Na,K-ATPase in skeletal muscle: two populations of β-spectrin control localization in the sarcolemma but not partitioning between the sarcolemma and the transverse tubules

The organization of the plasma membrane into distinct structural domains underlies the ability of cells to perform many specific functions, from maintaining ionic homeostasis to synaptic transmission. Understanding how such domains form has been difficult, even in bipolar cells like epithelia. It is especially challenging in excitable cells that, in addition to the usual set of intracellular membrane compartments, also contain many local plasmalemmal domains devoted to synaptic transmission, ion transport, and metabolic trafficking. We are studying skeletal muscle as a model excitable cell, as its plasma membrane, or sarcolemma, is accessible in biochemical amounts, repetitive in organization, and composed of fewer types of domains than the plasma membranes of neurons, with which it otherwise shares many similarities. Our goal is to understand the unusual structure and function of the sarcolemma at the molecular level. Here we focus on the role of the spectrin-based membrane skeleton in organizing the Na,K-ATPase in the sarcolemma and in the transverse tubule system, to which the sarcolemma is connected.

Spectrin is one of the major components of the membrane cytoskeleton responsible for maintaining the biconcave shape of the mammalian erythrocyte (reviewed by Bennett, 1990; Gallagher and Forget, 1993; Hassoun and Palek, 1996). In the erythrocyte, spectrin is composed of heterodimeric complexes of αI and βI subunits (Winkelmann and Forget, 1993), products of the human erythroid spectrin genes, SPTA1 and SPTB, respectively. αβ-Spectrin heterodimers can associate head-to-head to form tetramers and higher oligomers that can polymerize further by virtue of their ability to bind actin (Bennett, 1990; Pumplin and Bloch, 1993; Hartwig, 1995; Viel and Branton, 1996). This protein network is anchored to the erythrocyte membrane through its interaction with erythroid ankyrin, ankyrin 1, which links the β subunit of spectrin to the bicarbonate/chloride exchanger, an integral membrane protein. Mutation or loss of either spectrin or ankyrin results in hemolytic anemia, consistent with the idea that both proteins are vital to the proper support of the red cell membrane (Gallagher and Forget, 1993; Hassoun and Palek, 1996).

Spectrin is also a component of the membrane-associated cytoskeleton in skeletal muscle (Nelson and Lazarides, 1983; Craig and Pardo, 1983; Porter et al., 1992; Vybiral et al., 1992; Porter et al., 1997; Zhou et al., 1998), the importance of which has been underscored because it contains dystrophin, the protein missing in patients with Duchennes Muscular Dystrophy (Hoffman et al., 1987). Spectrin is present at the sarcolemma in two distinct populations, both of which contain the alternatively spliced product of βI-spectrin expressed by striated muscle (βIΣ2: Winkelmann et al., 1990). One population, selectively enriched at sites overlying Z-lines, contains the muscle form of β-spectrin associated with α-fodrin (αII-spectrin), expressed by the SPTAN1 gene. The other population, selectively enriched at sites overlying M-lines and in longitudinally oriented strands, contains muscle β-spectrin without significant amounts of an identifiable α subunit (Porter et al., 1997; Zhou et al., 1998). Viewed in longitudinal sections, the network of β-spectrin at the sarcolemma appears as a rectilinear grid or lattice (Porter et al., 1992; Porter et al., 1997; Williams and Bloch, 1999a,b), the elements of which have been termed ‘costameres’ (Pardo et al., 1983; Porter et al., 1992). Although the molecular cloning and sequencing of β-spectrin from muscle is not yet complete, the available information (Winkelmann et al., 1990; Weed, 1996; Zhou et al., 1998) suggests that the same isoform of βI-spectrin is present in the two populations of β-spectrin at the sarcolemma.

