Adducin Preferentially Recruits Spectrin to the Fast Growing Ends of Actin Filaments in a Complex Requiring the MARCKS-related Domain and a Newly Defined Oligomerization Domain*

Adducin is a protein associated with spectrin and actin in membrane skeletons of erythrocytes and possibly other cells. Adducin has activities in in vitro assays of association with the sides of actin filaments, capping the fast growing ends of actin filaments, and recruiting spectrin to actin filaments. This study presents evidence that adducin exhibits a preference for the fast growing ends of actin filaments for recruiting spectrin to actin and for direct association with actin. β-Adducin-(335–726) promoted recruitment of spectrin to gelsolin-sensitive sites at fast growing ends of actin filaments with half-maximal activity at 15 nmand to gelsolin-insensitive sites with half-maximal activity at 75 nm. β-Adducin-(335–726) also exhibited a preference for actin filament ends in direct binding assays; the half-maximal concentration for binding of adducin to gelsolin-sensitive sites at filament ends was 60 nm, and the K d for binding to lateral sites was 1.5 μm. The concentration of β-adducin-(335–726) of 60 nm required for half-maximal binding to filament ends is in the same range as the concentration of 150 nm required for half-maximal actin capping activity. All interactions of adducin with actin require the myristoylated alanine-rich protein kinase C substrate-related domain as well as a newly defined oligomerization site localized in the neck domain of adducin. Surprisingly, the head domain of adducin is not required for spectrin-actin interactions, although it could play a role in forming tetramers. The relative activities of adducin imply that an important role of adducin in cells is to form a complex with the fast growing ends of actin filaments that recruits spectrin and prevents addition or loss of actin subunits.

Adducin is a protein associated with spectrin and actin in membrane skeletons of erythrocytes and possibly other cells. Adducin has activities in in vitro assays of association with the sides of actin filaments, capping the fast growing ends of actin filaments, and recruiting spectrin to actin filaments. This study presents evidence that adducin exhibits a preference for the fast growing ends of actin filaments for recruiting spectrin to actin and for direct association with actin. ␤-Adducin-(335-726) promoted recruitment of spectrin to gelsolin-sensitive sites at fast growing ends of actin filaments with half-maximal activity at 15 nM and to gelsolin-insensitive sites with half-maximal activity at 75 nM. ␤-Adducin-(335-726) also exhibited a preference for actin filament ends in direct binding assays; the half-maximal concentration for binding of adducin to gelsolin-sensitive sites at filament ends was 60 nM, and the K d for binding to lateral sites was 1.5 M. The concentration of ␤-adducin-(335-726) of 60 nM required for half-maximal binding to filament ends is in the same range as the concentration of 150 nM required for half-maximal actin capping activity. All interactions of adducin with actin require the myristoylated alanine-rich protein kinase C substraterelated domain as well as a newly defined oligomerization site localized in the neck domain of adducin. Surprisingly, the head domain of adducin is not required for spectrin-actin interactions, although it could play a role in forming tetramers. The relative activities of adducin imply that an important role of adducin in cells is to form a complex with the fast growing ends of actin filaments that recruits spectrin and prevents addition or loss of actin subunits.
The ubiquitous expression of spectrin and spectrin-associated proteins in most cells of metazoan organisms raises the issue of how spectrin-based structures are assembled and regulated. The erythrocyte membrane skeleton is the best characterized example of a spectrin-actin network, although spectrin also is associated with membranes at sites of cell-cell contact in epithelial cells, along axons in the nervous system, and at specialized sites in striated muscle (1). Spectrin in erythrocytes is organized into a polygonal network formed by 5-7 spectrin molecules linked to short actin filaments about 40 nm in length (2)(3)(4). Spectrin-actin junctions contain a group of proteins that presumably have functions in promoting spectrin-actin inter-actions, forming connections with the membrane, and regulation of actin filament length (1,5).
Adducin is one of the proteins localized at spectrin-actin junctions (6) and was originally purified based on calmodulin binding activity (7). Adducin also is a substrate for protein kinase C and protein kinase A (8 -10). Adducin phosphorylated at the major protein kinase C site is localized in dendritic spines and axonal growth cones of cultured neurons, indicating adducin is a protein kinase C substrate in vivo (11). Adducin is encoded by three closely related genes termed ␣, ␤, and ␥ adducins which are expressed in many types of cells (7,9,12). Adducins all contain an N-terminal globular head domain, a neck domain, and a protease-sensitive tail domain with a Cterminal basic stretch of 22 amino acids with homology to the MARCKS 1 protein (9,12,13). The MARCKS-related domain of adducin contains the major sites for both protein kinase C phosphorylation and calmodulin binding (10). Adducin is a mixture of heterodimers and heterotetramers in solution, with tetramers formed by four head domains in contact with one another to form a globular core, and interacting tail domains extending away from the core (14). Adducin oligomers in erythrocytes comprise ␣/␤ subunits and in other cells include ␣/␥ as well as ␣/␤ combinations of subunits (7,9).
