Reassociation of ankyrin with band 3 in erythrocyte membranes and in lipid vesicles.

The binding of human erythrocyte ankyrin (band 2.1) to the erythrocyte membrane has been characterized by reassociating purified ankyrin with ankyrin-depleted inside-out vesicles. Ankyrin reassociates at high affinity with a limited number of protease-sensitive sites located only on the cytoplasmic side of the erythrocyte membrane. Depleting the vesicles of band 4.2 does not affect their binding capacity. A 45,000-dalton polypeptide derived from the cytoplasmic portion of band 3 competitively inhibits the binding of ankyrin to inside-out vesicles. Although the bulk of band 3 molecules appear to have the potential for binding ankyrin, nly a fraction of the band 3 molecules in native membranes or in reconstituted liposomes actually provides accessible high affinity ankyrin binding sites.

The binding of human erythrocyte ankyrin (band 2.1) to the erythrocyte membrane has been characterized by reassociating purified ankyrin with ankyrin-depleted inside-out vesicles. Ankyrin reassociates at high affinity with a limited number of protease-sensitive sites located only on the cytoplasmic side of the erythrocyte membrane. Depleting the vesicles of band 4.2 does not affect their binding capacity. A 45,000-dalton polypeptide derived from the cytoplasmic portion of band 3 competitively inhibits the binding of ankyrin to inside-out vesicles. Although the bulk of band 3 molecules appear to have the potential for binding ankyrin, only a fraction of the band 3 molecules in native membranes or in reconstituted liposomes actually provides accessible high affinity ankyrin binding sites.
Studies of the interactions between intrinsic and extrinsic proteins of the human erythrocyte membrane have shown that ankyrin (band 2.1') (1) and perhaps a family of related polypeptides (bands 2.2 to 2.6) (2, 3) provide high affinity binding sites that link spectrin (4) to the cytoplasmic surface of the erythrocyte membrane (5-7). Using purified components, the binding between ankyrin and spectrin has been c o n f i e d by studies of proteins in solution and by direct, electron microscopic visualization of the spectrinankyrin complex (8,9). These studies also showed that attachment of ankyrin to ankyrin-depleted vesicles enhanced the vesicles' spectrin binding properties, but the association between ankyrin and the cytoplasmic face of the membrane remained to be elucidated. Both Sheetz (10) and Bennett and Stenbuck (7) have shown that ankyrin is associated with bands 3 and 4.2 when erythrocyte ghosts or inside-out vesicles are treated with the nonionic detergent, Triton X-100, at various salt concentrations. Here we show that ankyrin reassociates at high affinity with a limited number of protease-sensitive sites located only on the cytoplasmic side of the erythrocyte membrane. This reas-' Nomenclature for human erythrocyte polypeptides according to Steck (22). sociation does not require band 4.2 and occurs at physiological ionic strength and pH. Competition experiments provide evidence that band 3 is the polypeptide that anchors ankyrin to the membrane.

EXPERIMENTAL PROCEDURES
Materials-Fresh human blood was obtained from the Northeast Regional Red Cross Blood Program and was used within 2 days. Diisopropyl fluorophosphate (DFP)? dithiothreitol, N -{Trisrhydroxymethyl]methyl-2-amino) ethanesulfonic acid (Tes), Sepharose 4B, ammonium sulfate, type 111, and Triton X-100 were obtained from Sigma, ultrapure urea from Schwarz/Mann, and DE52 from Whatman. DE52 was cycled at least two times in acid and base and equilibrated with 20 mM KC1 buffer. Bovine pancreatic cu-chymotrypsin (recrystallized three times from bovine pancreas) and trypsin (crystallized two times) were obtained from Sigma and dissolved in buffer just before use. '251-Bolton-Hunter reagent was purchased from Amersham or New England Nuclear. Phosphatidylcholine from egg yolk was obtained from Sigma and was used as received. SM-2 Bio-Beads were purchased from Bio-Rad, washed in methanol, and stored in the same buffer in which they were to be used.
