Interaction between the N-terminal Domain of Gastric H,K-ATPase and the Spectrin Binding Domain of Ankyrin III*

We screened a cDNA bank of rabbit gastric fundic mucosa by two-hybrid assays looking for binding partners of the N-terminal domain of the rabbit gastric H,K-ATPase. We extracted five clones sharing more than 90% sequence identity. The longest clone codes for a protein sharing a high identity (96 and 96.8%, respectively) with a fragment of the membrane domain, from Arg-835 to Ser-873, plus the major part of the “spectrin binding domain” going from Glu-874 to Leu-1455 of human and mouse ankyrin III. We conclude that the membrane and spectrin binding domains of the rabbit ankyrin III are candidates for the binding partner of the N-terminal domain of the rabbit gastric H,K-ATPase. To validate the ankyrin-ATPase interaction and to test its specificity, we produced both domains in yeast and bacteria, coimmunoprecipitated them with an anti-ATPase antibody, and copurified them by affinity chromatography. The sequence of rabbit ankyrin III was not known, and this is the first report demonstrating that the ankyrin III and the H,K-ATPase interact with no intermediate. The interaction involves the N-terminal domain of the ATPase on one hand and the spectrin binding domain of the ankyrin on the other.

The sequestration of molecules in intracellular compartments is a fundamental capacity of all living cells, and in polarized cells, the sequestration is directional. For instance, the plasma membrane that envelops all epithelial cells has two domains, basal and apical. Both correspond to different components. Membrane proteins must be targeted to one or the other domain after synthesis. The mechanisms of targeting are not yet fully understood, but the cell skeleton has a major role, at least by leading proteins to, and holding them in, their membrane domains (1)(2)(3).
Linkers are required for the interaction of some membrane proteins such as AE1 (anion exchange 1 or band 3 protein) and Na,K-ATPase with skeleton components. The AE1 and Na,K-ATPase linker is ankyrin (4 -6). Smith et al. (7) previously demonstrated the colocalization of ankyrin and H,K-ATPase in parietal cells and the coimmunoprecipitation of the gastric pump with exogenous erythrocyte ankyrin. This supports the idea that the gastric ATPase can bind ankyrin as its parent protein, the Na,K-ATPase, does. However, it is not known how ATPase and ankyrin interact and whether minor cellular components act as linkers.
In this study, we demonstrate that the rabbit gastric H,K-ATPase binds to ankyrin III. We map the binding site in the N-terminal (Nt) 1 domain of ATPase on one hand and in the spectrin binding domain of ankyrin III on the other hand. The spectrin binding domain of the rabbit ankyrin III is cloned and has more than 90% identity with the analogous domains of the mouse and human ankyrin. Fragments of rabbit ATPase and rabbit ankyrin III were both expressed to test the specificity of the interaction by coimmunoprecipitation and affinity chromatography.

EXPERIMENTAL PROCEDURES
Library Construction-Stomachs were collected from Nembutal-injected rabbits. Gastric fundic mucosa was scrapped and immediately frozen in liquid nitrogen. Samples were sent to CLONTECH Laboratories, Inc. for preparation of the cDNA bank with (dT) and random oligomers to limit the overexpression of 3Ј extremities of mRNA.
cDNA were ligated in the prey plasmid pGAD 10 using EcoRI adaptors and incorporated in DH5␣ Escherichia coli strain by efficiency transformation. The library had about 1.7 million independent clones, of which 30% were empty.
Construction of the Bait Plasmid-The DNA fragment for the first 98 residues of the rabbit gastric H,K-ATPase was obtained by PCR from the complete cDNA. The PCR conditions were as follows: (i) denaturation at 95°C for 10 min; (ii) 35 cycles each consisting of denaturation at 95°C for 1 min, annealing at 55°C for 1 min, and elongation at 72°C for 30 s using Vent polymerase (New England Biolabs); and (iii) a final elongation at 72°C for 10 min.
The PCR product with the expected size of 320 base pairs was digested with BamHI and SalI and purified on a 4% agarose gel with a QIAquick gel extraction kit (Qiagen) according to the manufacturer's instructions. It was inserted in the BamHI/SalI site of the pFBL23 plasmid (a gift of J. Camonis, Institut Curie, Paris, France), and the product was sequenced (Genome Express, Grenoble, France).
