Identification of a Binding Motif for Ankyrin on the (cid:97) -Subunit of Na

Cytoskeleton membrane associations are important for a variety of cellular functions. The anion exchanger of erythrocytes (AE1) and Na (cid:49) ,K (cid:49) -ATPase of polarized epithelial cells provide well studied examples of how integral membrane proteins are anchored via the linker molecule ankyrin to the spectrin-based membrane cy- toskeleton. In the present study we have generated several recombinant fragments of the large (third) cyto- plasmic domain (CD3) of Na (cid:49) ,K (cid:49) -ATPase to define binding sites of ankyrin on CD3 at a molecular level. We provide evidence that a cluster of four amino acids, ALLK, is essential for binding of ankyrin to both recom- binant CD3 and to native Na (cid:49) ,K (cid:49) -ATPase. Once bound, conformational changes might uncover further binding sites for ankyrin on Na (cid:49) ,K (cid:49) -ATPase. A motif related to the ALLK cluster is also present in the cytoplasmic do- main of AE1 where this sequence (ALLLK) turned out to be also important for ankyrin binding. These motifs are highly conserved during evolution of both Na (cid:49) ,K (cid:49) ATPase and AE1, further underlining their potential role in cytoskeleton to membrane linkage.

Cellular differentiation and several cellular functions depend to a large degree on the compartmentalization of particular membrane proteins such as receptors, adhesion molecules or ion translocating proteins to specialized domains of the cell surface (for review, see Refs. [1][2][3]. One mechanism of how certain membrane proteins are placed at specialized sites of the plasma membrane could be by linkage of their cytoplasmic domains to the cytoskeleton. The HCO 3 Ϫ , Cl Ϫ -exchanger of erythrocytes (anion exchanger 1, AE1) 1 and the sodium pump (Na ϩ ,K ϩ -ATPase) of transporting epithelia provide two rather well studied examples of how ion-translocating integral membrane proteins are tethered via specific linker molecules to the fibrous scaffold of spectrin and actin that extends underneath the plasma membrane of virtually all cell types of the body (for review, see Refs. [1][2][3]. Ankyrin is the main linker molecule that connects the cytoplasmic domains of both AE1 and Na ϩ ,K ϩ -ATPase to ␤-spectrin of the membrane scaffold (4 -7).
In transporting epithelia interaction of Na ϩ ,K ϩ -ATPase and AE1 with the spectrin-based membrane scaffold is considered important for the polarized restriction of these transporters to either the apical or the basolateral cell surface (5, 8 -12). Polarity of the Na ϩ ,K ϩ -ATPase has profound implications for the direction of the transport of sodium and several other ions and molecules across the epithelial layer. In various transporting epithelia (such as kidney tubules, parotid gland, retinal pigment epithelium, choroid plexus, Mardin Darby canine kidney (MDCK) cell line) Na ϩ ,K ϩ -ATPase is codistributed with ankyrin and can be copurified (coimmunoprecipitated) with ankyrin and spectrin (5, 9 -13). Binding of erythrocyte ankyrin to kidney Na ϩ ,K ϩ -ATPase in vitro (4,5,14) and competitive inhibition of this interaction by addition of the cytoplasmic domain of AE1 (5,13) suggests that there might be a common binding site on ankyrin involved in binding to both AE1 and Na ϩ ,K ϩ -ATPase. Binding of ankyrin fragments to the ␣-subunit of Na ϩ ,K ϩ -ATPase indicated the involvement of the AE1binding domain of ankyrin and, in addition, a further domain of ankyrin not involved in AE1 binding (6). These observations are compatible with a recent report describing binding of erythrocyte and MDCK ankyrin to two of the five putative cytoplasmic domains of Na ϩ ,K ϩ -ATPase, i.e. to cytoplasmic domains 2 and 3 (14). The present study was performed to obtain more detailed information about the ankyrin binding sites on Na ϩ ,K ϩ -ATPase, if possible at the amino acid level. We confined this study to the large cytoplasmic domain (CD3) because we found in a screening approach that a recombinant protein containing a portion of CD3 blocked binding of erythrocyte ankyrin to native kidney Na ϩ ,K ϩ -ATPase. The most striking outcome of this study is that a motif of four amino acid residues (ALLK) appears to be essential for ankyrin binding. A similar motif (ALLLK) occurs in the sequence of the cytoplasmic domain of AE1 where it appears to participate also in ankyrin binding.

