INTRODUCTION

Na,K-ATPase is a ubiquitous plasma membrane transporter that is responsible for the maintenance of the potassium and sodium homeostasis in animal cells. The minimal functional enzyme unit is composed of two subunits. The large  subunit has 10 putative transmembrane spans and carries out the catalytic and transport functions of the enzyme. The smaller  subunit is a type II glycoprotein with a short cytoplasmic domain, one transmembrane segment, and a large C-terminal ectodomain that comprises several N-glycosylation sites and three highly conserved disulfide bridges. The  subunit plays an essential role in the structural and functional maturation of the  subunit. Only  subunits that assemble with  subunits in the ER¹ are stably expressed, acquire their functional properties, and are transported to the plasma membrane (for review see Ref. 1). Upon assembly, the  subunit itself also undergoes a structural maturation that is reflected by an increased resistance against proteolysis (2).

The molecular mechanisms that guide the important process of assembly of the  subunit with the  subunit are still poorly understood. The structural requirements for - assembly have only been partially characterized (3-7). Several regions in the  subunit participate in the interaction with the  subunit. It is possible to distinguish, on the one hand, structural interaction sites including the ecto- and the transmembrane domains of the  subunit that mediate the maturation of the  subunit and, on the other hand, functional interaction sites such as the cytoplasmic N terminus that participate in the modulation of the transport activity of the functional enzyme (2).

Little is known how the  subunit acquires a conformation that is compatible with assembly, e.g. how it folds into a correct tertiary structure that exposes the structurally defined surfaces assuring the specific interaction with the  subunit. Molecular chaperones such as BiP (binding protein) are associated with the newly synthesized  subunit (8) suggesting that they might be important for the correct folding and/or in the ER retention of unassembled  subunits.

Because the  subunit, as many other membrane and secretory proteins, acquires N-linked sugars and forms intramolecular disulfide bonds during its synthesis in the ER, it appeared interesting to us to use this protein as a model protein to compare the role of these co-translational modifications in protein folding, assembly, and in the functional expression. The role of N-glycosylation and disulfide bonds in the folding of proteins has been extensively studied (for review, see Refs. 9 and 10) but rarely compared in the same protein. Their importance in the folding process varies. When N-glycosylation is prevented by site-directed mutagenesis, many but not all glycoproteins misfold, aggregate, and degrade in the ER (11). Disulfide bonds do or don't directly affect the folding process, but, in most cases, their formation stabilizes the folded form. In this study, we used the Xenopus oocyte expression system to study 1) the structural properties of mutant  subunits deficient in glycosylation or disulfide bonds, 2) their interaction with BiP, and 3) their ability to associate with  subunits and to be transported to the plasma membrane as functional · complexes. Our results show that both co-translational modifications play an important but qualitatively distinct role in the correct folding of the  subunit. Indeed, the interaction of the  subunit with molecular chaperones, its stability, and its association efficiency are differently affected after abolition of glycosylation sites or cysteine bonds.

EXPERIMENTAL PROCEDURES

Site-directed mutagenesis was performed according to the polymerase chain reaction method of Nelson and Long (12). A linearized pSD5 vector (13) containing a 1cDNA (pSD51) of Xenopus Na,K-ATPase (14) was used as a template for the preparation of cysteine mutants (see Fig. 1). First, single mutants in which Cys¹⁴⁹, Cys¹⁷⁵, or Cys²⁷⁶ were substituted by a serine residue were prepared. The DNAs of the single mutants served as templates to prepare double mutants in which Cys¹²⁶/Cys¹⁴⁹ (CC1), Cys¹⁵⁹/Cys¹⁷⁵ (CC2), or Cys²¹³/Cys²⁷⁶ (CC3), respectively, were substituted with serine residues. The mutated DNA fragments of the CC1 and CC2 mutants were introduced into the wild type pSD51 by using unique BamHI and HindIII restriction sites and that of the CC3 mutant by using HindIII and SmaI (present in the vector).

Site-directed mutagenesis of the 1 subunit of Xenopus Na,K-ATPase affecting cysteine residues implicated in disulfide bond formation and asparagine residues implicated in N-glycosylation.

