Determinants of pore folding in potassium channel biogenesis

Erin Delaney; Pooja Khanna; LiWei Tu; John M. Robinson; Carol Deutsch

doi:10.1073/pnas.1324274111

Edited by Richard W. Aldrich, The University of Texas at Austin, Austin, TX, and approved February 11, 2014 (received for review December 31, 2013)

Significance

The essential structural feature of canonical K, Na, and Ca channels is a permeation pathway created by four reentrant pore loops. This study provides insights into the biogenic determinants of this reentrant architecture, and therefore potential candidates for folding defects.

Abstract

Many ion channels, both selective and nonselective, have reentrant pore loops that contribute to the architecture of the permeation pathway. It is a fundamental feature of these diverse channels, regardless of whether they are gated by changes of membrane potential or by neurotransmitters, and is critical to function of the channel. Misfolding of the pore loop leads to loss of trafficking and expression of these channels on the cell surface. Mature tetrameric potassium channels contain an α-helix within the pore loop. We systematically mutated the “pore helix” residues of the channel Kv1.3 and assessed the ability of the monomer to fold into a tertiary reentrant loop. Our results show that pore loop residues form a canonical α-helix in the monomer early in biogenesis and that disruption of tertiary folding is caused by hydrophilic substitutions only along one face of this α-helix. These results provide insight into the determinants of the reentrant pore conformation, which is essential for ion channel function.

The functions of ion channels derive from their architecture, namely, their primary sequence and their secondary, tertiary, and quaternary folds. Flux of ions occurs through a transmembrane pore, which in a potassium-selective channel (K channel) is formed by the convergence of four subunits, each with a reentrant “pore loop” between two transmembrane helices. This intervening loop in each subunit contains a turret, a pore helix, a selectivity filter (SF), and a loop linking the SF to the terminal transmembrane helix (Fig. 1A). The pore loop’s contributions to permeation and gating are well known, yet determinants of its reentrant architecture, manifest in the monomer for a voltage-gated K (Kv) channel (1), have not been elucidated.

Fig. 1.

Accessibility of a cysteine reporter in the monomeric K channel pore loop. (A) Monomer of the Kv1.2/2.1 pore region from Protein Data Bank (PDB) ID code 2r9r (8). The turret is shown in blue, the pore helix in red, the selectivity filter (SF) in yellow, the loop preceding S6 (loop-S6) in green, and transmembrane segments S5 and S6 in light gray. The dark gray space-filling residue is equivalent to M390 in Kv1.3. (B) Sequence alignment of the pore helix from selected ion channels. Conserved aromatic residues (red box) are highlighted. (C) Sequence of Kv1.3 pore. Colors are as in A. M390C is bolded black. (D) Time course of cysteine modification (Left) and quantification of the kinetics (Right). M390C/Kv1.3(T1−) was isolated from microsomal membranes, pegylated (2 mM PEG-MAL) for the indicated times, and fractionated using SDS/PAGE (1). Numbers to the Right of the gel indicate unpegylated (0) and singly pegylated (1) protein. The unpegylated doublet is the parent protein in the ER-glycosylated form (upper band of doublet) and unglycosylated form (lower band of doublet). Three minor bands (∼15%) in the vicinity of 45 kDa are due to routine minor glycosylation of three consensus sites in the C terminus of Kv1.3. Data are plotted as percentage unpegylated and fit with a single exponential to obtain k_mod.

The pore helix (red, Fig. 1A) consists of 12 residues that are highly conserved across most K channels (Fig. 1B). Moreover, the aromatic residues in the middle of the pore helix are conserved even for such diverse ion channels as GluR2, NavAb, CNG, KIR, and BK channels. Do these conserved residues serve as determinants of pore loop tertiary folding? Mutations in this region cause disease, e.g., long QT syndrome, a potentially fatal condition (2, 3). Although trafficking defects have been implicated (3), the underlying secondary, tertiary, and/or quaternary folding defects have not been identified. Key to understanding pathology is our understanding of these folding events and determinants of pore formation, specifically of pore loop folding into a reentrant conformation.

In this paper, we probe the key determinants of pore loop tertiary structure at an early stage of biogenesis in the endoplasmic reticulum (ER)-resident monomer of Kv1.3. Our findings highlight critical determinants of secondary and tertiary folding that allow segments of a membrane protein to be correctly embedded in the bilayer, events of fundamental importance in biogenesis.

Results

To evaluate topological determinants of pore loop architecture in early Kv biogenesis, we used two strategies. First, we chose a T1-deleted Kv1.3 (Kv1.3 T1−), which is well translated in a cell-free rabbit reticulocyte lysate, integrated into microsomal membranes (ER membranes), and is monomeric under the in vitro translation conditions of our experiments, but can produce tetramers and functional channels under other conditions (4 ⇓–6). Second, as an indication of tertiary folding, we determined the accessibility of a reporter cysteine, M390C (Fig. 1A), engineered into the C terminus of the pore helix of an otherwise cysteine-free Kv1.3(T1−), which folds and functions (4, 7). Its accessibility will be low when the pore helix is embedded in a reentrant conformation, and high when the tertiary structure is disrupted (1). In the accessibility assay, we measure the kinetics of modification of a reporter cysteine by a mass-tagged maleimide reagent, PEG-maleimide (PEG-MAL). An accessible cysteine will have a relatively fast modification rate, whereas a buried one will have a relatively slow rate. We thus measured modification rate constants, k_mod, for M390C.

