Mapping the Binding Site on Ankyrin for the Voltage-dependent Sodium Channel from Brain*

Erythrocyte ankyrin is a member of a family of proteins that mediate the linkage between membrane proteins and the underlying spectrin-actin-based cy- toskeleton. Ankyrin has been shown to interact with a variety of integral membrane proteins such as the an- ion exchanger, the Na+K+-ATPase, and the voltage-dependent sodium channel (NaCh) in brain. To under- stand how ankyrin interacts with these proteins and maintains its specificity and high affinity for the volt-age-dependent NaCh, we have mapped the binding site on ankyrin for the NaCh by examining the binding of purified ankyrin subfragments, prepared by proteolytic cleavage, to the purified rat brain NaCh incor- porated into liposomes. laaI-Labeled ankyrin and the radiolabeled 89- and 43-kDa fragments of ankyrin bind to the NaCh with high affinities and with Kd values of 34, 22, and 63 nM, respectively, and have stoichiometries of approximately 1 mol/mol NaCh. The 72-kDa spectrin binding domain is inactive and does not bind to the NaCh. Dissection of ankyrin reveals that the 43-kDa domain retains all the binding prop-erties of native ankyrin to the NaCh. Analysis of the primary structure reveals that the NaCh binding site is confined to a domain of ankyrin consisting entirely of the 11 terminal 33-amino acid repeats and is distinct from the ankyrin domains that interact with spectrin and the Na+K*-ATPase. The voltage-gated sodium channel (NaCh)’ domain") was obtained from spectrin-depleted vesicles by digestion with a-chymotrypsin. To prepare the 89-kDa fragment, erythrocyte ankyrin (4 mg) was digested with 20 pg of TLCK a-chymotrypsin for 1 h at 4 "C in 20 ml of 10 mM sodium phosphate, 0.5 mM Na2EDTA, 1 mM NaN3, 0.05% Tween 20, pH 7.4 (Buffer A). The reaction was terminated by the addition of 5 mM PMSF and the mixture chromatographed on a 10-ml Mono Q anion exchange column equilibrated in Buffer A contain- ing 0.5 mM DTT. The column was washed with the equilibration buffer and eluted with a gradient of 0-0.5 M NaBr in Buffer A containing 0.5 mM DTT. Protein eluting at 0.2 M NaBr was pooled, adjusted to 1 mM MgClz and 1 M NaBr and applied to a hydroxyapatite column (3 ml) equilibrated with 10 mM sodium phosphate, 1 mM NaN3,0.5 mM DTT, 0.05% Tween 20, pH 7.4 (Buffer B). The column was washed in Buffer B containing 25 mM sodium phosphate and eluted with Buffer B containing 50 mM sodium phosphate. Protein eluting from the hydroxyapatite column was dialyzed against 10 mM sodium phosphate, 0.5 mM Na2EDTA, 1 mM NaN3, 0.5 mM DTT, 0.05% Tween 20, pH 7.2, in a microdialysis unit and applied to a Mono S column. The Mono S column was washed with Buffer A and eluted with a gradient of 0-0.5 M NaBr in Buffer A. The peak eluting at -0.2 M NaBr was concentrated by ultrafiltration and used directly for binding experiments or for further proteolytic digestion. The protein linear with silver. a-chymotryptic

Direct binding of '2SI-labeled ankyrin to the reconstituted NaCh shows that ankyrin binds tightly to the channel with 1:l mol stoichiometry, an interaction which depends on the cytoplasmic domain of the NaCh. Most importantly, the interaction of ankyrin with NaChs is selective since neither the neuronal y-aminobutyric acid receptor nor the dihydropyridine sensitive Ca2+ channel, both of which have different distributions on the neuron, immunoprecipitate or bind brain ankyrin. The idea that in mature nerve, NaChs are restrained by cytoskeletal linkages is consistent with recent immunofluorescent and electron microscopic studies in rat sciatic nerve which have shown that a specialized form of erythrocyte ankyrin is localized at the nodes of Ranvier, whereas the "brain" ankyrin isoform is uniformly distributed along the axon (4).
