Characterization of Disulfide Cross-links between Fragments of Proteolyzed Na,K-ATPase

This study characterizes disulfide cross-links between fragments of a well defined tryptic preparation of Na,K-ATPase, 19-kDa membranes solubilized with C12E10in conditions preserving an intact complex of fragments and Rb occlusion (Or, E., Goldshleger, R., Tal, D. M., and Karlish, S. J. D. (1996) Biochemistry 35, 6853–6864). Upon solubilization, cross-links form spontaneously between the β subunit, 19- and 11.7-kDa fragments of the α subunit, containing trans-membrane segments M7-M10 and M1/M2, respectively. Treatment with Cu2+-phenanthroline (CuP) improves efficiency of cross-linking. Sequencing and immunoblot analysis have shown that the cross-linked products consist of a mixture of β-19 kDa dimers (≈65%) and β-19 kDa–11.7 kDa trimers (≈35%). The α-β cross-link has been located within the 19-kDa fragment to a 6.5-kDa chymotryptic fragment containing M8, indicating that βCys44 is cross-linked to either Cys911 or Cys930. In addition, an internal cross-link between M9 and M10, Cys964-Cys983, has been found by sequencing tryptic fragments of the cross-linked product. The M1/M2-M7/M10 cross-link has not been identified directly. However, we propose that Cys983 in M10 is cross-linked either to Cys104 in M1 or internally to Cys964 in M9. Based on this study, cross-linking induced byo-phthalaldehyde (Or, E., Goldshleger, R., and Karlish, S. J. D. (1998) Biochemistry 37, 8197–8207), and information from the literature, we propose an approximate spatial organization of trans-membrane segments of the α and β subunits.

The spatial organization of trans-membrane segments of the catalytic subunits is unknown although such information is essential for defining the structure of the cation transport path. Recently published structures of Ca-ATPase and H-ATPase (12,13), at 8-Å resolution, show that the 10 trans-membrane segments are ␣-helices and most helices are tilted so that their spatial organization changes at different levels in the membrane. A tentative model of helix packing of Ca-ATPase has been proposed based on constraints of trans-membrane topology, site-directed mutagenesis, and disulfide cross-linking (12,14,15). Modeling of Na,K-ATPase has also been attempted based on hydrophobic labeling and prediction of helix orientation with respect to lipid (16). Despite these attempts, direct determination of helix proximity is largely lacking. Recently, covalent cross-linking experiments have begun to provide such specific information (2,15). This paper describes experiments utilizing CuP 1 -catalyzed S-S bridge formation. Older work (17)(18)(19)(20) showed that treatment of detergent-solubilized Na,K-ATPase leads primarily to a 1:1 ␣␤ cross-linked product. Identification of the cross-linked residues constitutes a formidable task for, whereas the ␤ subunit has only one free cysteine, Cys 44 , located within its single trans-membrane helix (21), the ␣ subunit contains 23 cysteines. The problem could be simplified by using 19-kDa membranes which are obtained by extensive tryptic digestion of Na,K-ATPase. This preparation consists of fragments of the ␣ subunit comprising mainly trans-membrane segments (M1/ M2, M3/M4, M5/M6, and M7/M10) connected by the external loops, and a partially cleaved ␤ subunit (8,22). The fragments contain only 10 cysteines located within or next to trans-membrane segments.
A potential problem in cross-linking membrane proteins is the possibility of intermolecular cross-linking via random collisions in the membrane. Detergent solubilization can overcome this problem because the soluble protein can be diluted. Indeed Sarvazyan et al. (23,24) have reported that treatment of digitonin-solubilized 19-kDa membranes with CuP yielded two cross-linked products, dimers of fragments containing M7/ M10 and M1/M2 (22 kDa:11 kDa) and trimers containing these two fragments with the ␤ subunit (␤:22 kDa:11 kDa) in 1:1:1 stoichiometries. These studies established that a cysteine residue within Asn 831 -Tyr 1016 of the ␣ subunit (Cys 911 , Cys 930 , Cys 964 , or Cys 983 ) is cross-linked to the ␤ subunit (Cys 44 ), but did not identify the cysteine. 19-kDa membranes lack ATP-dependent functions, but re-tain cation occlusion and ouabain binding (8,22,25). Recently we described a procedure for solubilizing 19-kDa membranes with the non-ionic detergent C 12 E 10 which preserves intact the complex of fragments with occluded Rb ions and bound ouabain (1). The intact soluble complex contains one copy of each fragment. After solubilizing 19-kDa membranes in the absence of Rb ions and ouabain, neither Rb occlusion nor the complex of fragments are intact (1). Because cross-linking in the latter condition, as in Ref. 23, might reveal non-native interactions between fragments, one of our objectives has been to characterize cross-linking of the C 12 E 10 -solubilized intact complex of fragments containing occluded Rb ions. A second objective has been to identify cross-linked cysteines.
