Subunit isoform selectivity in assembly of Na,K-ATPase α-β heterodimers

Densitometric quantification of the results presented in A was performed by dividing the density of the α 1 subunit band by the density of the corresponding PM YFP-β band. C-D. Proteins were extracted from MDCK cells expressing either YFP-β 1 (C) or YFP-β 2 (D) using extraction buffer containing 1% CHAPS and subject to successive rounds of immunoprecipitation using anti-α antibody followed by a final round of immunoprecipitation using anti-GFP/YFP antibody. Both extracted and immunoprecipitated proteins were treated with Endo H prior to SDS-PAGE. The majority of YFP-β subunits co-immunoprecipitated with α 1 were complex-type glycosylated. Almost all of the β subunits not assembled with α 1 were Endo H sensitive and immunoprecipitated in the final anti-GFP/YFP immunoprecipitation. PM –plasma membrane, Super2 – supernatant after the second round of immunoprecipitation.

2 muscle and plays a key role in the modulation of blood pressure in response to stress (3,14,15). In vivo human mutations in the α 2 and α 3 isoforms are associated with neurological diseases, familial hemiplegic migraine type 2 and rapidonset dystonia-parkinsonism (16). The α 4 isoform is required for sperm motility and fertility (10,11). The β 1 subunit plays an important role in intercellular adhesion in epithelia (17,18), and the β 2 subunit, or AMOG (adhesion molecule on glia), is important for adhesion and migration of neurons on glia (19). Decreased expression of the β 1 subunit is associated with cancer (reviewed in (20)), whereas abnormalities in expression and distribution of the β 2 subunit are linked to glioma and epilepsy (21)(22)(23)(24).
Therefore, it is clear that both α and β isoforms of the Na,K-ATPase have organ-and tissue-specific functions. However, very little is known about particular α-β heterodimers responsible for these roles. Transfection studies indicate that each of the four α subunit isoforms can assemble with each of the three β subunit isoforms and form a functional pump (6,7,25). These data imply that in cells co-expressing multiple Na,K-ATPase subunits isoforms, various α and β isoforms also assemble in different combinations, dependent on their relative cellular content. However, selective coimmunoprecipitation of the α 2 subunit, but not of the ubiquitously expressed α 1 subunit, with the β 2 subunit from mouse and rat brain (19, 26), as well as from heart and adrenal medullary cells of guinea-pigs and rats (26,27) suggest that the α 2 subunit is the preferred binding partner of the β 2 subunit. In support of this hypothesis, the tissue expression pattern of the β 2 subunit, mainly in muscle and nervous system, is similar to that of the α 2 subunit (3,7).
Here we show that not only does the β 2 subunit preferentially assemble with the α 2 subunit, but also the α 2 subunit is mostly associated with the β 2 subunit in mouse brain. In addition, by analyzing the competition of the β 1 or β 2 subunits for binding to the α 1 subunit in MDCK cells, we demonstrate that the β 1 subunit is a greatly preferred binding partner of the α 1 subunit compared to the β 2 subunit. The results of co-immunoprecipitation of α and β subunits from various detergent extracts of native tissues and cultured cells indicate that α 1 -β 1 and α 2 -β 2 heterodimers are more stable than α 1 -β 2 heterodimers. Therefore, there is selective assembly of the different α and β subunit isoforms with likely tissue-specific functional consequences.

EXPERIMENTAL PROCEDURES
Cell lines-The Na,K-ATPase dog β 1 or human β 2 subunits linked with their N-termini to YFP were constructed as described previously (28,29). Stable MDCK cell lines expressing YFP-β 1 ** and YFP-β 2 were obtained and maintained as described previously (30). Confocal microscopy-Confocal microscopy images were acquired using the Zeiss LSM 510 laser scanning confocal microscope and LSM 510 software, version 3.2. Primary antibodies used for immunofluorescent staining and Western blot analysis-For immunofluorescent staining, the monoclonal antibodies against the Na,K-ATPase α 1 subunit, clone C464.6 (Millipore) and against the Na,K-ATPase β 1 subunit, clone M17-P5-F11 (Affinity Bioreagents) and polyclonal antibodies against the Na,K-ATPase α 2 subunit, (Millipore) and against the Na,K-ATPase β 2 subunit (Millipore) were used. The polyclonal antibody against the Na,K-ATPase β 1 subunit (31), which was a generous gift of Dr. W. James Ball Jr. (University of Cincinnati), was used for Western blot analysis. Also, the following monoclonal antibodies were used for Western blot analysis: against GFP, clones 7.1 and 13.1, which also recognizes YFP (Roche Diagnostics), against the Na,K-ATPase α 1 subunit, clone C464.6 (Millipore), against the Na,K-ATPase α 3 subunit (Upstate), against the Na,K-ATPase β 2 subunit, clone 35 (BD Bioscience Pharmingen), and against the Na,K-ATPase β 3 subunit (Santa Cruz). Extraction of proteins from MDCK cells and mouse brain homogenates-Confluent MDCK cell