Identification of the Spectrin Subunit and Domains Required for Formation of Spectrin/Adducin/Actin Complexes*

Adducin is an actin-binding protein that has been pro- posed to function as a regulated assembly factor for the spectrin/actin network. This study has addressed the question of the subunit and domains of spectrin required for formation of spectrin/adducin/actin com- plexes in in vitro assays. Quantitative evidence is pre-sented that the (cid:98) -spectrin N-terminal domain plus the first two (cid:97) -helical domains are required for optimal participation of spectrin in spectrin/adducin/actin com- plexes. The (cid:97) subunit exhibited no detectable activity either alone or following association with (cid:98) -spectrin. The critical domains of (cid:98) -spectrin involved in complex formation were determined using recombinant proteins expressed in bacteria. The N-terminal domain (residues 1–313) of (cid:98) -spectrin associated with F-actin with a K d of 26 (cid:109) M , and promoted adducin binding to F-actin with half-maximal activation at 110 n M . Addition of the first (cid:97) -helical domain (residues 1–422) lowered the K d for F-actin by 4-fold to 6 (cid:109) M , but also reduced the capacity by 3-fold and had no effect on interaction with adducin. Further addition of the second (cid:97) -helical domain (resi-dues 1–528) did not alter binding to F-actin but resulted in a 2-fold increased activity in promoting adducin binding with half-maximal activation at 50 n M . Addition of up to eight additional (cid:97) -helical domains (residues 1–1388) resulted in no further change in F-actin binding or as- sociation with adducin. These results demonstrate an unanticipated role of the first repeat of (cid:98) -spectrin in actin binding activity and of the second repeat in asso- ciation with adducin/actin, and imply the possibility of an extended contact between adducin, spectrin, and ac- tin involving several actin subunits. but a role of repeat domains in actin binding has not been In this study, we report that the (cid:98) -spectrin N-terminal domain plus the first two (cid:97) -helical repeats are required for full activity in formation of a ternary complex with adducin and actin. We also present evidence that the first (cid:97) -helical repeat of (cid:98) -spectrin contributes to association with actin filaments. These findings support a model for the adducin/spectrin/actin complex with lateral association of spectrin and adducin tail domains along the actin filament and involving several actin subunits. into a 10-ml Mono Q and/or Mono S ion exchange column and eluted by a linear 0.05–0.5 M NaBr gradient. Typically a 3-liter culture can produce 2–20 mg of protein with over 95% purity. Assay of Spectrin/Actin/Adducin Complex Assembly—In vitro co- sedimentation assays of association of spectrin and adducin with F-actin were performed as described previously (Gardner and Bennett, 1987).

The spectrin-based membrane skeleton was first visualized as a detergent-insoluble assembly of proteins in erythrocytes (Yu et al., 1973). Spectrin and its associated proteins are now known to be expressed in cells of most metazoan organisms (reviewed by Bennett and Gilligan (1993)). Proposed functions of spectrin-based membrane skeletons include maintenance of mechanical stability of erythrocyte membranes, organization of integral proteins in specialized membrane domains, and regulation of vesicle-membrane interactions. Electron microscopy has resolved the erythrocyte membrane skeleton as a twodimensional network with 5-6 spectrin molecules linked to short actin filaments to form a sheet of 5-6-sided polygons (Byers and Branton, 1985;Liu et al., 1987;Shen et al., 1986). These striking images raise the issue of how such a regular network can be assembled and have focused attention on the accessory proteins localized at spectrin-actin junctions (reviewed by Bennett (1990)).
Adducin is localized at spectrin-actin junctions (Derick et al., 1992) and has been proposed to contribute to assembly of the spectrin-actin network (Gardner and Bennett, 1987). Adducin was initially discovered as a calmodulin-binding protein in erythrocytes ) and now is known to be expressed in brain as well as most other tissues Joshi et al., 1991). Adducin binds to spectrin-actin complexes with high affinity but binds to either spectrin or actin alone with low affinity. Adducin also promotes addition of a second spectrin molecule to spectrin-actin complexes in a calcium-calmodulin regulated manner Bennett, 1987, 1988). In epithelial tissues and cultured cells, adducin is co-localized with spectrin and actin at lateral domains of cellcell contact (Kaiser et al., 1989). Adducin is phosphorylated by protein kinase A and protein kinase C (Waseem and Palfrey, 1988;Ling et al., 1986) and is a candidate to participate in dynamic behavior of the spectrin skeleton.
