Selective Association of Spectrin with the Cytoplasmic Surface of Human Erythrocyte Plasma Membranes QUANTITATIVE DETERMINATION WITH

A specific association between spectrin and the inner surface of the human erythrocyte membrane has been examined by measuring the binding of purified [“‘Plspectrin to inside out, spectrin-depleted vesicles and to right side out ghost vesicles. Spectrin was labeled by incubating erythrocytes with ‘lzPl, and eluted from the ghost membranes by extraction in 0.3 mM NaPO,, pH 7.6. [“‘PlSpectrin was separated from actin and other proteins and isolated in a nonaggregated state as a S),,,,. = 7 S (in 0.3 mM NaPO,) orS&I,,p = 8 S (in 20 mM KCl, 0.3 mM NaPO,) protein after sedimentation on linear sucrose gradients. Binding of [:l’Plspectrin to inverted vesicles devoid of spectrin and actin was at least IO-fold greater than to right side out membranes, and exhibited different properties. Association with inside out vesicles was slow, was decreased to the value for right side out vesicles at high pH, or after heating spectrin above 50” prior to assay, and was saturable with increasing levels of spectrin. Binding to everted vesicles was rapid, unaffected by pH or by heating spectrin,

These experiments suggest that attachment of spectrin to the cytoplasmic surface of the membrane results from a selective protein-protein interaction which is independent of erythrocyte actin. A direct role of the major sialoglycoprotein or Band 3 as a membrane binding site appears unlikely.
Biological membranes are currently viewed, to a first approximation, as dynamic structures whose components are capable of rapid lateral diffusion within the two dimensional plane of a phospholipid bilayer (l-3). Membrane proteins, however, may exhibit a restricted distribution such as the static arrays observed in neuromuscular (4) and gap (5) junctions. Moreover, the lateral diffusion of membrane molecules can be nonrandom, as in formation of "caps" of labeled surface components of lymphocytes over one pole of the cell (6). These examples and others (reviewed in Ref. 7 and 8) indicate possibilities for long range interactions and regulation in the membrane which are not predicted on the basis of a highly fluid and disordered environment.
The human erythrocyte may provide a model system for studying the mechanisms that can influence long range interactions in cell membranes.
Evidence is accumulating that spectrin (Bands 1 and 2)l (g-111, and possibly erythrocyte actin (121, which are peripheral membrane proteins located on the cytoplasmic surface of erythrocyte ghosts (13), may be involved in control of the lateral distribution of the major membrane glycoproteins.
Antispectrin immunoglobulin trapped within erythrocyte ghosts induces clustering of colloidal iron hydroxide binding sites (14,15), which are thought to represent the sialoglycoproteins.

2754
[32P]Spectrin Bind&g to Inverted Human Erythrocyte Membranes ment of particles to form clusters may be induced by manipulation of pH or ionic strength (16). A direct association between spectrin (and possibly actin) and the intramembrane particles has been proposed (17) because it was found that conditions which promote precipitation of extracted spectrin and actin (low pH, CaZ+, Mg*+, basic proteins) also induce aggregation of particles in membranes partially depleted of spectrin. This hypothesis is supported by the recent demonstration that aggregation of particles reconstituted in liposomes proceeds only at low pH and after binding of spectrin and erythrocyte actin (18,19).
An alternative explanation for restraint of intramembrane particles is that the cytoplasmic portions of these structures are simply trapped within a meshwork of spectrin and actin without the formation of specific protein-protein associations (16,30,31). Little is known of the molecular nature of the spectrin-membrane association, the role, if any, erythrocyte actin plays in this interaction, or the details of the organization of spectrin and actin on the inner membrane surface. Thus, on the basis of available data, neither entrapment nor binding of spectrin to the intramembrane particles can be excluded. Some cross-linking studies (32) indicate a close proximity between spectrin and multiple membrane proteins, although the affinity and specificity of these interactions cannot be determined.
The present experiments have been directed towards elucidating in biochemical terms the nature of the attachment of spectrin to the erythrocyte membrane. The approach adopted here has been to measure the interaction of purified spectrin with inside out membrane vesicles depleted completely of spectrin and actin. A specific association between spectrin and the cytoplasmic surface of the membrane is characterized, and features are described strongly supporting a selective, proteinprotein interaction.
