Covalent Cross-links between the γ Subunit (FXYD2) and α and β Subunits of Na,K-ATPase

This study describes specific intramolecular covalent cross-linking of the γ to α and γ to β subunits of pig kidney Na,K-ATPase and rat γ to α co-expressed in HeLa cells. For this purpose pig γa and γb sequences were determined by cloning and mass spectrometry. Three bifunctional reagents were used: N-hydroxysuccinimidyl-4-azidosalicylic acid (NHS-ASA), disuccinimidyl tartrate (DST), and 1-ethyl-3-[3dimethylaminopropyl]carbodiimide (EDC). NHS-ASA induced α-γ, DST induced α-γ and β-γ, and EDC induced primarily β-γ cross-links. Specific proteolytic and Fe2+-catalyzed cleavages located NHS-ASA- and DST-induced α-γ cross-links on the cytoplasmic surface of the α subunit, downstream of His283 and upstream of Val440. Additional considerations indicated that the DST-induced and NHS-ASA-induced cross-links involve either Lys347 or Lys352 in the S4 stalk segment. Mutational analysis of the rat γ subunit expressed in HeLa cells showed that the DST-induced cross-link involves Lys55 and Lys56 in the cytoplasmic segment. DST and EDC induced two β-γ cross-links, a major one at the extracellular surface within the segment Gly143-Ser302 of the β subunit and another within Ala1-Arg142. Based on the cross-linking and other data on α-γ proximities, we modeled interactions of the transmembrane α-helix and an unstructured cytoplasmic segment SKRLRCGGKKHR of γ with a homology model of the pig α1 subunit. According to the model, the transmembrane segment fits in a groove between M2, M6, and M9, and the cytoplasmic segment interacts with loops L6/7 and L8/9 and stalk S5.

The Na,K-ATPase actively pumps Na ϩ ions out of cells and K ϩ ions into cells and maintains the characteristic transmembrane electrochemical gradients of Na ϩ and K ϩ ions. As could be expected for a protein with such a central physiological role, Na,K-ATPase is closely regulated at several levels.
Recently a unique mode of regulation of the Na,K-ATPase has been described (for reviews, see Refs. [1][2][3]. It involves interactions between the ␣/␤ complex and members of a family of seven short single span transmembrane proteins termed the FXYD proteins (4). Four members of the family, FXYD1 (PLM), 1 FXYD2 (␥), FXYD4 (CHIF), and FXYD7, are now known to interact specifically with the Na,K-ATPase and alter the pump kinetics in characteristic and different ways. The FXYD proteins show a highly tissue-specific expression pattern: ␥ is expressed in kidney, CHIF is expressed in kidney and colon, PLM is expressed in heart and skeletal muscle, and FXYD7 is expressed in brain. In kidney ␥ is expressed as two splice variants, ␥a and ␥b (4,5). Splicing in other FXYD proteins has not been detected at the protein level. The working hypothesis is that FXYD proteins function as tissue-specific modulators of Na,K-ATPase that adjust or fine-tune its kinetic behavior to the specific needs of the given tissue, cell type, or physiological state (1)(2)(3).
Functional interactions between the ␥ subunit and the Na,K-ATPase have now been studied extensively after coexpression with the ␣/␤ subunits in mammalian cells and Xenopus oocytes or by neutralizing interactions with a specific anti-␥ antibody (6 -10). In cultured mammalian cells ␥ raises apparent affinity for ATP by shifting the E 1 -E 2 conformational equilibrium toward E 1 , reduces apparent affinity for cytoplasmic Na ϩ by making cytoplasmic K ϩ a better competitor (8 -10), and slightly reduces extracellular K ϩ affinity (11). Anti-␥C (directed against the sequence KHRQVNEDEL at the C terminus of rat ␥) abrogates the effect of ␥ on the apparent ATP affinity in renal Na,K-ATPase or HeLa cells transfected with ␥ but not that on the K ϩ to Na ϩ antagonism (7,8,10). In oocytes, ␥ reduces affinity for cell Na ϩ and evokes a small increase in the extracellular K ϩ affinity, which varies with voltage (6). In HeLa and HEK293 cells and Xenopus oocytes small or insignificant differences in functional effects are found between ␥b and ␥a (Refs. 10 and 12, but see Ref. 13). In Xenopus oocytes (12) and mammalian cells (14) CHIF raises the apparent affinity for cell Na ϩ by 2-3-fold, the reverse effect to that of ␥. In HeLa cells, CHIF * This work was supported by the Minerva Foundation, the Israel Science Foundation, and the Weizmann Institute renal research fund. 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.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AY941203.
‡ ‡ Incumbent of the William Smithburg chair of biochemistry. To whom correspondence should be addressed. Tel.: 972-8-934-2278; Fax: 972-8-934-4118; E-mail: steven.karlish@weizmann.ac.il. has no effect on the affinity for external K ϩ or ATP (14), whereas in Xenopus oocytes an increased K 0.5 to external K ϩ was observed at high voltages and in the presence of external Na ϩ (12). The opposite functional effects of CHIF and ␥ on the apparent Na ϩ affinity are consistent with their different patterns of expression along the nephron and physiological roles (for reviews, see Refs. [1][2][3]. When expressed in Xenopus oocytes PLM interacts with both the ␣1␤1 and ␣2␤1 isoforms and decreases the internal Na ϩ affinity of the pump by about 2-fold and external K ϩ affinity by a small amount (15), whereas use of an anti-PLM antibody on choroid plexus membranes suggested that PLM might increase Na,K-ATPase activity (16). FXYD7 is the fourth family member whose functional interaction with the Na,K-ATPase has been demonstrated. In Xenopus oocytes FXYD7 decreases the apparent K ϩ affinity of the pump when expressed with ␣1␤1 or ␣2␤1 but not with ␣3␤1 (2).
By comparison with functional studies there is little information on structural interactions of FXYD proteins with ␣/␤ subunits. It is, however, becoming clear that there are multiple sites of interaction, involving both transmembrane segments and the extramembrane domains. The fact that the anti-␥C abrogates the effect of ␥ on the apparent ATP affinity but not that on the K ϩ to Na ϩ antagonism (8, 10) provided an initial indication. In addition, in HeLa cells expression of ␥ with either C-or N-terminal truncated sequences removes the effect on ATP affinity but does not affect the K ϩ to Na ϩ antagonism (17). In a recent systematic study of roles of the different segments, a series of ␥/CHIF chimeric molecules was prepared in which extracellular, transmembrane, and cytoplasmic sequences were interchanged (18). It was found that both the stability of the FXYD-␣/␤ complex in detergent and the effects on the apparent Na ϩ affinity were determined by the origin of the transmembrane segment. Interestingly, however, different residues appear to be involved in the stability and functional effects of the transmembrane segments. The functional role of the transmembrane segments has been confirmed in a study showing that peptides corresponding to the transmembrane segment of ␥ reduce apparent Na ϩ affinity of the ␣/␤ complex in HeLa cell membranes as found previously for full-length transfected ␥ (19). In short, extramembrane segments mediate the effect of ␥ on apparent ATP affinity, whereas transmembrane segments mediate the effect on cation affinities.
