Isoform-specific Monoclonal Antibodies to Na,K-ATPase α Subunits EVIDENCE FOR A TISSUE-SPECIFIC POST-TRANSLATIONAL MODIFICATION OF THE α SUBUNIT

Abstract Monoclonal antibodies to isoforms of the Na,K-ATPase have become important tools in the study of the enzyme's distribution, physiological roles, and gene regulation, and when their epitopes are defined, they are useful in the study of enzyme structure as well. Evidence is presented that the α3-specific antibody McBX3 recognizes an unusual epitope that is not present on α3 in the heart. The epitope, which is also found in kidney α1 from some species, was mapped to a site on the large intracellular loop near the ATP binding site. DNA sequencing of reverse transcribed-PCR products encompassing the corresponding regions from α3 from brain (where McBX3 recognizes α3) and heart demonstrated that the tissue difference in epitope is not due to alternative splicing of the mRNA. Instead, hydroxylamine sensitivity indicated that the antibody recognizes a post-translational modification. The epitope for a new antibody for α3, XVIF9-G10, was mapped to a site near the N terminus, a location analogous to the sites for the well-characterized antibodies McK1 (α1) and McB2 (α2). The antibody XVIF9-G10 reacted with the α3 of the heart as well as that of the brain; however, McBX3 and XVIF9-G10 both stained the same cellular structures in sections of the rat retina. A new α1-specific antibody, 6F, was characterized and mapped to another site near the N terminus; this antibody has broader species specificity than the other well-characterized α1 antibody, McK1.

The catalytic subunit of the Na,K-ATPase 1 has three principal isoforms, ␣1, ␣2, and ␣3 (1), each derived from a different gene (2). The sequence conservation is about 85% between isoforms and as high as 99% between vertebrate species for each isoform. Sequence differences between isoforms are concentrated in a few locations, most notably near the N terminus. Isoform-specific antisera have been obtained either against specific peptides (3,4) or by preadsorbing partially isoformspecific antisera with preparations containing the other isoforms (5,6). Monoclonal antibodies have the great advantages of unlimited supply, consistent quality, and low background, however.
The monoclonal antibody McK1 recognizes rat ␣1 and was mapped to the sequence DKKSKK near the N terminus by protein chemistry and comparison of species-specific sequence differences (7). Data are reported here with chimeras of ␣1 and ␣2 that confirm this epitope assignment with fine resolution. The antibody McB2 recognizes ␣2 and has a very broad species specificity (8). Its epitope has been mapped to the sequence GREYSPAATTAENG at the N terminus by protein chemistry and lambda expression libraries derived from the ␣2 cDNA. 2 The antibody McBX3 is largely specific for ␣3 in rat brain (8), but identification of its epitope has been problematic. Its location and unusual properties are described here.
Four monoclonal antibodies raised against chicken kidney in the laboratory of D. M. Fambrough (Johns Hopkins University) were re-examined for their usefulness as isoform-specific probes. Three of them appeared to be specific for ␣1 and had different ranges of species cross-reactivity. One of these, 6F, was investigated in more depth. Another monoclonal antibody, XVIF9-G10, raised against dog cardiac microsomes in the laboratory of K. P. Campbell (HHMI, University of Iowa), was shown here to be specific for the Na,K-ATPase ␣3 subunit in the rat and to be a useful reagent for this isoform in cardiac tissue, where the McBX3 antibody failed to bind.
The data illustrating the location of the epitope for McBX3 were presented at the Biophysical Society Meeting in 1991 (9).
Other antibodies were used as controls in epitope mapping studies. A peptide-directed antiserum was developed against a sequence at the N terminus of rat ␣3 (GDKKDDKSSPKKS) by M. J. Caplan and R. Mercer (3). Ab 38 is a peptide-directed antiserum against a sequence near the FITC binding site in the middle of rat ␣3 (SIHETEDPNDNRYLLVM), developed by Y-M. Hsu and G. Guidotti (4). Peptide-directed antisera against a different sequence of sheep ␣1 near the N terminus (KKKAK-KERDMDELKK) and against RRRPGGWVEKE near the C terminus were developed by W. J. Ball, Jr. (11). Anti-KETYY, a peptide at the C terminus, was developed by R. Bayer and J. Kyte (12). K1 is an antiserum raised against rat ␣1 subunit cut from a gel (5); it has compo-* This work was supported by National Institutes of Health Grant HL 36271. 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.
nents that stain both the N-and C-terminal halves.
Gel Electrophoresis-Electrophoresis on polyacrylamide gels by the Laemmli protocol and Western blot were performed as described previously (7). Detection was by chemiluminescence using luminol-based reagents from Amersham Corp. or Pierce or by alkaline phosphatase with indoxy phosphate/nitro blue tetrazolium as substrate. Tricine gels were by the method of Schä gger and Von Jagow (16).
Protein Fragmentation and Sequencing-Tryptic fingerprinting was performed during gel electrophoresis, essentially by the method of Cleveland et al. (17). Trypsin does not normally work in the presence of SDS, but digestion is paradoxically facilitated by soybean trypsin inhibitor (18). Mild trypsin digestion of native enzyme in the presence of 15 mM K ϩ was performed as described by Jo ͞rgensen and Farley (19). More extensive digestion by trypsin to generate a Rb ϩ -stabilized 19-kDa fragment was performed as described by Karlish et al. (20). Tryptic digestion in the presence of 100 mM ammonium bicarbonate was performed as described by Ovchinnikov et al. (21).
Cleavage at Asp-Pro by overnight incubation in 70% formic acid was performed as described by Landon (4,22). Cleavage by V8 protease was done in the presence of a small amount of SDS because Na,K-ATPase was completely resistant to this protease without the detergent; details are provided in the figure legends.
For sequence analysis, protein fragments were electrophoretically blotted onto polyvinylidene difluoride membrane (Immobilon P, Millipore) in 10 mM CAPS buffer (pH 11), containing 10% (v/v) methanol. After staining with Ponceau S, protein bands were excised from the membrane and washed extensively with MilliQ water. Amino acid analysis and Edman degradation were done at the Microchemistry Facility at Harvard University.
PCR and DNA Sequencing-Total RNA from rat brain or 10 -13-dayold primary cultures of rat cardiac myocytes was prepared with RNAgents Total RNA Isolation System (Promega). The procedure is based on guanidine thiocyanate disruption of nucleoprotein complexes followed by phenol/chloroform extraction of intact RNA. The cDNA was obtained using 2-3 g of total RNA, oligo(dT) as the priming oligonucleotide, and avian myeloblastosis virus reverse transcriptase (Promega).