The pegylation time course (Fig. 1D, Left) shows the disappearance of a lower doublet (“0” band, unpegylated protein; Methods) and appearance of pegylated protein (band “1”). The percent unpegylated protein (1 − Fpeg) monotonically decreases (right) and is well fit with a single-exponential function to give a k_mod of 0.43 M⁻¹⋅s⁻¹. Replicate experiments for M390C in the native Kv1.3(T1−) background gave an average k_mod of 0.74 ± 0.11 M⁻¹⋅s⁻¹ (n = 5), consistent with previous values for this residue in a folded monomer in the ER membrane and 10–50 times slower than that for exposed residues in the turret and loop-S6 (1).

M390C appears to be hindered, due to its distance from the surface in a reentrant conformation and its protein/lipid environment. Although the axial length along the helix is 18 Å (Fig. 2A), the distance of a given residue to the membrane surface will depend on the tilt angle of the helix in the membrane. However, another factor contributing to pore helix accessibility is the intramolecular contacts of the pore helix with transmembrane segments S5, S6, and the selectivity filter. Although this region is helical in the tetrameric crystal structure (8), its secondary conformation is undetermined for the monomer in the ER. To evaluate helicity of the monomeric “pore helix,” we used periodicity analysis. To experimentally assess periodicity, we probed solvent accessibility of pore helix residues using cysteine modification and an analysis of periodicity using a discrete Fourier transform (9, 10). Each residue was mutated to a cysteine and pegylated to determine k_mod (Fig. 2B and Table S1). Based on the relative k_mod values, there appears to be a less accessible face (blue arrow) and a more accessible face (orange arrows), with the exception of the two threonines at the bottom of the pore helix, which likely reside at intramolecular interfaces (8). A periodicity analysis of these k_mod values gives a periodicity index, α-PI, of 2.55 and a spectral peak at 114°, both consistent with an α-helix (for an α-helix, α-PI is ≥2). This cysteine scan provides clear evidence that this region, in the monomer, is a helix.

Fig. 2.

Location of pore helix residues. (A) Cartoon representation of two factors influencing residue accessibility: helical face and axial length. (B) Helical wheel diagram of Kv1.3 pore helix. Hydrophilic, hydrophobic, and acidic residues are indicated by circles, diamonds, and triangle, respectively. The numbers in red boxes are mean k_mod (molar⁻¹⋅second⁻¹; Table S1) for cysteine substitution at the indicated position. High rates are delineated by the orange arrows; low rates are mostly delineated by the blue arrow. The words “Buried” and “Exposed” indicate the deduced exposure of pore residues in a mature tetrameric Shaker Kv channel (15).

Assuming an α-helix (Fig. 2B) and using the Eisenberg hydrophobicity scale, HydroCalc (www.bbcm.univ.trieste.it/∼tossi/HydroCalc/HydroMCalc.html), the mean relative hydrophobic moment, calculated relative to that of a perfectly amphipathic peptide, is 0.12, which is very low. Thus, although this region is a helix, it does not manifest clearly hydrophobic and hydrophilic faces. Nonetheless, it has a solvent accessible face, and we thus postulated that the more accessible sites would better tolerate potentially disruptive side chains.

To this end, we first screened a series of pore helix mutants for expression of voltage-activated K⁺ current in Xenopus oocytes. Of the 46 mutations we engineered along the pore helix of Kv1.3, only 12 expressed current (Table S2). The absence of Kv current could arise from either a biophysical defect in gating or a biogenic defect. The latter category contains candidates for faulty formation of the pore loop architecture, which we explored by engineering a select group of the nonfunctional Kv1.3 mutations into M390C/Kv1.3(T1−) and measuring k_mod. Initially, we did alanine and glutamate scans (Fig. 3A, purple and red triangles, respectively, and Table S1). P380 is strategically poised to allow the bend for loop reentry and therefore was not mutated at any time in our study.

Fig. 3.

M390C modification rate constants and periodicity analysis. (A) Mean k_mod of 390C for peptides containing the indicated mutations. For the indicated single point mutations, k_mod for M390C was determined from a fit of the time course to a single-exponential function (P < 0.001). For mutants displaying a fractional decay, FD (fraction unpegylated protein), of >0.35 at 1 min (fast rate), the data were fit with a double-exponential function and a weighted average of the fast and slow rate constants used to calculate k_mod. All data are mean ± SEM (n = 3–5) or mean ± average deviation for n = 2 (Table S1). Where error bars are not visible, they are smaller than the dimensions of the symbol. The dashed blue line highlights the range (less than or equal to ∼2.5 M⁻¹⋅s⁻¹) of slow rates, not significantly different from slow rates obtained from a cysteine scan of the pore helix (Fig. 2B and Table S1). (B) Mean k_mod for 390C peptide with a Glu at the indicated pore helix residue. These values were used in the periodicity analysis. (C) Monomer of the Kv1.2/2.1 pore region as in Fig. 1A, with residues equivalent to F383 (black), A386 (blue), and V387 (green) shown as space-filling residues. (D) Normalized spectral density versus angular frequency. The discrete Fourier transform power spectrum was calculated from the natural logarithm of the mean modification rate constant of each pore helix residue (B), using the algorithm of Cornette et al. (9) (their equation 1). The calculated α-PI was 3.6 and the spectral peak is 97°.

