Identification of two regions of beta G spectrin that bind to distinct sites in brain membranes.

This study analyzed the complex interactions of intact spectrin with bovine brain membranes by evaluating membrane associations of defined regions of beta G spectrin, the subunit responsible for high affinity membrane binding. Two regions of beta G spectrin were expressed in bacteria and demonstrated to contain fully functional binding site(s) for a subset of spectrin-binding sites in brain membranes depleted of peripheral proteins. One region, located near the NH2 terminus, was comprised of 106-residue repeats and required repeats 2-7 for full activity. The other binding domain was located at the COOH terminus, which is the most variable between beta G and beta R spectrins, is distinct from the 106-residue repeats, and contains a pleckstrin homology domain. NH2-terminal beta spectrin polypeptides interacted with a membrane site(s) that recognized both brain and erythrocyte isoforms of spectrin, was inhibited by calcium/calmodulin, and was not blocked by the COOH-terminal polypeptide. The COOH-terminal region associated with a membrane site(s) that was specific for brain spectrin, was not inhibited by calcium/calmodulin, and was not blocked by the NH2-terminal polypeptide. These observations demonstrate membrane association of spectrin with at least two independent sites, which differ with regard to regulation by calcium/calmodulin and in selectivity for spectrin isoforms.

200 nm in length and contains two subunits, termed a and p, associated side-to-side and head-to-head to form heterotetramers. The ends of spectrin tetramers associate with actin filaments in vitro through interactions mediated by the actin-binding domain of the p subunit. The organization of spectrin and actin has been resolved in erythrocyte membranes where these proteins form a polygonal network with 5-6 spectrin molecules associated with short actin filaments. The spectrin-actin network is linked to the plasma membrane of erythrocytes through association of spectrin with peripheral proteins ankyrin and protein 4.1, which both interact with the cytoplasmic domain of the anion exchanger.
Membrane interactions of spectrin are not well characterized in cells other than erythrocytes. The paradigm of the erythrocyte is not widely applicable in detail since ankyrin mediates * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked indicate this fact.
"advertisement" in accordance with 18 U.S.C. Section 1734 solely to contact between spectrin and a variety of proteins in addition to the anion exchanger (Bennett, 1992). Moreover, ankyrin linkages constitute only a portion of spectrin-binding sites in membranes. Spectrin associates directly with integral proteins in brain membranes, which bind to p subunits of spectrin with KO values in the range of 5-50 n~ (Steiner and Bennett, 1988).
Certain of these ankyrin-independent sites for spectrin are under regulatory control by calcium, and are inhibited by calmodulin (Steiner et al., 19891, and by calpain cleavage of p spectrin (Hu and Bennett, 1991).
The purpose of this study was to identify sites in PC spectrinl that mediate contacts with brain membranes.
Two binding sites of pC spectrin were characterized which are located at NHz-terminal and COOH-terminal regions, and interact with distinct binding sites in brain membranes. These findings provide direct evidence for contacts between spectrin and multiple membrane sites. This information will be important in designing experiments to precisely evaluate functions of individual spectrin-membrane interactions in in vivo studies, and will provide reagents to isolate particular spectrin-binding proteins.
EXPERIMENTAL. PROCEDURES Materials-1261-Labeled Bolton-Hunter reagent was from ICN. Diisopropyl fluorophosphate, leupeptin, pepstatin A, benzamidine hydrochloride, GTP, lysozyme, phenylmethylsulfonyl fluoride, EDTA, EGTA, Tween 20, Triton X-100, and thioglycolic acid were from Sigma. Nitrocellulose paper and electrophoresis reagents were from Bio-Rad. Sucrose, ammonium sulfate, and urea were from SchwarzlMann. Isopropyl p-D-thiogalactopyranoside was from ICN. Deoxyribonuclease 1 was from U. S. Biochemical Corp.. BactoAgar, yeast extract, and Bactotryptone were from Difco Laboratories. NheI was from New England Biolabs and other restriction enzymes as well as bovine serum albumin were from Boehringer Mannheim. Taq polymerase was from Perkin Elmer. Human erythrocyte spectrin (Bennett, 1983), bovine brain spectrin , and the p subunit of brain spectrin  were purified as described. Calmodulin was isolated from bovine brain as described (Gopalakrishna and Anderson, 1982). Brain membranes depleted of peripheral membrane proteins by extraction with sodium hydroxide were prepared from bovine brain tissue as described .