The population of muscle β-spectrin that is associated with α-fodrin forms heterodimers that resemble the αβ-spectrin heterodimers of the erythrocyte (Porter et al., 1997). This similarity suggests that the association of β-spectrin/α-fodrin heteromers with the sarcolemma is likely to be mediated by ankyrin, several isoforms of which are present in skeletal muscle (Nelson and Lazarides, 1984; Flucher and Daniels, 1989; Birkenmeier et al., 1993; Birkenmeier et al., 1998; Devarajan et al., 1996; Zhou et al., 1997; Kordeli et al., 1998; Tuvia et al., 1999; Wood and Slater, 1998). Possible candidates for the integral sarcolemmal protein(s) to which ankyrin in turn binds include the chloride-bicarbonate exchanger, the Na,Ca-exchanger (Li et al., 1993), the voltage-gated Na channel (Srinivasan et al., 1988; Flucher and Daniels, 1989; Wood and Slater, 1998), cell adhesion proteins (Davis and Bennett, 1993), and the Na,K-ATPase (Nelson and Veshnock, 1987; Madreperla et al., 1989; Morrow et al., 1989; Davis and Bennett, 1990; Devarajan et al., 1994; Jordan et al., 1995; Thevananther et al., 1998; Zhang et al., 1998), all of which bind to ankyrin in other excitable cells. As β-spectrin alone is capable of binding ankyrin (Kennedy et al., 1991), it seems likely that the population of muscle β-spectrin that lacks an identifiable α subunit associates with the membrane similarly. How the two populations of β-spectrin form in skeletal muscle, and how they generate distinctive domains at the sarcolemma, are still poorly understood. One possibility is that different structures, approaching the sarcolemma from the Z and M-lines of the nearby contractile apparatus, interact selectively with different components of the membrane skeleton, helping to stabilize and localize the two spectrin populations. A second possibility is that the organization is imposed by domains of the muscle’s basal lamina (e.g. Colognato et al., 1999). Alternatively, the two populations of β-spectrin may bind to different ankyrins that in turn are linked to different integral proteins concentrated in discrete sarcolemmal domains.

We have used a combination of immunofluorescence and immunoprecipitation approaches, as well as mouse mutants that are missing βI-spectrin or the erythroid form of ankyrin (ankyrin 1 or Ank1; Ank_R), to determine the nature of the sarcolemmal complex containing β-spectrin. Our results indicate that both populations of β-spectrin are anchored to the Na,K-ATPase in the sarcolemma via ankyrin 3 (Ank3 or Ank_G). This suggests that the distribution of the two populations of β-spectrin into distinct sarcolemmal domains cannot readily be explained by an association with different sets of ankyrins and integral membrane proteins.

We further report on the subcellular distribution of the two α subunits of the Na,K-ATPase present in normal and mutant skeletal muscle fibers. Although both the α1 and α2 forms of the Na,K-ATPase are present at the sarcolemma, where they associate with muscle β-spectrin, only α2 is present in significant amounts in the transverse tubules (t-tubules; Lavoie et al., 1997). We show that this differential distribution is not altered in mutant muscle lacking β-spectrin. Thus, their association with the spectrin-based membrane skeleton at the sarcolemma does not influence the relative abilities of the α1 and α2 subunits of the Na,K-ATPase to partition into the t-tubules.

Unless otherwise specified, all materials were purchased from Sigma Chemical Co. (St Louis, MO) and were the highest grade available.

The rabbit antibody, 9050, was made against purified human erythrocyte β-spectrin (βIΣ1) and was affinity-purified over a column of erythrocyte β-spectrin and cross-adsorbed against β-fodrin (βII) and α-fodrin (αII), purified from bovine brain (Porter et al., 1997; Zhou et al., 1998). Additional antibodies to erythroid β-spectrin were generated in laying hens and purified with the EggStract kit (Promega, Madison, WI) and affinity purification, as described for 9050. Specificity was demonstrated by immunoblotting (J. A. Ursitti et al., unpublished). The generation and purification of rabbit anti-α-fodrin, 9053, has been described (Porter et al., 1997).

Monoclonal antibodies to the α1 and α2 subunits of the Na,K-ATPase, McK1 and McB2 (Felsenfeld and Sweadner, 1988; Urayama et al., 1989), were from Dr K. J. Sweadner (Massachusettes General Hospital, Boston, MA). Rabbit antibodies to the α1 and α2 subunits, and monoclonal mouse antibody to α1, were purchased from Upstate Biotechnologies (Lake Placid, NY). Antibodies to ankyrins 2 and 3 were generously provided by Dr V. Bennett (Howard Hughes Medical Institute, Duke University, Durham NC) and to ankyrin 3 by Dr J. Morrow (Department of Pathology, Yale University School of Medicine, New Haven CT). Rabbit antibodies to the spectrin-binding domain of ankyrin 1, p65, have been described (Zhou et al., 1997). Non-immune mouse monoclonal antibodies, MOPC21, were obtained from Sigma Chemical Co. (St Louis, MO).

Secondary antibodies included goat anti-rabbit, goat anti-mouse, donkey anti-mouse, and goat anti-sheep IgGs, and goat anti-chicken IgY, conjugated to fluorescein or tetramethylrhodamine for use in immunofluorescence experiments, or to alkaline phosphatase for use in immunoblotting. These antibodies, as well as non-immune sheep and rabbit sera, were from Jackson Immunoresearch Laboratories (West Grove, PA). All secondary antibodies were species-specific with minimal cross reactivity.