A linearized pGEM₂ vector or a pSD5 vector containing a 1cDNA (pGEM₂1, pSD51) of Xenopus Na,K-ATPase was used as a template for the preparation of the N2 or the N1 and the N3 glycosylation mutants, respectively (see Fig. 1). In single mutants Asn¹⁵⁸, Asn¹⁹³, or Asn²⁶⁵, respectively, were substituted by a glutamine residue. The mutated DNA fragments of the N1 and N2 mutants were introduced into the wild type pSD51 by using BamHI and PstI restriction sites and that of the N3 mutant by using HindIII and SmaI. For the construction of the double mutant N2,3, the mutated BamHI-PstI fragment of the pGEM1 N193Q mutant was introduced into the pSD51 N265Q (N3) mutant. For the preparation of the triple mutant N1,2,3, the mutated BamHI-HindIII fragment of the N158Q (N1) mutant was introduced into the pSD51 N193Q/N265Q (N2,3) mutant. All polymerase chain reaction-generated fragments were sequenced by dideoxy sequencing (15). All mutant 1 subunits were tested for their translation efficiency in a reticulocyte lysate (6).

Expression in Xenopus Oocytes and Immunoprecipitation of  and  Subunits of Na,K-ATPase

cRNAs encoding Xenopus 1 (14), 1 (14), and  mutants were obtained by in vitro transcription (16). Stage V and VI oocytes were obtained from Xenopus females (Noerdhoek, Republic of South Africa) as described (6). Routinely, 5-7 ng of  and/or 0.2-0.6 ng of cRNA were injected into oocytes. Oocytes were incubated in modified Barth's medium containing 0.6-2.5 mCi/ml [³⁵S]methionine (Amersham Corp.) for the times indicated in the figure legends and eventually subjected to different chase periods in the presence of 10 mM cold methionine. Microsomes were prepared as described (2), and the  and  subunits of Na,K-ATPase or BiP were immunoprecipitated under denaturing or nondenaturing conditions as described (6, 8) by using  antibodies produced against the N terminus of the Xenopus 1 subunit (17),  antibodies against the ectodomain of the Xenopus 1 subunit (17), or BiP antibodies produced against Xenopus BiP (8). In some instances, immunoprecipitates were subjected to endoglycosidase (Endo H) (Calbiochem) treatment as described (18). Dissociated immunocomplexes were separated by SDS-PAGE in the absence or presence of 2% -mercaptoethanol and the labeled proteins revealed by fluorography. Quantifications and determinations of the molecular mass of the immunoprecipitated bands were performed with an analytic program for electrophoretic images (Bio-1D) from Vilbert Lourmat (Marne La Vallée, France).

Proteolysis of  Subunits

Wild type and mutant cRNAs were injected into oocytes together with 3 µCi/oocyte of [³⁵S]methionine. Oocytes were incubated for 5 h before homogenization by 12 gentle strokes with a glass Teflon homogenizer in a solution containing in mM, 250 sucrose, 50 potassium acetate, 5 MgCl₂, 1 dithiothreitol, 50 Tris-HCl, pH 7.5. After addition of 10 mM CaCl₂, samples were incubated in the absence or presence of 200 mg/ml proteinase K (Merck) and in the absence or presence of 1% Triton X-100 for 1 h at 4 °C. Proteolysis was stopped with 5 mM phenylmethylsulfonyl fluoride and 3.7% hot SDS before immunoprecipitation of the samples.

[³H]Ouabain binding was essentially done as described (6). After cRNA injection, oocytes were incubated for 2-4 days. Oocytes were then loaded with Na⁺ by incubation for 2 h in a solution containing in mM, 110 NaCl, 10 Tris-HCl, pH 7.5. After recovery for 30 min in a solution containing in mM, 90 NaCl, 2 CaCl₂, 5 BaCl₂, 20 tetraethylammonium chloride, 5 MOPS, pH 7.4, the oocytes were transferred to a solution containing in mM, 50 NaCl, 50 N-methyl-D-glucamine-Hepes, 2 CaCl₂, 5 BaCl₂, 20 tetraethylammonium chloride, 5 MOPS, pH 7.4, and 1 µM [21,22-³H]ouabain (Amersham Corp.) was added. In some instances, increasing concentrations of KCl were added to study the competition of ouabain binding. In this case, the osmolality of the solution was maintained by adapting the concentration of N-methyl-D-glucamine-Hepes. After 12 min at 19 °C, oocytes were washed with a solution containing in mM, 90 NaCl, 2 CaCl_2, 5 BaCl₂, 5 MOPS, pH 7.4, before dissolution in 0.5% SDS and counting. Nonspecific ouabain binding determined in the presence of a 1000-fold excess of cold ouabain amounted to about 3-7% of the total binding.