Procedures-Spectrin and recombinant polypeptides were radiolabeled using Bolton-Hunter reagent as described (Bennett, 1983). Association of radiolabeled proteins with peripheral protein-depleted brain membranes was measured as described (Steiner and Bennett, 1988). Protein concentrations were determined by the methods of Bradford (1976) and Lowry et al. (1951) with bovine serum albumin as a standard. SDS-polyacrylamide gel electrophoresis and immunoblotting procedures were performed as described (Bennett and Davis, 1981). Molecular biology methods were performed essentially as described by Sambrook et al. (1989).
Nomenclature for spectrins is based on Hu et al. (1992) and is as follows: pR spectrin refers to p spectrin first characterized in erythrocytes and also expressed in a subset of neurons, skeletal muscle, and cardiac muscle; PG spectrin refers to the generally expressed spectrin present in most tissues except for mammalian erythrocytes; brain and erythrocyte spectrins refers to spectrin isolated from these tissues.  chain reaction was employed to amplify selected regions of cDNA encoding PC spectrin using cDNA contained in plasmids isolated by Hu et al. (1992). The expressed polypeptides have 3 additional residues a t the NH, terminus, and no additional residues at the COOH terminus. Primers for the polymerase chain reaction contained, in addition to spectrin sequences, a stop codon at the 3' end and restriction enzyme sites at the 5' end for NheI and a t the 3' end for EcoRI or XhoI. Primers also contained 4 additional bases beyond the restriction sites to permit efficient digestion of the products by restriction enzymes. The amplified products were then ligated into a PET plasmid with a T7 promotor (Studier et al., q 9 0 ) purchased from Promega (pGEMEX). Digestion of this plasmid with NheI deletes the region encoding a viral gene 10 protein that otherwise would be included on the amino termini of recombinant po1,ypeptides. Ligated plasmids were used to transform JM109, a nonexpressor bacterial strain lacking the T7 polymerase. Plasmids with inserts were subsequently isolated and transfected into BL21, an Escherichia coli strain with a T7 polymerase under control of the lacZ promoter (Studier et al., 1990). BL21 is deficient in the Ion protease and lacks the ompT outer membrane protease. The BL21 strain employed in this study also contains a plasmid, pLysS, that confers resistance to chloramphenicol and encodes T7 lysozyme which inhibits the T7 RNA polymerase and reduces expression of the recombinant proteins. Bacteria were isolated following culture and induction as described (Davis et al., 1991). Bacterial pellets were washed with 0.1 M NaCI, 10 m~ sodium phosphate buffer, pH 7.4, resuspended in 50 m~ sodium phosphate, 1 m NaEDTA, 25% sucrose, 0.5 mg/ml lysozyme, pH 8.4, with protease inhibitors (10 pg/ml pepstatin A and leupeptin, 0.5 m diisopropyl fluorophosphate, 5 m~ benzamidine), and were incubated for 30 min in ice. MgCl, was added to a final concentration of 10 m~ and DNase to 40 pg/ml and allowed to partially digest bacterial DNA during a 30-min incubation. Protein was extracted by the addition of 2 volumes of extraction buffer (200 m NaCI, 2 m NaEDTA, 1 m dithiothreitol, 20 m sodium