Only three residues, situated near the middle of the pore helix, exhibited enhanced modification kinetics at M390C when mutated, namely F383, A386, and V387. Ala produced little or no change in k_mod except at position 383 [k_mod is 5.10 ± 0.54 (n = 3)] and position 387 [k_mod is 4.05 ± 1.42 (n = 3)]. In both cases, rates are approximately fivefold higher than observed with the native residues and the Ala substitutions decrease the hydrophobicity of the side chain at these positions (11). The more provocative changes, however, occurred with a glutamate scan of the pore helix. Glu substitution gave a k_mod at 390C similar to that obtained for alanine, except at 383, 386, and 387. Glu produced an ∼18-, 8-, and 21-fold higher modification rate, respectively, compared with that for the native pore helix residues. We interpret this as an increased accessibility to PEG-MAL in the vicinity of 390C, consistent with improper folding of the pore helix. Despite the fact that 388 and 389 are further from the surface and near the bottom of the pore helix, they tolerate Glu [k_mod is 0.45 ± 0.08 (n = 3) and 0.65 ± 0.08 M⁻¹⋅s⁻¹ (n = 3), respectively, at M390C]. However, the neighboring 387 position does not, as manifest by the 21-fold increase in M390C k_mod for the V387E mutant compared with the native pore helix.

To focus attention on the glutamate perturbations, a plot of the glutamate scan alone is shown in Fig. 3B. The three strong folding determinants, F383, A386, and V387, are located on the same face of the pore helix in the native structure (Fig. 3C) and this face is relatively buried (Fig. 2B and Table S1). Glu substitution on the less accessible face has a dramatic effect on k_mod at M390C, whereas glutamate substitution on the more accessible face is tolerated with little or no change in k_mod at M390C. To evaluate whether Glu disruption has a helical pattern, we did a periodicity analysis of the k_mod data for M390C when Glu is substituted, one at a time at residues 381–389. Glu substitution exhibits a strong α-helical periodicity with a periodicity index of 3.6 (Fig. 3D). This index is evidence for a canonical α-helix, in which consecutive residues are separated by 100°, consistent with the observed spectral peak at 97°.

We interpret high values of k_mod as evidence that selected Glu mutants produce a perturbed tertiary conformation of the reentrant pore loop, due to either unraveling, rotation, and/or movement of the helix toward the surface. Before addressing these different scenarios, we first evaluated whether the enhanced rates were due to electrostatics or to characteristics of the Glu mass and geometry. We suggest the former because substitution at A386 with Trp (largest side chain) does not perturb k_mod (Fig. 3A, yellow triangle).

To more directly test the hypothesis that charged side chains at specific pore helix locations alter k_mod, we explored the effects of Lys and Asp at position 387 (Fig. 4A) and, for comparison, at 388 (Fig. 4B). All charged side chains at position 388 produced slow rates at M390C and were fit with a single-exponential function to get k_mod. In contrast, charged side chains at 387 produced fast rates that could not be well fit with a single-exponential function but were fit well with a double-exponential function and a weighted average of the fast and slow rate constants used to estimate k_mod (Methods). Substitution of a negatively charged, but smaller, Asp at 387 produces a similar (26-fold) increase in modification rate, consistent with a charge at this location disrupting the tertiary fold of the reentrant loop. Moreover, when we substituted positively charged Lys at this location, we obtained a similarly enhanced modification rate [24.5 ± 4.0 M⁻¹⋅s⁻¹ (n = 3)]. At positions 386 and 387, each locations where Glu and Asp produce an increased k_mod, substitution with Lys also produces an increased k_mod. At position 388, where Glu and Asp produce no change in rate, Lys also leaves the modification rate unchanged (Fig. 4C). Residues 387 and 388 are consecutive residues and, if present in a canonical α-helix, would be separated by 100° but in a similar axial plane. The surrounding tertiary environments must be different.

Fig. 4.

Kinetics of modification for V387 and V388 charge mutants. (A and B) Time course of modification of M390C/Kv1.3(T1−) mutated to the indicated residues. Pegylated and unpegylated protein bands are as in Fig. 1. Below each gel is the quantified disappearance of the unpegylated species fitted with an exponential function to give the indicated k_mod. The series of experiments illustrated in A are for mutants with fast modification rates; those in B have slow modification rates. (C) In the background of M390C, k_mod are means ± SEM, n = 3–4 for A386E, V387E, K, D, and V388E, K, D, and means ± average deviation for duplicate measurements for A386K and D.

Regarding the perturbed tertiary conformation of the monomeric pore, we probed other reporter locations. The accessibility of M390C, buried at the nadir of the pore helix, is expected to increase if the entire pore helix moves as a rigid rod toward the extracellular surface of the protein (tertiary unfolding of the reentrant loop) and/or if the pore helix itself unwinds its secondary structure and resides on the surface. Either scenario predicts that the accessibility of reporter residues all along the pore helix will increase upon introduction of charges at 383, 386, and 387. However, a residue in the turret that is already maximally exposed will remain relatively unaffected by a disrupting mutation. As predicted, control residue D371C in the exposed turret of the pore region, which has a high cysteine modification rate constant (1), is relatively unchanged when V387E is engineered into the 371C background [k_mod is 14.5 ± 0.1 M⁻¹⋅s⁻¹ (n = 2)]. We then examined a reporter cysteine, F383C, stationed midway along the pore helix on the same helix face as M390C but spaced ∼10 Å apart along the longitudinal axis of the helix (8), and measured k_mod for mutations A386E, V387E, and V388E. At this new location (closer to the surface), qualitatively similar results were obtained as those for a reporter at the bottom of the pore helix (Fig. S1), consistent with a movement of the entire pore helix toward a more superficial location.