To study the K⁺ activation of the Na,K-ATPase-mediated ⁸⁶Rb⁺ uptake, oocytes were injected with Xenopus wild type or mutant 1cRNA plus cRNA coding for 1 subunits of Bufo marinus Na,K-ATPase that is more ouabain resistant than the Xenopus 1 subunit (19). After loading of the oocytes with Na⁺ (see above) and recovery in a solution containing in mM, 90 NaCl, 1 CaCl₂, 5 BaCl₂, 10 Hepes, pH 7.4, oocytes were transferred to a solution containing varying concentrations of KCl, 90 mM TMA-Cl or 90 mM NaCl, 2 mM CaCl₂, 5 mM BaCl₂, 1 mM MgCl₂, 10 mM Hepes, pH 7.4, and 1 µM ouabain to inhibit the endogenous, oocyte Na,K-ATPase (19). After addition of 5 µCi/ml ⁸⁶Rb⁺ (Amersham Corp.), oocytes were incubated for 12 min at room temperature before washing in a solution containing in mM, 90 TMA-Cl, 1 CaCl₂, 1 MgCl₂, 10 KCl, Hepes, pH 7.4. To determine the ouabain sensitivity of the Na,K-ATPase-mediated ⁸⁶Rb⁺ uptake, oocytes were injected with Xenopus wild type or mutant 1cRNA plus Xenopus 1cRNA, and ⁸⁶Rb⁺ uptake measurements were performed in the presence of 90 mM NaCl and varying concentrations of ouabain. The parameters of the Hill equation V = V_max/[1+(K_1/2/CK)n_H were fitted to the data of the ⁸⁶Rb⁺ uptake (V) induced by various concentrations of K⁺ (CK) and yielded least-square estimates of the maximal uptake (V_max), the half-maximal activation constant (K1/2), and the Hill coefficient (n_H).

RESULTS

Effects of Abolition of Disulfide Bonds in the  Subunit of Na,K-ATPase on Its Initial Folding and Its Interaction with BiP

To assess the role in the initial folding of the three disulfide bonds present in the  subunit of oligomeric P-type ATPase, we introduced mutations in each S-S bond-forming cysteine pair (CC1, CC2, CC3, see Fig. 1) of Xenopus 1 subunits. The degree of misfolding of the disulfide bond-deficient mutants was probed by various tests after expression and metabolic labeling in Xenopus oocytes. In a first test, the structural characteristics of the immunoprecipitated, wild type  subunits and of the cysteine mutants were compared by performing SDS-PAGE under nonreducing conditions (Fig. 2). The abolition of each one of the disulfide bonds appears to have a distinct effect on the structure of the mutants since their migration in the absence of -mercaptoethanol differed slightly from that of the wild type and among each other (Fig. 2A, lanes 5-8). In comparison, the migration in the presence of -mercaptoethanol of wild type and mutant  subunits was generally slower and identical (Fig. 2A, lanes 1-4). Overexposure of a gel from a similar experiment (Fig. 2B) revealed in addition that a small fraction of individual Xenopus 1 subunits, which are retained in the ER (17), occur as disulfide-bonded multimers. However, aggregate formation seen with nonreduced samples is quantitatively and qualitatively similar to wild type (Fig. 2B, lane 5) and disulfide bond-deficient (lanes 6-8)  subunits. This result indicates that the lack of one cysteine bond does not enhance the tendency of individual  subunits to form inappropriate, intermolecular disulfide bonds and  aggregates.

Migration of Xenopus 1 cysteine mutants on SDS-polyacrylamide gels under nonreducing conditions.