Understanding the structural basis for the spectrin/adducin/ actin complex promises to provide insights into assembly and regulation of the membrane skeleton network. The domain structure of adducin includes an N-terminal 39-kDa globular domain connected by a 9-kDa neck domain to a 33-kDa protease-sensitive tail domain . Tail domains of both ␣ and ␤ adducin are responsible for binding of adducin to spectrin-actin complexes (Hughes and Bennett, 1995). Spectrin is composed of rod-shaped ␣ and ␤ subunits laterally associated in an anti-parallel orientation. The region of spectrin involved in association with actin contains the C terminus of the ␣ subunit and the N terminus of ␤-spectrin. The N terminus of ␤-spectrin contains a conserved actin-binding domain (Karinch et al., 1990), which is also found in a large family of actin-binding proteins (reviewed by Matsudaira (1991)). ␤-Spectrin also contains 17 repeats of a 106residue motif folded into triple ␣-helical units  followed by a C-terminal variable region (Hu et al., 1992;Byers et al., 1992;Winkelmann et al., 1990;1991). The 106-residue motif is also found in other actin-cross-linking proteins including ␣-actinin and dystrophin, but a role of repeat domains in actin binding has not been addressed.
In this study, we report that the ␤-spectrin N-terminal domain plus the first two ␣-helical repeats are required for full activity in formation of a ternary complex with adducin and actin. We also present evidence that the first ␣-helical repeat of ␤-spectrin contributes to association with actin filaments. These findings support a model for the adducin/spectrin/actin complex with lateral association of spectrin and adducin tail domains along the actin filament and involving several actin subunits.

Methods
Procedures-Determination of protein concentration was performed by the procedure of Bradford (1976) using bovine serum albumin as a standard. Alternatively, in the case of purified proteins, absorbance at 280 nm was measured and the concentration of protein was calculated using an extinction coefficient estimated from amino acid sequence by the methods of Gill and von Hippel (1989). Polyacrylamide gel electrophoresis was performed using 0.2% SDS in the buffers of Fairbanks et al. (1971) and 1.5-mm-thick 3.5-17% exponential gradient slab gels. Protein iodination was performed with Bolton-Hunter reagent (Bennett, 1983).
Purification of Proteins-Actin was isolated from acetone powder of rabbit skeletal muscle (Pardee and Spudich, 1982) with the modification that the chromatography step was performed on a Superose 12 column. Bovine brain spectrin was isolated from high salt extracts of brain membranes . ␣-Actinin was isolated from chicken gizzard (Feramisco and Burridge, 1980). Erythrocyte adducin was isolated from low salt extracts of human red cell membranes as described (Hughes and Bennett, 1995).
Expression of ␤-Spectrin Constructs in Escherichia coli-Human ␤ G spectrin cDNA was used as a template in subcloning (Hu et al., 1992). An NheI site followed by an ATG codon was included at the beginning of the PCR product. An EcoRI site following a stop codon was placed at the end of the PCR product. These restriction sites were used to clone the PCR products into the pGEMEX vector, a pET plasmid with a T7 promoter. The recombinant plasmids were transformed into BL21 (DE3/pLysS) bacterial strain. Overexpression of recombinant polypeptides in this E. coli host was induced by 0.5 mM isopropyl-1-␤-D-galactopyranoside (Studier et al., 1990). After 3 h of induction with isopropyl-1-␤-D-galactopyranoside, 3 liters of bacterial culture were centrifuged at 5,000 ϫ g for 10 min at 4°C.