To distinguish further the nature of the folding disruption, we chose one more reporter site, V388C, located on the opposite face of the helix (Fig. S1) and relatively accessible [10.1 ± 1.5 M⁻¹⋅s⁻¹ (n = 3)]. If a rigid-rod rotation occurs along the helical axis with charge substitution at 387, then we expect a decrease in V388C accessibility. However, this is not the case. The modification rate constant for V388C remains high [9.58 ± 0.74 M⁻¹⋅s⁻¹ (n = 2)] in the presence of a glutamate substitution at residue 387. The most parsimonious interpretation of all of our data is that substitution of a charged residue on the hindered face causes the helix to move toward the surface, disrupting the reentrant conformation.

In Kv1.3, mutation of either Trp384 or Trp385 to Ala, Glu, Leu, or Pro gave no K⁺ current expression (Table S2). Although an aromatic residue at this location appears to be necessary for function, and in HERG for trafficking to the cell surface (12), the detailed kinetics of side-chain substitution for Trps in the tertiary folding assay has not yet been assessed. We thus mutated residue W384 to Phe and determined k_mod at M390C. This mutant is homologous to Shaker W434F, which is predicted to fold because it is functional (13). W384F/M390C produces a k_mod of 1.02 ± 0.07 M⁻¹⋅s⁻¹ (n = 2) (Fig. S2), consistent with it being folded. Substitution of an Ala at either W384 or W385 produces no significant difference in k_mod. Other side chains are able to substitute in the folding reaction. In contrast, the double-mutant W384A/W385A exhibits a faster k_mod, suggesting that the double-alanine substitution alters the conformation of the pore helix relative to the SF and surrounding residues. The relative insensitivity of folding to single Trp substitution may reflect redundancy conferred by the second aromatic residue.

Discussion

In this study, we set out to explore determinants of pore helix tertiary folding to form the reentrant architecture of the Kv pore. Previously, we demonstrated that this architectural feature of channel pores is likely to be manifest in the monomer at early stages of biogenesis, mediated in part by its intramolecular interactions (1). However, two significant issues remained unaddressed. First, although the pore helix region residing within the exit port of the ribosomal tunnel is compact relative to a fully extended nascent peptide, its secondary structure in this location was unknown. Nor was the secondary structure of the pore helix known within a Kv monomer embedded in a lipid bilayer membrane. In this study, we show that the secondary structure is indeed a helix. Second, we did not know which residues of the pore helix region were key determinants of the tertiary reentrant conformation of the monomer. In this study, we show that three pore helix residues play critical roles in stabilizing this reentrant conformation.

To address these questions, we used a twofold strategy: cysteine scanning and introduction of residues that might disrupt the reentrant conformation. To probe the environment on each face of a putative helix, we cysteine scanned all along the pore helix region and determined accessibility from the kinetics of modification by PEG-MAL, a method previously demonstrated to reflect relative accessibility (14). This systematic scan shows that one face of the pore helix is less accessible to a cysteine reagent than the other. We further confirmed this observation by an analysis of the periodic pattern of the modification rates along the pore helix, using a Fourier power spectrum (9). The spectrum for this cysteine scan has a prominent peak in the vicinity of 100° and a periodicity index of 2.55. Values of α-PI > 2 are supportive evidence for a two-faced α-helix (9). Interestingly, the pattern of high and low modification rates of residues along the accessible and inaccessible faces is consistent with results from Lü and Miller (15), who used a cysteine scan and Ag⁺ binding as an assay for accessibility in the functional tetramer of Shaker. Residues that produced a decrease in conductance when treated with Ag⁺ were deemed “exposed.” Those producing little or no change in conductance were deemed “buried.” In the Kv1.3 monomer, residues with relatively high k_mod values correlate well with exposed residues in the tetramer, whereas relatively low k_mod values are consistent with buried residues in the tetramer. This again supports the notion that the pore helix architecture is primarily established in the monomer and is fine-tuned in the tetramer by additional intermolecular interactions. Moreover, the accessibility of pore helix residues is not simply correlated with their hydrophobicity. As noted above, the pore helix itself is only poorly amphipathic, yet there is a clear demarcation between hindered and accessible faces (Fig. 2B).

We predicted that residues on the hindered face of the pore helix would not tolerate hydrophilic substitutions, and indeed that is what we observed, especially when using negatively charged glutamate substitutions. For our systematic glutamate scan, we used M390C as a reporter. We interpret high values of k_mod for this residue, in response to specific glutamate substitutions, as evidence that the pore helix has departed from its buried reentrant position. In fact, glutamate substitutions at only three positions (F383, A386, and V387) significantly raise k_mod, and these residues are squarely in the middle of the hindered face of the pore helix. This conclusion is further supported by a periodicity analysis (Fig. 3D). In this case, α-PI is 3.6, well above the value of 2, and therefore consistent with a helical phenotype. Moreover, a reporter cysteine at the bottom of the pore helix (M390C), and one near the top (F383C), show the same pattern in k_mod when glutamates are substituted into the pore helix at disruptive positions.

