Subunit interactions in the Na,K-ATPase explored with the yeast two-hybrid system.

Subunit interactions of the alpha1- and beta1-subunits of the chicken Na,K-ATPase were explored with the yeast two-hybrid system. Gal4-fusion proteins containing domains of the alpha1- and beta1-subunits were designed for examining both intersubunit and intrasubunit protein-protein interactions. Regions of the alpha- and beta-subunits known to be involved in alpha-beta-subunit assembly were positive in two-hybrid assay, supporting the validity of the assays. A library of beta-subunit ectodomains with C-terminal truncations was screened to find the maximal truncation retaining an interaction with the alpha-subunit extracellular H7H8 loop (where H7 refers to the seventh membrane span, and so on). The maximal truncation removed all the cysteines involved in disulfide bridges, leaving only 63 amino acids of the beta-subunit ectodomain. Scanning alanine mutagenesis led to identification of an evolutionarily conserved sequence of four amino acids (SYGQ) in the extracellular H7H8 loop of the alpha-subunit that is crucial to alpha-beta-intersubunit interactions. Oligomerization studies with single domains failed to detect self-association of either of the two large cytosolic loops (H2H3 and H4H5) within the alpha-subunit. However, evidence was found for an interaction between these two cytoplasmic loops.

The Na,K-ATPase is an integral membrane protein that plays a central role in ionic homeostasis in animals by mediating the translocation of Na ϩ and K ϩ ions across the plasma membrane against their electrochemical gradients. The active Na,K-ATPase is a heterodimer comprised of a 100-kDa ␣-subunit that spans the plasma membrane 10 times, and a 40 -60-kDa glycoprotein ␤-subunit that has a short cytoplasmic Nterminal domain, a single transmembrane domain, and a large extracellular domain. Both subunits are required for Na ϩ and K ϩ ion transport (1)(2)(3). The ␣-subunit contains the cation binding sites and the sites of ATP binding and phosphorylation, and it is therefore sometimes referred to as the catalytic subunit. The ␤-subunit is involved in the structural and functional maturation of the holoenzyme (4,5) and transport to the plasma membrane (2,6), and it appears to influence K ϩ sensitivity (7,8). The ␣and ␤-subunits assemble in a noncovalent, glycosylation-independent manner during or soon after biosynthesis (9,10), and assembly is required for exit from the endoplasmic reticulum (11).
Identification of domains involved in assembly of Na,K-ATPase subunits has been approached in previous studies by immune precipitation experiments that have involved co-expression of truncated ␤-subunits (12,13) and chimeras between the Na,K-ATPase ␣-subunit and either sarcoplasmic/endoplasmic Ca-ATPase (14 -16) or the gastric H,K-ATPase catalytic subunit (17)(18)(19). Expression of Na,K/Ca-ATPase chimeric catalytic subunits together with the avian ␤-subunit in mammalian cells, usually in the T7 RNA polymerase-based expression system (20), has allowed us to define a 26-amino acid segment within an extracellular loop of the Na,K-ATPase ␣-subunit that is necessary and sufficient for assembly of the chimeras with the Na,K-ATPase ␤-subunit. To define further the specific amino acids involved in Na,K-ATPase subunit interactions, we have employed the yeast two-hybrid assay system (21,22).
There is evidence that the Na,K-ATPase exists as an (␣-␤) 2 heterotetramer in cell membranes, at least during some portion of the transport cycle (23)(24)(25). Thus, one would expect that, in addition to sites of ␣-␤-subunit assembly, there must be sites at which ␣-␤-subunit heterodimers interact to form the native tetramers. Blanco et al. (26) demonstrated that ␣-␤ oligomers containing two different isoforms of the ␣-subunit could be purified by immune precipitation from detergent-solubilized rat brain membranes and that ␣-␣ dimers formed when two ␣-subunit isoforms were co-expressed in Sf-9 insect cells. In studies reported here, we used the yeast two-hybrid system to screen for intrasubunit interactions in the major cytosolic domains of the ␣-subunit and in the extracellular domain of the ␤-subunit.