Where do FXYD proteins interact with the ␣/␤ complex? On the basis of cryoelectron microscopy of renal Na,K-ATPase electron densities were assigned as transmembrane helices of ␣, ␤, and ␥ subunits. The ␥ subunit helix was proposed to lie in a groove bounded by M2, M6, and M9 of the ␣ subunit (20). A denaturation study suggested that ␥ might interact in the M8 -M10 region (21). Recently a role for M9 of the ␣ subunit has been inferred from effects of mutants in M9 on stability of ␣/␤-␥, ␣/␤-CHIF, or ␣/␤-FXYD7 complexes and their functional consequences studied in Xenopus oocytes (22). Leu 964 and Phe 967 were important for stability of the complexes, whereas Phe 956 and Glu 960 were required for mediation of effects of the FXYD protein on K ϩ affinity. Thus, stabilizing and functional interactions were separable as found also in mutational studies in CHIF/␥ chimera (23). Interestingly the Phe 956 and Glu 960 mutations did not alter effects of ␥ and CHIF on Na ϩ affinity, implying that still other interactions in the transmembrane segment mediate these effects. Modeling of the FXYD helix was consistent with docking in the groove between M2, M6, and M9.
The present work has utilized a different approach, namely covalent cross-linking, to obtain direct evidence for proximities between the ␥ subunit and ␣ and ␤ subunits of pig kidney Na,K-ATPase and in HeLa cells expressing rat ␣ and ␥ subunits. We have used a variety of bifunctional reagents with different chemical specificities and arm lengths, optimized cross-linking efficiency, and determined the approximate positions of observed ␣-␥ and ␥-␤ cross-links. Based on the inferred position of ␣-␥ cross-links in the cytoplasmic domains of the ␣ and ␥ subunits, we have modeled the ␣-␥ interactions.

Pig Kidney Na,K-ATPase: Preparations and Detergent Solubilization
Partially purified Na,K-ATPase from pig kidney outer medulla was prepared as described previously (24). Extensively trypsinized 19-kDa membranes were prepared as described previously (25). Membranes were solubilized by the non-ionic detergent C 12 E 10 (polyoxyethylene 10-lauryl ether) in the presence of Rb ϩ plus ouabain or Na ϩ plus oligomycin as described previously (14).

HeLa Cell Expression of ␥a or ␥b
HeLa cells overexpressing the rat ␣1 subunit of Na,K-ATPase were kindly provided by Dr. J. B Lingrel, University of Cincinnati College of Medicine, Cincinnati, OH. Cells were transfected with wild type and mutated rat ␥ constructs subcloned into pIREShyg. Transfection was done using Polyfect (Qiagen) according to the manufacturer's instructions. Colonies stably expressing ␥ proteins were selected in 400 g/ml hygromycin B and tested for maximal expression of ␥ by Western blotting.
Covalent Cross-linking NHS-ASA-Purified pig kidney Na,K-ATPase or 19-kDa membranes were suspended in 10 mM sodium borate, pH 9.5, 130 mM NaCl or 10 mM HEPES, pH 8, 130 mM NaCl to 0.5 mg/ml protein concentration. NHS-ASA (26) dissolved in Me 2 SO was added to 0.25-1 mM in three aliquots and incubated at room temperature in the dark for 30 min. The reaction was quenched by 50 mM unbuffered Tris. The pH of the suspension was restored to near neutral with 100 mM HEPES, pH 7.4, and the suspension was illuminated for 2 min with a xenon lamp (150 watts, fitted with a filter cutting off light below 300 nm). Where indicated, the membranes were pelleted and resuspended in the detergent solubilization buffer: 25 mM imidazole, pH 7.5, 1 mM EDTA and either 10 mM RbCl plus 5 mM ouabain (pig kidney enzyme) or 20 mM NaCl plus oligomycin 0.1 mg/ml (pig kidney enzyme, HeLa cells). C 12 E 10 was added at 1 mg/ml. The soluble fraction was separated by centrifugation at 100,000 ϫ g and was illuminated with the UV lamp for 2 min. The sample was denatured with 2% SDS, and protein was precipitated by 4 volumes of methanol:diethyl ether (2:1, v/v). Pellets were dried in N 2 and suspended in gel sample buffer.
DST-Purified kidney enzyme was suspended in 10 mM HEPES, pH 8, 250 mM NaCl or 30 mM Rb ϩ plus 220 mM choline at 0.5-1 mg/ml protein, or the Na,K-ATPase was solubilized with C 12 E 10 in 10 mM HEPES, pH 8, in the presence of 10 mM RbCl plus 5 mM ouabain or 20 mM NaCl plus oligomycin. HeLa cell membranes were solubilized with C 12 E 10 in the presence of Na ϩ /oligomycin. 19-kDa membranes were suspended in 10 mM RbCl, 10 mM Na-HEPES, pH 7.4. DST (27) in Me 2 SO was added to a final concentration of 2 mM followed by incubation for 30 min at room temperature. The reaction was quenched with 50 mM unbuffered Tris.
EDC-Purified kidney enzyme was suspended in 100 mM MES, pH 6, 250 mM NaCl or 30 mM Rb ϩ plus 220 mM choline to 0.5-1 mg/ml protein. Alternatively the Na,K-ATPase was solubilized with 1 mg/ml C 12 E 10 in media containing 100 mM MES, pH 6, 10 mM RbCl plus 5 mM ouabain or 20 mM NaCl plus oligomycin. 19-kDa membranes were suspended in 100 mM MES, pH 6, 10 mM RbCl, and 140 mM choline chloride. 1 mM EDC (28) and 5 mM N-hydroxysuccinimide (NHS) dissolved in water were added, and the mixture was incubated for 2 h at room temperature. The cross-linking was stopped by addition of 10 mM hydroxylamine.

Deglycosylation
Cross-linked native enzyme or 19-kDa membranes (30 -40 g) were treated with PNGase F (125 units, New England Biolabs) for 24 h at 37°C in a medium containing 20 mM Tris⅐HCl, pH 8.0, 10 mM RbCl after denaturation in the buffers supplied with the PNGase. The digestion was arrested by addition of concentrated gel sample buffer.