The patterns of expression and cellular localization of adducin are consistent with a role in interaction with spectrin in erythrocytes as well as other cells. Adducin is localized at spectrin-actin junctions in mature erythrocytes (6) and is expressed early in erythropoiesis at a time when the spectrinactin network is forming (15). Adducin and spectrin are both concentrated at sites of cell-cell contact in epithelial tissues and at dendritic spines and axon growth cones of cultured neurons (11,16). Expression of a dominant-negative form of spectrin that inhibits assembly of spectrin tetramers results in loss of epithelial polarity and disassociation of adducin from the plasma membrane (17). Hts, an adducin-related protein in Drosophila, is required for the formation of ring canals, an actin-rich structure, and is co-localized with spectrin in developing germline cells (18,19). Spectrin and adducin also have been observed associated with a dynactin complex which contains an actin-related protein (20).
Adducin activities determined in in vitro assays include recruiting spectrin to actin filaments (21,22), bundling actin filaments (23,24), and capping the fast growing ends of actin filaments (25). Each of these individual adducin activities may reflect different aspects of adducin function in living cells. In this study we have quantitated the relative affinities of adducin for actin capping, direct binding to actin, as well as recruiting spectrin to the sides and fast growing ends of actin filaments. The conclusion of these experiments is that adducin preferentially associates with and recruits spectrin to the fast growing ends of actin filaments. We also demonstrate that adducin interactions with spectrin and actin require the MARCKS-related domain of adducin as well as a newly defined oligomerization site localized in the neck domain of adducin.

Methods
Procedures-Protein concentration was determined by the procedure of Bradford (26). SDS-polyacrylamide gel electrophoresis was performed using 0.2% SDS with buffers of Fairbanks et al. (27) on 1.5-mm thick 3.5-17% exponential gradient gels. Quantitation of proteins by the pyridine dye elution method was basically as described (28). Briefly, the bands corresponding to the targeted proteins were cut out from the Coomassie Blue-stained SDS-polyacrylamide gel and the dyes were extracted in 25% pyridine. The relative amounts of protein were measured by reading the absorbance at 595 nm of dye/pyridine solution.
Protein Purification-Actin was purified from acetone powder of rabbit skeletal muscle (29) with a modification that actin monomer was isolated by gel filtration chromatography on a Superose 12 column before the final polymerization step. Bovine brain spectrin was isolated following high salt extraction from brain membranes (30). Red blood cell adducin was purified from the low salt extract of human erythrocyte membranes as described previously (14). Purified human plasma gelsolin was purchased from Cytoskeleton. Purified recombinant human plasma gelsolin expressed from Escherichia coli was a generous gift from Dr. Paul A. Janmey, Harvard Medical School, Boston. ␤-Adducin constructs were generated using full-length human ␤-adducin cDNA (12) as a polymerase chain reaction template and were subcloned into a Studier plasmid with a T7 promoter (31). Each adducin polypeptide has three additional N-terminal amino acids (Met-Ala-Ser). Procedures for subcloning and bacterial expression of the constructs were as described (32). The expressed proteins were purified as follows. Bacterial pellets were resuspended in 100 ml of 50 mM sodium phosphate (pH 7.4), 1 mM EGTA, 25% sucrose, 10 mM MgCl 2 , 0.04 mg/ml DNase I, and protease inhibitors including 1 mM 4-(2-aminoethyl)-benzenesulfonyl fluoride, 10 mM benzamidine, 0.01 mg/ml pepstatin A. The suspension was mixed with 200 ml of extraction buffer containing 200 mM NaCl, 20 mM sodium phosphate (pH 7.4), 2 mM EDTA, 1% Triton X-100, 1 mM DTT, and protease inhibitors as mentioned above. Subsequently the lysate was forced through a 20-gauge needle once and centrifuged for 20 min at 5,000 ϫ g. The recombinant proteins were in the Triton-soluble fraction, which was then loaded onto a 15-ml S-Sepharose column. The protein was eluted with 0.5 M NaCl in 10 mM sodium phosphate (pH 7.4), 1 mM EDTA, 1 mM NaN 3 , 0.05% Tween 20, 1 mM DTT. Optimal column fractions were pooled and dialyzed against the same elution buffer but with 50 mM NaCl. The protein was loaded onto a 10-ml Mono Q high pressure liquid chromatography column and eluted by a linear 0.05-0.5 M NaCl gradient. Generally, a 2-liter culture can result in 5-10 mg of protein with over 90% purity.
Preparation of Immobilized Actin-Purified F-actin was incubated with EZ-Link TM Biotin-BMCC for 5 h with a 5:1 molar ratio of biotin versus actin. Then the biotinylated F-actin was depolymerized by dialysis for 48 h against G buffer containing 2 mM Tris (pH 8.0), 0.2 mM CaCl 2 , 0.2 mM ATP, 0.5 mM DTT. Avidin beads were prepared as follows. NeutrAvidin was coupled to epoxide-beads for 48 h in a coupling buffer containing 10 mM HEPES (pH 7.0), 1 M NaCl, 1 mM sodium EGTA, 1 mM NaN 3 . The reaction was quenched by a buffer containing 100 mM Tris (pH 8.0), 1 M NaCl, 1 mM EGTA, 1 mM NaN 3 and then the beads were washed twice with coupling buffer. Subsequently the biotinylated and depolymerized actin was incubated with avidin beads at 4°C for 5 h. The amount of actin bound to the beads (usually 4 -9 g per l of packed beads) was monitored by measuring free actin before and after the coupling using Bradford assays. The immobilized actin were washed twice by a buffer containing 2 mM Tris (pH 8.0), 50 mM KCl, 2 mM MgCl 2 , 1 mM ATP, 0.5 mM DTT. In the final step, beads coupled actin and un-biotinylated actin monomer (with a ratio of 1:9) were co-polymerized in 2 mM Tris (pH 8.0), 50 mM KCl, 2 mM MgCl 2 , 1 mM ATP, 0.5 mM dithiothreitol.
