Molecular compatibility of the channel gate and the N terminus of S5 segment for voltage-gated channel activity.

Voltage-gated ion channels are modular proteins designed by the structural linkage of a voltage sensor and a pore domain. The functional coupling of these two protein modules is a subject of intense research. A major focus has been directed to decipher the role of the S4-S5 linker and the C-end of the inner pore helix in channel gating. However, the contribution of the cytosolic N terminus of S5 remains elusive. To address this issue, we used a chimeric subunit that linked the voltage sensor of the Shaker channel to the prokaryotic KcsA pore domain (denoted as Shaker-KcsA). This chimera preserved the Shaker sequences at both the N terminus of S5 and the C-end of S6. Chimeric Shaker-KcsA subunits did not form functional homomeric channels but were synthesized, folded, and trafficked to the cell surface, as evidenced by their co-assembly with Shaker wild type subunits. Sequential substitution of Shaker amino acids at the C-end of S6 and the N terminus of S5 by the corresponding KcsA created voltage-sensitive channels with voltage-dependent properties that asymptotically approached those of the wild type Shaker channel. Noteworthy, substitution of the region encompassing Phe(401)-Phe(404) at the N-end of Shaker S5 by KcsA residues resulted in a significant gain in voltage sensitivity of the chimeras. Furthermore, analysis of channel function at high [K(+)](o) revealed that the Phe(401)-Phe(404) region is an important molecular determinant for competent coupling of voltage sensing and pore opening. Taken together, these findings indicate that complete replacement of Shaker S5 and S6 by KcsA M1 and M2 is required for voltage-dependent gating of the prokaryotic channel. In addition, our results imply that the region encompassing Phe(401)-Phe(404) in Shaker is involved in protein-protein interactions with the voltage sensor, and signal to the Phe(401) in the S5 segment as a key molecular determinant to pair the voltage sensor and the pore domain.

Channel proteins constitute a functionally important class of membrane proteins that mediate the transmembrane passage of ions and other small molecules in their thermodynamically favorable direction. The voltage-gated ion channel superfamily, which includes Na ϩ , Ca 2ϩ , and K ϩ channels, shows a very high sequence similarity suggesting a similar molecular architecture. Among this family of channels, voltage-gated K ϩ channels (Kv) 1 are involved in a host of cellular processes from setting the resting membrane potential and shaping the action potential waveform and frequency to controlling synaptic strength (1). A canonical voltage-gated K ϩ channel consists of four ␣-subunits that are assembled to form the ion-permeation pathway across the membrane. In this context, a prototypical K ϩ channel ␣-subunit is formed by an N-and C-terminal domains and six transmembrane helices (S1-S6). The Kv channel family shows a subunit modular organization, constituted by a tetramerization domain, a voltage sensor, and a permeation pathway domain (2).
The structural basis for the channel ionic selectivity has been established in greatest detail for K ϩ channels because of the high resolution structure of the prokaryotic K ϩ -selective KcsA channel (3). The crystallization of other prokaryotic K ϩ channels such as KirBac (4), MthK (5), and KvAP (6) have been useful to recognize regions in the protein structure that are important to gate the channel in response to activating stimuli. In particular, the crystal structure of the KvAP channel from Aeropyrum pernix has provided the first high resolution view of a full-length, voltage-gated K ϩ channel (6). Unexpectedly, the crystal structure of KvAP revealed a structural arrangement for the voltage sensor dramatically different from the conventional models. In the crystal, the voltage sensor is located at the channel cytosolic perimeter, and adopts a "paddle-like" conformation (6,7). In this model, S4 and part of S3 constitute the so-called paddle, which crosses the whole cellular membrane in response to membrane depolarization, thus providing a novel gating mechanism that notably departs from the conventional voltage-sensing models proposed for eukaryotic K ϩ channels (7,8). Nonetheless, the paddle model does not appear to account for functional data on the voltage gating of eukaryotic Kv channels (9 -15).
Although the high resolution structures of prokaryotic K ϩ channels have supported many general features of the pore domains, the functional coupling of the voltage sensor and pore module remains poorly understood. Recently, it was shown as a critical role of the S4 -S5 linker and the C-terminal end of the inner pore helix in channel gating (16 -20). Our and others groups (21)(22)(23) have used a chimeric-based approach to identify amino acid motifs involved in the transduction of the electrical stimuli by the sensor module into a conformational change of the pore module. A chimerical strategy that combined the voltage sensor module of the eukaryotic Shaker channel and the pore domain of the prokaryotic KcsA protein, showed that the chimeric channels are synthesized, folded, and trafficked to the membrane (24). Furthermore, voltage sensitivity and K ϩ selectivity may be provided to the chimeric channels when the C terminus of the eukaryotic protein is present in the chimera, demonstrating that the ion conduction pore is conserved among K ϩ channels (22). Additional studies on the functional coupling of the pore and voltage sensor domains suggested the critical importance of the structural compatibility of the S4-S5 linker and the C-end of the inner pore helix in channel gating (23).