Chymotryptic Cleavages
Purified kidney enzyme or NHS-ASA-cross-linked membranes were suspended in 10 mM HEPES, pH 7.4, containing either 20 mM RbCl or 20 mM NaCl (0.1 mg/ml protein) and incubated with 1:20 chymotrypsin (mg/mg of protein) for 15 min (see Refs. 29 and 30). The enzymatic digestion was stopped by 10 volumes of ice-cold HEPES buffer containing 1 mM phenylmethylsulfonyl fluoride and 150 mM RbCl.

Fe 2ϩ -catalyzed Oxidative Cleavage
Purified kidney enzyme or NHS-ASA-cross-linked membranes were suspended in 10 mM HEPES, pH 7.4, in the presence of either 130 mM Na ϩ E 1 Na or 130 mM NaCl, and 40 M AMPPNP. The enzyme in E 1 Na conformation was incubated with 5 mM ascorbic acid, 5 mM H 2 O 2 , 5 M Fe 2ϩ for 15-20 min, and the cleavage was stopped by addition of sample buffer containing 5 mM Desferal (see Refs. [31][32][33].

Mass Spectrometry
Pig ␥a and ␥b were extracted from gels with organic solvents, or gel pieces were subjected to tryptic digestion as described previously (5). The sequences were determined by a combination of mass measurements of intact subunits and tryptic peptides and direct sequencing by tandem mass spectrometry as described previously (5). Mass spectra were acquired on an Q-STAR Pulsar i electrospray-quadrupole time-offlight tandem mass spectrometer containing a quadrupole collision cell (MDS-Sciex) and equipped with a nanoelectrospray source (Proxeon Biosystems).

Peptide Fingerprinting
Peptide fingerprinting was performed on a Bruker Reflex III TM matrix-assisted laser desorption ionization time-of-flight mass spectrometer (Bruker, Bremen, Germany) equipped with a delayed extraction ion source, a reflector, and a 337 nm nitrogen laser.

Cloning of Pig ␥a
The EST data base contains only one expressed sequence tag entry corresponding to pig ␥a (BX674298). To verify the amino acid sequence of this protein, pig ␥a was cloned by reverse transcription-PCR. Pig kidney RNA was reverse transcribed from the poly(A) tail and then amplified using the primers GCAGGAAGAGGGCAGTGG (5Ј) and TGCGATGGGGGCACAGCCGA (3Ј). The 270-bp product was sequenced from both ends and found to correspond to ␥a. To exclude PCR errors, sequencing was repeated for two independent reactions of reverse transcription and amplification. The nucleotide sequence of pig ␥a has been deposited in GenBank TM under accession number AY941203.

Modeling the ␣-␥ Interaction
Homology Modeling of Pig ␣1 Subunit-The alignment used by Li and co-workers (22) for Ca-ATPase and Bufo Na,K ␣ subunits was modified to the appropriate Ca-ATPase and pig Na,K ␣1 subunit alignments. The sequence alignment was converted to a structure using the homology-modeling program package Modeler6 v. 2 with default settings (35,36) and the crystal structure of Ca-ATPase in the ATP-bound form (37,38) as template. 50 structures were generated using different initial random velocities for all atoms. The best model was chosen based on the in-built objective function, which uses stereochemical parameters as well as similarity to the template to rank the models.
Positioning the Transmembrane Helix of the ␥ Subunit-Following the approach by Li and co-workers (22), the TM helix of ␥ was designated to include residues Arg 26 -Ser 46 and was manually positioned between helices M2, M6, and M9. Starting from this initial conformation, 50,000 different random conformations were generated by ran-domly rotating the helix 0 -360°around the long axis; translating it Ϫ3 to ϩ5 Å toward or away from the center of the bundle, respectively; translating the helix Ϯ10 Å along its long axis; and tilting the helix relative to the fixed angles of the TM helices of the ␣ subunit by Ϯ20°. The coordinate transformations were done using the program package CNS (39) with the OPLS (40) parameter set. Each such generated conformation was subjected to 50 steps of Powell minimization, and the Van der Waals interaction energy of the ␥-helix with the ␣ subunit was calculated. All degrees of freedom were sampled in 20 bins per parameter. The average energy of each bin was evaluated independently by using Boltzmann averaging of all conformations within this bin. After Boltzmann averaging the minimal average energy of the parameter in question was taken as that at the center of the bin of lowest energy. A related procedure has been shown to successfully reproduce structures of four-helix bundles (41,42).
Interactive Modeling of the Cytoplasmic Segment of the ␥ Subunit-A secondary structure prediction by Jpred2 (43) revealed that the region starting from residue Ser 47 has no preference for any secondary structure. Therefore the sequence SKRLRCGGKKHR (pig ␥a sequence) was modeled interactively as a random coil using the DeepView modeling package (44) with the following constraints. 1) The Ramachandran plot should be fulfilled. 2) Neither backbone nor side chains of ␥ should clash with the ␣ subunit.
3) The positive charges of Lys 54 and Lys 55 and of His 56 and Arg 57 should interact with negatively charged residues of the ␣ subunit.
In a first step the / angles of the backbone were adjusted so that the segment came into contact with the ␣ subunit without backbone clashes and so that side chains of acidic residues of the ␣ subunit came into proximity with the side chains of Lys 54 , Lys 55 , His 56 , and Arg 57 . In a second step, a rotameric search of each individual contact side chain for both ␣ and ␥ subunits was performed to optimize the side-chain packing. The two steps were iterated until no clashes occurred.
Refinement-The initial model was refined with a short molecular dynamics protocol in vacuo using the molecular dynamics package Gromacs (45,46) and the Gromos96 vacuum force field (47). To relieve steric strain, 5000 steps of steepest descent minimization were performed followed by 100 ps of molecular dynamics at 300 K with a restrained backbone followed by another 100 ps of free molecular dynamics without any restraints. In a final step, 5000 steps of conjugate gradient minimization were performed. Although the complex membrane-water interface is not well reproduced by a simulation in vacuo, the emphasis of our calculation was less on observing the dynamics of the system and more on relieving local stress induced by the interactive modeling. After the short run, the root mean square deviation between the initial and final ␥ conformation was less than 2 Å. Fig. 1 shows structures of the three cross-linkers used in this study. NHS-ASA is a heterobifunctional, 8-Å spacer, photoactivated reagent. The NHS ester reacts with lysine residues, whereas the arylnitrene can react with double bonds, insert into C-H and N-H bonds, or undergo subsequent ring expansion to react with a nucleophile (e.g. primary amines and Cross-linking ␥ to ␣ and ␤ Subunits of Na,K-ATPase thiols). DST is a homobifunctional, lysine-specific reagent with a spacer arm of 6.4 Å. EDC activates carboxyls that couple with primary amines to generate peptide bonds, producing a zero length cross-link. The EDC cross-linking is done in the presence of NHS, which increases the yield (28).