Generation of Various cDNAs for Fusion Protein Expression-Vari-
ous portions of the nucleotide sequence encoding the large cytoplasmic domain (domain 3, CD3) of the ␣-subunit of rat Na ϩ ,K ϩ -ATPase (15) were generated by polymerase chain reaction using full-length ␣-subunit cDNA kindly provided by R. Levenson (Yale University, New Haven, CT). Primer 1 (nucleotides 1417-1436) and primer 2 (nucleotides 2315-2332) generated a 915-nucleotide cDNA corresponding to amino acids 387-691 that cover Ϸ90% of the sequence of the large cytoplasmic domain (CD3-1). Combination of primer 2 with either primer 3 (nucleotides 1600 -1616) or primer 4 (nucleotides 1657-1673) resulted in 5Ј-truncated cDNAs corresponding to amino acid residues 447-691 (CD3-4) and 466 -691 (CD3-5), respectively. Two additional cDNAs were obtained by exonuclease III truncation of the 5Ј end of CD3-1 following the protocol of Henikoff (16). The resulting cDNAs corresponded to amino acid residues 410 -691 (CD3-2) and 421-691 (CD3-3), respectively. Two further cDNAs corresponding to amino acid residues 514 -691 (CD3-6) and 387-514 (CD3-7) were obtained by BamHI and EcoRI cleavage of CD3-1 cDNA cloned in vector pRSET A/B/C (Invitrogen, San Diego, CA). A clone covering a 1149-nucleotide fragment of human brain ankyrin (clone 307, nucleotides 5099 -6247) (17) was kindly provided by Dr. V. Bennett and Dr. E. Otto * Supported by grants of the Deutsche Forschungsgemeinschaft (to R. K. and D. D.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Site-directed Mutagenesis-Point mutation of CD3-7 at position 1632 (A 3 G) was performed to change the triplet AAG (that codes for lysine) to the triplet GAG that codes for glutamic acid (CD3-7*). Mutagenesis was carried out as described in detail elsewhere (19). For generation of the single-stranded uracil-containing cDNA used by this method, Escherichia coli strain CJ 236 deficient in dUTPase and uracil-N-glycosylase was transformed with CD3-7 cDNA cloned in pRSET-A plasmids (20). Single-stranded plasmid DNA purified from this E. coli strain infected with bacteriophage M13KO7 (21) was used for site-directed mutagenesis.
Expression and Purification of Bacterial Fusion Proteins-All amplified cDNAs were subcloned in frame in a pRSET-A or pRSET-C vector (Invitrogen) and transformed in E. coli XL 1 blue using the CaCl 2 procedure (20). After checking all pRSET constructs by restriction mapping and DNA sequencing (22), expression and purification of fusion proteins was performed according to the manufacture's manual (Invitrogen). Briefly, transformed bacteria were grown up to an OD of 0.3 at 595 nm. Then isopropyl-␤-D-thiogalactoside was added to a final concentration of 1 mM, and 1 h later the transcription and translation of the recombinant cDNAs was initiated by the addition of the bacteriophage T7/M13 at a multiplicity of infection of 5-10 plaque-forming units/cell. After further growth of 4 -5 h, the bacteria of a 100-ml suspension were sedimented, lysed, and homogenized in 15 ml of buffer containing 6 M guanidinum HCl, 20 mM sodium phosphate, and 0.5 M NaCl, pH 8. After removal of insoluble material (500 ϫ g, 15 min), the supernatant was loaded onto a Ni 2ϩ -preloaded chelating Sepharose (Pharmacia, Uppsala, Sweden) column (1.2 ϫ 25 cm; flow rate, 1.5 ml/min) equilibrated with urea buffer (8 M urea, 0.5 M NaCl, 20 mM sodium phosphate, pH 8). Contaminating bacterial proteins were removed by washing the column with urea buffer at pH 8 and 6. Fusion proteins were eluted with urea buffer at pH 4, dialyzed at 4°C against 0.1 mM EDTA, 10 mM Tris/HCl (pH 8), and concentrated by dialysis against solid polyethylene glycol 20,000 (Roth, Karlsruhe, FRG). A typical yield was 50 -90 g of fusion protein per ml of bacterial suspension.
Purity of the preparation was assayed by SDS-PAGE (10%). Removal of fusion peptides was performed by cleavage with enterokinase (23) that requires 10 mM CaCl 2 for full activity. After dialysis of fusion proteins (concentration of up to 120 g/ml) with 10 mM CaCl 2 in 10 mM Tris/HCl (pH 8 at RT), 12.5 g/ml enterokinase (Boehringer Mannheim) was added, and cleavage was allowed to proceed for 12-20 h at 37°C.