All purification steps were performed at 4°C and were monitored by SDS-polyacrylamide gel electrophoresis. Each overexpressed construct had the expression level of over 50% of total proteins and the mobility corresponding to the predicted molecular mass. Precise procedures for purification varied for different constructs as well as different prepara-tions of the same construct. In general, the bacterial pellet was lysed by freezing/thawing once and resuspended in 100 -200 ml of 50 mM sodium phosphate, pH 7.4, 1 mM NaEGTA, 25% sucrose, with protease inhibitors (1 mM 4-(2-Aminoethyl)-benzenesulfonyl, 10 mM benzamidine, 10 g/ml pepstatin A, 200 g/ml phenylmethylsulfonyl fluoride as final concentrations), and incubated on ice for 30 min. DNase was added to a final concentration of 40 g/ml and MgCl 2 to 10 mM during the incubation. Then protein was extracted by addition of 2 volumes of Triton containing buffer (200 mM NaCl, 20 mM sodium phosphate, pH 7.4, 2 mM NaEDTA, 1% Triton X-100, 1 mM DTT, and protease inhibitors as mentioned above). The suspension was forced through a 18-gauge needle once and centrifuged for 20 min at 5,000 ϫ g. Usually 50 -90% of construct proteins were in the Triton-soluble fraction, which was processed as described next. The supernatant of Triton-extracted lysate was precipitated by addition of solid (NH 4 ) 2 SO 4 to achieve 60% saturation (3.9 mg/ml starting solution). After incubation for 30 min on ice, the precipitated proteins were pelleted by centrifugation for 30 min at 5,000 ϫ g. The protein pellet was solubilized in 20 -30 ml of 1 M NaBr, 10 mM sodium phosphate, pH 7.4, 1 mM NaEDTA, 1 mM NaN 3 , 0.05% Tween 20, 0.5 mM DTT at a protein concentration of 0.5-3 mg of protein/ml and was dialyzed against the same buffer for 6 h before it was loaded into a 1.5-liter Superose 12 column equilibrated with the same buffer. Fractions containing the construct protein were pooled and dialyzed against 50 mM NaBr, 10 mM sodium phosphate, pH 7.4, 1 mM NaEDTA, 1 mM NaN 3 , 0.05% Tween 20, 0.5 mM DTT for 6 h. Then the protein was loaded into a 10-ml Mono Q and/or Mono S ion exchange column and eluted by a linear 0.05-0.5 M NaBr gradient. Typically a 3-liter culture can produce 2-20 mg of protein with over 95% purity.
Assay of Spectrin/Actin/Adducin Complex Assembly-In vitro cosedimentation assays of association of spectrin and adducin with Factin were performed as described previously (Gardner and Bennett, 1987).

Identification of the Spectrin Subunit Required for Association with
Adducin/Actin-The first step in mapping the spectrin contribution to spectrin/adducin/actin complexes was to determine which subunit(s) of spectrin are involved. The approach involved determining the activity of native spectrin and isolated spectrin subunits (see "Methods") in promoting adducin binding to actin (Fig. 1). Native spectrin tetramer increased binding of 125 I-labeled adducin to F-actin in a saturable manner with half-maximal activation at about 10 nM spectrin. Adducin can associate with actin independently from spectrin, and this is a high capacity but low affinity binding with a K d of 28 M (Gardner and Bennett, 1987). Assays in this study were performed with 5 nM adducin and 1.5-2 M F-actin. Under these conditions, the extent of binding of adducin to actin is 2-3-fold higher in the presence of spectrin. The concentration required for half-maximal activation of adducin binding is used in this study as a measure of activity of spectrin in stabilizing spectrin/adducin/actin complexes. Double-reciprocal plots of this data are linear (Fig. 1B) and allow estimates of concentrations required for half-maximal activation.
␣-Spectrin exhibited essentially no activity in this assay. The lack of activity of the ␣ subunit is not due to aggregation since a gel filtration column profile of ␣-spectrin showed that over 90% of this subunit was in the non-aggregated fractions (data not shown). ␤-Spectrin, in contrast to the inactive ␣ subunit, increased binding of 125 I-labeled adducin to F-actin in a saturable manner with half-maximal activation at 22 nM. This apparent affinity is lower than that of native spectrin (10 nM), but still displays high affinity binding. In other experiments, the concentration required for half-maximal activation for the ␤ subunit varied from 20 -40 nM. ␤-Spectrin exhibited only half the extent of activation of the native spectrin tetramer. The basis of reduced extent and reduced affinity of ␤-spectrin compared to spectrin tetramer is not known. However, ␤-spectrin differs from native spectrin in several respects that could be relevant: (a) ␤-spectrin is a monomer while native spectrin is a ␣ 2 ␤ 2 tetramer, and ␤-spectrin therefore only has a single actin-binding site compared to two sites in the native tetramer; (b) monomeric ␤-spectrin lacks lateral contacts with an ␣ subunit. The question of whether ␣-spectrin contributes to ␤-spectrin activity in the complex will be addressed below (see Fig. 8). The conclusion from the data in Fig. 1 is that, to a first approximation, ability to associate with adducin/actin complexes resides in ␤-spectrin.
