INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

OVER THE 11 years since the cloning of the human cystic fibrosis transmembrane conductance regulator (CFTR) gene, progress has been made in defining portions of the protein comprising the pore-lining domains and residues within those domains that play important roles in establishing the biophysical character of open CFTR Cl channels (38). Several approaches have been applied to this study, including identification of transmembrane (TM) domains and residues therein that contribute to features such as open channel block by organic drugs, single-channel conductance in Cl, accessibility to chemical modification, and selectivity between monovalent anions. Although some portions of the CFTR peptide have been studied extensively, we do not yet have a clear picture of how these TM domains and amino acids work together to determine the electrostatic profile and physical dimensions of the permeation pathway.

It is presently not known how many TM domains contribute to the conduction pathway of CFTR. Indeed, the oligomeric structure of the functional channel is unclear. Despite several recent studies (10, 12, 28, 35, 43, 48), it is unclear whether a single CFTR channel is formed by a single CFTR peptide (see Ref. 29). The possibility exists that the functional CFTR channel is constructed from a dimer of CFTR peptides. A dual-pore model has also been suggested (46) wherein the amino-terminal and carboxy-terminal halves of the protein each confer Cl-conducting pores with different characteristics. For the purposes of the present study, we assumed that functional channels are made from a single CFTR peptide and confer only a single pore (one-channel, one-pore hypothesis), although we also interpret our data in light of the dual-pore configuration. Based on previous work (29) in TM6, TM11, and TM12 and preliminary data in TM5, we predicted that these four TMs contribute to the pore. This notion suggests a similar architecture to that of the substrate-binding domain of the related P-glycoprotein (25). TM6 is clearly the best studied of the TM domains in CFTR, having received so much attention because it is the helix that includes a greater number of charged amino acids than any other helix. However, because it is unlikely that the conduction pathway of this channel protein is constructed from only one or two TM helices, other domains must also contribute amino acids to the pore. It is not yet possible to identify homologous positions in any two helices that may lie across the pore from each other.

One of the parameters that describes a channel to a particularly high degree of detail is its ability to select between ions of similar charge. This functional distinction arises from a structural arrangement that is finely tuned to provide this specification (11). In contrast to other investigators (19), we do not believe that CFTR exhibits a well-defined selectivity filter similar to that found in voltage-gated ion channels but rather that selectivity arises from the generalized character of the amino acids lining the pore (39) in combination with a region that discriminates between anions of various sizes as described herein. It is generally accepted that CFTR exhibits only weak selectivity (9), especially compared with the strong selectivity between monovalent ions exhibited by the well-studied voltage-gated ion channels. However, the CFTR channel is capable of some degree of discrimination between halides and between Cl and larger polyatomic anions (e.g., Ref. 23).

From what portions of the protein does anion selectivity arise? The determinants of anion selectivity in CFTR have been studied by several investigators (29), again emphasizing primarily amino acids in TM6. Because most studies (e.g., Refs. 3, 6, 19, 28) have investigated either the effects of only a single substitution at a few positions or the effects of several substitutions at only one position, it has been difficult to assemble the data into a structural framework. To gauge the importance of mutations studied previously and new mutations presented as included here, we have begun to compare the results of a common substitution at several positions in TM6. Furthermore, because the anions that permeate CFTR are three-dimensional entities, the structures that confer selectivity (and anion binding sites) should be considered in three dimensions as well. Hence, it is important to compare the effects of mutations at positions in one TM domain with mutations at similar or homologous positions in other TM domains that may line the pore. Only this broad-scale approach will allow comparisons of the magnitude of effects at various positions and in more than one TM domain, making it possible to look for patterns in the data.

In the present work, we describe the initial steps in this study. We used macroscopic recording of CFTR currents in Xenopus oocytes to study selectivity between Cl and a wide range of test anions in wild-type (WT) CFTR and 12 CFTR variants. Our first objective was to add to the two-dimensional view of the pore by comparing the effects of equivalent substitutions at multiple positions along the length of TM6. To this end, we assayed the effects of mutations at K335, T338, and T339 (22-24, 26) and provide new information for mutations at S341. The second objective was to build toward a three-dimensional view by studying the results from mutations at comparable positions in TM6 and TM12. Our results suggest that the determinants of selectivity are distributed along the length of the pore, arguing against the presence of a classic selectivity filter in this channel. However, we also show that the region of strong discrimination between monovalent anions near the middle of TM6 extends to S341. Selectivity between monovalent and divalent anions was affected most strongly by mutations at the cytoplasmic end of this region. These data are consistent with a narrowing of the pore from the extracellular vestibule toward the middle of TM6. Mutations at one position in TM12 match the pattern reflected by similar mutations in TM6, suggesting strongly that this TM domain also contributes to the pore of CFTR. Portions of these data have been presented in abstract form (30-32).