Female Sprague-Dawley rats, aged 6 months to 1 yr (Zivic-Miller, Zelienople, PA) were used. Ankyrin 1-deficient mice (nb/nb: White et al., 1990), β-spectrin-deficient mice (ja/ja: Bodine et al., 1984; Bloom et al., 1994), and age-matched controls were bred and raised in the Barker laboratory. As soon as they suckled, newborn ja/ja mice were transfused through the superficial temporal vein with 0.1 ml of concentrated erythrocytes, obtained from blood drawn from the retroorbital sinus of a C57BL/6J (B6)-+/+ mouse. Transfusion allowed the ja/ja mice, which normally die within 7 days after birth, to survive for an average of 3.7 months (Kaysser et al., 1997).

Animals were anesthetized and sacrificed by perfusion fixation, as described (Williams and Bloch, 1999a). The tibialis anterior (TA), extensor digitorum longus (EDL) and quadriceps muscles were removed and incubated for an additional 5 minutes in 2% paraformaldehyde in PBS. Tissue was blotted dry, snap frozen and cryosectioned (20 μm thickness). Samples were collected on slides coated with a solution of 0.5% gelatin, 0.05% chromium potassium sulfate, and stored at −70°C. Tissue for cross sections (20 μm) was obtained without perfusion or fixation, but was otherwise handled as above. The preparation of unfixed longitudinal sections has been reported (Williams and Bloch, 1999a).

Sections were incubated in PBS/BSA (PBS containing 1 mg/ml BSA, 10 mM sodium azide) for 15 minutes and then in primary antibody in PBS/BSA for 2 hours at room temperature, or overnight at 4°C. When sheep antibodies to the dihydropyridine receptor (DHPR) were used, solutions contained 0.01% Triton X-100. After washing, samples were incubated for 1 hour in secondary antibodies in PBS/BSA, washed again and mounted (Williams and Bloch, 1999a). Samples were observed with a Zeiss 410 confocal laser scanning microscope equipped with a ×63, NA 1.4 plan-apochromatic objective. The pinholes for fluorescein and tetramethylrhodamine were 18. All labeling was shown to be specific through the use of the appropriate non-immune controls (e.g. Fig. 1).

To generate figures, images were arranged and labeled with Corel Draw 6 (Corel Corporation Ltd, Ottawa, Ontario). Insets were prepared with Metamorph (Universal Imaging, West Chester, PA) and magnified 2-fold with Corel Draw 6.

Sprague-Dawley rats, anesthetized as above, were perfused (Porter et al., 1997). The major muscle groups from the hindlimb were dissected and frozen in liquid nitrogen. While submerged in liquid nitrogen, the sample was ground to a fine powder with a mortar and pestle. This powder was suspended in a solution (modified from Hoffman et al., 1989) containing 1% deoxycholate, 1% NP-40, 10 mM sodium phosphate, 0.14 M NaCl, 2 mM EDTA, pH 6.8, supplemented with protease inhibitors (Porter et al., 1992), homogenized in a Polytron PT 10/35 Brinkmann homogenizer at 4°C for 1 minute (4× 15 seconds), and incubated for 1 hour at 4°C. Insoluble material was removed by centrifugation at 16,000 RPM for 1 hour (4°C, SS-34 rotor, Sorvall Instruments, Newton, CT), and the supernatant was stored at −70°C.

T-tubule membranes (fraction F3) and sarcolemmal fractions (fraction PF6) were isolated from rabbit skeletal muscle, as described (Dombrowksi et al., 1996).

Proteins were separated by SDS-PAGE on 5-15% acrylamide minigels (Hoefer, San Francisco, CA) as described (Laemmli, 1970), except that samples to be tested for the Na,K-ATPase were incubated in sample buffer at 37°C for 15 minutes instead of boiling. Molecular mass standards were acquired from Bethesda Research Laboratories (Bethesda, MD). Some gels were visualized with Coomassie Brilliant Blue or silver staining. For immunoblotting, proteins were transferred to nitrocellulose (Burnette, 1981). Blots were incubated briefly in milk-PTA (3% milk solids 10 mM NaN₃, 0.5% Tween-20 in PBS) and then overnight at 4°C or for 2 hours at room temperature in primary antibody in milk-PTA. After washing, samples were incubated with secondary antibodies in milk-PTA for 1 hour at room temperature. Bound antibody was visualized chromogenically (Kirkegaard and Perry, Gaithersburg, MD) or by chemiluminescence (‘Western Light Detection’ kit, Tropix Laboratories, Bedford, MA).