Here, we have further addressed this issue to gain molecular and mechanistic information on the cross-talk between the voltage sensor and pore domains. We constructed a chimera where the S5-P-S6 region of the Shaker channel was replaced by the KcsA M1-M2 counterpart. The prokaryotic channel was inserted between the junction of exons 8 and 9 at the N-end of the S5 domain of the eukaryotic channel, and the junction of the proline-valine-proline (PVP) motif (Fig. 1). Unexpectedly, this chimera, which preserved the Shaker sequences at the N-end of S5 and C-end of S6, was not functional, although channel activity could be rescued when co-expressed with wild type Shaker subunits. Replacement of the eukaryotic amino acids at the S6 by their prokaryotic counterparts produced active channels gated by voltage and/or [K ϩ ] o . Additional swapping of the S5 sequence gave rise to chimeric channels with voltage-dependent gating that approach that characteristic of the Shaker channel, and identify the Phe 401 -Phe 404 region in Shaker as a key molecular determinant for coupling the voltage sensor and the pore domain.

EXPERIMENTAL PROCEDURES
Molecular Biology of Chimeras Design and Mutations-Standard molecular biological techniques were applied as described (25). Shaker inactivating removed (Sh4IR) and Shaker inactivating (Sh4In) were a gift of L. Toro (UCLA) (26), and KcsA was provided by S. Choe (Salk Institute). The Shaker-KcsA was constructed by replacing the region of Sh4IR encompassing from the junction of exons 8 and 9 at the S5 segment to the PVP domain at the S6 segment (amino acids 405 to 472), with that corresponding to KcsA (amino acids 39 to 105). PCR was used to introduce NdeI and KpnI sites in Shaker, respectively, at amino acids 363-364 and 431-432. Two oligonucleotide primers were utilized to amplify KcsA from amino acid 39 through the stop codon while simultaneously introducing a blunt end at amino acid 39, and a KpnI at amino acid 105, which was subsequently used to subclone KcsA into the Shaker/pBluescript construct. Thereafter, the regions S6 ( 473 PVPVIVS 479 ) and S5 ( 398 LLIFFLF 404 ) containing the amino acid sequences of the Shaker channel were stepwise replaced by their KcsA counterparts ( 106 VTAALAT 112 and 31 AATVLLV 37 ) by site-specific mutagenesis (QuikChange Site-directed Mutagenesis Kit, Stratagene). The Ile 405 amino acid is conserved in the two sequences. The sequences of the transferred and mutated segments were verified by both restriction analysis and automated DNA sequencing. For in vitro transcription, chimeric and wild type channel clones were linearized and used as a template using the mMESSAGE mMACHINE kit (Ambion, Austin, TX).
Electrophysiological Recordings-Xenopus oocytes were defolliculated using the calcium-free Barth's solution, collagenase (2 mg/ml), and slow agitation (50 -60 rpm) for 1-2 h. Oocytes were stored at 18°C for 12 h before the injection. In vitro transcribed RNA was injected into Xenopus oocytes (5-10 ng/oocyte) as described (27). Electrophysiological recordings were made 2-4 days after injection. Two electrode voltage clamp was performed using an Electrode Voltage Clamp amplifier (TEC 10CD, NPI Electronic, Tamm, Germany). Throughout the experiments oocytes were continuously perfused with an external solution containing (in mM): 1 MgCl 2 , 0.3 CaCl 2 , 10 Na-HEPES, pH 7.5. According to the type of experiments carried out, 3, 10, 30, and 100 mM KCl was also used, and N-methyl-D-glucamine was added to keep constant the ionic strength. All recordings were obtained at room temperature ϳ22°C.