Mapping Adducin and Actin Binding Domains in ␤-Spectrin-␤-Spectrin constructs were evaluated to determine the minimal region sufficient for full activity of ␤-spectrin in forming spectrin/adducin/actin complexes (Fig. 2). The N-terminal domain (resides 1-313) includes the actin-binding domain (residues 47-186) (Karinch et al., 1990) as well as all the N-terminal sequence before the first ␣-helical domain. Increasing numbers of ␣-helical domains were added to the N-terminal domain of ␤-spectrin based on definition of Drosophila ␤-spectrin segments (Byers et al., 1992). The ␣-helical domains can be folded properly and individually if their boundaries correspond to folding units (Winograd et al., 1991). Constructs were expressed in bacteria and purified by ion exchange chromatography and gel filtration. Proper folding was evaluated by circular dichroism spectroscopy, which provides a sensitive measure of ␣-helical secondary structure. The constructs exhibited spectra corresponding to 60 -70% ␣-helix (results not shown), which is consistent with the helix composition of native spectrin (Calvert, et al., 1980;Yan et al., 1993). For some constructs, protease resistance also was evaluated. The results of circular dichroism spectra and protease resistance combined with behavior on gel filtration indicated that the bacterially expressed ␤-spectrin constructs were properly folded and unaggregated.
␤-Spectrin constructs were evaluated for activity in promoting binding of adducin to F-actin (Figs. 1, 3, and 4). A construct containing the N-terminal domain plus 10 ␣-helical domains (residues 1-1388) was equivalent to the intact native ␤ subunit in terms of the concentration required for half-maximal activation, although residues 1-1388 exhibited a greater extent of activation (Fig. 1). ␤-Spectrin constructs also were equivalently active that contained the N-terminal domain plus either 2, 3, 4, or 10 ␣-helical domains (Fig. 3). Therefore, the N-terminal domain plus the first two triple helical domains is sufficient for full activity of ␤-spectrin. However, deletion of the second ␣-helical domain reduced the apparent affinity 2-fold resulting in half-maximal activation at 120 -140 nM (Fig. 4). Additional deletion of the first ␣-helical domain resulted in no further FIG. 1. The effect of ␣-spectrin, ␤-spectrin, native spectrin tetramer, and ␤-spectrin residues 1-1388 on binding of 125 I-labeled adducin to F-actin. A, 125 I-labeled adducin (5 nM, 127,500 cpm/pmol), and polymerized rabbit skeletal muscle actin (2.0 M) were incubated with increasing concentrations of bovine brain ␣-spectrin, ␤-spectrin, spectrin tetramer, or ␤-spectrin construct encompassing residue 1-1388. The incubation was performed in a 100-l volume for 2 h at 4°C in a buffer containing 30 mM HEPES, 50 mM KCl, 2 mM MgCl 2 , 1 mM NaEGTA, 10% sucrose, 0.05% Tween-20, 2 mg/ml bovine serum albumin, 0.5 mM ATP, 0.2 mM DTT, pH 7.0. Actin filaments and associated proteins were collected by layering samples onto equal volume of 20% sucrose dissolved in the same incubation buffer in Ti 42.2 centrifuge tubes and centrifuging for 1 h at 25,000 rpm. Supernatant and pellet were separated and analyzed for 125 I radiation in a ␥ counter. Data come from duplicate determinations and are given as mean Ϯ S.D. The sedimentation of adducin without actin (0.10 nmol/mol actin) has been subtracted as background. F-actin concentrations represent total actin in the assay and are not corrected for monomeric actin. S.D. is the sum of standard derivation of duplicate data determination and that of duplicate background determinations. B, the data of A are expressed as a double-reciprocal plot of 1/spectrin-dependent binding of 125 I-labeled adducin to F-actin versus 1/free spectrin construct. Spectrin-dependent binding of 125 I-labeled adducin is calculated as total binding of 125 Ilabeled adducin to actin (mean of duplicate determinations in A) corrected for the amount of adducin that sedimented with actin in the absence of spectrin (0.47 nmol/mol actin). The concentration of spectrin required for half-maximal activation of adducin binding equals the reciprocal of x axis intercept.