Actin Co-sedimentation Assay-The co-sedimentation assay using free actin filaments was described previously (14,22). The co-sedimentation assay using immobilized actin was performed as follows. 125 I-Labeled ligand (adducin or spectrin) was incubated with immobilized actin (0.4 M actin) in a 60-l volume for 1 h at 4°C in a buffer containing 30 mM HEPES (pH 7.0), 50 mM KCl, 2 mM MgCl 2 , 1 mM EGTA, 10% sucrose, 0.05% Tween 20, 2 mg/ml bovine serum albumin, 0.5 mM ATP, 0.2 mM DTT, 0.5 mM NaN 3 . The incubated mixtures were then layered onto 200 l of 20% sucrose dissolved in the same incubation buffer in 400-l microcentrifuge tubes and centrifuged for 10 min at 4,000 ϫ g. Supernatants and pellets were separated by freezing the tubes on dry ice and cutting the tips of the tubes. The radioactivity of both the supernatant and the pellets was analyzed with a gamma counter. The amount of actin in pellets was determined in parallel experiments under the same conditions.
Actin Polymerization Assay-Pyrene-labeled actin was prepared as described (33). G buffer (2 mM Tris (pH 8.0), 0.2 mM CaCl 2 , 0.2 mM ATP, 0.5 mM DTT) was used in all assays. The assay quantitated inhibition of actin polymerization at barbed ends utilizing the method of Pollard (34) in which rapid polymerization was initiated using F-actin nuclei. Briefly, 4 M G-actin was mixed with adducin, and actin polymerization was initiated by adding 0.25 volume of 1.25 M F-actin nuclei in 10 mM Tris (pH 8.0), 250 mM KCl, 5 mM MgCl 2 . Actin polymerization was followed by monitoring the increased pyrene fluorescence of labeled F-actin (excitation at 365 nm and emission of 407 nm) during 30 -210 s after initiation of the polymerization using a spectrofluorimeter. The sample temperature was maintained at a constant 25°C for all experiments using a circulating water bath.
Electron Microscopy-Immobilized actin or free actin in a buffer containing 30 mM HEPES (pH 7.0), 50 mM KCl, 2 mM MgCl 2 , 1 mM EGTA, 0.05% Tween 20, 0.5 mM ATP, 0.5 mM NaN 3 was applied to a 400-mesh copper grid coated with carbon film for 30 s and washed by 10 -20 drops of 1% uranyl acetate. The grid was then air-dried for 5 min and examined by a Philips EM 301 electron microscope.

Use of Immobilized Actin to Study Adducin/Actin/Spectrin
Interactions-Adducin/actin/spectrin interactions were previously monitored by a binding assay using actin filaments as the sedimentation matrices (14,21,22,32). Recent studies showed that adducin caps the fast growing ends of actin filaments (25). In order to evaluate directly interactions of adducin with actin filament ends, we developed a technique for measuring binding to short actin filaments with relatively more ends than in the conventional assay. Biotinylated actin was coupled to avidin beads (referred to as "immobilized actin"). Beads were selected to be (a) impermeable to proteins to minimize the background due to trapping, (b) small (0.35 m) to maximize surface/volume ratio, and (c) of appropriate density (1.1 g/ml) to sediment through a sucrose barrier. Immobilized actin sediments at 4,000 ϫ g for 10 min, and free actin requires high centrifugation forces (up to 80,000 ϫ g for 60 min) to pellet. Compared with the conventional assay using free actin, the immobilized actin assay can reduce the background that results from sedimentation of ligands independent of actin.
The filament lengths of immobilized actin were compared with those of uncoupled actin filaments (referred to as "free actin" in this article). Fields of negatively stained immobilized actin contain much fewer extended actin filaments than a sample of free actin, even though both preparations contain equal concentrations of actin (Fig. 1A). It suggests that immobilized actin has more short filaments which are coupled to the beads. Moreover, the number of actin filament ends are determined quantitatively by using gelsolin to cap the fast growing ends of actin filaments (35). Actin/gelsolin co-sedimentation was performed in a buffer at pH 7.0 and calcium-free to minimize actin filament severing activity of gelsolin (36,37). Gelsolin does not bind to beads alone but binds to immobilized actin with a ratio of approximately 1 gelsolin/15 actin monomers (Fig. 1C, lanes 3 and 4). In contrast, gelsolin binds to free actin with a ratio below 1 gelsolin/250 actin subunits (Fig. 1C, lanes 1 and 2), suggesting much fewer filaments ends available in free actin samples.