Erythrocyte and brain ankyrin are members of a family of proteins that associate with a variety of integral membrane proteins such as the anion transporter (5), the Na+K'-ATPase (6,7), lymphocyte adhesion antigen PgP-1 (8), as well as the spectrin-actin cytoskeletal network (9,10). Because ankyrin is able to interact and associate with a diverse set of membrane proteins, there is great interest in determining the common structural motifs that mediate these interactions. Although the cytoplasmic domain of the anion transporter is able to compete with the ankyrin-NaCh interaction (3), there is no primary sequence homology (10) between these two proteins. The primary sequence for several erythrocyte (11) and brain ankyrins (12) have recently been determined from cloning of the cDNAs. Ankyrins are proteins of molecular mass -220 kDa folded into three independent domains which include an 89-kDa NH2-terminal domain (131, followed by a 72-kDa domain (14), and a 55-kDa regulatory domain at the carboxyl terminus (15).
To approach the question as to how ankyrin might interact with a diverse set of membrane proteins and still maintain its specificity for the NaCh, we sought to define the binding site of ankyrin for the NaCh by examining the binding of ankyrin subfragments prepared by proteolytic cleavage. We show here that a 43-kDa subfragment derived from the NH2-terminal 89-kDa fragment containing 11 of the 22 tandem 33-amino acid repeats of ankyrin is sufficient to bind to the NaCh and is distinct from the domains that interact with spectrin and the Na'K+-ATPase.

Materials and
Methods-Frozen rat brains were obtained from Pel Freez. Lipids (DOPC, PS, and PE) were obtained from Avanti Polar Lipids. TLCK-treated a-chymotrypsin, Staphylococcus aurezu V8 protease, and other chemicals were obtained from Sigma. 13H]Saxitoxin (STX) (specific activity 30 Ci/mmol) was obtained from Amersham Corp. All other chemicals were the highest commercial grade quality available.
Purification of the Sodium Channel-NaChs were purified from frozen rat brains by sequential chromatography on DEAE-Sephadex, 7483 hydroxyapatite, and wheat germ agglutinin (WGA)-Sepharose columns as described (16,17). Two to three hundred pg of protein was routinely obtained from the WGA step. In order to completely deplete spectrin and ankyrin from preparations of purified NaCh the WGA Sepharose was washed with 300 ml of 0.4 M NaCl, 20 mM HEPES, Tris, pH 7.4,0.1% (w/v) Triton X-100, and 0.025% phosphatidylcholine before elution with N-acetylglucosamine. Fractions from the WGA-Sepharose column (5 pg) were analyzed by SDS-gel electrophoresis on 4-15% linear gradient polyacrylamide gels and stained by ammonical silver (18). The purified NaCh protein routinely appeared as a single broad band with a molecular mass of -260 kDa.
[3H]STX binding to the purified NaCh was determined as described (19) by rapid centrifugation on Sephadex G-50 fine column. The average specific activity from the eight preparations of purified NaCh used in this study was 2,200 f 100 pmol. Both the purity and the [3H]STX binding activities of NaChs obtained either from fresh or frozen rat brains were comparable. Therefore, frozen rat brains were routinely used.
Reconstitution of Purified Sodium Chonnek-Purified NaChs were reconstituted into unilamellar phospholipid vesicles using a D0PC:PE:PS lipid ratio of 5030:15 and a protein:lipid ratio of 1:100. Triton X-100 was removed by gentle shaking overnight with Bio-Beads SM-2. Vesicles were dialyzed against iso-osmotic WGA elution buffer to remove residual N-acetylglucosamine. The protein concentration of vesicles was typically 200-300 pg/ml. Each preparation of reconstituted NaChs was assayed for [3H]STX binding activity by adding vesicles to an incubation mixture of 18 nM [3H]STX in 100 mM choline chloride, 5 mM KCl, 7.8 mM CaC12, 1.3 mM MgSO,, 20 mM Tris/HCl, pH 7.4, containing 0.1% bovine serum albumin. Samples were mixed and incubated for 60 min at 4 "C and the vesicles rapidly collected by filtration on Whatman GF/B filters. Filters were washed three times with ice-cold buffer and the radioactivity bound to filters determined by liquid scintillation counting. Nonspecific binding was measured in the presence of 10 p~ tetrodotoxin and subtracted from the total.
The fraction of reconstituted NaChs oriented with their cytoplasmic domains at the external surface was measured by disruption of the vesicles with 2% Triton X-100 and 0.05% phosphatidylcholine and retitration with 18 nM [3H]STX (17). Data from the eight preparations of reconstituted NaChs used in this study showed a 1.45 & 0.07 increase in the number of the [3H]STX binding sites indicating that 45 & 7% of the channels are oriented with their cytoplasmic surfaces exposed.