monolayers grown in 6-well plates were rinsed twice with ice cold PBS and incubated with 200 l/well of the extraction buffer at 4ºC for 30 min followed by scraping cells. Mouse brain homogenates containing 600 g protein in 500 l of 150 mM NaCl in 50 mM Tris pH 7.5 were incubated with 500 l of the extraction buffer at 4ºC for 30 min. The extraction buffer contained 150 mM NaCl in 50 mM Tris pH 7.5 and the 2x concentration of the indicated detergent(s). When DOC was used alone, no NaCl was added to the extraction buffer. Prior to using, the extraction buffer was mixed with Complete Protease Inhibitor Cocktail (Roche Diagnostics), 1 tablet/50 ml. Cell extracts were clarified by centrifugation (15,000 g, 10 min) at 4ºC. Where indicated, protein extracts were treated by PNGase F from Flavobacterium meningosepticum (New England BioLabs) or by Endo H from Streptomyces plicatus (Glyco-Prozyme Inc.) according to manufacturers' instructions prior to loading on SDS-PAGE. Immunoprecipitation-Protein extracts from MDCK cells or from mouse brain homogenates (100-300 µg protein) were incubated with 30 µl of the protein A-agarose suspension (Roche Diagnostics) in a total volume 1 ml of the extraction buffer at 4°C with continuous rotation for at least 3 hours (or overnight) to remove the components that non-specifically bind to protein A. The pre-cleared cell extract was mixed with 2 µl of polyclonal antibodies against GFP, which also recognize YFP (Clontech), or 10 µl of polyclonal antibodies against the Na,K-ATPase α 2 subunit (CHEMICON International), or 10 µl of polyclonal antibodies against the Na,K-ATPase α subunit (32) and incubated with continuous rotation at 4° C for 60 min. After addition of 30 µl of the Protein A-agarose suspension, the mixture was incubated at 4° C with continuous rotation overnight. The beadadherent complexes were washed 3 times on the beads and then eluted as described previously (33).
Multi-round immunoprecipitation from MDCK cells expressing either YFP-β 1 or YFPβ 2 was performed as described above with the exception that, after the first round of immunoprecipitation using 30 µg protein in 1% CHAPS and 5 µl anti-α antibody, the unbound proteins in the supernatant were collected and incubated again with 5 µl anti-α antibody in a second round of immunoprecipitation. A third round of immunoprecipitation was performed by using 4 µl anti GFP/YFP antibody from the supernatant after the second round of immunoprecipitation.
Where indicated, the bead-adherent proteins were treated with PNGase F or with Endo H. Deglycosylation by PNGase F was performed by incubation of the bead-adherent proteins with 1 µl PNGase F in 30 μl of 50 mM sodium phosphate, pH 7.5 containing 1% NP-40 and at 37°C for 1 hour. Digestion by Endo H was performed by incubation of the bead-adherent proteins with 3 μl of Endo H in 30 μl of 50 mM sodium citrate/phosphate, pH 5.5 containing 1% NP-40 at 37°C for 3 hours. After incubation with glycosidases, the reaction mixture was separated from the beads. The adherent proteins were eluted from the beads by incubation in 30 μl of 2x SDS-PAGE sample buffer for 5 min at 80°C. To account for possible dissociation of immunoprecipitated proteins from the beads during deglycosylation, the eluted proteins were combined with the reaction mixture. After separation by SDS-PAGE, the immunoprecipitated and co-immunoprecipitated proteins were analyzed by Western blot by using appropriate antibodies. Isolation of basolateral plasma membrane proteins of MDCK cells using surface-specific biotinylation-Cells were maintained for 6 days after becoming confluent in transwell inserts. Biotinylation and isolations of basolateral surface proteins was performed according to previously described procedures (34)(35)(36). Western blot analysis-1-10 g of proteins extracted from MDCK cells, microsomal membranes isolated from animal tissues in SDS-PAGE sample buffer, or 5-20 μl of proteins eluted from the Protein A-conjugated agarose beads were loaded onto 4-12% gradient SDS-PAGE gels (Invitrogen). Proteins were separated by SDS-PAGE, transferred onto a nitrocellulose membrane (BioRad) and detected by Western blot analysis as described previously (33). Immunoblots were quantified by densitometry using Zeiss LSM 510 software, version 3.2. Immunofluorescent staining-MDCK cells were fixed by incubation with 3.75% formaldehyde in PBS for 15 min at 37ºC and permeabilized by incubation with 0.1 % Triton X-100 for 5 min. Fixed cells or frozen tissue sections on FDA Standard frozen tissue rat or human arrays (BioChain) were incubated with Dako Protein Block Serum-Free solution (Dako Corporation) for 30 min. Immunostaining was performed by 1hr-incubation with the monoclonal antibodies followed by 1hr-incubation with Alexa Fluor 633 or Alexa Fluor 488 conjugated anti-mouse or anti-rabbit antibodies (Invitrogen). Statistical analysis-was performed using Student's t-test (GraphPad Prism 4 software and Microsoft Excel). Statistical significance and number of experiments are specified in the figure legends.

