Report

Molecular Architecture of the KvAP Voltage-Dependent K+ Channel in a Lipid Bilayer

See allHide authors and affiliations

Science  15 Oct 2004:
Vol. 306, Issue 5695, pp. 491-495
DOI: 10.1126/science.1101373

Abstract

We have analyzed the local structure and dynamics of the prokaryotic voltage-dependent K+ channel (KvAP) at 0 millivolts, using site-directed spin labeling and electron paramagnetic resonance spectroscopy. We show that the S4 segment is located at the protein/lipid interface, with most of its charges protected from the lipid environment. Structurally, S4 is highly dynamic and is separated into two short helices by a flexible linker. Accessibility and dynamics data indicate that the S1 segment is surrounded by other parts of the protein. We propose that S1 is at the contact interface between the voltage-sensing and pore domains. These results establish the general principles of voltage-dependent channel structure in a biological membrane.

Voltage-dependent channels are composed of two functionally linked but structurally independent domains (14). The pore domain is responsible for ion selectivity and contains the channel gate, whereas a voltage-sensing domain (segments S1 to S4) alters the conformation of the gate in response to changes in transmembrane voltage. Crystal structures of three different prokaryotic K+ channels have elegantly demonstrated the common architecture of the pore domain (57), as well as the basic principles underlying permeation and selectivity for K+ ions (8, 9). Furthermore, they point to at least one plausible mechanism for the opening of the intracellular gate (10, 11). There is much less agreement in relation to the structure and conformations of the voltage-sensor domain.

Voltage sensing is linked to structural rearrangements of the S4 segment, a transmembrane helix containing positive charges every three residues. Based on functional and indirect structural analyses, the general consensus has been that the S4 segment and its charges must be isolated from the low dielectric of the membrane by a shield of protein. In one explicit model, this shield surrounds the S4 segment, which moves across an aqueous “gating pore” or “canaliculi” in response to changes in the electric field (1214). However, structural and mechanistic models derived from the recent crystal structures of KvAP and its isolated voltage sensor (15) appear to run contrary to these concepts. Obtained as complexes with an Fab antibody fragment, the new structures led to the suggestion that S4 and parts of S3 form a stable hairpin (the “paddle”) that is located at the periphery of the channel, exposed to the membrane lipid (15). This multicharged hairpin would act as a hydrophobic cation moving across the membrane and pulling on the activation gate, thus opening the channel (16).

To evaluate these seemingly incompatible models, we have used site-directed spin labeling and electron paramagnetic resonance (EPR) spectroscopy (1719) to measure the structural dynamics of reconstituted KvAP. Experiments were carried out on a set of mutants comprising the entire voltage-sensing domain (20) under conditions that promote a deep inactivated state, in which the voltage sensor is presumably in a conformation similar to that in the open state. Our data were evaluated within the framework of four distinct structural references: the KvAP crystal structure (15), the structure of the KvAP isolated voltage sensor (15), a structural model (21) based on the open KvAP paddle model (16), and models with “canonical” voltage-sensor arrangements (22, 23).

The fully processed EPR data set for the whole sensor domain is shown in Fig. 1 (24). Figure 1A shows two representative sets of spectra illustrating the overall distribution of line shapes for regions of S1 and S4 (including all charged positions). The spectra revealed populations of mixed dynamics (particularly in the less mobile positions in S4), suggesting that the sensor as a whole is highly flexible when compared with the overall rigidity of the pore domain as observed in KcsA (25, 26). Figure 1B shows the residue environmental parameter profiles for probe mobility Embedded Image (black circles), O2 accessibility ΠO2 (red squares), and Ni++ chelate complex (NiEdda) accessibility ΠNiEdda (blue triangles) from reconstituted full-length KvAP. We found clear, well-defined boundaries to the individual transmembrane (TM) segments, both in terms of probe dynamics and exposure to the aqueous environment in the connecting loops (Fig. 1B, top and bottom). Loops tend to be highly mobile and readily accessible to NiEdda, suggesting an unambiguous transmembrane orientation for all TM segments. This topological arrangement is incompatible with the disposition of both S1 and S2 segments in the crystal structure (Fig. 1C), as well as with that proposed for S2 in the structural model of open KvAP (15, 16).

Fig. 1.

