Subunit Folding and Assembly Steps Are Interspersed during Shaker Potassium Channel Biogenesis*

In the voltage-dependent Shaker K+ channel, distinct regions of the protein form the voltage sensor, contribute to the permeation pathway, and recognize compatible subunits for assembly. To investigate channel biogenesis, we disrupted the formation of these discrete functional domains with mutations, including an amino-terminal deletion, Δ97–196, which is likely to disrupt subunit oligomerization; D316K and K374E, which prevent proper folding of the voltage sensor; and E418K and C462K, which are likely to disrupt pore formation. We determined whether these mutant subunits undergo three previously identified assembly events as follows: 1) tetramerization of Shaker subunits, 2) assembly of Shaker (α) and cytoplasmic β subunits, and 3) association of the amino and carboxyl termini of adjacent Shaker subunits. Δ97–196 subunits failed to establish any of these quaternary interactions. The Δ97–196 deletion also prevented formation of the pore. The other mutant subunits assembled into tetramers and associated with the β subunit but did not establish proximity between the amino and carboxyl termini of adjacent subunits. The results indicate that oligomerization mediated by the amino terminus is required for subsequent pore formation and either precedes or is independent of folding of the voltage sensor. In contrast, the amino and carboxyl termini of adjacent subunits associate late during biogenesis. Because subunits with folding defects oligomerize, we conclude that Shaker channels need not assemble from pre-folded monomers. Furthermore, association with native subunits can weakly promote the proper folding of some mutant subunits, suggesting that steps of folding and assembly alternate during channel biogenesis.

In the voltage-dependent Shaker K ؉ channel, distinct regions of the protein form the voltage sensor, contribute to the permeation pathway, and recognize compatible subunits for assembly. To investigate channel biogenesis, we disrupted the formation of these discrete functional domains with mutations, including an aminoterminal deletion, ⌬97-196, which is likely to disrupt subunit oligomerization; D316K and K374E, which prevent proper folding of the voltage sensor; and E418K and C462K, which are likely to disrupt pore formation. We determined whether these mutant subunits undergo three previously identified assembly events as follows: 1) tetramerization of Shaker subunits, 2) assembly of Shaker (␣) and cytoplasmic ␤ subunits, and 3) association of the amino and carboxyl termini of adjacent Shaker subunits. ⌬97-196 subunits failed to establish any of these quaternary interactions. The ⌬97-196 deletion also prevented formation of the pore. The other mutant subunits assembled into tetramers and associated with the ␤ subunit but did not establish proximity between the amino and carboxyl termini of adjacent subunits. The results indicate that oligomerization mediated by the amino terminus is required for subsequent pore formation and either precedes or is independent of folding of the voltage sensor. In contrast, the amino and carboxyl termini of adjacent subunits associate late during biogenesis. Because subunits with folding defects oligomerize, we conclude that Shaker channels need not assemble from pre-folded monomers. Furthermore, association with native subunits can weakly promote the proper folding of some mutant subunits, suggesting that steps of folding and assembly alternate during channel biogenesis.
The Shaker K ϩ channel is a member of the superfamily of voltage-dependent cation channels that control the excitability of nerve and muscle (1). In response to depolarization of the membrane, Shaker channels open, conduct K ϩ , and then close by a fast inactivation process. During the biogenesis of voltagedependent K ϩ channels such as Shaker, four ␣ subunits assemble to form the aqueous pore for K ϩ permeation (2)(3)(4)(5). The ␣ subunits also associate with cytoplasmic ␤ subunits that regulate channel function (6 -8).
Many of the functional properties of the Shaker channel have been mapped onto the primary structure of the protein by electrophysiological and biochemical analysis of site-directed mutants. These studies suggest that the protein is organized in a modular fashion, with different regions responsible for subunit assembly, activation gating, ion conduction, and fast inactivation (Fig. 1A). A conserved amino-terminal region corresponding to residues 97-196 has been implicated in controlling ␣-␣ and ␣-␤ subunit oligomerization (9 -12). Transmembrane segments S2, S3, and S4, which interact structurally and contain charged residues that sense changes in membrane potential, contribute to the voltage sensor (13)(14)(15)(16)(17)(18). Upon depolarization, the voltage sensor undergoes conformational changes that lead to opening of the K ϩ -selective pore (17). The P region, portions of the S5 and S6 transmembrane segments, and the S4-S5 cytoplasmic loop contribute to the pore, which is located in the center of four Shaker subunits (19 -24). Upon opening, the wild-type channel rapidly inactivates by a ball and chain mechanism, in which an amino-terminal region between residues 6 and 46 blocks the inner mouth of the pore, preventing further conduction (25)(26)(27).
