Probing sodium channel cytoplasmic domain structure. Evidence for the interaction of the rSkM1 amino and carboxyl termini.

Epitopes for monoclonal antibodies directed against the purified adult rat skeletal muscle sodium channel (rSkM1) were localized using channel proteolysis and fusion proteins. The interactions between these and other monoclonal antibodies with site-specific polyclonal antibodies were used to investigate the spatial relationships among rSkM1 cytoplasmic segments. Competition between antibodies for binding was performed using a solution-phase assay in which solubilized channel protein retains many of the biophysical characteristics of the rSkM1 protein in vivo. Our results support a model in which: 1) the amino terminus assumes a rigid structure having a fixed orientation with respect to other intracellular segments; 2) the interdomain 2-3 region is centrally located on the cytoplasmic surface of the channel, extends farther into the cytoplasm, and has an intermediate degree of flexibility; 3) the beginning of the amino terminus and end of the carboxyl terminus specifically interact with each other; and 4) domains 1 and 4 are adjacent. The sequences responsible for the interaction of the amino and carboxyl termini were identified by demonstrating the specific binding of a synthetic peptide encompassing the first 30 residues of the rSkM1 amino terminus to a fusion protein containing the rSkM1 carboxyl terminus.

Epitopes for monoclonal antibodies directed against the purified adult rat skeletal muscle sodium channel (rSkM1) were localized using channel proteolysis and fusion proteins. The interactions between these and other monoclonal antibodies with site-specific polyclonal antibodies were used to investigate the spatial relationships among rSkM1 cytoplasmic segments. Competition between antibodies for binding was performed using a solution-phase assay in which solubilized channel protein retains many of the biophysical characteristics of the rSkM1 protein in vivo. Our results support a model in which: 1) the amino terminus assumes a rigid structure having a fixed orientation with respect to other intracellular segments; 2) the interdomain 2-3 region is centrally located on the cytoplasmic surface of the channel, extends farther into the cytoplasm, and has an intermediate degree of flexibility; 3) the beginning of the amino terminus and end of the carboxyl terminus specifically interact with each other; and 4) domains 1 and 4 are adjacent. The sequences responsible for the interaction of the amino and carboxyl termini were identified by demonstrating the specific binding of a synthetic peptide encompassing the first 30 residues of the rSkM1 amino terminus to a fusion protein containing the rSkM1 carboxyl terminus.
In studies of voltage-dependent ion channels, a major goal is to correlate specific aspects of protein structure with channel function. In order to attain this goal, an accurate model of channel tertiary structure is required. Sequence information for a variety of voltage-dependent sodium channels is now available, providing the basis for several models of channel tertiary structure (1)(2)(3)(4)(5). All current models postulate the presence of four membrane-embedded homologous domains joined by cytoplasmic linking and terminal sequences. Although the models were initially based largely on theoretical considerations, various aspects have been tested using a variety of molecular, biochemical, and immunological techniques, and their general features have been validated.
In previous studies, we used a combination of limited proteolysis and antibody binding to provide experimental support for the presence of four compact repeat domains in the skeletal muscle sodium channel, to identify the topography of the regions that join and flank these domains, and to probe the relative orientation of the large extramembrane cytoplasmic elements (6 -11). This work also allowed us to map the location of epitopes for monoclonal antibodies we had previously generated against purified sodium channel protein. Our binding studies divided these antibodies into two large groups based on mutually exclusive competition (11). While epitopes in each group were typically clustered in similar regions of the channel sequence, in several cases we found monoclonals in the same group that recognized epitopes widely separated in the primary sequence but presumably brought together in the native protein by folding of the polypeptide backbone (6,12). Identification of similar interactions that reflect tertiary rather than primary structure can provide useful information about the structural organization of the channel's cytoplasmic domains and form the basis for the work reported here.
In this study, we first localize epitopes for additional monoclonals against the channel. Competition between our monoclonal panel and polyclonal antibodies developed against defined channel oligopeptides is then used to probe the spatial organization of epitopes in the native channel. Using this approach, we have developed a model for the organization of the channel cytoplasmic domains and have identified a specific interaction between the channel's amino and carboxyl termini.