In the two-hybrid system assay, plasmids are constructed that encode two hybrid proteins: one consists of the DNA-binding domain of the transcription factor Gal4 fused to one test protein, X, and the other consists of the Gal4 activation domain fused to another test protein, Y. These plasmids are transformed into a Saccharomyces cerevisiae strain that contains reporter genes whose regulatory region contains Gal4 binding sites. Either hybrid protein alone must be unable to activate transcription of the reporter genes. The DNA-binding domain hybrid should not activate transcription because it does not provide the activation function, whereas the activation domain hybrid also should not activate transcription because it cannot localize to the Gal4 binding sites. Interaction of the two test proteins reconstitutes the function of the Gal4 transcription factor and results in expression of the reporter genes, which are detected by assays for the reporter gene products. For our studies, a set of plasmids encoding Gal4-fusion proteins that contained elements of the Na,K-ATPase ␣and ␤-subunits were constructed and used in the two-hybrid system to explore intersubunit interactions and to look for intrasubunit interactions. The results of these experiments are presented in this report.

EXPERIMENTAL PROCEDURES
Construction of Plasmids Encoding Hybrid Proteins-All of the hybrid constructs were created using amplification by polymerase chain * 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. This research was supported by NINDS Grant NS23241 from the National Institutes of Health.
‡ To whom correspondence should be addressed: Dept. of Biology, The reaction (PCR). 1 The PCR reactions contained 10 ng of template pBluescript SK ϩ plasmid (Stratagene) containing a cDNA encoding either the chicken Na,K-ATPase ␣1or ␤1-subunit, 100 ng of each primer (see below), 1 unit of Perkin-Elmer Taq DNA polymerase, 50 mM KCl, 20 mM Tris-HCl, pH 8.3, 1.5 mM MgCl 2 , 0.001% gelatin, and 0.2 mM of each of the four deoxynucleotide triphosphates (Pharmacia Ultrapure), in a reaction volume of 0.1 ml, overlaid with 50 l of mineral oil (Sigma). Amplification was performed for 30 cycles with a temperature profile of 1 min at 95°C, 1 min at 42 or 50°C, and 1 min at 72°C. All of the PCR fragments were digested with the appropriate restriction enzymes (BamHI, BglII, NcoI, and/or SmaI) overnight at room temperature for SmaI digests or at 37°C for the other restriction enzymes. The digested PCR products were purified by agarose gel electrophoresis, electroeluted, ligated overnight at 15°C into both the pAS2 and pACT2 vectors for use in the yeast two-hybrid system (27), and transformed into Escherichia coli DH5␣ competent cells (Boehringer Mannheim). Ampicillin-resistant colonies were screened for the presence of the PCR fragment by restriction analysis of their plasmids. The nucleotide sequences of candidate plasmids were determined, and the desired plasmids were used in the yeast transformations described below.
The DNA encoding the ␤-subunit ectodomain was constructed using the PCR primers TC150 (5Ј-ACGTCCCGGGGAATTTGAACCCAAG-TAC-3Ј) and TC110 (5Ј-ATCGGGATCCGCTGCTTTTTATGTCAAATT-3Ј). This domain contains amino acids Glu 63 to Ser 304 of the chicken Na,K-ATPase ␤1-subunit, numbering from the N-terminal Ala of the mature protein. The DNA encoding the extracellular ␤X149 domain was constructed using PCR primers TC150 and TC191 (5Ј-TACGTC-CCGGGATCCTGGACAGGGATGAGATAGGGGTTG-3Ј). This domain is a 93-amino acid C-terminal truncation of the ␤1 extracellular domain and contains amino acids Glu 63 to Val 211 of the chicken Na,K-ATPase ␤1-subunit. The DNA encoding the extracellular ␤X96 domain was constructed using PCR primers TC150 and TC190 (5Ј-TACGTCCCGG-GATCCCAGTTCTCCAGCCACTCACGTTTG-3Ј). This domain is a 146amino acid C-terminal truncation of the ␤1 extracellular domain containing amino acids Glu 63 to Asn 158 of the chicken Na,K-ATPase ␤1-subunit.