phosphate, 1% Triton X-100, plus protease inhibitors, pH 7.4). The suspension was forced through a 20-gauge needle, followed by centrifugation (20 min at 5000 rpm) to collect the supernatant which contained the expressed polypeptides. Polypeptides were precipitated with ammonium sulfate a t 60% saturation, and purified by successive steps of gel filtration on Superose 12 in the presence of 1 M NaBr, 1 m dithiothreitol, 10 m~ sodium phosphate, 1 m NaEDTA, 1 m NaN3, and 0.05% Tween 20, ion exchange chromatography on Mono Q and Mono S columns, and in some cases hydroxylapatite chromatography. Pure proteins were dialyzed against 100 m~ NaCl, 10 m~ sodium phos-C -$j Protein, nY tides derived from PC spectrin, residues 376-1246 (Panel B ) and residues 2-2333 (Panel C) by the corresponding unlabeled hc. 2. Displacement of binding to bovine brain membranes of lmI-labeled brain spectrin (Panel A), and recombinant polypeg spectrin and recombinant polypeptides. Various concentrations of unlabeled brain spectrin and recombinant polypeptides were incubated for 30 min at 2 "C with 10 pg/ml bovine brain membranes depleted of peripheral membrane proteins ("Methods") in a buffer containing 100 m~ NaCI, 10 m Hepes, 1 m NaEGTA, 1 m M NaN3, 3 mg/ml bovine serum albumin. 12sI-Labeled proteins (5 n~; 1.5-3.5 x los cpdpmol) were then added in a final volume of 0.2 ml and the incubation continued for 120 min. Membrane-bound and free radiolabeled proteins were separated in 0.4-ml Microfuge tubes by sedimentation of membranes for 20 min a t 5,000 x g through a bamer of 10% sucrose dissolved in assay buffer. The tubes were nonspecific binding by either subtracting values obtained with heat-denatured protein in the case of spectrin, or values obtained with 640 IIM frozen, and tips, containing membrane-bound ligands, and tops, containing free ligand, were assayed for lZsI. The data were corrected for unlabeled polypeptides for recombinant polypeptides. Data are presented as percent of control values determined in the absence of unlabeled polypeptides. phate, 1 m~ NdDTA, 1 m~ dithiothreitol, 0.05% 'heen 20, 10% sucrose, 1 m~ NaNs. and stored frozen at -70 "C.

Identi/kation of NH2-terminal and COOH-terminal Regions of pa Spectrin Capable of Binding to Bmin Membmnee'ho
regions corresponding to residues 376-1246 (containing complete repeats 2-43) and COOH-terminal residues 2060-2333 of a; spectrin were expressed in bacteria and evaluated for ability to bind to sites in brain membranes (Fig. 1). The rationale for selection of residues 376-1246 was that a calpain fragment of a: spectrin missing the NH2-terminal 40 kDa was capable of binding to brain membranes, although with reduced affinity (Hu and Bennett, 1991). These observations suggested that at least one component of the binding sit&) would be located adjacent to the NH2-terminal actin-binding domain. COOHterminal residues were selected since the COOH-terminal do-Protein, nM main is the most variable between & and PR spectrins and is ma. 4. Displacement of binding of 1 S W a b d e d brain epectrin a candidate site to mediate &-specific interactions previously to brain membranes by recombipept p o l s p e p t i~ derived lrom noted in brain membranes (Steiner and Bennett, 1988).