The polymerase chain reaction (PCR) was performed in 10 l of buffer containing 50 mM Tris-Cl (pH 8.3), 2 mM MgCl 2 , 0.25 mg/ml bovine serum albumin, 200 M dNTPs, the 5Ј-and 3Ј-specific primers (100 -200 nM each), and 1 unit of Taq polymerase. PCR products were separated by electrophoresis in 1.0% low melting agarose in Tris acetate-EDTA (TAE) buffer. Bands of the expected size were excised and purified using the Magic PCR Preps DNA Purification System (Promega).
Enzymatic and Chemical Modification of the McBX3 Binding Site-Deglycosylation was performed as follows. 100 g of Na,K-ATPase resuspended at 2 mg/ml in 0.25 M sodium phosphate buffer (pH 7.4) was denatured first with SDS and ␤-mercaptoethanol (at final concentrations of 0.1 and 1%, respectively) for 30 min at 37°C, followed by addition of 1-2 units of glycosidase F (Boehringer Mannheim). In some experiments up to 0.5% Triton X-100 or 1% Nonidet P-40 was added to the reaction mixture in order to preserve glycosidase activity. The reaction was performed overnight at 37°C. The samples were precipitated with ice-cold methanol and redissolved in SDS electrophoresis sample buffer.
Galactosylation was performed according to Ref. 24, with 100 g of Na,K-ATPase resuspended in a buffer containing 50 mM Tris-Cl (pH 7.2), 5 mM MnCl 2 , 1 mM 5ЈAMP, 40 mM galactose, 0.2% aprotinin, and 5% glycerol. The mixture was incubated with 0.2 units of galactosyltransferase (Sigma) (autogalactosylated beforehand) for 30 min at 37°C followed by 60 min at 4°C. To stop the reaction, the samples were diluted with 30 volumes of ice-cold buffer, containing 25 mM imidazole and 1 mM EDTA, and centrifuged at 100,000 ϫ g (Beckman, SW 50.1 rotor) for 60 min at 4°C. Pellets containing membrane-bound modified enzyme as well as control pellets were resuspended in sucrose/EDTA/ Tris buffer, and samples were dissolved in gel buffer and run on a 10% Laemmli gel.
Treatment with hydroxylamine was performed both under basic and neutral conditions in different experiments (25).
Tissue Sections and Immunofluorescence-Pieces of fresh rat retina were fixed by immersion in PLP (paraformaldehyde/lysine/periodate) for 1 h at room temperature, followed by 30 min in lysine PO 4 , and then were soaked in 30% sucrose overnight at 4°C. Then they were mounted in Tissue-Tek, frozen in liquid nitrogen, and 10-m sections were cut on a cryostat. Dried sections were permeabilized with 0.3% Triton X-100 in Tris-buffered saline, 5% goat serum, for 30 min at room temperature, stained, and examined by epifluorescence microscopy as described previously (26). Fig. 1, A-E, compares the reactivities of five different monoclonal antibodies to Na,K-ATPase ␣ subunits with preparations of membranes from rat renal A-E, blots were prepared from SDS gels containing 1-3 g of protein/ lane for kidney (K) and axolemma (Ax) samples, and 35-50 g of protein/lane for adult rat heart (AH) and newborn rat heart (NbH) samples. Hybridoma supernatants were used for staining at dilutions of 1:10 to 1:50, and detection was by alkaline phosphatase-conjugated secondary antibodies. Below the blots is shown the Na,K-ATPase ␣ isoform composition of each tissue as determined by other means (see text). F, blots were prepared for samples of membranes from rat brain (RB), macaque brain (MB), and macaque heart (MH); 3 g of protein for brain and 40 g for heart. G, blots were prepared for samples of membranes from rat brain (B), 3 g/lane, and newborn rat heart (H), 50 g/lane. The ␣3-specific anti-peptide antisera GDKKD and SIHET recognize segments near the N terminus and the center of the ␣3 polypeptide, respectively. medulla, brainstem axolemma, adult cardiac ventricle, and newborn cardiac ventricle. All of the tissues have some ␣1, but there is very little in brainstem axolemma. Only axolemma and adult ventricle have ␣2, while only axolemma and newborn ventricle have ␣3. The antibodies McK1, McB2, and McBX3 ( Fig. 1, A, C, and D) were originally shown to be specific for ␣1, ␣2, and ␣3 by several lines of experimental evidence. First, they recognized proteins on gels with the same mobilities as each isoform translated in an in vitro reticulocyte expression system (8). Second, the presence of reactive bands correlated with the presence of each mRNA species in brain (␣1, ␣2, and ␣3), skeletal muscle (␣2 and ␣1), and kidney (␣1) (1). Last, expression of rat ␣1 in Madin Darby canine kidney cells resulted in detection of protein by McK1 4 ; expression of ␣2 mRNA in differentiating 3T3-L1 cells correlated with the appearance of an McB2-reactive ␣ subunit (27); and McBX3 antibody stained retinal photoreceptors (28), which had been shown to express exclusively ␣3 mRNA (29). Fig. 1, B and E, shows the reactivities of two new antibodies. Antibody 6F was similar to McK1 in recognizing only ␣1. Antibody XVIF9-G10 recognized only ␣3. A clear difference between McBX3 and XVIF9-G10, however, was that XVIF9-G10 reacted with ␣3 in newborn rat heart, while McBX3 ( Fig. 1D) inexplicably did not. Fig. 1F illustrates that this peculiarity of McBX3 was not unique to the rat heart. Macaque brain and cardiac ventricle samples reacted equally well with a peptide-directed antiserum specific for ␣3, anti-GDKKDDKSSPKKS, while McBX3 recognized the ␣3 only in macaque brain, not heart. Similar results were obtained with dog and human heart samples (data not shown); these species are known to express ␣3 in the heart (30 -32). In Fig. 1G it can be seen that antibody XVIF9-G10 and two peptide-directed antisera specific for ␣3, anti-GDKK-DDKSSPKKS and anti-SIHETEDPNDNRYLLVM, all recognized newborn rat cardiac ␣3.

Comparison of Antibodies-
The specificity of the antibodies 6F, XVIF9-G10, and McBX3 was studied to better understand their utility as probes and to investigate the apparent tissue-specific heterogeneity in the ␣3 isoform.