Little is known about the steps of subunit folding and assembly that generate functional K ϩ channels. In this study, we investigate the pathway of channel biogenesis, focusing on assembly of the subunits, folding of the voltage sensor, and formation of the pore.
The Shaker protein is made as an immature, core-glycosylated precursor in the ER, 1 where it folds, acquires hallmarks of the native structure, and assembles into tetramers (8,28). In the ER, Shaker ␣ subunits also assemble with co-expressed, cytoplasmic ␤ subunits (8). Upon transfer to the Golgi apparatus and processing of the oligosaccharide residues, the immature Shaker precursor is converted to the mature product (8,28). The immature and mature forms of the protein can be readily distinguished by their different electrophoretic mobilities. Maturation of the wild-type protein occurs efficiently in such diverse expression systems as Xenopus oocytes and cultured human embryonic kidney (HEK293T) cells (28,29). Maturation is highly correlated with folding and assembly of Shaker subunits into a native conformation (13,18).
A variety of site-directed mutations block maturation of the Shaker protein, trapping it in the ER in an immature form (13,18,30). Such mutations are likely to disrupt proper folding or assembly of the channel protein, making it a substrate for the quality control and retention system of the ER (reviewed in Refs. 31 and 32). Previously, we have used mutations that block maturation to identify likely structural interactions in the Shaker channel (13,18). For instance, the mutant proteins D316K and K374E do not mature or form functional channels. These defects can be rescued by combining the two mutations in one subunit. The double mutant subunit K374E/D316K matures and incorporates efficiently into active channels, indicating that a native structure has been restored (18). The two mutations do not suppress each other if they are located on separate subunits. These results indicate that D316K and K374E disrupt biogenesis by preventing the formation of a specific tertiary interaction between positions 316 and 374, located in the voltage sensor in transmembrane segments S3 and S4, respectively (18).
In addition to providing insights into protein packing, mutations that disrupt folding and assembly provide an opportunity to investigate the pathway of channel biogenesis. Therefore, we have disrupted the formation of different functional domains by mutagenesis and determined the effects of these mutations on subunit assembly. We focused on five mutations that block maturation, located in three different functional domains. A deletion in the amino terminus, ⌬97-196, removes a region previously implicated in subunit recognition and assembly (9,10). The charge reversal mutations D316K and K374E disrupt folding of the voltage sensor, as described above (18). The mutations E418K in the S5-P loop and C462K in S6 are located in the region of the pore and are good candidates to disrupt its formation. These mutants were evaluated for their ability to establish three previously identified quaternary interactions as follows: 1) tetramerization of Shaker ␣ subunits; 2) assembly of Shaker ␣ subunits and auxiliary ␤ subunits (6,8); and 3) association between the amino and carboxyl termini of adjacent ␣ subunits, known to be a structural feature of the native channel (5). The results of this investigation predict a tentative sequence of events during biogenesis and suggest that Shaker channels do not assemble from prefolded monomers. Rather, it is likely that folding and assembly steps alternate during channel biogenesis.

EXPERIMENTAL PROCEDURES
Mutagenesis and Expression of Shaker Protein-Point mutations and ⌬97-196, a deletion of amino acids 97-196, were engineered into a pBluescript II (Stratagene, La Jolla, CA) subclone of the Shaker B cDNA (33) using polymerase chain reaction methods (34,35). Mutated regions were transferred into the wild-type subclone and verified by dideoxy sequencing. The Sh1-246 construct has been described previously (36).
For expression in Xenopus laevis oocytes, cRNA was transcribed using the mMESSAGE mMACHINE RNA transcription kit (Ambion, Austin, TX). Oocytes were prepared and injected as described previously (37). The concentration of cRNA was determined by measuring absorption at 260 nm and confirmed by electrophoresis on denaturing agarose gels with an RNA ladder of known concentration (Life Technologies, Inc.).
For transfection of mammalian cells, wild-type and mutant Shaker cDNA were transferred into a plasmid vector, pcDNA1/AMP (Invitrogen, Carlsbad, CA). The Kv␤2 clone in the pRGB4 vector was the kind gift of Dr. J. Trimmer (38). DNA used for transfections was purified on Qiagen columns (Qiagen, Chatsworth, CA).
Cell Culture and Biochemistry-Human embryonic kidney cells (HEK293T) (39) were grown, transfected as described previously (28), and collected at 36 -48 h post-transfection. For metabolic labeling, transfected cells were incubated for at least 30 min in methionine-and cysteine-free medium and subsequently pulsed for 20 min in 250 Ci/ml [ 35 S]methionine and -cysteine (EasyTag Express Protein Labeling Mix, NEN Life Science Products). Cells were chased in complete media supplemented with 5 mM methionine. Ice-cold lysis buffer (1% Triton X-100, 150 mM NaCl, 50 mM HEPES-NaOH, pH 7.4), supplemented with protease inhibitors as described previously (28), was added directly to the tissue culture plates. The plates were rotated at 4°C for 10 min, and the cells were collected by brief trituration. Cell lysates were centrifuged at 100,000 ϫ g for 30 min at 4°C to remove insoluble material.