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Preparation of oocytes and cRNA. WT CFTR was subcloned into the pAlter vector (Promega, Madison, WI). Mutants K335E, K335F, T338A, T339A, S341A, S341T, T1134A, and T1134F were prepared as previously described (33). Mutants K335A, T338E, and T1134E were prepared with the QuickChange protocol (Stratagene, La Jolla, CA). S341E CFTR was a generous gift from D. Dawson (Oregon Health Sciences University, Portland, OR). All mutant constructs were verified by sequencing across the entire open reading frame before use. Stage V-VI oocytes were isolated from female Xenopus and were incubated at 18°C in a modified Liebovitz L-15 medium with the addition of HEPES (pH 7.5), gentamicin, and penicillin-streptomycin. Capped transcripts (mMessage mMachine, Ambion, Austin, TX) for WT CFTR and for each of the variants (2-35 ng for most and 340 ng for T1134E CFTR) along with 0.6 ng of ₂-adrenergic receptor cRNA were injected into each oocyte. Recordings were made at room temperature 42-96 h after injection.

Electrophysiology. Standard two-electrode voltage-clamp techniques were used to study whole cell currents. Electrodes were pulled in four stages from borosilicate glass (Sutter Instrument, Novato, CA) and filled with 3 M KCl. Pipette resistances measured 0.5-1.4 M in standard ND96 bath solution that contained (in mM) 96 NaCl, 2 KCl, 1 MgCl₂, and 5 HEPES (pH 7.5). Currents were acquired with a GeneClamp 500 amplifier and pCLAMP software (Axon Instruments, Foster City, CA); the corner frequency was 500 Hz. For selectivity experiments, NaCl was replaced with the Na⁺ salt of each of the anions studied. This resulted in the retention of a residual 4 mM Cl from the K⁺ and Mg²⁺ salts. Data were corrected for junction potentials at the ground bridge (3 M KCl in 3% agar), which ranged from 0.2 to 2.4 mV as determined with a free-flowing KCl electrode. CFTR channels in oocytes were activated by exposure to isoproterenol (Iso) and alternately assayed in the presence of a Cl-containing bath or a substitute anion. Substitutions were always made in the same order and for a 1-min duration. Monovalent test anions, in order of use, included acetate, bromide (Br), gluconate, glutamate, iodide (I), nitrate (NO), isethionate, perchlorate (ClO), and thiocyanate (SCN). Fluoride was excluded from this list due to the low solubility of MgF₂. We also assayed selectivity between Cl and one divalent anion, thiosulfate (S₂O). Exposure to these test anions was kept brief to avoid alteration of cytoplasmic Cl concentration (49). Data for each substitute anion were bracketed with the data for Cl plus Iso before and after the substitute anion (e.g., Cl plus Iso; acetate; Cl plus Iso; Br; Cl plus Iso). Relative permeability [anion x-to-Cl permeability (P_x/P_Cl)] and relative conductance [anion x-to-Cl conductance (G_x/G_Cl)] for each substitute anion (anion x) were calculated with the average data for the preceding and subsequent exposures to Cl. This procedure allowed us to compare the G_x/G_Cl values for several anions by controlling for the changes in activation during the experiment.

View larger version (16K): [in this window] [in a new window]   Fig. 1.   Selectivity of the wild-type (WT) cystic fibrosis transmembrane conductance regulator (CFTR) channel in oocytes. Macroscopic current-voltage (I-V) relationships were generated from currents obtained with hyperpolarizing voltage ramps between +60 and 80 mV. Raw currents are shown before correction for junction potentials. All data are from a single representative oocyte; currents in substitute anions were normalized to the initial current in Cl to account for changes in activation level during the experiment (see MATERIALS AND METHODS). A: currents in the presence of Cl and small monovalent anions. B: the same data for Cl and currents in the presence of large monovalent anions and the divalent anion thiosulfate (S2O). Glut, glutamate; Acet, acetate; Gluc, gluconate; Iseth, isethionate; Vm, membrane potential. Substitution solutions contained 96 mM substitute anion plus 4 mM Cl.

View this table: [in this window] [in a new window]   Table 5.   V in ND96 bath solution for WT and mutant CFTRs

View this table: [in this window] [in a new window]   Table 6.   Selectivity between Cl and the divalent anion S2O

View larger version (14K): [in this window] [in a new window]   Fig. 2.   Lyotropic permselectivity between monovalent anions in WT CFTR. A: relative permeabilities for most anions were determined by the change in Gibbs free energy on hydration [hydration energy (hydrGo)]. Relative permeability (Px/PCl) values for I and ClO were lower than expected from this relationship. B: relative conductance (Gx/GCl) values did not exhibit a clear dependence on hydration energy.