Voltage-sensor environmental data set. (A) EPR spectra of spin-labeled mutants from selected regions of S1 and S4 segments. Spectra represent 100 G scans. Dotted lines mark the location of restricted (blue) and highly dynamic motional components of the spectra. Arrowheads point to the small amounts of residual free spin label. (B) Environmental parameter profiles: mobility parameter ΔH0–1 (black circles), oxygen accessibility parameter ΠO2 (red squares), and NiEdda accessibility parameter ΠNiEdda (blue triangles). Grayed areas represent assignments derived from the KvAP crystal structure. The dotted line is the average mobility for the entire segment. Arrows in the S4 segment point to residues that display a loss in α-helical periodicity. Gating charges in S4 are highlighted by circles. (C) Environmental parameters (ΔH0–1, ΠO2, and ΠNiEdda) mapped onto a molecular surface rendering of the full-length KvAP crystal structure. (Top) Ribbon representation of the full-length KvAP crystal structure (two subunits are shown for clarity). Individual transmembrane segments are color coded as follows: S1, red; S2, yellow; S3, green; and S4, blue. Black spheres show the location of cysteine mutants for which we were unable to obtain data. Molecular surface and data mapping was done using Chimera (35, 36).

In the bilayer-embedded regions, average motional dynamics gradually increases from S1 through S3 to S4, with a somewhat parallel increase in O2 accessibility, in which S4 clearly emerges as the TM segment with the largest O2 accessibility. In S4, the α-helical periodicity in Embedded Image and ΠO2 is interrupted around residues 128 to 130, dividing the segment into two smaller helices (S4a and S4b) joined by a somewhat flexible linker (Fig. 1B, black arrows). In addition, we find that a large number of residues in S3 are exposed to collisions with NiEdda (Fig. 1B, bottom), although they are well within the confines of the putative transmembrane region, according to the structure of the isolated sensor.

Although there is widespread agreement that the full-length KvAP structure is distorted by the crystallization conditions, mapping the EPR-determined environmental parameters directly onto it provides a structural context to evaluate the degree of distortion (Fig. 1C). Local dynamics do not reveal many gross discrepancies with our experimental results (with the exception of S1), yet it is clear that among the regions most exposed to O2 (S3 and S4 and parts of S2), few appear to be within the expected bilayer boundaries, according to the location of the pore domain. The situation is precisely the opposite for positions exposed to NiEdda, where the loops and portions of S3 appear almost in the center of the bilayer.

Mapping the same data set onto the structure of the isolated sensor domain (Fig. 2) revealed a better spatial correlation, which suggests that the structure of the isolated sensor is a reasonable representation of its conformation as part of the full-length channel. The water/lipid interface is well defined from the NiEdda map and comprises crescent-shaped surfaces in both ends of the sensor domain, along its long axis. As expected, these are populated by residues from the N terminus, the S1-S2 loop, the S2-S3 loop, and S3b and the extended C-terminal end of S4. Mapped O2 and mobility data also help position the sensor relative to the lipid and the pore domain. S1 and portions of S2 outline a motionally restricted, low O2 accessibility area that represents the likely docking surface of the sensor to the pore domain. At the opposite side, S4 and S3a form an X-shaped, highly dynamic, and O2-accessible area expected to be at the protein/lipid interface.

Fig. 2.

EPR data mapped onto the isolated voltage-sensor structure. Yellow outlines demarcate the regions of the sensor with the largest accessibility to O2 or NiEdda. The dashed black lines point to the presumed limits of the lipid bilayer. Details are identical to those in Fig. 1C.

A more detailed investigation of the likely arrangement of the individual TM segments can be obtained from frequency and vector analysis of the residue accessibility parameters (Fig. 3). In each case (Fig. 3, A to C), the resultant vectors displaying the orientation of the accessible surfaces are shown both in the context of helical wheel representations and mapped in three dimensions on the individual segment structures derived from the isolated sensor. S1 and S2 offer contrasting examples of this analysis (Fig. 3A). In S1, the sum vectors for O2 accessibility (Embedded Image) and mobility (Embedded Image) are negligible, which strongly suggests that S1 is likely surrounded by other regions of the channel. In S2, Embedded Image and Embedded Image are robust and essentially in phase, pointing to a well-defined lipid-exposed region toward the center of the helix. The very large sum vectors in S3a and S3b for both ΠNiEdda and Embedded Image are also in phase (Fig. 3B), which is consistent with the idea that the putative aqueous crevices are large enough to allow substantial motions of the spin label. Indeed, the NiEdda accessibility profile shows that a stretch of only about 14 residues in S3 is fully inaccessible to NiEdda, corresponding to a slab of about 20 Å or less across the membrane. Moreover, because the ΠNiEdda parameter estimates effective collisions for a probe of 5 to 6 Å, it is likely that actual water penetration into these putative crevices is more extensive than that revealed by the ΠNiEdda parameter.