Electrophysiology-For electrophysiological analysis, Xenopus oocytes were injected with 0.25-5 ng of cRNA. Whole cell currents were recorded at room temperature (22-25°C) in modified Barth's saline (37), using a two-electrode voltage clamp (Warner Electronics, Hamden, CT). Data were acquired using pCLAMP version 5.5.1 software (Axon Instruments, Foster City, CA) and analyzed using pCLAMP and Microsoft Excel for Macintosh (version 5.0a). Statistical analysis was performed using InStat version 2.0.1 or Microsoft Excel. The significance of dominant negative experiments was determined using the Kruskal-Wallis non-parametric analysis of variance followed by Dunn's multiple comparisons.
Oocyte Biochemistry-For biochemical analysis of protein, each oocyte was injected with 50 -75 ng of cRNA and 500 nCi of [ 35 S]methionine (in vitro translation grade; ICN, Irvine, CA). After 48 h incubation, the cells were lysed using a procedure modified from that described by Hollmann et al. (40). Briefly, oocytes were homogenized in buffer H (1% Triton X-100, 100 mM NaCl, 20 mM Tris-HCl, pH 7.4) supplemented with protease inhibitors and rocked gently for 15 min at 4°C. The majority of yolk proteins were removed by spinning for 2 min at maximum speed in a microcentrifuge, and the remaining insoluble material was removed by centrifugation at 100,000 ϫ g for 30 min at 4°C. Alternatively, oocyte membranes were isolated and solubilized in 2% Lubrol-PX as described previously (29). The protein was then subjected to immunoprecipitation (29). In some cases, oocytes were oxidized with 1 mM iodine prior to lysis as described previously (5).
Cross-linking-Solubilized Shaker protein was incubated with 100 M 3,3Ј-dithiobis(sulfosuccinimidyl-propionate) (DTSSP) (Pierce) for 30 min at room temperature. The reaction was quenched by the addition of concentrated Tris-HCl, pH 8, to a final concentration of 100 mM, followed by incubation for another 30 min. Alternatively, intact Xenopus oocytes expressing Shaker protein were suspended in modified phosphate-buffered saline, pH 7.4 (5), and incubated with 1 mM bismaleimidohexane (BMH) (Pierce), a membrane-permeable cross-linking reagent, for 10 min at room temperature. The reaction was quenched by the addition of 5 mM N-ethylmaleimide, followed by incubation for another 10 min. Immunoprecipitation, Electrophoresis, and Fluorography-Immunoprecipitation of Shaker or Kv␤2 protein was performed as described previously (29) using an antibody directed against the Shaker-␤-galactosidase fusion protein (kind gift of Dr. L. Jan) (41) or a peptide fragment of the Kv␤2 subunit (kind gift of Dr. J. Trimmer) (42). Protein samples were resuspended in Laemmli sample buffer including either 10% 2-mercaptoethanol (reducing conditions) or 16 mM iodoacetamide (non-reducing conditions) and boiled for 3 min prior to electrophoresis on denaturing 5 or 10% polyacrylamide gels with 3 or 4% stacking gels (28). Gels were stained with Coomassie Blue, soaked in fluorographic enhancer, dried, and exposed to film.

Mutations in Three Functional Domains Prevent Shaker
Maturation-Shaker proteins mutated in different functional domains were expressed and metabolically labeled in Xenopus oocytes, immunoprecipitated, and subjected to electrophoresis and fluorography (Fig. 1B). As previously reported (26,30), the Shaker-IR mutant, which contains a deletion of the inactivation ball and chain (residues 6 -46), matures normally and generates functional channels. In contrast, mutations D316K in S3 and K374E in S4 prevent maturation by disrupting the folding of the voltage sensor (18). We have now characterized the maturation of Shaker proteins with mutations in the pore region (E418K in the S5-P loop and C462K in S6) and in the amino-terminal domain implicated in subunit recognition and assembly (⌬97-196). The ⌬97-196, E418K, and C462K proteins did not form functional channels (data not shown) and failed to mature (Fig. 1B). These results indicate that proper formation of the assembly domain, the voltage sensor, and the pore are required for maturation, whereas the inactivation ball and chain can be deleted with no apparent detrimental effects on biogenesis.