Purification of Ankyrin-Ankyrin was purified from outdated human blood as described (5). Purified fractions from the sucrose gradient were collected and concentrated to -4 mg/ml. Alternatively, fractions from the DE-52 column were concentrated using CM-30 ultrafiltration membranes which removed most of the 29-kDa band.
Purification of Ankyrin Proteolytic Subfragments-Three proteolytic fragments were obtained from purified ankyrin as described (13)(14)(15)20). An 89-kDa domain from the NH2-terminal sequence of ankyrin was obtained by a-chymotryptic digestion, a 43-kDa fragment (comprising the COOH-terminal fragment of the 89-kDa domain) was obtained by S. aureus V8 protease digestion of the 89-kDa domain, and a 62.5-kDa fragment (apparent molecular mass of 72 ADa and is referred to as "72-kDa domain") was obtained from spectrin-depleted vesicles by digestion with a-chymotrypsin.
To prepare the 89-kDa fragment, erythrocyte ankyrin (4 mg) was digested with 20 pg of TLCK a-chymotrypsin for 1 h at 4 "C in 20 ml of 10 mM sodium phosphate, 0.5 mM Na2EDTA, 1 mM NaN3, 0.05% Tween 20, pH 7.4 (Buffer A). The reaction was terminated by the addition of 5 mM PMSF and the mixture chromatographed on a 10ml Mono Q anion exchange column equilibrated in Buffer A containing 0.5 mM DTT. The column was washed with the equilibration buffer and eluted with a gradient of 0-0.5 M NaBr in Buffer A containing 0.5 mM DTT. Protein eluting at 0.2 M NaBr was pooled, adjusted to 1 mM MgClz and 1 M NaBr and applied to a hydroxyapatite column (3 ml) equilibrated with 10 mM sodium phosphate, 1 mM NaN3,0.5 mM DTT, 0.05% Tween 20, pH 7.4 (Buffer B). The column was washed in Buffer B containing 25 mM sodium phosphate and eluted with Buffer B containing 50 mM sodium phosphate. Protein eluting from the hydroxyapatite column was dialyzed against 10 mM sodium phosphate, 0.5 mM Na2EDTA, 1 mM NaN3, 0.5 mM DTT, 0.05% Tween 20, pH 7.2, in a microdialysis unit and applied to a Mono S column. The Mono S column was washed with Buffer A and eluted with a gradient of 0-0.5 M NaBr in Buffer A. The peak eluting a t -0.2 M NaBr was concentrated by ultrafiltration and used directly for binding experiments or for further proteolytic digestion. The protein was analyzed on 4-15% linear gradient polyacrylamide gels and stained with silver. The 89-kDa domain was the major protein obtained by a-chymotryptic digestion of ankyrin.
To prepare the 43-kDa COOH-terminal subfragment of the 89-kDa fragment, purified 89-kDa protein (1 mg/ml) was digested with 5 pg/ml S. aureus V8 protease for 3 h at 37 'C in 0.1 M NaCl dissolved in Buffer A. The digest was applied to a Mono S column and eluted with a linear gradient of 0-0.5 M NaBr in Buffer A. The 43-kDa fragment was then concentrated to 100 pg/ml. The purified protein was analyzed on 615% linear gradient polyacrylamide gels stained with silver. Because attempts to concentrate the 43-kDa fragment following chromatography led to aggregation, peak fractions from the Mono S column were used directly for radiolabeling and binding experiments. Fragments of lower molecular masses -32 kDa and moving at the dye front occasionally appeared on storage of purified 43-kDa preparations. Hence these binding assays were carried out directly after purifications.
The 72-kDa spectrin-binding fragment of ankyrin was purified from spectrin-depleted inside-out vesicles by a-chymotrypsin treatment (14,21). Spectrin-depleted erythrocyte ghosts were prepared by incubating erythrocyte ghosts for 30 min at 37 "C in 0.3 M sodium phosphate, pH 7.5. Ghosts were washed once with 7.5 mM sodium phosphate buffer, pH, 7.5, and resuspended in 200 ml of ice-cold sodium phosphate buffer containing 0.2 mM DTT. Vesicles were digested with 1 pg/ml TLCK a-chymotrypsin for 45 min at 4 "C, and the digestion was terminated by the addition of PMSF to a final concentration of 200 pg/ml. The digested protein was absorbed to 20 ml of DE-52 cellulose equilibrated with 7.5 mM sodium phosphate, pH 7.5. The slurry was mixed for 15 min and poured into a column and washed with 300 ml of 7.5 mM phosphate buffer with 0.2 mM DTT and 50 pg/ml PMSF. The column was washed with 75 mM KC1 in phosphate buffer with 0.2 mM DTT and eluted with 200 mM KC1 in the same buffer. Purity of the eluted protein was analyzed on silver-stained 4-15% linear gradient polyacrylamide gels (18).