RESULTS
Determination of the Pig ␥ Sequence by Cloning and Mass Spectrometry-Knowledge of the sequence of the pig ␥a and ␥b subunits is required for localization of the cross-links and modeling. Sequences were inferred by cloning of ␥a and mass spectrometry of tryptic fragments of ␥a and ␥b, utilizing methods described previously for analysis of rat ␥a and ␥b (5). Fig. 2 shows the inferred sequences of pig ␥a and ␥b aligned with known human and rat ␥a sequences. The underlined segments, corresponding to about 50% coverage, were determined by direct sequencing of the fragments by electrospray ionizationtandem mass spectrometry and are identical to the sequence inferred by cloning. The mass spectrometry showed that the initiator methionine is absent from the N terminus of the ␥a protein, whereas the N-terminal sequence of ␥b begins with Ac-MDRWYL or MDRWYL as found previously for rat ␥b. cDNA sequencing suggested a single nucleotide polymorphism resulting in the eighth amino acid of ␥a being either Gly or Asp. However, no such polymorphism was detected by mass spectrometry at the protein level. The predicted masses of ␥a (without initiator methionine) and ␥b (with acetylated initiator methionine) are 6954 and 7200 Da, respectively. Although the samples were more heterogeneous than observed for rat ␥a and ␥b, the measured masses of the most abundant species of the intact reduced and carbamidomethylated ␥a and ␥b, 7027.9 and 7288.8 Da, respectively, correspond to the predicted masses of singly oxidized ␥a and doubly oxidized ␥b. 2 NHS-ASA-induced ␣-␥ Cross-link on Pig Kidney Na, K-ATPase- Fig. 3 shows representative experiments demonstrating a specific intramolecular cross-link between ␣ and ␥ subunits and controls to exclude possible artifacts. The pig kidney Na,K-ATPase was first incubated with NHS-ASA in the dark at alkaline pH, and then the pH was restored to 7.4 prior to illumination with UV light. Optimization experiments showed the most efficient cross-linking with 1 mM NHS-ASA, pH 9.5, in the dark reaction and an illumination time of 2 min. In the lane marked NHS-ASA in Fig. 3A, the ␣-␥ cross-link was visualized in an immunoblot using an anti-␥C antibody. In the lane marked C, the enzyme was first dissolved in SDS prior to the reaction with NHS-ASA. The lack of cross-linking in this condition indicates that a native structure of the protein is required and excludes the possibility of unselective ␣-␥ cross-linking induced by the NHS-ASA during solubilization with SDS prior to application to gels. Another possible artifact, particularly for membrane proteins at a high concentration such as the Na,K-ATPase, is that molecules that have reacted with the bifunctional reagent can collide randomly with other molecules in the membrane and produce intermolecular cross-links. The experiment in Fig. 3B excluded this possibility. The renal enzyme was first incubated with NHS-ASA, the pH was restored to 7.4, and the membranes were then dissolved in a non-ionic detergent, C 12 E 10 , in the dark prior to illumination of the C 12 E 10 -soluble protein. Solubilization was done in the presence either of Rb ϩ plus ouabain or Na ϩ plus oligomycin, conditions known to preserve Rb ϩ or Na ϩ occlusion, respectively, and a native structure of the protein (14). When dissolved in the detergent, the protein concentration was diluted by at least 4 orders of magnitude by comparison with that in the membrane-bound state, thus greatly reducing the chance of random collision. The observation of the cross-link even after solubilization in C 12 E 10 provides a strong indication for a specific intramolecular cross-link. Cross-linking was more efficient in the presence of Na ϩ /oligomycin compared with Rb ϩ /ouabain, suggesting a possible conformation dependence. The latter possibility was tested more systematically in Fig. 3C, which looked at the effect of the presence of either Rb ϩ or Na ϩ ions in either the dark or light stages, and as seen, Na ϩ ions in only the dark stage amplified the crosslinking efficiency. In these optimal conditions, which were used for all subsequent cross-linking experiments, up to 50% of the ␥ subunit could be cross-linked. Fig. 3D shows the use of 19-kDa membranes to assign the sidedness of the cross-link. 19-kDa membranes are produced by extensive tryptic digestion of renal Na,K-ATPase. The cytoplasmic domains of the ␣ subunit are removed, leaving membranebound fragments (25,48). The ␤ subunit is partially cleaved to 16-kDa and 50-kDa fragments, and ␥a and ␥b subunits are intact (49) (see the schematic model in Fig. 5A). Fig. 3D shows that treatment of 19-kDa membranes with NHS-ASA produced no cross-linked bands of the ␥ subunits. Thus in native Na,K-ATPase the cross-link must be located on a cytoplasmic residue of the ␣ subunit, which has been removed in 19-kDa membranes.
A similar conclusion on sidedness comes from an observation that the anti-␥C antibody blocked cross-linking with NHS-ASA (Fig. 4). The ␣-␥ cross-link was largely suppressed, and a faster running band recognizing the anti-␥C appeared. This new immunoreactive band was shown to be the heavy chain of the antibody itself because it was adsorbed to protein A beads after denaturation with SDS. The small amount of the remaining ␣-␥ cross-linked protein was not adsorbed and remained in the supernatant. The results of this experiment suggest strongly that the cross-link lies near the epitope of the ␥ subunit on the cytoplasmic surface.
Figs. 6 and 7 show how selective chymotryptic and Fe 2ϩcatalyzed oxidative cleavages of the control and cross-linked enzyme were used to locate the position of the NHS-ASA crosslink to a limited segment of the ␣ subunit (see the cleavage sites in Fig. 5B). The strategy was to demonstrate which known fragments of the ␣ subunit do or do not also recognize the anti-␥C antibody. 3 As seen in Fig. 6A, in an E 2 Rb conformation, 2 The error of the mass measurement of the intact protein is about 0.05-0.1% (or 3-7 Da). 3 Potentially a caveat to this approach is that unrelated intramolecular cross-links in the ␣ subunit could affect the cleavage pattern. In practice, however, the observation of similar cleavage fragments in the cross-linked and uncross-linked protein shows that this is not a serious problem.