Deletion of Amino-terminal Assembly Domain Prevents Oligomerization of Shaker
Subunits-In the ⌬97-196 mutant, the amino-terminal region implicated in controlling subunit assembly has been deleted ( Fig. 1). The ability of ⌬97-196 subunits to oligomerize was assessed in Xenopus oocytes using chemical cross-linking and dominant negative experiments.
The wild-type and ⌬97-196 proteins were solubilized in the detergent Lubrol and incubated with the reagent DTSSP, which reversibly cross-links primary amino groups. Upon treatment of the wild-type protein with DTSSP, three high molecular weight adducts corresponding to a dimer, trimer, and tetramer of Shaker protein were detected under non-reducing conditions ( Fig. 2A, lanes 1). Under reducing conditions, which cleave the DTSSP cross-links, only the monomer band was observed. Upon treatment of the ⌬97-196 protein with DTSSP, no adducts were observed under either reducing or non-reducing conditions ( Fig. 2A, lanes 2). Only the monomer band was detected. Significantly, no aggregated protein was observed at the top of the gel. These results demonstrate that the amino-terminal region between residues 97 and 196 is critical for the formation of detergent-stable tetramers of Shaker subunits. Furthermore, the data indicate that the ⌬97-196 protein does not aggregate nonspecifically in detergent solution.
To determine whether ⌬97-196 subunits associate in the undisrupted environment of the ER membrane, ⌬97-196 was expressed in Xenopus oocytes and treated in situ with the membrane-permeable cross-linking reagent bismaleimidohex- In lane 1, both the mature (upper band) and immature (lower band) forms of the Shaker-IR protein are visible. For other mutant proteins, only the immature form of the protein is detected. Compared with the other mutants, the immature forms of Shaker-IR and ⌬97-196 migrate with increased mobility due to the deletion of 41 and 100 amino acid residues, respectively. The wild-type and mutant proteins accumulate to different steady state levels, which may reflect differences in turnover kinetics.
FIG. 2. ⌬97-196 subunits fail to oligomerize. A, solubilized wildtype and ⌬97-196 protein was subjected to cross-linking with DTSSP, followed by immunoprecipitation, electrophoresis, and fluorography. Wild-type protein formed three high molecular weight adducts that were present under non-reducing (Non-Red) but not reducing (Red) conditions (lanes 1). Apparent molecular masses: monomer (1x), 110 kDa; dimer (2x), 215 kDa; trimer (3x), 285 kDa; and tetramer (4x), 335 kDa. No adducts were detected upon cross-linking of ⌬97-196 protein (lanes 2). Apparent molecular mass: monomer of ⌬97-196 (denoted by ⌬), 78 kDa. A representative experiment is shown, n ϭ 2. B, oocytes expressing the N259Q/N263Q (lanes 1) or ⌬97-196 (lanes 2) proteins were incubated with BMH or with vehicle (DMSO) alone. High molecular weight oligomers were detected for N259Q/N263Q subunits but not for ⌬97-196 subunits. Although BMH is often used as a sulfhydrylspecific cross-linker, we have observed that this reagent reacts at pH 7.4 with both sulfhydryl and amino groups in the Shaker protein. 2 C, Shaker-IR and ⌬97-196 subunits were co-expressed in the indicated molar ratios. Currents were evoked by pulsing from a holding potential of Ϫ80 mV to ϩ40 mV for 100 ms. Current amplitudes at ϩ40 mV were determined and normalized to the control value obtained by injection of Shaker-IR alone. Histogram bars show the mean Ϯ S.E., n ϭ 5-10 for each injection ratio. The Shaker-IR control bars show a normalized S.E. as an indication of the variability in control measurements. A nonparametric analysis of variance (Kruskal-Wallis) demonstrated no statistical significant difference between the injection ratios (p ϭ 0.837).
ane (BMH). The mutant N259Q/N263Q, which makes unglycosylated Shaker subunits that assemble into active cell-surface channels (29), was used as a control. Following immunoprecipitation and electrophoresis, high molecular weight adducts of N259Q/N263Q subunits were observed (Fig. 2B, lanes 1). In contrast, no adducts or aggregates of ⌬97-196 subunits were detected (Fig. 2B, lanes 2). These results suggest that ⌬97-196 subunits do not associate or aggregate in the ER membrane.
To investigate whether ⌬97-196 subunits can associate with native Shaker subunits in vivo, we used a dominant negative strategy. It has been shown previously that fragments of Shaker protein that include the putative amino-terminal assembly domain suppress the functional activity of co-expressed, native, full-length subunits (9,36). This is because the aminoterminal fragment associates with the intact Shaker protein, preventing its incorporation into active cell-surface channels.
cRNA encoding the Shaker-IR subunit, which forms functional channels lacking N-type (fast) inactivation (25), was injected alone or was co-injected with cRNA encoding the ⌬97-196 subunit in molar ratios ranging from 1:1 to 1:10. Shaker-IR current amplitudes in the absence or presence of co-expressed ⌬97-196 subunits were determined at ϩ40 mV using a twoelectrode voltage clamp. No dominant negative effect was observed over the entire range of expression ratios (Fig. 2C). These results demonstrate that the ⌬97-196 subunit is unable to oligomerize with native Shaker subunits. We conclude that the ⌬97-196 mutation prevents the assembly of Shaker subunits.