Ghosts (11) were prepared from fresh human blood after washing the erythrocytes three times in phosphate-buffered saline (150 mM NaCl, 5 mM NaP04, 1 mM Na*EDTA, and 3 mM NaNa, pH 7.5). The subsequent lysing and washing solution was 5 mM NaP04, pH 7.6, containing 1 mM Na,EDTA. Ghosts were incubated overnight in phosphate-buffered saline with 0.4 mM (0.007556, v/v) isopropyl fluorophosphate at 0°C to remove band 6 and to inactivate endogenous proteases. After washing in excess phosphate-buffered saline, membranes were stored at 0°C in a 20 mM KC1 buffer (20 mM KCI, 1 mM sodium phosphate, 1 mM NazEDTA, and 3 mM sodium azide, pH 7.6).
Resealed ghosts were prepared by warming the band 6-depleted ghost membranes for 15 min at 37°C. Permeability to macromolecules was ascertained by mixing ghosts with 370 (w/v) dextran (84,000 daltons, Sigma) dissolved in 20 mM KC1 buffer. After this treatment, ghosts appeared bright by phase contrast microscopy, indicating their impermeability to the added dextran (12).
Preparation of Membrane Vesicles-Right-side-out vesicles were prepared by forcing erythrocyte ghosts in 20 mM KC1 buffer through a 26-gauge hypodermic needle. The resultant vesicles were washed with 20 mM KC1 buffer and resuspended in this same buffer containing 0.4 mM DFP. To prepare inside-out vesicles (131, band &depleted ghosts were washed twice at 4°C with 15 volumes of a low salt buffer containing 1 mM Tes and 0.1 mM NaZEDTA, pH 8, and then incubated at 37OC for 30 min in 15 volumes of this low salt buffer containing 0.4 n w DFP. The resultant vesicles were centrifuged at 4°C (30 min at 40,000 X g) and the pellets were washed once more in approximately 20 volumes of low salt buffer (without DFP), then in 20 mM KC1 buffer, and finally resuspended in 20 mM KC1 buffer to the original volume of the membrane pellet prior to removal of band 6 (final protein concentration approximately 2 mg/ml). The inside-out orientation of membrane faces was ascertained by freeze-fracture electron microscopy (13).

11965
previously described (8) and then resuspended in 20 mM KC1 buffer to their original volume.
KCI-Urea-stripped Vesicles-To remove bands 2.1, 4.1, and 4.2, inside-out vesicles were incubated for 30 min at 37°C in 20 volumes of 1 M KC1 buffer containing 2.5 M urea and 0.1 M glycine, pH 7.6, with 0.4 mM DFP. The suspension was centrifuged at 40,000 x g for 30 min, the supernatant discarded, and the pellet resuspended in a small volume of 20 mM KC1 buffer containing 1 mM dithiothreitol. After dilution to 20 volumes with 20 mM KC1 buffer, the vesicles were again centrifuged, the supernatant discarded, and the pellets resuspended to their original volume in the same buffer. The membrane suspension was then dialyzed for 3 days at 4°C against 20 mM KC1 buffer containing 0.2 mM dithiothreitol.
Isolation of Ankyrin-Inside-out vesicles were f i t incubated in a buffer (pH 6.75) containing 1 M KCI, 0.4 M urea, 25 mM NazEDTA, and 0.05 M glycine to selectively remove band 4.1. After 30 min on ice, vesicles were pelleted and then resuspended in cold 1 M KC1 buffer (1 M KCI, 25 mM NaP04, and 25 mM NazEDTA, pH 7.6). The pH of the final suspension was adjusted to 7.6 with 1 M sodium borate (pH 8.5) if necessary and made at least 0.4 KIM in DFP. After incubation at 37°C for 30 min to release ankyrin (a), the membranes were pelleted (225,000 X g for 20 min at 4°C) and the supernatant containing ankyrin was dialyzed overnight against 70 to 100 volumes of 20 mM KC1 buffer containing 0.2 mM dithiothreitol. Ankyrin was purified on a 1-ml DEAE-cellulose (DE52) column using 115 mM KC1 to elute contaminants and 195 RIM KC1 to elute ankyrin. The peak fractions containing ankyrin were dialyzed extensively against 20 mM KC1 buffer containing 0.2 mM dithiothreitol and then made 0.4 mM in DFP. Small amounts (less than 20% of the total protein content) of the related polypeptides, bands 2.2 to 2.6, co-purified with ankyrin.