␣1-Specific Antibodies McK1 and 6F-The McK1 binding site was originally mapped to the first 32 amino acids of rat ␣1 by proteolytic fragmentation, and its epitope was predicted to be DKKSKK based on sequence differences among species (7). The antibody was shown to have greatly reduced affinity for mouse and human ␣1, which have DKKGKK at that site, and to have negligible affinity for pig and sheep ␣1, which have a twoamino acid deletion (SK) at that site and the sequence DKKK instead. No other sequence differences were present in the first 32 amino acids that could account for the antibody's specificity. The accuracy of the epitope identification was confirmed in the experiment shown in Fig. 2. Deletions of ␣1 and chimeras of ␣1 and ␣2 were tested for reactivity with McK1 in Western blot. McK1 reacted well with full-length ␣1, but it failed to react when the first 16 or 32 amino acids were deleted, when the first 27 amino acids were replaced by 25 from ␣2, or when the first 15 amino acids were replaced by 14 from ␣2. Most persuasively, it failed to react when the only alteration of ␣1 was the replacement of amino acids GDKKS (four amino acids from the predicted core of the epitope) by NGGGK from ␣2. Antibody 6F, which was shown in Fig. 1 to react with only ␣1 in the rat, was one of several monoclonal antibodies raised against chicken Na,K-ATPase in the lab of D. M. Fambrough. Two others also appeared to be specific for ␣1 on the basis of tissue reactivity and the electrophoretic mobility of the bands recognized (data not shown), but one (2F) reacted very poorly with rat Na,K-ATPase, and the other (3B) was weaker than 6F and, unlike 6F, was negative in ELISA against rat kidney membranes (data not shown). A fourth antibody, ␣5, was not isoform-specific. The reactivity of these antibodies with Na,K-ATPases from a variety of species is included in Table I. It can be seen that 6F recognizes ␣1 in a number of species in which McK1 does not bind.
Proteolytic fingerprinting narrowed the location of the 6F epitope (Fig. 3). Proteolysis occurs during gel electrophoresis when trypsin is mixed with soybean trypsin inhibitor (18). The protease-inhibitor complex is stabilized against SDS denaturation, allowing digestion of the ␣ subunit to occur at selected sites that bear no relationship to the trypsin-sensitive sites of native enzyme. When the resulting fragments were probed with antibodies known to bind near the N terminus, identical fingerprints were obtained for ␣1 (McK1), ␣2 (McB2), and ␣3 (GDKK) (Fig. 3A). In the same experiment, antibody 3B also gave the same fingerprint, whereas antibody ␣5 stained different fragments. Antibody ␣5 is independently known to bind somewhere between amino acids 554 and 715 from its ability to recognize Na,K-ATPase-H,K-ATPase and Na,K-ATPase-Ca 2ϩ -ATPase chimeras (34,35). In a different experiment, 6F stained the same set of fragments as McK1 (Fig. 3B). The smallest fragment stained by both antibodies was 17.5-kDa, or approximately 160 amino acids. The 6F epitope must consequently lie within this distance of the N terminus.
Of the nine species whose kidney or electric organ Na,K-4 R. Levenson and K. J. Sweadner, unpublished data.

FIG. 2.
McK1 reactivity with ␣1/␣2 chimeras. Deletions of ␣1 and chimeras of rat ␣1 and ␣2 were made by Dr. L. K. Lane, University of Cincinnati, to investigate the functional role of the N-terminal sequence difference between isoforms (33). They were cloned into a pRc/CMV vector and expressed in HeLa by stable transfection. Membrane preparations treated with NaI were used for Western blots at 10 g of total protein/lane by Drs. A. Therien and R. Blostein, Montreal General Hospital, and the blots were stained with McK1. Of the constructs tested, only intact ␣1 was stained, although all of the constructs showed comparable expressed ATPase activity. The figure shows the sequence beginning at glycine, which is amino acid 6 if the initiator methionine is termed number 1. Normally, both ␣1 and ␣2 are processed during biosynthesis to remove the first five amino acids (MGKGV) including the methionine; that is assumed to occur also during expression of the deletions and chimeras but has not been experimentally verified.
ATPases were tested for reactivity with 6F, only sheep ␣1 and Xenopus ␣1 failed to react (Table I). The amino acid sequences of the tested ␣1s were aligned to look for differences that could plausibly cause this species specificity, as well as the lack of recognition of rat ␣2 and ␣3, as shown.
The reactive pig enzyme differs from the nonreactive sheep enzyme at only three amino acids in the region defined by the fingerprint. The only credible candidate for the epitope site is a substitution at amino acid 55; serine is found in most reactive species, but asparagine replaces it in sheep. However, the sequence of Torpedo Na,K-ATPase is threonine-glutamine, and yet it reacts with 6F reasonably well. It is plausible that threonine-glutamine is a more conservative substitution for serinearginine than asparagine-arginine and that it could be accommodated in the antibody binding site. Xenopus ␣1, rat ␣2, and rat ␣3 have even more different substitutions. Although the lysine in ␣2 might be expected to be tolerated as well as glutamine as a substitute for arginine, valine as a substitute for threonine may be detrimental because of the loss of a polar hydroxyl group.  Fig. 1. To further determine their species and isoform specificity, they were tested in Western blots against membrane fractions from brain, kidney, and electric organ of a number of animal species for which Na,K-ATPase cDNA sequences are known. Table I shows the results. McBX3 reacted well with ␣3 in brain samples from rat, human, and chicken. In contrast, XVIF9-G10 reacted well with ␣3 in rat and human samples but not at all with that from chicken. XVIF9-G10 did not react with any of the ␣1-containing samples. 5 Proteolytic fingerprinting by trypsin-soybean trypsin inhibitor complex during SDS-gel electrophoresis was used to map the location of the XVIF9-G10 epitope (Fig. 3C). The fingerprint stained by XVIF9-G10 was indistinguishable from that of the ␣3 anti-peptide antibody (anti-GDKKDDKSSPKKS) that binds near the N terminus and completely unrelated to the fingerprint of the Ab 38 ␣3 anti-peptide antibody (anti-SI-HETEDPNDNRYLLVM), which binds near the middle of ␣3. The 17.5-kDa smallest fragment stained by XVIF9-G10 encompasses not only the N-terminal variable region but also transmembrane segments M1 and M2, with the short, extracellularly facing variable region between them.