Amino-terminal Assembly Domain Is Required for Pore Formation-As demonstrated above, the interaction between amino-terminal assembly domains is critical for the oligomerization of Shaker subunits. In the native channel, however, additional intersubunit interactions must occur. For instance, a subunit is obliged to contact its neighbors around the K ϩselective, aqueous pore. To determine whether subunit interactions around the pore can be detected in ⌬97-196, we took advantage of the M448C mutation. Met-448 is located in a narrow part of the P region, at or near the center of symmetry of the pore (Fig. 1A) (43). Upon substitution with cysteine, an intersubunit disulfide bond between two 448 residues forms to generate a covalent dimer of Shaker subunits (43). As shown in Fig. 3, this reaction occurs spontaneously whether the M448C protein is located on the cell surface or is retained in the ER by treatment with brefeldin A plus nocodazole.
The M448C mutation was inserted into the ⌬97-196 subunit, and the construct was expressed in Xenopus oocytes. As a control, the M448C mutation was also inserted into the unglycosylated mutant N259Q/N263Q (29). Intact cells were treated with N-ethylmaleimide to protect free sulfhydryl groups prior to lysis and solubilization. The ⌬97-196/M448C and N259Q/ N263Q/M448C proteins were immunoprecipitated and subjected to electrophoresis under reducing and non-reducing conditions (Fig. 3). In contrast to the N259Q/N263Q/M448C protein, no disulfide-linked dimers of ⌬97-196/M448C protein were detected. These results indicate that prior oligomerization of the 97-196 domain is required for pore formation and suggest that other subunit interactions which must occur in the native state, including those around the pore, are insufficient to drive stable assembly of the pore.
Voltage Sensor and Pore Region Mutant Subunits Form Detergent-stable Tetramers-Mutations D316K and K374E prevent proper folding of the voltage sensor, whereas mutations E418K and C462K are likely to disrupt pore formation. To determine whether these mutant subunits oligomerize, they were solubilized from oocyte membranes with Lubrol and incubated with the reversible cross-linking reagent DTSSP. In addition to the immature monomer, three adducts were observed under non-reducing conditions (Fig. 4, lanes 2-5). No larger adducts were detected. The adducts corresponded to a dimer, trimer, and tetramer of immature mutant subunits, as verified by DTSSP cross-linking of the unglycosylated Shaker mutant, N259Q/N263Q. The N259Q/N263Q monomer and adducts (Fig. 4, lane 1) migrated slightly faster than the coreglycosylated monomers and adducts formed by the D316K, K374E, E418K, and C462K subunits. The small differences in mobility reflect the lack of carbohydrate residues on the N259Q/N263Q protein (29). These results indicate that nonmaturing subunits with mutations in the voltage sensor and the pore region form detergent-stable tetramers. Significantly, these subunits do not form larger aggregates in detergent solution.
Association of Voltage Sensor or Pore Region Mutant Subunits Requires Assembly Domain-To investigate subunit interactions of voltage sensor and pore mutants in vivo, D316K, K374E, E418K, or C462K were co-expressed with Shaker-IR for dominant negative experiments. In the presence of each of the mutants, the current amplitude was significantly suppressed (Fig. 5). To determine whether this dominant negative effect was mediated by the amino-terminal assembly domain, the ⌬97-196 deletion was combined with each of the point mutations, and the resulting subunits were co-expressed with Shaker-IR. The dominant negative effect was eliminated upon deletion of amino acids 97-196 in each of the mutant constructs (Fig. 5). These results indicate that mutations in the voltage sensor and pore regions do not prevent the assembly of Shaker subunits and, furthermore, that assembly is mediated specifically by the amino-terminal domain between residues 97 and 196. The data are incompatible with the idea that associations between the immature mutants and Shaker-IR are due to nonspecific aggregation.
Assembly with Native Subunits Can Promote Cell-surface Expression of Some Mutant Subunits-Association with voltage sensor and pore region mutant subunits traps Shaker-IR subunits in the ER, leading to the dominant negative effect described above. To determine whether, conversely, association with Shaker-IR subunits can promote the cell-surface expression of voltage sensor and pore region mutant subunits, we used an inactivation tagging strategy (2,18). The voltage sensor and pore region mutants contain the inactivation ball and chain located at the amino terminus of the protein (Fig. 1A) (25,26). Co-assembly of mutant and Shaker-IR subunits into functional, cell-surface channels would be indicated by the presence of an inactivating component of current superimposed on the non-inactivating current characteristic of Shaker-IR channels. Because one ball and chain per tetrameric channel is sufficient to produce N-type inactivation (44), inactivation tagging provides a sensitive way to detect the presence of the mutant subunits in active channels.