The immobilized actin assay produces results consistent with those of free actin assay but with a dramatically high signal expressed in terms of signal per actin subunit (Fig. 2). Spectrin recruiting activity of adducin was half-maximal at 50 nM adducin for both free actin and the immobilized actin cosedimentation assay. This affinity for spectrin recruitment of adducin is also comparable with values measured previously (14). However, adducin in the immobilized actin assay exhibited a 15-20 times higher signal per actin subunit than the free actin assay. The results support our design of immobilized actin as a tool to study interactions that are related to the ends of actin filaments.
Adducin Polypeptides Containing Only Neck and Tail Domains Are Functionally Equivalent to Native Adducin in Spectrin Recruiting and Actin Capping Activities-Contributions of individual domains to adducin function were evaluated using recombinant polypeptides expressed in bacteria (Fig. 3). We focused on ␤-adducin as an initial step and have confirmed the major findings with ␣-adducin. The domain boundaries of adducin subunits were inferred from patterns of polypeptides produced by limited proteolytic digestion of erythrocyte and brain adducin (13,22). Mild chymotryptic digestion of adducin yields an N-terminal fragment of 48 kDa (residues 1-436), whereas tryptic digestion produces an N-terminal fragment of 39 kDa (residues 1-354), implying a chymotrypsin-resistant domain of 80 residues (355-435) which is referred to as the neck domain. Twenty additional N-terminal residues were added to constructs encompassing the neck domain to compensate for possible errors in estimates of molecular weight by SDS-electrophoresis (Fig. 3). The tail construct (residues 409 -726) designed previously by Hughes and Bennett (14) lacks most of the neck domain. Thus ␤-adducin residues 335-436 and 437-726 are designated to be the neck and the tail domains, respectively, and were expressed in bacteria with a three amino acid addition at the N terminus but otherwise no additional residues (see "Methods"). The head domain (residues 1-354) was not included due to the limited solubility when expressed in bacteria (data not shown).
Activities of ␤-adducin polypeptides were compared with native erythrocyte adducin in recruitment of spectrin to actin filaments ( Fig. 4) and preventing polymerization of actin ( Fig.  5). ␤-Adducin-(335-726) and native erythrocyte adducin exhibited an equivalent binding extent and binding affinity (halfmaximal at 50 nM) of spectrin recruiting activity expressed per adducin polypeptide chain (Fig. 4). The capping activities of adducin domains were studied by the pyrene actin polymerization assay (Fig. 5). ␤-Adducin-(335-726) blocked 70 -80% of the actin polymerization, an extent similar with that of native erythrocyte adducin. However, the estimated K cap for ␤-335-726 is 150 nM, and the K cap of erythrocyte adducin is about 80 nM. Both values are within the range of K cap ϭ 100 nM (calculated for adducin dimers) of erythrocyte adducin in the previous report (25).
These results demonstrate that activities of adducin with regard to spectrin recruiting and actin capping are located in neck-tail domains of a single subunit without apparent requirement for the head domain or heterodimers/tetramers. The neck-tail of ␣-adducin, the counterpart of ␤-adducin-(335-726), exhibited similar spectrin recruiting and actin capping activities (not shown). An experimental benefit of activity of the neck-tail polypeptides is that these recombinant proteins could be employed as an equivalent of native adducin for further analysis of adducin structure and function.
The MARCKS-related Domain of Adducin Is Necessary but Not Sufficient for Interactions with Spectrin and Actin-␤- Adducin-(335-694), which lacks the MARCKS-related domain, almost completely lacked activity in promoting association of spectrin with actin filaments (Fig. 4). ␤-Adducin-(335-694) also exhibited an 80% reduction in the extent of actin capping activity (Fig. 5) and in actin binding activity (Fig. 6A). The residual activities of ␤-adducin-(335-694) in actin capping and actin binding assays are in the range of nonspecific effects. The MARCKS-related domain is not sufficient for activities, however, since a construct, ␤-adducin residues 437-726, with the MARCKS-related domain and the rest of the tail, but lacking the neck domain, also exhibited minimal spectrin recruiting activity, actin capping activity, and actin binding activity (Figs. 4 -6). These experiments are the first evidence that the MARCKS-related region is necessary (but not sufficient) for adducin functions. A role for this domain is consistent with the previous findings that calmodulin, which binds to the MARCKS-related domain (10), inhibits spectrin recruitment (21) and actin capping (25). Moreover, phosphorylation of the MARCKS-related domain by protein kinase C also inhibits actin capping and spectrin recruiting activities (11). The MARCKS-related domain is highly conserved in all adducin subunits, so these results suggest a general principle that this domain is essential for adducin/actin interactions.