EXPERIMENTAL PROCEDURES Na,K-ATPase was prepared from pig kidney red outer medulla (26) and stored at Ϫ80°C. Protein and ATPase activity were determined as described (26). Specific activities were 13-18 units/mg of protein. Before use enzyme was thawed and dialyzed overnight at 4°C against 1000 volumes of 25 mM histidine, 1 mM EDTA(Tris), pH 7.0. 19-kDa membranes were prepared by tryptic digestion of Na,K-ATPase with trypsin (22), and resuspended at 3 mg/ml in 2 mM RbCl, 25 mM imidazole, 1 mM EDTA, pH 7.5. The specific Rb occlusion capacity was 6 to 6.5 nmol/mg of protein. 19-kDa membranes were solubilized with C 12 E 10 at a ratio of 2.2 (w/w) as described by Or et al. (1). Before cross-linking with CuP the pH was adjusted to 8.0 with RbOH (final concentration, 0.8 mM).
Cross-linking Catalyzed by CuP-The solubilized preparation (0.4 mg/ml) was treated at 20 -22°C with CuP (final concentration: 0.5/2.5 mM), added in 5 aliquots every 12 min. After 1 h solid urea was added to 2 M and the pH was readjusted to 8.0 with solid Tris base. Free cysteines were blocked by adding iodoacetamide to 40 mM and free Cu 2ϩ was chelated with 10 mM EDTA(Tris). After 30 min at 20 -22°C the mixture was acidified to pH Ϸ 6.0 with acetic acid and protein was precipitated with 4 volumes of methanol/ether (2:1) and stored at Ϫ20°C overnight. When analytical amounts were cross-linked, urea, Tris base, and acetic acid were omitted.
Gel Electrophoresis-Precipitated protein was collected by centrifugation at 9700 ϫ g for 1 h at 4°C, dried under a stream of nitrogen, and dissolved in loading buffer. Samples were resolved by Tricine-SDS-PAGE as in Refs. 22 and 27. Either 14-cm short or 23-cm long gels were used. In nonreducing conditions glutathione and mercaptoethanol were omitted from sample and running buffers.
Purification of CuP-catalyzed Cross-linked Products-Precipitated protein (15 mg) was dissolved in 2 ml of nonreducing sample buffer and resolved on two 1.5-mm thick long 10% Tricine gels. Gels were briefly fixed, stained, and destained and bands of interest were cut out and equilibrated with 660 mM Tris-Cl, pH 8.9, for 1 h. Protein was eluted into 50 mM NH 4 HCO 3 , 0.1% SDS using a Bio-Rad Model 422 Electro-Eluter, run at 8 mA/tube overnight. Eluted protein was precipitated with 4 volumes of methanol at Ϫ20°C after adjusting the pH to 7.0 with acetic acid. Pellets were collected by centrifugation and dissolved in 50 mM Tris-Cl, pH 8.0, 0.1% SDS. Samples were analyzed on a minigel to check the amounts and purities of the eluted proteins.
Digestion with N-Glycosidase F-Free or cross-linked ␤ subunit (0.1 mg/ml) in 50 mM phosphate buffer, pH 7.5, containing also 0.01% SDS and 0.7% octyl glucoside, was treated with 2500 units of N-glycosidase F at 37°C for 20 h. The reaction was stopped by adding loading buffer.