To further characterize the structural effect of the abolition of disulfide bonds, homogenates of oocytes expressing wild type or mutant  subunits were subjected to proteolysis in the absence or presence of Triton X-100. Proteinase K did not attack the wild type  subunit (Fig. 3A, lanes 1 and 2) but digested the three mutants by about 40% (lanes 5, 8, and 11). Probably due to partial cleavage of the cytoplasmic N terminus (2), protease digestion also produced a slight shift in the molecular mass of the residual mutant proteins (lanes 5, 8, and 11) and to a lesser extent in that of the wild type  subunits (lane 2). Finally, in the presence of Triton X-100, which opens closed membrane vesicles and exposes the ectodomain of the  subunit, only the wild type (lane 3) but not the mutant  subunits (lanes 6, 9 and 12) is partially resistant to proteinase K digestion. Thus, abolition of each one of the disulfide bonds produces a structural effect on the  subunit that is reflected by an increased protease sensitivity.

Sensitivity of  cysteine mutants to degradation by proteinase K and cellular proteases and their interaction with BiP.

We have previously shown that Xenopus 1 when expressed in Xenopus oocytes without an  subunit is slowly degraded in or close to the ER (17). Despite the increased protease sensitivity, the disulfide bond-deficient mutants showed a similar rate of cellular degradation to the wild type  subunits (Fig. 3B). Furthermore, the wild type as well as the newly synthesized CC1, CC2, and CC3 mutants interacted with BiP and the chaperone interaction correlated with the degradation rate of the  subunits (compare Fig. 3, C with B).

Thus altogether, these data indicate that abolition of only one disulfide bond in the  subunit leads to a significant change in its tertiary structure which, however, does not influence the cellular half-life of the proteins or their interaction with BiP.

Association of  Cysteine Mutants with  Subunits and Formation of Functional · Complexes

To assess whether the cysteine mutants can adopt an assembly-competent conformation, we co-expressed the mutants with  subunits in oocytes and performed immunoprecipitations with an -antibody under nondenaturing conditions that preserve - interactions. As expected, wild type  subunits co-precipitated with  subunits either in their core-glycosylated form after a pulse (Fig. 4A, lane 3) or in their fully glycosylated form after a chase period (lane 4) reflecting the transport of the · complexes from the ER to the plasma membrane (18). Furthermore,  subunits that were expressed together with  subunits became stabilized (lanes 3 and 4) in contrast to  subunits that were synthesized alone in the oocyte (lanes 1 and 2). The CC1, CC2, and CC3 mutants associated, although inefficiently, with co-expressed  subunits during a pulse period (lanes 5, 7 and 9), but they did not visibly stabilize them (lanes 6, 8 and 10). As shown for the CC1 mutant, the cysteine mutants also associated transiently with the endogenous, oocyte  subunits (lanes 11 and 12) that represent a stable pool of unassembled  subunits (15). Results obtained on the conformation and on the ability to assemble with  subunits were similar for mutants affected in only one of the cysteine residues forming the disulfide bonds (data not shown).

Assembly of  cysteine mutants with the  subunit of Na,K-ATPase.

During the chase period, the CC2 and CC3 mutants co-expressed with  subunits were degraded in a pre-Golgi compartment in their core-glycosylated form, similar to when they are expressed alone (compare Fig. 4B, lanes 7-10 to lanes 11 and 12 and to Fig. 3B). On the other hand, a small portion of the CC1 mutants became fully glycosylated when they were co-expressed with  subunits (Fig. 4B, lane 6) but not when they were expressed alone (lane 12). This result suggested that in contrast to CC2 and CC3 mutants, the CC1 mutant is still able, although inefficiently, to form · complexes that can leave the ER. To verify this observation, we tested the expression of functional pumps at the cell surface by measuring [³H]ouabain binding and Na,K-ATPase-mediated ⁸⁶Rb⁺ uptake in oocytes co-expressing  and wild type  or one of the  mutants. Oocytes expressing wild type · complexes produced about three or four times more functional Na⁺,K⁺ pumps at the cell surface than oocytes expressing  subunits alone (Fig. 5, A and B). On the other hand, a similar number of Na⁺,K⁺ pumps to -expressing oocytes was found in oocytes co-expressing  subunits and the CC2 or the CC3 mutant. In agreement with the biochemical data shown in Fig. 4, the CC1 mutant, however, produced a small but significant increase in the number of functional pumps at the cell surface. This increase was only observed when the CC1 mutant was co-expressed with  subunits (Fig. 5).