Triton Extraction and Reconstitution of Membrane Polypeptides-Triton X-100 extracts were prepared from ghosts and from KCI-urea-stripped vesicles by incubating the membranes in phosphate-buffered saline (3.2 mg/ml and 1.5 mg/ml of protein for ghosts and stripped inside-out vesicles, respectively) with an equal volume of 1% (w/v) Triton X-100 in phosphate-buffered saline. After 20 min on ice, the suspension was centrifuged at 225,000 X g for 60 min. The clear supernatant solution (approximately 0.5 mg/ml of protein) was added to a tube lined with a film of egg lecithin prepared by drying a chloroform solution of the liquid in a NZ atmosphere. The final phospholipid/protein ratio was approximately 101 (w/w). After the lipids dissolved, the clear mixture was combined with SM-2 Bio-Beads and was agitated for 4 h at 4°C. During this time, the Triton X-100 adsorbed to the beads and the lipids with associated proteins came out of solution as vesicles. Protein aggregates and multilamellar vesicles were pelleted by centrifugation at 10,000 X g for 20 min. The supernatant containing unilamellar and some multilamellar vesicles was centrifuged (225,000 X g for 90 rnin) and the pelleted vesicles were fused by alternatively freezing and thawing the pellet three times.
Protein orientation and accessibility in reconstituted membranes were estimated by proteolysis with trypsin, which only cleaves the cytoplasmic portion of band 3 (14). Membranes were incubated with 10 p g / d of trypsin in 20 mM KC1 buffer (37OC for 30 min) and then with 5 mM DFP for an additional 30 min at the same temperature. After an overnight incubation at O' C, the samples were dissolved with preheated (65°C) gel sample buffer for electrophoresis. Densitometric scans of stained gels indicated that approximately 50% (+lo%) of the protein was accessible to the enzyme.
Limited Proteolysis of Inside-out Vesicles and Triton Extracts-To ascertain whether ankyrin interactions with inside-out vesicles were dependent on membrane-associated polypeptides, vesicles (1.8 mg/ml of protein) that had been stripped with 1 M KC1 were incubated for 1 h at 0°C in the presence or absence of a-chymotrypsin (1 pg/ ml). The suspension was then diluted with an equal volume of DFP (IO mM in 5 mM NaP04 and 1 mM NazEDTA, adjusted to final pH 7.2) and warmed to 37°C for 30 min to increase the reactivity of the protease inhibitor. Vesicles were diluted with 10 volumes of phosphate-buffered saline (to minimize adsorption of protease to membranes), pelleted (30 min at 40,000 X g), and resuspended in phosphate-buffered saline containing 0.4 mM DFP. After overnight incubation on ice, vesicles were pelleted again, washed twice in 20 volumes of 20 mM KC1 buffer, and then resuspended to their original volume in the same buffer containing 0.4 mM DFP.
To generate water-soluble fragments from band 3 that had been solubilized in Triton, 5 pg of a-chymotrypsin was added to 1 ml of a clarified 1% Triton X-100 extract (see above). After incubation for 30 min at O'c, an equal volume of 10 mM DFP in 5 mM NaP04 and 1 mM Na2EDTA, final pH 7.2, was added. After an overnight incubation at 0°C. the extract was loaded onto a 1-ml DEAE-column at 4°C. Residual band 3 was eluted with 150 m~ KC1 buffer containing 1% Triton X-100 and then the column was washed to remove Triton using 200 mM KC1 buffer. When the 280-nm absorbance was stable and equivalent to that of buffer without detergent, water-soluble fragments of band 3, as well as small amounts of residual intact band 3, were eluted with 270 mM KC1 buffer containing 0.2 mM dithiothreitol. Peak fractions were pooled, dialyzed against 20 mM KC1 buffer containing 0.2 m~ dithiothreitol, and clarified by centrifugation at 225,000 x g for 60 rnin).