Sequencing-Bands of interest were cut out of the gel, equilibrated in 660 mM Tris-Cl, pH 8.9, for 1 h and transferred to PVDF by either of two procedures. (a) Bands were separated on a second 1.5-mm thick short 10% Tricine gel and electroblotted onto PVDF (22). (b) Bands were cut into small pieces, and protein was eluted into 1.7 ml of 1 M Tris-Cl, 0.1% SDS, pH 8.45, by mixing on a rotating wheel for 20 h. Supernatants were transferred into 3-ml syringes connected to 13-mm Swinnex filter holders (Millipore) and eluted proteins were blotted onto PVDF (Immobillon P SQ Millipore) by centrifugation at 290 ϫ g for 25 min. Sequencing of peptides blotted onto strips of PVDF was done on an Applied Biosystems Model 475A protein sequencer with an on-line Model 120A phenylthiohydantoin analyzer. Each reported sequence was done at least twice.
Materials-1,10-Phenanthroline-HCl was from BDH Chemicals, L-1tosylamido-2-phenylethyl chloromethyl ketone-trypsin (bovine pancreas) was from Worthington, ␣-chymotrypsin was from Merck, Nglycosidase F was from New England Biolabs, and octyl glucoside from Calbiochem. Trypsin inhibitor (type 1-S from soybean), phenylmethylsulfonyl fluoride, iodoacetamide, ouabain, thioglycolate, Tricine, C 12 E 10 , and molecular weight markers (SDS17 and VII-L) were from Sigma. Electrophoresis grade reagents for SDS-PAGE and PVDF paper were from Bio-Rad and Millipore, respectively. [␥-32 P]ATP was from Amersham. Antisera recognizing Asn 889 -Gln 903 and Lys 1012 -Tyr 1016 (anti-KETYY) and the catalytic subunit of PKA were gifts from colleagues (see "Acknowledgments"). Antisera were raised against an 11.7-kDa peptide containing M1/M2 (N-terminal Asp 68 ) and a 16-kDa fragment of the ␤ subunit (N-terminal Ala 5 ) as described previously (1). Fig. 1 presents a representative experiment showing components of C 12 E 10 -solubilized 19-kDa membranes in reducing or nonreducing conditions and also cross-linked products after treatment with CuP. In reducing conditions (Red), untreated C 12 E 10 -solubilized 19-kDa membranes reveal the standard 19-kDa and smaller (8 -12 kDa) fragments of the ␣ subunit (the latter are not well resolved on this 10% gel), an intact ␤ subunit or its Ϸ50-kDa glycosylated and 16-kDa fragments (22). In nonreducing conditions (Non-red) one major new band appeared, 70 -80 kDa, and the intensity of the 19-kDa fragment was less than in reducing conditions. The Ϸ50and 16-kDa fragments were not seen because they are internally cross-linked between Cys 125 and Cys 148 . After treatment of C 12 E 10 -solubilized 19-kDa membranes with CuP (Sol), the intensity of the 70 -80-kDa band rose and that of the 19-kDa peptide and ␤ subunit decreased further (compare Sol with non-Red), suggesting that this band contains both the 19-kDa peptide and ␤ subunit. About 63% of the ␤ subunit underwent cross-linking based on the weight ratio of the cross-linked product and remaining ␤ subunit extracted from gels (Ϸ3.1, w/w), and on the composition of the cross-linked product (see Table I). Raising CuP concentration or incubation time did not improve the yield (not shown). CuP treatment of unsolubilized 19-kDa membranes, containing occluded Rb ions, did not produce the 70 -80-kDa cross-linked band (Mem). After denaturation of 19-kDa membranes with SDS, CuP treatment did not produce the 70 -80-kDa band (SDS), excluding the possibility that the cross-link is the product of a nonspecific association of its components after termination of the reaction. A small amount of a higher molecular weight band (Fig. 1, asterisk) was observed after C 12 E 10 solubilization but the amount was not increased by CuP treatment. This band was also observed with intact 19-kDa membranes. It may represent the product of inter-molecular cross-linking between adjacent complexes, and therefore it was not investigated further. In conditions producing the 70 -80-kDa band the remaining 19-kDa fragment ran a little faster than in the other conditions, implying that the peptide is more compact (compare Non-red and Sol with Mem, and SDS). This hints at the possibility of an internal cross-link (see also Table II). Fig. 2 presents an immunoblot using anti-KETYY to detect cross-links in intact 19-kDa membranes. After CuP treatment in the presence of Rb ions (ϩRb) the antibody recognized the 19-kDa fragment but no higher molecular mass species, confirming the absence of cross-links in intact 19-kDa membranes seen in Fig. 1. By contrast, after CuP treatment in the absence of Rb ions (ϪRb) the amount of 19-kDa peptide was reduced, and a number of bands appeared (30.8 -66 kDa), as well as a smear of material in the upper half of the gel. Using antibodies raised against fragments containing M1/M2, M3/M4, and M5/M6 all these bands were recognized by anti-M1/M2 and some of them also by anti-M3/M4 or anti-M5/M6 (not shown). This heterogeneity suggests that the cross-links might be nonspecific. In fact the same pattern of anti-KETYY recognition was found for 19-kDa membranes in which Rb occlusion had been thermally inactivated (30, 31) (Fig. 2, Th.in). As discussed below these cross-links cannot be assumed to represent proximities between fragments in a native state.