Aliquots of rat muscle homogenate containing 1 mg protein were incubated overnight with mouse IgG bound to Sepharose beads, to remove nonspecifically bound protein. (Early experiments had shown a nonspecific reaction of mouse secondary antibodies with a ∼100 kDa protein in immunoblots of unboiled samples that was eliminated by this step.) Beads were pelleted by centrifugation (14,000 rpm, 5 seconds, Eppendorf 5415 centrifuge). The pellet was mixed with a volume of sample buffer (Laemmli, 1970) equal to 1/3 of the original homogenate volume. The remaining supernatant was then precipitated overnight with Protein A-Sepharose beads (Pharmacia, LKB, Uppsala, Sweden), to remove more non-specific binding components. After centrifugation, the remaining supernatant was incubated with Protein A-Sepharose beads bound overnight at 4°C to the appropriate antibodies. Aliquots containing 10 μg of rabbit antibody or the equivalent in normal rabbit serum were used for each mg protein in the original homogenate. Beads were mixed overnight at 4°C with supernatant. After centrifugation, the pellet was washed and mixed with SDS-PAGE sample buffer, as above, and divided into two equal fractions. One fraction was incubated at 37°C for 15 minutes; the other was boiled for 5 minutes. Beads and sample buffer were separated by low speed centrifugation through a Centricon filter (pore size, 0.45 μm: Amicon, Danvers, MA).

We used the affinity-purified antibody, 9050, for immunoprecipitation of β-spectrin. For sequential immunoprecipitations, homogenates were precipitated first with 9053 anti-α-fodrin, then with 9053 again, to assure that all the α-fodrin was removed, and finally with 9050, to precipitate the remaining β-spectrin (Porter et al., 1997).

Our experiments were designed to elucidate the relationship between the two isoforms of the Na,K-ATPase expressed in skeletal muscle, and the two distinct populations of β-spectrin present at the sarcolemma. We chose to study the Na,K-ATPase because it has been associated with the spectrin-based membrane skeleton in other cells, and isoform-specific antibodies are available to the two forms, α1 and α2, expressed in skeletal muscle (Orlowski and Lingrel, 1988). We focus on fast twitch muscle, because it displays clearly defined structures containing the two populations of muscle β-spectrin (Williams and Bloch, 1999a,b), one over Z-lines that is composed of β-spectrin and α-fodrin, the other over M-lines and in longitudinal strands that contains β-spectrin without any identifiable α subunit (Porter et al., 1997). We use immunofluorescence and immunoprecipitation protocols, as well as mouse mutants that lack key components of the membrane skeleton, to study the ability of the α1 and α2 subunits of the Na,K-ATPase to associate with ankyrin 3 and spectrin at the sarcolemma, and their ability to partition between the sarcolemma and the transverse tubules of myofibers.

Double immunofluorescence labeling experiments were used to examine the distribution of the α1 and α2 subunits of the Na,K-ATPase and β-spectrin in paraformaldehyde-fixed, longitudinal cryosections of rat extensor digitorum longus (EDL) muscle. Primary antibodies were visualized with species-specific secondary antibodies coupled to fluorescein and tetramethylrhodamine. As reported previously (see Introduction for references), muscle β-spectrin is concentrated in costameres, present at the sarcolemma over Z and M-lines and in longitudinal strands (Fig. 1A and inset). The distribution of the α1 subunit of the Na,K-ATPase was nearly identical to that of β-spectrin, with high concentrations at the costameres and no significant staining in the intercostameric regions (Fig. 1B; see Williams and Bloch, 1999b). This was confirmed by computer-generated composite pictures (Fig. 1C). We obtained similar results when we compared the distribution of the α2 subunit of the Na,K-ATPase with β-spectrin (Fig. 1D-F). We never observed regions of the sarcolemma that contained significant amounts of β-spectrin without also containing the α1 or α2 subunits of the Na,K-ATPase, or that contained significant amounts of one of the subunits of the Na,K-ATPase without also containing β-spectrin. Controls omitting one of the primary antibodies showed that this colocalization was not due to fluorescence ‘bleed-through’ or species cross-reactivity by secondary antibodies (e.g. Fig. 1J-L). As the primary antibodies react specifically with their respective antigens (see below), all the labeling was therefore specific. These results suggest that both the α1 and α2 subunits of the Na,K-ATPase concentrate with β-spectrin in costameres of fast twitch myofibers of the rat.

We next examined ankyrin 3, which has been associated with the Na,K-ATPase in other tissues. As reported above for the α subunits of the Na,K-ATPase, ankyrin 3 colocalized with β-spectrin (Fig. 1G-I), suggesting that both ankyrin 3 and the two α subunits of the Na,K-ATPase are concentrated in all regions of costameres, together with muscle β-spectrin. Ankyrin 3 and the Na,K-ATPase are therefore likely to associate with both populations of β-spectrin (see Introduction), the population over Z-lines bound to α-fodrin, and the population over M-lines and in longitudinal domains that does not contain significant amounts of α-fodrin.