ai speetrin residuem 376-1246 and residues w)80-8353. Binding of Expressed polypeptide^ were evaluated for proper folding by branes was measured (Fig. 2) in the presence of various concentrations lmI-labeled brain spectM (2 m, 3.5 x 106 cpdpmol) to brain memcircular dichroism SpeCbScoPY and sensitivity to Proteam diof unlabeled brain spectrin (O), epeetrin (aa 376-1246) (0). betk gestion (data not shown). The circular dichroism spectrum of spectrin (aa 2060-2333) ( 0 , and a combination of both spectrin polyresidues 376-1246 was consistent with equivalent or greater a peptides (D. Data are Presented Pe-t of Control V a l u e s deterhelical content than native spectrin. The spectrum of the COOH-terminal domain, in contrast, had a much lower a helical content. Both NHz-terminal and COOH-terminal polypep Data from displacement experiments of NH&xmhal and tides were relatively resistant to digestion with chymotrypsin COOH-terminal polypeptides in Fig. 2 were used to calculate under conditions where heat-denatured NH2-terminal polypep amounts of polypeptide bound at each concentration of unlatide was completely digested. A further examination of folding beled polypeptide, and these values are presented as Scatchard was not necessary since the polypeptides were fully functional plots (Fig. 3). The plot for spectrin (Fig. 34) includes a high in binding assays when compared to intact fl spectrin (see be-affinity component with an apparent KO of 2 n~, and a lower low).
affinity portion with apparent KO values ranging from 30 to 100 NH+xminal and COOH-terminal PC polypeptides were m, as noted previously (Steiner and Bennett, 1988). The caevaluated for activity in association with brain membranes pacity of brain membranes for spectrin was in the range of 100 depleted of ankyrin and other peripheral membrane proteins pmol/mg membrane protein, although this value is imprecise by alkaline extraction (Fig. 2). Association of radiolabeled poly-due to experimental limitations in obtaining binding data at peptides with brain membranes was displaced by unlabeled high concentrations of spectrin. The plot for the NH2-terminal polypeptide in each case (Fig. 2). Half-maximal displacement polypeptide contains a class of sites with a KO of 35 m (Fig. occurred with 25 I~M NH&mnhal polypeptide, and 40 m 3B). The plot for the COOH-terminal polypeptide also could be COOH-terminal polypeptide (Fig. 2). Binding data in these ex-fitted to a single class of sites with a KO also of 35 m (Fig. 3C).
periments were corrected for nonsaturable interactions by sub-The capacity of brain membranes for both polypeptides was traction of values obtained with a 128-fold excess of unlabeled approximately 100 pmoVmg membrane protein.
brain membranes (Fig. 4). NH2-terminal and COOH-terminal It was important in evaluating activities of recombinant P spectrin polypeptides to compare them to isolated brain P spectrin in displacement assays (Fig. 5). An assumption in these experiments was that /3 spectrin isolated from forebrain is predominately comprised of the PC isoform of spectrin. This assumption is based on the fact that PC is expressed in high levels in forebrain and is the major component of brain spectrin (Davis and Bennett, 1983;Hu et al., 1992). PR Spectrin is expressed in only a minor subset of neurons in the forebrain (Lambert and Bennett, 19931, although cerebellum does express substantial levels (Riederer et al., 1986). Brain p spectrin was nearly equivalent to the NHz-terminal recombinant polypeptide, and actually less active than COOH-terminal polypeptide in displacing binding of these polypeptides to brain membranes (Fig. 5). This result indicates that the recombinant polypeptides contain fully functional binding site($ for a subset of spectrin-binding proteins in brain membranes. Spectrin and recombinant P spectrin polypeptides exhibited lower activity in binding assays than native spectrin tetramer (Figs. 3-5). The reduced activity of isolated P spectrin could result from the fact that native spectrin contains a as Membrane binding of radiolabeled proteins 5 m; 1.5-2.5 x lo5 cpdpmol) was measured as in Fig. 2. Data are presented as percent of control values determined in the absence of unlabeled polypeptides. well as P subunits, and potentially is multivalent with two 0 subunits.
NH2-terminal and COOH-terminal PG Polypeptides Bind to Distinct Sites in Brain Membranes-Several different types of experiments demonstrate that NHz-terminal and COOH-terminal recombinant PG polypeptides associate with distinct sites and that these sites differ in specificity for spectrin isoforms as well as regulation by calmodulin. Evidence for distinct sites is provided by the lack of ability of unlabeled COOH-terminal polypeptide to displace binding of radiolabeled NHz-terminal polypeptide to membranes (Fig. 6). Similarly, unlabeled NHZterminal polypeptide lacks activity in displacing binding of COOH-terminal polypeptide (Fig. 6).