Yeast Transformation and ␤-Galactosidase Assays-Fusion genes were introduced into a yeast reporter strain by the lithium acetate transformation procedure of Gietz et al. (28). Transformants were allowed to grow at 30°C, usually for 2-4 days, until colonies were large enough to assay for ␤-galactosidase activity. Transformant cells were then plated directly onto sterile Whatman number 1 filters that had been layered onto selective growth media. After colonies had grown, the filters were assayed for ␤-galactosidase activity. The cells were permeabilized by a cycle of freezing the filters in liquid nitrogen and thawing to room temperature. Each filter is then soaked with 2 ml of Z-buffer (CLONTECHMatchmaker protocol manual) containing 5-bromo-4chloro-3-indolyl-␤-D-galactoside. The filters were then placed in a covered plastic container at room temperature and checked periodically for the appearance of blue colonies. Blue colonies appeared between 30 min and 12 h. The filters were then dried and photographed to record the data.
Truncation of the Extracellular Domain of the ␤-Subunit-Truncations of the DNA encoding the extracellular domain of the Na,K-ATPase ␤-subunit were performed on the pACT␤ plasmid. Exonuclease III digests were performed for 12 min during which samples were removed and the reactions quenched every 2 min. Plasmid DNA was digested with two restriction enzymes, cleaving the DNA just 3Ј of the ␤-subunit coding DNA and leaving a protective 3Ј overhang at the termination sequence and a 5Ј overhang in the direction of the ␤-subunit coding sequence. The 3Ј overhang is resistant to exonuclease III digestion. The cut DNA, 5 g, was dissolved in 50 l of reaction buffer (New England Biolabs) (66 mM Tris-HCl, pH 8.0, and 0.66 mM MgCl 2 ), 100 units of exonuclease III enzyme (New England Biolabs) were added, and the reaction mixture was incubated at 37°C. The exonuclease III digestion rate under these conditions is approximately 300 bases per min. At 2-min intervals, 5 l of aliquots were removed, mixed with STOP solution (0.2 M NaCl, 5 mM EDTA, pH 8.0), and incubated at 70°C for 10 min to inactivate the exonuclease III enzyme. The digested DNA was then ethanol-precipitated and resuspended in 20 l of mung bean nuclease buffer (New England Biolabs) (50 mM sodium acetate, pH 5.0, 30 mM NaCl, 1 mM ZnSO 4 ). Ten units of mung bean nuclease (New England Biolabs) were added, and the samples were incubated at 30°C for 15 min to digest the single-stranded DNA and create blunt ends for ligation. The DNA was again precipitated in ethanol and resuspended in 20 l of Tris-EDTA buffer, and blunt-end ligations were performed overnight at 16°C to generate small libraries of pACT␤ truncations. Yeast cells were co-transformed with aliquots of these pACT␤ truncations and pAS ␣EC49.

The Sodium Pump Is Compatible for Use in the Two-hybrid
System-Previous immunoprecipitation studies using chimeric Ca-ATPase/Na,K-ATPase ␣-subunits and chimeric DPPIV/ Na,K-ATPase ␤-subunits have demonstrated that a stretch of 26 amino acids (called EC26) (15), within the H7H8 extracellular loop of the ␣-subunit, and the ectodomain of the ␤-subunit (13) are required for ␣-␤-subunit assembly. In the present study, initial two-hybrid experiments with Gal4-fusion proteins containing the entire H7H8 extracellular loop (called EC49) of the ␣-subunit and the extracellular domain of the ␤-subunit showed that these fusion proteins form protein-protein interactions resulting in activation of transcription of the ␤-galactosidase reporter gene. Fig. 1 is a diagram of the Na,K-ATPase, indicating the regions of the ␣and ␤-subunits that were analyzed in two-hybrid assays. Table I shows the results indicating that neither pAS ␣EC49 nor pACT␤ alone activates transcription. Table II shows that when both the pAS ␣EC49 and pACT␤ are co-expressed, ␤-galactosidase transcription is activated. These results demonstrate that these ␣ and ␤ peptides as fusion proteins in yeast retain their ability to assemble and therefore are compatible for use in the two-hybrid system.