FIG. 2. Schematic model of ␤-spectrin constructs and Coomassie Blue-stained SDS gels of the purified ␤-spectrin construct proteins. The region marked N refers to the N-terminal residues 1-313. The region marked C refers to C-terminal about 200 residues. The ABD region within 1-313 is the actin binding domain representing residues 47-186 (Karinch et al., 1990). The numbers in the model, from 1 to 17, refer to repeat domain numbers of ␤-spectrin. The name of each construct is designated by the number of amino acid residues that the construct encompasses. The constructs containing different domains were expressed and purified as described under "Experimental Procedures." Purified proteins (1.5 g/lane) were electrophoresed on SDSpolyacrylamide gel and stained with Coomassie Blue. change in activity. A construct containing the first and second ␣-helical domains but lacking the N-terminal domain exhibited no detectable activity in promoting binding of adducin to F-actin.
These results indicate that the ␤-spectrin N-terminal domain plus the first two ␣-helical domains are required for full activity in forming a spectrin/adducin/actin complex, but the N-terminal domain alone or combined with the first ␣-helical domain still is about 50% as active (Fig. 5). The differences in adducin-dependent activities of ␤-spectrin constructs could be dependent on association with F-actin and/or on direct interaction with adducin. To determine which possibility is the basis for the 2-fold difference, it was necessary to evaluate actin binding affinity of these constructs.
Effect of ␣-Helical Domains on Actin Binding Activity of the ␤-spectrin N-terminal Domain-The N-terminal domain alone or combined with the first ␣-helical domain were labeled and assayed for activity in co-sedimenting with F-actin (Fig. 6A). Since spectrin and actin alone associate with a relatively low K d in the micromolar range, 10 -20 M concentrations of spectrin polypeptides were required to approach saturation. Such concentrations of spectrin are not obtainable with native spectrin due to solubility problems and due to formation of bundles of actin filaments, but could be achieved with recombinant spectrin polypeptides. Double-reciprocal plots of the binding data for the two constructs revealed differences in capacity and affinity for actin (Fig. 6). The N-terminal domain alone associated with F-actin with a K d of 26 M, and a capacity of 1 molecule/1-2 F-actin subunits. The fact that the capacity for spectrin domain was less than the theoretical 1:1 ratio could be due to negative cooperativity at high occupancy and/or partial denaturation of F-actin. The possibility that recombinant spectrin polypeptides altered the actual amount of polymerized actin is ruled out by a parallel assay using 125 I-labeled actin incubated with different concentration of spectrin constructs. The results showed no effect of spectrin construct on the actual amount of F-actin recovered in the pellets under assay conditions (data not shown). The N-terminal domain plus the first ␣-helical domain exhibited a 4-fold higher affinity but a 3-fold lower capacity. The difference in affinity and capacity for this construct were confirmed in four separate experiments, including different preparations of F-actin and of spectrin polypeptides. Thus, while the absolute values for capacity were not FIG. 3. ␤-Spectrin constructs 1-1388, 1-777, 1-643, and 1-528 have the same effect on promoting 125 I-labeled adducin binding to F-actin. 125 I-Labeled adducin (5 nM, 127,000 cpm/pmol) was incubated with F-actin (2.0 M) in the presence of increasing concentration of ␤-spectrin 1-1388 (Ⅺ), 1-777 (å), 1-643 (Ç), or 1-528 (E). The co-sedimentation assay is described in Fig. 1. The data are expressed as the means of duplicate determinations. The difference between each pair of duplicate determinations is less than 5%. The adducin cosedimented with actin alone (0.38 nmol/mol actin) and adducin sedimented without actin (0.11 nmol/mol actin) have been subtracted.

FIG. 5. Concentrations of ␤-spectrin constructs required for half-maximal activation of binding of 125 I-labeled adducin to F-actin.
Values for the concentrations required for half-maximal activation were determined by the x axis intercept of the double-reciprocal plot of 1/spectrin construct-dependent binding of 125 I-labeled adducin to F-actin versus 1/spectrin construct. The data were collected from the same assays as noted in Figs. 3 and 4. Each value reported is the result of one independent binding assay. rigorously determined, the difference in capacity between constructs was a consistent finding.