Brain membrane sites for NHz-terminal and COOH-terminal polypeptides differ with respect to specificity for brain and erythrocyte spectrin. Binding of the NHz-terminal polypeptide is displaced almost equivalently by spectrin isolated from human erythrocytes or bovine brain (Fig. 7). In contrast, binding of the COOH-terminal polypeptide is displaced by brain spectrin, but not by erythrocyte spectrin (Fig. 7). The finding that the membrane site(s) that recognize the NHz-terminal polypeptide do not distinguish between erythrocyte and brain spectrin is not surprising in view of the high degree of sequence similarity between PR and PC spectrins in their NHz-terminal regions (Hu et al., 1992). The lack of ability of brain membrane sites that bind the COOH-terminal polypeptide to recognize erythrocyte spectrin also is consistent with the complete diver- variant of spectrin expressed in erythrocytes.
Brain membrane sites that recognize NH2-terminal and COO€€-terminal PG spectrin polypeptides also differ with respect to regulation by calmodulin. Binding of the NH2-terminal & spectrin poiypeptide was inhibited by 76% by calmodulin in the presence of calcium, while binding of the COOH-terminal polypeptide wae unageeted (Fig. 8). Half-' 1 inhibition of binding of the N H r t e f m i n a l polypeptide occurred at 160 n~ calmodulin. The membrane site(s1 recognized by the NH2-termind polypeptide thus are responsible for the calmodulin-sensitive binding ofbrain spectrin noted previously in brain membranes (Steiner et aL, 1989). Binding of the isolated N H r terminal polypeptide is inhibited by lower concentrations of calmodulin than that ofintact spectrin, which requires about 1 p~ for half-maximal inhibition (Steiner et al., 1989). The in-  2-8,3-8,4-8, and 5-8 (Fig. 9) were compared in terms of activity in displacing binding of radiolabeled NHz-terminal polypeptide 376-1246 (repeats 2-8) to brain membranes (Fig.  10). Repeat boundaries are based on the folding unit defined by Winograd and Branton (1991) for a spectrin. Deletion of residues 376-623, corregpondiag to a portion of repeat 1 and all of repeat 2, i n d the concentration required for half-maximal inhibition from 65 m in this experiment to 125 m. Further deletion of residues 623-60!3, corresponding to moat of repeat 3, caused a small inmease in concentration required for 60% displacement to 150 m. However, deletion of residues 609-739 corresponding to loss ofrepeat 4 resulted in a polypeptide with a greater than 10-fold increase in concentration required for 50% displacement (Fig. 10). These results suggest that repeats 2 and 3 contribute to activity but that repeat 4 is essential.
Activity of a polypeptide containing residues 1-777, corresponding to the actin-binding domain, repeata 1-4, and a portion of repeat 6, was evaluated to determine ifrepeats 24 were d c i e n t for biuding and ifthe & -b i n d i n g domain or repeat 1 contributed significantly to membrane binding (Fig. 10). This polypeptide was about 6-fold less active than spectrin residues 376-1246 (Fig. 10). Reduced activity of this polypeptide is most likely not due to gross misfolding, since the polypeptide was not aggregated and was resistant to protease digestion (data not shown). The partial activity of the actin-binding domain plus repeats 1-4 indicates that repeat 4 is not sufficient for high-affinity binding, and that the actin-binding domain is not involved in membrane interactions of /3 spectrin measured in these assays.
Involvement of repeats COOH-terminal to repeats 2-4 was explored using constructs with various COOH-terminal deletions (Fig. 11). Deletion of residues 1037-1246 containing a portion of repeat 7, all of repeat 8, as well as part of repeat 9 had little effect on activity of polypeptides that retained repeats 2 4 (data not shown). Deletion of residues 1037-1246 did reduce by about %fold the activity of a smaller polypeptide retaiuing only repeats 4-6 (Fig. 11). These results indicate that, to a first approximation, residues 1037-1246 do not play a major role in association with the membrane site.