We used immunoprecipitation to study the association of the Na,K-ATPase and ankyrin 3 with muscle β-spectrin. Immunoprecipitates generated from muscle homogenates with affinity-purified antibodies to β-spectrin (see Materials and Methods) were collected and examined by immunoblotting for ankyrin 3 and for each of the α subunits of the Na,K-ATPase.

Precipitation with anti-β-spectrin concentrated β-spectrin in the pellet (Fig. 2A, lane 2; band at ∼265 kDa). The pellet also contained significant amounts of ankyrin 3 (Fig. 2D, lane 2; band at ∼190 kDa; see also Thevananther et al., 1998) and the α1 and α2 subunits of the Na,K-ATPase (Fig. 2B,C, lanes 2; bands at ∼100 kDa). We could not detect ankyrin 1 or ankyrin 2 in the immunoprecipitate, however (not shown). The presence of β-spectrin, ankyrin 3 and the two isoforms of the Na,K-ATPase in the immunoprecipitate was specific, as they could not be detected in control immunoprecipitates generated with normal rabbit serum (Fig. 2A-E, lanes 1), nor were they labeled in the immunoblots by a non-immune mouse (Fig. 2E, lane 2) or rabbit IgG (not shown). These results suggest that muscle β-spectrin, ankyrin 3 and both the α1 and α2 subunits of the Na,K-ATPase are associated in a complex in vivo.

The presence of the Na,K-ATPase and ankyrin 3 at all costameric regions suggests that these proteins interact not only with the population of β-spectrin at Z-lines, associated with α-fodrin, but also with the population of β-spectrin at M-lines and in longitudinal domains, which do not contain significant amounts of α-fodrin or any other identifiable α-spectrin-like subunit (Porter et al., 1997; Zhou et al., 1998). We employed sequential immunoprecipitation to examine this further. We used antibodies to α-fodrin in two rounds of immunoprecipitation to obtain the β-spectrin/α-fodrin heteromers, and then immunoprecipitated with antibodies to β-spectrin to isolate β-spectrin that is free of α-fodrin (Porter et al., 1997). The pellets were analyzed, as above, for the presence of β-spectrin, ankyrin 3, and the α1 and α2 subunits of the Na,K-ATPase, as well as for α-fodrin. As controls we used precipitates generated with non-immune sera.

The first immunoprecipitation with anti-α-fodrin concentrated both β-spectrin and α-fodrin in the pellet (Fig. 3A,B, lanes 2), as reported (Porter et al., 1997). The second precipitation with anti-α-fodrin again concentrated β-spectrin and α-fodrin in the pellet, although in apparently smaller amounts than in the previous precipitation (Fig. 3A,B, lanes 3), consistent with the nearly complete removal of the α-fodrin from the supernatant (not shown). The final precipitation with anti-β-spectrin concentrated most of the remaining β-spectrin in the pellet (Fig. 3A, lane 4). Neither α-fodrin nor β-spectrin was precipitated non-specifically, as pellets generated with non-immune rabbit serum did not contain either protein (Fig. 3A,B, lanes 1). Thus, sequential immunoprecipitation separates a population of β-spectrin that is associated with α-fodrin from one that is not (Porter et al., 1997).

We probed these samples with antibodies to the α1 and α2 subunits of the Na,K-ATPase (Fig. 3C,D) and to ankyrin 3 (Fig. 3E). Both subunits of the Na,K-ATPase were present in the precipitates generated by anti-α-fodrin (Fig. 3C,D lanes 2,3). The final immunoprecipitation with anti-β-spectrin also concentrated these proteins in the pellet (Fig. 3C,D, lanes 4). Ankyrin 3 was also present in the two immunoprecipitates generated by anti-α-fodrin (Fig. 3E, lanes 2,3) and in the immunoprecipitate generated by anti-β-spectrin (Fig. 3E, lane 4). Neither of the α subunits of the Na,K-ATPase nor ankyrin 3 were concentrated in the non-immune precipitate (Fig. 3C-E, lanes 1). These results show that ankyrin 3 and both the α1 and α2 subunits of the Na,K-ATPase associate specifically with both populations of β-spectrin in skeletal muscle. To our knowledge, this is the first time that the biochemical association of ankyrin 3 with βI-spectrin has been demonstrated to occur in situ.