Library Screening-Yeast strain HF7c (a donation of J. Camonis, Institut Curie, Paris, France) was transformed to Trp prototrophy with pFBL23-Nt using the lithium acetate method (8). The resulting strain was further transformed to Trp, Leu, His prototrophy with the pGAD10 library plasmid. The transformation mix (100 l) was plated on a 10-cm Petri dish and incubated at 30°C for 5 days. Aliquots of each mix were taken before plating and used to determine the transformation efficiency on Trp-and Leu-free medium.
Hisϩ colonies were screened for ␤-galactosidase activity using the filter assay method. Positively transformed colonies eliciting 50-unit absorbance at 600 nm were blotted onto Whatman 3MM paper. The filter was frozen in liquid nitrogen and thawed at room temperature. The freeze-thaw cycle was repeated three times. The filter was soaked in 60 mM Na 2 HPO 4 , 40 mM NaH 2 PO 4 , 10 mM KCl, and 1 mM MgCl 2 containing 2 mg/ml 5-bromo-4-chloro-3-indolyl ␤-D-galactopyranoside (X-gal) and then incubated overnight at 37°C. The clones expressing the ␤Ϫgalactosidase were selected only when the Lex A DNA binding domain-Nt fusion was present.
Plasmids were extracted and recovered by electroporation in the leu Ϫ E. coli strain 1066 (a gift of C. Cullin, Center Génétique Moléculaire, CNRS, Gif sur Yvette, France). The plasmids were then used to transform the HF7c strain alone or with pFBL23 in the presence of Nt or a control protein (laminin, hydrophobic protein). The clones expressing the Nt of H,K-ATPase were selected when they were positive for Trp, Leu, His prototrophy and ␤-galactosidase activity. The plasmids were sequenced (Genome Express).
Coimmunoprecipitation of the Two-hybrid Proteins Expressed from Yeast-Yeast extracts were prepared from liquid cultures in SD-Trp-Leu-His medium (CLONTECH Laboratories, Inc.). Cultures were grown to an absorbance at 600 nm of 0.8. The yeast extracts were centrifuged (5,000 ϫ g for 5 min), and the pellets were washed in 50 mM Na 2 HPO 4 , 300 mM NaCl, pH 8. The yeast extracts were lysed by vortexing with glass beads (reference number 854 170/1, Braun Biotech International). Cell debris was discarded by centrifugation (20,000 ϫ g for 20 min), and the supernatants were incubated for 4 h at 4°C in the presence of 10 g of the anti-Nt H,K-ATPase antibody (95A3). Then, 10 g of IgG2 against mouse IgG were added and incubated for 4 h at 4°C. Immune complexes were collected on protein A-Sepharose (Amersham Pharmacia Biotech), separated by SDS-PAGE, and blotted onto nitrocellulose.
Construction of Vectors Expressing the Nt of H,K-ATPase-The DNA sequence encoding for the Nt (amino acids 1-98) was obtained from the complete cDNA of rabbit gastric H,K-ATPase cloned in Bluescript vector by PCR amplification. The PCR conditions were the same as for the bait plasmid. The final product of 320 base pairs was digested with NcoI and BglII and purified on a 4% agarose gel with a QIAquick gel extraction kit (Qiagen) using the manufacturer's instructions. This cDNA was inserted between the restriction sites NcoI and BglII of the prokaryotic expression vector pQE 60 (Qiagen) to generate a plasmid encoding for an in-frame fusion protein composed of Nt and 6 histidines in the C-terminal extremity. The plasmid with the insert was checked by sequencing (Genome Express).