EXPERIMENTAL PROCEDURES
Materials-Materials for the preparation of oligopeptides and antibodies, the isolation of crude membranes, and for the purification of membrane proteins were obtained from sources previously identified (8). DEAE-Sephadex (A25-120), wheat germ agglutinin-agarose, protease inhibitors, and one set of prestained molecular mass standards (26.2-180 kDa) were obtained from Sigma. Another set of prestained molecular mass standards (15-110 kDa) was obtained from Bio-Rad. 125 I-protein A and 125 I-labeled goat anti-mouse IgG were from ICN Radiochemicals (Irvine, CA). The pMal fusion protein and purification system was obtained from New England Biolabs.
Preparation of Antisera-Monoclonal antibodies were previously generated by immunizing mice with purified detergent-solubilized rat skeletal muscle sodium channel protein (13). Oligopeptides were synthesized and polyclonal antibodies generated against these synthetic oligopeptides using methods detailed in earlier publications (8 -10). All oligopeptides correspond to regions of the rat skeletal muscle sodium channel protein sequence ( Fig. 1 and Table I). The carboxyl-terminal cysteine residue which terminates each peptide is not part of the naturally occurring sequence and was added to assist in coupling peptide to carrier protein prior to rabbit immunization.
Sodium Channel Protein Purification-Preparation of muscle surface membranes containing unproteolyzed sodium channels was identical to that described previously (8). Sodium channel protein was solubilized using Nonidet P-40 and then purified using sequential ion exchange and lectin affinity chromatography (8). Purified solubilized sodium channel protein was kept at 4°C and used within 2 weeks.
Antibody Binding Assay-Purified rSkM1 protein (5 pmol) was adsorbed to 25 l of wheat germ lectin-Sepharose 4B by gentle agitation for 1 h at room temperature. The resin was pelleted and washed with * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
phosphate-buffered saline, pH 7.4 (PBS). 1 Monoclonal or polyclonal antisera was added, incubated with resin, and samples were removed at 30-min intervals over 6 h. After pelleting and washing, the resin was incubated with 5 Ci of 125 I-labeled goat anti-mouse IgG (monoclonal antibodies) or 5 Ci of 125 I-protein A (polyclonal antibodies) for 2 h at room temperature. The resin was then washed with PBS and immobilized radioactivity quantitated in a gamma counter.
Antibody Competition Assay-Purified rSkM1 protein (5 pmol) was added to 25 l of wheat germ lectin-Sepharose 4B and incubated for 1 h at room temperature. The resin was gently pelleted and washed with PBS. Serial dilutions of polyclonal antibody were added and incubated with resin for 2 h. The resin was washed again and then incubated with a fixed concentration of secondary antibody (monoclonal supernatant) for 4 h at room temperature. Following pelleting and washing, the resin was incubated with 5 Ci of 125 I-labeled goat anti-mouse IgG for 2 h at room temperature and washed with PBS; immobilized radioactivity was quantitated in a gamma counter.
Fusion Protein Constructs-Fusion proteins expressed from subcloned fragments of the rSkM1 and rSkM2 (also known as rH1) interdomain 2-3 or carboxyl-terminal regions were expressed and purified using the pMal protein fusion and purification system (New England Biolabs). Briefly, the cloned gene was inserted downstream from the malE gene encoding the maltose-binding protein (MBP). This construct directs the expression of a fusion protein containing MBP at its aminoterminal end and a fragment of the cloned sodium channel interdomain or carboxyl-terminal region at the carboxyl-terminal end. Details of construct synthesis and the methods used to generate different length fragments of the rSkM1 interdomain 2-3 and rSkM1 and rSkM2 carboxyl-terminal regions are given in another publication (14). Western blots of these fusion proteins incubated with monoclonal antibodies F/C11 and B/D6 were used to localize the epitopes for these antibodies (see Sun et al. (14) for details).
Peptide Binding to Interdomain 2-3 or Carboxyl-terminal Fusion Proteins-Fusion protein constructs (0.5 nmol) bound to amylose resin were incubated with 50 nmol of synthetic peptide for 4 h at room temperature or overnight at 4°C. Resin was washed five times with PBS, 0.25% Tween 20, incubated with either A/B2 or L/D3 monoclonal supernatant (dilution 1:10) for 4 h, washed again as above, and then incubated with 0.8 Ci of 125 I-labeled goat anti-mouse IgG for 4 h at room temperature or overnight at 4°C. The resin was washed five times with PBS, 0.25% Tween 20 over 1 h before radioactivity bound to resin was quantitated by gamma counting.