Radiolabeling of Ankyrin and Ankyrin-derived Proteolytic Fragments-Ankyrin, the 89-kDa and the 43-kDa proteolytic fragments of ankyrin were radiolabeled with '261-labeled Bolton Hunter reagent as described (5,21). '=I-Labeled proteins were used within 1 week of their preparation.
Protein Determinations-Protein concentration was routinely determined by the method of Bradford (22). The protein concentration of reconstituted vesicles was determined by solubilizing vesicles with 2.5% Triton X-100 and determining protein by the Peterson assay (23).
Binding of '=I-Ankyrin and Ankyrin-derived Proteolytic Fragments to NaCh Vesicles-Vesicles containing purified NaChs in binding buffer (0.4 M NaC1,25 mM HEPES/Tris, 0.05% Triton X-100, 0.01% phosphatidylcholine, pH 7.4, containing 1 mM PMSF, 1 mM 1,lOphenanthroline, 1 mM iodoacetamide, 1 pM leupeptin, 1 pM pepstatin A, and 1 p~ aprotinin) were incubated with various concentrations of '=I-labeled ankyrin or '=I-labeled ankyrin subfragments for 6 h at 4 "C with gentle shaking. Vesicles were then rapidly filtered on GF/ B glass fiber filters presoaked in binding buffer containing 0.1% bovine serum albumin and washed once with 5 ml of binding buffer to remove unbound ankyrin. Alternatively, after equilibration, vesicles were separated from unbound ankyrin by rapid centrifugation through Sephacryl S-300 mini-columns (2 ml) from which vesicles were obtained in the void volume. Although both assays yielded identical results, the GF/B filtration assays were convenient for the large number of samples tested and in recovery of Iz5I-labeled ankyrin bound to the NaCh following filtration. To recover NaCh-ankyrin complexes, after filtration, 1.0 ml of Buffer C containing 6% Triton X-100 was added directly to filters, and proteoliposomes were sohbilized for 20 min on ice. The supernatant was removed and adsorbed to 0.5 ml of packed WGA-Sepharose 4B for 12 h at 4 "C. The supernatant was removed, the gel washed two times with 1 ml of Buffer C, and the NaCh-ankyrin complexes subsequently eluted with 1 ml of Buffer C containing 50 mM N-acetylglucosamine. Assays were performed in duplicate and aliquots were counted in a y-counter.

RESULTS AND DISCUSSION
NaChs were purified from rat brain to high purity and on analytical polyacrylamide gels purified NaChs consisted of a single diffusely staining band of molecular mass 260 kDa (Fig.  1, lane A ) . In NaChs used from the final stage of purification, >95% of the ankyrin and spectrin associated with NaChs Sodium Channel Binding Site on Ankyrin 7485 were removed. In all preparations used, as before (3), neither the 36-nor 33-kDa polypeptides, which have been reported to accompany some preparations of NaChs (16), were observed in our preparations. After reconstitution of NaChs into liposomes, [3H]STX binding showed that -50% of the NaCh had their cytoplasmic surface oriented toward the outside of the vesicle, accessible to ankyrin binding. Ankyrin was purified to homogeneity from erythrocytes and on 4-15% silverstained linear gradient SDS-polyacrylamide gels had a molecular mass of 220 kDa with no evidence of low molecular mass breakdown products (Fig. 1, lane B ) . In some preparations a faster migrating band -195 kDa beneath the 220-kDa ankyrin band could be discerned on silver-stained gels, which may correspond to the calpain-derived 195-kDa fragment (15)

200-
= 97 - Although the nitrocellulose binding and dot blot assay that we initially developed to explore the interaction between reconstituted NaChs and ankyrin is rapid and highly reproducible (3), the sample volume is small and recovery of the complexes is not possible. Therefore, as a first step we wanted to develop an approach that potentially would permit recovery of ankyrin-NaCh complexes for further purification and to identify the domains that participated in these interactions. Although we have constructed high affinity and capacity columns with anti-NaCh monoclonal and polyclonal antibodies generated in our laboratory (24), we were able to recover only 40-50% of the NaCh-ankyrin complexes using these antibody affinity columns. Because the NaCh is the only protein contained within the vesicle and because of its affinity for wheat germ agglutinin, we reasoned that '9-labeled NaCh-ankyrin complexes could be recovered by absorption to WGA-Sepharose beads after solubilization of the proteoliposomes with Triton X-100. Based upon the amount of NaCh applied to the column and amounts of '251-ankyrin bound in peak fractions, virtual quantitative recovery of lZ5I-labeled ankyrin, 89-kDa, and 43-kDa ankyrin subfragment-NaCh complexes could be obtained and with similar efficiencies.