Expression of the Nt of H,K-ATPase-JS 238 E. coli strain (a gift of O. Fayet, Laboratoire de Microbiologie et Génétique Moléculaire, Toulouse, France) overexpressing the lactose operon repressor (Lac lq mutation) was transformed by the CaCl 2 method with pQE 60 containing the Nt of the ATPase. Cells were grown at 37°C in LB medium containing 100 g of ampicillin until the absorbance at 600 nm reached 0.8 units. Expression of the fusion protein was induced by adding 1 mM isopropyl-1-thio-␤-D-galactopyranoside for 4 h at 30°C. Bacteria were harvested at 4°C by centrifugation at 5,000 ϫ g for 20 min. Pellets were suspended in 50 mM Na 2 HPO 4 , 300 mM NaCl, pH 8 in the presence of the protease inhibitor mixture (P 8215, Sigma). Bacteria were lysed by sonication (14 m for 1 min) and centrifuged at 20,000 ϫ g for 20 min at 4°C. The supernatant was used for immunoprecipitations and affinity chromatographies.
Construction of the Vector Expressing the Spectrin Binding Domain-The DNA sequence encoding for the spectrin binding domain (SBD) was obtained by the PCR amplification of a clone isolated by the two-hybrid screening. The PCR conditions were as follows: (i) denaturation at 95°C for 10 min; (ii) 35 cycles each consisting of denaturation at 95°C for 1 min, annealing at 55°C for 1 min, and elongation at 72°C for 2 min with the Pfu polymerase (reference number 60025251, Stratagene); and (iii) a final elongation at 72°C for 10 min.
The PCR product of 1850 base pairs was digested with BamHI and Bsu36I and purified on a 1% agarose gel with a QIAquick gel extraction kit (Qiagen) using the manufacturer's protocol. Then the full cDNA was inserted between the BamHI/Bsu36I restriction sites of the yeast expression vector pYeF2 2A (a gift of Center Génétique Moléculaire, CNRS, Gif sur Yvette, France) to generate a plasmid encoding for the SBD under the dependence of a galactose-inducible promoter (pGAL10-CYC1) with an adenine selection marker. The plasmid with insert was sequenced (Genome Express).
Expression of the Spectrin Binding Domain-Expression was performed in yeast strain BMA 64 (a donation of Center Génétique Moléculaire, CNRS, Gif sur Yvette, France). Yeast were grown at 30°C overnight in a 100-ml flask containing 25 ml of WO medium (yeast nitrogen base (Life Technologies, Inc. 25685-033), 1.43 g/liter; ammonium sulfate, 5 g/liter; galactose, 20 g/liter) complemented with 20 g/ml uracil, 20 g/ml tryptophan, 100 g/ml leucine, and 10 g/ml histidine. The following day, 250 ml of the same medium were seeded and incubated at 30°C up to the absorbance at 600 nm of 0.8. Yeast were pelleted by centrifugation (5,000 ϫ g for 10 min), suspended in 2 ml of sonication buffer with 5 l of protease inhibitor mixture (P 8215, Sigma), and then broken by vigorous vortexing with glass beads (854 170/1, Braun Biotech International). Cell debris were discarded by centrifugation at 4°C (15,000 ϫ g for 10 min).
Immunoprecipitation-Four mixtures were prepared and incubated at 4°C for 4 h. These mixtures were as follows: (i) 100 l of bacterial extract containing the Nt plus 2 ml of yeast extract containing the SBD; (ii) 100 l of bacterial extract containing the Nt plus 2 ml of yeast extract without the SBD; (iii) 100 l of bacterial extract without the Nt plus 2 ml of yeast extract containing the SBD; and (iv) 100 l of bacterial extract without the Nt plus 2 ml of yeast extract without the SBD.
Ten g of 95A3 were added and incubated for 4 h at 4°C. Then, 10 g of IgG2 anti-mouse IgG were added for 4 h at 4°C. Immune complexes were collected on protein A-Sepharose (Amersham Pharmacia Biotech), separated by SDS-PAGE, blotted onto nitrocellulose by electro-elution, and analyzed.