What kinds of conformational changes might cause these increases in k_mod in response to glutamate substitutions at positions 383, 386, and 387? Three possibilities are unraveling, rotation, or translocation of the pore helix toward the surface. Rotation of the helix at an interface between lipid and water is ruled out by using reporter cysteines on the opposite faces of the pore helix. If glutamate substitution causes an increase of k_mod at M390C, then a rotation of the pore helix along its axis would cause a decrease, viewed from the opposite face of the helix at V388C, which does not happen. The results in total suggest that disruptive mutants cause the entire pore helix to move to a more superficial (i.e., accessible) location with respect to the bilayer. Equivalently, the exposure of the reporter residues to hydrophilic reagents could be caused by conformational changes in the environment surrounding the pore helix. Whether it maintains its helical conformation in an accessible location is not clear.

A fair question to ask is why misfolding would have a helical pattern. We imagine tertiary folding as a sequential process. The pore helix is likely to have a helical secondary structure before it begins to enter and fold into the bilayer. This tertiary folding is accompanied by an energetic penalty if a glutamate side chain encounters a hydrophobic environment. As a consequence, the buried reentrant conformation is destabilized, and the pore helix tends to move to a more superficial (and higher dielectric) location. Note that, besides increasing k_mod, the mutations F383E, A386E, and V387E often show double-exponential time courses of modification (Fig. 4A). This could be explained if only a fraction of the monomers are in a disrupted superficial location in these glutamate mutants. Consistent with this hypothesis, all slow rate constants (0.46 ± 0.61; n = 24) from double-exponential fits to the charge-mutant kinetics are within the range of k_mod determined for nonperturbed residues in the glutamate scan, i.e., k_mod values below the dashed line in Fig. 3A. The slow component compares well with rates from a tertiary folded state.

The glutamate scan was part of a more extensive examination of residues that either tolerate or disrupt the reentrant conformation. The results together show that hydrophilic, especially charged, substitutions at positions F383, A386, and V387 raise k_mod values, again consistent with the energetic cost of burying such side chains in a hydrophobic environment. In general, the electrophysiology of these mutants is also consistent with this scenario, in that increasingly hydrophilic substitutions at these positions (e.g., F→A→E) increase k_mod monotonically and the probability of the channel being nonfunctional. Charged side chains (Glu, Lys, Asp) may be tolerated at most pore helix positions because water can reach these locations in the monomeric structure and stabilize an aqueous crevice (pocket of high dielectric) during this stage of biogenesis. Other locations, namely 383, 386, and 387, may not accommodate water in the tertiary fold of the pore helix, and thus charge substitution leads to conformational rearrangement of the pore region. Substitution of a bulky and hydrophobic Trp at 386 does not perturb the tertiary fold, again consistent with charge rather than sterics governing the energetics of the folded state. However, another possible consideration is protonation of the glutamate side chain to avoid the cost of burying a charge in a low dielectric. However, there is also an energetic cost to protonating the carboxylate group. Nonetheless, there is precedent for protonation of a pore helix glutamate of KcsA, E71 (equivalent to our residue 388), for which the pK_a is >7.5 (16).

Many cation-selective channels have structurally homologous reentrant pore loops. This includes channels selective for either K⁺ or Na⁺ ions, channels that poorly discriminate these two cations, and channels gated either by changes of membrane potential or by neurotransmitters. A well-conserved feature of all of these channels is a hydrophobic pore helix containing two or three consecutive aromatic residues near its middle (Fig. 1B). These aromatic residues create an “aromatic cuff” (17) important for both biogenesis and function. K channels typically have two consecutive tryptophan residues, both of which are critical for normal function (12, 18, 19) (Table S2). Conservation of aromatics suggests an important role of these residues, either for biogenesis or for channel function. However, mutation of either of these two residues individually to alanine does not disrupt the reentrant conformation of the pore helix, viewed from the reporter M390C. It is only when both are mutated to alanine within the same construct that k_mod increases substantially. We suggest, therefore, that the pore helix evolved with two consecutive tryptophans as a form of robust redundancy to protect against errors in channel biogenesis. We further hypothesize that the folded monomer is stabilized in part by extensive interactions of these tryptophans, both with other pore helix residues and with SF, S5, and S6 residues. These interactions presumably help to anchor the pore helix in its folded conformation and additionally may explain why mutations of these residues typically affect channel gating. The energetics of these stabilizing interactions remains to be elucidated. Our results, obtained from a variety of point mutations of both aromatic and other pore helix residues, suggest that hydrophobicity is a stabilizing factor in general. Indeed, earlier molecular-dynamic simulations of the monomer embedded in a bilayer show a preponderance of lipid molecules in close proximity to most pore helix residues, including F383 and W384 (1).

We show here that charged side chains destabilize a reentrant conformation at only three locations along the inaccessible face of the pore helix, validating its helical conformation in the biogenic monomer and suggesting that this immature pore helix lies along a hydrophilic–hydrophobic interface in the ER. These findings help elucidate prerequisites for pore formation in channel biogenesis.

Methods

Constructs, in Vitro Translation.