Analogous studies have been reported previously for binding of aldolase (33) and glyceraldehyde-3phosphate dehydrogenase (34)(35)(36)  at least 90% of these are inverted by the morphological criteria first noted by Steck et al. (44). Such membranes are stored at 0" and used for binding studies for up to 4 days after preparation. Preparation of Right Side Out Vesicles-Ghosts from 2 ml of packed erythrocytes are suspended in 5 ml of 20 mosM NaPO,, pH 7.6, and passed forcefully through a 26-gauge l/z inch hypodermic needle (44). This procedure is repeated four to five times or until membranes are reduced to small (<lp,~) vesicles as visualized by phase microscopy.
Following vesiculation, the membranes are pelleted and resuspended in 20 mM KCl, 0.7 mM NaPO,, pH 7.6,0.5 mM NaN, at a concentration of 1 to 2 mg/ml. These vesicles are at least 70% right side out when viewed by freeze-fracture electron microscopy (Fig. 1B), and are identical to erythrocyte ghosts when analyzed by SDS-polyacrylamide electrophoresis (not shown).

Procedures
Membrane protein was estimated by the method of Lowry et al. (46) with bovine serum albumin as a standard, whileespectrin concentrations were determined by absorbance at 280 nm assuming an Ai% of 10.1 (ll), or by intrinsic fluorescence (excite 295 nm, emit at 338 nm, 10 nm band widths).
Occasionally it was necessary to compare the protein content of spectrin and membranes directly, in which case the Lowry method was used. SDS-polyacrylamide electrophoresis was performed essentially as described by Fairbanks et al. (40). Freeze-fracture electron microscopy was conducted as previously described (17)(18)(19).

RESULTS
The elution of spectrin and erythrocyte actin from erythrocyte ghosts at low ionic strength in the absence of divalent metal ions has been noted previously (g-12,16,40,47-49).
The quantitative dependence of spectrin dissociation on salt concentration at pH 7.6 ( Fig. 21, indicates an abrupt transition in stability of the membrane. spectrin complex below about 10 mM KCl; erythrocyte actin is eluted in a similar manner (not shown). These measurements were performed after a 25min incubation at 37", at which time elution is essentially complete (30 min at 20,000 x g). The resulting membrane pellets were analyzed by electrophoresis in the presence of 0.2% SDS on 5% polyacrylamide gels. Following staining with Coomassie blue, the gels were scanned at 550 nm, and the area under peaks corresponding to spectrin (Bands 1 and 2) was estimated relative to that of Band 3. The values are expressed as the percentage (area Bands 1 + S/area Band 3) of a control sample maintained at 0" in 20 mM KCl, and were determined in duplicate with an average half-range of ?8%. Controls, in which the extracted spectrin was centrifuged without membranes, indicated that less than 4% of spectrin was pelleted through the 20% sucrose when centrifuged in the absence of membrane.
(40).3 A similar dependence on ionic strength is evident at lower temperatures, although the time required for equilibrium is increased. Qualitatively similar results have been obtained with other salts, including NaCl, NaPO,, and even the ionic detergent, sodium cholate. Although fresh blood has been used in this study, nearly identical measurements have been made with blood aged for 7 weeks at 4". This procedure for eluting spectrin and erythrocyte actin avoids use of EDTA, which, in the absence of Ca"+ or Mg2+, promotes irreversible loss of function of muscle actin (50).

Rate Zonal Sedimentation of Spectrin
Spectrin and actin, eluted from erythrocyte ghosts at low ionic strength, sediment as two separate proteins on linear sucrose gradients. More than 50% of the spectrin migrates at a s&h w of about 8 in 20 mM KC1 and 7 at low salt ( Fig. 3A), assuming a partial specific volume of 0.73 (47). Some rapidly sedimenting protein which may represent aggregated material is present under both conditions. Analysis of gradient fractions by SDS-polyacrylamide electrophoresis demonstrates that the peaks of spectrin in 20 mu KC1 (Fig. 4E) and in low salt (not shown) contain no detectable actin or significant amounts of protein other than Bands 1 and 2. Erythrocyte actin is distributed in Fractions 11 and 12 in both gradients. Sedimentation on sucrose gradients provides a simple means of purifying spectrin to electrophoretic homogeneity in a nonaggregated form; such preparations are used exclusively in this report. Spectrin was isolated from erythrocyte ghosts by incubation in 0.3 mM NaPO,, pH 7.6 (see "Methods").