The next experiments indicated that the site of XVIF9-G10 binding is close to the N terminus. In Fig. 4A, trypsin was used to digest purified rat brain Na,K-ATPase for different lengths of time. Binding of McBX3 (shown below to bind in the middle of the protein) was affected very little. When a replicate blot was stained with XVIF9-G10, it could be seen that its epitope was almost completely removed at the earliest time point, when there was no detectable change in the size of ␣3. A shift in molecular weight of 5000 in the McBX3-stained band could easily have been detected, implying that the XVIF9-G10 epitope lies within 50 amino acids of the N terminus. In Fig. 4B a similar experiment was performed, except that rat axolemma membranes were used, with a lower concentration of trypsin that resulted in slower digestion. Three replicate blots were prepared and stained with McBX3, XVIF9-G10, and the peptide-directed antiserum against ␣3 (anti-GDKKDDKSSPKKS, at the N terminus). The loss of stain for each antibody was quantified by densitometry. The results indicated that there was very little loss of McBX3 binding during the digestion, but staining by XVIF9-G10 and the anti-peptide antibody declined in parallel.
The fact that XVIF9-G10 binds close to the N terminus and that it reacts with rat and human ␣3 but not with chicken ␣3 (Table I) allows prediction of the location of the epitope. Sequence comparison of rat, human, and chicken ␣3 at the N terminus shows that chicken has sequence differences, particularly deletions, that could plausibly account for the failure of XVIF9-G10 to bind. 5 It should be noted that in the laboratory where XVIF9-G10 was developed, it was observed to stain rabbit and human skeletal muscle by immunofluorescence (C. Leveille and K. P. Campbell, unpublished observations). In our experiments (not shown), XVIF9-G10 did not react with Na,K-ATPase in rat skeletal muscle. ␣3 is not known to be expressed at significant levels in rat skeletal muscle, but human skeletal muscle has been observed to contain ␣3 as a minor component (36), and the pattern of immunofluorescent stain with XVIF9-G10 suggested that only certain muscle fibers were positive. Rabbit tissues have not been studied for ␣ isoform content to our knowledge, but the immunofluorescence seen with XVIF9-G10 indicated extensive labeling, approximately equivalent to that seen with an antibody to ␣2 (37). Further investigation is needed to determine whether XVIF9-G10 reacts with ␣2 in this species or whether skeletal muscle in the rabbit expresses ␣3.

FIG. 3. Tryptic fingerprints; digestion in SDS.
A, rat brain microsomes were the source of all three Na,K-ATPase isoforms for this experiment. Trypsin mixed with soybean trypsin inhibitor was added to the samples marked ϩ before electrophoresis, and digestion occurred during electrophoresis. The resulting blots were stained with the indicated monoclonal antibodies and with the ␣3-specific antiserum GD-KKDDKSSPKKS. In most of the blots, prestained molecular weight markers are visible on the left, with molecular masses of 116, 84, 58, 48.5, 36.5, and 26.5 kDa (Sigma 7b). B, 6F was from a different experiment in which rat kidney microsomes were the substrate for digestion. With heavier gel loading, a broader range of fragment sizes was seen, but the smallest ones were indistinguishable from those stained by McK1, McB2, 3B, and GDKK. C, similarly digested rat brain microsomes were stained with anti-GDKKDDKSSPKKS, anti-SIHETED-PNDNRYLLVM, and monoclonal antibody XVIF9-G10.

MGDKKDDKSSPKKSKAKERRDLDD
This segment overlaps with the peptide used to produce the polyclonal anti-␣3 antibody (anti-GDKKDDKSSPKKS). It is notable that we have already observed that this polyclonal antibody fails to recognize chicken ␣3 (32). Sequence differences are also found at the H1-H2 extracellular loop, but cleavage that far from the N terminus would have been detected as a size change in Fig. 4. Antibody McBX3: Epitope Mapping and Characterization-Although McBX3 reacted as expected with ␣3 from rat, human, and chicken brain (Table I), a more complex picture emerged when it was tried on a panel of brain microsomes from additional mammalian species. Preferential reaction with ␣3 was seen in rat, human, and mouse brain samples, accompanied by weak cross-reactivity for brain ␣1. For rabbit, guinea pig, hamster, sheep, and calf brain samples, however, reactivity for ␣3 was weaker and reactivity for ␣1 was stronger, resulting in the approximately equal staining of two well-separated bands on blots of gels (data not shown). It was obvious that McBX3 was not ␣3-specific in all species, although staining at the position of ␣2 (intermediate between ␣1 and ␣3) was never seen. McBX3 was also tested against kidney and electric organ samples as sources of ␣1 (Table I). It reacted well with the ␣1 subunit from chicken, pig, and sheep. When the amino acid sequences of the Na,K-ATPases tested with McBX3 were aligned to look for differences that could account for the species and isoform specificity, no sequence could be found.
The observation that a monoclonal antibody was capable of recognizing Na,K-ATPase ␣3 subunit in one tissue (brain) but not another (heart) suggests the existence of an underlying isoform heterogeneity that was not suspected before. Such heterogeneity could arise from either mRNA splicing or from protein modification. Protein fragmentation was used to determine the approximate location of the epitope to facilitate molecular analysis.
Initial experiments, performed on Na,K-ATPase purified from the rat brainstem axolemma (enriched in ␣3), showed that McBX3 bound in the C-terminal half of the protein. Cleavage at asparagine-proline (which is found at three sites in rat ␣3) by formic acid produced a 40-kDa fragment from the C-terminal end of ␣3 that was stained by McBX3 (data not shown). Cleavage at tryptophan residues with N-chlorosuccinimide produced a family of McBX3-reactive large fragments that were also derived from the C-terminal half of the protein (data not shown). Further mapping experiments with McBX3 were performed with pig kidney ␣1, a strongly reactive Na,K-ATPase, because of the greater ease of preparing large quantities of purer enzyme. In pig ␣1, the McBX3 epitope was also in the C-terminal half; Fig. 5 illustrates the fragments generated by cleavage at the single Asp-Pro site (amino acids 586 -587, pig ␣1 numbering). A polyclonal antibody (K1) against ␣1 stained both the 60-kDa N-terminal and 40-kDa C-terminal fragments. A peptide-directed antiserum against the N terminus stained the 60-kDa fragment, and McBX3 stained the 40-kDa fragment. Other minor fragments, most of which were stained by both antibodies, are artifacts resulting from the heavy loading of the lanes containing digested enzyme.