COOH-terminal deletions beyond residue 1037 significantly reduced activity in polypeptides that retained repeats 4 and 5. Deletion of residues 960-1037 corresponding to about 80% of repeat 7 resulted in a 4-fold loss of activity, while deletion of repeat 6 resulted in a greater than lO-fold loss of activity (Fig.  11). These results provide evidence for a role in membrane association of repeata 6 and 7 in addition to repeats 2-4.

DISCUSSION
Measurements of association of spectrin with brain membranes depleted of peripheral proteins have suggested heterogeneity in ankyrin-independent binding sites for spectrin that vary in affinity for spectrin, and selectivity for spectrin isoforms (Steiner and Bennett, 1988;Steiner et al., 1989). This study approached the problem of dissecting the complex interactions of intact spectrin by evaluating associations of defined regions of one of the spectrin subunits. Two regions of PC spectrin were identified that associate with distinct and most likely A B C D Mr X 10-3

28-
independent membrane sites. The NH2-terminal region requires spectrin repeats 2-7 for full activity, interacts with a membrane site(s) that recognizes both brain and erythrocyte isoforms of spectrin, and is inhibited by calciudcalmodulin. The COOH-terminal region, which is comprised of nonrepeat sequence, associates with membrane sites that are specific for the general isoform of spectrin, and are not inhibited by calciudcalmodulin. These observations with expressed /3 spectrin polypeptides provide strong evidence for association of spectrin with multiple membrane sites, and provide the tools for future characterization of spectrin binding sites.
The two regions of PC spectrin implicated in membrane interactions are separated from each other by about 60 nm along the folded spectrin molecule, and presumably could associate simultaneously with both attachment sites. It will be important to determine if the membrane-binding sites for spectrin are co-expressed in the same cells and if they are co-localized in the same membrane domains. In any case, the potential for interactions of spectrin with both calmodulin-sensitive and calmodulin-insensitive contacts has interesting implications for regulation of spectrin-based structures in brain. Signals leading to elevation of calcium ion and dissociation of one class of spectrin interactions would have no effect on the calmodulin-insensitive contacts. This difference in regulation would allow specific assembly-disassembly events to occur either in particular cells, terminal domain with membranes could result from independent contacts with more than one membrane site. Alternatively, a single membrane site may contact an extended portion of P spectrin and require multiple repeats for complete activity. A prediction based on two or more independent contacts is that a truncated polypeptide reduced in activity for one of the sites should bind with unaltered affinity for the second site. This prediction was tested by comparing ability of a full-length polypeptide (residues 376-1246) and a truncated polypeptide (residues 609-1246) to displace binding of labeled truncated polypeptide (residues 609-1246) (Fig. 12). The truncated polypeptide was less active than the full-length polypeptide in displacing binding of labeled truncated polypeptide. This result does not support the presence of independent membrane sites, and is consistent with a single membrane site that requires same spectrin, depending on the degree of overlap in expression of the two types of sites.