All mutations were made in a pSp64T/Kv1.3 T1(−) background construct using Stratagene QuikChange Kit. For details of molecular biology methods and in vitro translation see SI Appendix. Kv1.3 T1(−) RNAs were translated for 1 h at 30 °C in the presence of canine microsomal membranes (Promega), using a rabbit reticulocyte lysate translation system (Promega) with 2 mM DTT, an amino acid mixture minus methionine, and [³⁵S]cysteine/methionine (4 μL/50 μL translation mixture; ∼10 μCi/μL; Environmental Health and Radiation Safety, University of Pennsylvania, Philadelphia, PA) in a total reaction volume of 50 μL, according to the Promega Protocol and Application Guide.

Tertiary Structure Experiments.

A 5-μL sample of the 50-μL translation mixture was removed for assay on the gel; the remainder was diluted into a final volume of 500 μL of 1× PBS* (Gibco CaCl₂- and MgCl₂-free Dulbecco’s PBS, pH 7.4, supplemented with 4 mM MgCl₂), and 2 mM DTT (Invitrogen NuPAGE Sample Reducing Agent), and centrifuged (Beckman Optima TLX Ultracentrifuge, Beckman TLA 100.3 rotor) through a sucrose cushion (120 μL; 0.5 M sucrose, 100 mM KCl, 5 mM MgCl₂, 50 mM Hepes, and 1 mM DTT; pH 7.5; to remove globin protein) for 7 min at 192,500 × g at 4 °C to isolate ribosome-bound membranes in the pellet. These pellets, containing membrane-embedded peptides, were resuspended on ice with 100–375 μL of 0.1% Anatrace n-dodecyl-β-d-maltopyranoside, Anagrade, in PBS* with 1 mM DTT and incubated 1 h on ice to dissolve membranes. Although the resulting mixed micelle is a different environment from what the protein encounters in the ER membrane, it nonetheless approximates gross features of the environment presented by a lipid bilayer (1). Samples were centrifuged at 210,000 × g for 35 min at 4 °C and the supernatant containing solubilized peptides was transferred to a fresh tube, then aliquoted 50 μL per tube. An equal volume of PBS* solution containing PEG-MAL was added (final PEG-MAL was 2 mM, or 1 mM for very fast modification rates; final DTT was 0.5 mM), incubated on ice for 0–2 h, and quenched with 100-fold excess DTT at ambient temperature for 10 min.

Data Analysis.

The kinetics of pegylation was calculated as the disappearance of unpegylated protein normalized to total protein (1 − Fpeg) in each lane on the gel. The unpegylated protein was quantified as the sum of the two bottom-most bands (Fig. 1D, band 0), rather than as the appearance of pegylated protein. This obviates a calculation error due to a pegylated band (band 1) that may not be easily resolved from contaminating higher M_r bands (which migrate near band 1) derived from glycosylation of three C-terminal glycosylation consensus sites. These sites get glycosylated due to a flipped S6 topology (typically ∼15–20%), which occurs naturally in ER membranes, even without detergent solubilization (20). Thus, we consider only correctly folded, unpegylated protein, i.e., band 0 and its disappearance, to calculate k_mod of a reporter cysteine in the properly folded pore. To calculate the rate constant, k_mod, the time course of pegylation was fit to a single-exponential function, as judged by the goodness-of-fit parameter of a single-exponential function (P < 0.001). We defined the fractional decay of unpegylated protein as FD = ((1 − Fpeg)_max − (1 − Fpeg)_1′)/((1 − Fpeg)_max – (1 − Fpeg)_min). Mutants displaying a FD of greater than 0.35 at 1 min (fast rate) were fit with a double-exponential function and a weighted average of the fast and slow rate constants was used to calculate k_mod.

To determine the helicity of the pore helix in the monomer, we calculated the power spectral density of this region of the monomer (Fig. 3D) using a discrete Fourier transform (9, 10) (SI Appendix). The code for a SciLab program to do this calculation was kindly provided by Dr. Yu Zhou (Department of Anesthesiology, Washington University, St. Louis, MO). Besides the power spectrum, we determined the α-PI, which we define as the fractional weight of the power spectrum between 85° and 115°. Values of α-PI greater than 2 are consistent with an α-helix (9). We used two different datasets as inputs for this calculation: (i) natural logarithms of modification rate constants from a cysteine scan of residues 382–389, and (ii) natural logarithms of modification rate constants from a Glu scan with M390C as the reporter. For this analysis, we used k_mod values for residues 381–389. The values of α-PI are given in the text.

Acknowledgments

We thank Dr. Richard Horn for careful reading of the manuscript and helpful discussion. This work was supported by National Institutes of Health Grant GM 52302 (to C.D.).

Footnotes

↵¹E.D. and P.K. contributed equally to this work.
↵²To whom correspondence should be addressed. E-mail: cjd{at}mail.med.upenn.edu.

Author contributions: C.D. designed research; E.D., P.K., and J.M.R. performed research; E.D., P.K., L.T., J.M.R., and C.D. analyzed data; and C.D. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1324274111/-/DCSupplemental.