Complete incubation mixtures were swirled gently at 4°C for 30 min. One hundred fifty microliters from each mixture was then transferred to a 4 0 0 4 microfuge tube that contained 10% (w/v) sucrose dissolved in incubation buffer. An air space separated the 150-pl sample at the top of the tube from the 2 5 0 4 sucrose solution at the bottom of the tube. After a total incubation time of 60 min, centrifugation was initiated (18,000 rpm, Sorvall SS-34 rotor) to separate the free 2.1 from the bound 2.1 which sedimented with the vesicles. Calculation of bound and free ligand included a correction for the fraction of the incubation mixture which was not sampled.

Preparation and Selective Extraction of Inside-out Vesicles-Treatment of ghosts with Tes-EDTA-DFP at 37OC
released 80 to 9 0 % of the spectrin and actin (Fig. 1, a and b).  within sealed vesicles or unfragmented ghosts because it was not affected by high salt extraction (Fig. 1, c and d) or by proteolysis of the inside-out vesicles with chymotrypsin (see Fig. 66) (5). Analysis of freeze-fractured replicas (Fig. 2) indicated that more than 854 of the vesicles prepared with Tes-EDTA-DFP were inside-out.
Inside-out vesicles incubated with 1 M KC1 at 37°C (henceforth referred to as KC1-stripped vesicles) released 65 to 80% of band 4.1 and ankyrin and a variable amount (typically 20 to 30%) of band 4.2; inside-out vesicles incubated in 1 M KC1 and 2.5 M urea released 70 to 80% of bands 4.1 and ankyrin, as well as 70 to 80% of band 4.2 (Fig. 1, c and d). Band 4.1 was selectively extracted using 1 M KC1 and 0.4 M urea at 0°C so that the ankyrin subsequently extracted at 37°C with KC1 could be purified by a single passage over DEAE (Fig. 1, e to  h ) .
Ankyrin Binding to Inside-out Vesicles-We expected that ankyrin would reassociate with inside-out vesicles which had been stripped of this polypeptide without denaturing its binding sites. Preliminary experiments indicated that KCI-stripped vesicles bound ankyrin more effectively than NaOH-stripped membranes (prepared as described in Ref. 15). KCI-stripped vesicles were, therefore, used to characterize the binding of purified ankyrin.
T o compare the binding capacity of variably-stripped vesicles, the membrane concentrations were always adjusted to compare the same number of vesicles (as estimated from band 3 content) and scaled per milligram equivalent. The milligram equivalent was defined as the quantity of inside-out vesicles that contained 0.33 mg of band 3, as estimated from densitometric scans of SDS-polyacrylamide gels. One-milligram equivalent of inside-out vesicles contained 1 mg of total protein, while 1-mg equivalent of KC1-stripped vesicles contained approximately 0.8 mg of total protein.