Examples of current traces obtained following co-expression of Shaker-IR and mutant subunits are shown in Fig. 6. No evidence of incorporation was obtained for C462K subunits, consistent with the strong dominant negative effect of this mutation. Thus, even a single C462K subunit prevented formation of a native structure. In contrast, a small inactivating component of current was detected upon co-expression of D316K, K374E, or E418K subunits with Shaker-IR (Fig. 6 or data not shown), indicating that mutant subunits were occasionally incorporated into functional channels with Shaker-IR subunits. Incorporation into active channels occurred at a low efficiency, as evidenced by the dominant negative effect on peak current amplitude apparent in the same traces.
For comparison, Shaker-IR subunits were also co-expressed with wild-type subunits or with a truncated amino-terminal fragment corresponding to residues 1-246 (Sh1-246) (Fig. 6). The random association of wild-type and IR subunits expressed at a 1:1 ratio was highly efficient, leading to a much larger inactivating component than co-expression of IR and mutant subunits. Co-expression with the amino-terminal fragment had a strong dominant negative effect on Shaker-IR expression but produced no inactivating current at any injection ratio (Fig. 6) (36). This result is significant because application of a short peptide containing the inactivation ball conferred inactivation on Shaker-IR channels in an excised patch of membrane (26). In our experiments, however, the larger Sh1-246 fragment did not incorporate into functional channels or act as a soluble FIG. 6. Incorporation of mutant subunits into Shaker-IR channels. Shaker-IR subunits were co-expressed at ratios of 1:1, 1:5, and 1:10 as indicated with wild-type, D316K, or C462K subunits, or with an amino-terminal fragment, Sh1-246, which contains residues 1-246 of Shaker plus 8 additional amino acid residues (36). For co-expression of Shaker-IR with D316K or C462K, representative current traces from the experiments shown in Fig. 5 are shown. Currents were elicited using a two-electrode voltage clamp by pulsing for 100 ms from Ϫ80 mV to potentials from Ϫ60 to ϩ80 mV in 20 mV increments. The current amplitude scale bars on the right represent 5 A.
The inactivation ball and chain is intact in ⌬97-196 subunits, but upon co-expression of IR and ⌬97-196 subunits, no inactivating component of current was detected (data not shown). This result is consistent with the conclusion that the 97-196 deletion prevents subunit assembly.
Voltage Sensor and Pore Region Mutant Subunits Associate with Cytoplasmic ␤ Subunit-K ϩ channel ␣ subunits associate with cytoplasmic ␤ subunits that modulate the functional properties of the channel (reviewed in Ref. 45). Assembly with ␤ subunits is mediated by residues corresponding to 178 -187 in the Shaker amino terminus (11,12). The ability of mutant subunits to associate with the rat Kv␤2 subunit was assessed by reciprocal co-immunoprecipitation experiments.
Wild-type or mutant Shaker protein was expressed in HEK293T cells at a 3:1 ratio with the ␤ subunit. Following metabolic labeling and cell lysis, the samples were divided and precipitated with antibodies directed against Shaker or Kv␤2. The wild-type Shaker protein and the D316K, K374E, E418K, and C462K mutants were detected upon precipitation with either Shaker or Kv␤2 antibodies (Fig. 7). Similarly, the Kv␤2 protein, with an apparent molecular mass of 39 kDa, could also be immunoprecipitated with either antibody (Fig. 7). In contrast, the ⌬97-196 protein, in which the residues critical for ␤ subunit association have been deleted (11,12), was not immunoprecipitated by Kv␤2 antibodies (Fig. 7).
To determine whether nonspecific interactions between Shaker and ␤ subunits occurred during solubilization and immunoprecipitation, cells were transfected with either wild-type or Kv␤2 cDNA, metabolically labeled, and separately lysed. The cell lysates were pooled prior to immunoprecipitation. Under these conditions, Shaker protein was precipitated only with Shaker antibodies, and in the reciprocal experiment, the ␤ subunit was precipitated only with the Kv␤2 antibody (Fig. 7). These results demonstrate that Shaker subunits with mutations in the voltage sensor or pore region assemble specifically with ␤ subunits and that the association is mediated by residues in the Shaker amino terminus.