CuP-catalyzed Cross-linked Peptides-
The components of the 70 -80-kDa cross-linked product have been identified by sequencing (Table I) and immunoblots (e.g Fig. 3). Immunoblots showed that both spontaneously formed and CuP-induced cross-linked product contain the ␤ subunit 19-kDa peptide (M7-M10) and an 11.7-kDa peptide (M1/M2) (not shown). N-terminal sequencing clearly identified four fragments (Table I). Two sequences are derived from the ␤ subunit (N terminus Ala 5 in 19-kDa membranes, Ref. 22), which is partially cleaved to fragments of 16 kDa (N terminus Ala 5 ) and   19-kDa membranes (0.5 mg/ml) were suspended in 25 mM imidazole, 1 mM EDTA, pH 7.5, without (ϪRb, Th.in) or with 30 mM RbCl (ϩRb). Membranes were either kept on ice (ϪRb) or were incubated at 37°C for 25 min (ϩRb, Th.in). All samples were then treated with CuP as described under "Experimental Procedures," and then dissolved in 2% SDS and treated with 40 mM iodoacetamide and 10 mM EDTA for 30 min at room temperature. Protein was precipitated with 4 volumes of methanol. The samples were resolved on a 10% gel in nonreducing conditions, transferred to PVDF paper, and then probed with the anti-KETYY antibody. Ϸ50 kDa (N terminus, Gly 143 ). The sequence with N terminus Ala 5 represents both the 16-kDa fragment and intact ␤ subunit, and about 65% is cleaved judging by the yields of 26 and 17 pmol/cycle, respectively. Two sequences are derived from the ␣ subunit: 19 kDa (N terminus Asn 831 , 17 pmol/cycle) and 11.7 kDa (N terminus Asp 68 , 9 pmol/cycle). Note that the yields of the components are not stoichiometric. The excess ␤ subunit (26 pmol/cycle) over the 19-kDa fragment (17 pmol/cycle) implies that the cross-linked product was not completely resolved from free ␤ subunit. The different yields of the 19-and 11.7-kDa fragments could imply that the cross-linked product consists of either a mixture of ␤:19-kDa dimers (Ϸ65%) and ␤:19 kDa:11.7 kDa trimers (Ϸ35%) or a mixture of dimers, ␤:19 kDa and ␤:11.7 kDa, which are not well resolved on the gel. Deglycosylation of the ␤ subunit has allowed us to distinguish between these alternatives. The deglycosylated cross-linked product was clearly resolved into two bands of 62-and 54 kDa (Fig.  3A, Coom). Both bands were recognized by anti-␤ antibodies and anti-KETYY but only the upper 62-kDa band was recognized also by anti-M1/M2. Thus the cross-linked product consists of a mixture of ␤:19-kDa dimers and ␤:19 kDa:11.7-kDa trimers.
Identification of Cross-linked Residues and Segments-The following experiments demonstrate cross-linking of Cys 44 of the ␤ subunit to a cysteine in M8 (Cys 911 or Cys 930 ), and an internal cross-link between Cys 964 (M9) and Cys 983 (M10). We have not been able to identify the M7/M10-M1/M2 cross-link, probably because the ␤:19 kDa:11.7-kDa trimer was formed in too low a yield (see "Discussion").