T o determine what percentage of our purified ankyrin was capable of reassociating with the KC1-stripped vesicles, the labeled ligand was incubated at a single concentration with increasing amounts of stripped vesicles (Fig. 3a). At least 704 of the labeled ankyrin bound to the KC1-stripped vesicles; less than 20% bound to the heat-denatured vesicles although no aggregation of membranes or change in the electrophoretic profiie of the membrane polypeptides was apparent after heat denaturation (not shown). The relationship between the concentration of native vesicles and the percentage of ankyrin which was bound was approximately linear below 40-pg equivalents of membrane protein (200-pg equivalents/ml in the reassociation assay). For this reason, our reassociation assays were consistently carried out with between 10 and 30-pg equivalents/sample. KCI-stripped vesicles bound '"I-labeled ank-yrin with high affinity (Kd = approximately 5 to 8 X lo-' M, Fig. 3, h and c ) . Analysis according to Scatchard (16) showed that the high affinity binding to stripped vesicles saturated at approximately 95 pg of ank-win/mg equivalent of membrane protein. This amounted to approximately 1 mol of ank.win/8 mol of band 3. In some experiments, of which Fig. 3 is an example, curvature of the Scatchard plots at high ankyrin concentrations hinted at the presence of additional lower affinity ankyrin binding sites. Heat-denatured vesicles bound little '"Iankyrin and resealed ghosts bound even less. The binding to heat-denatured vesicles and to resealed ghosts exhibited low affinity and did not appear to be saturable. Inside-out vesicles that had not been stripped of their native ankyrin bound less labeled ankyrin with lower apparent affinity than did the KCIstripped vesicles. This result would be expected if added labeled and native unlabeled ankyrin competed for the same binding sites.
T o confirm that the labeling process had not modified the reassociation characteristics of ankyrin, competition experiments were performed with unlabeled ankyrin (Fig. 4). Unlabeled ankyrin competed with labeled ankyrin ( K , = 11 pg/ml or 5 X lo-" M ) and up to 90% of the "'I-ankyrin binding was eliminated by a 65-fold excess of unlabeled ankyrin. This competitive activity was totally lost if the unlabeled ankyrin was heated a t 65°C for as little as 2 min after the addition of excess crystalline dithiothreitol to preclude oxidative crosslinking (data not shown). Since the apparent K, derived from these experiments is similar to the K,, calculated from the binding of labeled ligand (Fig. 3c), it is apparent that iodination of the molecule has not significantly changed the reassociation characteristics of ankyrin.
Although most reassociation assays were carried out at 0-4"C, we compared the rate of binding and the saturation levels achieved a t 0-4OC versus 20°C and 37°C. The protease inhibitor DFP was included in these assays and quantification of the labeled polypeptides on sodium dodecyl sulfate-polyacrylamide gels following the reassociation assays indicated that proteolytic degradation of the sensitive ank-yrin polypeptide had not occurred. As expected, the 20°C and 37°C samples reached equilibrium faster than the cold samples, but exhibited a slight decrease (about 10%) in the maximum number of binding sites. The binding affinity was not affected by temperature. These experiments also indicated that at 0-4"C. equilibrium binding was reached within 45 to 60 min.
The effect of ionic strength on the reassociation of ankyrin was also tested (Fig. 5). In contrast to the binding of spectrin to spectrin-depleted inside-out vesicles, where binding affinity is increased at least 3-fold in phosphate-buffered saline compared to 20 mM buffer,:' the affinity of ankyrin for stripped membranes was the same at these two salt concentrations. As expected, binding was substantially reduced in 0. To assess the nature of the high affinity binding site for ankyrin, KC1-stripped vesicles were treated briefly with achymotrypsin (Fig. 6, a and b). This enzyme readily cleaves the intrinsic erythrocyte membrane polypeptide, band 3 (14), which comprises the major polypeptide component of KClstripped membranes. Band 4.2, which is also prominent in these membranes, is not noticeably affected by brief treatment with a-chymotrypsin.
Treating the vesicles with a-chymotrypsin substantially reduced their ability to bind ankyrin (Fig. 7 ) . Because of the extreme sensitivity of both membrane-bound and soluble ankyrin to proteolysis, we performed two experiments to confirm that the labeled ankyrin was not degraded by residual protease activity associated with the vesicles. In the first, proteolyzed vesicles were incubated with labeled ankyrin and then unproteolyzed vesicles were added for a second incubation period. Samples which contained proteolyzed vesicles for the first incubation and additonal unproteolyzed vesicles for the second incubation bound an amount of ankyrin nearly equivalent Binding to heat-denatured vesicles has been subtracted to correct for nonspecific binding. to the sum of ankyrin bound by the two kinds of vesicles alone (Fig. 7). Thus, residual chymotrypsin activity could not account for the reduction in ankyrin binding to chymotrypsintreated vesicles. A second experiment to control for proteolysis used sodium dodecyl sulfate-polyacrylamide gel electrophoresis to determine the amount of radioactivity in the relevant stained bands (not shown). No differences were found between samples which contained protease-treated vesicles and control samples which contained only unproteolyzed vesicles.