␣EC49 (H7H8) Interacts with the Extracellular ␤X149 and ␤X96 Truncated Domains-Hamrick et al. (13) showed that the avian Na,K-ATPase ␤-subunit, truncated by 92 or 146 residues from the C terminus, remained competent to form ␣⅐␤ complexes when expressed in mammalian cells. As an extension of these experiments, we constructed GAL4-fusion proteins containing deletions in the extracellular ␤ domain that were the same as those of Hamrick et al. (13), that is deletions of 92 and 146 amino acids from the C terminus. These ␤ deletions were named ␤X149 and ␤X96 because they retain 149 and 96 aminoacyl residues of the ␤-subunit ectodomain, respectively (see Fig. 2). These deletion constructs co-expressed with the ␣EC49 (H7H8) fusion protein yielded positive results in the two-hybrid assay (Table II), suggesting that the N-terminal 96 amino acids of the extracellular domain of the ␤-subunit (Glu 63 -Asn 158 ) are sufficient to form a protein-protein interface with the EC49 extracellular loop (H7H8) of the ␣-subunit.
Further Truncations of the ␤-Subunit Extracellular Domain-To determine the minimal C-terminal extent of the ␤-subunit involved in protein-protein interactions with the ␣EC49 (H7H8) ␣-domain, we performed a series of exonuclease III digests on the pACT␤ construct from the 3Ј end. The pools of pACT␤ deletions were transformed into yeast together with the pAS ␣EC49 (H7H8) construct. The co-transformants were plated onto media lacking histidine. If the co-expressed fusion proteins interacted, they would cause transcription of a histidine reporter gene, allowing those co-transformants to grow on histidine-deficient medium. Two separate experiments were performed, resulting in a total of 25 transformant colonies after 12 min of exonuclease III digestion. The DNA from each of these 25 colonies was isolated and used as a template in PCR reactions primed by the primers for the ␤X96 construct (TC150 and TC190). This PCR screen was used to look for any deletion that encoded less than the 96 amino acids of the extracellular ␤ domain already identified. The results of this screening indicated that three of the ␤ deletions contained fewer than 96 amino acids of the extracellular ␤ domain. Then, a set of nested oligonucleotide primers was used to estimate the extent of coding region remaining in the three ␤ deletion clones. From this experiment, it appeared that the shortest ␤ truncation encoded no more than 63 amino acids and possibly as few as 61. Finally, to test this conclusion, site-directed mutagenesis was performed on the pACT␤X96 plasmid, introducing a stop codon after the codon for Asp 125 of the ␤-subunit ectodomain. This construct, pACT␤X63, co-expressed with pAS ␣EC49, yielded positive results in the yeast two-hybrid system, confirming that Glu 63 to Asp 125 is a sufficient extent of the extracellular ␤-subunit domain for interaction with the ␣-subunit EC49 loop (H7H8). These results suggest that the three disulfide loops of the ␤-subunit are not required for interaction with ␣EC49 (H7H8) and that no more than the 63 amino acids adjacent to the transmembrane domain are necessary for this interaction. This region is shaded in the diagram of the ␤-subunit in Fig. 2.