Fig. 3.

Frequency and vector analysis of environmental data point to the likely arrangement of the individual TM segments. (A to C) The orientation of the sum vector for accessibility Embedded Image and mobility data Embedded Image in S1-S2 (A), S3 (B), and S4 (C) as shown on helical projections (circle), with a single residue serving as reference point (black dot). In S4, the orientation of the charged residues is represented by blue dots. The shaded area highlights the degree of eccentricity for the complete set of accessibility data relative to the maximal accessibility vector. In each case, the sum vector pointing to the direction of highest accessibility is also plotted in three dimensions (yellow arrows) in relation to the structure of each segment as it appears in the isolated voltage sensor. Accessibility to O2 or NiEdda has been color coded onto the backbone worm of each TM segment. In segments S3 and S4, helical wheels are also shown for the two individual helices between the linker (S3a, S3b, S4a, and S4b) and the relevant regions highlighted by a gray dotted rectangle around the structure. In the S4 segment, the position of the charged residues is shown by blue dots around the unitary circle. (D) Conceptual model illustrating the effect of twisting S4a and S4b (relative to each other) on the location of charged groups at the protein-membrane interface.

Based on the loss of α-helical periodicity around residue 129, the calculated Embedded Image and Embedded Image for both S4a and S4b are large and essentially in phase (Fig. 3C), as expected from an S4 with an extensive lipid-accessible surface. Additional data about the peripheral location of S4, based on fluorescence measurements of aggregated channel, are provided in (27). Given the overwhelming evidence for the peripheral location of S4, a key question to ask is what type of environment surrounds the charged residues. Most of the S4 charges in KvAP (R123, R126, R133, and R136) are not exposed to the environment, as indicated by low probe mobility and negligible accessibility to either O2 or NiEdda (Fig. 3C, blue numerals). R120 appears partially protected, although it is at the edge of the nonaccessible region in the helical wheel (Fig. 3C). Finally, R117 is located in the accessible region of the helical wheel and shows fairly high motional freedom, but its intermediate O2 and NiEdda accessibility argues for a location close to the water/lipid interface. In the presence of the S4a-S4b linker, the two resulting helices appear to be twisted ∼90° in relation to their accessible surface, an arrangement that places the majority of the charges on the same face of the segment, in contrast with the natural helical screw arrangement of charges expected if the S4 were a straight helix (Fig. 3D).

The present data set provides an opportunity to evaluate explicit open-state models of potassium channel structure. To this end, we have considered two recent versions of the “canonical” model from the laboratories of Benoit Roux and Diane Papazian [the LPR model (23)] and Robert Guy [the DHG model (22)] (Fig. 4A, data shown for the LPR model only), plus a paddle-like structural model (21) based on an interpretation of the KvAP open “structure” of Jiang et al. (Fig. 4D) (15). [Details on the construction of the model are given in (27)]. Therefore, these coordinates are an approximation of the original paddle model (which does not include an explicit position for S1) and thus should serve only as a guide for the general evaluation of the model in relation to the present data set. When considering the water-exposed surfaces, the canonical models show remarkable compatibility with the ΠNiEdda experimental data, as expected from the transmembrane topology of all helical segments (Fig. 4B). However, the agreement is not favorable for the paddle model (Fig. 4E), which places extensive areas of NiEdda-accessible residues well within the low dielectric regions of the bilayer. The bulk of these discrepancies derive from the unusual placement of S2 as a band surrounding the pore domain, approximately parallel to the plane of the bilayer. The lipid-exposed regions were evaluated more quantitatively by correlating the average of the experimentally determined O2 accessibility in each TM segment with the solvent accessibility calculated by a hard-sphere scanning method (Fig. 4, C and F) (28). In this case, the canonical models fail to produce a positive correlation with the calculated accessibilities, primarily as a result of the shielding of S4 away from the periphery of the molecule (Fig. 4C). On the other hand, the paddle model shows significant correlation to the expected TM segment accessibilities, a consequence of the location of the paddle at the channel/lipid interface (Fig. 4F).