One approach involved tryptic digestion of the 70 -80-kDa product and sequencing of cross-linked fragments. Crosslinked fragments were identified as bands present in nonreducing but absent in reducing conditions (Fig. 4). Several such bands were observed (marked **, *, a and b). The bands marked with an asterisk (*) were not sequenced for they were observed also in digests of the ␤ subunit and represent fragments cross-linked by internal S-S bridges (not shown). The two fragments a, 9.3 kDa, and b, 7.7 kDa, were not observed in digests of the ␤ subunit and were sequenced (Table II). Fragment a contained three peptides derived only from the ␣ subunit, with N termini Met 973 (M10), Asn 944 , and Ile 946 (M9), respectively. The combined average yield of fragments containing M9 is 12.6 pmol/cycle while that of the fragment containing M10 is 11.7 pmol/cycle. These yields are measured against a background of about 1 pmol/cycle, i.e. with a possible error of about 10%. Thus, the experiment demonstrates that M9 and M10 are present in equimolar proportions, suggesting that Cys 964 within M9 and Cys 983 within M10 are cross-linked to each other. Sequencing of fragment b led to essentially the same result suggesting that it is a truncated version of fragment a (probably cleaved at Arg 998 or Arg 1003 ).
The evidence for the internal cross-link Cys 964 -Cys 983 in Table II excludes either residue as a partner of Cys 44 of the ␤ subunit, unless one makes an unlikely assumption that Cys 44 can form S-S bridges with different cysteines in the 19-kDa peptide. Thus, one could predict that Cys 44 is cross-linked to either Cys 911 or Cys 930 in M8. However, direct evidence was not obtained because we detected no cross-linked band containing sequences from both ␣ and ␤ subunits. A hypothesis to explain the paradox could be that the broad band (**) in Fig. 4 contains cross-linked fragments of both ␣ and ␤ subunits, but the latter are also cross-linked by the internal S-S bridges to other glycosylated fragments of the ␤ subunit. Sequencing of this band (**) showed indeed that it consists of a heterogenous mixture of peptides which precludes simple interpretation (not shown).
The hypothesis just proposed was tested as follows (Figs. 5, 6, and Table III). The 70 -80-kDa cross-linked product, or the 19-kDa fragment, were first incubated with protein kinase A and [␥-32 P]ATP, in order to phosphorylate Ser 936 in the PKA site RRNS (32,33), and the labeled proteins were then digested with chymotrypsin. 2 The digestion products were resolved on a 16.5% gel, blotted onto PVDF paper, which was first autoradiographed (Fig. 5, 32 P i ) and then also immunostained with an antibody recognizing residues Asn 889 -Gln 903 in the loop between M7 and M8 (Fig. 5, L7-8). For greater ease of understanding, the scheme in Fig. 5 marks the positions of the Asn 889 -Gln 903 epitope and the phosphorylation site of PKA in relation to the trans-membrane segments, M7-M10. In nonreducing conditions only broad, poorly defined bands of 32 P ilabeled material (*) were observed (Non-red). However, in reducing conditions, a discreet 32 P i -labeled band of 6.5 kDa appeared (Red). The same 6.5-kDa band appeared also in a chymotryptic digest of 32 P i -labeled 19-kDa fragment resolved under reducing conditions (19 kDa). Thus the 6.5-kDa band is a fragment of the ␣ subunit containing Ser 936 which is crosslinked to the broad band of material in nonreducing conditions and is released by reduction of an S-S bridge. By contrast to the result with PKA phosphorylation of Ser 936 , the anti-Asn 889 -Gln 903 antibody recognized an 8.0-kDa chymotryptic fragment, the size of which was not affected by reduction (Fig. 5, L7-8). Thus, this fragment is not a cross-linked species and does not contain the PKA site at Ser 936 (compare L7-8 with 32 P i ).