Scanning Alanine Mutagenesis of the ␣EC49 Domain-To identify individual amino acids involved in the ␣-␤ intersubunit interaction, we constructed Gal4-fusions with altered forms of the EC49 ␣-subunit domain (H7H8) that contained  amino acid mutations to alanine. We chose to use the EC49 context since the EC26 fusion protein activated transcription on its own (Table I). Sequence alignments of the aminoacyl resides in the EC49 (H7H8) domain of the ␣-subunit revealed a cluster of residues well-conserved among all the known ␣-subunits of the Na,K-ATPase and H, K-ATPase families (the families in which ␣-␤ heterodimers appear to be the functional units) (Fig. 3).
To identify aminoacyl residues within the H7H8 loop that might be involved in ␣-␤-subunit interactions, a set of four alanine scanning variants of the EC49 region were constructed (see Fig. 3), and these constructs were co-expressed with the Gal4-␤-subunit constructs pACT␤, pACT␤X149, and pACT␤X96 in yeast two-hybrid assays. Table III shows the results of the two-hybrid assays. The fusion proteins EC49 Ala1, EC49 Ala3, and EC49 Ala4 all demonstrate positive protein-protein interactions with all three ␤ fusion proteins. These results suggest that the 12 mutated aminoacyl residues may not be crucial to protein-protein interactions between the ␣and ␤-subunits. However, the fusion protein EC49 Ala2 did not show positive protein-protein interactions with any of the three Gal4-␤-subunit fusion proteins. These results suggest that the group of four highly conserved amino acids (SYGQ) is directly involved in protein-protein interactions between the ␣and ␤-subunits.
A Search for Homotypic ␣-Subunit or ␤-Subunit Interactions-There is substantial evidence that the Na,K-ATPase exists largely as (␣-␤) 2 dimers in cell membranes. The yeast two-hybrid system was used to seek elements of the ␣-subunit that might be involved in ␣-␣ interactions; similarly, we sought evidence for ␤-␤ interactions. The two largest individual cytosolic domains of the ␣-subunit (see Fig. 1) were tested for self-association: ␣ cytoplasmic loop 1 (H2H3) with itself, and ␣ cytoplasmic loop 2 (H4H5) with itself. The results in Table IV demonstrate that in two-hybrid assays no ␣-␣ self-associations occurred. Likewise, tests for oligomerization involving the ectodomain of the ␤-subunit were mostly negative (Table IV).
Only the ␤X96 fusion protein demonstrated any oligomerization interactions with itself. These results are inconclusive as to whether biologically significant ␤-␤-subunit interactions occur. The lack of positive results with the larger ectodomain suggests that ␤-␤ interactions are not biologically meaningful. The full-length ␤-subunit ectodomain could not be tested for ␤-␤ interactions because the ␤-subunit ectodomain fused to the Gal4 DNA-binding domain was positive when expressed in yeast by itself.
The Two Large Cytoplasmic Loops (H2H3 and H4H5) of the ␣-Subunit May Interact-The two large cytoplasmic loops of the ␣-subunit were also tested for interaction with each other in the yeast two-hybrid system. The ␣ cytoplasmic loop 1 (H2H3) and ␣ cytoplasmic loop 2 (H4H5) were positive for interaction when ␣ cytoplasmic loop 1 (H2H3) was fused to the activation domain and ␣ cytoplasmic loop 2 (H4H5) was fused to the DNA-binding domain of Gal4 but not when in the reverse orientation (Table IV). These results do support the occurrence of interactions between the H2H3 and the H4H5 loops of the ␣-subunit but are less compelling than would be the case if positive results had been obtained with the loops in both fusion configurations. DISCUSSION While the major application of the two-hybrid system has been for screening cDNA libraries to find clones encoding proteins that bind some target protein, the same methodology is useful for identifying domains or amino acids involved in interactions between proteins that are known to interact. Many combinations of proteins have been used successfully in the two-hybrid assay. These combinations include many nuclear,  (50,51) are also shown. The initial segment of the ␤-subunit ectodomain, shaded, includes all of the residues encoded in pACT␤X63 and is the minimal length of ␤-subunit ectodomain found in this study to interact with the H7H8 loop of the ␣-subunit. The C-terminal residues of truncated ␤-subunit ectodomains ␤X96 (Asn 158 ) and ␤X149 (Val 211 ) are drawn as squares.