Digestion of pig kidney ␣1 with trypsin was used next to map the site (Fig. 6). Digestion in the presence of K ϩ is known to cleave ␣1 initially at Arg-438 (a site named T1) and more slowly at Lys-30 (T2) (19). The band labeled 1 is the 55-kDa C-terminal fragment and that marked 5 is the 40-kDa Nterminal fragment, shown stained here with antibodies against the N and C termini. We observed that upon more extensive digestion, two additional cleavages on the C-terminal side of T1 occurred (bands marked 2 and 3), and we used N-terminal sequence analysis of the smaller and more stable 42-kDa fragment (band 3) to determine that the cleavage was at Arg-589 (N-terminal sequence AAVPDAVGK). This fragment was positive for McBX3, as well as for a peptide-directed antiserum against the C terminus, indicating that the epitope lies between Ala-590 and the C terminus. A small amount of a 19-kDa fragment can also be seen in (labeled band 4), stained only by the antibody against the C terminus; this was characterized further below.
Extensive digestion by trypsin in the presence of Rb ϩ has been shown to result in the removal of most of the cytoplasmic portion of the ␣ subunit, leaving a complex of transmembrane hairpins in the membrane plus a larger fragment beginning at

K-ATPases from different species and tissues
Membrane preparations from the indicated sources were electrophoresed on SDS gels, blotted to nitrocellulose, and stained with monoclonal antibodies. The relative intensity of stain ranged from strong (ϩϩϩ) to barely detectable (ϩ/Ϫ); Ϫ indicates no detectable stain.
The H,K-ATPase of gastric mucosa was detected with antiserum HK-9 (a generous gift of M. J. Caplan (Yale)) as a positive control. For those wishing to detect the Na,K-ATPases of Torpedo or Xenopus, an antibody designated 2F12 (courtesy of M. Schachner, ETH Zü rich) gave good reaction on Western blots (data not shown). 2F12 reacted also with rat ␣2 and ␣3, but did not recognize chicken ␣1.
Asn-831 (NPKTD) and ending at the C terminus ("19-kDa membranes" (20)). Fig. 7 shows the staining of pig kidney enzyme digested in this way, after resolution of small fragments on a Tricine gel. An antiserum against an N-terminal peptide stained only large fragments, products of incomplete digestion. An antiserum against the C terminus prominently stained the 19-kDa fragment (as shown here, this fragment's apparent molecular weight is sometimes higher, depending on the gel and the M r standards used; the calculated M r is approximately 21,000). McBX3, on the other hand, stained minor fragments at approximately 21 and 15-kDa, but did not stain the 19-kDa fragment at all. This restricts the epitope site to between Ala-590 and Asn-831, eliminating the last 186 amino acids of the C-terminal end of ␣1.
Tryptic digestion in the presence of NH 4 ϩ is known to produce cleavages at some sites that are protected in Rb ϩ , and so it was tried as an alternative method. This resulted in the release of a 27-kDa soluble fragment that reacted with McBX3 (Fig. 8,  band 1). Inspection of the sequence of ␣1 and the position of transmembrane spans indicates that the only place that such a large soluble fragment could originate is the large intracellular loop between H4 and H5. This loop, which extends from Lys-342 to Lys-767, includes the first 164 amino acids of the positive formic acid-generated fragment beginning at Pro-587, as well as the tryptic fragment beginning at Ala-590, thus narrowing the epitope location to between Ala-590 and Lys-767. Not all of the epitope was solubilized in this procedure; the membrane-bound fraction (Fig. 8, band labeled 2 in lane 3) contained the same McBX3-reactive 42-kDa band seen in Fig. 6 above.
The 27-kDa fragment of Fig. 8 was the smallest one obtained in good yield with trypsin as the digesting enzyme, possibly due to a tryptic cleavage site in or near the antibody combining site.
Consequently, we attempted use of a protease with different specificity. In our hands, Staphylococcus aureus V8 protease, which cleaves preferentially at glutamate residues, did not digest detergent-free Na,K-ATPase at all, but extensive digestion of pig kidney Na,K-ATPase in 0.04% SDS allowed identification of a water-soluble 10.5-kDa fragment on Tricine gels that was still reactive with McBX3. Fig. 9 shows that numerous fragments were obtained because of incomplete digestion, but the smallest stained fragment, resolved on Tricine gels, was used for sequencing. The absence of smaller stained fragments implies the presence of cleavage sites in the epitope, since the gel system can resolve peptides as small as 3 kDa when the acrylamide has a high percentage of cross-linker. Cleavage was at Glu-560, giving the N-terminal sequence XFQXDTDDVN, corresponding to (E)GFQFDTDDVN. This fragment overlaps with the sequence 590 -767 deduced above, since it is predicted, based on size, to extend from Gly-561 to Glu-669. Sequences of FIG. 4. Tryptic digestion of ␣3 removes the XVIF9-G10 binding site. In this figure, trypsin was used to lightly digest undenatured, SDS-purified enzyme. A, rat brain Na,K-ATPase was digested by trypsin at a ratio of 150:1 in 15 mM K ϩ , sucrose/Tris/EDTA buffer, 37°C. Trypsin inhibitor was added, and gel samples were acidified to prevent further digestion during gel electrophoresis. Lane 1, molecular weight markers (Sigma 7B); lane 2, undigested control; lanes 3-5, digestion for 1, 5, and 10 min. B, rat axolemma membranes were digested by trypsin at a ratio of 200:1 in the same medium at 37°C, and digestion was stopped at 30 s, 1, 3, and 10 min. The amount of antibody stain remaining at the position of undigested ␣3 was quantified with a Molecular Dynamics 300A scanning densitometer. two minor comigrating peptides could be followed in this experiment: XRKIXEFXXX, corresponding to (E)QRKIVEFTCH at Glu-907, which could be ruled out as a candidate for McBX3 binding because it lies within the 19-kDa fragment of Fig. 7, and XFGLGGYPGF, corresponding to (E)YFGLGGYPGF at Glu-228 of the Na,K-ATPase ␤ subunit.
With knowledge of the general location of McBX3 binding, we were able to ask whether alternative splicing of the ␣3 mRNA could account for the failure of McBX3 to recognize ␣3 in the heart. The region of the epitope corresponds to nucleotides 1900 -2220 in rat ␣3 (23). This includes sequence from exons 13 to 15. Fig. 10 shows a schematic of the rat brain ␣3 cDNA for this region and marks the location of introns (as determined for human ␣3 (38)). Two different primer pairs were chosen to permit isolation of ␣3-derived DNA (avoiding ␣1 or ␣2) by PCR. cDNA was prepared from rat brain and from newborn rat cardiac myocytes. Both primer pairs resulted in the production of DNA fragments of identical sizes from the two tissue sources. A panel of seven restriction enzymes (indicated in Fig. 10) was tested for their ability to detect differences between the PCR products from brain and heart; identical patterns were obtained for all seven (data not shown). Finally, DNA sequence was obtained for DNA fragments isolated by PCR from the cardiac myocyte preparation. For the locations indicated in Fig.  10, the sequence was identical to that already reported for rat brain ␣3. The data showed conclusively that authentic ␣3 is expressed in newborn rat cardiac myocytes. Because the PCR products straddled introns 8 through 18, the data rule out alternative splicing of exons 13-15. The forward and reverse primers were all in different exons, making it very unlikely that an alternatively spliced gene product would have been missed.