The COOH-terminal domain of PC spectrin was selected for these experiments based on several features that suggest potential for protein interactions. A globular knob most likely corresponding to the COOH-terminal domain has been visualized in the midregion of spectrin tetramers (Dubreuil et al., 1990) and is in an excellent location to mediate interactions between spectrin and other structures. The sequence of the COOH-terminal domain diverges from the 106-residue repeats that comprise the major portion of both (Y and /? subunits of spectrin, and is the most variable between P spectrin isoforms (Winkelmann et al., 1990;Hu et al., 1992). The COOH-terminal domain thus is a likely candidate to mediate isoform-specific interactions of spectrins. Another interesting feature of the COOH-terminal domain is a 106-residue region (residues 2099-2305) with homology to sequences termed pleckstrin homology domains, which are present in a diverse group of proteins including signaling proteins ras-GAP, ras-GRF, son of sevenless, and pleckstrin, a major substrate for protein kinase C in platelets (Mayer et al., 1993;Haslam et al., 1993). Pleckstrin homology domains have been proposed, by analogy with SH2 and SH3 domains, to participate in protein interactions potentially regulated by phosphorylation, and to have a role in signal transduction. It will be of interest to determine if the membrane interactions of PC spectrin COOH-terminal domain detected in this study are mediated by the pleckstrin homology motif.  Fig. 9). Binding of lZ5I-labeled recombinant PC spectrin residues 376-1246 (5 n~; 2 x lo5 cpdpmol) to brain membranes was measured in the presence of various concentrations of unlabeled recombinant polypeptides as described in the legend to Fig. 2 FIG. 11. Effect of COOH-terminal deletions on activity in displacement of binding to brain membranes of 1261-labeled recombinant PC spectrin residues 609-1246. Recombinant polypeptides were derived from pc spectrin residues 609-1246 (O), residues 609-1036 (Dl, residues 609-949 (m), and residues 609-847 (0) (see Fig. 9).
Binding of 1251-labeled recombinant PC spectrin residues 609-1246 (10 I " ; lo5 cpdpmol) to brain membranes was measured in the presence of various concentrations of unlabeled recombinant polypeptides as described in the legend to Fig. 2. Data are presented as percent of control values determined in the absence of unlabeled polypeptides.
Protein, nM   FIG. 12. Comparison of activity of recombinant polypeptides derived from PC spectrin residues 376-1246 (0) and residues 609-1246 (0) in displacing binding of lmI-labeled Pa spectrin residues 809-1246 to brain membranes. Binding of 12SI-labeled recombinant pc spectrin residues 609-1246 (10 m; lo6 cpdpmol) to brain membranes was measured in the presence of various concentrations of unlabeled recombinant polypeptides as described in the legend to Fig. 2. Data are presented as percent of control values determined in the absence of unlabeled polypeptides.
Repeats 2-7 required for full activity of the calmodulin-sensitive NH2-terminal region extend approximately 20 nm along the length of spectrin (Figs. 10 and 11). Requirement of multiple repeats for high affinity binding could, in principle, result from participation of several repeat domains in correct folding of recombinant polypeptides. A conformational role for multiple repeats is not likely, however, since each 106-residue repeat of spectrin is believed to represent a n independent folding unit (Winograd and Branton, 1991). Moreover, the circular dichroism spectrum of the inactive recombinant polypeptide, residues 739-1246, was almost identical to the spectra of active recombinant polypeptides (data not shown). Another explanation for such a large apparent binding site is that association with the membrane involves a n extended contact along the surface of spectrin. It also is possible that contact sites between distant residues of p spectrin are formed due to bending of spectrin. In support of this idea, many of the images of spectrine visualized by electron microscopy exhibit a kink 20-30 nm from the ends of spectrin tetramers (Bennett et al., 1982;Dubreuil et al., 1990), in the position occupied by the portion of sequence implicated in membrane binding.
The capacity of brain membranes for recombinant p spectrin polypeptides is comparable to amounts of native spectrin actually present in brain membranes. Binding capacities were in the range of 100 pmoVmg of membrane protein for membranes depleted of peripheral membrane proteins, which corresponds to about 20 pmoVmg of total, unextracted membranes. Spectrin, for comparison, is present in 30 pmoVmg of total brain membrane (Bennett et al., 1982). Candidates for spectrin-binding proteins currently include N-CAM180 (Pollerberg et al., 1987), and CD45, a membrane-spanning protein with a cytoplasmic domain containing tyrosine phosphatase activity (Lokeshwar and Bourguignon, 1992). I t should be possible with the recombinant p spectrin polypeptides characterized in this study to evaluate these proteins for specific interactions as well as systematically search for heretofore unidentified spectrinbinding proteins.