Purification of Ankyrin, Cytoplasmic Domain of AE1, and Na ϩ ,K ϩ -ATPase-Ankyrin and the cytoplasmic domain of AE1 were purified from one unit of concentrated human erythrocytes exactly as described by Bennett and Stenbuck (24) and Bennett (25). Na ϩ ,K ϩ -ATPase-enriched membrane vesicles were isolated from the outer medulla of four pig kidneys as described in detail elsewhere (26). Purity of proteins was analyzed by SDS-PAGE (see Fig. 1).
Biotinylation of Ankyrin-Peak fractions from the anion exchange purification step containing 0.2-1 mg/ml ankyrin were dialyzed at 4°C against 20 mM sodium borate (pH 8.5). Samples of 500 l were used for biotinylation at 4°C by subsequent addition of 1 l of biotin-N-hydroxysuccimid ester (Pierce) taken from a stock of 12 mg/ml in dimethyl sulfoxide. After 2 h, the reaction was stopped by addition of 50 l of Tris/HCl (1 M, pH 8, at RT). 30 min later, uncoupled substances were removed by dialysis against PBS, pH 7.4.
Binding of Ankyrin to Na ϩ ,K ϩ -ATPase-25 l of packed Na ϩ ,K ϩ -ATPase vesicles (Ϸ20 g of protein) were suspended in 500 l of PBS (pH 7.4, 37°C) containing 5% bovine serum albumin (BSA, grade V, Sigma). After 30 min, 500 ng of biotinylated ankyrin were added to this volume (5⅐10 Ϫ9 M) and allowed to bind to Na ϩ ,K ϩ -ATPase for 30 min at 37°C. Afterward, the mixture of Na ϩ ,K ϩ -ATPase and biotinylated ankyrin was layered on top of a 10% sucrose cushion (800 l in PBS) in a 1.5-ml Eppendorf tube and centrifuged at RT for 8,000 ϫ g. Tubes were then frozen in liquid nitrogen, and the pellet-containing tips of the tubes were cut off with a razor blade. Proteins of the pellet were solubilized with 100 l of SDS sample buffer (29) until completely solved and then immobilized on nitrocellulose filters, washed with M-PBS, and subjected to biotin detection via streptavidin-HRP as described above. Controls included all steps in the absence Na ϩ ,K ϩ -ATPase vesicles.
In a further cosedimentation assay recombinant fragment 7 of the cytoplasmic domain 3 of Na ϩ ,K ϩ -ATPase (CD3-7) and its mutated counterpart (CD3-7*), respectively, were added at a concentration of 18 g/ml to 1 g/ml biotinylated ankyrin in PBS, 5% BSA. After 30 min at 37°C this mixture was used for binding studies with Na ϩ ,K ϩ -ATPase vesicles as described above.

RESULTS
Localization and electrophoretic characterization of recombinant fragments of Na ϩ ,K ϩ -ATPase and other proteins used for binding studies are documented in Fig. 1.
Specific binding of ankyrin to the large (third) cytoplasmic domain (CD3) of Na ϩ ,K ϩ -ATPase was shown by dot blot assay in which a bacterially expressed large portion of CD3 (CD3-1, that covers almost the entire length of CD3) (Fig. 1) was immobilized on nitrocellulose (solid phase) and incubated with biotinylated ankyrin (soluble phase). No binding of ankyrin was observed to BSA, soluble bacterial proteins, a bacterially expressed ankyrin fragment carrying the same short fusion peptide as CD3-1, and the fusion peptide itself (Fig. 2).
To further identify the binding site of ankyrin on CD3-1, six fragments of CD3-1 were generated (CD3-2 to CD3-7) and subjected to dot blot assay using the same protocol. Ankyrin bound to CD3 fragments 2, 3, and 7, but not to CD3 fragments 4, 5, and 6 (compare Fig. 3 with Fig. 1). After removal of the fusion portion, binding occurred also to CD3-4 (abbreviated as ⌬CD3-4) but not to ⌬CD3-5, indicating that the 19 amino acid residues 447-465 of Na ϩ ,K ϩ -ATPase contain a site essential for ankyrin binding (see also "Discussion").
Comparison of this sequence stretch (VAG-DASESALLKCIEVCCG) with the sequences of rat, mouse, and human erythrocyte AE1 (30) revealed a short common motif (ALLK/ALLLK) shared by the cytoplasmic domains of both Na ϩ ,K ϩ -ATPase and AE1. To test the possibility that the ALLK motif of Na ϩ ,K ϩ -ATPase is essential for ankyrin binding we generated a point-mutated variant of CD3 fragment 7 (CD3-7*), in which lysine 458 (K) was replaced by glutamic acid (E). As shown in Fig. 4, biotinylated ankyrin bound to nonmutated CD3-7, but not to CD3-7* in which ALLK was mutated to ALLE.