Electrodes were made with hematocrite glass capillaries and pulled with a P-97 puller (Sutter Ins. Co., Novato, CA); they were filled with 1 M KCl buffered with 10 mM TES and typically had resistance of 300 -500 K⍀. The currents were sampled at 4 -5 kHz after filtering at 1 kHz. Leak subtraction was accomplished with two inverted quarter amplitude pre-pulses that were scaled and subtracted from the test pulse (P/4). The standard voltage protocol consisted of a family of 600-ms long depolarizing pulses from Ϫ120 mV and ϩ120 mV with 20 mV steps, from a holding potential (V h ) of Ϫ80 mV. Shaker data were usually obtained in a similar manner by using a family of 200-ms long depolarizing pulses from Ϫ120 to ϩ120 mV, ⌬V ϭ 20 mV, and V h ϭ Ϫ80 mV. The G-V curves were obtained by converting the maximal current values from the family step stimulations to conductance by using the relation G ϭ I/(V Ϫ E K ), where G is the conductance (S), I (A) is the peak current recording, V is the command pulse potential, and E K is the theoretical K ϩ reversal potential obtained with the Nernst equation considering the [K ϩ ] i of 120 mM (T ϭ 295 K) (28). Conductance values were normalized and fitted to a two-state Boltzmann distribution of the form, where G max is the maximal conductance, V 0.5 is the value of the membrane potential at which 50% of the maximal conductance is reached, and a n is the slope of the G/V curve. C denotes the fraction of the voltage-independent conductance. Data were acquired and analyzed using Pulse/PulseFit 8.11 (HEKA Elektronik, Lambrecht Germany), Origin 7.0 SM0 (OriginLab Corp., Southampton, MA), and ANA (Pusch M., IBF, CNR, Genoa, Italy). Data are shown as mean Ϯ S.D., with n (number of oocytes) Ն 5. Molecular Modeling-The crystallographic structure of the full KvAP protein (Protein Data Bank code 1ORQ) was used as a starting structure for simulations. The structure was edited with Swiss Protein Data Bank viewer 3.7 (29) and Insight II (Biosym/MSI). The structural model was constructed by the regularization of the S3 and S4 ␣-helices and the reorganization of the S1 to S4 helices to form a compact assembly: S1 to S3 were oriented following the crystal structure of the isolated voltage sensor (Ref. 6, Protein Data Bank code 1ORS). The extracellular ends of S1 and S3 are ϳ31 Å from the external opening of the pore, whereas the S3-S4 linker is located Ϸ19 Å from the pore, as indicated by sitedirected cysteine mutation findings and blockade with a series of compounds varying in length (30). Localization of the S4 relative to S5 was tested in terms of energy with FOLDX (foldx.embl.de), and taking into account functional data suggesting that S4 is at the interface between adjacent subunits interacting with S5 (31), the analysis showed better energy values when S4 was located at 40 -45 degrees with respect to the main axis of the pore. This is consistent with the notion that S5 is slightly tilted with respect to the pore axis (31). For this reason, a Յ40 degree tilting angle for S4 produces a strong clash with the C-terminal of the adjacent S5 and/or with the N terminus of the S5 within its subunit. Conversely, a Ͼ45 degree angle for S4 results in a loss of the interaction with adjacent S5. Thus, we favor a location of the S4 segment at the interface between adjacent subunits, and tilted Ϸ45 degrees from the pore axis, thus establishing a direct interaction with the S5 helix of the contiguous monomer. In support to this tenet, a recent model reports that the S3b-S4 voltage sensor paddle is closed to the C-terminal of S5 of the adjacent subunit in the resting state, which is fully consistent with this conformation (32).
The orientation and optimization of the side chains corresponding to this interaction were carried out in two steps: first, those residues making van der Waals clashes were selected and fitted with "Quick and Dirty" algorithms; second, the model was energy minimized (100 steps of steepest descent and 100 of conjugate gradient, cut-off of 10 Å for non-bonded interactions) with Insight II, to relax the backbone and side chains. The model was tested in terms of energy with Fold-X (33), which evaluates the properties of the structure, such as its atomic contact map, the accessibility of its atoms and residues, the backbone dihedral angles, and to the H-bond and electrostatic networks of the protein. In addition, the model was evaluated with PROCHECK (34) showing a Ramachandran plot with 91.3% of the residues in most favored regions, and 8.6% in additional allowed regions. Molecular graphics were created with the program Pymol (www.pymol.org).

Design and Heterologous Expression of a Chimeric Shaker-
KcsA Channel-Our previous chimeric channels containing the full-length KcsA into the background of the voltage-sensing module of mKv1.1 or Sh4IR did not produce functional voltagedependent channels, although they were normally trafficked to the cell surface (24). In contrast, functional channels could be rescued when the C terminus of the Shaker channel was included into the chimera (22). To investigate the protein domain important for endowing voltage-sensing activity to the Shaker-KcsA channels, we designed a chimera by replacing the Ile 405 -Pro 473 region of Shaker with the Ile 38 -Leu 105 segment of KcsA (Fig. 1). The rationale of this chimera considered the insertion of KcsA between the junction of exons 8 and 9 at the N terminus of Shaker S5, and the conservation of the PVP motif at the S6 of the eukaryotic channel. Thus, an important feature of this chimeric subunit is the preserved Shaker amino acids at the N terminus of the S5 segment and the putative residues that structure the gate of the eukaryotic channel at the C-end of S6 ( Fig. 1).