While it appeared that the presence of intact band 3 was correlated with the competence of stripped membranes to bind ankyrin, we investigated the possible role of band 4.2 in this reassociation phenomenon. Inside-out vesicles were selectively extracted to remove either little or most of the band KCI-stripped vesicles were incubated with buffer ( a ) or with 1 pg/ml of a-chymotrypsin ( h ) for 60 min at 0°C. After inactivating the protease (see "Experimental Procedures"), the vesicles were used for the binding experiments shown in Fig. 7 . Polypeptides released into the supernatant by a-chymotrypsin were separated on DEAE-cellulose (c) and the approximately 45,000-dalton fragment and approximately 38,000-dalton subfragment attributable to the cytoplasmic domain of band 3 (2. 4) were used for the competition experiment shown in Fig. 9. Alternatively, Triton extracts from ghosts ( d ) were incubated with 5 pg/rnl of a-chymotrypsin for 30 min at 0°C (e). After inactivating the protease, the approximately 45,000-dalton fragment (t, was purified by chromatography on IIEAE-cellulose. A fragment of identical molecular weight and purity was also obtained by achymotrypsin digestion of the Triton X-100 extract derived from KCIurea-stripped vesicles (g). Fig. 1, c and d ) .

4.2, along with ankyrin and band 4.1 (see
Binding of ankyrin was unaffected by these treatments (Fig.  8). We concluded that band 4.2 is not required for the high affinity binding of ankyrin to KC1-stripped vesicles.
Competition experiments were used to confirm that band 3 was the membrane binding site for ankyrin. The 45,000-dal-ton4 water-soluble polypeptide that is known to be a fragment derived from the cytoplasmic portion of band 3 (2, 14) competed effectively with KC1-stripped vesicles for the binding of ankyrin (Fig. 9), whereas neither band 6 nor heat-treated 45,000-dalton fragment (2 min a t 100°C) showed any effect. However, the apparent inhibition constant ( K , = approximately 5 X 10" M, Fig. 9) suggests that the band 3 fragment binds to ankyrin approximately 10 times less avidly than does ankyrin to the intact membrane site.
Because band 3 is present in a 4-to 8-fold molar excess over ankyrin monomers in the intact erythrocyte membrane (17), and because the reassociation of ankyrin with the KClstripped vesicles was limited to a similar stoichiometry, it became apparent that some mechanism, perhaps denaturation, must be limiting the interaction of ankyrin with band 3 in the membrane. The most likely step during which a large subset of band 3 molecules might be denatured, leaving a minor subset completely protected, occurs during the elution of spectrin and actin with low salt at 37OC. This step occurs before ankyrin is stripped from the vesicles. We reasoned that We refer to the 45,000-and 38.000-dalton polypeptic fragments of band 3 without implying that these are exact molecular weights.
if this low ionic strength treatment was able to denature unoccupied ank-yrin binding sites, those binding sites that were occupied by ank-yrin might remain undenatured. However, low ionic strength treatment of stripped inside-out vesicles that lacked ankyrin had absolutely no effect on the subsequent reassociation of ankyrin (data not shown).
T o determine whether discrete subpopulations of band 3 accounted for the limited binding of ankyrin, we attempted to separate the putative binding-incompetent population of band 3 molecules from the binding-competent molecules. Ghosts were extracted with detergent (0.5% Triton X-100) in 150 mM KCl, an ionic strength which is known to maximize both spectrin-ankyrin and ankyrin-membrane interactions. The showing protease sensitivity of ankyrin binding sites. "."I-ankyrin was incubated with 25yg equivalents of n-chymotrypsin-treated vesicles (A), or with control vesicles not exposed to enzyme (0). or sequentially with a-chymotrypsin-treated vesicles followed by control vesicles (0).