We further tested the idea that the α1 and α2 subunits of the Na,K-ATPase associate with both populations of muscle β-spectrin by examining the sarcolemma of two mouse mutants, ja/ja and nb/nb, which lack βI-spectrin and ankyrin 1, respectively (Bodine et al., 1984; White et al., 1990; Bloom et al., 1994). Based on our results with rat muscle, we predicted that both forms of the Na,K-ATPase would be nearly uniformly distributed at the sarcolemma of ja/ja mice but would be found normally in costameres of nb/nb mice.

Double immunofluorescence labeling of the sarcolemma of fast twitch, EDL muscle of wild-type mice showed the costameric pattern described above. Muscle β-spectrin, recognized with affinity-purified chicken antibodies, was concentrated in a rectilinear array composed of longitudinal strands and elements overlying Z-lines and M-lines (Fig. 4A,E). Both the α1 and α2 subunits of the Na,K-ATPase were concentrated in the same structures (Fig. 4B,F). We observed similar patterns in EDL fibers from the nb/nb mouse (Fig. 4C,D,G,H), suggesting that ankyrin 1 is not necessary for the organization of β-spectrin or either isoform of the Na,K-ATPase at the sarcolemma.

The organization of the sarcolemma in the ja/ja mouse was significantly altered, however. As expected from the nature of the ja/ja mutation (Bloom et al., 1994), we detected no β-spectrin at the sarcolemma (Fig. 4K), which did, however, label with antibodies to the α subunits of the Na,K-ATPase (Fig. 4I,J). Unlike the wild type, both the α1 and the α2 subunits of the Na,K-ATPase were more broadly distributed in the sarcolemma of ja/ja muscle (Fig. 4I,J). Indeed, the α1 subunit appeared nearly uniformly distributed (Fig. 4I). The α2 subunit also redistributed, as it was clearly not confined to the rectilinear lattice of costameres in ja/ja muscle, although it could still be detected in irregular longitudinal strands and in Z-lines (Fig. 4J). The α2 subunit is also present intracellularly (Fig. 6D), so its apparent presence at Z-lines may be due to the inability of the confocal microscope to distinguish between structures in the membrane and in nearby myofibrils. The irregularity of the longitudinal strands containing α2 suggests the presence of folds, but the possibility remains that some of the longitudinal domains of costameres are stable in ja/ja muscle. Nevertheless, our results show that the sarcolemmal organization of both the α1 and the α2 subunits of the Na,K-ATPase is altered when β-spectrin is absent. We obtained similar results on muscle fibers that had been fixed in situ and on fibers that were cryosectioned from unfixed tissue, suggesting that the altered distribution of the Na,K-ATPase was not caused during the collecting or processing of tissue samples. These results suggest that β-spectrin is required for the localization of both the α1 and the α2 forms of the Na,K-ATPase at costameres.

In addition to labeling at the sarcolemma, antibodies to the α2 subunit of the Na,K-ATPase, but not the α1 subunit, labeled structures in the myoplasm. The α2 subunit has been found in subcellular fractions enriched in t-tubules, while both the α1 and the α2 subunits are present in sarcolemmal fractions (Lavoie et al., 1997). We confirmed this observation in rabbit calf muscle by immunoblotting fractions isolated from sarcolemma and t-tubular membrane (Fig. 5). The t-tubule fraction, identified by the presence of high concentrations of the dihydropyridine receptor (DHPR: Fig. 5C) contained only the α2 subunit and had no detectable α1 subunit (Fig. 5A,B, lanes 1). The sarcolemmal fraction, which lacked detectable DHPR, contained both α subunits of the Na,K-ATPase (Fig. 5A,B, lanes 2).

We also used double immunofluorescence labeling of unfixed cross sections of rat EDL muscle to examine the subcellular distribution of the α1 and α2 subunits of the Na,K-ATPase. Antibodies to the α2 (Fig. 6D) but not the α1 (Fig. 6A) subunit revealed a distinctive reticular pattern in the sarcoplasm that co-labeled with antibodies to the DHPR (Fig. 6D-F), consistent with the presence of the α2 subunit of the Na,K-ATPase in t-tubules. Antibodies to the α1 subunit of the Na,K-ATPase did not show significant intracellular labeling and so gave no overlap with DHPR (Fig. 6A-C). Thus, morphological studies confirm the results of subcellular fractionation.