Intriguingly, heterologous expression of the Shaker-KcsA chimera in Xenopus oocytes did not produce voltage-gated ionic currents, neither when 3 nor 100 mM [K ϩ ] o were utilized. We used high [K ϩ ] o in these studies because previous chimeras and mutants exhibited remarkable ionic currents in this condition (22,23). The lack of functional expression of the chimera was not overcome by larger hyper-or depolarizing steps (from Ϫ150 to 180 mV) nor by injecting increasing amounts of cRNA or by co-injecting the Shaker ␣ and ␤ subunits (data not shown). To study if the biogenesis and surface expression of the chimeric subunit was affected, we evaluated its ability to interact with Shaker wild type channel subunits in a co-expression experiment. For this purpose, we used both the Sh4In and the Sh4IR as wild type subunits. Amphibian oocytes were co-injected with chimeric and wild type subunits, and the channel activity of the presumed heteromeric assemblies was recorded. As illustrated in Fig. 2, A and B (insets), co-expression of wild type subunits with chimeric subunits at a ratio of 1:3 (cRNA w/w) notably reduced (Ն80%) the current amplitude of Shaker wild type channels indicating a dominant negative phenotype of the chimera. Note that the presence of the chimeric subunit did not affect the fast-inactivating characteristic of Sh4In subunits ( Fig. 2A, inset). Analysis of the changes in current size of Shaker wild type channels as a function of the fraction of chimeric cRNA co-injected showed a gradual decrease in the voltage-evoked ionic current as the amount of chimeric subunit was augmented (Fig. 2, A and B). The dominant negative phenotype of the chimera suggests that it co-assembles with wild type subunits giving rise to heteromeric channels. However, the dominant negative effect is much milder than that exhibited by our previous chimera (24), suggesting that heteromeric channels may be functional and respond to voltage changes.
To investigate this notion, we co-injected the Sh4In wild type and chimera subunit at a ratio of 1:3 and measured the voltagegated responses. As illustrated in Fig. 2C, in the co-injected oocytes voltage-gated currents were characterized by a voltagedependent fast-inactivating phase followed by a non-inactivating current. The voltage-dependent increase of the peak-tosteady state current ratio shown by homomeric Sh4In was drastically reduced in the heteromer Sh4In:Shaker-KcsA assemblies, primarily because of the increase in the magnitude of the non-inactivating component (Fig. 2D). Because the chimeric subunit was obtained in the Sh4IR background, this result suggests that the non-inactivating, voltage-dependent component corresponds to heteromeric channels that contain the chimera. This notion is further substantiated by the Ϸ50 mV shift toward depolarizing potentials of the conductance to voltage curve (G-V), and the 50% decrease in gating valence of the Sh4IR:Shaker-KcsA heteromers, as compared with homomeric Sh4IR channels (Fig. 2E). Taken together, these findings indicate that chimeric Shaker-KcsA subunits co-assemble with wild type monomers. Thus, the chimera is efficiently synthesized, folded, and trafficked to the plasma membrane. Consequently, the lack of function on homomeric Shaker-KcsA assemblies may arise from uncoupling the voltage sensor and channel gate.
Mutations at the C-end of the S6 Segment in the Shaker-KcsA Subunit Lead to Functional Channel-Comparison of the amino acid sequence of our Shaker-KcsA chimera with that designed by Lu et al. (22) reveals differences at the N terminus of segment S5 and the C-end of segment S6 (Fig. 1). The main difference is that our Shaker-KcsA chimera preserves the sequences of the Shaker channel instead of those of KcsA, suggesting that functional coupling of both modules depends on the molecular compatibility of these subunit regions. Hence, to investigate the molecular requirements of functional coupling, we replaced the Shaker amino acids at both the C-end of S6 and the N-end of S5 of the Shaker-KcsA chimera by those corresponding to KcsA.
First, we focused on the channel intracellular gate, and sequentially mutated the Pro 473 -Ser 479 sequence to Val 106 -Thr 112 in the background of the Shaker-KcsA chimera. Chimeric mutant channels were heterologously expressed in Xenopus oocytes and the voltage-gated activity was recorded at 3 mM [K ϩ ] o (Fig. 3 A, schematic representation of the chimera Shaker-KcsA containing the voltage sensor of Sh4IR and the KcsA pore. The chimera was obtained by replacing the Ile 405 I-Pro 473 region of Shaker with Ile 38 -Leu 105 of KcsA. The chimera conserved the S4-S5 loop, the cytoplasmic part of the outer helix (S5), the C-end of the inner helix (S6), and the cytoplasmic C-terminal end of Sh4IR. B, sequential substitution of Shaker residues at both the N terminus of S5 and C-end of S6 by those corresponding to KcsA, to transfer the complete M1 and M2 segments of the prokaryotic channel to the chimera. conditions tested (data not shown). The lack of channel activity of these mutants was not because of abrogation of protein biosynthesis, as demonstrated by Western immunoblot analy-sis of whole cell extracts (data not shown). However, additional mutation of Val 476 to Ala and Val 478 to Ala gave rise to voltagegated responses that were larger than those obtained from Reduction of the voltage-dependent channel activity of Sh4In (A) and Sh4IR (B) as function