View larger version (14K): [in this window] [in a new window]   Fig. 3.   Gx/GCl data identify anions that bind. The relationship between Gx/GCl and anion radius is plotted for all monovalent anions studied and for the divalent anion S2O. The substitute anions fall into three categories: ions that are highly conductive (on the dashed line), ions that are too large to traverse the pore (on the solid line), and ions that bind in the pore to inhibit conductance of the residual Cl (below the lines). The baseline conductance due to the 4 mM residual Cl in all solutions lies at the level of the solid line. The intersection of the solid and dashed lines suggests a pore diameter of ~5 Å.

View larger version (14K): [in this window] [in a new window]   Fig. 4.   Dependence of macroscopic current on extracellular Cl concentration ([Cl]). Oocytes expressing WT CFTR were studied in ND96 control solution (100 mM Cl) and after partial or complete substitution of Cl with either gluconate or S2O. Fractional current was calculated after a step to +80 mV. Values are means ± SD; n = 5 oocytes. The data in gluconate, which does not block the pore from the extracellular side, suggest that the half-maximal inhibitory concentration for Cl under these conditions is ~5 mM. The decreased conductance in S2O-containing solutions indicates that this anion blocks the current carried by the residual Cl.

View larger version (22K): [in this window] [in a new window]   Fig. 5.   Pore-domain mutations affect the shape of macroscopic I-V curves. A: data from WT CFTR, indicating mild outward rectification under these conditions. B-D: I-V relationships from oocytes expressing mutants at K335, T338 or T339, or S341, respectively, in TM6. E: I-V relationships from oocytes expressing mutants at T1134 in TM12. In each case, data were generated with a hyperpolarizing ramp from 1 representative oocyte in ND96 bath solution.

View larger version (25K): [in this window] [in a new window]   Fig. 6.   Normalized Px/PCl (A) and Gx/GCl (B) for all anions in the alanine-substituted mutants. Px/PCl and Gx/GCl values for each anion x in each mutant were calculated for T1134A, K335A, T338A, T339A, and S341A CFTRs and normalized to Px/PCl and Gx/GCl values for WT CFTR. Solid lines at unity, Px/PCl and Gx/GCl in WT CFTR; dashed lines, ±2 SD of the WT data. Data points were plotted as a function of distance down the pore from the extracellular end at left to the middle of the pore at the right. Points that fall outside the dashed lines represent a large change in Px/PCl or Gx/GCl for that ion. Ions are listed from top to bottom in order of increasing ionic radius.

View larger version (13K): [in this window] [in a new window]   Fig. 7.   Mutations that affect binding of small anions also affect binding of anionic pore blockers. Introduction of an acidic amino acid (glutamic acid) near the binding site for small anions, at S341 in TM6, also destabilizes binding of diphenylamine-2-carboxylate (DPC). Currents carried by WT CFTR or S341E CFTR were measured in the absence of drug and in the presence of 100 (for WT CFTR) or 500 (for S341E CFTR) µM DPC applied to the bathing solution. Shown are fractional currents remaining after 7-min exposure to DPC compared with currents measured in the absence of DPC (I/Io) after steps to membrane potentials between 140 and +80 mV. Values are means ± SD; n = 6 oocytes. WT CFTR exhibited voltage-dependent block by DPC at hyperpolarizing potentials. S341E CFTR was not blocked by DPC, even at high concentrations.

View larger version (20K): [in this window] [in a new window]   Fig. 8.   Relative affinity [(Px/PCl)/(Gx/GCl)] for small monovalent anions for the alanine- and glutamic acid-substituted mutants. Data for each variant were plotted as a function of distance down the pore, with the extracellular end on the left and the middle of the pore on the right. Nos. in parentheses, range of relative affinity values for each variant calculated as the ratio of the highest to the lowest. A: alanine substitutions. B: glutamic acid substitutions. , Acetate; , ClO; , Cl; , I; , Br; , NO3; hexagon, SCN. WT CFTR was very effective at discriminating between monovalent anions. Both alanine and glutamic acid mutations at T338 and S341 diminished the discriminating power of the pore.