If alternative splicing is ruled out, there are two complementary hypotheses to explain the properties of McBX3: that unmodified protein contains the epitope and modification blocks the binding in negative tissues, or that a protein modification is found only in positive tissues. Attempts were made to remove candidate protein modifications to see whether McBX3 binding could be abolished or restored.
The catalytic subunit of the dog kidney Na,K-ATPase has recently been demonstrated to contain covalently linked carbohydrate, both an N-linked N-acetylglucosamine residue and unidentified O-linked sugars (24,39). Rat kidney and brain FIG. 7. McBX3 does not stain the 19-kDa C-terminal fragment. More extensive tryptic digestion in Rb ϩ maximized the recovery of the 19-kDa fragment. Pig kidney Na,K-ATPase, resuspended at 1.5 mg/ml in a medium containing 25 mM imidazole (pH 7.4), 10 mM RbCl, and 1 mM EDTA, was incubated with trypsin at a ratio of 10:1 for 1 h at 37°C. The reaction was stopped by the addition of 3 mM diisopropylfluorophosphate in a buffer containing 25 mM imidazole, 2 mM RbCl, and 1 mM EDTA, and the sample was centrifuged at 100,000 ϫ g for 1 h at 4°C. The pellet, "19 kDa membranes," was washed twice and finally resuspended in the same buffer. Three replicate blots of digested and control enzyme were stained with antisera against the N and C termini (anti-KKKAKKERDMDELKK and anti-KETYY, respectively) and McBX3. The markers were Rainbow (Amersham Corp.), low molecular weight range.
FIG. 8. McBX3 stains a large soluble fragment. Pig kidney Na,K-ATPase (1 mg/ml in 100 mM NH 4 HCO 3 buffer (pH 7.4)) was incubated with trypsin at a ratio of 100:1 for 10 min at 37°C, and the reaction was stopped by addition of soybean trypsin inhibitor. The sample was diluted with 10 volumes of cold 100 mM NH 4 HCO 3 and centrifuged at 100,000 ϫ g (Beckman, SW-50.1 rotor) for 90 min at 4°C. The supernatants were pooled and lyophilized, and the pellet was washed twice with the same buffer; both were finally resuspended in sucrose/EDTA/ Tris buffer. The gel was a Tricine gel. Lane 1, undigested control; lane 2, soluble fragments from the supernatant; lane 3, membrane-associated fragments. Two different molecular weight marker mixtures were used as in Fig. 6.   FIG. 9. V8 protease digestion fragments stained by McBX3. Purified pig kidney ATPase was digested with endoproteinase V8 from Staphylococcus aureus (Boehringer Mannheim). Na,K-ATPase, resuspended in 100 mM NH 4 HCO 3 (pH 7.4) containing 0.04% SDS, was incubated with V8 protease at a ratio of 50:1 for 30 min at 37°C. The reaction was stopped by addition of 3 mM diisopropylfluorophosphate (Sigma) in 100 mM NH 4 HCO 3 , and water-insoluble fragments were removed by centrifugation at 100,000 ϫ g (Beckman, SW 50.1 rotor) for 90 min at 4°C. The pellet containing membrane-bound fragments was washed twice with the same buffer. The supernatants containing watersoluble fragments were pooled, lyophilized, redissolved in water, and subjected to additional digestion with V8 protease at a final concentration of 4%. To stop the reaction, 3 mM diisopropylfluorophosphate was added after 2 h of incubation at 37°C. Fragments from the supernatant were resolved on a Tricine gel, and the blot was stained with McBX3. The molecular weight markers were Sigma 17 S, with sizes indicated on the figure. The smallest McBX3-stained fragment (marked with an asterisk) was cut from unstained portions of the same blot and sequenced.
Na,K-ATPase ␣ subunits have also been reported to react with certain sugar-specific lectins (40). Curiously, the N-linked sugar (N-acetylglucosamine) in dog kidney Na,K-ATPase was found to be on the cytoplasmic surface (39). Such a post-translational modification would be a plausible candidate for McBX3 immunoreactivity, since glycosylation frequently differs in different tissues, and dog kidney ␣1 is one of the Na,K-ATPases that reacts well with McBX3. There is a consensus sequence for N-linked glycosylation in the region 590 -669: N631 (NETVED), a highly conserved site. It is unlikely that N-acetylglucosamine is the epitope for McBX3, however, for two reasons. Treatment of pig kidney Na,K-ATPase with peptide Nglycosidase F removed carbohydrate from the ␤ subunit, as demonstrated by a shift in its gel electrophoretic mobility, but it had no effect on the ability of McBX3 to react with the ␣ subunit (data not shown). Conversely, treatment of pig kidney Na,K-ATPase with galactosyltransferase and UDP-galactose, which was shown by Pedemonte and Kaplan (24,39) to introduce approximately stoichiometric amounts of galactose into the N-linked N-acetylglucosamine group, also did not affect McBX3 binding (data not shown). Either or both of these manipulations would be expected to affect antibody binding. If, conversely, antibody bound to unmodified protein in rat brain and pig kidney and N-acetylglucosamine blocked the site on ␣3 in heart, then dog kidney microsomes, which have the sugar group, should have been negative for McBX3. The data do not rule out glycosylation of a different nature, however, such as O-linked glycosylation.
There is another consensus sequence for N-linked glycosylation in the region of the epitope, NHT is unique to rat ␣3 at Asn-670. However, treatment of rat cardiac ventricle membranes with peptide N-glycosidase F did not increase McBX3 binding nor was there any effect of growing cardiac myocytes in tunicamycin (data not shown).