The  . 4) or actin capping activity (Fig. 5). These findings suggest that the neck domain is necessary but not sufficient in itself for adducin interactions with spectrin and actin. Adducin is known to associate into dimers and tetramers (7,14,22), suggesting the possibility that one role of the neck domain could be in formation of oligomers. A possible correlation between adducin domains and oligomerization was evaluated by chemical crosslinking experiments. Fig. 7A shows that polypeptides containing the neck plus tail and neck alone were cross-linked into oligomers. The major form of oligomer is dimer, but trimer, tetramer, and higher oligomers also are evident. In contrast, the construct lacking the neck region, ␤-adducin-(437-726), was not cross-linked and remained a monomer. The cross-FIG. 2. Adducin-dependent recruitment of spectrin to free actin or immobilized actin. 125 I-Labeled spectrin (1 nM, 284,000 cpm/pmol) and increasing concentrations of erythrocyte adducin were incubated with free actin (2.0 M) or immobilized actin (0.4 M) at 4°C for 1 h. The assay procedure was described under "Experimental Procedures." In the assay with free actin, the amount of actin in the final pellet (67.2 pmol/tube) was measured in a parallel experiment under the same conditions. The sedimentation of spectrin in the absence of adducin (about 15% of total spectrin) was subtracted as background. The inset shows the same binding curve determined with free actin with an enlarged scale of y axis. In the assay with immobilized actin, the actin in the pellet (8.3 pmol/tube) was determined in a parallel assay. The sedimentation of spectrin in the absence of immobilized actin (less than 1% total spectrin) was determined by the spectrin bound to equal amount of beads lacking actin and was subtracted as background. All the data points are averages of triplicate measurements. linking reaction reflects specific protein-protein interactions because the majority of the cross-linked products were abolished in the presence of 6 M urea. It is noteworthy that native erythrocyte adducin forms tetramers as the major cross-linked products (Fig. 7A), suggesting that additional oligomerization sites must exist which further connect ␣and ␤-adducin dimers into hetero-oligomers (Fig. 7A).
The oligomeric state of the neck domain of ␤-adducin was further analyzed by sedimentation equilibrium experiments (Fig. 7B). The measured molecular mass was 32 Ϯ 6 kDa (mean Ϯ S.D., n ϭ 4). Compared with the monomer molecular mass of 12 kDa, the result indicates that the protein forms oligomers in solution. The average molecular mass is consistent with either trimers, a mixture of dimers and tetramers, or a dimer-tetramer equilibrium. Cross-linking results reveal dimers, trimers, and tetramers although trimers may result from partially cross-linked tetramer (Fig. 7A). The best fit model presents a dimer-tetramer association that can fit all the sedimentation profiles well (not shown).
Fast The association of adducin with actin filament ends was not apparent in previous actin binding assays, presumably due to the low capacity compared with lateral binding sites. Gelsolin, which associates with the fast growing ends of actin filaments (35) and co-sediments with immobilized actin (Fig. 1), was used as a probe to distinguish the small number of adducin-binding sites on the ends of actin filaments from the much larger number of sites derived from the sides of actin filaments. No actin severing activity occurred under assay conditions, since neither gelsolin nor gelsolin-actin 1:1 complex changed the amount of actin sedimented with the beads (Table I). Moreover, the possibility that gelsolin effects could be complicated by direct interaction between gelsolin and either spectrin or adducin was excluded because neither spectrin nor adducin bound to gelsolin beads (data not shown).
Gelsolin blocked up to 70 -80% binding of 50 nM adducin to actin with half-maximal inhibition at about 100 nM gelsolin (Fig. 8A). Increasing gelsolin concentration up to 1.2 M did not further enhance the inhibition, indicating a class of adducinbinding sites that was resistant to gelsolin. These results support the interpretation that gelsolin abolished association of adducin with actin filament ends, and the remaining gelsolinresistant binding corresponds to the high capacity lateral association with actin filament subunits.
Since adducin caps actin filaments and recruits spectrin to F-actin, adducin may also recruit spectrin to actin filament ends. Gelsolin sensitivity was used to evaluate targeting of spectrin to actin filament ends (Fig. 8B). Gelsolin inhibited 70 -80% spectrin recruitment with half-maximal inhibition at 200 nM gelsolin or 60 nM gelsolin-actin complex (Fig. 8B). These values are consistent with the results of Janmey and co-workers (38) who measured the actin capping activity with a K cap of 100 nM for gelsolin and 10 nM for purified gelsolin-actin complex. The gelsolin sensitivity of spectrin recruitment suggests that fast growing ends of F-actin are involved in the spectrin/ a Supernatant and pellet from the assay were separated and analyzed in SDS gel. The relative amounts of actin were determined by pyridine dye elution method (see "Methods"). b 0.4 M gelsolin (ϩgelsolin) or only buffer (control) were incubated with immobilized actin before the assay. adducin/actin interaction and under these conditions represents the major class of sites. The gelsolin-resistant sites presumably are due to recruitment of spectrin to the sides of actin filaments.
Purified gelsolin from two different sources (see "Methods") consistently showed the same inhibition for actin binding and spectrin recruitment of adducin. Despite the view that gelsolin may not bind to F-actin ends in calcium-free buffers (39), our results are consistent with the reports that gelsolin binds to F-actin ends with a relatively low affinity but does not sever F-actin under these conditions (38).