Functional expression of Na⁺,K⁺ pump complexes containing  cysteine mutants.

In conclusion, despite the considerable perturbation of the structural integrity of the ectodomain of the  subunit by the abolition of only one disulfide bond, all the mutants are capable to associate transiently with  subunits. However, only the CC1 mutant affected in the most N-terminal disulfide bond is partially able to form stable and functional · complexes. This result indicates that the C-terminal domain containing the last two disulfide bonds is particularly important to ensure the acquisition of a conformation compatible with an efficient association of the subunit with the  subunit.

Effects of Abolition of Glycosylation Sites in the  Subunit of Na,K-ATPase on Its Initial Folding and Its Interaction with BiP

All  isoforms of the Na,K-ATPase as well as the  subunits of H,K-ATPases are subjected to N-glycosylation during their synthesis. To assess the role of the sugar moieties in the initial folding of  subunits, we prepared mutants in which one (N1, N2, or N3), two (N2, 3) or three (N1,2,3) of the potential glycosylation sites in the 1 subunit of Xenopus Na,K-ATPase were abolished by replacement of the consensus asparagine residues with glutamine (see Fig. 1). Expression of these mutants in Xenopus oocytes revealed that all three potential glycosylation sites are used. Single, double, and triple mutants exhibited a progressively decreasing molecular mass (Fig. 6). For so far undefined reasons, but perhaps due to differential trimming, the core-glycosylated, single mutants N1, N2, and N3, which each lacks one of the glycosylation sites, exhibited slightly different molecular masses (compare lanes 3, 5, and 7). However, after deglycosylation with Endo H, all mutants as well as the wild type  subunit showed the same molecular mass that was identical to that of the triple mutant lacking all glycosylation sites (compare lanes 2, 4, 6, 8, and 10, to lanes 11 and 12) or to that of the  subunit synthesized in a reticulocyte lysate (data not shown).

State of glycosylation of Xenopus 1 mutants affected in one, two, or three consensus N-glycosylation sites.

To assess the effect of glycosylation on the structural stability of the newly synthesized  subunit, we subjected the triple N1,2,3 mutant synthesized in oocytes to proteolysis in the presence or absence of Triton X-100. In contrast to the wild type  subunit (Fig. 7A, lanes 1-3), the triple glycosylation mutant was completely degraded by proteinase K in the presence of Triton X-100 (lanes 4-6) indicating that the sugar chains either influence the folding of the  subunit or directly protect it from proteolytic attack.

Sensitivity of -glycosylation mutants to degradation by proteinase K and cellular proteases and their interaction with BiP.

With the exception of the N1,2,3 mutant, all glycosylation mutants expressed without  subunits in the oocyte were degraded during a chase period of 3 days similar to the wild type  subunit (Fig. 7B). Interestingly, the nonglycosylated N1,2,3 mutant remained stable over the whole chase period (Fig. 7C). During its stable expression in the ER, the N1,2,3 mutant remained associated with BiP (Fig. 7D). This result provides further support for the previously raised hypothesis (8) that interaction with BiP might be important for the ER retention of partially misfolded or overexpressed Na,K-ATPase  subunits until they are recognized by a pre-Golgi degradation system. On the other hand, a comparison of the gel migration of nonreduced samples (Fig. 7E) reveals no increase in the formation of disulfide-bonded aggregates by nonglycosylated  subunits that could explain their high stability.

Association of  Glycosylation Mutants with  Subunits and Formation of Functional · Complexes

The ability of the glycosylation mutants to form stable · complexes was assessed by performing immunoprecipitations under nondenaturing conditions on microsomes of pulse/chase-labeled oocytes expressing  and  mutants. After the pulse period, all glycosylation mutants co-precipitated with an  antibody indicating that they were able to associate with the  subunit (Fig. 8A). The  mutants with one or two residual glycosylation sites were in their core-glycosylated (Endo H-sensitive) form (Fig. 8, A and B). After a chase of 3 days, the  mutants remained associated with the  subunit (Fig. 8C), and with the exception of the N1,2,3 mutant that is nonglycosylated, the other mutants with residual glycosylation were now mainly processed to the fully glycosylated (Endo H-resistant) form (Fig. 8, C and D) indicating that the · complexes had been transported through a distal Golgi compartment. In comparison to the residually glycosylated  mutants that stabilized the  subunit similarly to the wild type  subunit (Fig. 8C, compare lanes 2-5 to lane 1), the nonglycosylated  mutant only achieved about 50% (average of two experiments) stabilization of the subunit (Fig. 8C, compare lane 6 to lane 1). Thus, these data indicate that the acquisition of an assembly-competent conformation of the  subunit depends on the presence of the sugars. However, the presence of only one sugar chain is necessary and sufficient to ensure an efficient association with the  subunit.