In principle, the 6.5-kDa 32 P i -labeled cross-linked fragment might extend forward from Ser 936 and include M9 (containing Cys 964 ) or backward from Ser 936 and include M8 (containing Cys 911 and Cys 930 ). In order to distinguish between these possibilities we have sequenced the fragment, repeating the experiment of Fig. 5 with 50-fold more protein (Fig. 6). The CuP cross-linked product (0.3 mg) was labeled with 32 P i using PKA, digested with chymotrypsin, and protein was separated on a 16.5/6% gel. Most of the digest was separated under reducing conditions (Red), but for comparison a portion was separated in nonreducing conditions (Non-red). Because a large amount of protein was loaded onto each lane, the labeled fragment could not be expected to run as a sharp band even in the reducing conditions, as in Fig. 5, and indeed the autoradiograph in Fig.  6 (Red) showed a relatively broad band of labeled material in the region of 6 -7 kDa. The broad labeled band was transferred to PVDF, and sequenced (Table III). A mixture of two sequences was found, N terminus Glu 902 corresponding to a fragment containing M8 and N terminus Arg 972 corresponding to a fragment containing M10, respectively. Only the former fragment contains Ser 936 , and hence the radioactive label. Thus, Figs. 5, 6, and Table III demonstrate that Cys 44 of the ␤ subunit is cross-linked to either Cys 911 or Cys 930 in M8 (see "Discussion").

Cross-linked Fragments of 19-kDa Membranes
CuP treatment of 19-kDa membranes solubilized with C 12 E 10 in the presence of Rb ions and ouabain led to appearance of one major cross-linked band, which consists of a mixture of ␤:19-kDa dimers (Ϸ65%) and ␤:19 kDa:11.7-kDa trimers (Ϸ35%) (Figs. 1 and 3, Table I). These cross-links reflect proximities of fragments within the intact detergent-solubilized complex. The fact that cross-links were formed in detergent-solubilized but not in native 19-kDa membranes could imply either that the detergent induced a degree of rearrange-ment of trans-membrane segments which permit cross-links between the relevant cysteines or that cysteines embedded in lipid, such as Cys 44 of the ␤ subunit, are not reactive in native membranes due to insufficient exposure to oxygen and CuP, and only become exposed to oxygen or CuP after solubilization. However, the crucial point is that Rb occlusion is fully preserved in these C 12 E 10 -solubilized 19-kDa membranes (1). Therefore the organization of trans-membrane segments in the soluble complex of fragments, and proximities revealed by cross-linking (M8/M␤, M9/M10), must be essentially similar to those in native 19-kDa membranes and Na,K-ATPase.
CuP treatment of digitonin-solubilized proteolyzed dog kidney Na,K-ATPase produces cross-linked fragments corresponding to a trimer of ␤:22 kDa:11 kDa (1:1:1) and, in lower amounts, a dimer of 22 kDa:11 kDa 3 (23). Our study confirms the finding of a ␤:19 kDa:11.7-kDa trimer, although in lower yield than a ␤:19-kDa dimer, but a 19 kDa:11.7-kDa dimer was not observed (see Fig. 1). Presumably, the similarities and differences between the two sets of observations reflect the state of the soluble complex. Solubilization of pig kidney 19-kDa membranes with C 12 E 10 , in the absence of Rb ions and ouabain, leaves the ␤-M7/M10 pair tightly associated but the M5/M6 and M3/M4 fragments dissociate, while the M1/M2 fragment shows an intermediate degree of interaction with the ␤:19-kDa pair (1). Thus, solubilization of 19-kDa membranes by digitonin in the absence of Rb ions and ouabain (23) should not have preserved an intact complex. Tighter association of the 11-kDa fragment with the ␤:22-kDa pair could explain why FIG. 5. Phosphorylation and chymotryptic digestion of the CuP-catalyzed cross-linked product. The CuP cross-linked product (10 g, 0.25 mg/ml) was labeled with 15 Ci of [␥-32 P]ATP using PKA (see "Experimental Procedures"). The protein was precipitated with methanol and then digested with chymotrypsin at 0.2 mg/ml for 1.5 h. The digest was divided and either reduced (Red) or not (Non-red). The 19-kDa fragment (2 g, 0.1 mg/ml) was also labeled with [␥-32 P]ATP, digested with chymotrypsin, and then reduced. Samples were resolved on a short 16.5% T, 6% C gel, and blotted onto PVDF. The sheet was autoradiographed using a Fuji BAS 1000 PhosphorImager ( 32 P i ) and then immunostained with an antibody recognizing residues Asn 889 -Gln 903 (L7-L8). a ␤:22 kDa:11-kDa trimer was observed even in the absence of Rb ions and ouabain (23). CuP treatment of unsolubilized 19-kDa membranes revealed a heterogeneous mixture of cross-linked products containing the 19-kDa fragment, in the absence of Rb ions, and suppression of the cross-links in the presence of Rb ions (Fig. 2). These cross-links are probably the products of non-native interactions since the same fragments were observed after thermal inactivation which disorganizes the fragments (30, 31). K ϩ , Na ϩ , or ouabain protect against CuP-catalyzed cross-linking of fragments of unsolubilized 19-kDa membranes (24). These ligands also protect 19-kDa membranes against thermal inactivation of Rb occlusion (30, 31). Thus protection against CuP-catalyzed cross-linking presumably reflects cation-or ouabain-induced stabilizing interactions between fragments which prevent their disorganization.