Measurements of affinity and capacity for gelsolin-sensitive adducin/actin interactions were performed (Fig. 9). Gelsolinsensitive actin binding of ␤-adducin-(335-736) occurred with half-maximal activity at 60 nM and a capacity of 0.06 ␤-adducin-(335-736)/actin subunit, which is in contrast to a K d of 1.5 M and capacity of 1 adducin/actin for adducin/actin binding in the absence of gelsolin (Fig. 6). The concentration of 60 nM required for half-maximal gelsolin-sensitive binding is close to the K cap of 150 nM determined in the actin capping assay (Fig.  5). Given that gelsolin-insensitive sites are the majority of total sites (estimated as 94% from Figs. 6 and 9), the K d ϭ 1.5 M approximates the affinity of gelsolin-insensitive adducin/actin binding. Assuming the gelsolin-sensitive sites represent actin filament ends, ␤-adducin-(335-726) has a 25-fold higher affinity for actin at the ends of filaments compared with the sides of filaments. The capacity of 0.06 ␤-adducin-(335-726) monomers/ actin subunit (0.03 dimers/actin monomer) implies that in the bead assay actin filaments contain one exposed end for every 33 monomers, which is consistent with filament lengths determined by gelsolin-binding sites and visualized by electron microscopy (Fig. 1). sucrose, 0.05% Tween 20, 2 mg/ml bovine serum albumin, 0.5 mM ATP, 0.2 mM DTT, 0.5 mM NaN 3 . Then 125 I-labeled ␤-adducin-(335-726) (50 nM, 42,000 cpm/pmol) was added, and the mixture was incubated at 4°C for 1 more hour. The subsequent assay procedure was described under "Methods." The percent of control refers to adducin binding to immobilized actin in the presence of gelsolin as a percentage of adducin binding to actin without gelsolin. Each data point is the mean of duplicate with less then 10% variation between each pair of measurements. B, displacement of adducin-dependent spectrin recruitment to immobilized actin by gelsolin and gelsolin-actin complex. Immobilized actin (400 nM as final) were incubated with increasing concentrations of gelsolin or gelsolin-actin complex at 4°C for 1 h. The gelsolin-actin complex was formed by incubating equal molar amounts of gelsolin and G-actin at 4°C in the presence of 1 mM CaCl 2 for 2 h and then adding 2 mM EGTA. The gelsolin/actin complex was not further purified and therefore was a mixture of gelsolin-actin complex, free gelsolin, and G-actin. ␤-Adducin-(335-726) (50 nM final) and 125 I-labeled spectrin (1 nM final, 1.7 ϫ 10 6 cpm/pmol) were added to the actin/gelsolin mixture and incubated for 1 more hour. The subsequent binding assay was described under "Methods." All the data points are the mean of duplicate measurements with variation of less than 10%. . The binding assay was described under "Methods." The data are presented as nmol of adducin/nmol of sedimented actin, where adducin sedimented without actin has been measured for each adducin concentration and has been subtracted as background. The data are a representative of two other separate assays with similar results. Each point is obtained by averaging triplicate measurements. B, gelsolin-sensitive actin binding activity of adducin. The data were calculated from A by subtracting the binding of ␤-adducin-(335-726) to actin in the presence of gelsolin from binding in the absence of gelsolin.

Spectrin Is Preferentially Recruited by Adducin to Gelsolinsensitive Sites on Actin Filaments-
The spectrin recruiting activity of ␤-adducin-(335-726) was measured as a function of adducin concentration in the presence and absence of gelsolin (Fig. 10). The gelsolin-sensitive component of spectrin recruiting exhibits a half-maximal activation at 15 nM ␤-adducin-(335-726), whereas the gelsolin-insensitive spectrin recruiting was half-maximal at 75 nM ␤-adducin-(335-726). The extent of gelsolin-sensitive spectrin recruitment reached a maximum of 1 site per 1,000 actin monomers at 50 nM adducin. The extent of spectrin recruitment depends on the spectrin concentration, which was only 1 nM in this assay and does not represent the actual capacity for spectrin. Nevertheless, the fact that gelsolin-sensitive spectrin recruitment reached a maximum whereas the gelsolin-insensitive recruitment continued to increase indicates a limited number of gelsolin-sensitive sites that were considerably less than the number of gelsolin-insensitive sites. The ratio of gelsolin-sensitive/gelsolin-insensitive spectrin recruitment varied from about 5:1 or higher, at ␤-adducin-(335-726) concentrations below 25 nM, to 1:1 at ␤-adducin-(335-726) concentrations of 200 nM. These results indicate that at low adducin concentrations spectrin is prefer-entially recruited to gelsolin-sensitive sites at the fast growing ends of actin filaments.
The ␤-adducin-(335-726) concentration of 60 nM required for half-maximal activation of gelsolin-sensitive adducin/actin binary binding is greater than the concentration of 15 nM required for gelsolin-sensitive spectrin/adducin/actin ternary binding. The increased activity of adducin in the ternary system implies co-operativity of spectrin/adducin/actin interactions. Spectrin and adducin independently bind to actin, and presumably all three proteins interact in a ternary complex. The ternary complex therefore would be expected to exhibit a greater stability than a binary complex of adducin-actin. This ternary complex interaction may also explain the difference that gelsolin-insensitive spectrin binding to adducin/actin has half-maximal activation at 75 nM ␤-adducin-(335-726), whereas the gelsolin-insensitive adducin/actin binding occurs with half-maximal activation at 1.5 M ␤-adducin-(335-726).