KC1 was added at a final concentration of 20 mM to a portion of the extract, and 0.8-ml aliquots were layered onto linear sucrose gradients (see "Methods") containing either 0.3 mM NaPO,, pH 7.6 (0) or 20 mM KCl, 0.3 mM NaPO,, pH 7.6 (0). Following centrifugation for 18 h at 40,000 r-pm, fractions (0.8 ml) were collected (A). The peak fraction from the gradient containing 20 mM KC1 (0)  The basis of the ionic strength-dependent change in sedimentation velocity for spectrin is not known. The change is reversible (Fig. 3B), and other proteins examined, including bovine serum albumin, beef liver catalase, human IgG, and erythrocyte actin exhibited no change in sedimentation behavior under these conditions. Other experiments with spectrin in low and high ionic strength media4 also indicate sedimentation changes in the analytical ultracentrifuge as well as shape changes of the molecules visualized by electron microscopy. The important finding, in terms of this study, is that spectrin which has been isolated and purified undergoes a change in state under conditions identical to those which favor dissociation of spectrin from the membrane. The reversibility of this change suggested that reassociation of purified spectrin with its membrane site(s) could be achieved under appropriate conditions. Detection of Spectrin Binding to Inverted Vesicles -Inside out vesicles provide a membrane surface with an appropriate orientation to measure reassociation of spectrin (44, 51, 52), and it is possible to prepare such vesicles completely lacking spectrin and actin (Fig. 4). At least 90% of these membranes are inverted by morphological criteria (Fig. IA), although they sediment in 10% dextran (TllO), and in initial experiments released a significant amount of lZI-labeled wheat germ agglutinin (M, = 22,000) which had been bound to the external surface of the membrane (42). The spectrin-depleted inside out vesicles thus are permeable to relatively large molecules and may not provide the selective access to the inner surface of the membrane obtained by Reck and co-workers (51, 52). However, the possibility of nonspecific binding to the external membrane surface is evaluated by using as controls vesicles which are at least 70% right side out (Fig. IB) vesicles contain a full complement of spectrin and actin, and may be unsealed since they sediment in 10% dextran TllO. It is unlikely that the presence of in vivo bound spectrin will interfere with possible binding to the external surface of these vesicles since this spectrin is known to be localized on the cytoplasmic face of the membrane (13). Preliminary results demonstrated that binding of spectrin to inverted vesicles could be detected by SDS-polyacrylamide electrophoresis.
Bands 1 and 2 recombined in approximately equal ratios, and the binding was abolished completely by prior digestion of vesicles with trypsin. A reassociation assay was subsequently developed on the basis of radioactivity measurements.
Many reports have demonstrated a cyclic AMP-independent phosphorylation of a protein co-migrating with Band 2 of spectrin following incubation of erythrocyte ghosts with [y-3*P1ATP (53)(54)(55)(56)(57)(58). Furthermore, approximately 2 atoms of bound phosphate/polypeptide chain have been reported in preparations of spectrin purified in the presence of sodium deoxycholate (59). These findings indicate spectrin may be phosphorylated in vivo, and suggest that radiolabeled, native spectrin may be obtained by incubating erythrocytes with 32P, followed by the usual purification procedure. Spectrin prepared in this way exhibits a major peak of radioactivity comigrating with Band 2, and a smaller amount (about 20%) ["'PISpectrin  (8 Fg,5,500 cpm) were analyzed by electrophoresis on 5% polyacrylamide gels in the presence of 0.2% SDS. One set of gels was sliced into l-mm sections without staining and the other was stained with Coomassie blue. The arrows and numerals represent the position of protein bands on the gels stained with Coomassie blue. TD, tracking dye. Counts (cpm) (above) were determined from the total radioactivity recovered from the gels.
with Band 1 (Fig. 5, Table I). This "2P-labeled protein is eluted from erythrocyte membranes under the same conditions as spectrin (Fig. 12) and co-migrates with spectrin during sedimentation on sucrose gradients, and on gel filtration on 4% agarose (not shown). When spectrin preparations containing the ""P-protein are incubated with inside out vesicles, the specific activity of spectrin which reassociates with membranes is the same as that of the unbound and starting material (Table I). Furthermore, the relative distribution of label between Bands 1 and 2 is not altered in membrane-bound spectrin (Table I). Thus, the 32P may reasonably be assumed to be incorporated into spectrin, and to provide a suitable label for binding studies.