Assembly of -glycosylation mutants with the  subunit of Na,K-ATPase.

The cell surface expression of · complexes containing the N1,2,3 mutant was estimated by [³H]ouabain binding on intact oocytes combined with pulse/chase experiments. Denaturing immunoprecipitations of the labeled, wild type  subunit expressed without the  subunit showed that a small fraction of the  subunit population became fully glycosylated during the chase (Fig. 9A, lanes 3 and 4) due to its association with endogenous, oocyte  subunits (15). The fraction of fully glycosylated  subunits was more important when the  subunit was expressed together with  subunits (lanes 7 and 8). In contrast to the wild type  subunit, the nonglycosylated  subunit showed, as expected, no change in its molecular mass after the chase when it was expressed alone (lanes 5 and 6) or together with  subunits (lanes 9 and 10). The N1,2,3 mutant co-precipitated with the subunit during the pulse period similarly to the wild type  subunit but, as shown above, it was not able to stabilize the  subunit to a similar extent during the chase period (Fig. 9B, compare lanes 5 and 6 to lanes 3 and 4).

Cell surface expression of Na⁺,K⁺ pump complexes containing nonglycosylated 1 mutants.

Expressed alone, the wild type  subunit and the N1,2,3 mutant increased the expression of the number of Na⁺,K⁺ pumps at the cell surface about 2-fold compared with noninjected oocytes due to association with the endogenous  subunit (Fig. 9C). When co-expressed with  subunits, the wild type  subunit allowed for about a 4-fold increase and the N1,2,3 mutant for about a 3.5-fold increase in the number of Na⁺,K⁺ pumps (Fig. 9C). A comparison between the stabilization of the  subunit by the wild type or mutant  subunit and the expression of Na⁺,K⁺ pumps at the cell surface consistently showed that the expression of ·N1,2,3 mutant complexes was more important than expected from the ability of the  mutant to stabilize the  subunit. While the expression of ·N1.2,3 mutant complexes, assessed by ouabain binding, was only about 20% less than that of ·wild type  complexes (Fig. 9C), the stabilization of ·N1,2,3 complexes was about 50% lower (average of two experiments) than that of ·wild type complexes (Fig. 9B, lanes 3-6). This discrepancy might be explained by a regulated expression of Na⁺,K⁺ pumps in Xenopus oocytes that only permits the expression of a limited number of pumps at the cell surface and that would establish an intracellular pool of pumps (20).

Another possibility to explain the differences between the expression and the stabilization of the ·N1,2,3 complexes might be that the mutant complexes have a different ouabain sensitivity than the ·wild type  complexes. This latter possibility incited us to study the effect of the sugars in the  subunit on several functional properties of the cell surface-expressed · complexes. The apparent K⁺ affinity (K1/2) of Bufo Na,K-ATPase, measured as the half-maximal K⁺ activation of the Na⁺,K⁺ pump-mediated ⁸⁶Rb⁺ flux, was about 300 µM in the absence of extracellular Na⁺ and 1 mM in the presence of extracellular Na⁺ for wild type as well as for mutant, nonglycosylated Na⁺,K⁺ pumps (Fig. 10, A and B) These values are similar to those previously measured by electrophysiological means (21). Furthermore, the affinity for ouabain of the Xenopus Na⁺,K⁺ pumps containing the nonglycosylated mutant  subunit was similar to that of the glycosylated wild type pumps. The inhibition constant (K_i) measured as the half-maximal concentration needed to inhibit the Na,K-ATPase-mediated ⁸⁶Rb⁺ uptake was close to 50 nM for both types of pumps (Fig. 10C). Finally, no significant differences could be observed between the wild type and the nonglycosylated Xenopus pumps in the K⁺ concentrations needed to compete with ouabain binding (Fig. 10D).