Affinity Chromatography-Four columns of 500 l of nickel resins (superflow, Qiagen) were prepared and equilibrated with 50 mM Na 2 HPO 4 , 300 mM NaCl, pH 8. In columns A and B, 10 ml of the bacterial extract containing the Nt domain of the ATPase were poured. In C and D, 10 ml of the bacterial extract free of the Nt domain of ATPase were poured. All columns were washed according to the manufacturer's protocol. Three liters of a yeast culture containing the SBD fragment of ankyrin were poured in columns A and C, and 3 liters of the same culture medium free of SBD were poured in columns B and D. All columns were rinsed with 50 mM Na 2 HPO 4 , 300 mM NaCl, pH 8 and eluted with 500 mM imidazole. The four eluates were analyzed by SDS-PAGE and Western blots.
Western Blots-Nitrocellulose sheets were incubated in 1ϫ phosphate-buffered saline, 5% bovine serum albumin, 0.1% Tween 20 in the presence of antibodies, either the 95A3 mouse IgG monoclonal antibody against the Nt of the H,K-ATPase or the rabbit IgG polyclonal antibody against the SBD of ankyrin III (a gift of K. Kordeli, Laboratoire de Biologie Cellulaire, Institut J. Monod, Faculté de Jussieu, Paris, France). For staining, secondary species antibodies conjugated with horseradish peroxidase were used, and the stain was revealed with the Pierce chemiluminescence kit (reference number 34080, Pierce) using the manufacturer's protocol.

Screening of the Gastric cDNA Bank
Seven million gastric cDNA clones were screened against the Nt fragment of the H,K-ATPase by the two-hybrid assays; 7 million is four times the amount of independent clones of the cDNA bank and should thus be statistically sufficient to screen all the different cDNA fragments. Selected were 144 positive clones, which amount to 20 positive clones per million. Five clones had more than 90% identity. Three were identical (1,791 base pairs) and corresponded to 597 aa. The two others were shorter and overlapped the previous sequence on 498 aa (1,494 base pairs). Those fragments have a high identity (Ͼ96%) with the "membrane domain" of the human and mouse ankyrin III from Arg-835 to Ser-873 and with the major part of the spectrin binding domain going from Glu-874 to Leu-1455 (9).
By looking for a protein interacting with the N-terminal domain of the rabbit gastric H,K-ATPase, we selected the membrane and spectrin binding domains of the rabbit ankyrin III. The sequence shows little variation between the human and mouse ankyrin III sequences (Fig. 1).

Characterization of the ATPase-Ankyrin Interaction
Coimmunoprecipitation of Fusion Proteins-To check whether the activation of the reporter genes of the two-hybrid assay is the consequence of a specific interaction between the N-terminal domain of the ATPase and the ankyrin fragment, we first tested whether the Nt domain (expressed as Lex A DNA binding domain-Nt) and the SBD domain (expressed as SBD-gal4-activating domain) can coprecipitate when they are incubated in the presence of a specific anti-ATPase antibody. Both proteins are recovered in the same precipitate (Fig. 2,  lanes 1 and 4). This precipitate is absent when the extracts of the bait Lex A DNA binding domain-Nt cotransfected with gal4-activating domain (SBD-free) and of the prey SBD-gal4activating domain cotransfected with Lex A DNA binding domain (Nt of ATPase-free) are used. This absence indicates that the immunoreaction is specific. Thus the activation of the reporter gene of the two-hybrid assays is due to the direct interaction of the Nt domain of the gastric ATPase with the ankyrin fragment.
Coimmunoprecipitation of the Isolated Domains-To corroborate the previous conclusions, both proteins were expressed in the absence of their fusion partners: the activating and binding domains. The SBD (597 aa) was expressed in yeast, and the Nt of the ATPase (98 aa) was expressed in E. coli. After mixing the yeast and E. coli extracts, the Nt anti-ATPase antibody (95A3) was added. Precipitation was induced as described under "Experimental Procedures." The SBD of ankyrin was recovered in the ATPase precipitate (Fig. 3, lanes 1 and 2). If the ATPase is absent, the ankyrin fragment is not precipitated, which implicates that the coprecipitation requires a specific interaction (Fig. 3, lanes 3-6).