RESULTS
Monoclonal Epitope Localization-Having localized the epitopes for representative members of one group of monoclonal antibodies (A/B2 and L/D3) to the amino terminus of rSkM1 in a previous study (12), our initial goal was to identify the epitopes for two representative members (F/C11 and B/D6) of a second group of monoclonal antibodies. Regional localization was obtained by comparing the pattern of proteolyzed rSkM1 sodium channel fragments that bound these antibodies to the patterns already described for sodium channel fragments visualized by antibodies directed against known epitopes distributed throughout the channel primary structure (8). The pattern of fragments as well as the limit peptide identified by antibodies F/C11 and B/D6 indicates that the epitopes for these two monoclonal antibodies lie within a region encompassing interdomain 2-3 and domain 3 (Fig. 2).
These epitopes were further localized to the interdomain 2-3 region on the basis of antibody binding to a fusion protein that contained residues 794-1014 in the rSkM1 sequence. Both antibodies reacted specifically with this fusion protein both on Western blots and in radioimmunoassay. We refined this localization using fusion proteins containing successively smaller fragments of the interdomain 2-3 region. These fusion proteins were constructed using either naturally occurring restriction sites in the coding sequence or synthetic primers in conjunction with the polymerase chain reaction (see Sun et al. (14) for details). Our binding data (Fig. 3) restricts the epitope for F/C11 to residues 865-875 (mid-portion of interdomain 2-3) and the epitope for B/D6 to residues 965-975 (carboxyl-terminal half of interdomain 2-3 just prior to domain 3) in the rSkM1 sequence.
Antibody Competition Studies-To investigate the spatial relationships of sodium channel intracellular segments, four representative monoclonal antibodies were assayed for binding competition with a panel of polyclonal antisera for which defined epitopes are already known (Table I). Measurements were made with solubilized purified rSkM1 sodium channel protein in a mixed micellar form that retains native channel structure, as indicated by toxin binding activity (15), sensitivity to proteolytic enzymes (8), and capacity for functional reconstitution (16,17). Channel protein was immobilized to wheat germ-Sepharose beads. We first measured the kinetics of 1 The abbreviations used are: PBS, phosphate-buffered saline; MBP, maltose-binding protein.  Table I for details of the residues comprising each antibody epitope.

TABLE I Antibodies
List of monoclonal and polyclonal antibodies used in this study with the location and the specific residues comprising the epitope for each antibody in the rSkM1 sequence. Fig. 1  antibody binding to immobilized sodium channel protein to ensure that all measurements were performed under equilibrium conditions. Binding of individual monoclonal antibodies to the channel reached 90% of final values within 15 min of incubation at 22°C. When no competition was seen between a test monoclonal antibody and a polyclonal antiserum, polyclonal antibody binding prior to the addition of monoclonal antibody had no effect on the kinetics of monoclonal antibody binding. As expected, when competition between antibodies was observed, prior incubation with a polyclonal antibody frequently slowed the rate at which a subsequently applied monoclonal antibody bound, although in all cases binding appeared to reach equilibrium within four hours at 22°C. Polyclonal antibodies directed against epitopes in the amino and carboxyl termini and each of the three interdomain regions were then examined individually for interactions with the four representative monoclonal antibodies. The binding of aminoterminal monoclonal antibodies A/B2 and L/D3 (residues 1-6 and 19 -24, respectively) (12) was not affected by prior equili-bration of the channel with polyclonal antibody I-31 (residues 31-46) (Fig. 4A). This result was unexpected, since the epitopes for all three antibodies lie within the first 46 amino acids of the rSkM1 amino terminus. A/B2 and L/D3 were then separately screened for competition with polyclonal antibodies directed against specific regions in the interdomain 1-2 (I-467), interdomain 2-3 (B-30), interdomain 3-4 (R-12), and carboxyl terminus (B-23). A/B2 did not interact with any of these sitedirected polyclonal antibodies. For L/D3, however, competition was observed with both B-30 (interdomain 2-3) and B-23 (proximal carboxyl terminus) ( Table II). The amount of monoclonal antibody bound at 4 h of incubation declined to approximately 35% of control as the concentration of polyclonal antiserum used for preincubation increased from 1:100,000 to 1:1.