Functional properties of Na,K-ATPase · complexes containing nonglycosylated  subunits.

Thus, these data demonstrate that the presence of at least one sugar moiety is necessary for the correct folding of the 1 subunit of Na,K-ATPase permitting its efficient assembly with the  subunit. On the other hand, our functional tests did not permit us to reveal a role of the sugars in the functional properties of the mature enzyme expressed at the cell surface.

DISCUSSION

The present study reveals the importance of co-translational modifications for the initial folding of the  subunit of Na,K-ATPase. We show that disulfide bond formation and to a lesser extent core glycosylation during the synthesis of the  subunit are needed to render it competent for the assembly with the catalytic  subunit.

Six cysteine residues in the ectodomain of the  subunits of the Na,K-ATPase and the H,K-ATPase, the positions of which are highly conserved, form disulfide bonds in a sequential pattern (22-24). Cleavage of these disulfide bonds by reducing agents leads to structural and functional changes in purified enzyme preparations (25-28) suggesting that they are important to maintain the  subunit in a conformation that is compatible with a correct association with the  subunit. In this study, we show that prevention of the formation of only one of the three disulfide bonds in the  subunit of Na,K-ATPase indeed is sufficient to abolish its ability to correctly assemble with  subunits and as a consequence to impede the maturation and efficient expression of functional pumps.

Acquisition of disulfide bridges is catalyzed by protein disulfide isomerase in the ER, and it is thought to limit the number of folding states a protein can adopt during or after synthesis, to guide the folding toward a final state and ultimately to contribute to the stability of the correctly folded state (for reviews see Refs. 10 and 29). Thus, the observed inability of an efficient assembly of the  subunit of Na,K-ATPase or of oligomeric, viral proteins (for review see Ref. 9) in which cysteine residues involved in disulfide bond formation are changed by site-directed mutagenesis is probably due to destabilization and increased misfolding of the newly synthesized proteins. However, in the cysteine-deficient  subunit, misfolding is not accompanied, as in many other proteins, by the formation of intermolecular disulfide-bonded aggregates or by a more pronounced interaction with the molecular chaperone BiP (for review, see Ref. 10). Possibly, the lack of only one cysteine residue in  subunits is not enough for the formation of covalent aggregates or for a global misfolding but leads to local structural perturbations that impede the acquisition of a completely folded association domain. Such partial misfolding might explain why cysteine bond-deficient  subunits can transiently but not permanently associate with  subunits.

Our data indicate that the relative importance of the three disulfide bonds in the  subunit on the acquisition of an assembly-competent form is different. Indeed, while  mutants, affected in the formation of disulfide bonds Cys¹⁵⁹-Cys¹⁷⁵ (numbering refers to the 1 subunit of Xenopus laevis) or Cys²¹³-Cys²⁷⁶, completely lose their ability to assemble permanently with the  subunit, the  mutant affected in the formation of the most N-terminal disulfide bond Cys¹²⁶-Cys¹⁴⁹ permits the formation of a small number of functional · complexes at the cell surface. Thus, the domain comprised between the two most C-terminal disulfide bonds is able to adopt a structure that partially permits correct subunit interaction. Probably this domain in the  subunit is more compact as suggested by its lower sensitivity to reduction by -mercaptoethanol that first breaks the least stable disulfide bond Cys¹²⁶-Cys¹⁴⁹ in purified enzyme preparations (26). On the basis of these results, it has been speculated that the most N-terminal disulfide bond might function as a kind of hinge that allows establishment of close contact between the hydrophobic, membrane-anchored domain and the hydrophilic C-terminal part of the  subunit that contains the sugar chains.

Our data showing the structural importance of the two most C-terminal disulfide bonds in the  subunit are in agreement with similar studies performed with cysteine mutants of Torpedo  subunits. However, our results obtained with the  mutant lacking the first N-terminal disulfide bond differ from those of Noguchi et al. (30) who reported that Torpedo  subunits lacking this disulfide bond are able to assemble and stabilize  subunits but cannot form functional Na⁺,K⁺ pumps. Although we cannot completely rule out that species differences are responsible for the discrepancies in the results, it is more likely that they are due to differences in the experimental approach. Indeed, Noguchi et al. (30) performed continuous pulse labeling protocols to analyze - interaction. In contrast, in this study, we performed pulse/chase protocols that are the only means to distinguish between a short and a long term association of  subunits. Furthermore, it is likely that, in contrast to the highly sensitive ⁸⁶Rb⁺ uptake measurements used in our study, the Na,K-ATPase assay used by Noguchi et al. (30) was not sensitive enough to detect a small fraction of functional pumps formed with mutant  subunits and/or that sodium thiocyanate used during sample preparation led to dissociation and inactivation of the loosely assembled · mutant complexes.