pACT ␣EC49 ϩ cytoplasmic, mitochondrial, and viral proteins but only a few membrane-associated proteins (22). To our knowledge, this study is the first to utilize the two-hybrid system to define domains of subunit interaction in a multi-subunit plasma membrane protein. It is important to note some caveats about the studies reported here. First, ␣and ␤-subunit interactions that are essential to the processes involved in assembly of the Na,K-ATPase in the endoplasmic reticulum need not be entirely the same as sites of subunit interaction in the mature, active enzyme. Second, while studies on assembly identify domains and residues that are important to the ␣ and ␤ contacts, they do not distinguish those elements that indirectly regulate the ␣-␤ interface from those that are the interface. This same caveat, of course, applies to virtually all studies that involve analysis of perturbations in protein structure. Third, while these experiments reveal some domains and residues that are at interfaces, they do not identify all relevant interfaces. We know, for example, that subunit assembly of the Na,K-ATPase is more efficient when the ␤-subunit includes its cytoplasmic and membrane-spanning domains, although these are not absolutely necessary for assembly (13,30). ␤-Subunit Interactions with the ␣-Subunit H7H8 Loop-To define features of the ␤-subunit necessary and sufficient for assembly with the ␣-subunit, Renaud et al. (12) began by making mutations in the avian ␤1-subunit and assaying for the ability of mutant ␤-subunits to assemble with mouse ␣-subunits in transfected mouse L-cells. These experiments showed that the cytosolic 33 amino acids at the N terminus of the ␤-subunit were not required for assembly. A set of small deletions in the membrane-spanning region of the ␤-subunit also failed to inhibit assembly, although most of these deletions prevented the assembled ␣⅐␤ complexes from leaving the endoplasmic reticulum. Hamrick et al. (13) expanded these studies by examining the assembly potential of C-terminally truncated ␤-subunits. Truncation of the ␤-subunit by 92 or 146 amino acids failed to abolish assembly with the ␣-subunit, although the yields of assembled complexes in the immunoprecipitation experiments were clearly decreased. These results suggested a region between aminoacyl residues 126 and 170 might be especially important for assembly. However, subunit assembly was not detected between avian ␣-subunits and a chimera consisting of the cytosolic and membrane spanning domains of   DPPIV and the ␤-subunit extracellular domain truncated by 146 amino acids. The combined effect of replacement of the cytosolic and membrane-spanning domains together with truncation of the C terminus by 146 residues either decreased the efficiency of subunit assembly or the stability of the assembled complexes during isolation (i.e. solubilization and immunoprecipitation) to the point where no evidence of assembly remained. However, the ectodomain of the ␤-subunit, as a secretory protein, remained capable of forming stable complexes with the ␣-subunit (31).
In the experiments reported here, an attempt was made to define the minimal ␣-␤ interaction domains further, by using the yeast two-hybrid system, which might be sensitive enough to detect weaker interactions. Indeed, positive results were found when Gal4 fusions containing deletions in the extracellular ␤ domain matching those created by of Hamrick et al. (13) were co-expressed with the EC49 fusion protein. This is the first time that the regions of the ␣and ␤-subunit that had separately been defined as sufficient for subunit assembly were actually shown to interact with each other. Previously it had been found (a) that ␤-subunits truncated at their C terminus by 92 or 146 residues could assemble with the entire ␣-subunit, and (b) that a Na,K-ATPase/Ca-ATPase chimera in which the only residues contributed by the Na,K-ATPase were 26 amino acids in the H7H8 loop could assemble with the entire extracellular domain of the ␤-subunit (15,16). The results of the two-hybrid assay indicate that these minimal assembly domains do interact with each other. That is the segment from Glu 63 to Asn 158 of the ␤-subunit is sufficient to form a proteinprotein interface with the EC49 extracellular loop (H7H8) of the ␣-subunit.