30-
When NaCh vesicles were incubated with varying concentrations of '251-labeled ankyrin, ankyrin bound in a concentration-dependent and -saturable manner (Fig. 2). The interaction was specific since '"I-ankyrin could be displaced by unlabeled ankyrin. As described earlier (3), control experiments in which vesicles were used without NaCh protein or where the cytoplasmic surface of the NaCh was removed by proteolysis did not bind '251-ankyrin or ankyrin subfragments. Analysis of the binding isotherm by Scatchard analysis shows that '251-labeled ankyrin binds to the NaCh with a K d of 34 nM and a capacity of 1.3 mol of ankyrin/mol of NaCh after correction for the orientation of NaChs in vesicles determined by [3H]STX binding. Although the equilibrium binding measured in these experiments is consistent with earlier measurements, the slightly lower binding affinity of ankyrin for the NaCh observed in these experiments, compared with the previously published value of 20 nM (3) (32) (right p a n e l ) . from differences in the addition of supplemental detergent and/or increases in ionic strength required to carry out these assays. Similar reductions in the binding of ankyrin to the anion exchanger in the presence of detergent or high ionic strength have been described (20). In addition, differences in lipid composition (DOPC uersus the D0PC:PE:PS vesicles used here) or slight increase in density of NaChs reconstituted into vesicles used for this study could also contribute to the differences observed.
Although a calpain-like 195-kDa fragment of ankyrin lacking a portion of the COOH terminus accompanied some preparations of ankyrin, we did not observe any differences between ankyrin preparations in either the stoichiometry or binding affinities to the NaCh. This is likely related to the negligible role that the COOH terminus plays in mediating ankyrin-NaCh interactions (see below).
To understand the general role of ankyrin as a linker molecule between the cytoskeleton and several membrane proteins, we sought to dissect the domains of ankyrin that mediate its association with the NaCh. Mapping the domain structure of ankyrin by proteolytic cleavage and recent analysis of the primary structure of human erythrocyte ankyrin have shown that the molecule consists of three major domains: a 89-kDa NHz-terminal domain followed by a 62-kDa spectrin binding domain and a 55-kDa regulatory domain at the carboxyl terminus of the protein (Fig. 3). Among the three domains, the 89-kDa domain is unique in that together with some flanking sequences tandem 33-amino acid repeats of ankyrin. Previous work (20) has shown that this purified 89-kDa domain not only competes with ankyrin -for binding to ankyrin-depleted inside-out vesicles but also binds with enhanced affinity to the reconstituted anion exchanger. In contrast the Na+K+-ATPase, a membrane protein that binds ankyrin with high affinity, has a very low affinity for the 89-kDa fragment (25). The question arises, then, as to which domain of ankyrin binds to the NaCh.
Each of the subfragments were tested for their ability to displace lZ5I-ankyrin from its interaction with the NaCh. As shown in Fig. 4, the 89-kDa fragment displaces lZ5I-labeled ankyrin from NaChs with a half-maximal displacement of about 20 nM. At equivalent concentrations, the 89 kDa was more effective than native ankyrin =50 nM) in these displacement assays. At all concentrations tested the 72-kDa internal domain did not compete for ankyrin binding to the NaCh, and this is consistent with our previous report using nitrocellulose binding assays which showed that the 72-kDa spectrin domain does not compete for NaCh-ankyrin interactions (3). In addition, the lack of competition of the 72-kDa spectrin binding domain to NaCh-containing vesicles confirms that the binding of ankyrin and the 89-kDa fragment is not due to residual spectrin associated with the NaCh.