References

↵
1. Gajewski C,
2. Dagcan A,
3. Roux B,
4. Deutsch C
(2011) Biogenesis of the pore architecture of a voltage-gated potassium channel. Proc Natl Acad Sci USA 108(8):3240–3245.
OpenUrl Abstract/FREE Full Text
↵
1. Huang FD,
2. Chen J,
3. Lin M,
4. Keating MT,
5. Sanguinetti MC
(2001) Long-QT syndrome-associated missense mutations in the pore helix of the HERG potassium channel. Circulation 104(9):1071–1075.
OpenUrl Abstract/FREE Full Text
↵
1. Anderson CL,
2. et al.
(2006) Most LQT2 mutations reduce Kv11.1 (hERG) current by a class 2 (trafficking-deficient) mechanism. Circulation 113(3):365–373.
OpenUrl Abstract/FREE Full Text
↵
1. Lu J,
2. Deutsch C
(2001) Pegylation: A method for assessing topological accessibilities in Kv1.3. Biochemistry 40(44):13288–13301.
OpenUrl CrossRef PubMed
↵
1. Tu L,
2. et al.
(1996) Voltage-gated K⁺ channels contain multiple intersubunit association sites. J Biol Chem 271(31):18904–18911.
OpenUrl Abstract/FREE Full Text
↵
1. Tu L,
2. Santarelli V,
3. Deutsch C
(1995) Truncated K⁺ channel DNA sequences specifically suppress lymphocyte K+ channel gene expression. Biophys J 68(1):147–156.
OpenUrl CrossRef PubMed
↵
1. Robinson JM,
2. Deutsch C
(2005) Coupled tertiary folding and oligomerization of the T1 domain of Kv channels. Neuron 45(2):223–232.
OpenUrl CrossRef PubMed
↵
1. Long SB,
2. Tao X,
3. Campbell EB,
4. MacKinnon R
(2007) Atomic structure of a voltage-dependent K⁺ channel in a lipid membrane-like environment. Nature 450(7168):376–382.
OpenUrl CrossRef PubMed
↵
1. Cornette JL,
2. et al.
(1987) Hydrophobicity scales and computational techniques for detecting amphipathic structures in proteins. J Mol Biol 195(3):659–685.
OpenUrl CrossRef PubMed
↵
1. Zhou Y,
2. Xia XM,
3. Lingle CJ
(2011) Cysteine scanning and modification reveal major differences between BK channels and Kv channels in the inner pore region. Proc Natl Acad Sci USA 108(29):12161–12166.
OpenUrl Abstract/FREE Full Text
↵
1. Wimley WC,
2. White SH
(1996) Experimentally determined hydrophobicity scale for proteins at membrane interfaces. Nat Struct Biol 3(10):842–848.
OpenUrl CrossRef PubMed
↵
1. Ferrer T,
2. et al.
(2011) Molecular coupling in the human ether-a-go-go-related gene-1 (hERG1) K⁺ channel inactivation pathway. J Biol Chem 286(45):39091–39099.
OpenUrl Abstract/FREE Full Text
↵
1. Perozo E,
2. MacKinnon R,
3. Bezanilla F,
4. Stefani E
(1993) Gating currents from a nonconducting mutant reveal open-closed conformations in Shaker K⁺ channels. Neuron 11(2):353–358.
OpenUrl CrossRef PubMed
↵
1. Lu J,
2. Deutsch C
(2005) Secondary structure formation of a transmembrane segment in Kv channels. Biochemistry 44(23):8230–8243.
OpenUrl CrossRef PubMed
↵
1. Lü Q,
2. Miller C
(1995) Silver as a probe of pore-forming residues in a potassium channel. Science 268(5208):304–307.
OpenUrl Abstract/FREE Full Text
↵
1. Bhate MP,
2. McDermott AE
(2012) Protonation state of E71 in KcsA and its role for channel collapse and inactivation. Proc Natl Acad Sci USA 109(38):15265–15270.
OpenUrl Abstract/FREE Full Text
↵
1. Doyle DA,
2. et al.
(1998) The structure of the potassium channel: Molecular basis of K⁺ conduction and selectivity. Science 280(5360):69–77.
OpenUrl Abstract/FREE Full Text
↵
1. Yifrach O,
2. MacKinnon R
(2002) Energetics of pore opening in a voltage-gated K⁺ channel. Cell 111(2):231–239.
OpenUrl CrossRef PubMed
↵
1. Cordero-Morales JF,
2. et al.
(2006) Molecular determinants of gating at the potassium-channel selectivity filter. Nat Struct Mol Biol 13(4):311–318.
OpenUrl CrossRef PubMed
↵
1. Tu L,
2. Wang J,
3. Helm A,
4. Skach WR,
5. Deutsch C
(2000) Transmembrane biogenesis of Kv1.3. Biochemistry 39(4):824–836.
OpenUrl CrossRef PubMed

Article Alerts

Email Article

Citation Tools

Request Permissions

Article Classifications

Biological Sciences
Physiology

Proceedings of the National Academy of Sciences: 111 (12)

Table of Contents

Submit

[1] ↵
Gajewski C,
Dagcan A,
Roux B,
Deutsch C
(2011) Biogenesis of the pore architecture of a voltage-gated potassium channel. Proc Natl Acad Sci USA 108(8):3240–3245.
OpenUrl Abstract/FREE Full Text

[2] Gajewski C,

[3] Dagcan A,

[4] Roux B,

[5] Deutsch C

[6] ↵
Huang FD,
Chen J,
Lin M,
Keating MT,
Sanguinetti MC
(2001) Long-QT syndrome-associated missense mutations in the pore helix of the HERG potassium channel. Circulation 104(9):1071–1075.
OpenUrl Abstract/FREE Full Text

[7] Huang FD,

[8] Chen J,

[9] Lin M,

[10] Keating MT,

[11] Sanguinetti MC

[12] ↵
Anderson CL,
et al.
(2006) Most LQT2 mutations reduce Kv11.1 (hERG) current by a class 2 (trafficking-deficient) mechanism. Circulation 113(3):365–373.
OpenUrl Abstract/FREE Full Text

[13] Anderson CL,

[14] et al.