By screening a library of cDNAs encoding C-terminal truncations of the ␤-subunit ectodomain, we found that an even shorter segment of the ␤-subunit contained an ␣-subunit binding site. This region was estimated by PCR amplification to correspond approximately to residues Glu 63 to Phe 123 . These results were substantiated by showing that Glu 63 to Asp 125 interacts with the EC49 ␣-domain (H7H8). These results are remarkable in showing that none of the three disulfide loops of the ␤-subunit is critical in protein-protein interactions with the EC49 loop (H7H8) of the ␣-subunit.
␣-Subunit Interactions with the ␤-Subunit Ectodomain-There are several isoforms of each subunit of the Na,K-ATPase (see reviews 25,32,33). Assembly between the various ␣ isoforms and ␤ isoforms from the same species has been demonstrated directly with immunological methods (16, 34 -37) and indirectly by detection of functional pumps (11, 36 -39). Formation of interspecies hybrid ␣⅐␤ complexes has also been demonstrated (2, 7, 12, 13, 40 -42). Furthermore, co-expression of Na,K-ATPase ␣1-subunits and H,K-ATPase ␤-subunits resulted in functional hybrid ␣⅐␤ complexes (6,39). Finally, Na,K-ATPase/Ca-ATPase catalytic subunit chimeras containing only 26 amino acids of the Na,K-ATPase H7H8 loop were shown to assemble with the gastric H,K-ATPase ␤-subunit as well as with the ␤1 and ␤2 isoforms of the Na,K-ATPase ␤-subunit (16). The results of these studies suggest common assembly domains in the Na,K-ATPase and the H,K-ATPase subunits, with the ␣-subunit domain lying within the H7H8 loop. There is a stretch of 26 aminoacyl residues within this domain that is well conserved in evolution between Na,K-ATPase and H,K-ATPase ␣-subunits. Twelve of the 26 residues are identical in all ␣-subunits. These residues might be expected to include a general motif for interaction with the ectodomain of ␤-subunits. The replacement of critical aminoacyl residues by alanine should eliminate protein-protein interactions involving these contacting amino acids.
Protein-protein interfaces are made up of a mixture of hydrophobic and hydrophilic residues. They are usually well paired so that hydrogen bond donors and acceptors are matched along with the hydrophobic groups (43). Alanine is chosen as a generic replacement residue because it is the most common amino acid in proteins, and it is found within buried and exposed positions and in all manner of secondary structures (43). Alanine does not supply new hydrogen bonding, sterically bulky, or unusually hydrophobic side chains (43). Alanine substitutions reduce the functional comparisons among the mutants to a common standard state. To increase the efficiency of analysis one can mutate amino acid groups to alanine in clusters ranging from 2 to 5 residues within segments of 10 to 15 residues (43). This allows one to determine quickly which clustered mutants are most disruptive and subsequently dissect them to identify the important residues. The scanning mutational approach directly tests only the importance of side chains; information about main chain interactions remains unknown. Main chain interactions are common among protein-inhibitor complexes but less so in subunit-subunit and antibody-antigen interactions (43).
With this alanine-scanning strategy, the fusion proteins EC49 Ala1, EC49 Ala3, and EC49 Ala4 all demonstrate positive protein-protein interactions with the full ␤-subunit ectodomain and the two truncated ␤ fusion proteins. However, the fusion protein EC49 Ala2 did not evidence interaction with any of the three ␤ fusion proteins. This result suggests that the group of four highly conserved amino acids, SYGQ, in the ␣-subunit H7H8 loop includes residues directly involved in protein-protein interactions between the ␣and ␤-subunits.