Identity of Cross-linked Regions
The conclusion that Cys 911 or Cys 930 in M8, is cross-linked to Cys 44 of the ␤ subunit rests on a combination of direct evidence and exclusion of alternatives. The inference of the internal cross-link (Fig. 4, Table II, Fig. 1) suggest that M9 and M10 form a hairpin with Cys 964 and Cys 983 juxtaposed, and excludes both residues as partners for the ␤ subunit. Direct evidence has been obtained by utilizing PKA to label Ser 936 with 32 P i (32,33). After chymotryptic digestion of the 32 P i -labeled crosslinked product, reducing conditions released a 6.5-kDa 32 P ilabeled fragment, with N terminus Glu 902 (Figs. 5 and 6, Red; Table III). This 6.5-kDa fragment includes Cys 911 and Cys 930 in M8 and ends before Cys 964 in M9. In nonreducing conditions the fragment is cross-linked to a mixture of partially digested glycosylated fragments of the ␤ subunit, held together by internal S-S bridges (Figs. 4 -6).
Proximity of M8 and M␤ is compatible with prior evidence that ␣ and ␤ subunit interact strongly at the extracellular surface primarily within the short sequence SYGQ outside M8 (34,35). It is also compatible with our finding that o-phthalaldehyde cross-links ␣ and ␤ subunits near M8 (2). Although we cannot say which cysteine in M8 is cross-linked to the ␤ subunit, Cys 911 is a more likely candidate. A helical wheel representation of M␤ (Fig. 7) reveals a sector of about 100°with short or non-hydrophobic side chains (Gly 43 , Cys 44 , Gly 47 , Gly 51 , Gln 54 , shaded boxes) near the extracellular surface. The nonhydrophobic sector could participate in a protein-protein interface with the remaining hydrophobic surface facing the lipid. Thus M␤ may contact M8, including Cys 911 , near the extracellular surface.
The cross-link between M7/M10 and M1/M2 fragments has not been identified. Based on topological considerations, Sarvazyan et al. (23) proposed that Cys 983 -Cys 104 (M1-M10) or Cys 138 -Cys 930 (M2-M8) are likely pairs. At first sight a Cys 983 -Cys 104 cross-link appears incompatible with the Cys 983 -Cys 964 internal cross-link (Table II) short extracellular loops. Obviously, Fig. 8 depicts only an approximate arrangement, due to lack of detailed information on helix proximity and tilt. The major objective is to illustrate the structural constraints discussed below.