of the relative amount of chimera Shaker-KcsA. Insets in panels A and B depict representative current responses of each construct alone and co-injected with Shaker-KcsA at a ratio 1:3 (w/w). Bars are representative of the peak current amplitudes normalized with respect to those obtained from homomeric Shaker channels (n Ն 4). Whole oocyte currents were measured 2 days after injection with a voltage protocol consisting of a square pulse to ϩ50 mV of 300-ms duration from a holding potential (V h ) of Ϫ80 mV. C, co-expression of Sh4In with Shaker-KcsA at a ratio of 1:3 (w/w) generates functional channels that mediate a voltage-dependent, non-inactivating current. Channel activity was elicited from Ϫ80 to ϩ60 mV, with 20 mV potential increments. D, Changes of the peak current (I p )/steady state current (Iss) as a function of the applied potentials of mutants shown in panel C. E, G-V relationships for Sh4IR and Sh4IR with Shaker-KcsA at a ratio of 1:3 (w/w). Conductance changes were obtained from the respective I-V relationships using G ϭ I/(V m Ϫ E K ), where V m is the stimulation potential value and E K is the theoretical reversal potential for K ϩ , calculated with the Nernst relation considering [K ϩ ] int ϭ 120 mM. [K ϩ ] o was 100 mM. Solid lines depict the best fit to a Boltzmann distribution. V 0.5 values were Ϫ20 Ϯ 2 mV for Sh4IR and ϩ31 Ϯ 4 mV for Sh4IR:Shaker-KcsA chimera. The slope values (a n (mV)) of the G-V curve were 14 Ϯ 2 mV for Sh4IR and 29 Ϯ 3 mV for Sh4IR:Shaker-KcsA chimeras. All values are mean Ϯ S.D. with n Ն 5.
non-injected oocytes, although much smaller than currents from homomeric wild type channels (Fig. 3, A and B (left)). The G-V was obtained to study the voltage-dependent gating of this chimera (Fig. 3C). Noteworthy, the channel exhibits a significant conductance at negative potentials, indicating the presence of a strong voltage-independent component. In accordance, the conductance change as a function of the applied voltage can barely be described as a two-state model (V 0.5 ϭ 23 Ϯ 10 mV and a n ϭ 24 Ϯ 9 mV), with a voltage-independent component C of 0.60. Thus, partial restoration of the KcsA M2 segment in the background of the Shaker-KcsA chimeras produce channels that exhibit a rather voltage-independent channel gating.
We next mutated residues Ile 477 to Leu and Ser 479 to Thr to recapitulate the entire M2 segment of KcsA into the chimera background (Fig. 1). These mutations produced a chimeric subunit that also assembled into functional channels (Fig. 3B,  right). The G-V relationship depicts a voltage-dependent change in conductance that can readily be described with a two-state Boltzmann distribution, with a V 0.5 of 82 Ϯ 11 mV, a slope (a n ) of 21 Ϯ 8 mV, and a modest voltage-independent component C of 0.06 (Fig. 3C). Note that the V 0.5 of the chimera is remarkably Ն85 mV more depolarized than that of wild type Shaker channels (V 0.5 ϭ Ϫ6.5 Ϯ 0.3 mV, Fig. 3C). Taken together, these results imply that the C-end of the inner pore helix is primarily involved in defining the energetics of the channel gate. Mutations in this domain create functional K ϩselective channels that either gate in a voltage-independent manner or require high voltages to be activated.
Mutations at the N Terminus of S5 Segment in the Shaker-KcsA Chimera Creates Voltage-dependent Homomeric Channels-Because restoring the KcsA gate region in the chimera background gave rise to channels with low voltage sensitivity, we questioned whether the amino acid sequence at the N terminus of S5 contributes to couple the voltage sensor to the channel gate. To address this issue, we sequentially mutated residues encompassing Leu 398 -Phe 404 to the corresponding amino acids of KcsA (Ala 31 -Val 37 ) in the background of the Shaker-KcsA chimera that contains the KcsA gate (Fig. 1,  bottom). Replacement of Leu 398 and Leu 399 by Ala and Ile 400 to Thr produced a chimera that required high voltage for activation (V 0.5 ϭ 101 Ϯ 20 mV, a n ϭ 26 Ϯ 8 mV, C ϭ 0.02) (Fig. 4, A and C). In marked contrast, subsequent mutation of Phe 401 to Val yielded voltage-activated chimeric subunits whose voltagedependent conductance curve was shifted by Ϸ85 mV toward hyperpolarizing potentials (V 0.5 ϭ 14 Ϯ 8 mV) without affecting the apparent gating valence (a n ϭ 25 Ϯ 6) and with a modest voltage-independent component (C ϭ 0.02) (Fig. 4, A and C). Thus, this result strongly implies an improved coupling between the voltage sensor and the channel gate. Additional mutation of Phe 402 to Leu did not further hyperpolarize the G-V curve (V 0.5 of 24 Ϯ 3 mV, a n of 25 Ϯ 3 mV and C ϭ 0.02) (Fig. 4, B and C). However, replacement of Phe 404 by Val, a change that completely transfers the KcsA M1 sequence, created channels with further hyperpolarized V 0.5 to 6 Ϯ 1 mV, and exhibited an a n ϭ 18 Ϯ 1 mV with a modest voltageindependent component C of 0.10 (Fig. 4, A and C). Taken together, stepwise transference of KcsA M1 residues at the N-end of the chimera S5 segment created voltage-dependent channels with a V 0.5 that remarkably approaches that of the Shaker channel (Fig. 5A), without virtually affecting the gating valence (Fig. 5B). Therefore, these findings illustrate that the complete replacement of Shaker S5 by KcsA M1 is required for efficient voltage-dependent coupling, and indicates that residue Phe 401 is a molecular determinant of competent voltagedependent gating of chimeric channels.