RESULTS
Distribution of the Na,K-ATPase α 1 , α 2 , β 1 and β 2 subunits in rat and human tissues. Normal rat or human frozen tissue arrays (BioChain) that contained sections of adrenal gland, brain, breast, colon, esophagus, heart, kidney, liver, skeletal muscle, nerve and ovary were used to perform immunofluorescent double staining of the Na,K-ATPase α 1 and α 2 subunits. The Na,K-ATPase α 1 and α 2 subunits were differentially distributed in several organs, particularly in peripheral nerves and kidney, which are shown as examples ( Fig. 1 and Fig.  2A). The α 2 subunit was found predominantly in perineurium that surrounds fascicles, while the α 1 subunit was also present in epineurium that surrounds nerves (Fig. 1). In the kidney, the α 1 subunit was detected in tubules, while the α 2 subunit was seen predominantly in renal arteriole and glomerulus ( Fig. 2A). Double staining of the α 1 and β 2 subunits showed that tissue distribution of the β 2 subunit is similar to that of the α 2 , but not of the α 1 subunit in both peripheral nerve and kidney sections ( Fig. 1 and Fig. 2A). Also, double staining of the β 1 and β 2 subunits showed that distribution of the β 1 subunit in the nerve is similar to that of the α 1 , but not of the β 2 subunit (Fig. 1B). Staining of the α 1 , α 2 and β 2 subunits in rat brain cortex showed both co-expression of these isoforms in the same cells and cell-specific expression (Fig.  2B). Co-expression of α 1 and α 2 or α 1 and β 2 is evident from co-localization of red and green fluorescence, producing a yellow color on the merged images, while the differential expression of these isoforms is seen from separated green and red spots on the merged images. Coexpression of the β 1 isoform with both α 2 and β 2 isoforms in the same cells was also seen in human brain sections (not shown).
Subunit composition of the Na,K-ATPase αβ heterodimers in mouse brain. Western blot analysis detected the presence of six different isoforms of the Na,K-ATPase subunits in mouse brain extracts (Fig. 3), consistent with previously published results for rat brain (37). To validate the antibodies used to detect β subunits isoforms, we performed Western blot analysis of mouse brain extracts incubated with or without PNGase F that cleaves N-glycans attached to the β subunits. The bands at 33-35 kDa, which were expected for deglycosylated β 1 , β 2 and β 3 subunits were detected in samples treated with PNGase F by using the antibodies listed in Experimental Procedures (Fig. 3A). In control samples, these antibodies detected bands at 45-50 kDa (Fig. 3A).
Immunoprecipitation of the β 2 subunit from 1% NP40/0.5% DOC extracts of mouse brain homogenates resulted in co-precipitation of the α 2 isoform and of a minor amount of the α 3 isoform, but not of the α 1 subunit (Fig. 3B). These results are consistent with detection of the α 2 subunit, but not of the α 1 subunit, in the fraction isolated from mouse brain by immunoaffinity chromatography using the antibody against the β 2 subunit (38).
Immunoprecipitation of the Na,K-ATPase α 2 subunit resulted in co-precipitation of the β 2 subunit, but not of the β 1 subunit (Fig. 3C). A minor amount of the β 3 subunit was detected in the immunoprecipitated fraction (Fig. 3C). No α 1 or α 3 subunits were detected in the immunoprecipitated fraction, confirming specificity of the anti-α 2 antibody. In contrast, the antibody raised against the 4/5 loop of the Na,K-ATPase α 1 subunit that is relatively homologous in other α isoforms (32) immunoprecipitated all three α isoforms, α 1 , α 2 and α 3 and all three β subunit isoforms, from mouse brain extracts (Fig. 3C) confirming its non-selectivity, as expected To evaluate stability of the various Na,K-ATPase α-β complexes, we compared coimmunoprecipitates of the β subunits with α subunits from mouse brain extracts obtained by using increasing concentrations of the non-ionic detergent, DDM. The amount of the α 2 subunit immunoprecipitated with specific anti-α 2 antibodies was similar in 0.5%, 1% and 2% DDM (Fig. 4A, left panel). In contrast, the amount of the β 2 subunits, which were coprecipitated with the α 2 subunit, gradually decreased with increasing detergent concentration (Fig. 4A, left panel and Fig. 4B), showing a partial dissociation of the α 2 -β 2 complex by DDM. No co-precipitation of the β 1 or β 3 subunit was found at any DDM concentration tested.
The total amount of all three α subunit isoforms immunoprecipitated by non-specific anti-α antibodies was similar in 0.5%, 1% and 2% DDM extracts (Fig. 4A, right panel). The amount of the β 2 subunit co-precipitated by nonspecific anti-α antibodies gradually decreased with increasing detergent concentration (  Fig. 4B), strongly suggesting that precipitation of the β 2 subunit by non-specific anti-α antibodies reflects α 2 -β 2 complexes. This conclusion is consistent with co-precipitation of the α 2 , but not of the α 1 subunit, with the β 2 subunit (Fig. 3C). The amount of the β 1 subunit that was co-precipitated with α subunits by the non-specific anti-α antibodies was not affected by detergent concentration (Fig. 4A, right panel). Only a slight decrease in the amount of the β 3 subunit co-precipitated with α subunits was observed with an increase in DDM concentration (Fig.  4A, right panel and Fig. 4B). These results indicate that complexes of the α 1 or α 3 subunit with either β 1 or β 3 subunit are preserved at all tested DDM concentrations. Therefore, both α 2 and β 2 subunits predominantly associate with each other, but not with other partner subunits in mouse brain.