We examined muscle from the ja/ja mouse to learn if interactions with the spectrin-based membrane skeleton at the sarcolemma plays a role in excluding the α1 subunit of the Na,K-ATPase from the t-tubules. Frozen cryosections of EDL muscle from ja/ja mice were double labeled with antibodies to the α1 or α2 subunits of the Na,K-ATPase, together with antibodies to the DHPR, to mark t-tubules (not shown). As with controls (Fig. 6G), the α2 subunit was present both at the sarcolemma and in the t-tubules of ja/ja muscle (Fig. 6H), while the α1 subunit remained restricted to the sarcolemma (Fig. 6I,J). Thus, there is no significant change in the subcellular distribution of either form of the Na,K-ATPase when β-spectrin is absent from the sarcolemmal membrane skeleton.

The formation of distinctive membrane domains requires that particular membrane proteins be delivered to and retained in the plasma membrane or in appropriate subcellular membrane fractions and, often, that these proteins be concentrated together into characteristic regions or structures. The cytoskeleton has been proposed to play a role in both processes, first by helping to deliver proteins to particular membrane systems, and then by interacting with proteins to concentrate them into distinctive structures or domains. These processes are not understood in excitable cells, nor have they been studied with proteins, like the Na,K-ATPase, that are targeted to more than one membrane system and that can accumulate in distinctive structures in those membranes. Here we study the role in these processes of one part of the cytoskeleton, the spectrin-based membrane skeleton. We show that β-spectrin is required for the accumulation of both forms of the Na,K-ATPase into costameres at the sarcolemma, but not for the differential partitioning of the α1 and α2 forms of the Na,K-ATPase between the sarcolemmal and the t-tubular membranes.

Association of spectrins, including the general tissue homolog, fodrin, with the Na,K-ATPase has been studied extensively in renal epithelium (Nelson and Veshnock, 1987; Morrow et al., 1989; Nelson and Hammerton, 1989; Davis and Bennett, 1990; see also Madreperla et al., 1989; Smith et al., 1993). The predominant epithelial spectrins are heterotetramers of α- and β-fodrin ([αIIβII]₂) that associate with the Na,K-ATPase through ankyrin 3 (Ank_G: Nelson and Hammerton, 1989; Koob et al., 1990; Hu et al., 1995; Peters et al., 1995; Thevananther et al., 1998). By contrast, mature, fast twitch skeletal muscle fibers express several ankyrins (see Introduction for references), as well as α-fodrin and an excess of β-spectrin (Porter et al., 1997; Zhou et al., 1998), but no β-fodrin (Weed, 1996; Zhou et al., 1998; Williams et al., 2000). Despite these differences, our results suggest that the Na,K-ATPase at the sarcolemma associates quite efficiently with ankyrin 3 and muscle β-spectrin.

Indeed, our immunofluorescence studies indicate that both the α1 and α2 forms of the Na,K-ATPase are restricted to costameres, and that this restriction requires β-spectrin, whether or not it is associated with α-fodrin. In wild-type muscle little or no labeling of the Na,K-ATPase could be detected in sarcolemmal regions outside of costameres, whereas in ja/ja samples the Na,K-ATPase (especially the α1 subunit) seemed nearly uniformly distributed in the sarcolemma. Our immunoprecipitation protocols confirmed the association of both subunits of the Na,K-ATPase with β-spectrin. Surprisingly, considering the number of studies of the Na,K-ATPase and its association with the membrane skeleton in epithelia, ours appears to be the first definitive demonstration that the Na,K-ATPase can associate with βI-spectrin in mammalian cells (see also Vladimirova et al., 1998; Williams and Bloch, 1999a).

Our results further suggest that the linkage of the Na,K-ATPase to the spectrin-based membrane skeleton is mediated to a significant extent by ankyrin 3, which associates with muscle β-spectrin whether or not it is bound to α-fodrin. Indeed, our results are the first to demonstrate biochemically the association in vivo of βI-spectrin (the product of the erythroid spectrin gene) with ankyrin 3. We have not yet detected either ankyrin 1 or ankyrin 2 in our immunoprecipitates (not shown), but this result is difficult to interpret. Both ankyrin 1 and 2 are present at the sarcolemma (Zhou et al., 1997; Tuvia et al., 1999) and should be able to bind simultaneously to integral membrane proteins and to β-spectrin. These ankyrins may dissociate more readily than ankyrin 3 under the conditions we used, or they may be selectively degraded in muscle homogenates. Although this question warrants further study, our current evidence suggests that the linkage of the Na,K-ATPase to spectrin is mediated predominantly by ankyrin 3.