Binding of I:'2P]Spectrin to Membrane Vesicles
Assay -In order to analyze quantitatively the reassociation of spectrin with membranes, it was essential to devise a means of separating free and membrane-bound spectrin which is rapid, complete, reproducible, and adaptable to small volumes and multiple samples. The assay employed here involves sedi- mentation of membranes through a barrier of 20% sucrose in 0.5ml microfuge tubes. More than 90% of the vesicles are pelleted after brief centrifugation at 20,000 x g, and many samples can be run simultaneously with appropriate adapters. The contamination with free spectrin is negligible, and the "nonspecific" contribution to binding is usually less than 15% (see below).
Specificity of Binding-An important feature of a meaningful reconstitution of spectrin is that binding should be restricted to the inner surface of the membrane.
Binding of lS2P1spectrin, measured under standard conditions, is at least lo-fold greater for inverted spectrin-depleted vesicles than right side out spectrin-containing membranes (Fig. 6). Furthermore, the association with inverted vesicles exhibits significant qualitative differences from that with everted membranes (see below), and over 80% of the binding capacity of inverted vesicles is lost following pretreatment of membranes with 0.1 M acetic acid (Fig. 61, or after mild digestion with trypsin (Table II). The loss of binding with acid could be due to denaturation of the "receptor" site(s) for spectrin or to elution of the peripheral membrane proteins which occurs with this treatment (60)." General Features -The binding of ["2]spectrin to inverted vesicles progresses slowly at o", and is maximal at about 90 min, whereas the interaction with right side out vesicles is nearly complete at the earliest time examined (12 min) (Fig.  7). Electrophoresis of membrane-bound and free [32P]spectrin following a 90-min incubation demonstrates no significant dephosphorylation, or redistribution of label (Table II). Also, no proteolysis of spectrin or membrane proteins can be detected. The binding of spectrin to inverted vesicles increases linearly with membrane protein within the range of about 5 to 40 pg (Fig. 6).
The binding of spectrin to inverted vesicles, but not right side out membranes, is affected by pH (Fig. 8). Association is greatest at pH 6.6, decreases by about 50% at pH 7.6, and declines progressively to about the same extent as obtained with right side out vesicles at pH values above 11. Buffers with pH less than 6.5 were not tested since spectrin may precipitate under these conditions (17,49). The enhanced binding at lower pH is due to an increased affinity of the spectrin-membrane interaction, with little change in the number of membrane binding sites estimated by Scatchard plots (see below). Pretreatment of [""Plspectrin at temperatures above 51" nearly abolishes the subsequent binding to inverted vesicles, while the association with right side out membranes is decreased only slightly (Fig. 9). The loss of binding capacity occurs abruptly between 45 and 51", and the curve resembles the cooperative thermal denaturation transitions for other proteins (61). The extent of binding of [32Plspectrin to right side out membranes is equal to that of inside out vesicles under conditions which presumably prevent biologically relevant interactions, and thus provides an estimate of artifactual "binding" due to trapping or adsorption. Therefore, in some experiments, the binding to everted vesicles was substracted from the values for inverted membranes to determine the amount of specific binding.   the plot is linear and the K,, is about 45 pg/ml (Fig. 1OB) Membrane-bound radioactivity was determined (see "Methods"), and the specific binding to inverted vesicles calculated by subtracting the values for binding to right side out membranes, which did not exceed 5 pg/mg of membrane protein.
is not clear. Negative cooperativity appears unlikely since a Hill plot is linear with a slope of nearly unity (Fig. lOB, inset), although more data are required to establish this point.
Competitive displacement of ["'Plspectrin binding by purified, unlabeled spectrin (Fig. 11) demonstrates that binding is saturable, and provides another means of estimating the affnity of the interaction. Fifty per cent of the binding is displaced by about 27 pg/ml of spectrin at pH 7.6, which is thus a minimal value for the K,, under these conditions.

Effect of Zonic Strength and Divalent Metal
Ions on Association and Dissociation of I:"PlSpectrin Dissociation of reconstituted ["2P]spectrin from inverted vesicles depends on ionic strength (Fig. 12) just as does elution of spectrin from erythrocyte ghosts (Fig. 2). Association of [:"P]spectrin with inverted vesicles also depends critically on the salt concentration.
