 2 mM EDTA, pH 12, partially denatures band 3 and prevents the subunit dissociation events that we observe upon incubation of KI-stripped inside-out vesicles at 37°C. The results of our studies provide a possible resolution to the controversy regarding the extent of interaction between the membrane-spanning and cytoplasmic domains of band 3. Evidence in favor of little or no contact between the two domains include: (i) the two domains unfold independently when examined in situ by differential scanning calorimetry (52,53), (ii) the thermostabilities of the two domains can be independently regulated by changes in pH and ligand binding (52,53), (iii) the two domains rotate in situ at different rates (54, 55), (iv) proteinases of different specificities can cleave band 3 between the two domains, yielding isolated membrane and cytoplasmic domains that display no affinity for each other (5), (v) the isolated membrane-spanning domain catalyzes anion transport similar to uncleaved band 3 (55)(56)(57)(58)(59), and (vi) the isolated cytoplasmic domain retains its binding sites for ankyrin, the glycolytic enzymes, band 4.1, band 4.2, hemoglobin, and hemichromes (12, 18, 20, 60 -64).
Arguments that support some type of communication between the two major band 3 domains include: (i) hemichrome binding to the cytoplasmic domain induces autologous antibody binding to an extracellular epitope in the membrane-spanning domain (9 -14); (ii) anion transport rates can be modulated by perturbations of the cytoplasmic domain, including disulfide cross-linking (65), Ca 2ϩ binding (66 -68), band 4.2 association (69), hemoglobin binding (70), and phosphorylation (71); (iii) the affinity of the cytoplasmic domain for ankyrin decreases at least 10-fold upon cleavage of the domain from the membrane (36); (iv) the cytoplasmic domain becomes thermodynamically more stable upon cleavage from its membrane-spanning counterpart (53); (v) mutations in the membrane-spanning domain can alter interactions with the membrane skeleton (72-78); (vi) mutations in the cytoplasmic domain can affect anion transport (79), and DIDS labeling of the membrane domain affect phosphorylation (80), hemoglobin binding (81), and Ca 2ϩ interactions (82) of the cytoplasmic domain. Since intact band 3 exists in situ in a reversible equilibrium between tetramers, dimers, and possibly monomers (44,51,(83)(84)(85), and since changes in this equilibrium can dramatically affect ankyrin binding, it seems highly reasonable that the above instances of communication between the two domains might alternatively be mediated by changes in subunit interaction without any direct involvement of interdomain contact. Indeed, autologous antibody recognition of an exoplasmic epitope on band 3 has already been shown to arise specifically from hemichrome-induced aggregation of the cytoplasmic domain of band 3 (9 -14), and mutations in the membrane-spanning domain of band 3 that result in altered cell morphologies are frequently associated with abnormal band 3 aggregation (76,86).
Finally, if the rates of change of the band 3 dimer 3 tetramer distribution in vivo are similar to those we have measured in Fig. 2, it is unlikely that new ankyrin-binding sites on band 3 can be rapidly generated in vivo. Instead, when a localized region of ankyrin-band 3 linkages are transiently broken during some mechanical deformation, it would seem more likely that new bridges might preferentially re-establish with preexisting band 3 tetramers (i.e. former ankyrin binding sites) rather than with dimers that must be slowly associated into receptive tetramers. Fortunately, band 3 oligomers dissociate only slowly, assuring that no significant loss of ankyrin-binding sites will occur during transient rearrangements of the membrane skeleton. However, because band 3 will eventually dissociate to its nonbinding states of association in the absence of ankyrin, it would be predicted that the number of available ankyrin sites in normal membrane preparations will never significantly exceed the number of ankyrin molecules in vivo. This expectation has, in fact, been confirmed by many researchers (21, 36 -38).