Although addition of the second ␣-helical domain to the N-terminal domain plus the first ␣-helical domain increased the concentration required for half-maximal activation of adducin binding 2-fold, this addition had no measurable effect on affinity for actin (Fig. 7). Moreover, further addition of up to eight ␣-helical domains caused little change in the affinity for actin. ␣-helical domain one alone or combined with domain two in the absence of the N-terminal domain exhibited no measurable binding to actin.
These results support the following conclusions. The N-terminal domain, as expected, is essential for primary binding between ␤-spectrin and actin. Also, the N-terminal domain contains a high affinity site for adducin. However, the first ␣-helical domain has an unexpected interaction with actin, since addition of this domain increased the affinity of the Nterminal domain for actin by 4-fold and reduced the capacity by 3-fold with little change in the concentration required for halfmaximal activation of adducin binding. Increased affinity due to spectrin dimerization rather than interaction with actin is unlikely based on the finding of reduced capacity, as well as lack of positive cooperativity and lack of evidence for ␤-spectrin self-association. The second ␣-helical domain contributes to interaction of ␤-spectrin with adducin independent of actin, since addition of the second domain adds little to actin affinity, but does increase stability of adducin/spectrin/actin complexes.
Effect of ␣-Spectrin on Association of ␤-Spectrin with Adducin/Actin Complexes-The N-terminal domain of ␤-spectrin is closely aligned with the C-terminal domain of ␣-spectrin to form lateral interchain binding between spectrin subunits (Speicher et al., 1992;Viel and Branton, 1994). ␣-Spectrin could therefore affect the association of ␤-spectrin with adducin and actin. ␤-Spectrin residues 1-1388 (N-terminal domain plus 10 ␣-helical domains) is equivalent to native ␤-spectrin in terms of forming adducin/actin complexes (Fig. 1) and contains the sites for association with ␣-spectrin (Speicher et al., 1992;Viel and Branton, 1994). ␤-Spectrin 1-1388 lacks the site required for end-end association with the N-terminal domain of ␣-spectrin, and therefore will form laterally associated heterodimers with ␣-spectrin but will not form end-end tetramers. ␤-Spectrin 1-1388 was therefore used to determine effects of ␣-spectrin on adducin/actin interactions (Fig. 8).
Effect of Adducin on Binding of ␣-Actinin to F-actin-The N-terminal and first two ␣-helical domains of ␤-spectrin exhibit sequence similarity to ␣-actinin (Byers et al., 1989). It was of interest to determine if ␣-actinin also shared ability to interact with adducin. The effect of adducin on binding of chicken smooth muscle ␣-actinin and spectrin to F-actin were compared in Fig. 9. Labeled ␣-actinin associated with F-actin in the absence of adducin, indicating that ␣-actinin was active (data not shown). However, adducin had no effect on binding of ␣-actinin to F-actin even though adducin did promote binding of spectrin with half-maximal stimulation at 40 nM adducin. FIG. 6. Actin binding activity of ␤-spectrin constructs. A, association of 125 I-labeled ␤-spectrin 1-313 (E) or 1-422 (å) to F-actin. 125 I-Labeled ␤-spectrin constructs were mixed with unlabeled construct protein to reach the specific activities (8,900 cpm/mol for 1-313 and 15,000 cpm/mol for 1-422). Serially diluted construct proteins were incubated with actin (3.0 M) in a 60-l volume. The incubation buffer and the following sedimentation steps are the same as described in Fig.  1. Data are represented as spectrin bound to actin (mol/mol F-actin), where spectrin sedimented without actin has been measured for individual spectrin concentrations and has been subtracted. The F-actin is determined by using 125 I-labeled F-actin in a parallel assay to measure the actual amount of sedimented actin as F-actin concentration (data not shown). The experiment is the representative of three others with similar results. B, the data of A are expressed as a double-reciprocal plot. The K d between spectrin construct and F-actin is determined by the reciprocal of x axis intercept of double-reciprocal plot. The binding capacity of construct to F-actin is determined by the reciprocal of y axis intercept.
FIG. 7. K d for association between ␤-spectrin constructs and F-actin. The same methods delineated in Fig. 6 are followed to determine the K d of several construct proteins. Each value reported is the result of one independent binding assay.
Adducin therefore is selective for spectrin/actin complexes and is not likely to stabilize actin interactions of ␣-actinin. Other proteins with an actin-binding domain share even less similarity with ␤-spectrin residues 1-528 than ␣-actinin. Thus dystrophin and other members of the family are not likely to have actin binding activity promoted by adducin either, although further experiments are required to demonstrate this issue.