Virtually no binding occurs with 0.3 mM NaPO,, pH 7.6, while binding is maximal at 10 mM KCl, (Fig. 12). Magnesium and calcium chloride in the range of 1 to 4 mM have little influence on binding in the presence of 20 mM KCl; binding is also not markedly affected by EDTA (Table  III). In the presence of low salt, however, 1 mM magnesium ion stimulates binding by almost 20-fold, to the same level obtained with 20 mM KC1 (Table III). Magnesium at 1 mM is also as effective as 10 to 20 mM KC1 in preventing dissociation of [:'"P]spectrin (not shown). Magnesium does not induce gross aggregation of spectrin under these conditions since gel filtration on Sepharose 4B of purified spectrin in the presence of 5 mM MgC12, 20 mM KCl, 0.7 mM NaPO,, pH 7.6, reveals no protein in the void volume and no significant displacement of the eluted volume compared to runs performed in the absence of MgCl,.

Effect of Digestion
with Trypsin -Digestion of inverted vesicles with 0.1 pg/ml of trypsin for 30 min at 24", followed by addition of phenylmethylsulfonyl fluoride and thorough wash- discuss some of the special problems in dealing with this protein before considering the binding data. Spectrin has generally been obtained from a low ionic strength extract of erythrocyte ghosts which has been noted to form filaments or fibrils with addition of 50 to 100 mM salt and/ or 1 to 5 mM Ca*+ or Mg2+ (10, 12, 48, 6'7, 68). Such crude extracts contain Band 5 (M, = 43,000) in levels as high as lo%, which is about equimolar to the amount of spectrin, i.e. Bands 1 and 2, in the extract. Band 5 resembles muscle actin in that it forms filaments with the addition of salt, and is decorated with heavy meromyosin (12). Although erythrocyte actin may be separated from spectrin by gel filtration under some conditions" (11,47,59), others have reported co-migration of these proteins (48, 69) as well as evidence for functionally significant interactions between spectrin and actin (12,69). Band 5 and spectrin have very similar isoelectric points, and are both precipitated over a similar range of pH values (17). Thus, separation of spectrin from actin can be difficult but is essential if the ambiguities of filament formation and association with other peripheral proteins are to be distinguished from the intrinsic behavior of spectin. effect on binding and dissociation of [""Plspectrin as 10 to 20 mM KCl, which is in accord with reports that divalent metals prevent elution of spectrin from ghosts (40, 49).
Binding of [32P]spectrin to inverted vesicles is usually lofold greater than that to right side out membranes, and exhibits significantly different properties. The association with inside out vesicles occurs slowly (Fig. 7), is affected markedly by pH (Fig. 8), is abolished by preincubating Y-Plspectrin at temperatures greater than 50" (Fig. 9), and is saturable at high concentrations of spectrin (Fig. 10). In contrast, binding to right side out membranes is rapid, is relatively unaltered by pH or thermal denaturation of ["*P]spectrin, and rises linearly with increasing levels of ["2Plspectrin.
The physical state of spectrin in solution is presently uncertain. Cross-linking studies (11,31) indicate association between Bands 1 and 2 rather than between 1 and 1 or 2 and 2, but the possibility of a higher polymer of (l-2), or (l-2), G (l-%, + 1 transitions cannot be excluded. The stability and relative proportions of (l-2), and free 1 and 2 is not clear (49,59,70). A further difficulty is that isolated spectrin readily selfassociates to form large aggregates (11,47), although this may be prevented by the presence of detergents (59).
The equilibrium constant for the [:'2P]spectrin-membrane association is about 45 pg/ml at pH 7.6, while at pH 6.5, the high affinity component of binding occurs with a K,, of 4 pgl ml" (see Fig. 10). These concentrations correspond to roughly 10m7 M and IO-# M, respectively, if a molecular weight of 460,000 is assumed, and are in the same range as obtained for association of glyceraldehyde-3-phosphate-dehydrogenase with erythrocyte membranes (34, 36). Inside out vesicles bind maximally about 200 pg of spectrin/mg of membrane protein' at either pH, which is close to the amount of spectrin present in erythrocyte ghosts (30). Unlabeled spectrin competes effectively for ["2P]spectrin binding (Fig. 11).