Plasma membrane and intracellular distribution of the Na,K-ATPase α 1 , β 1 and β 2 in MDCK cells. To test whether the α 1 subunit preferentially assembles with the β 1 or β 2 isoform, the content of α 1 -assembled and α 1unbound β subunits was compared in stable MDCK cell lines expressing either YFP-β 1 or YFP-β 2 . The major endogenous Na,K-ATPase isoforms in these cells are α 1 and β 1 . Both YFPβ 1 and YFP-β 2 were co-localized with the endogenous α 1 subunit in the lateral membranes, as detected by immunofluorescence (Fig. 5A), consistent with previously reported results (29). In addition, both YFP-β 1 and YFP-β 2 , but not the α 1 subunit, were found inside the cells (Fig. 5A), indicating these intracellular forms of YFP-β 1 and YFP-β 2 are not assembled with the α 1 subunit. This intracellular retention was greater in dispersed cells than in confluent monolayers (Fig. 5B), and it showed co-localization with the ER marker (Fig. 5C). The fraction of YFP-β 2 localized in the ER was greater than that of YFP-β 1 (Fig. 5A-B).
To quantify the levels of the endogenous and exogenous Na,K-ATPase subunits in the ER and the plasma membrane in the two cell lines, we analyzed total cell lysates and plasma membrane fractions by Western blot analysis. In both cell lines, the YFP-linked β subunit was detected as two bands in total lysates (Fig. 6A). These two bands represent differentially Nglycosylated species of fusion proteins. Only a single band, which corresponded to the upper band in cell lysate, was found in the basolateral membrane (Fig. 6A), indicating that the lower band in cell lysate corresponds to the intracellular fraction of YFP-β 1 or YFP-β 2 . As shown previously, this intracellular form of either YFP-β 1 or YFP-β 2 mostly represents the ER-resident fraction of each fusion protein (34). The relative amount of this ER-resident fraction appears to be greater for YFP-β 2 than for YFP-β 1 (Fig. 6A, left lanes). However, direct densitometric quantification of the two fractions of YFP-β 2 was not possible because of a significant overlap between two bands on SDS-PAGE (Fig. 6A, top panel, left lane).
To separate these two bands, we used Endo H, which is known to remove high-mannoseand hybrid-, but not the complex-type Nglycans, from glycoproteins. Treatment of total cell lysates or biotinylated proteins with EndoH resulted in a slight increase in electrophoretic mobility of the plasma membrane fraction of YFP-β 2 , but not of YFP-β 1 (Fig. 6A). The ERresident form in cell lysates was completely deglycosylated by Endo H, producing a band at ~60 kDa that corresponds to the protein core molecular mass of YFP-β (Fig. 6A). As a result, a better separation of the plasma membrane and ER fractions on SDS-PAGE was observed in Endo H treated cell lysates, allowing densitometric quantification (Fig. 6B).
This quantification confirmed the greater ER retention of YFP-β 2 (49% of total cellular content) as compared to YFP-β 1 (31% of total cellular content) in mature cell monolayers. The difference in the relative content of the ER form of YFP-β 2 and YFP-β 1 was more prominent in immature cell monolayers, 80% and 38% of total cellular content, respectively (Fig. 6A-B). In contrast, the levels of YFP-β 1 and YFP-β 2 in the basolateral membrane were similar in the two transfected cell lines (Fig. 6C). Also, the levels of the endogenous α 1 and β 1 subunits were similar in YFP-β 1 -and YFP-β 2 -expressing MDCK cell lines (Fig. 6C).
Therefore, the amount of α 1 -unassembled YFP-β 2 in the ER is significantly greater than that of YFP-β 1 , while the quantities of α 1assembled YFP-β 1 and YFP-β 2 in the plasma membrane are similar ( Fig. 5 and 8), suggesting preferential assembly of the α 1 subunit with the β 1 isoform rather than with the β 2 isoform.
Alignment of the β 1 subunit with the β 2 subunit shows that ten α 1 -interacting residues in the β 1 subunit predicted by the high resolution structure of the α 1 -β 1 Na,K-ATPase (39) differ from the corresponding residues in the β 2 subunit (Fig. 7B). These differences are expected to weaken the α-β interaction. For example, the presence of an asparagine residue in the β 2 subunit instead of Arg86 would not allow effective interaction with Glu122, Glu124 and Trp894 of the α 1 subunit (Fig. 7B), consistent with the lower stability of α 1 -β 2 complexes as compared to that of the α 1 -β 1 complexes.
The α 1 subunits were predominantly located in the plasma membrane, whereas both YFP-β 1 and YFP-β 2 were also found in the ER (Fig. 5A). Therefore, the amount of co-precipitated α 1 subunits should be compared to the amount of the plasma membrane fraction of YFP-β proteins, but not to the total amount of immunoprecipitated YFP-β. To better separate the plasma membrane and ER-resident fractions, the immunoprecipitated proteins were treated with Endo H prior to SDS-PAGE (Fig. 8A). Western blot analysis followed by densitometric quantification, which allowed calculation of the amount of co-precipitated α 1 subunits relative to the amount of the plasma membrane fractions of immunoprecipitated YFP-β 1 or YFP-β 2 . This quantification showed that the amount of the YFP-β 2 -bound α 1 subunits was 15 % less than the amount of YFP-β 1 -bound α 1 subunits (Fig.  8B). Similar results were obtained in 1% CHAPS (not shown).