Reassociation with Band 3
limited population of band 3 molecules that was solubilized from ghosts under this condition was assumed to contain band 3 molecules which were unassociated with the membrane cytoskeleton (10). These band 3 molecules were likely to be the binding-incompetent molecules, if such molecules existed. The competence of this limited population of band 3 molecules  to bind ankyrin was tested by using the 45,000-dalton watersoluble fragment derived from these Triton-solubilized molecules (Fig. 6, d to f ) in binding competition experiments. The fragment derived from this limited population of band 3 inhibited binding of ankyrin to KC1-stripped vesicles as effectively (Kc z approximately 20 pg/ml or 5 X M, Fig. 10) as the fragment derived under identical conditions from the total population of band 3 solubilized by Triton X-100 from KClurea-stripped vesicles ( Fig. 6g and Fig. 10) or the fragment derived from the total population of band 3 still in the membrane (Fig. 9).
To ascertain directly whether the limited population of band 3, solubilized from ghosts with detergent at physiological ionic strength, was capable of the same high affinity interaction with ankyrin that we had observed with KC1-stripped vesicles, band 3 in Triton extracts (Fig. 6) was reconstituted into bilayer vesicles (Fig. 2 b ) . The amount of accessible band 3 in the reconstituted membranes was estimated from sodium dodecyl sulfate-polyacrylamide gels of reconstituted vesicles that had been exhaustively digested with trypsin. We ascertained that 50% (*lo%) of the band 3 reconstituted into the vesicles was cleaved by trypsin. Because band 3 is cleaved by trypsin only at its cytoplasmic face, this indicated that only 50% of the band 3 in the vesicles was appropriately oriented to serve as a binding site for ankyrin. Binding isotherms expressed in terms of accessible band 3 reconstituted into the vesicles (Fig. 11) show that at low ankyrin concentrations, the limited population of band 3 molecules extracted from ghosts by Triton in 150 mM KC1 binds ankyrin with the same or higher affinity as the total population of band 3 derived from KC1-stripped vesicles. However, at higher ligand concentrations, many more binding sites are accessible to ankyrin, albeit at a lower apparent binding affinity, in the reconstituted membranes. These extra binding sites as well as the high affinity sites were both destroyed by heat denaturation of the membranes.

DISCUSSION
Our results show that ankyrin binds with high affinity to a limited number of heat-sensitive and protease-sensitive sites which are located on the cytoplasmic surface of the erythrocyte membrane. Competition experiments using the 45,000dalton cytoplasmic fragment of band 3 indicate that these sites are part of the cytoplasmic portion of the transmembrane protein, band 3 (18). Thus, our studies using inside-out vesicles, purified ankyrin, and the water-soluble fragment of band 3 accord with and clarify studies of Bennett and Stenbuck (7) which show that ankyrin is bound to a complex of band 3 and band 4.2 in detergent extracts of erythrocyte membranes.
The band 6 and band 4.2 polypeptides also have binding sites on band 3 (19) and their attachment could also modulate the interaction between band 3 and ankyrin. However, band 6 does not bind to inside-out vesicles at physiological ionic strength and pH (20) and does not affect the binding of ankyrin to membranes under such conditions (Fig. 9). Pretreatment of our vesicles with urea to extract band 4.2 had no effect on the binding of ankyrin (Fig. 8). Thus, under the conditions employed in this study, band 4.2 also appears to play no role in binding ankyrin to the membrane or to band 3.
Extrapolation of our binding data indicates that at saturating levels of ankyrin, the stoichiometry of ankyrin that binds with high affinity to band 3 in the vesicles is approximately 1 mol of ankyrin/8 mol of band 3. Although this is nearly the same stoichiometry found in freshly prepared erythrocyte ghosts (17), the basis of this stoichiometry is puzzling if, as observed in Triton solutions, 1 band 3 molecule can bind 1 ankyrin molecule (7). Since the Polypeptide moieties of the band 3 molecules are thought to be identical (7,14,21), one would expect all of them to provide potential binding sites for ankyrin. Limited synthesis of ankyrin could account for the stoichiometry in the native cell, but other reasons must be sought to explain why less than 15% of the band 3 molecules in our inside-out vesicles appear to provide binding sites for ankyrin.