DISCUSSION
This study presents quantitative evidence that the ␤-spectrin N-terminal domain plus the first two ␣-helical domains are responsible for participation of spectrin in spectrin/adducin/ actin complexes formed in in vitro assays. ␣-Spectrin exhibited no detectable activity either alone or following association with ␤-spectrin 1-1388. The primary activity of ␤-spectrin in association with actin and adducin/actin complexes resides in the N-terminal domain. However, the first ␣-helical domain also is involved in association with actin and increases the K d of the N-terminal domain for F-actin by 4-fold. The second ␣-helical domain of ␤-spectrin contributed little to actin-binding but did increase by 2-fold the activity of the N-terminal domain plus first ␣-helical domain in stabilizing spectrin/adducin/actin complexes. Although the region of ␤-spectrin required for adducin/ actin binding activity is homologous to the N terminus of ␣-actinin, ␣-actinin is not affected by adducin in actin-binding assay. Thus the interactions between spectrin and adducin are specific for spectrin and are not likely shared by other members of the spectrin/␣-actinin/dystrophin family of actin-binding proteins.
Involvement of ␤-spectrin and its actin-binding N-terminal domain in an actin-related complex was anticipated in view of studies with spectrin (Karinch et al., 1990) and other members of related actin-binding proteins (Mimura and Asano, 1987;Fabbrizio et al., 1993;Hemmings et al., 1992;Lebart et al., 1993;Levine et al., 1992;McGough et al., 1994). However, a role for the first two ␣-helical domains of ␤-spectrin in interactions with actin and adducin has not been reported previously, and has interesting implications for arrangement of components of spectrin/actin and spectrin/adducin/actin complexes (see below).
The use of monomeric forms of ␤-spectrin permits measurement of the actin binding affinity without complication of actin cross-linking activities. Our result shows that ␤-spectrin Nterminal residue 1-422 has full actin binding affinity with a K d ϭ 6 M, which is the first documented affinity for ␤-spectrin binding to actin. Consistent with this value are the actin binding affinities measured for filamin N-terminal 190-kDa proteolytic fragment (K d ϭ 3 M) (Gorlin et al., 1990), ␣-actinin N-terminal residue 1-269 (K d ϭ 4 M) (Way et al., 1992b), and dystrophin N-terminal residues 1-431 or smaller peptides (K d ϭ 1-5 M) (Fabbrizio et al., 1993). Different binding assay systems might cause variations, such as the dystrophin actin binding affinity ranging from 0.1 M (Corrado et al., 1994) to 44 M (Way et al., 1992 a). ␤-Spectrin residues 1-313 have a low actin binding affinity K d ϭ 26 M, comparable to the dystrophin-related protein, utrophin, which has a N-terminal 261amino acid actin-binding domain that associates with actin with a K d ϭ 19 M (Winder et al., 1995). These results combined with the present study indicate that the full actin binding affinity of this family of proteins is in the micromolar range and the domains outside the homologous actin-binding domain may be required to attain maximal activity.
The finding that addition of the first ␣-helical domain both enhances affinity of the N-terminal domain and reduces the capacity of actin filaments suggests a model for spectrin/actin complexes shown in Fig. 10. The dimensions of most of the protein subunits and segments have been measured by a variety of approaches and these dimensions are drawn in scale. According to this scheme, the N-terminal domain of spectrin binds to the interface between two actin subunits, as described for the homologous domain of ␣-actinin . In addition, optimal binding also involves contact between the first ␣-helical domain and at least one additional actin subunit. The ␣-helical domain thus stabilizes a spectrin-actin complex but also inhibits binding of immediately adjacent spectrin molecules through steric hindrance. The first ␣-helical domain has no detectable independent actin-binding site, suggesting two possibilities. Either the interaction with actin is too weak to be measured experimentally (i.e. a K d greater than 100 M), or a FIG. 8. Effect of ␣-spectrin on activity of ␤-spectrin (1-1388) in promoting binding of 125 I-labeled adducin to F-actin. The dimer of ␣-spectrin/␤-spectrin 1-1388 (q ) was formed and purified as described under "Experimental Procedures." ␣-Spectrin (65 g, ç), ␤-spectrin residues 1-1388 (40 g, f) were both denatured and then renatured to establish parallel conditions with dimer preparation. The adducin promoting activity of ␣-spectrin, ␤-spectrin residues 1-1388, or dimer ␣/␤ 1-1388 was assayed as noted in Fig. 1. The adducin binding to actin alone (0.40 nmol/mol actin) has been subtracted from spectrindependent adducin binding to F-actin.