If the lower amount of YFP-β 2 -bound α 1 subunit was related to partial disruption of the α 1 -β 2 complex by detergent, the detergent extract must contain the α 1 -unbound plasma membrane forms of YFP-β 2 subunits. To determine the presence of α 1 -unbound β subunits in 1 % CHAPS extracts of both YFP-β 1 -and YFP-β 2expressing MDCK cells, multi-round immunoprecipitation using anti-α antibody was performed.
The first round of immunoprecipitation pulled down a vast majority of the α 1 subunits. Co-precipitated fractions of YFP-β 1 or YFP-β 2 contained predominantly mature forms and minor amounts of the ER-resident immature forms (Fig. 8C-D). The second round of immunoprecipitation using anti-α antibody precipitated the rest of the α 1 subunits and mature YFP-β 1 or YFP-β 2 . The third round of immunoprecipitation using anti-GFP antibody pulled down almost exclusively the immature forms of YFP-β 1 or YFP-β 2 and no α 1 subunits (Fig. 8C-D). These results confirm that both YFP-β 1 and YFP-β 2 are assembled with the α 1 subunits in the plasma membrane, while the majority of the ER-resident YFP-β 1 and YFP-β 2 are not assembled with the α 1 subunits (Fig. 8C-D). In addition, the data demonstrate that the (YFP-β 2 )-α 1 complex is fully preserved in 1% CHAPS extracts.

DISCUSSION
The β 1 , but not the β 2 isoform, is a preferred binding partner of the α 1 subunit. YFP-β 1 and YFP-β 2 stably expressed in MDCK cells are predominantly distributed between the basolateral plasma membrane and ER (Fig. 5). All the β subunits present in the basolateral membrane are α 1 -assembled, as we showed previously for the endogenous β 1 subunits (40) and now for both exogenous YFP-β 1 and YFP-β 2 (Fig. 8). On the other hand, the majority of the ER-resident β subunits are not bound to α 1 subunits (34,40), (Fig. 5B and Fig. 8C-D). These results are consistent with the previous finding showing that both YFP-β 1 and YFP-β 2 compete with the endogenous β 1 subunits for binding to the limited amount of the endogenous α 1 subunits (34). Only the α 1 -assembled β subunits exit the ER, while the unassembled subunits are retained in the ER and rapidly degraded (34). As a result of this competition, a fraction of α 1 -β 1 heterodimers exported from the ER is replaced by α 1 -(YFP-β) heterodimers, explaining the decrease in the amount of the mature endogenous β 1 subunit in the basolateral membrane in both YFP-β 1 -and YFP-β 2expressing cell lines as compared to nontransfected MDCK cells (Fig. 6C). YFP-β 1 -and YFP-β 2 -expressing cell lines have similar quantities of exogenous and endogenous Na,K-ATPase subunits in the plasma membrane (Fig.  6C), whereas the abundance of the ER-located α 1 -unbound YFP-β 2 is greater than that of YFP-β 1 (Fig. 5 and Fig. 6A-B). Since the ER retention of YFP-β 1 or YFP-β 2 is due to their competition with the same number of endogenous β 1 subunits for binding to the α 1 subunit, these results imply that the β 2 subunit has lower affinity for the α 1 subunit than does the β 1 subunit.
This interpretation, however, is complicated by the fact that α 1 -unassembled β subunits and β-unassembled α 1 subunits are not freely floating in the ER. Instead, the orphan α and β subunits are bound to ER chaperones (33,41), which facilitate normal folding of the subunits and possibly the assembly process per se. The β 2 , but not the β 1 , subunit persistently binds calnexin in the ER, suggesting that it undergoes repeated calnexin-assisted folding prior to its assembly with the α subunit (29,33). Persistent calnexin binding to glycoproteins is dependent on repeated cycles of de-and re-glucosylation of glycoprotein N-glycans by the ER glucosidase and UGGT1, respectively (42). It is possible that the β 2 subunit is not completely folded by the time when calnexin is dissociated from deglucosylated β 2 subunit and thus is recognized by the folding sensing enzyme, UGGT1 (43). UGGT1 re-glucosylates the β 2 subunit Nglycans, which results in repeated calnexin binding. However, it cannot be excluded that calnexin-free β 2 subunits bind UGGT1 not because they are misfolded, but because they fail to assemble with the α 1 subunits due to their lower α 1 -binding affinity as compared to that of the β 1 subunits. The UGGT1-mediated reglucosylation of the β 2 subunit would then induce its re-binding to calnexin. Therefore, both longer association with calnexin and greater accumulation in the ER of the β 2 subunit as compared to those of the β 1 subunit could result from either the longer time required for folding of the β 2 subunit or its lower affinity to the α 1 subunit. The lower binding affinity of the β 2 subunit toward the α 1 subunit is anticipated from the differences between the α 1 -interacting residues of the β 1 subunit and the corresponding residues of the β 2 subunit (Fig. 7B).