DISCUSSION
This study presents evidence that adducin exhibits a preference for the fast growing ends of actin filaments for recruitment of spectrin to actin and for direct association with actin. ␤-Adducin-(335-726) promoted recruitment of spectrin to gelsolin-sensitive sites at fast growing ends of actin filaments with a half-maximal activity of 15 nM and to gelsolin-insensitive sites with half-maximal activity of 75 nM. ␤-Adducin-(335-726) also exhibited a preference for actin filament ends in direct binding assays; the half-maximal concentration for binding to filament ends was 60 nM, and the K d for total actin binding was 1.5 M. The concentration of ␤-adducin-(335-726) of 60 nM required for half-maximal binding to filament ends is in the same range as the concentration of 150 nM required for halfmaximal actin capping activity. All interactions of adducin with actin require the MARCKS-related domain as well as an oligomerization site defined for the first time in the neck domain. Surprisingly, the head domain of adducin is not required for spectrin-actin interactions, although it could play a role in forming tetramers. This information is summarized in a schematic model, which is drawn with the assumption that the tail domains lie along the groove of actin filaments (Fig. 11). A previous model depicted that adducin capped F-actin with the tails toward the slow growing ends of actin (25). Because of the new information from this study, which include the necessity of MARCKS-related domain and the sufficiency of neck-tail for the actin capping activity of adducin, we propose the model in which adducin tails oriented toward fast growing ends of Factin and MARCKS-related domains mediate direct contact with actin ends (Fig. 11).
The relative activities of adducin imply that an important role of adducin in cells is to form a complex with the fast growing ends of actin filaments that recruits spectrin and prevents addition or loss of actin subunits. Adducin is the first example of an actin-capping protein that recruits other proteins to actin filament ends and could represent a new class of assembly factor with the function of integrating actin into other cell structures. The number of spectrin molecules recruited per adducin is not yet known. However, a possibility consistent with available data is that one spectrin tetramer is stabilized for each adducin dimer, and two spectrin molecules are associated with adducin tetramers. In this case, spectrin tetramers, which contain two actin binding domains, would form a onedimensional chain which could be the structural precursor to the two-dimensional network of the mature erythrocyte membrane skeleton (2)(3)(4). Additional spectrin molecules associated with lateral sites along actin filaments may be stabilized by protein 4.1, which is present in approximately one copy per spectrin dimer. Actin filaments in erythrocyte membrane skel- were added, and the mixture was incubated for another hour. Subsequent steps were described under "Experimental Procedures." Spectrin pelleted with actin (0.10 pmol of spectrin/nmol of actin) in the absence actin/adducin were subtracted as background. All the data points are the mean of triplicate measurements. B, the gelsolin-sensitive spectrin recruiting activity of adducin was calculated from A by subtracting the adducin-dependent spectrin binding to actin in the presence of gelsolin from that in the absence of gelsolin.
etons are likely to contain tropomyosin, associated along the filament groove (40), and tropomodulin, which binds tropomyosin and caps the slow growing ends of actin filaments (41), as well as dematin (42). It will be of interest to evaluate association of adducin with actin complexed with these accessory proteins. A prediction is that adducin will be excluded from sites on actin filaments coated with tropomyosin and will be confined to filament ends (5).
The fast growing ends of actin filaments of erythrocyte ghosts or isolated spectrin-actin complexes include a population capable of supporting addition of actin monomers and therefore are not capped (2,43). Fowler and colleagues (44) have recently found that actin filament ends in intact erythrocytes are inaccessible to the actin-capping protein CapZ, which is exclusively localized in the cytosol. However, extraction of adducin in low ionic strength buffers lacking magnesium exposed binding sites for CapZ. These findings considered together with the actin capping activity of adducin are consistent with capping of actin filaments in unlysed erythrocytes by adducin (44).
A technical innovation that facilitated this study was an actin sedimentation assay using immobilized actin. The increase in signal is interpreted as resulting from the decrease in filament length and an accompanying increase in the number of actin filament ends (Figs. 1 and 2). Possible reasons for the presence of short filaments are as follows: 1) immobilized actin forms nuclei for actin polymerization with each immobilized actin subunit participating in one filament, and 2) shear force induces fragmentation of beads-coupled long actin filaments. Relatively few direct measurements of protein interactions with actin filament ends have been reported in the literature, with the majority of studies focusing on inhibition of actin polymerization. The assay described here could be useful for direct analysis of interactions of other actin-capping proteins with filament ends.
An oligomerization domain of adducin was mapped to the neck region, encompassing residues 335-436 of ␤-adducin (Fig.  7). Within the stretch of 335-436, residues 360 -386 are highly conserved among human ␣and ␤-adducin subunits (12), rat ␥-adducin (9), and Drosophila Hts (18). This 360 -386 region has the highest probability in full-length ␤-adducin to form an amphiphilic ␣-helix with hydrophobic residues distributed along one side (not shown). Amphiphilic ␣-helices can form coiled-coil structures and mediate protein oligomerization, and the region 360 -386 is one candidate for contacts in oligomers. In support of a role for the region 360 -386, mutation of methionine 369 to proline reduced oligomerization activity of adducin neck-tail (not shown). The fact that the neck domain forms tetramers as well as dimers indicates that subunit contacts in addition to a single coiled-coil interaction also are likely to be involved in oligomerization. The precisely controlled 1:1 ratio of ␣/␤ subunits of erythrocyte adducin (7) could be a consequence of the large excess of ␣over ␤-subunits (12) and lack of stability of ␣-homodimers. Alternatively, some mechanism could exist that specifies assembly of adducin heterodimers. The head domain of adducin may be responsible for additional subunit contacts in heterodimers or tetramers. In support of the idea of subunit contacts in addition to the neck domain, cross-linking of adducin followed by proteolysis results in a 160-kDa tetramer of 40-kDa monomers that lack neck domains (14).