A Search for a Region Involved in Dimerization of ␣⅐␤ Complexes-Although structural studies of the sodium pump support a subunit stoichiometry of one ␣-subunit to one ␤-subunit, the exact quaternary structure is still in debate. The formation of a higher order enzyme complex is supported by studies of ␣-␣ interactions among the Na,K-ATPase isoforms in rat brain and among rat ␣-subunits expressed in virally infected Sf-9 insect cells (26). Expression of a truncated ␣1 isoform with the fulllength ␣-subunit demonstrated that the C-terminal half of the ␣-subunit is required for ␣-␣-subunit oligomerization in insect Sf-9 cells. Through the use of chimeras between the catalytic ␣-subunits of the Na,K-ATPase and H,K-ATPase, the region involved in ␣-␣ interaction was further defined as lying between residues Gly 554 and Pro 785 in the central, cytosolic loop (44). Cross-linking studies by Sarvazyan et al. (45) indicate ␣-␣ associations between the N-terminal H1-H2 and C-terminal H8-H10 segments of the Na,K-ATPase ␣-subunit, with the most probable interacting helices being the H1-H10 pair and the H2-H8 pair. It is not known whether these contacts involve intra-or inter-␣-subunit interactions.
The results of ␣-␣ oligomerization experiments performed with the two-hybrid system did not reveal any self-association of either large cytoplasmic loop of the ␣-subunit, even though ␣ cytoplasmic loop 2 contained the residues Gly 554 to Pro 785 , implicated in ␣-␣ oligomerization by Koster et al. (44). One possible rationalization of the differing results is that ␣-␣ dimerization involves elements of ␣-subunit structure that are not intrinsic to an isolated ␣-subunit cytosolic loop, such as would be present in Gal4 fusion proteins. There is evidence to support this interpretation in recent experiments of Froehlich and colleagues (46), experiments that suggest that dimerization of ␣⅐␤ complexes to form (␣⅐␤) 2 complexes may involve conformational states that occur transiently during the transport cycle.
␣-Subunit Cytoplasmic Loop 1 (H2H3) and Cytoplasmic Loop 2 (H4H5) Interactions-The cytoplasmic loop 2 of the ␣-subunit (H4H5) contains the phosphorylation and nucleotide binding sites. However, some mutations within cytoplasmic loop 1 (H2H3) influence the ATPase activity and vanadate sensitivity of the Na,K-ATPase, suggesting that loop 1 may interact with loop 2 (H4H5). The possibility of loop 1-loop 2 interaction was tested in two-hybrid assays. The assays were positive when cytoplasmic loop 1 (H2H3) was fused to the activation domain of Gal4 transcription factor, and cytoplasmic loop 2 (H4H5) was fused to the DNA-binding domain but not vice versa. In twohybrid assays of defined protein combinations, one orientation of the hybrids (i.e. protein X fused to the DNA-binding domain and protein Y to the activation domain) often activates transcription much more efficiently than the reverse hybrids (29). This may reflect differences between the levels of expression or stability of hybrids containing X and those containing Y. Transcription is optimal when the activation domain hybrid is in excess over the DNA-binding domain hybrid (29). When the reverse is true, DNA-binding domain hybrids bound to the reporter gene promoters are less likely to be engaged in the X-Y protein-protein interaction and therefore may not give positive results. Since ␣-subunit loop 1 and loop 2 interactions were seen with only one orientation, this constitutes positive but weak evidence for inter-loop interaction.
Future Studies-There are additional protein-protein interactions within the Na,K-ATPase that can be approached with two-hybrid studies. For example, the present studies do not include a search for interactions involving the cytosolic domain of the ␤-subunit nor the N-terminal, C-terminal, and H6H7 and H8H9 cytosolic loops of the ␣-subunit. In addition, the Na,K-ATPase is known to interact with ankyrin (47-49), and there is evidence that the ␤2 isoform ␤-subunit expressed by glial cells may interact with "receptors" on some central nervous system neurons (34). The yeast two-hybrid system appears to be a promising approach not only for defining the subunit assembly domains more completely but also for observing other proteinprotein interactions that involve the Na,K-ATPase.