Fig. 4.

Evaluation of structural models of KvAP. The explicit coordinates of three structural models of open KvAP were used to evaluate overall concordance with experimental accessibility data. (A) Ribbon representation of the canonical model (results for the Laine-Papazian-Roux model are shown). Color coding of the TM segments is the same as in Fig. 2C. (B) Map of NiEdda-accessible residues onto a solvent accessible surface. The dotted line points to the approximate location of the water/lipid interface. Black ovals highlight regions of high NiEdda accessibility putatively embedded in the bilayer. (C) Correlation between the average solvent accessibility (28) of the lipid-embedded region of the LPR model, with the average experimentally determined O2 accessibility in the same set of residues. Each point corresponds to individual TM segments; the error bars represent SDs. The gray line represents a 1:1 correlation. (D) Ribbon representation of the paddle model, as above. (E) Map of NiEdda-accessible residues onto a solvent accessible surface. (F) Correlation between the average solvent accessibility of the lipid-embedded region of the paddle model with the average experimentally determined O2 accessibility in the same set of residues. Details as in (C). In parts (B) and (E), the solvent accessible surface was calculated and color mapped with the program Chimera (35, 36). (G) Side view of a helix-packing model of KvAP based on the present data set showing the arrangement of the TM segments relative to the membrane and the pore domain. The model highlights the peripheral location of the S4 segment and points to S1 as the segment most likely to be surrounded by protein.

We found that many of these discrepancies can be reduced or eliminated by reasonably simple reorientations of the sensor domain structure relative to the pore domain. In the case of both canonical models, a rotation of ∼100° to 120° degrees about the long axis of the sensor domain helps correct major discrepancies in lipid-exposed accessibilities. In the paddle model, tilting the sensor domain ∼50° to 60° toward the pore domain (plus internal repositioning of S1 and S2) places the majority of the water-exposed regions above the water/lipid interface, preserving the disposition of the lipid-exposed areas. This first-order approximation simply docks the current isolated sensor structure onto the pore domain, although it is likely that the sensor structure will be somewhat different in the context of the full-length channel. The resulting model (Fig. 4G) places the S4 segment squarely at the protein/lipid interface, in agreement with the paddle model. The voltage sensor contacts the pore domain in the immediate vicinity of the S1 segment, whose restricted dynamics and low accessibility imply that it is mostly surrounded by protein. Furthermore, the S4 segment behaves as two α helices, S4a and S4b (Fig. 3), connected by a short linker (residues 129 to 131). The influence of this linker region on the functional behavior of the sensor remains to be established, but it might point to the presence of a hinge that leads to differential rearrangements of these two regions of S4 in response to voltage changes.

Our data support the general idea of the S4 charges being shielded from the low dielectric environment of the membrane (at least for the open/inactivated conformation), in agreement with general concepts behind earlier voltage-sensor models. However, the present results are incompatible with models that place the S4 segment in a cocoon of surrounding protein that protects it from the lipid environment and generates a so-called “S4 channel” or “canaliculi.” The specific pattern of dynamics and lipid accessibilities for S1 to S3 also makes our data incompatible with more recent transitional models in which the S4 segment might be partially exposed to lipid (22, 29). At the moment, we do not have enough constraints to differentiate between a paddle-like model and one in which the S4 segment is indeed peripheral but is also part of deep aqueous crevices, helping to “focus” the transmembrane voltage field during channel activation (30, 31). This will require information regarding the conformation of the sensor in the closed state.

Our model is also partially at odds with the results from studies of perturbation analyses in eukaryotic Kv channels (3234). Although there is agreement regarding the local environment surrounding S2 and S3, these studies have concluded that S1 is likely located at the periphery of the channel, in contrast with the present results. We do not fully understand the origin of this discrepancy, although the direct translation of functional data into structural parameters is not always straightforward. Although further structural analyses will be required to solve these issues and ultimately define the multiple conformations of membrane-embedded Kv channels, the present KvAP model represents a starting point for the analysis of the structure and conformational changes underlying voltage-dependent gating.

Supporting Online Material

www.sciencemag.org/cgi/content/full/306/5695/491/DC1

Materials and Methods

Figs. S1 to S3

References

Movie S1

References and Notes

View Abstract

Navigate This Article