To further examine the relationship between substitutions at the N terminus of S5 and the C-end of S6 and the gating properties, we calculated the free energy difference between the closed and open states at 0 mV (⌬G o ) from V 0.5 and the apparent gating valence (z) of the G-V relationship. Notice that ⌬G o quantity does not account for the voltage-independent conductance because it relies only on the voltage dependence of the channel (20). features: (i) at variance with Sh4IR, all chimeras exhibit ⌬G o Ͼ 0 kcal/mol; (ii) mutation of Phe 401 to Val in the N terminus of S5 reduces the ⌬G o ϳ 1.5 kcal/mol, with no significant energetic gain for the additional substitutions in the S5 N terminus of the chimera. Accordingly, this analysis shows that in the Shaker N terminus of S5 and, in particular, residue Phe 401 are critical molecular determinants for coupling voltage sensing to channel opening.
Extracellular K ϩ Modulates the Channel Activity of Shaker-KcsA Chimeras-Extracellular K ϩ modulates the activity of Kv channels primarily by stabilizing the open state of the channel, as evidenced by the remarkable prolongation of tail current kinetics at high [K ϩ ] o (35). In addition, high [K ϩ ] o favors channel gating of Kv and chimeric channels (22,23,36). Thus, to further understand the functionality of our chimeras we recorded the voltage-dependent ionic currents at 100 mM [K ϩ ] o ( Fig. 6). At variance with non-injected oocytes and Shaker wild type, a raise in the [K ϩ ] o augmented the ionic currents of most of the chimeras. Noteworthy, current to voltage (I-V) relationships at high [K ϩ ] o show that the chimeras did not exhibit the strong outward rectification characteristic of Sh4IR channels (Fig. 6, B and D), consistent with an altered coupling between both protein modules in the chimera. Two of the chimeras S6(VTAAIAS) and S5(AATVFLF) displayed a virtually linear I-V, indicating that these channels equally conduct at negative and positive potentials. Indeed, the ratio of ionic currents elicited at Ϫ80 and 80 mV (I Ϫ80 mV /I 80 mV ) was Ն0. 45. In contrast, the currents of wild type Shaker channels modestly decrease and the ratio I Ϫ80 mv /I 80 mv was unchanged when the [K ϩ ] o was increased to 100 mM. Analysis of the G-V relationships for these chimeras also shows a significant increment in the fraction of the voltage-independent conductance at high [K ϩ ] o (Fig. 7A). Furthermore, conduction through these chimeras could not be abrogated at strong negative potentials suggesting that the energy released by K ϩ binding to the protein suffices to open the channel. Note, however, that K ϩ selectivity was virtually  A and B, values of the V 0.5 (mV) and a n (mV) for the chimeras and Sh4IR, respectively. Values were obtained from the fit of the G-V curves to a two-state Boltzmann distribution. All values are mean Ϯ S.E. with n Ն 5. C, bar graph of the free energy (⌬G o ) for all chimeras and Shaker wild type. The ⌬G o was obtained from ⌬G o ϭ zFV 0.5 , where F is the Faraday constant (0.023 Kcal/mol mV), V 0.5 is the voltage required to activate half-maximal conductance, and z is the apparent gating valence obtained from: z ϭ 25.69 mV/a n , where a n is the slope of the G-V curve. unaffected, as shown by the change in the reversal potential upon increasing the [K ϩ ] o to 100 mM (Fig. 6, B and D).  (Table I)  a Voltage-dependent properties of chimeric channels at 100 mM ͓K ϩ ͔ o . V 0.5 (mV), a n (mV), and C were obtained from the fitting of G-V curves to a Boltzmann distribution. C denotes the fraction of voltage-independent conductance. Data are given as mean Ϯ S.E., with n Ն 6.