Spectrin utilized in this study was routinely purified by sedimentation on linear sucrose gradients in which Bands 1 and 2 migrate together as a single peak well separated from Band 5 and aggregated material. The availability of pure, nonaggregated spectrin which can be labeled with [""PIphosphate, and a simple and rapid means of separating free and membrane-bound spectrin make binding studies technically possible. Evaluating such data is more difficult, however. No measurable membrane function is known to be lost or modified by elution of spectrin, and it is likely that in the intact cell spectrin performs a structural role which cannot be reproduced once vesicles are formed. Furthermore, spectrin is presumably attached to the membrane in uiuo, and the binding process examined here has no demonstrated relevance to any known physiological event. In view of the lack of functional correlates of spectrin binding, it is essential to establish other criteria which will distinguish specific and possibly biologically relevant, interactions from nonspecific adsorption. These features would be expected for a selective association resembling the attachment of spectrin in erythrocytes. (a) The ionic dependence of dissociation and association of YPlspectrin should be the same as in native membranes. (b) Binding should be limited to the inner surface of the membrane. (c) Binding should be abolished by treatments of spectrin, and possibly vesicles, which denature proteins. (d) Binding should be saturable at a level similar to the amount of spectrin present in viuo, and exhibit an appropriate affinity.
These features of spectrin binding are necessary aspects of a specific association, but each one is insufficient in itself to provide a clear demonstration that actual reconstitution is occurring.
However, considered together, the experiments strongly support the hypothesis that spectrin is rebinding to the original membrane site to which it is bound in the erythrocyte ghost. If such an interpretation is accepted, the loss of binding following digestion of vesicles with trypsin and pretreatment with dilute acetic acid indicates that the attachment of spectrin to the membrane is due primarily to a protein-protein association. Functionally important interactions between spectrin and phospholipids may also occur in addition. Although spectrin does not associate with liposomes composed of phosphatidylcholine (18,19), binding to extracted erythrocyte lipids has been detected." This interaction is not abolished by heat-denaturing spectrin, and its possible biological significance is not clear.
The binding of ["2P]spectrin observed in this study demonstrates properties consistent with these expectations. YPlSpectrin dissociates and associates with a nearly identical dependence on salt concentrations as observed for elution of spectrin from ghosts (Fig. 2, 12). MgCl, at 1 mM has the same The affinity for binding of spectrin (lo-' to 10mH M) suggests a relatively rapid rate of dissociation from its membrane site(s); a half-life of about 15 min at O", for example, can be calculated from the time course (Fig. 7) for a spectrin membrane complex at pH 7.6 with a K,, of lo-' M. The apparent stability of the attachment of spectrin in unsealed ghosts indicates that additional interactions of spectrin with itself or other proteins (or both) may be operative in viva but are not reconstituted under our experimental conditions. Thus, because the actual conhguration of spectrin in erythrocytes may be intermediate between a loosely bound, self-sufficient network and a form irreversibly attached to the membrane, any controls which spectrin exerts on the topography of transmembrane elements (7,(14)(15)(16)(17)(18)(19) should be considered in dynamic terms where the residence time of spectrin with its membrane site is finite and may be subject to regulation. It is pertinent that a significant lateral motion has been observed for intrinsic erythrocyte proteins labeled with fluorescein, and that the rate of movement depends on the metabolic state of the cell7 Moreover, a measurable rate of rotational diffusion of eosin-labeled Band 3 6 D. M. Shotton and D. Branton, manuscript submitted for publi-' V. Fowler and D. Branton, manuscript submitted for publicacation. tion.
has recently been reported (71). In view of these considerations, it is of considerable interest to determine which membrane protein(s) are involved in the interaction with spectrin. The possible candidates for a binding site are limited presumably to the proteins or glycoproteins present in at least 2 x lo" copies/erythrocyte, which is the approximate number of spectrin molecules (i.e. Bands 1 and 2) (30, 40). Nearly total loss of binding following proteolysis of inverted vesicles is obtained with no significant change in the periodic acid-Schiff-staining profile. After partial digestion only half the binding capacity is lost, but Band 3 is almost completely reduced to a membrane-bound, 50,000-dalton peptide, and high molecular weight proteins in the region of band 2.1 disappear. Further digestion does not affect the 50,000dalton fragment of Band 3, but does abolish binding. These results suggest that neither the major sialoglycoproteins nor Bands 3 or 2.1 serve directly as spectrin binding sites, although these polypeptides could provide a primary anchorage for an intermediate protein to which spectrin is bound directly.
A polypeptide has been partially purified from the tryptic digest of inverted vesicles which blocks 50% of spectrin binding at a concentration of 35 pg/ml; essentially no inhibition is obtained with the purified 45,000-dalton portion of Band 3, released in soluble form by trypsin (64), even at levels as high as 300 pg/ml.8 Competition for spectrin binding thus provides an assay to monitor isolation procedures, and it should be possible with this approach to identify and purify the protein(s) involved in the association of spectrin with the membrane.