Affinity Chromatography-To further check the specificity of the interaction, we tested whether the Nt-SBD complex was recovered by affinity chromatography on nickel resin (Fig. 4). In two columns (A and B), we poured the bacterial extract containing the Nt fragment of the ATPase tagged with 6 histi- dines (Fig. 4, a and b, lane 4). In two other columns (C and D) we poured the bacterial extract free of the ATPase fragment (Fig. 4c, lane 4; result not shown for column D, because no protein was eluted from this column). Then, different yeast extracts were added. Those containing the ankyrin fragment (Fig. 4, a and c, lane 3) were loaded on columns A and C, and those free of the ankyrin fragment (Fig. 4b, lane 3 and unshown result (column D)) were loaded on columns B and D.
All columns were then rinsed and eluted. The presence of the Nt fragment of the ATPase and of the SBD fragment of ankyrin in the elution was demonstrated by SDS-PAGE and by Western blots. The fractions eluted from column A contain both proteins (Fig. 4a, lanes 1 and 2). Similar bands are absent from the elution of column C (Fig. 4c, lanes 1 and 2) and of column D (result not shown). Therefore, the Nt domain of the ATPase binds to the nickel resin and is able to hold back the SBD fragment. Moreover, the Western blot of the fractions loaded on, and eluted from, columns B, C, and D (Fig. 4, b and c; results from column D are not shown) argue for the specificity of the interaction because, in the absence of the Nt, of the SBD, or of both fragments, no immunoreactive protein is held back on the columns, as revealed by the antibodies.
Therefore, we conclude that the interaction between ATPase and ankyrin fragments is specific. The Nt domain of the gastric ATPase and the SBD of rabbit ankyrin III are responsible for this interaction. Because the Nt and SBD fragments are the only proteins recovered by affinity chromatography, we further conclude that the interaction is direct and does not require a third partner (Fig. 5). DISCUSSION Looking for cytoplasm partners of the N-terminal domain of the rabbit gastric H,K-ATPase, we screened a cDNA bank of rabbit fundic mucosa by two-hybrid assays. We identified the SBD of rabbit ankyrin III. Previous studies strongly suggested an interaction between the H,K-ATPase and an ankyrin. Smith et al. (7) demonstrated by indirect immunofluorescence and by confocal microscopy on isolated gastric glands that the H,K-ATPase colocalized with ankyrin in resting and stimulated parietal cells. In the resting cells, the distribution of both proteins is diffuse in the cytoplasm, whereas after stimulation, they are both located at the apical membrane. In the same work, Smith et al. (7) demonstrated that exogenous erythrocyte ankyrin coimmunoprecipitates with the parietal cell microsomes. Because the main protein of gastric microsomes is the proton pump, this H,K-ATPase was a good candidate for an interaction with ankyrin.
In this work we make the first direct demonstration that the ankyrin III and the H,K-ATPase interact and that their interaction does not require a third partner. Moreover, we demonstrate that the 80 residues of the N terminus of the ␣ subunit of the gastric H,K-ATPase bind to a fragment of ankyrin that contains the spectrin binding domain of ankyrin.
It has been demonstrated that the Na,K-ATPase and ankyrin interact by way of several sites (10). The ATPase sites were mapped on the second (aa 140 -290) and catalytic (aa 345-784) loops of the ␣ subunit. It was further demonstrated that the ATPase interacts with the repeat domain (89 kDa) and the SBD of ankyrin (10). The existence of at least two sites explains the high affinity (K d , 25 nM) of the complex, and the destruction of immunoprecipitate probed with the anti-SBD polyclonal antibody and a secondary anti-rabbit antibody conjugated with horseradish peroxidase. The stained band is the SBD fragment. Lane 2, immunoprecipitate probed with the 95A3 anti-monoclonal antibody and a secondary anti-mouse antibody conjugated with horseradish peroxidase. The stained band is the Nt fragment of the ATPase. Lane 3, immunoprecipitate from the fraction containing only the Nt of H,K-ATPase probed with the anti-SBD polyclonal antibody and a secondary anti-rabbit antibody conjugated with horseradish peroxidase. Lane 4, immunoprecipitate from the fraction containing only the Nt of H,K-ATPase probed with the 95A3 antibody and a secondary anti-mouse antibody conjugated with horseradish peroxidase. The stained band is the Nt fragment of the ATPase. Lane 5, immunoprecipitate from the fraction containing only the SBD of rabbit ankyrin III probed with the anti-SBD polyclonal antibody and a secondary anti-rabbit antibody conjugated with horseradish peroxidase. Lane 6, immunoprecipitate from the fraction containing only the SBD of rabbit ankyrin III probed with the 95A3 antibody and a secondary anti-mouse antibody conjugated with horseradish peroxidase. one site decreases the affinity constant by 20-fold (5).