Non-maturing Mutants Lack Key Structural Hallmark of the Native State-Although Shaker subunits are not normally linked by disulfide bonds, two cytoplasmic cysteine residues, Cys-96 in the amino terminus and Cys-505 in the carboxyl terminus, form an intersubunit disulfide bond in intact cells exposed to oxidizing conditions (5). Oxidation results in a characteristic pattern of adducts representing a dimer, trimer, linear tetramer, and circular tetramer of Shaker protein (5). Oxidation does not alter the functional properties of the channel, indicating that cysteines 96 and 505 are in close proximity in the native state (5). Therefore, disulfide bond formation provides an assay for the native state of the protein. The 96/505 disulfide bond can be catalyzed in wild-type Shaker protein located in the ER, where wild-type Shaker channels fold and assemble (8).
To determine whether mutant subunits form the 96/505 disulfide bond, ⌬97-196, D316K, K374E, E418K, and C462K subunits were expressed, metabolically labeled, and oxidized with iodine in Xenopus oocytes (Fig. 8). Positive and negative controls were provided by the Shaker wild-type and C96S subunits, respectively. As previously reported, the C96S mutation prevents disulfide bond formation (5). The high molecular weight adducts characteristic of the 96/505 reaction were not detected upon iodine treatment of the single mutants D316K and K374E, which prevent proper folding of the voltage sensor (Fig. 8A, lanes 3 and 4). However, when the folding defect was rescued in the double mutant combination K374E/D316K (18), formation of the 96/505 disulfide bond was also restored (Fig.  8A, lane 5), resulting in the four high molecular weight adducts characteristic of the 96/505 disulfide bond (5). The mutations ⌬97-196, E418K, and C462K also prevented the oxidation of Cys-96 and Cys-505 (Fig. 8B, lanes 2-4). These results indicate that mutations in the assembly domain, the voltage sensor, and the pore region prevent establishment of the proximity between Cys-96 and Cys-505 in adjacent subunits. Because this proximity is a structural hallmark of the native state, these results also confirm that the mutant proteins fail to adopt a native conformation.

DISCUSSION
The Pathway of Shaker Channel Biogenesis-Our results suggest that Shaker channel biogenesis can be divided into FIG. 7. Interactions of wild-type and mutant Shaker subunits with Kv␤2. Wild-type and mutant Shaker subunits and the Kv␤2 protein were co-expressed and metabolically labeled in HEK293T cells. Following cell lysis, the samples were divided and precipitated with antibodies directed against Shaker (anti-␣) or Kv␤2 (anti-␤). Immunoprecipitated samples were subjected to electrophoresis and fluorography. Lane 1, wild-type Shaker and Kv␤2 proteins were expressed in separate cells, and lysates were pooled prior to immunoprecipitation. several phases (Fig. 9). Membrane insertion, glycosylation, and subunit oligomerization occur early in the process. Intermediate events include formation of the voltage sensor and the pore. Late during biogenesis, the amino and carboxyl termini of adjacent subunits come into proximity, establishing an interaction that persists in the native structure.
All of the non-native Shaker proteins were efficiently targeted and inserted into the ER membrane as evidenced by their uniform core glycosylation. 2 For the voltage sensor and pore mutants, dominant negative experiments indicate that the amino terminus is correctly located in the cytoplasm, where it mediates assembly with Shaker-IR subunits. We conclude that the mutant proteins adopt the correct membrane topology at least from the amino terminus through the S1-S2 loop. In the ER, the mutant proteins are substrates for glucose trimming and interact transiently with the chaperone calnexin. 2 Core glycosylation is known to occur co-translationally, and glucose trimming and calnexin association are also early events in the biogenesis of other membrane proteins (46). These events are likely to occur early during the process of Shaker biogenesis and do not require prior folding or assembly of the subunits.
Our data indicate that prefolded monomers are not required for oligomerization of Shaker subunits. In particular, proper formation of tertiary structural interactions in the voltage sensor is not required for the formation of Shaker tetramers or for the assembly of Shaker and ␤ subunits. During the biogenesis of the wild-type channel, subunit assembly mediated by the amino-terminal domain could precede folding steps such as voltage sensor formation or, alternatively, assembly and folding of the voltage sensor could occur independently.
Oligomerization of Shaker subunits mediated by the aminoterminal domain is required for formation of the pore. The P region mutation M448C, which leads to the spontaneous oxidation of intersubunit disulfide bonds, was used to assay for pore formation in ⌬97-196 subunits. No subunit interactions around the pore were detected in ⌬97-196 subunits. The M448C reaction was not used to determine whether pore formation occurs in subunits that disrupt folding of the voltage sensor. This is because the spontaneous M448C reaction can occur between subunits held in close proximity in tetramers by the amino-terminal assembly domain during the long duration of our biochemical experiments (data not shown). For this reason, it does not provide a reliable assay for pore formation in the voltage sensor mutants.