FIG. 9. Comparison of the adducin effects on promoting spectrin and ␣-actinin binding to F-actin. 125 I-Labeled spectrin (50 nM, 38,500 cpm/pmol, f) or 125 I-labeled ␣-actinin (50 nM, 23,200 cpm/pmol, ç) was incubated with F-actin (5 M) and increasing concentrations of adducin in a 80 l volume at 24°C for 1 h. The incubation buffer and the following sedimentation steps are the same as described in Fig. 1. Adducin dependent ligand (spectrin or ␣-actinin) binding is expressed as the ligand associated to F-actin in the presence of adducin subtracted by the ligand associated to F-actin alone. The data express the mean of duplicate. The difference between each pair of duplicate determinations is less than 5%. Spectrin binding to actin alone (2.43 nmol/mol actin) and ␣-actinin binding to actin alone (0.10 nmol/mol actin) have been subtracted. secondary actin binding site is created after the N-terminal domain binds to actin and causes conformational change in actin and/or spectrin.
Adducin participates in spectrin/actin complexes through extended tail domains 300 residues in length, which are predominantly a random-coil but associate with each other (Hughes and Bennett, 1995). The length of the tail domains is not known precisely, but easily could extend the distance of 140 Å predicted for the length of ␤-spectrin N-terminal domain plus two ␣-helical domains. Consideration of the configuration of the adducin tail and involvement of ␣-helical domains in spectrin/ actin/adducin complexes suggest a model where adducin tail domains in association with ␤-spectrin extend along both front and back of the long axis of actin filaments (Fig. 10). An appealing feature of this model is that it immediately suggests how adducin could recruit one and possibly more spectrin molecules to the complex, as predicted from binding assays (Gardner and Bennett, 1987;Bennett et al., 1988). Adducin has recently been discovered to cap the fast-growing end of actin filaments in an interaction that requires both the head domain and the spectrin/actin-binding tail domain . This observation implies that adducin tails extend from the fast growing end of actin filament. It will be of interest in future experiments to evaluate possible overlap in binding sites on actin filaments between tropomyosin and adducin tails, and to determine the placement of other proteins such as dematin and tropomodulin, which are located at spectrin-actin junctions.
The predicted lateral association of spectrin along actin filaments has not been noted previously in electron micrographs of membrane skeletons, but would be difficult to resolve since the length of 140 Å at both ends of spectrin amounts to only about 15% of the extended length of a spectrin tetramer of 2000 Å. It is of interest that rotary-shadowed images of spectrin reveal a reproducible kink about 300 Å from each end of the tetramer (Bennett et al., 1982), which could be relevant to participation of the first portion of spectrin in lateral association with actin filaments.
␤-Spectrin residues 1-528 exhibits sequence homology to the N terminus of ␣-actinin with 45% identity in residues 1-313, and 38% identity in the first two ␣-helical domains. However, ␣-actinin does not interact with adducin ( Fig. 9) and could be viewed as a naturally occurring mutant of ␤-spectrin, which lacks adducin binding activity but retains native folding. Comparison of the ␣-helical domains of ␤-spectrin and ␣-actinin indicates that the several long non-conserved regions are the predicted coil regions between helix A and B in both the first and the second ␣-helical domains (both are 9 -10 amino acid stretches without any similarities) as well as the end of the helix A of the second domain. These differences in primary structure are candidates to explain the functional differences between spectrin and ␣-actinin with respect to adducin. It will be of interest in future experiments to design mutations targeting these potential ternary binding sites to study their functional consequence in the formation of spectrin/actin/adducin complexes in cells.
␤-Spectrin residues 1-528 can be used in future experiments as a monomeric form of native spectrin with full activity in association with actin and adducin, but which avoids complications due to actin filament cross-linking activity and limiting solubility. This new reagent makes it possible to fully saturate actin filaments with spectrin and adducin presumably in a homogeneous orientation. Therefore a high resolution structure of spectrin/adducin/actin complex can, in principle, be solved by image analysis of decorated actin filaments as has been reported for ␣-actinin .