FIG. 2. Alignment of pig ␥a and ␥b sequences with human and rat ␥a. The pig ␥a amino acid sequence was deduced by sequencing a reverse transcription-PCR product. Underlined segments were determined by full sequencing of tryptic peptides using mass spectrometry (see Ref. 5). Mass spectrometry also provided the unique ␥b N-terminal sequence.
Cross-linking ␥ to ␣ and ␤ Subunits of Na,K-ATPase two well characterized chymotryptic fragments, Val 440 -Tyr 1016 and Ala 589 -Try 1016 , were detected with anti-KETYY, which recognizes the C terminus of the ␣ subunit. Neither of these fragments was recognized by the anti-␥C. The fragments that were recognized by the anti-␥C are identifiable as (a) the complementary fragment Gly 1 -Val 440 and (b) probably a secondary cleavage fragment of the complement to Gly 1 -Ala 589 (Fig. 6,  asterisk). The experiment shows that the cross-link must be located upstream of Val 440 . In an E 1 Na conformation a major fragment Ile 263 -Tyr 1016 was detected by anti-KETYY (Fig. 6B). Closer inspection showed that it consists of two fragments, which correspond to the chymotryptic fragments of the uncross-linked and cross-linked ␣ subunit, respectively. The upper of the two fragments was also recognized by anti-␥C. Thus the experiment shows that the cross-link lies within this fragment and is located downstream of Ile 263 . The Fe 2ϩ cleavage experiments of Fig. 7, A and B, confirmed and extended these conclusions. In an E 1 Na conformation, Fe 2ϩ /ascorbate/H 2 O 2 treatment produced two fragments of the ␣ subunit recognized by anti-KETYY, Glu 80 -Tyr 1016 and His 283 -Tyr 1016 (32,50). Cleavage of the cross-linked product produced the same fragments, including a broader band, presumably a doublet corresponding to the His 283 -Tyr 1016 fragment of uncross-linked and cross-linked ␣ subunit, respectively. The upper His 283 -Tyr 1016 fragment was also recognized by anti-␥C. The complementary fragment Gly 1 -His 283 was recognized by antibody 6H, which has an epitope near the N terminus of the ␣ subunit, but was not recognized by anti-␥C. The experiment indicates the presence of the cross-link in the His 283 -Tyr 1016 fragment and lo- The renal Na,K-ATPase was treated with NHS-ASA at pH 9.5 in the dark, the pH was restored to 7.4, the protein was dissolved in C 12 E 10 , anti-␥C was added at 1:10 dilution where indicated, and then the suspension was illuminated. After illumination an aliquot was treated with 1% SDS to disrupt non-covalent complexes; then 10 volumes of 50 mM Tris⅐HCl, pH 7.4, 5 mM EDTA, 150 mM NaCl, and 1% Triton X-100 (TENT solution) was added followed by 30 l of protein A beads; and the slurry was incubated overnight at 4°C. The supernatant was removed by centrifugation, and protein was precipitated with methanol:ether. Protein bound to the beads was released by gel sample buffer. C, control; Ab, antibody; Sup, supernatant.
cates it downstream of His 283 . Fig. 7B shows results of incubation with AMPPNP-Fe/ascorbate/H 2 O 2 , which produced two well characterized fragments recognized by anti-KETYY, Val 440 -Tyr 1016 and Val 712 -Tyr 1016 (33,50), in both the control and cross-linked preparations. Neither of these fragments was recognized by anti-␥C. A fragment that was not recognized by anti-␥C, but was recognized by the anti-N-terminal antibody, was identified as Gly 1 -Val 712 . This experiment confirms that the cross-link lies upstream of Val 440 . In summary, the inferences from Figs. 6 and 7 are that the cross-link is located upstream of His 283 and downstream of Val 440 on the cytoplasmic side. We tried but were not able to locate the cross-link more exactly by direct methods such as mass spectrometry of the digested cross-linked polypeptide. However, as argued in the "Discussion," there is good reason to think the cross-link is located in the cytoplasmic stalk S4, emerging from transmembrane segment M4.
DST-induced ␣-␥ and ␤-␥ Cross-links on Pig Kidney Na, K-ATPase-The immunoblot, probed with anti-␥C, in Fig. 8 shows that DST induced a prominent ␣-␥ cross-link and a less prominent ␤-␥ cross-link. The figure also presents the same controls for specificity as discussed in relation to NHS-ASA. The bands were identified as ␣-␥ and ␤-␥ cross-links by their mobility on the gels just above uncross-linked ␣ or ␤ subunits and in the case of the ␤-␥ cross-link by the change of mobility after deglycosylation (seen in Figs. 9 and 10). The ␣-␥ crosslink was amplified in the Na ϩ -containing compared with the Rb ϩ -containing medium as found also for NHS-ASA. The ␤-␥ cross-link was usually unaffected by the ionic composition of the medium (although it appears to be amplified in Fig. 8,  lane 4). The ␣-␥ cross-link was more efficient at pH 9 compared with pH 8 (Fig. 8, right-hand lanes) consistent with reactivity of DST with lysine residues. The anti-␥C did not block the DST-induced ␣-␥ cross-link unlike the result with NHS-ASA, suggesting that the sites of NHS-ASA-and DSTinduced cross-linking are not identical (see the "Discussion"). Fig. 9 presents chymotryptic and Fe 2ϩ cleavage data, similar to that described for NHS-ASA. The chymotryptic cleavage on the cross-linked protein in the E 2 Rb conformation produced the two bands Val 440 -Tyr 1016 and Ala 589 -Tyr 1016 , but neither band was recognized by anti-␥C. However, the band marked with the asterisk that ran similarly to that produced in the digest of the NHS-ASA cross-linked protein FIG. 6. Specific chymotryptic cleavage of the NHS-ASA-induced ␣-␥ cross-link. Membrane-bound Na,K-ATPase was treated with NHS-ASA as described under "Materials and Methods." Aliquots of the untreated (C) and NHS-ASA-treated enzyme were digested with ␣-chymotrypsin (CHY) in Rb ϩ -containing media (A) or Na ϩ -containing media (B). Samples were applied to gels, and immunoblots were developed either with anti-KETYY or anti-␥C. *, unknown chymotryptic fragment.

FIG. 7. Specific Fe 2؉ -catalyzed oxidative cleavage of the NHS-ASA-induced ␣-␥ cross-link.