Analysis. Reversal potentials (V_rev) for Cl (V) and for each test anion were used to calculate P_x/P_Cl values according to the Goldman-Hodgkin-Katz equation in the following form
where [Cl]_r and [Cl]_t are the concentrations of Cl in the reference and test solutions, respectively; [x]_t is the concentration of anion x in the test solution (96 mM); V_r is the change in V_rev; z is the valence; R is the gas constant; T is the absolute temperature; and F is the Faraday constant (49). Relative chord conductances for anion entry were calculated from the change in current over a voltage range from V_rev to V_rev+25 mV. By analyzing only the outward currents, we isolated the data for currents carried by the mixture of Cl and each substitute anion under known conditions.

Statistical comparisons. Unless otherwise noted, all values are means ± SE. Statistical analysis was performed with the Wilcoxon rank sum test after appropriate tests for normality and variance (SigmaStat, Jandel Scientific, San Rafael, CA). In Tables 1-6, significance is indicated as P  0.001.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Characterization of the WT CFTR pore: lyotropic selectivity and anion binding. We studied CFTR currents in the presence of Cl and nine substitute monovalent anions selected to span a wide range of radii (Table 1). Figure 1 compares currents generated from a hyperpolarizing voltage protocol in solutions containing Cl as the sole anion or solutions containing 4 mM Cl and 96 mM substitute anion for a representative oocyte expressing WT CFTR. V_rev values for currents in the presence of Br, NO, and SCN were more negative than those for currents in the presence of Cl, whereas all other ions studied reversed at potentials depolarized from the V. V_rev data (summarized in Table 1) were used to calculate relative permeabilities for each anion (Table 2). Our results indicate that for most anions, P_x/P_Cl values in WT CFTR were determined according to hydration energies (Fig. 2, Table 3), as expected for a channel with a lyotropic selectivity sequence (39, 41, 45). Two anions, I and ClO, exhibited P_x/P_Cl values lower than expected from this relationship. This behavior was absent in similar studies of the oocyte endogenous Ca²⁺-activated Cl channels (34), indicating that there may be distinctive interactions of these anions with the pore of CFTR. Relative permeabilities to isethionate, glutamate, and gluconate are very low, as if these ions are too large to traverse the pore.

Rationale for the positions studied. For the purpose of this study, we assayed the effects of mutations at five positions chosen on the basis of 1) predictions made by comparing the sequence of TM6 and TM12 with the sequences of ligand-gated anion channels (29, 33), 2) the results of mutations at these positions on blockade by diphenylamine-2-carboxylate (DPC) and/or 5-nitro-2-(3-phenylpropylamino)benzoic acid (NPPB) (33, 50), and 3) previous studies by other investigators (24, 26). In TM6, we studied mutations at K335, T338, and S341, all of which were predicted to face into the pore based on our proposed alignments (29) and conservation of positioning of polar residues. However, on the basis of cysteine-scanning mutagenesis, Cheung and Akabas (5, 6) concluded that T338 does not face the pore. We also studied T339 as a control position; we do not believe that this amino acid faces into the pore based on the insensitivity of blockade by DPC to mutations at this site (33). In TM12, we studied mutations at T1134, predicted to lie at a position slightly more extracellular than K335. Interpretation of the results of these mutations is predicated on the assumption that their effects are limited to the site mutated without propagation to distant sites. However, we recognize the possibility of nonselective effects (see Refs. 1, 4). For instance, the R347D mutation has been shown to have nonspecific effects due to destruction of a salt bridge (8). It is also possible that mutation of a site in TM6 does not affect the interaction of anions with the pore at that site but affects the interaction of anions with the pore at a site in an adjacent TM domain.

Effects of alanine substitution on selectivity between monovalent anions. We have begun to use alanine substitution to determine the effect of the loss of side-chain functionality at each site listed in Rationale for the positions studied to identify the residues that contribute to selectivity in WT CFTR. Alanine-scanning mutagenesis avoids a priori bias regarding the location of residues that are "likely" to contribute to the pore. Because alanine is small and fairly unreactive, introduction of alanine is usually well tolerated in both soluble proteins and integral membrane proteins (e.g., Ref. 40). With alanine substitution, we expected to find a range in the magnitude of changes in P_x/P_Cl. However, the changes should be largest for substitutions in regions of the pore that contribute the most to discrimination between anions.