The significance of this observation was further tested by a competitive binding assay where binding of ankyrin to native kidney Na ϩ ,K ϩ -ATPase was tested in the absence and presence of CD3-7 and CD3-7*, respectively. As expected, binding (cosedimentation) of ankyrin to Na ϩ ,K ϩ -ATPase vesicles was inhibited by addition of nonmutated CD3-7, but was not inhibited by the mutated variant CD3-7* (Fig. 5). These experiments show that the N-terminal third of cytoplasmic domain 3 of Na ϩ ,K ϩ -ATPase (CD3-7) serves as binding site for ankyrin and that the ALLK motif within this fragment appears to be essential for ankyrin binding.
In a further series of experiments we addressed the question whether the ALLK motif is directly involved in ankyrin binding or whether this motif plays an indirect role in affecting the conformation of the cytoplasmic domain. As shown in Fig. 6, the ALLK peptide inhibited binding of ankyrin to immobilized CD3-7, indicating that the ALLK motif is directly involved in binding of ankyrin to this portion of the cytoplasmic domain of Na ϩ ,K ϩ -ATPase. To determine whether the ALLLK cluster in the cytoplasmic domain of the erythrocyte AE1 is also essential for ankyrin binding, we extended binding studies to the purified cytoplasmic domain of AE1 (CD-AE1). Since the ALLLK peptide was too hydrophobic (hardly soluble), we used the extended AE1 sequence LRALLLKHSH. Both peptides (ALLK and LRALLLKHSH) inhibited binding of ankyrin to immobilized CD-AE1 and also to CD3-7 of Na ϩ ,K ϩ -ATPase. Nonsense and control peptides did not interfere with binding of ankyrin to either the CD-AE1 or CD3-7. This indicates that the ALL(L)K clusters on both Na ϩ ,K ϩ -ATPase and AE1 participate directly in ankyrin binding. FIG. 1. Position and electrophoretic  (SDS-PAGE) characterization of recombinant fragments of CD3 of rat Na ؉ ,K ؉ -ATPase (␣1-subunit) used for binding studies. Numbering of amino acid residues follows Herrera et al. (15). CD3-7* carries a point mutation in which lysine 458 was exchanged against glutamic acid. SDS-PAGE of purified pig kidney Na ϩ ,K ϩ -ATPase and human erythrocyte ankyrin is also shown.  . 1). No binding occurs to CD3-4, CD3-5, and CD3-6. However, after removal of the fusion peptide by enterokinase cleavage (⌬CD3-4, ⌬CD3-5) binding is seen to ⌬CD3-4 but not to ⌬CD3-5, indicating that the binding site for ankyrin on CD3 is located between amino acid residues 447 and 465.

DISCUSSION
Interactions of integral membrane proteins with components of the cytoskeleton are probably important for diverse cellular functions, as for example for the assembly of specific plasmalemmal adhesion domains and for the generation or maintenance of cellular polarity (1)(2)(3)31). The erythrocyte AE1 and kidney Na ϩ ,K ϩ -ATPase provide well studied examples of such cytoskeleton-membrane associations. Both integral membrane proteins are linked via ankyrin to the spectrin-based membrane cytoskeleton. The main ankyrin binding site on AE1 appears to be located in the midportion of the cytoplasmic domain (7,32). In Na ϩ ,K ϩ -ATPase, cytoplasmic loops 2 and 3 (CD2, CD3) have been implicated in ankyrin binding (14).
In the present study we were able to identify a main binding site for ankyrin on CD3 of Na ϩ ,K ϩ -ATPase. Generation of various CD3 fragments expressed by E. coli narrowed the protein binding site of ankyrin to a stretch of 19 amino acids (VAGDASESALLKCIEVCCG). This stretch contains the only site of homology between Na ϩ ,K ϩ -ATPase and the cytoplasmic domain of AE1, that consists of the cluster ALLK (Na ϩ ,K ϩ -ATPase) or ALLLK (AE1) (5,13). Several lines of evidence were provided indicating that the ALLK motif is essential for binding of ankyrin to Na ϩ ,K ϩ -ATPase (Fig. 7): (a) no ankyrin binding was obtained with those CD3-fragments of Na ϩ ,K ϩ -ATPase that lack the ALLK motif; (b) mutation of ALLK to ALLE in the CD3-7 fragment (CD3-7*) abolished binding of ankyrin; (c) the nonmutated CD3-7 fragment inhibited binding (cosedimentation) of ankyrin to native Na ϩ ,K ϩ -ATPase, whereas the mutated analogue (CD3-7*) did not interfere with ankyrin binding; and (e) both the ALLK peptide of Na ϩ ,K ϩ -ATPase and the ALLLK peptide of AE1 inhibited binding of ankyrin to CD3-7, whereas nonsense peptides (LAKL, WAGHRPTLGP) did not interfere with ankyrin binding.