The α 1 -β 2 heterodimers are less resistant to the disruptive effect of various detergents than the α 1 -β 1 complexes. Co-immunoprecipitation of the Na,K-ATPase α 1 subunit with YFP-β 2 was observed only in selected detergents, and the amount of co-precipitated α 1 subunit varied in these detergents (Fig. 7), showing that the α 1 -β 2 complex is partially or completely disrupted by the majority of tested detergents. On the other hand, the amount of the Na,K-ATPase α 1 subunit that co-precipitated with YFP-β 1 is the same in all tested detergents, indicating that the α 1 -β 1 complex is fully preserved in these detergents (Fig. 7). These results are in agreement with previously reported disruption of α 1 -β 2 , but not of the α 1 -β 1 or α 2 -β 2 complexes formed in Xenopus oocytes by Triton X-100 (44,45).
Selective formation of the α 2 -β 2 Na,K-ATPase in mouse brain. The β 2 subunit of the Na,K-ATPase was first discovered as AMOG (19). Immunoaffinity purification of AMOG from mouse brain by using AMOG-specific antibody resulted in co-purification of a 100 kDa protein that later was identified as the Na,K-ATPase α 2 subunit (and possibly α 3 subunit), but not the α 1 subunit (38). Consistent with these data, we found that immunoprecipitation of the β 2 subunit from mouse brain resulted in coimmunoprecipitation of the α 2 , but not of the α 1 subunit (Fig.  3B). Conversely, immunoprecipitation of the α 2 subunit selectively co-precipitated the β 2 subunit (Fig.  3C and Fig. 4). Six isoforms of the Na,K-ATPase are expressed in the brain (9,37). Both α 2 and β 2 subunits are predominantly expressed in glial cells (37), so the formation of the α 2 -β 2 complexes is, in part, due to the cell-specific coexpression of the two isoforms. However, both α 2 and β 2 subunits are also found in subsets of neurons (37). Similarly, the α 1 and β 1 isoforms are expressed in both glial cells and neurons. With the exception of the neuron-specific α 3 subunit, other isoforms are expressed in both neurons and glial cells (37,(46)(47)(48). Even though there are cell-and region-specific differences in expression of various isoforms, many cell types in brain contain multiple Na,K-ATPase subunit isoforms (9,37,46). Accordingly, co-expression of the α 1 , α 2 and β 2 subunits in the same cells in rat brain cortex is detected here by immunofluorescence (Fig. 2B). Therefore, the preferential formation of the α 2 -β 2 complexes in the brain is determined not only by cell-specific co-expression of these isoforms, but also by their binding preferences. Preferential formation of the α 2 -β 2 was also detected in heart and adrenal medullary cells, where the α 1 subunit is more abundant than the α 2 subunit (26,27), emphasizing preferential β 2 subunit binding to the α 2 subunit.
Interestingly, the α 2 -β 2 complexes are less stable than complexes of the α 1 or α 3 subunit with either β 1 or β 3 subunit (Fig. 4). Recent studies have demonstrated that α 2 -β 1 complexes are less stable to heat and detergents than α 1 -β 1 or α 3 -β 1 complexes perhaps due to weaker interactions of the α 2 subunit with phosphatidylserine, which stabilizes the protein (49). Thus, it is possible that detergent-mediated disruption of α 2 -β 2 complexes (Fig. 4) is the result of displacement of selectively bound phosphatidylserine.
Preferential assembly of α 2 and β 2 isoforms in the brain may have several implications. The β subunits are known to modify the kinetic properties of the α isoforms. The α 2 -β 2 heterodimer has the lowest K + affinity among nine different α-β heterodimers formed by each of the three α subunit isoforms (α 1 -α 3 ) and each of the three β subunit isoforms (β 1 -β 3 ) (25). Thus, assembly of the α 2 subunit preferentially with the β 2 isoform may be crucial to restore external K + homeostasis after a series of action potentials in the nervous system, since the α 2 -β 2 heterodimer would respond to an increase in external K + because of its low K + affinity (25). In astrocytes, the Na,K-ATPase α 2 subunits form complexes with different glutamate transporters, and glutamate inward transport is inhibited by ouabain, suggesting a specific role of the Na,K-ATPase α 2 subunit in reuptake of glutamate from the synaptic cleft (50). Since the activity of the least K + -sensitive α 2 -β 2 isoform would increase more at the elevated external K + concentration, assembly with the β 2 subunit may be important for the specific role of the Na,K-ATPase α 2 subunit in glutamate clearance.
Considerable evidence exists for the presence of endogenous ouabain-like molecules in mammalian tissues that may serve to regulate Na,K-ATPase activity (3,14,51). Particularly, a signaling role of the Na + /K + -ATPase has been demonstrated in regulating synaptic plasticity and dendritic growth in cortical neurons (52). The human α 1 , α 2 and α 3 isoforms have similar ouabain affinities (25). However, the lowest K + affinity of the α 2 -β 2 isoform implies that this heterodimer has the lowest K + /ouabain antagonism as compared to other α-β heterodimer isoforms (25). So, at physiological K + concentrations, ouabain and endogenous ouabain-like compounds may predominantly bind to the α 2 -β 2 isoform and to a lesser extent to other complexes and thus specifically regulate the α 2 -dependent signaling pathways.