b Characteristics of the ͓K ϩ ͔ o -activated ionic currents. Dose-response curves were fitted to a Michaelis-Menten binding isotherm: I ϭ I 0K ϩ (I max /1 ϩ (͓K ϩ ͔ o /EC 50 )), n H , where I max is the maximum current, I 0K is the ͓K ϩ ͔ o -independent current component, EC 50 is the ͓K ϩ ͔ o concentration, which elicits the half-maximal response, and n H is the Hill coefficient. Data are given as mean Ϯ S.E., with n Ն 4. expressed in Xenopus oocytes. To this end, oocytes were held at Ϫ80 mV and bathed with Ringers solution containing 3 mM [K ϩ ] o . As indicated, the solution was changed to Ringers containing 10, 30, and 100 mM [K ϩ ] o (Fig. 8A). Exposure to high [K ϩ ] o was interspersed with washes with 3 mM [K ϩ ] o. Fig. 8A illustrates that oocytes expressing the Shaker channel did not respond to increments in the [K ϩ ] o . In marked contrast, raises in [K ϩ ] o elicited, in a concentration dependent manner, large non-inactivating, inward ionic currents from oocytes expressing chimeras S6(VTAAIAS), S5(AATVFLF), and S5(AA-TVLLF). Analysis of the [K ϩ ] o -activated currents of all chimera revealed that these three species exhibited the largest [K ϩ ] oactivated ionic currents (Fig. 8B). Dose-response relationships for [K ϩ ] o -evoked currents were best described with a sigmoidal curve, which could be fitted to a Michaelis-Menten binding isotherm (Fig. 8C). This analysis depicts a [K ϩ ] o concentration able to activate half-maximal response of ϳ40 mM for all chimeras, and reveals that chimeras mainly differed in the [K ϩ ] o -independent ionic current (I 0K ) and the maximum current evoked by the cation (Table I). Both values were largest for the S6(VTAAIAS) and S5(AATVFLF) chimeras. Thus, these findings imply that mutation of Val 478 and Phe 401 have an important impact on the energetic stability of the closed state or the activation energy of the closed to open state. Taken together, these results indicate that external K ϩ may act as an activating agonist of chimeric subunits, primarily those showing voltage-independent gating activity. DISCUSSION The aim of this study was to gain further insights on the molecular determinant involved in coupling a voltage sensor and a pore domain. Our rationale considered the use of a gain-of-function approach aimed at endowing the voltage-insensitive prokaryotic KcsA channel with voltage dependence, by linking it to the proposed voltage sensor module of the eukaryotic Shaker channel. For this purpose, KcsA was in- serted between the junction of exons 8 and 9 at the N terminus of Shaker S5, and the conserved PVP motif at S6 of the eukaryotic channel (Fig. 1). Thus, the Shaker-KcsA chimera conserved the Shaker residues at both the N-end of S5 and the C-end of S6. Intriguingly, homomeric chimeric channels were not gated by voltage. The lack of channel activity was not because of a defect in protein biogenesis and trafficking, because heteromeric assemblies of Shaker wild type and chimera subunits give rise to voltage-gated channels. Thus, the absence of channel activity of the chimera was because of uncoupling of the voltage sensor and the activation gate. Hence, to determine which were the molecular determinants in both the C-end of S6 and N terminus of S5 that are critical for precise coupling of both the protein modules, we sequentially replaced the Shaker amino acids at both regions by the corresponding KcsA. This strategy gave rise to several findings: (i) the PVP domain present in Shaker S6 does not contribute to voltage sensing, consistent with other reports (37-39); (ii) mutation of the channel gate (Val 478 ) gives rise to voltage-independent, [K ϩ ] o -de-pendent channel activity; (iii) the C-end region of S6 modulates the stability and energetics of the channel gate but does not participate in voltage sensing; and (iv) the N terminus domain of S5 is critical for endowing strong voltage sensitivity to the chimeras. Noteworthy, sequential substitution of Shaker residues at the N-end of S5 by those of KcsA created channels with voltage-dependent gating properties that progressively approached those characteristic of the eukaryotic channel. Furthermore, the most salient contribution of our study is the identification of region Phe 401 -Phe 404 in the N terminus of the S5 segment as a critical structural determinant for providing competent voltage-gated channel activity to the chimeras. Indeed, replacement of Phe 401 by Val notably hyperpolarized the V 0.5 and reduced the ⌬G o for activation, implying that the amino acid at this position is involved in protein-protein interactions with residues of the S4 segment. In addition, analysis of voltage-activated currents at high [K ϩ ] o revealed that Phe 402 and Phe 404 are involved in the fine tuning of voltage gating, as demonstrated by the [K ϩ ] o -dependent hyperpolarization of FIG. 9. A conventional type of structural model of Kv channels is consistent with the interaction of the N terminus of S5 and C-ends of S6 and S4. A, top, crystal structure of the homotetrameric KvAP channel in the closed state. Bottom, enlargement of the S3-S5 region showing that S4 does not interact with S5 in the structure. B, top, structural model constructed by regularization of the S3 and S4 ␣-helices, and the reorganization of the S1-S4 domain to form a compact assembly. Bottom, the S4 ␣-helix is located at the interface between adjacent subunits. Residues at the C-end of S4 and the N terminus of S5 are highlighted. their V 0.5 and a decrease of the variation of free energy for channel opening. The [K ϩ ] o -induced hyperpolarization of the G-V relationship is similar to that observed for Shaker wild type channels, and is consistent with a role of Phe 402 and Phe 404 in protein-protein interactions that contribute to stabilize the tetrameric channel. Therefore, our gain-of-function strategy has unveiled that Phe 401 is a key molecular determinant of the coupling of voltage sensitivity, and that positions Phe 402 and Phe 404 additionally contribute to finely pair the voltage sensor and the pore domains. These results are consistent with observations showing that mutation of Phe 401 and Phe 402 affects the voltage sensitivity of Shaker channels (16 -18), and indicate that the role of the segment encompassing Phe 401 -Phe 404 plays a more important role in voltage sensing than previously recognized. Overall, these results complement those obtained by Lu et al. (23), indicating that the Shaker S4-S5 linker and the sequence around the C-terminal end of S6 must be preserved to obtain functional Shaker-KcsA chimeras. We additionally found that the sequence at the N terminus end of S5 is critical for competent voltage gating.