Membrane-bound Na,K-ATPase was treated with NHS-ASA as described under "Materials and Methods." Aliquots of the untreated (C) and NHS-ASA-treated enzyme were cleaved with Fe 2ϩ /ascorbate/H 2 O 2 in Na ϩ -containing media (E 1 Na conformation) (A) or Na ϩ -and AMPPNP-containing media (E 1 AMPPNP-Fe conformation) (B) as described under "Materials and Methods." Samples were applied to gels, and immunoblots were developed either with anti-KETYY, anti-N terminus (6H), or anti-␥C.
FIG. 8. DST-induced ␣-␥ and ␤-␥ cross-links. Membrane-bound Na,K-ATPase was treated with 2 mM DST in Na ϩ -containing medium at pH 8 or 9 (right) or in a Rb ϩ -or Na ϩ -containing medium at pH 9 (left), or the protein was first solubilized with C 12 E 10 in the presence of Rb ϩ /ouabain (ou.) or Na ϩ /oligomycin (oli.) (middle) as described under "Materials and Methods." Controls (C) refer to enzyme treated with 2% SDS and then with DST. was recognized by anti-␥C. The chymotryptic cleavage in the E 1 Na conformation produced the expected bands Ile 263 -Tyr 1016 , and this band was recognized by the anti-␥C. Similarly Fe 2ϩ -catalyzed cleavage in the E 1 Na conformation produced the expected band His 283 -Tyr 1016 , and again this band was recognized by the anti-␥C. In short these cleavage experiments indicate that the DST-induced cross-link is located in the same region as the NHS-ASA cross-link, downstream of His 283 and upstream of Val 440 . Fig. 10 presents an experiment on DST cross-linking of 19-kDa membranes, utilizing anti-␥C and anti-␤ 16-kDa and anti-␤ 50-kDa antibodies. This shows clearly that all the bands recognized by anti-␥C were also recognized by anti-␤ antibodies. Thus, only ␤-␥ cross-links and no ␣-␥ cross-links were observed in 19-kDa membranes, as found also with NHS-ASA. The major cross-links were associated with the intact ␤ or the 50-kDa glycosylated fragment, and there were two minor bands, which represent cross-links with the 16-kDa fragment of ␤. Accordingly the mobility of the upper two bands increased upon deglycosylation with PNGase. The anti-␤ 16-kDa antibody also recognized an additional band, which was crosslinked with the 19-kDa fragment of ␣, as detected with anti-KETYY (not shown) and is irrelevant to the current study. The experiment shows (a) that the ␣-␥ cross-link on native enzyme is located on the cytoplasmic side on a lysine residue that is absent from 19-kDa membranes and (b) the major ␤-␥ crosslink is located to the 50-kDa extracellular domain of the ␤ subunit, and there are minor cross-links to the 16-kDa fragment.
␣-␥ Cross-links in HeLa Cells: Locating a DST-induced Cross-link in the ␥ Subunit by Mutational Analysis-As described previously, the rat ␥ subunit has been expressed together with the rat ␣ subunit (and endogenous human ␤ subunit) in HeLa cells and used extensively for analysis of the functional effects of ␥ (8,10,17,23,51). Since DST is lysine-lysine specific, one or more lysine residues in the cytoplasmic segment of ␥ must be involved in the cross-link, and the possible candidates are Lys 46 , Lys 55 , and Lys 56 (rat numbering, see Fig. 2). The data in Fig. 11 show the effects of the single or combined Lys to Arg, Lys to Ala, or Lys to His mutations in ␥b or ␥a on DST-or NHS-ASA-induced crosslinks. DST or NHS-ASA cross-linking was carried out on the HeLa cell membranes solubilized in C 12 E 10 in the medium containing Na ϩ /oligomycin as described under "Materials and Methods." Fig. 11A shows that DST induced a specific ␣-␥b cross-link (compare lanes 1 and 2). The cross-link was not affected by the K46R mutant but was largely blocked in the triple mutation K46R/K55A/K56A. By contrast the NHS-ASA cross-link was not blocked (Fig. 11B), indicating that NHS-ASA does not cross-link via a lysine residue on the ␥ subunit. The latter also serves as a control showing also that the ␣-␤-␥ complex is intact in the detergent-solubilized membranes containing the triple mutant. Fig. 11C presents another experiment showing that the DST cross-link was also largely blocked in the double mutant K55A/K56A. Finally the single mutants K55H and K56H and double mutant K55H/ K56H of ␥a were expressed in the HeLa cells (Fig. 11D). In this case the DST-induced cross-link was partially reduced in each of the individual mutants and largely blocked in the double mutant. Thus, the conclusion from Fig. 11 is that the DST induces the ␣-␥ cross-link about equally on Lys 55 FIG. 9. Specific chymotryptic and Fe 2؉ -catalyzed cleavage of DST-induced ␣-␥ cross-link. Membrane-bound Na,K-ATPase was treated with DST as described under "Materials and Methods." Aliquots of the untreated (C) and DST-treated enzyme were digested with ␣-chymotrypsin (Chy) in Rb ϩ -containing or Na ϩ -containing media (A) or cleaved with Fe 2ϩ /ascorbate/H 2 O 2 in a Na ϩ -containing medium (B). A sample of the enzyme digested with chymotrypsin was also treated with PNGase. Samples were applied to gels, and immunoblots were developed either with anti-KETYY or anti-␥C. Degly., deglycosylated. *, unknown chymotryptic fragment.

FIG. 10. DST-induced ␤-␥ cross-links in 19-kDa membranes.
19-kDa membranes suspended in a Rb ϩ -containing medium at pH 7.4 were treated with 2 mM DST for 30 min at room temperature. Controls (C) refer to 19-kDa membranes pretreated with 2% SDS and then with DST. The blots were probed with anti-␥C, anti-␤50, or anti-␤16 antibodies. An aliquot was also treated overnight with PNGase prior to application to the gel. Degly., deglycosylated. and Lys 56 of rat ␥ and a proximal lysine residue in the ␣ subunit.