[15] ↵
Lu J,
Deutsch C
(2001) Pegylation: A method for assessing topological accessibilities in Kv1.3. Biochemistry 40(44):13288–13301.
OpenUrl CrossRef PubMed

[16] Lu J,

[17] Deutsch C

[18] ↵
Tu L,
et al.
(1996) Voltage-gated K⁺ channels contain multiple intersubunit association sites. J Biol Chem 271(31):18904–18911.
OpenUrl Abstract/FREE Full Text

[19] Tu L,

[20] et al.

[21] ↵
Tu L,
Santarelli V,
Deutsch C
(1995) Truncated K⁺ channel DNA sequences specifically suppress lymphocyte K+ channel gene expression. Biophys J 68(1):147–156.
OpenUrl CrossRef PubMed

[22] Tu L,

[23] Santarelli V,

[24] Deutsch C

[25] ↵
Robinson JM,
Deutsch C
(2005) Coupled tertiary folding and oligomerization of the T1 domain of Kv channels. Neuron 45(2):223–232.
OpenUrl CrossRef PubMed

[26] Robinson JM,

[27] Deutsch C

[28] ↵
Long SB,
Tao X,
Campbell EB,
MacKinnon R
(2007) Atomic structure of a voltage-dependent K⁺ channel in a lipid membrane-like environment. Nature 450(7168):376–382.
OpenUrl CrossRef PubMed

[29] Long SB,

[30] Tao X,

[31] Campbell EB,

[32] MacKinnon R

[33] ↵
Cornette JL,
et al.
(1987) Hydrophobicity scales and computational techniques for detecting amphipathic structures in proteins. J Mol Biol 195(3):659–685.
OpenUrl CrossRef PubMed

[34] Cornette JL,

[35] et al.

[36] ↵
Zhou Y,
Xia XM,
Lingle CJ
(2011) Cysteine scanning and modification reveal major differences between BK channels and Kv channels in the inner pore region. Proc Natl Acad Sci USA 108(29):12161–12166.
OpenUrl Abstract/FREE Full Text

[37] Zhou Y,

[38] Xia XM,

[39] Lingle CJ

[40] ↵
Wimley WC,
White SH
(1996) Experimentally determined hydrophobicity scale for proteins at membrane interfaces. Nat Struct Biol 3(10):842–848.
OpenUrl CrossRef PubMed

[41] Wimley WC,

[42] White SH

[43] ↵
Ferrer T,
et al.
(2011) Molecular coupling in the human ether-a-go-go-related gene-1 (hERG1) K⁺ channel inactivation pathway. J Biol Chem 286(45):39091–39099.
OpenUrl Abstract/FREE Full Text

[44] Ferrer T,

[45] et al.

[46] ↵
Perozo E,
MacKinnon R,
Bezanilla F,
Stefani E
(1993) Gating currents from a nonconducting mutant reveal open-closed conformations in Shaker K⁺ channels. Neuron 11(2):353–358.
OpenUrl CrossRef PubMed

[47] Perozo E,

[48] MacKinnon R,

[49] Bezanilla F,

[50] Stefani E

[51] ↵
Lu J,
Deutsch C
(2005) Secondary structure formation of a transmembrane segment in Kv channels. Biochemistry 44(23):8230–8243.
OpenUrl CrossRef PubMed

[52] Lu J,

[53] Deutsch C

[54] ↵
Lü Q,
Miller C
(1995) Silver as a probe of pore-forming residues in a potassium channel. Science 268(5208):304–307.
OpenUrl Abstract/FREE Full Text

[55] Lü Q,

[56] Miller C

[57] ↵
Bhate MP,
McDermott AE
(2012) Protonation state of E71 in KcsA and its role for channel collapse and inactivation. Proc Natl Acad Sci USA 109(38):15265–15270.
OpenUrl Abstract/FREE Full Text

[58] Bhate MP,

[59] McDermott AE

[60] ↵
Doyle DA,
et al.
(1998) The structure of the potassium channel: Molecular basis of K⁺ conduction and selectivity. Science 280(5360):69–77.
OpenUrl Abstract/FREE Full Text

[61] Doyle DA,

[62] et al.

[63] ↵
Yifrach O,
MacKinnon R
(2002) Energetics of pore opening in a voltage-gated K⁺ channel. Cell 111(2):231–239.
OpenUrl CrossRef PubMed

[64] Yifrach O,

[65] MacKinnon R

[66] ↵
Cordero-Morales JF,
et al.
(2006) Molecular determinants of gating at the potassium-channel selectivity filter. Nat Struct Mol Biol 13(4):311–318.
OpenUrl CrossRef PubMed

[67] Cordero-Morales JF,

[68] et al.

[69] ↵
Tu L,
Wang J,
Helm A,
Skach WR,
Deutsch C
(2000) Transmembrane biogenesis of Kv1.3. Biochemistry 39(4):824–836.
OpenUrl CrossRef PubMed

[70] Tu L,

[71] Wang J,

[72] Helm A,

[73] Skach WR,

[74] Deutsch C

Main menu

User menu

Search

Determinants of pore folding in potassium channel biogenesis

Significance

Abstract

Results

Discussion