The thick dashed line in Fig. 8 demarcates separate domains comprising M1 to M6 and M7 to M10 with M␤, respectively, evidence for which is provided by the following observations. (a) Compared with class II eukaryotic P-type pumps with 10 transmembrane helices, prokaryotic class I P-type heavy metal pumps contain the equivalent of only the first six helices and conserved cytoplasmic regions (with two extra helices at the N-terminal side). It has been proposed that sequences corresponding to M1 to M6 of class II pumps represent a core structural and functional unit for cation transport and energy transduction, while the extra four C-terminal segments of class II pumps evolved to serve additional functions (3,36). (b) We have shown recently that Fe 2ϩ (37) and Cu 2ϩ (38) ions catalyze selective oxidative cleavages of Na,K-ATPase, providing information on spatial organization around the bound metal ions. Iron-dependent cleavages of the ␣ subunit occur at the cytoplasmic surface in conserved sequences located between M2 and M3 and between M4 and M5 (37). Copper-dependent cleavages occur at the extracellular surface between M7 and M8, to a small extent between M5 and M6, and between M9 and M10, and in the ␤ subunit (38). A comparison of iron-and coppercatalyzed cleavages is highly suggestive of division of the ␣ subunit into two domains comprising M1-M6 and M7-M10/M␤, respectively (38). (c) High resistance to digestion by selective and unselective proteases of the C-terminal fragment (M7-M10) of Na,K-ATPase (8,22) and H,K-ATPase (39) implies that it is folded into a compact domain.
Division into domains of M1-M6 and M7-M10 provides a strong constraint on the way helices can or cannot be packed. For example, M7 and M8 are unlikely to be widely separated in the molecule although they are connected by a relatively long extracellular loop (Ϸ40 residues) which, in principle, could permit such a separation. The positions allocated to M1-M10 of the ␣ subunit and M␤ are based on the following considerations.
M7, M8, and M␤-Proximity of M7 to M5 is strongly implied by our recent finding of a cross-link between cytoplasmic segments just preceding M7 and M5 (2). M6 and M7 are connected by a fairly short cytoplasmic loop, and charged residues within the loop may form the entrance to occlusion sites in M4, M5, and M6 (42). Therefore M7 is also placed near M6. As argued above, M7 and M8 are unlikely to be widely separated and proximity of M8 to M4, M5 and M6, is suggested by mutations of Glu 908 in Ca-ATPase (9,43). Thus M7 and M8 are placed together. M␤ interacts with M8 as required by the CuP crosslinking (␤Cys 44 -Cys 911 or perhaps ␤Cys 44 -Cys 930 ), and the evidence for interaction between ␤ and ␣ subunits near the entrance to M8 (35,2). As seen in Fig. 7 a likely surface of interaction of M␤ occupies a sector of only 100°with the remaining hydrophobic sector facing the lipid. This is suggestive of the triangular contact in Fig. 8 with M␤ placed between M7 and M8.
M1, M2, and M3-The major constraints on the locations of these helices are the indications for the two domains and the necessity for M3 to be close to M4 and M2 to M1. We have proposed above that M1 is cross-linked either to M10 (Cys 104 -Cys 983 ) or M9 (Cys 104 -Cys 964 ), the M1-M10 interaction (Cys104-Cys983) depicted in Fig. 8 is a more likely possibility due to other constraints on the interactions of M9 referred to in the previous paragraph.
Comparison of Fig. 8 with the helix packing model of Ca-ATPase (12) shows a general similarity in positions of M1-M7 as well as a major difference in that M8, M9, and M10 are placed on opposites sides of the molecules. Conceivably the position of M7 relative to M8-M10 is affected by the ␤ subunit which interacts strongly with the L7/8. However, it is equally likely that the uncertainty is due to lack of detailed information. Two structural constraints used in Ref. 12, namely proximity of M4-M6 based on S-S bridges (15) and orientation of nonconserved residues of trans-membrane helices toward the lipid, fit either arrangement. A third line of evidence involved placing helices identified as the cytoplasmic stalk, S2-S5, above M2-M5 in the middle of the molecule (12). Without comparable structural data for Na,K-ATPase it is not known whether this constraint applies. Specific predictions in Fig. 8, are that M4 and M6 are largely surrounded by neighboring helices rather than lipid (see Ref. 47), M1 is close to M10, and there is no proximity between M1/M2 and M7/M8. Clearly, it now becomes necessary to devise a means to test specific predictions and assumptions in order to refine the models.  (2) or cysteine residues cross-linked by CuP in this study are joined by gray lines. The gray rectangle in the ectodomain of the ␤ subunit represents a region which interacts within the extracellular loop between M7 to M8 (SYGQ). The division between two membrane domains is depicted by the thick dashed line.