EDC-induced ␤-␥ Cross-link on Pig Kidney Na,K-ATPase-Finally Fig. 12A shows that EDC induced primarily ␤-␥ crosslinks in either membrane-bound or detergent-soluble pig kidney Na,K-ATPase with only minor ␣-␥ cross-linking and little or no difference in Rb ϩ -containing or Na ϩ -containing media. The experiment with 19-kDa membranes in Fig. 12B shows essentially the same features as found for DST-induced crosslinking, namely all bands recognized by the anti-␥C were also recognized by anti-␤ antibodies. The EDC induced cross-links to both the 50-kDa fragment and 16-kDa fragments, confirming the existence of two separate ␤-␥ cross-linking positions. DISCUSSION The first point to make is that NHS-ASA and DST crosslink the same regions of the ␥ and ␣ subunits at the cytoplasmic side, whereas DST and EDC cross-link the same regions of ␥ and ␤ subunits at the extracellular side. The fact that bifunctional reagents with different chemical specificities cross-link the ␥ subunit with ␣ or ␤ subunits in both pig and rat and the same regions of cross-linking are involved in the different ␣-␥ and ␤-␥ cross-links provides a strong indication that they are specific intramolecular cross-links and that they occur in regions of subunit interactions. Although NHS-ASA is an efficient cross-linker and was very useful for establishing criteria for specific cross-links and the initial analysis of the site of the cross-link, a detailed analysis was complicated by the lack of knowledge of which chain, ␣ or ␥, contains the reacted lysine. Use of DST, which creates specific lysine-lysine cross-links, avoids this ambiguity and eventually provided more detailed information.
Locating the ␣-␥ Cross-link-The mutation work with HeLa cells showed that the DST-induced cross-link involved Lys 55 and Lys 56 in the cytoplasmic segment. Conversely the analysis showed that NHS-ASA cross-link does not involve lysines on the ␥ subunit, and thus the lysine modified in the dark reaction must be located on the ␣ subunit. Because both DST and NHS-ASA are N-hydroxysuccinimide esters, the same lysine on the ␣ subunit could be involved in the DSTinduced cross-link and the dark reaction of NHS-ASA. We have not attempted to identify the residue of ␥ involved in the NHS-ASA light reaction by mutational analysis because the nitrene is not specific and could react with more than one proximal residue.
Both DST-and NHS-ASA-induced ␣-␥ cross-links were amplified in Na ϩ -containing media, which stabilizes the E 1 Na conformation. Presumably the ␣-␥ interaction is responsible for the functional effect of ␥ on the apparent ATP affinity for Na,K-ATPase, which has been shown to be an indirect result of stabilization of the E 1 conformation (10). Amplification of the cross-link in E 1 is consistent with this interpretation for if ␥ interacts with ␣ to stabilize E 1 , the principle of linked equilibria requires that stabilization of E 1 by a different ligand (Na ϩ ions) should strengthen the ␣-␥ interaction. The functional effect of ␥ on the apparent ATP affinity, like the NHS-ASAmediated cross-link, was abrogated by the anti-␥C. Inhibition of the NHS-ASA-induced cross-link by the anti-␥C is suggestive of a position nearer the C terminus than Lys 54 or Lys 55 within the epitope His 56 -Lys 64 .
The combination of extensive and controlled cleavages of the DST-induced and NHS-ASA-induced cross-linked products showed that in both cases the cross-link on the ␣ subunit lies at the cytoplasmic side, downstream of His 283 and upstream of Val 440 . In reality, however, there are more stringent constraints than revealed directly by the experiments. First, His 283 is itself located at the cytoplasmic entrance of M3 (see Fig. 5), and because neither NHS-ASA-nor DSTmediated cross-links can be in M3, the extracellular L3-4, or M4, one can infer that they are, in fact, located after the cytoplasmic exit of M4 and before Val 440 . Second, because the DST-mediated cross-link on the ␥ subunit is located on Lys 54 or Lys 55 , 10 or 11 residues from the cytoplasmic exit of the transmembrane segment (L4 -5) (pig numbering, see Fig. 2), the cross-linked residues in the ␣ subunit are unlikely to be much further away from the cytoplasmic exit of M4, i.e. they are likely to be located within the S4 cytoplasmic stalk segment (Thr 338 -Glu 358 ). Third, there are only three candidate lysines within 15 residues of the exit of M4, Lys 342 , Lys 347 , and Lys 352 , and the only other lysines upstream of Val 440 , Lys 370 , Lys 406 , and Lys 438 , are 33, 69, and 101 residues distant from the end of M4, respectively. Fourth, of the three most likely candidates, Lys 342 , Lys 347 , and Lys 352 , Lys 342 can be excluded. This arises from the fact that Lys 342 is the first tryptic cleavage site after M4. The C-terminal residue of the M3-M4 fragment of 19-kDa membranes is not exactly known, although Arg 346 or Arg 343 are likely possibilities; but in any case, the M3-M4 fragment must include Lys 342 . Since no DST-or NHS-ASA-induced ␣-␥ cross-link was detected in 19-kDa membranes, Lys 342 cannot be involved in the crosslink of the native enzyme. Thus, we are left with the conclusion that the most likely lysine residue for DST-mediated cross-linking on the ␣ subunit is Lys 347 or Lys 352 . Similarly the lysine residue modified by NHS-ASA cross-link is predicted to be Lys 347 or Lys 352 .
Modeling ␣-␥ Interactions-We attempted to model the interaction of the ␥ subunit and a homology model of pig ␣1 subunit using information on ␣-␥ proximities and interactions. Fig. 13A (ribbons) and Fig. 13B (surface) shows the proposed general disposition of the ␥ and ␣ subunits. Fig. 14 shows details of proposed proximities and interactions of transmembrane (A) and cytoplasmic segment (B), respectively. An important consideration is that secondary structure predictions (Jrpred, protein predict, nnpredict, and psipred at the Swiss-Prot web site at www.expasy.org/tools/) FIG. 12. EDC-induced ␤-␥ cross-links in intact renal Na,K-ATPase and 19-kDa membranes. A, membrane-bound or C 12 E 10solubilized Na,K-ATPase was treated with 1 mM EDC plus 5 mM NHS at pH 6 for 2 h at room temperature (see "Materials and Methods"). Controls (C) refer to enzyme pretreated with 2% SDS and then with EDC. The immunoblot was developed with anti-␥C. B, 19-kDa membranes suspended in a Rb ϩ -containing medium at pH 6 were treated with 1 mM EDC plus 5 mM NHS for 2 h at room temperature. Controls (C) refer to enzyme pretreated with 2% SDS and then with EDC. Aliquots were treated overnight with PNGase prior to application to the gel. The blots were probed with anti-␥C, anti-␤50, or anti-␤16 antibodies. ou., ouabain; oli., oligomycin. all indicate that, whereas the transmembrane segment of the ␥ subunit is ␣-helix, the cytoplasmic sequence is unstructured or random coil. The inference from the current work that either Lys 54 or Lys 55 (pig numbering) of ␥ is cross-linked at 6 -8 Å distance from Lys 347 or Lys 352 in S4 of the ␣ subunit strongly supports a location of the transmembrane of ␥ in the groove between M2, M9, and M6, overlooked by M4, as proposed previously (20,22) and also seen in Fig. 13A. Therefore the first step was to optimize docking of the transmembrane segment. The next step was the manual interactive modeling of the unstructured tail SKRLRCGGKKHR. In addition to the cross-linking requirements, the modeling utilized two further constraints, namely that 1) the KKHR residues are known to be necessary for the ␣-␥ interaction (12) and 2) the ␥ subunit in 19-kDa membranes is intact (49) indicating that the KKHR residues are well protected from tryptic digestion, presumably by interacting with the ␣ subunit. The final model, encompassing both the TM segment as well as the tail (Figs. 13 (overview) and 14 (detailed)), was obtained after a short equilibration molecular dynamics calculation in vacuo as described under "Materials and Methods." In the process of docking of the transmembrane segment the tilt and translations were well defined with a single energy minimum. However, two nearly indistinguishable minima were observed for the rotation around the helix long axis. Only one of these minima (slightly higher in energy) placed residues Ala 33 , Ile 44 , and Ile 43 in contact with the transmembrane segments of the ␣ subunits. Because prior observations, based on mutations, showed that Ala 33 , Ile 44 , and Ile 43 are important for ␣-␥ interactions (23), the minimum with slightly higher energy was chosen to place the helix (seen in Fig. 14A). In the final model, the Phe 949 , Glu 953 , Leu 957 , and Phe 960 in M9 are in contact with ␥ in line with the recent mutational work (22). Additional TM contacts involve M6 and M2. The model in Fig.  14A is, in essence, a four-helix bundle consisting of M2, M6, M9, and TM␥ (41). It is similar to the recently published model (22), although the position of the transmembrane segment of ␥ is somewhat different.