When monoclonal antibodies against epitopes in the interdomain 2-3 region (F/C11 and B/D6) were examined for competition with the same panel of polyclonal antibodies, a second pattern of interactions was observed. In this case, each of the polyclonal antibodies interfered with the binding of both F/C11 and B/D6 (Fig. 4A and Table II). For each of these monoclonalpolyclonal antibody pairs, competition was incomplete and again increased gradually over a wide range of antibody concentrations, suggesting that antibodies bound to the two epitopes might occupy partially overlapping volumes but that neither antibody completely blocked access to the other epitope.
A third pattern of competition was consistently observed when polyclonal antibody I-1771 (directed against residues 1771-1791 at the distal end of the carboxyl terminus) was assessed with respect to monoclonal antibodies directed against epitopes located at or near the beginning of the amino terminus. I-1771 inhibited the binding of amino-terminal monoclonal antibodies A/B2 and L/D3 with a steep concentration dependence, decreasing monoclonal antibody binding by 80% over a single log unit of polyclonal antiserum dilution (Fig.  4B). This pattern of competition suggests mutually exclusive antibody binding to these two epitopes. This does not represent antibody cross-reactivity between the amino-and carboxylterminal epitopes since we have previously shown that each antibody recognizes only channel fragments containing its cognate epitope (8). A more plausible explanation is that the amino and carboxyl termini lie in close proximity and, perhaps, physically interact.
We tested this hypothesis by examining the interactions between fragments of various channel cytoplasmic segments in a solution phase binding assay. Fusion proteins containing the bacterial MBP joined at its carboxyl-terminal end to various rSkM1 or rSkM2 cytoplasmic segments were prepared. Fusion proteins containing the rSkM1 or rSkM2 carboxyl termini, the rSkM1 interdomain 2-3 region, or MBP alone were immobilized on amylose resin and incubated with a synthetic peptide corresponding to residues 1-30 of the rSkM1 amino acid sequence (1-30 peptide), the region containing the A/B2 and L/D3 epitopes. Bound peptide was identified by subsequent incubation of the peptide-fusion protein-resin complex with either A/B2 or L/D3 followed by 125 I-labeled goat anti-mouse IgG.
We found no specific binding of the 1-30 peptide to fusion proteins containing the rSkM2 carboxyl terminus, the rSkM1 interdomain 2-3, or the MBP alone. However, the peptide bound specifically and with high affinity to the rSkM1 sodium channel carboxyl-terminal fusion protein (Fig. 5). DISCUSSION Based on previous estimates (18), the absence of competition between two antibodies for binding suggests that their epitopes are located more than 3.5 nm apart or are constrained to face in opposite directions. True competitive binding occurs when the epitopes are in close physical proximity. For intermediate sep- FIG. 2. Initial localization of F/C11 and B/D6 epitopes. Western blot depicting nonproteolyzed (A) and limit digests (B) of rSkM1 protein developed with the indicated antibody. Limit digests were obtained by treating 5 pmol of purified sodium channel protein for 120 min with 1-chloro-3-tosylamido-7-amino-2-hepanone-␣-chymotrypsin (0.5 g/ml) at room temperature (8). All antibodies identify the 276-kDa ␣ subunit in the nonproteolyzed samples. The pattern of the limit digest indicates that the epitopes for the two unknown monoclonal antibodies (F/C11 and B/D6) are located on the same limit fragments as the epitope for B-30 (i.e. interdomain 2-3 and domain 3).
FIG. 3. Localization of monoclonal epitopes using fusion proteins. Schematic diagram of the panel of fusion proteins used to localize the epitopes for F/C11 and B/D6. AA refers to amino acid residues in the rSkM1 protein sequence. A "ϩ" indicates that the respective antibody identified the indicated fusion protein in radioimmunoassays and on Western blots, while a "Ϫ" indicates that there was no reactivity of the indicated fusion protein with the respective antibody. arations (ϳ3.5 nm), one antibody may sterically hinder the approach of a second to its epitope while not preventing its ultimate binding. Since conformational changes induced by the binding of one antibody that reduce the affinity of another antibody at a remote epitope appear to occur infrequently (18), we are able to place constraints on the organization of the epitopes under study within the roughly 9-nm diameter envelope of the sodium channel protein (see Barchi (19) for a review of channel physical properties).