Natural in vivo mutations in the α 2 subunit are associated with familial hemiplegic migraine and epilepsy (16). Most of these mutations cause functional defects in active Na + and K + transport and impaired clearance of extracellular K + or glutamate due to either the impairment of maturation and hence plasma membrane delivery of the enzyme, or the loss of the catalytic activity (16,53). It is known that neurological diseases, particularly epilepsy, are closely associated with the ER stress-related retention of essential ion transporters in the ER (54)(55)(56)(57). We showed recently that the β 2 isoform is much more sensitive to the ER stress than the β 1 isoform (33). Since the β subunit is essential for maturation of the Na,K-ATPase α-β heterodimers (58) and the α 2 selectively forms a complex with the β 2 isoform in the brain, it is possible that stress-induced impairment of the β 2 subunit folding in the ER increases the ER retention of the α 2 subunit, which decreases the Na,K-ATPase ion transport activity of the α 2 β 2 Na,K-ATPase and thus contributes to epilepsy. Consistent with this hypothesis, abnormalities in distribution of the β 2 subunit are linked to epilepsy (21,22). Therefore, the selectivity of α-β assembly, which is determined both by cell-specific expression and by isoform-specific binding preferences of α and β subunits, is crucial for cell-and tissue-specific functions of the Na,K-ATPase. Fig. 1. Localization of the Na,K-ATPase β 2 subunit in rat sciatic nerve and kidney sections is similar to that of the Na,K-ATPase α 2 subunit, but not of the Na,K-ATPase α 1 subunit or β 1 subunit. A, Frozen sections of rat sciatic nerve were double stained by using mouse antibodies against α 1 subunit (green) and either rabbit antibodies against α 2 subunit (red) or rabbit antibodies against β 2 subunit (red). B, Frozen sections of human trigeminal nerve were double stained by using by using either mouse antibodies against α 1 subunit (green) or mouse antibodies against β 1 subunit (green) and rabbit antibodies against β 2 subunit (red). Anti-mouse Alexa Fluor 488 conjugated secondary antibodies were used to detect anti-α 1 and anti-β 1 primary antibodies and anti-rabbit Alexa Fluor 633 conjugated secondary antibodies were used to detect anti-β 2 and anti-α 2 primary antibodies. Fig. 2. Localization of the Na,K-ATPase α 1 , α 2 and β 2 subunits in rat kidney and brain sections. Frozen sections of rat kidney (A) and rat brain (B) were double stained by using mouse antibodies against α 1 subunit (green) and either rabbit antibodies against α 2 subunit (red) or rabbit antibodies against β 2 subunit (red). Anti-mouse Alexa Fluor 488 conjugated secondary antibodies were used to detect anti-α 1 primary antibodies and anti-rabbit Alexa Fluor 633 conjugated secondary antibodies were used to detect anti-β 2 and anti-α 2 primary antibodies. Insets in the right panels (B) show 5-fold zoomed images Fig. 3. The Na,K-ATPase α 2 and β 2 subunits are selectively co-immunoprecipitated from mouse brain extracts. Proteins were extracted from mouse brain homogenate by using 1%NP40/0.5%DOC. A, The antibodies against the Na,K-ATPase β 1 , β 2 and β 3 isoforms were validated by Western blot analysis of mouse brain extracts pre-incubated with or without PNGase F that cleaves N-glycans from the β isoforms subunits and, hence, results in an increase in electrophoretic mobility of the subunits. B, Western blot analysis of the immunoprecipitated β 2 subunit and co-immunoprecipitated α subunit isoforms shows that the α 2 subunit is preferentially co-precipitated with the β 2 subunit. Input lanes contain 4% and 10% of the extract used for immunoprecipitation on α and β blots, respectively. C, Western blot analysis of proteins immunoprecipitated and co-immunoprecipitated by using either the α 2 -specific antibodies (left panels) or the α-non-specific antibodies (right panels) shows selective co-immunoprecipitation of the β 2 subunit with the α 2 subunit. Input lanes contain 10% of the extract used for immunoprecipitation. To prevent an overlap of the β subunit bands with the heavy chain band of the antibodies used for immunoprecipitation, the immunoprecipitated proteins were treated with PNGase F prior to SDS-PAGE. Inputs are HCh, heavy chain; IP, immunoprecipitation; WB -Western blot; DGdeglycosylated. Fig. 4. The Na,K-ATPase α 1 -β 2 complex is less stable than the Na,K-ATPase α-β 1 or α-β 3 complexes in detergent extracts obtained from mouse brain membranes. A, Various concentrations of n-dodecyl β-D-maltoside (DDM) were used to extract proteins from mouse brain homogenate. Western blot analysis of proteins immunoprecipitated and co-immunoprecipitated by using either the α 2 -specific antibodies (left panels) or the α-non-specific antibodies (right panels) shows a stepwise decrease in the amount of β 2 subunits co-immunoprecipitated by using both antibodies, but not in the amount of β 1 or β 3 subunits coimmunoprecipitated by using α-non-specific antibodies, with increasing detergent concentrations. B, Densitometric