The placement of the transmembrane segment served as an anchor for the manual, interactive modeling of the unstructured tail. One clear implication of the positioning of the TM segment between M2, M9, and M6 is that Lys 352 can be excluded as a cross-linking partner for Lys 54 or Lys 55 . Lys 352 faces the opposite side of the protein with respect to ␥ as seen in the ribbon diagram (Fig. 13A). The distance between the C-terminal end of the TM helix and Lys 352 is too far. Thus Lys 352 is not seen in the surface diagram of Fig. 13B that shows that only Lys 347 is at the same surface and thus close enough to be cross-linked to the cytoplasmic segment of the ␥ subunit.
Two different models for the positioning of the unstructured tail were evaluated. One places the tail as a rather straight extension from the TM segment toward Lys 347 , whereas the other follows a groove in the protein surface to the right as seen in Fig. 13B.
The straight model appears less desirable based on the following considerations. First, there is no obvious groove, extending straight up from the transmembrane segments, in which the tail can be placed and protected from tryptic digestion in 19-kDa membranes. Second, the high charge density of the unstructured tail, KKHR, cannot be satisfactorily compensated for. Only two acidic residues, Asp 745 and Asp 746 , are in close proximity, and one of these (Asp 746 ) appears to be involved in a salt bridge with Arg 589 . Third, only Lys 54 , and not Lys 55 , is at suitable distance to allow a cross-link to Lys 347 .
On the other hand, the model, which turns to the right, has features that render it more attractive. First, the tail lies in a groove, can make more extensive interactions, and may be protected from tryptic digestion. Second, positioning the tail in this groove automatically positions Lys 54 and Lys 55 in close enough proximity to the presumed cross-linking partner Lys 347 . Depending on the side-chain positioning and details of the backbone conformations, the distance between the terminal amine moieties is between 4 and 10 Å, and a reasonable conformation puts Lys 54 at a distance of 7.5 Å and Lys 55 at a distance of 6.3 Å from Lys 347 . Third, close to Lys 347 on S4, there are four acidic residues, namely Glu 756 and Glu 757 on S5 as well as Glu 821 and Asp 823 on the L6/7 loop, capable of interacting with Lys 54 , Lys 55 , His 56 , and Arg 57 of ␥. Thus, all positive charges are compensated by acidic residues. His 56 is also in close contact with Asn 910 of L8/9. Although we have no specific evidence, the sequence of the ␥ beyond Arg 57 , PINEDEL up to the C terminus, should interact with the ␣ subunit because truncation of the four residues EDEL abrogates the effect of ␥ on the K m for ATP (17). However, this interaction must be relatively weak because it includes the epitope of the anti-␥C antibody, which binds the ␥ subunit in the native protein (7).
It is important to emphasize that, whereas the model explains economically the available experimental data, the data are not sufficiently detailed to unambiguously place all the residues in context. The main value of this model is that it proposes testable hypotheses. It is nevertheless of interest that phospholamban, the regulator of sarcoplasmic/endoplasmic reticulum Ca-ATPase, has been shown to interact with the L6/7 cytoplasmic loop (52).
␤-␥ Cross-links-The major DST-and EDC-induced ␤-␥ cross-link is located in the extracellular domain of the ␤ subunit. There is a single lysine residue, Lys 11 , in the extracellular segment of ␥. Therefore, in the case of the DST, Lys 11 must be cross-linked to a lysine residue downstream of Arg 142 , the tryptic cleavage site producing the 50-and 16-kDa fragment in 19-kDa membranes (48). The less prominent DST-and EDC-mediated cross-links of ␥-␤ are located in the 16-kDa fragment, which contains the cytoplasmic N-terminal segment, transmembrane segment, and extracellular segment up to Arg 142 . In principle, these cross-links could be located at either surface of the 16-kDa fragment. However, in the case of DST, it is more likely that a lysine in the extracellular segment upstream of Arg 142 is the partner of the extracellular Lys 11 of ␥ because a cytoplasmic lysine of ␥ is involved in the ␣-␥ cross-link, and the ␥ subunit appears to fit into the groove between M2, M6, and M9 at the cytoplasmic side. The ␤ subunit is known to interact strongly with the ␣ subunit in the L7/8 loop (53) within residues 894 SYGQ (54), and Tyr 901 , Val 907 , and Cys 911 are also important (55). Crosslinking shows that TM␤ is close to TM8 of ␣, and in a previous study we showed that the 16-kDa fragment of the ␤ subunit can be covalently cross-linked to the ␣ subunit in L7/8 (within Tyr 895 -Tyr 901 ) (56), and Cu 2ϩ -catalyzed oxidative cleavages indicated that two regions of the ␤ subunit interact with the L7/8 loop (within residues 90 -115 and 194 -205, respectively) (57). The first of these sequences lies within the sequence span of the 16-kDa fragment (Lys 5 -Arg 142 ). Thus, one can propose that the extracellular domain of the ␥ subunits comes into proximity with the ␤ subunit near the site of the ␣-␤ interaction.