For most monoclonal-polyclonal pairs, maximum competition decreased specific binding to ϳ 35% of control values even with the highest concentrations of competing antisera. While this could reflect a measuring artifact resulting from a reduction in k on for the second antibody, this is not the case here since binding rates were determined directly and measurements were made under equilibrium conditions. Several other explanations must be considered. First, a portion of the solubilized sodium channel protein used in the binding assay may be denatured during preparation or storage, resulting in the spatial separation of epitopes which otherwise are close together. Alternatively, sodium channel protein may exist in different conformations, only some of which position the epitopes close to one another. Finally, variable post-translational modification of the sodium channel protein (e.g. phosphorylation) may prevent quantitative monoclonal antibody binding.
It is possible that some of the purified channel protein may be sufficiently denatured to allow separation of epitopes, even though we have shown in the past that most remain functional (16,17). However, since some monoclonal-polyclonal pairs produce greater binding inhibition (Ͼ80%), it is more likely that FIG. 4. Three patterns of antibody competition observed in this study. Polyclonal antibody was first incubated with immobilized sodium channel protein in a solution phase assay at the indicated dilutions of cell culture supernatants. After washing, one of four monoclonal antibodies was incubated with immobilized sodium channel and bound monoclonal antibody quantitated by the binding of iodinated goat anti-mouse IgG. All experiments were repeated twice with three samples per experimental point. Data are reported as the average of three samples with error bars indicating standard deviation. A, amino-terminal polyclonal antibody I-31 sterically hinders the binding of monoclonal antibodies F/C11 and B/D6 to the interdomain 2-3 region (approximately 900 residues distant from the amino terminus) but has no effect on the binding of two monoclonal antibodies (A/B2 and L/D3) whose epitopes are located immediately adjacent in the amino terminus of the channel. These data support a model in which the amino terminus is in a compact, folded, rigid structure and the 2-3 interdomain region is centrally located. B, binding of I-1771 inhibits subsequent A/B2 and L/D3 binding with a steep concentration dependence, suggesting that the near amino-and distal carboxyl termini of the sodium channel are located less than 3.5 nm apart in the native channel.

Summary of antibody competition results
Summary of the presence (ϩ) or absence (Ϫ) of antibody competition between the indicated monoclonal and polyclonal antibodies. All experiments were performed at least in triplicate with three samples per concentration. Competition was deemed to be present when a decrease of at least 50% in specific binding was observed in a pattern of decreasing monoclonal antibody binding with increasing polyclonal antibody titer. AA ϭ amino acid.  -1840), the rSkM1 interdomain 2-3 region (residues 794-1017), the rSkM2/rH1 carboxyl terminus (residues 1791-2018), and the maltose binding protein alone were immobilized on amylose resin and incubated with a synthetic peptide comprising residues 1-30 of the rSkM1 sodium channel amino terminus (I 1-30). This peptide bound specifically and with high affinity only to the rSkM1 carboxyl-terminal fusion protein, providing experimental support for the hypothesis that these two channel segments interact in vivo.
the lower levels of inhibition seen with other pairs simply reflect incomplete block as expected for intermediately separated epitopes where multiple polyclonal antibody molecules with varying affinities may need to bind in order to completely occlude the second epitope. Finally, we have no data that addresses the possible role of alternate channel conformations or variable post-translational modification in this process.
Predictions Concerning the Amino Terminus-Several pieces of data suggest a rigid structure for the amino terminus (Fig.  6). These include: 1) the absence of competition between I-31 and either A/B2 or L/D3 despite the fact that all three epitopes are located within the first 46 residues of the channel sequence; 2) the lack of competition between A/B2 and all site-directed polyclonal antibodies except I-1771 (see below); and 3) the relative resistance of the amino terminus to exogenous proteolysis (8).