The MARCKS-related domain is demonstrated in this paper to be necessary for interactions of adducin with spectrin and actin. The MARCKS-related domain of adducin also is the site of calmodulin binding and contains the site of phosphorylation by protein kinase C, which both regulate spectrin recruiting and actin capping activities of adducin (10,11,21,25). The polybasic MARCKS-related domain of adducin is a candidate to provide a direct contact with actin and to participate in ionic interactions with negatively charged residues on actin. The idea of an electrostatic interaction between adducin and actin is supported by the finding that spectrin recruiting activity of adducin is salt-sensitive. 2 The tropomyosin/actin interaction is mediated by the charged residues on the sides of actin filaments (45), suggesting the possibility that adducin could also bind to actin filaments in a similar manner and perhaps compete with tropomyosin. Possible residues involved in a complex with adducin at the fast growing ends of actin filaments include Asp-288 and Asp-286. These residues have been proposed to participate in salt bridges between actin monomers in actin filaments (46,47). The increased stability of adducin complexes with the ends of actin filaments could result from participation of both lateral contacts and the contacts with residues such as Asp-288/Asp-286 at filament ends. The atomic structures of actin complexed with gelsolin segment 1 and with profilin implicate apolar residues of actin (48,49). However, profilin also binds charged residues of actin (Asp-288, Glu-361) (49) which may be the targets of the MARCKS-related domain. These considerations suggest that steric inhibition rather than direct overlap of binding sites is likely to be a major factor in 2 X. Li, unpublished data.
FIG. 11. A schematic model of adducin association with actin and spectrin. The model for native adducin tetramer is derived from previous studies (14). The gelsolin-sensitive actin binding activity of adducin is of high affinity (half-maximal binding at 60 nM) and low capacity (0.06 adducin/actin), consistent with actin capping activity. The gelsolin-insensitive actin binding has low affinity (K d ϭ 1.5 M) but a capacity as high as 1 adducin monomer/1 actin subunit, suggesting binding to sides of actin filaments. Spectrin is recruited to the fast growing ends of F-actin by adducin at a half-maximal activation of 15 nM adducin (K act ϭ 15 nM). Spectrin is also recruited to the sides of actin filaments with a K act ϭ 75 nM adducin. The details of adducin-actin complexes at the fast growing end of one actin filament are illustrated. The neck domains of adducin mediate the formation of oligomers, at least homodimers. Two homodimers (␣ and ␤) further form a tetramer. The parallel orientation of adducin subunits in the oligomers is arbitrary. The MARCKS-related domain is inferred to contact directly to actin subunits with 1:1 capacity. Both the neck and the MARCKSrelated domain are required for the full activities of adducin. the gelsolin inhibition of adducin/actin binding (Fig. 8).
Adducin is unusual in its ability to associate with both the sides and ends of actin filaments. Tensin is another protein with multiple actin binding activities and caps F-actin, binds along the sides, and bundles the filaments (50). However, the actin binding and actin capping activities of tensin are located in different domains. The unique feature of adducin is that the same regions, residues 335-726, including both an oligomerization domain and the MARCKS-related domain, are required for actin capping, direct actin binding, and spectrin recruiting activities.
Polybasic domains similar to the MARCKS-related domain of adducin are present in the MARCKS protein family and the NR1 subunit of the NMDA (N-methyl-D-aspartate) receptor. The polybasic domain of the MARCKS protein also is the target for both calmodulin binding and phosphorylation and is directly involved in actin binding and cross-linking (51). The polybasic domain of the NMDA receptor is the site for high affinity calmodulin binding and phosphorylation by protein kinase C (52). The findings of this study suggest the hypothesis that the NR1 subunit of the NMDA receptor also may interact with actin and possibly promote either spectrin-actin or ␣-actinin-actin interactions (53).
Competition between gelsolin and adducin, utilized in this study as a probe for adducin interactions with the fast growing ends of actin filaments, may also occur in vivo considering the ubiquitous expression of adducin and gelsolin (9,12,54). A consequence of gelsolin occupation of actin filament ends instead of adducin would be displacement of spectrin and disassembly of spectrin-actin networks. A variety of pathways regulate actin capping activities, such as calcium dependence of gelsolin (35,55), phosphoinositide inhibition for gelsolin and Cap Z (56,57), and calcium/calmodulin inhibition and phosphorylation inhibition for adducin (11,25). A recent study suggested that phosphorylation of adducin by Rho kinase enhances the adducin/actin interaction (58). These considerations suggest that adducin could contribute to a coordinated and elaborate regulation of actin-based structures.