A central question arises: can these functional data be reconciled with currently established K ϩ channel models of voltage gating? Three classes of models have been considered for the activation of voltage-gated K ϩ channels, the conventional, the "paddle," and the transporter model (6,40,41). Both the conventional and transporter models are structurally similar, differing essentially on how the S4 transmembrane segment moves in response to a change in voltage. In contrast, the paddle model was proposed from the high resolution x-ray structure of the TM domain of KvAP. At variance, with the conventional models, in the KvAP channel the S4-charged helical segment and portions of S3 are forming a paddle that lies at the periphery of the channel, parallel to the intracellular membrane-water interface (6,7). During depolarization, the paddle-like motif moves across the membrane toward the extracellular side, thus triggering channel opening (7). However, the KvAP has failed to accommodate many functional properties of eukaryotic Kv channels, as well as of recent findings (2,9,11,12,15,17,40,(42)(43)(44)(45). Hence, it could be that the KvAP structure may correspond to a distorted nonfunctional channel as suggested by recent structural data from electron paramagnetic resonance spectroscopy (46). Nonetheless, the KvAP structure provides a structural framework that could be used to gain insights on channel gating. Hence, to interpret at a structural level our functional observations, first we used the KvAP crystal structure (Fig. 9A). The KvAP model places the S4 and S5 relatively distant, precluding a direct interaction between both protein segments (Fig. 9A, bottom). This structural arrangement appears inconsistent with our functional observations that suggest a plausible interaction with the S5 and S6 and/or S4. Thus, we next used the KvAP structural information as a scaffold, to create a channel model compatible with our functional observations. The structural model is akin to those reported by other groups for Kv channels (14,15,47), except for the interaction of the C terminus of S4 of one subunit and the N terminus of S5 of the adjacent monomer (Fig. 9B). To accomplish this spatial closeness, the S4 was positioned at the interface between adjacent subunits, and tilted by 45 degrees from the axis of the pore, thus approximating it to the N terminus of S5 (Fig. 9B, bottom). This conformation is consistent with results from structure-function in eukaryotic K ϩ channels (17,31). In this model, the closest proximity between both domains is located around residues corresponding to Phe 401 -Phe 404 in our chimera (Fig. 9B, bottom). Note that Phe 401 is close to Arg 363 in S4 and Phe 402 is near hydrophobic residues Leu 367 and Leu 371 in S4. This structural arrangement is consistent with the strong impact on voltage sensitivity of mutating these amino acids, and with the assigned role of S5 in regulating the final steps in activation that leads to channel opening (16,48). Thus, our functional findings on a hybrid Kv/KcsA channel also question the structural arrangement that emerged from the KvAP crystals, and favor a more conventional type of conformation for Kv channels. Although our structural model may account for most of the functional findings in the eukaryotic Kv channel, it does not favor a particular mode of voltage gating, being consistent with both upward translation of S4 or its tilting within the membrane. Our proposed model is also compatible with a recently proposed mode of voltage gating where the S3b-S4 voltage sensor paddle interacts with the C-end S5 segment of the adjacent subunit in the resting position, and flips across the intersubunit interface to interact with the S5 of its own subunit (31). Accordingly, coupling of the voltage sensor movement to the channel gate depends primarily on the flexibility of the S4-S5 linker and the interactions of the S5 N terminus, consistent with our results. Taken together all these findings suggest the notion that sensitive coupling of the voltage sensor to the pore domain requires a molecular compatibility between the N-and C-terminal regions of the S5 segment. A mismatch in the sequence results in a decrease in channel voltage sensitivity. Future studies will unravel the precise mechanism of how the electrical movement of the paddle induces the mechanical opening of the channel gate.