quantification of the results shown in A was performed by dividing the signal from the β antibody by the corresponding signal of the α antibody. A comparative graph shows these ratios as a percentage of the ratio obtained in 0.5% DDM. IP, immunoprecipitation; WB -Western blot; DGdeglycosylated by PNGase F prior to SDS-PAGE. Both YFP-β 1 and YFP-β 2 (green) are co-localized with the endogenous α 1 subunit (red) in the lateral membranes, but not inside the cells as detected by immunostaining of fixed cells using the monoclonal antibody against the Na,K-ATPase α 1 subunit (A-B). The intracellular retention of α 1 -unassembled YFP-β 2 is more prominent than that of YFP-β 1 and more evident in dispersed colonies than in confluent monolayers. This intracellular fraction of YFP-β 2 (green) shows co-localization with the ER (red) as detected by transient expression of the fluorescent ER marker, DsRed2-ER (C). N-nucleous; PMplasma membrane. Fig. 6. The greater intracellular retention of YFP-β 2 than YFP-β 1 is not associated with its higher level in the plasma membrane. A-B, The comparative Western blot analysis (A) of total cell lysates and basolateral biotinylated proteins (BL membrane) of MDCK cells stably expressing either YFP-β 1 or YFPβ 2 shows that the upper band found in either cell lysate represents the mature plasma membrane fraction (BLM), while the lower band corresponds to the intracellular fraction ER-resident fraction (ER). Treatment with endoglycosidase H (Endo H) resulted in a slight increase in electrophoretic mobility of BLM YFP-β 2 , but not of BLM YFP-β 1 , and a major increase in electrophoretic mobility of ER YFP-β 1 and ER YFP-β 2 . This allows a better separation of BLM and IC fractions of YFP-β 1 or YFP-β 2 on SDS-PAGE and their densitometric quantification (B). C, Western blot analysis of proteins isolated by basolateral surface-selective biotinylation show that stable expression of either YFP-β 1 or YFP-β 2 in MDCK cells resulted in a significant decrease in the amount of the endogenous Na,K-ATPase β 1 subunits in the basolateral membranes, but did not change the level of the α 1 subunits, as compared to nontransfected cells (NT cells). YFP-β 1 and YFP-β 2 are present in the basolateral membrane at similar levels in the two transfected cell lines. Nnucleus; PMbasolateral plasma membrane. Fig. 7. The Na,K-ATPase α 1 -β 1 complex is more stable than the Na,K-ATPase α 1 -β 2 complex in detergent extracts from MDCK cells. A, MDCK cells stably expressing either YFP-β 1 or YFP-β 2 were lysed by incubation with the extraction buffer containing an appropriate detergent (as indicated). After scraping the cells and removing non-extracted material by centrifugation, YFP-linked β 1 or β 2 subunits were immunoprecipitated. Immunoprecipitated YFP-β 1 or YFP-β 2 and co-immunoprecipitated α 1 subunit were analyzed by Western blot. Co-immunoprecipitation of the Na,K-ATPase α 1 subunit with YFP-β 1 was detected in all tested detergents. In contrast, co-immunoprecipitation of the Na,K-ATPase α 1 subunit with YFP-β 2 was observed only in selected detergents. B, A model of the Na,K-ATPase α 1 and β 1 subunits based on the crystal structure of the sodium-potassium pump at 2.4 A resolution (2ZXE) (39) shows the α 1 -interacting residues in the β 1 subunit that are different in the β 2 subunit (yellow labels and yellow halos). The corresponding β 1 -interacting residues in the α 1 subunit are indicated in light-blue. A close-up view of Arg86 of the β 1 subunit and its interacting residues in the α 1 subunit is shown in the right panel. Fig. 8. The Na,K-ATPase α 1 -β 2 complex is preserved in digitonin and CHAPS extracts from MDCK cells. A. Proteins were extracted from MDCK cells expressing YFP-β 1 , or YFP-β 2 , or YFP-linked bile acid transporter (YFP-NTCP), with the extraction buffer containing 1% digitonin. YFP-linked β 1 or β 2 subunits were immunoprecipitated and treated with Endo H followed by elution of proteins from the beads. Co-immunoprecipitation of the Na,K-ATPase α 1 subunit was detected with YFP-β 1 and with YFPβ 2 , but not with YFP-NTCP, indicating that there is no non-specific precipitation of the α 1 subunit. B. Densitometric quantification of the results presented in A was performed by dividing the density of the α 1 subunit band by the density of the corresponding PM YFP-β band. C-D. Proteins were extracted from MDCK cells expressing either YFP-β 1 (C) or YFP-β 2 (D) using extraction buffer containing 1% CHAPS and subject to successive rounds of immunoprecipitation using anti-α antibody followed by a final round of immunoprecipitation using anti-GFP/YFP antibody. Both extracted and immunoprecipitated proteins were treated with Endo H prior to SDS-PAGE. The majority of YFP-β subunits co-immunoprecipitated with α 1 were complex-type glycosylated. Almost all of the β subunits not assembled with α 1 were Endo H sensitive and immunoprecipitated in the final anti-GFP/YFP immunoprecipitation. PM -plasma membrane, Super2supernatant after the second round of immunoprecipitation.