Measurement of the direct binding of '251-labeled 89-kDa domain to NaChs reconstituted into vesicles shows that the 89-kDa fragment retains its binding activity and binds in a concentration-dependent manner with a K d of 22 nM (Fig. 5 ) compared with the K d of 34 nM of native ankyrin measured under similar conditions (Fig. 2). Both unlabeled ankyrin and 89-kDa protein compete effectively for binding of the lZ5Ilabeled 89-kDa domain, indicating that both bind to the same or overlapping sites on NaCh. Scatchard analysis of 89-kDa fragment binding is consistent with a single class of high affinity sites on the NaCh with a maximum binding capacity of 1.1 mol of 89 kDa/mol of NaCh. Although there is evidence that the binding of 89-kDa ankyrin subfragment binds to erythrocyte membranes with either negative cooperativity or to two different affinity sites (20), Scatchard analysis of the binding of both ankyrin and the 89-kDa fragment to the  NaCh show no evidence for such binding. Although there is some scatter in the data that is suggestive of two closely related sites that do not differ significantly in either affinity or capacity, the data are best fit to a single class of interacting sites with approximately equal capacities.

Binding
To further dissect regions of the 89-kDa NHAerminal domain that interact with NaChs, we prepared a 43-kDa fragment of the 89-kDa domain by digestion with V8 protease followed by purification by Mono S chromatography (Fig. 6).
Analysis of the primary structure shows that the 43-kDa fragment consists entirely of 33-amino acid repeats containing repeats 12-22 (11). In displacement assays the 43-kDa fragment competes with native ankyrin for the NaCh interaction, and in direct binding the 43-kDa fragment binds to a single class of noninteracting sites on NaCh vesicles with a Kd of 63 nM and a maximal capacity of 1.2 mol of 43-kDa fragment/ mol of NaCh (Fig. 6). Unlabeled ankyrin, the 89-kDa domain, and 43-kDa fragment proteins all were able to displace the 43-kDa fragment (Fig. 6). Even though removal of the NH2terminal 403 amino acids from the 89-kDa domain yields a

Sodium Channel Binding
Site on Ankyrin fragment with a moderate reduction in the affinity, retention of binding to the NaCh suggests that the carboxyl terminal region, containing 11 of the 22 tandem 33-amino acid repeats of ankyrin is sufficient to mediate the interaction with the cytoplasmic domain of the NaCh and contains most if not all of the information for binding to the NaCh. Although the 43-kDa domain binds to the NaCh without requirement of other regions of the 89-kDa domain, the reduced affinity possibly reflects a need for the flanking sequences in order to fold a stable native ankyrin structure (15). Alternatively, it is possible that the high affinity interaction between ankyrin and the NaCh involves some of the nonrepeat domains of the NHz terminus of the 89-kDa domain or other multiple contact sites.
In each of the fragments that retained binding, the stoichiometry of NaCh-ankyrin complexes remained essentially unchanged. The stoichiometry was 1.3 mol of ankyrin bound per mol of NaCh using intact ankyrin, considering that 50% of the NaCh molecules are oriented with their cytoplasmic domains on the surface of the vesicles, accessible to ankyrin binding. The 89-kDa domain bound to NaCh vesicles with a stoichiometry of 1.1:l and the 43-kDa fragment with an approximate stoichiometry of 1.2:l.
Ankyrin is a multifunctional protein whose primary sequence has revealed a structure which is uniquely suited to mediate the interaction between spectrin and several diverse membrane proteins. In this work we have localized the NaCh binding domain on ankyrin to the 43-kDa carboxyl terminal portion of the 22 tandem 33-amino acid repeat motif. Like the NaCh, both the 89-and 43-kDa domains of ankyrin also are sufficient to mediate the high affinity binding of ankyrin to the anion transporter (20, 25,26). In contrast, the Na+K+-ATPase, which has a high affinity for ankyrin (6,7), interacts weakly with the 89-kDa domain of ankyrin, requiring regions of the spectrin binding domain of ankyrin to reconstitute the high affinity interaction (25). Thus the interaction between ankyrin and the brain NaCh and the domains through which it interacts appear to be similar to those of the anion transporter. There are other similarities between these two membrane proteins. In previous work (3) we demonstrated that there is some structural homology in the cytoplasmic regions of these membrane proteins for the ankyrin binding site, because the cytoplasmic domain derived from the anion exchanger competed for NaCh-ankyrin interactions. The homology is possibly at the level of secondary and tertiary structure, since there is no primary structural homology between these two proteins in their cytoplasmic domains. Taken together, these observations are consistent with the idea that both the NaCh and the anion exchanger evolved to interact with a similar domain of ankyrin, analogous to the lock and key of an antibody-antigen reaction. The Na+K+-ATPase, on the other hand, which interacts with ankyrin has differentially evolved and requires additional regions of ankyrin to confer high specificity.