As with most cell surface proteins, all  isoforms of Na,K- or H,K-ATPase are glycoproteins and possess 3-7 N-linked sugar chains (for review see Ref. 31). The functional role of the glyco-moieties is poorly understood. In this study, we confirm previous reports (32-34) that nonglycosylated  subunits are able to assemble with  subunits and permit the expression of functional Na⁺,K⁺ pumps at the cell surface. Nevertheless, our data show that in the  subunit of Na,K-ATPase, as in many other glycoproteins (for review see Ref. 11), inhibition of glycosylation using tunicamycin (35) or elimination of the consensus sequences for N-linked glycosylation by site-directed mutagenesis leads to folding problems of the newly synthesized protein. These are reflected by an increased sensitivity of the  subunit against proteolysis and a decreased efficiency in the assembly with  subunits. As reported for vesicular stomatitis virus G-glycoprotein or influenza hemagglutinin, the effects of elimination of carbohydrate sites in the  subunit is additive (for references, see Ref. 9). No single oligosaccharide is essential for the correct folding, but all sugars must be eliminated to significantly affect the assembly competence.

Addition of N-linked sugar chains occurs as soon as the nascent polypeptide emerges in the ER lumen and thus occurs, in most cases, before folding and assembly. The addition of the large, hydrophilic carbohydrate chains facilitates the ongoing folding process of the nascent chain by favoring the recruitment of discrete polypeptide domains to the surface of the protein (for review and references see Ref. 36). In addition to local effects on folding, it has been proposed that core sugars have a global effect on the solubility of newly synthesized proteins that counteracts the tendency to form irreversible aggregates (37).

Although the degradation rates of wild type and sugar-deficient glycoproteins have not rigorously been compared, proteins in which the sugar moieties have been eliminated can be degraded in the ER (38). The nonglycosylated  subunit, however, is very resistant to ER degradation. The high stability of the nonglycosylated mutant compared with that of the wild type, glycosylated  subunit is surprising, since the presence rather than the absence of sugars would be expected to protect from proteolytic attack. In contrast to the prediction mentioned above, the nonglycosylated  subunit does not form disulfide-bonded aggregates that could explain its resistance to cellular degradation. On the other hand, it is permanently associated with BiP, perhaps in aggregates formed by hydrophobic interactions that could impede the recognition of the nonglycosylated  subunit by the pre-Golgi degradation pathway. Alternatively, the  subunit of Na,K-ATPase might be an example of a protein whose quality control and degradation depends on glycosylation. A similar hypothesis has recently been raised for carboxypeptidase Y whose unglycosylated species is stably retained in the ER (39). An important player in the quality control of glycoproteins is calnexin that binds to newly synthesized monoglucosylated N-linked oligosaccharides. A role for glycoprotein trimming and/or calnexin interaction is suggested by a recent study that reports that a mutant, secretory ₁-antitrypsin Z is degraded by the proteasome by first digesting the cytoplasmic tail of calnexin molecules that are associated with ₁-antitrypsin Z (40).

In contrast to disulfide bonds, complex-type sugar chains are often not important for overall stability and function of mature glycoproteins (for references see Ref. 11). In agreement, enzymatic deglycosylation of the  subunit in purified Na,K-ATPase has little effect on the enzyme activity or ouabain binding (41). Our studies using  subunits in which the glycosylation sites had been abolished by site-directed mutagenesis also do not reveal significant effects of the sugars in the 1 subunit on either the K⁺ activation or the ouabain binding of functional Na⁺,K⁺ pumps expressed at the cell surface. The question remains open whether complex-type oligosaccharides in the 1 subunit of Na,K-ATPase might represent specific recognition sites for interaction with external ligands, as suggested for many glycoproteins (42).