Further support for the involvement of ALLK in ankyrin binding to Na ϩ ,K ϩ -ATPase came from experiments in which CD3-4 was used for binding studies. In this CD3 portion the ALLK motif is located very close to the N terminus (VAG-DASESALLK . . .). Binding of ankyrin to CD3-4 occurred only when the fusion peptide was removed by enterokinase cleavage (⌬CD3-4) (Fig. 3), whereas in all other CD3 fragments the fusion portion had no effect on ankyrin binding. The fusion peptide alone did also not interfere with ankyrin binding. The inability of the uncleaved CD3-4 fusion protein to bind ankyrin can be tentatively interpreted by steric hindrance caused by the close proximity between the ALLK motif and the polyhistidine containing fusion portion.
The ALLK/ALLLK motif turned out to be also essential for binding of ankyrin to the purified cytoplasmic domain of AE1 as indicated by inhibition of ankyrin binding to CD-AE1 in the presence of both ALLK and the ALLLK-containing peptide LRALLLKHSH.
Taken together, these data indicate that the ALLK/ALLLK clusters of the cytoplasmic domains of Na ϩ ,K ϩ -ATPase and AE1, respectively, are directly involved in ankyrin binding. This notion is further supported by sequence data showing that the ALL(L)K motif of AE1 is an evolutionary highly conserved part in all vertebrates studied so far (fish, chicken, and all mammalian species) (30,(33)(34)(35). The same holds true for the ALLK motif of Na ϩ ,K ϩ -ATPase which has been conserved without any change from Drosophila to humans (15, 36 -40).
On the basis of these data it appears that the ALLK motif on Na ϩ ,K ϩ -ATPase serves as a primary docking site for ankyrin. Once bound, conformational changes within CD3 might uncover further binding sites located on the second cytoplasmic domain of Na ϩ ,K ϩ -ATPase (14). This model for ankyrin bind- FIG. 6. Influence of the ALLK motif on binding of ankyrin to the cytoplasmic domains of Na ؉ ,K ؉ -ATPase (fragment CD3-7) and of the erythrocyte anion exchanger (CD-AE1). Binding of biotinylated ankyrin (0.5 g/ml Ϸ 2.5⅐10 Ϫ9 M) to 10 g/dot of immobilized CD3-7, CD-AE1, and BSA was assayed in the absence and presence of the following peptides (10 Ϫ6 M): ALLK (potential ankyrin binding motif on CD3-7 of Na ϩ ,K ϩ -ATPase), LRALLLKHSH (sequence of CD-AE1 containing the related motif ALLLK), LAKL (nonsense peptide as control for ALLK), and WAGARPTLGP (control peptide, portion of the exoplasmic Z-loop of anion exchanger 2). The ALL(L)K motifs of Na ϩ ,K ϩ -ATPase and AE1 inhibit binding of ankyrin to both CD3-7 of Na ϩ ,K ϩ -ATPase and the cytoplasmic domain of AE1.
FIG. 7. Diagrammatic summary of the experimental background for the main conclusion of this study: the motif ALLK on CD3 of Na ϩ ,K ϩ -ATPase ␣-subunit appears to be essential for ankyrin binding. Binding of ankyrin to the ALLK motif on CD3 might uncover further binding sites located on CD2.
ing to Na ϩ ,K ϩ -ATPase is also compatible with the observation that ankyrin can bind to Na ϩ ,K ϩ -ATPase with both the AE1binding and spectrin-binding fragments (6), and that two different portions on AE1 appear to crosstalk in ankyrin binding (32).
In view of the location of the ALLK motif in close proximity to the ATPase center of Na ϩ ,K ϩ -ATPase (41) it is tempting to speculate that such complex interactions between ankyrin and Na ϩ ,K ϩ -ATPase might be somehow regulated by the activity of the pump, in that the active pump might fully expose the ALLK motif thereby facilitating its attachment to the membrane cytoskeleton and its exposure at the cell surface. The inactive (dead) pump, on the other hand, might no longer expose the ALLK motif and hence might become prone to cytoskeleton detachment and removal from the cell surface. Such a mechanism might explain the low half-life of apically delivered inactive Na ϩ ,K ϩ -ATPase in MDCK cells (42).