In spite of the biochemical evidence suggesting that most band 3 molecules share a common polypeptide sequence, we considered the possibility that not all band 3 molecules contain an ankyrin binding site. We extracted ghosts with detergent under conditions where only those band 3 molecules that were not linked to ankyrin would have been solubilized. After being reconstituted into lipid vesicles, these "noncytoskeletal" band 3 molecules bound substantial amounts of ankyrin with an affinity similar to that exhibited by the total band 3 population in stripped inside-out vesicles, although negative cooperativity or the presence of additional lower affinity ankyrin binding sites was suggested by the curved Scatchard plot (Fig. 11 b). Furthermore, a 45,000-dalton, water-soluble fragment of the "noncytoskeletal" band 3 inhibited the binding of labeled ankyrin to inside-out vesicles as effectively as the fragment derived directly from intact, stripped membranes or from a Triton extract of the latter. Thus, those band 3 molecules which are not bound to ankyrin in the membrane do have the capacity to bind ankyrin with high affinity.
One explanation why less than 15% of the band 3 molecules in our inside-out vesicles appear to provide binding sites,for ankyrin is that this estimate assumes all of the band 3 in our KC1-stripped vesicles is accessible. When one takes into account the presence of right-side-out vesicles (10 to 15%) and the presence of residual ankyrin (20 to 30% of the native ankyrin content), accessible band 3 may be only 50% of the total band 3 content. But, even when this correction is made, not all of the accessible band 3 molecules in our inside-out vesicles provide binding sites for ankyrin. At most, only 1 out of 4 band 3 molecules in our inside-out vesicles or reconstituted vesicles binds labeled ankyrin with high affinity.
One plausible explanation for the limited number of ankyrin binding sites in stripped vesicles is that steric factors further limit accessibility to the potential binding sites. There is general agreement that band 3 forms a stable dimer in the erythrocyte membrane (22) in Triton X-100 extracts (20) and after reconstitution into lipid vesicles (23). Cross-linking studies show that higher order oligomers may also exist (24, 25), and recent reanalyses of fluorescence depolarization data (26) and of the number of intramembrane particles in the erythrocyte membrane (27) strongly suggest that band 3 may exist as a tetramer. If band 3 exists in membranes as a tetramer, the self-association of band 3 dimers to form tetramers could effectively shield all but one binding site for ankyrin or create a new, shared binding complex. The latter hypothesis is consistent with the lack of obvious negative cooperativity when ankyrin reassociates with stripped membranes and the relatively poor inhibition of ankyrin binding by the monomeric cytoplasmic fragment of band 3. The apparent negative cooperativity seen in ankyrin reassociation with reconstituted band 3 is consistent with incomplete reformation of band 3 oligomers after detergent treatment.
In spite of the fact that ankyrin binds with high afflnity and specificity to band 3 sites on the cytoplasmic surface of ankyrin-depleted inside-out vesicles, evidence is required to show that this reassociation fully reconstitutes the native association of ankyrin with the membrane. Reconstitution of function is one of the most useful criteria for establishing the biological relevance of ligand binding. The only known function of ankyrin is related to its spectrin binding capacity (8,28). Thus far, our investigations have shown that reassociating ankyrin with KC1-stripped vesicles which do not bind spectrin does in fact reconstitute some of the binding capacity (8)." But further studies in which ["'Plspectrin was reassociated with membrane-bound '"I-ankyrin suggest that as much as 50% of the membrane-associated ankyrin may be unavailable for spectrin binding.5 Thus, the factors that limit the association of ankyrin to its attachment sites on the membrane and that restrict the spectrin binding capacity of membrane-associated ankyrin remain subjects for future research.