In brain, two distinct isoforms of ankyrin have been detected. Based upon immunological differences (4) and from recent primary sequence analysis of cloned brain ankyrins (12), brain ankyrin contains the 22 tandem 33-amino acid repeats, each of which closely resemble the repeats of erythrocyte ankyrin. However, brain ankyrin is less active in displacing ankyrin from its anion exchanger binding site (26), indicating that association of ankyrin with the NaCh is not exclusively mediated by the 33-amino acid repeat and that additional specificity is likely built into some of the flanking sequences. Recent studies have also shown that the erythrocyte and brain ankyrins have differences in cellular expression and localization. In nerve, the erythrocyte form is expressed primarily in axons of neurons and at specialized cell domains such as the node of Ranvier, whereas brain ankyrin isoforms are present in both nerve and glial cells and are not found at nodes of Ranvier ,(4). The differential distribution of these ankyrin isoforms is consistent with the distribution of the NaCh in nerve and may be pertinent to the differential distribution of the NaCh and Na+K+-ATPase on the nerve cell surface. In nerve cells it is known that the NaCh is localized to axon hillocks and nodes of Ranvier together with some isoforms of the Na+K+-ATPase (27). An interesting issue is whether the erythrocyte and brain isoforms of ankyrin serve to mediate specific and separate interactions with the NaCh and the Na+K+-ATPase. The association of these membrane proteins with different domains on ankyrin could serve to spatially and functionally separate these two proteins within the same domain on the axon membrane.
The 22 tandem 33-amino acid repeat sequence is a highly conserved motif present in all ankyrins discovered so far and is also present in several proteins involved in cell differentiation such as the Drosophila Notch protein (28) and lin 12 protein of Caenorhabditis elegans (29). Recent work has localized this structural motif to transcription factors that bind to "GA"-rich DNA sequences and referred to as GA binding proteins (30). These repeats, termed ankyrin repeats, may constitute a new class of protein-protein interactions and may have a broad role as a molecular tether in binding proteins of several types, such as nuclear proteins that help regulate gene expression as well as cytoplasmic proteins that participate in cell signaling pathways (31). Most striking are the 33-amino acid repeats which are required to mediate the interaction of ankyrin and the voltage-dependent NaCh and anion exchanger. It is important of note that the Na+K+-ATPase does not interact with this domain and has diverged in its recognition site. Although the erythrocyte ankyrins and brain ankyrins are encoded by different genes on separate chromosomes (11,12), the conserved NHz terminus comprising the tandem repeats suggests that all ankyrins have a basic functional role to perform, mediated by this region of the molecule. Within the motif itself, 15 amino acid residues are highly conserved and the remaining 18 are divergent. Since ankyrin has been demonstrated to bind to diverse membrane proteins such as the anion transporter, the Na+K+-ATPase, and the NaCh proteins, with no significant shared homologous domains, the interactions may be through the divergent amino acids in the motif rather than the conserved ones, where the conserved residues serve to mediate the initial low affinity interactions. These divergent amino acid residues may generate different structural elements (secondary or tertiary) capable of mediating the interaction of different types of membrane proteins. Although the 72-kDa and the COOH-terminal domain of ankyrin do not have a role in specifying the association of ankyrin with the NaCh, these domains have been shown to be important for interactions with spectrin (14) and the Na+K+-ATPase (25). For example, a spliced variant of ankyrin, ankyrin 2.2, has been shown to have a greater affinity for spectrin and for the anion exchanger (15). Recent cloning of the brain ankyrin isoforms have shown that the COOHterminal domains distinguish these ankyrin isoforms from their erythrocyte counterpart and are regions that are likely to confer additional specificity to their distribution in neurons and their interactions with other membrane proteins.