The next experiment indicates that McBX3 recognizes a modification in pig ␣1, rather than unmodified protein, and suggests that the modification is linked to the protein by an ester bond. McBX3 reactivity was abolished by treatment of the protein with basic hydroxylamine (Fig. 11). The same concentration and pH of Tris buffer had no effect on antibody binding (data not shown). Under the conditions used, hydroxylamine cleaves polypeptide chains at asparagine-glycine (25). Asn-Gly dipeptides occur at amino acids 167-168, 202-203, and 869 -870 in pig ␣1. Staining of a replicate blot with an antiserum against an epitope near the N terminus demonstrated that protein fragmentation occurred, and the fragments were present in high enough yield to be stained. The two fragments of 19 and 23 kDa stained by the N-terminal antibody probably represented cleavage at 168 and 203. In other experiments, a fragment of 89-kDa, corresponding to peptide 204-1016, was stained by anti-KETYY against the C terminus (data not shown). This, or a fragment of approximately 73 kDa predicted to contain the McBX3 epitope (amino acids 204 -869), should have been easily visualized in Fig. 11 given the strong stain of the untreated control.
Validation of the ␣3 Specificity of XVIF9-G10 and McBX3-Because the McBX3 antibody has been used to localize ␣3 in the nervous system using immunofluorescence (26,28), but is now known to not react with all ␣3 subunits, it was of considerable interest to determine whether XVIF9-G10 would detect anything different. Cryostat sections of the retina were stained with McBX3 and XVIF9-G10, and the same pattern of stain FIG. 11. Hydroxylamine abolishes McBX3 binding. 100 g of pig kidney Na,K-ATPase in sucrose/EDTA/Tris buffer was mixed with an equal volume of 2 M hydroxylamine (pH 10.5) and incubated for 24 h at 37°C. The resulting peptide fragments were precipitated with 12.5% trichloroacetic acid, and redissolved in SDS-electrophoresis sample buffer, and electrophoresed on a 10% acrylamide Laemmli gel. The molecular weight markers were from Bio-Rad. Replicate blots were stained with an antiserum against the N-terminal peptide KKKAK-KERDMDELKK and with McBX3. The bands marked with asterisks are the most abundant small N-terminal fragments.
FIG. 10. Diagram of PCR results. A, schematic map of rat ␣3 cDNA. The hatched box represents the mapped region of the epitope. Here and throughout the text, the numbering used for the ␣3 nucleotide sequence is according to Shull (23). The position of introns is indicated with vertical arrows (38). The ␣3 genespecific oligonucleotides that were used as PCR and sequencing primers are shown by horizontal arrows. Below them are the four fragments obtained in PCR amplifications of both brain and cardiac myocyte cDNA. B, the lower half of the figure zooms in on the sequence from nucleotides 1767 to 2656. Restriction sites FokI (F), HinfI (H), BauI (B), NlaIV (N), Sau3aI (S), AluI (A), and BstY (BY) are indicated; identical maps were obtained from PCR products from brain and heart cells. Below the map is the region of interest defined by protein fragmentation; below that is illustrated the portions of the heart-derived PCR product that were confirmed directly by sequencing.
was seen (Fig. 12, A and B). Similar results were obtained in sections of the cerebellum (data not shown), where the Purkinje neurons have characteristically high levels of ␣3 mRNA (41). In both retina and cerebellum, the distribution of ␣1 and ␣2 is distinctively different from that of ␣3 (26,28). Thus, both monoclonal antibodies recognize known concentrations of ␣3 in the nervous system, validating the previous localization studies, and no new ␣3-containing cells were observed.

DISCUSSION
Specificity of the Antibodies-The ion transport ATPase gene family is fairly large and still expanding, but the common structural features of the ␣ subunits permit the prediction that three isoform-specific antibodies (McK1, McB2, and XVIF9-G10) will not cross-react with other ATPases. For all of the Na,K-ATPases and H,K-ATPases so far described, sequence homology begins 25-35 amino acids after the initiator methionine. Although the N-terminal variable segments tend to have a high proportion of positively charged amino acids, there is little real homology apart from short stretches of amino acids. Sequence variation between invertebrate Na,K-ATPase ␣ subunits is even higher at the N terminus than for vertebrate Na,K-ATPases and includes some large differences in total length. It is even less likely, therefore, that the isoform-specific antibodies described here will work in invertebrates.
All of the specific antibodies (McK1, McB2, 6F, and XVIF9-G10) had low backgrounds when used on Western blots of membrane preparations. All of the antibodies recognized enzyme in ELISA (data not shown), but binding to plastic can cause partial denaturation. McK1, McB2, and XVIF9-G10 all bind in the first 30 amino acids, and all were useful for immunocytochemistry in conventionally fixed tissue. 6F is predicted to bind around amino acid 55 and was not (data not shown); perhaps this site is buried in native enzyme. This antibody has been shown to stain fixed sections of the rat kidney faintly, but the staining is considerably increased by prior treatment of the sections with SDS (42).
It is interesting that another monoclonal antibody, M7-PB-E9, has been mapped to a site within the same region as McBX3 (43). This antibody, like McBX3, reacts with rat ␣3 and sheep and pig ␣1 but much less well with rat ␣1. The epitope was deduced to include Ser-646 and Asp-652, based on the comparison of sequences for rat ␣1, ␣2, and ␣3, and ␣1s from sheep, pig, chicken, Xenopus, and Torpedo. The binding of McBX3 matches this specificity exactly if only these sequences are considered, but examination of two additional sequences eliminates this site for McBX3: dog ␣1 has a substitution Arg-646, which should be even more disruptive than the Asn-646 found in rat, Xenopus, and Torpedo ␣1s, and yet it is recognized well, and human ␣1 has the same sequence as sheep and pig ␣1 and rat ␣3, and yet it reacts weakly with McBX3.
The similarity between the two antibodies is striking, however, and it would be desirable to know if M7-PB-E9 reacts with the ␣3 in the heart. The deduced location of the epitope is in the vicinity of the binding site for ATP. M7-PB-E9 partially inhibits ATPase activity and stimulates the rate of ouabain binding and has been concluded to enhance the E1 to E2 conformational transition (44). Similar studies have not been done for McBX3, which is an IgM of lower affinity.
The region where both antibodies bind is apparently wellexposed on the surface of the protein for stimulating the immune system. If M7-PB-E9 actually recognized the same posttranslational modification as McBX3, it would suggest that the structure (whether injected as rat ␣3 or sheep ␣1) has excellent immunogenicity.