Adult skeletal muscle expresses two of the three known α subunits of the Na,K-ATPase, α1 and α2 (Orlowski and Lingrel, 1988). It has been reported (Lavoie et al., 1997) that, while both forms were present in the sarcolemma, only α2 was present in significant amounts in the t-tubules of skeletal muscle. We have confirmed this observation by immunofluorescence techniques, as well as by subcellular fractionation. The basis for the differential partitioning of the α1 and α2 subunits of the Na,K-ATPase between the sarcolemma and the t-tubules is still not understood, but our results suggest that interactions with the spectrin-based membrane skeleton are not involved. Both forms of the Na,K-ATPase associate with spectrin at the sarcolemma, and both become more uniformly distributed in the sarcolemma of ja/ja mice. Nevertheless, the α1 form of the Na,K-ATPase remains restricted to the sarcolemma in ja/ja muscle. This observation can be explained in several ways. (i) The α1 form of the Na,K-ATPase can indeed enter the t-tubule system in skeletal myofibers of the ja/ja mouse, but once in the t-tubule, it is rapidly removed or degraded. (ii) Although they are more uniformly distributed in the sarcolemma of the ja/ja mouse, both forms of the Na,K-ATPase remain bound to other integral or peripheral sarcolemmal proteins that prevent their movement into the t-tubule system. (iii) A diffusion barrier, located near the junction of the sarcolemma with the t-tubules, restricts the movement of sarcolemmal proteins into the t-tubular membrane. We cannot now distinguish among these possibilities. Whichever explanation applies, our results clearly show that the spectrin-based membrane skeleton at the sarcolemma does not play a significant role in determining the distribution of the Na,K-ATPase between the sarcolemma and the t-tubules. It seems likely that other factors, perhaps including targeting sequences in the α subunits of the Na,K-ATPase and their association with different β subunits, direct the α2 but not the α1 subunit to the t-tubules. This process is likely to be regulated by hormonal signaling and by the metabolic state of the muscle (e.g. Lavoie et al., 1996; Tsakiridis et al., 1996).

In summary, our results clearly demonstrate the presence in skeletal muscle of a complex of the Na,K-ATPase, ankyrin 3 and muscle β-spectrin, with or without a paired α-spectrin-like subunit. This complex is responsible for the localization of both the α1 and α2 isoforms of the Na,K-ATPase to costameres at the sarcolemma of fast twitch myofibers, but it appears to play no significant role in the partitioning of the α1 and α2 forms of the Na,K-ATPase between the sarcolemma and the t-tubules.

Our results do not readily explain why β-spectrin/α-fodrin heteromers concentrate over Z-lines, while β-spectrin without a paired α subunit concentrates over M-lines and in longitudinal elements. Our findings suggest that any model proposing that the differential distribution of the two populations of spectrin is due to their binding to distinct complexes of ankyrins and integral membrane proteins is highly unlikely. We now favor a model in which the differential distribution of the two populations of spectrin at costameres is determined by links between the sarcolemma and the contractile apparatus of nearby myofibrils (Fig. 7).

In our model, these links are made by intermediate filaments, which have been shown to connect the sarcolemma to the contractile apparatus at the level of Z- and M-lines (Pierobon-Bormioli, 1981; Street, 1983; Shear and Bloch, 1985). Desmin concentrates selectively around Z disks (Lazarides, 1978; Granger and Lazarides, 1979; Richardson et al., 1981). It is also present at the sarcolemma at Z line domains, but as it does not associate to a significant extent with longitudinal or M line domains (A. O’Neill et al., unpublished), we have proposed an additional structure that performs this function (‘connectors’ in Fig. 7). We postulate further that the desmin filaments promote the accumulation of the β-spectrin/α-fodrin heteromers at Z line domains, while the ‘connectors’ promote the accumulation of the population of β-spectrin that lacks an α subunit at longitudinal and M line domains. Intermediate filaments have been shown to bind spectrin and ankyrin (Langley and Cohen, 1986; Langley and Cohen, 1987; Georgatos and Blobel, 1987), consistent with the possibility that these links are established by direct interactions of the filaments with the spectrin-based membrane skeleton. The concentration of spectrin and its associated proteins into costameres should in turn concentrate the Na,K-ATPase. Rigorous testing of this model will help to reveal the identity and function of the proteins that organize the sarcolemma and how they interact simultaneously with the contractile apparatus and the membrane at costameres. As changes in the organization of costameres have been linked to muscular dystrophy (Porter et al., 1992; Ehmer et al., 1997; Williams and Bloch, 1999a), the factors that organize the sarcolemma are likely to serve an important function in the physiology of muscle.

We are grateful to A. O’Neill for her assistance, to Drs K. Sweadner, V. Bennett and J. S. Morrow for their generous gifts of antibodies, and to Drs N. C. Porter, D. W. Pumplin, W. R. Randall, M. P. Blaustein and M. F. Schneider for useful discussions. Our research has been supported by grants to R. J. Bloch from the National Institutes of Health (NS 17282, HL64304) and from the Muscular Dystrophy Association.