Folding of the voltage sensor and formation of the pore are likely to be intermediate events in biogenesis. Subunits with mutations in the voltage sensor and pore region oligomerize but do not establish proximity between adjacent amino and carboxyl termini. Additional experiments will be required to dissect the relationship between folding of the voltage sensor and formation of the pore.
The establishment of proximity between the amino and carboxyl termini of adjacent subunits is disrupted by all of the mutations we have analyzed, suggesting that this structural hallmark of the native state is formed late during biogenesis. In the wild-type protein, disulfide bond formation between Cys-96 and Cys-505 can be detected within 10 -20 min of the start of translation, while the protein still resides in the ER (8). 2 Relationship between Folding and Assembly of Shaker Subunits-The observation that voltage sensor mutations do not prevent tetramerization of Shaker subunits or assembly with ␤ subunits indicates that assembly does not require prefolded monomers. Whether tetramerization is required for proper folding of the voltage sensor has not been addressed by our experiments. Our results indicate that assembly and voltage sensor formation are either sequential or occur independently. Of these two possibilities, the data are more consistent with the idea that assembly and folding steps alternate during the biogenesis of Shaker channels. We have observed that folding mutants that are unable to form active homotetramers can co-assemble, albeit very inefficiently, with IR subunits into functional heterotetramers. This implies that association with native IR subunits occasionally promotes proper folding of some of the mutant subunits. It seems likely that this effect requires assembly of IR and mutant subunits before folding of the individual subunits is complete. It is worth noting that the low level of incorporation of mutant subunits into otherwise IR channels differs significantly from rescue by second site mutations, which is extremely efficient (13,18).
Folding and assembly steps are interspersed during the biogenesis of the nicotinic acetylcholine receptor, another channelforming, multi-subunit membrane protein (47). In this case, formation of specific assembly intermediates precedes the acquisition of toxin-binding sites and antibody epitopes that are FIG. 9. Tentative sequence of events during Shaker biogenesis in the ER. None of the mutations analyzed block membrane insertion or core glycosylation, which are likely to occur co-translationally. Analysis of mutant proteins indicates that subsequent steps of Shaker biogenesis can be divided into early, intermediate, and late stages. The ⌬97-196 mutation blocks the pathway at an early stage, preventing subunit assembly, pore formation, and the generation of the native structure (see Figs. 2, 3, and 8). In contrast, D316K, K374E, E418K, and C462K subunits assemble but block pore formation and/or folding of the voltage sensor and the subsequent generation of the native state (see . The data do not resolve the relationship between folding of the voltage sensor and formation of the pore. Furthermore, tetramerization mediated by the amino-terminal region could either precede or occur independently of voltage sensor formation. present in folded but not unfolded subunits. It has been reported that the Kv␤2 subunit increases the efficiency of maturation of rat Kv1.2 subunits, which are homologous to Shaker subunits (48). We have previously shown that co-expression of the Kv␤2 and Shaker proteins does not increase the efficiency or rate of Shaker maturation (8). Our data indicate that assembly with Kv␤2 likewise did not promote the maturation of the non-native mutant subunits used in our experiments.
Role of the Amino-terminal Assembly Domain in Biogenesis-Formation of functional channels comprising Shaker, mouse Kv1.1, or Aplysia Kv1.1a subunits requires the aminoterminal assembly domain (10,49,50). In contrast, aminoterminal deletions in rat Kv1.3, human Kv1.4, or in the Shab family member, Kv2.1, do not prevent functional assembly (51)(52)(53). Interactions between transmembrane segments are likely to promote the oligomerization of Kv1.3 subunits lacking the assembly domain (53,54). In contrast, we found no evidence that such interactions suffice for the assembly of Shaker subunits, although in the native channel, subunit interactions must occur within the plane of the membrane to account for ion permeation.
Subunits of the nicotinic acetylcholine receptor also contain amino-terminal domains that mediate subunit recognition and assembly (55). In both K ϩ channel and acetylcholine receptor subunits, the assembly domain emerges from the ribosome before any transmembrane segments, the assembly domains oligomerize with high affinity, and assembled subunits can be detected as soon as protein synthesis is completed (47,55,56). From these findings, it has been proposed that assembly of K ϩ channels and acetylcholine receptors may occur co-translationally (47,56). Our findings support the conclusion that tetramerization of Shaker subunits occurs early during biogenesis. It is unknown, however, whether nascent channel subunits are able to assemble before release from the ribosome. Current understanding of the mechanism of co-translational insertion and integration of membrane proteins would not account for the coordinated insertion of assembled subunits attached to different ribosomes (57). It is possible that co-translational assembly is prevented by specific chaperone molecules or by a low frequency of collisions between nascent chains.