Our data support a model in which the first 46 amino acids of the amino terminus forms an arc, with the A/B2 epitope (residues 1-6) facing away from, the L/D3 epitope (residues 19 -24) facing partly toward, and the I-31 epitope (residues 31-46) facing directly toward the centrally located interdomain 2-3 region (Fig. 6). Evidence supporting this hypothesis includes the absence of interaction between A/B2 and all polyclonal antibodies except I-1771, the partial interaction of L/D3 with B-30 (interdomain 2-3) and B-23 (beginning of carboxyl terminus), and the interaction between I-31 and monoclonal antibodies F/C11 and B/D6 (interdomain 2-3 region).
Predictions Concerning the Interdomain 2-3 Region-All of our polyclonal antibodies interfere with monoclonal antibody binding to the interdomain 2-3 region, suggesting that this region is centrally located on the cytoplasmic surface of the channel within a 3-3.5 nm distance from each of the polyclonal epitopes (Fig. 6). In our proteolysis study, the interdomain 2-3 region, while not the largest, was the most accessible of the interdomain regions to exogenous proteolysis using soluble proteases (8), consistent with a model in which the interdomain 2-3 region extends away from the membrane embedded portion of the channel into the cytoplasmic aqueous phase.
We have shown that A/B2 and L/D3 demonstrated competition with each other for binding to immobilized sodium channel protein (11). However, unlabeled F/C11 competed with neither A/B2 nor L/D3 while unlabeled B/D6 demonstrated partial competition with A/B2 and L/D3. No competition was observed between labeled F/C11 and unlabeled B/D6. These data are consistent with our topologic model of the sodium channel amino terminus and interdomain 2-3 regions.
The absence of competition between monoclonals to two adjacent interdomain 2-3 epitopes (F/C11 and B/D6) indicates that these two epitopes are oriented in different directions and, although separated by approximately 100 residues, may have a restricted range of motion. The observation that this region is the most sensitive to proteolysis of the interdomain regions supports a model in which the amino-terminal half of the interdomain 2-3 region extends into the cytoplasm, away from the membrane embedded domains (Fig. 6). The partial inhibition by A/B2 and L/D3 on B/D6 binding, the absence of inhibition by these two monoclonal antibodies on either B-30 or F/C11 binding, and the competition between I-31 and both F/C11 and B/D6 all suggest that the B/D6 epitope (located at the carboxyl-terminal end of the interdomain 2-3 region) is oriented toward and/or is closer to the amino terminus than are the F/C11 or B-30 epitopes (located at or just beyond the midportion of the interdomain 2-3 region) (Fig. 6). Testing these structural hypotheses will require further dissection of cytoplasmic domain topology using competition studies with antibody Fab fragments, physical probes such as fluorescence energy transfer, and direct imaging techniques such as electron diffraction.
Binding of Amino to Carboxyl Terminus-Our data suggest that the channel amino and carboxyl termini are closely interrelated in the tertiary structure of the channel. In order to accommodate this interaction, we suggest a model in which domains 1 and 4 are adjacent in the tertiary structure. This is consistent with previous binding studies which demonstrated competition between ␣ scorpion toxin and site-specific antibodies directed against the S5-S6 regions of domains 1 and 4 in the rat brain sodium channel (20). Although the functional significance of this interaction is unknown, binding between the amino and carboxyl termini could play a role in channel assembly or in stabilization of channel tertiary structure. Another possibility is suggested by our previous finding that A/B2 and L/D3 differentially label channels in the surface and Ttubular membranes of fast and slow skeletal muscle fibers (21,22). It is possible that the rigid surface formed by the interaction of amino-and carboxyl-channel segments creates a binding site for cytoskeletal proteins that contribute to the subcellular localization of channels in different membrane environments (12). FIG. 6. Cartoon depicting our model of sodium channel cytoplasmic domain structure. The figure on the left reflects the view obtained from inside the cell, looking up at the portions of the sodium channel protein which extend into the cytoplasm. The figure on the right represents a side view of the sodium channel protein. Specific points to observe include: (a) the interaction of the carboxyl terminus (lightest shade of gray) with the arch-shaped amino terminus (darkest shade of gray), presenting a face extending away from the bulk of the intracellular mass of the channel; (b) the centrally located interdomain 2-3 region with a fixed orientation to the remainder of the channel's intracellular segments; (c) the relative organization of each antibody epitope in three-dimensional space. See text for details.