Protein Fragmentation of the Na,K-ATPase-It is shown here (Fig. 3) that proteolytic fingerprinting with trypsin/soybean trypsin inhibitor gives very similar patterns for the three Na,K-ATPase isoforms when probed with antibodies that bind near the N terminus. It is notable that the smallest N-terminal fragments are 17.5 kDa and that the epitopes near the N terminus are not cleaved off despite the high content of positively charged amino acids. The existence of digestion hot spots that are shared by the three isoforms is interesting in view of the fact that trypsin digestion of undenatured ␣1, ␣2, and ␣3 does not proceed at the same sites despite the conservation of the sites described for ␣1 (45). The existence of reproducible digestion hot spots in the presence of SDS suggests that there may be regions where the detergent does not coat the protein uniformly, i.e. SDS association may be disrupted either by hydrophobic stretches (leading to alternative pseudo-micellar structures), or by post-translational modifications, or by particular amino acid sequences (clusters of positive or negative charge, or bulky amino acids). Cleavage to generate the 17.5-kDa smallest N-terminal fragment can be estimated to occur at KSSK, which is found immediately after the end of the 2nd hydrophobic transmembrane segment. The pattern of fragments seen with antibodies that bind near the center of the protein, such as Ab 38 (Fig. 3C), indicates that there are many cleavage sites away from the protected N-terminal region.
Limited tryptic digestion of membrane-bound ␣1 Na,K-ATPase has been a reliable method for producing defined protein fragments by ligand-sensitive cleavage at sites known as T1, T2, and T3 (19); the same cleavages occur in ␣1 from several different species. Here we show an additional cleavage of pig ␣1 at Arg-589 after digestion in the presence of K ϩ . Abbott and Ball (43) saw the same cleavage of sheep ␣1 in similar conditions and used it to help map antibody M7-PB-E9. We have suggested that the site be termed T4 (46).
Tissue-specific Modification of Na,K-ATPase Isoforms-The puzzling properties of antibody McBX3 emerged when attempts were made to detect ␣3 in the rat heart. In the newborn rat heart, a Na,K-ATPase ␣ form reactive with an antiserum (Ax2) raised against axolemma Na,K-ATPase was observed to be down-regulated during postnatal maturation (47), correlating with the disappearance of mRNA for ␣3 (48). The same  XVIF9-G10 (B) were used as primary antibodies on fixed sections of the rat retina, with goat anti-mouse secondary antibody coupled to rhodamine as a secondary antibody. Outer segments of the photoreceptors (OS), where rhodopsin is located, were not stained. Bright stain of the inner segments (IS) is characteristic of the high levels of ␣3 located there to pump out Na ϩ that enters the cell with the dark current. Photoreceptor cell bodies (P) are stained only very faintly by either antibody. Stain at the outer plexiform layer (OPL) varied from experiment to experiment with both antibodies but was usually detected. In the inner nuclear layer (ONL), bipolar cells are found mostly in the outer two-thirds, and these were stained with both antibodies. Poorly stained cells in the inner third of this layer are predominantly amacrine cells and Mueller glia. The inner plexiform layer (IPL) is the synaptic region for bipolar cells, amacrine cells, and the retinal ganglion cells, and it was stained uniformly and lightly by both antibodies. The distributions of ␣1 and ␣2 in the retina are quite different (28). band did not, however, react with McBX3, despite the fact that Ax2 antiserum and McBX3 stained the same cellular structures in the rat cerebellum (26). ␣3 protein in the newborn rat heart was later detected with a peptide-directed polyclonal antibody (15). Coca-Prados and colleagues (49) observed a similar discrepancy when ␣3 was detected with an anti-fusion protein antiserum in blots of the pars plana of the bovine ocular ciliary epithelium, but ␣3 was not seen in the pars plana by immunofluorescence with McBX3. In the adjacent pars plicata, both McBX3 and the antiserum worked, suggesting a tissuespecific difference in the McBX3 epitope.
The evidence presented here indicates that rat ␣3, and ␣1 in some species, can be post-translationally modified at the site of binding of McBX3. The effect of hydroxylamine in abolishing McBX3 binding suggests that the modification constitutes a principal part of McBX3's epitope and that the modification is either absent or structurally different wherever the antibody fails to bind. This appears to be the case in the heart of rat, dog, macaque, and human, all of which express ␣3 (32). The modification appears to be present in the kidneys of some species. We can speculate that it is absent in the bovine ciliary epithelium pars plana and present in the pars plicata.
There are few known post-translational modifications of the Na,K-ATPase. The first five amino acids of the N terminus of ␣1 and ␣2 are cleaved during biosynthesis, and the N terminus of rat brain ␣3 appears to be blocked (50). Unconventional glycosylation utilizing N-acetylglucosamine has been reported on the cytoplasmic surface of dog ␣1 (24,39). The location of the cytoplasmic sugar in Na,K-ATPase is not known, but sugar has been demonstrated to occur in the gastric H,K-ATPase ␣ subunit in a 27-kDa soluble tryptic peptide beginning at Asn-369 (NLEAV) and a 15-kDa peptide beginning at Ile-455 (IVIGD) (51). The predicted C terminus would be approximately at amino acid 600 in the H,K-ATPase. This overlaps slightly, if at all, with the deduced epitope location for McBX3 of 590 -669. Identity of the proposed post-translational modification is not known, as there are many possible structures. Ideally, a fragment carrying the epitope should be purified in high enough yield to submit to mass spectroscopy, but this is presently impractical.
Hydroxylamine can catalyze other hydrolysis reactions, including removing the phosphate from aspartyl-phosphate in the Na,K-ATPase active site (52). The active site phosphorylation site at Asp-369 (CSDK) is too labile to be an epitope, however, and is not near the location of McBX3 binding. The Asn-Gly dipeptide bonds themselves are also distant from the region of interest, making these unlikely epitope possibilities. At neutral pH hydroxylamine can also remove fatty acyl groups from cysteine (53), but we observed that hydroxylamine at neutral pH did not remove McBX3 binding (data not shown). ADP-ribosylation of arginine can also be ruled out, since this modification is sensitive to 0.5 M hydroxylamine at pH 7.0 (54). Further information about the nature of the McBX3 epitope is not available, but acylations at any of the nine serine or threonine residues in the region are theoretically possible, including O-linked glycosylation and palmitoylation.
An interesting possibility is that the proposed post-translational modification has an impact on the functional properties of the Na,K-ATPase. The literature on substrate affinities for Na,K-ATPase isoforms is exceptionally controversial (55)(56)(57)(58), and it has been directly demonstrated that ␣1␤1 expressed in different tissues or cells can have different properties (59,60). While these functional differences may arise from differences in lipid composition or in second messenger regulation, the possibility of differences in post-translational modification should be borne in mind. The segment modified is predicted to be in ␤-pleated sheet (61). A modification like fatty acylation might favor close association of the domain with the membrane or with hydrophobic regions of the protein, or might influence ␣-␣ interactions in the membrane.