The gastric H,K-ATPase and the Na,K-ATPase are two homologous ionic pumps. They have a similar structure, with two subunits sharing a high degree of sequence identity (60 -65% on the whole sequence). Homology is high, especially in the domains carrying the ankyrin Na,K-ATPase interaction sites. Therefore, the gastric H,K-ATPase could have ankyrin binding sites on the second loop (aa 155-300) and/or the catalytic domain (aa 357-777). Indeed, the catalytic domain of the H,K-ATPase has a 19-residue fragment conserved in the Na,K-ATPase and in the band 3 protein also. This fragment has been identified as an ankyrin binding site (11).
By two-hybrid assays, we tested whether loop II and the catalytic domain of the gastric ATPase interact with the SBD of ankyrin III. The ATPase fragments (residues Gln-161 to Thr-293 for loop II and Thr-357 to Arg-777 for the catalytic loop) failed to interact with the SBD. Two explanations can be offered. Either none of those H,K-ATPase domains binds to the SBD of ankyrin III, or the two-hybrid assays are inefficient. The latter hypothesis cannot be ruled out because, if the Nterminal of the ATPase is short and polar and thus has a good probability to be expressed as a native fold, loop II and the catalytic domain are longer and more hydrophobic. The expression of hydrophobic fragments often leads to inclusion bodies because the native fold is not reached. 2 Lastly, it could be argued that this part of the H,K-ATPase binds to other domains of ankyrin (repeat domain, regulatory domain) that we did not use in our assays.
The N-terminal domain of the H,K-ATPase and the spectrin binding domain of ankyrin interact. This interaction is specific to the gastric pump, because the N terminus of the Na,K-ATPase does not interact. Actually, no interaction of the N terminus of the Na,K-ATPase was detected when the whole ankyrin III was used. The N-terminal fragments of both ionic pumps are the most different parts of their sequences. This could indicate different functions. The N terminus of the H,K-ATPase could be a regulating element in the activation of parietal cells by acting in the translocation of the pump to the apical membrane. Indeed, this phenomenon is specific to the H,K-ATPase. It relies upon the recruitment of the gastric parietal cell skeleton by as yet unclear mechanisms. Ankyrin III could be a piece of the puzzle. FIG. 4. Western blots of the 11% SDS-PAGE of the fractions eluted from columns A, B, and C. a, column A was loaded with the bacterial extract containing the Nt of H,K-ATPase and rinsed, and the yeast extract containing the SBD of rabbit ankyrin III was poured. Lane 1, eluted fraction probed with the anti-SBD of rabbit ankyrin III polyclonal antibody and a secondary anti-rabbit antibody conjugated with horseradish peroxidase. Lane 2, eluted fraction probed with the anti-Nt of H,K-ATPase monoclonal antibody 95A3 and a secondary anti-mouse antibody conjugated with horseradish peroxidase. Lane 3, the initial material probed with the anti-SBD polyclonal antibody and a secondary anti-rabbit antibody conjugated with horseradish peroxidase. Lane 4, the initial material probed with the anti-H,K-ATPase monoclonal antibody 95A3 and a secondary anti-mouse antibody conjugated with horseradish peroxidase. Lane 5, the initial material probed with the secondary anti-rabbit antibody conjugated with horseradish peroxidase alone. Lane 6, the initial material probed with the secondary anti-mouse antibody conjugated with horseradish peroxidase alone. b, column B was loaded with the bacterial extract containing the Nt of H,K-ATPase and rinsed, and the yeast extract free of the SBD fragment was poured. Lanes are as in a. c, column C was loaded with the bacterial extract free of the Nt of H,K-ATPase and rinsed, and then the yeast extract containing the SBD fragment was poured. Lanes are as in a.