Effects of glutamic acid substitution on selectivity between monovalent anions. The alanine substitution experiments described in Effects of alanine substitution on selectivity between monovalent anions identified a region of strong discrimination between monovalent anions. Discrimination between anions could result from differences in energy barrier heights (reflected in P_x/P_Cl) or from differences in well depths (reflected in G_x/G_Cl). Because CFTR exhibits a strongly lyotropic selectivity sequence, it may be appropriate to consider that energy barriers to permeation in this channel mostly reflect the energy of dehydration rather than a physical barrier in the pore walls. In contrast, we know that anion binding makes an important contribution to selectivity in CFTR (Fig. 3, Table 4) (26). We reasoned, therefore, that disruption of anion binding by the introduction of a negative charge at specific residues would identify the positions that may determine selectivity by contributing to anion binding sites. Hence we introduced glutamic acid residues at the same sites that we studied with alanine substitution and asked whether the mutant channels were affected in their ability to discriminate between monovalent anions. This model-dependent approach relies on the assumption that placement of a negative charge near anion binding sites would greatly destabilize anion binding, whereas placement of a negative charge at a distance from anion binding sites would have a smaller effect.

Spatial dependence of discriminating power. If S341 and T338 lie in the region of highest discrimination between monovalent anions, we would expect that mutations here would have the greatest effect on the ability of CFTR channels to distinguish between monovalent anions. To test this notion, we determined the relative affinity (18) for each anion in WT CFTR and for each of the alanine and glutamic acid substitution mutants as relative affinity [(P_x/P_Cl)/(G_x/G_Cl)]. Intuitively, relative affinity for ion x could be elevated by mutations that promote anion entry into the channel (reflected as an increase in P_x/P_Cl) and/or promote an increase in anion binding (reflected as a decrease in G_x/G_Cl). For example, SCN accesses the WT pore easily and binds tightly, resulting in a high value for relative affinity (13.4 in WT CFTR). In contrast, acetate experiences a barrier to pore entry and then does not bind well, resulting in a low value for relative affinity (0.3 in WT CFTR). For each CFTR variant, the degree of spread between the lowest value for relative affinity (usually acetate) and the highest value for relative affinity (usually SCN) provides a measure of the discriminating power of the pore. A channel that cannot discriminate between monovalent anions would exhibit a narrow range of relative affinities. We excluded isethionate, glutamate, and gluconate from this analysis because G_x/G_Cl values for these anions are not clearly indicative of their intrinsic behavior due to the conductance arising from residual Cl.

Dependence on the direction of anion movement. Other investigators (41) have described hysteresis in I-V relationships obtained from patches expressing WT CFTR under bi-ionic conditions with I on one surface and Cl on the other. In those experiments, P_I/P_Cl was initially >2 but fell to below unity after time. This change in P_I/P_Cl was dependent on the voltage protocol applied to the patch and was interpreted as a shift from an I-permeant state to an I-blocked state. Given these previous results, we compared our results calculated from hyperpolarizing ramps with data calculated from depolarizing ramps. In WT CFTR, both P_I/P_Cl and P_ClO4/P_Cl were protocol dependent, although not to such a strong degree as described by others (41) (P_I/P_Cl was 0.52 vs. 0.36 and P_ClO4/P_Cl was 0.18 vs. 0.10 for depolarizing vs. hyperpolarizing ramps, respectively). This behavior was retained in K335F and T1134F CFTR but lost in all other mutants. In contrast, mutation T338A induced significant hysteresis for all three of the large anions studied. P_x/P_Cl values for isethionate, glutamate, and gluconate were ~0.13 for depolarizing ramps and ~0.07 for hyperpolarizing ramps. When Cl exit at negative potentials preceded entry of large anions at positive potentials, the P_x/P_Cl values calculated for those large anions were greater than when the voltage protocol was reversed. The depolarizing ramps begin with a 50-ms step to 80 mV, which induces strong inward current (Cl exit) and should result in significant loading of Cl into the pore from the cytoplasmic solution. The hyperpolarizing ramps began with a 50-ms step to +60 mV, which should induce depletion of Cl from the pore. The observed shift in P_x/P_Cl for large anions suggests that occupancy by Cl of an anion binding site that is cytoplasmic to the pore constriction reduced the energy barrier height for entry of large anions from the extracellular side.

Effects of mutations on monovalent-to-divalent selectivity. Very little is known about the selectivity between monovalent and divalent anions in CFTR because no polyvalent anions other than ATP⁴ (23) have been studied in this channel. We have begun to identify the determinants of selectivity according to valence by studying CFTR currents in the presence of thiosulfate (S₂O). When the ND96 bath solution was exchanged for one containing 96 mM Na₂S₂O₃ (with all else remaining equal), both inward and outward currents in WT CFTR were greatly reduced (Fig. 1) and the V_rev shifted to the right to 22.52 ± 1.98 mV (n = 16). P_S2O3/P_Cl was very low under these conditions and was highly variable in both WT CFTR and the CFTR variants. On the other hand, G_S2O3/G_Cl was well above zero (Table 6) and could be reliably measured in WT CFTR and the CFTR variants. Hence we focused on the effects of mutations in terms of G_x/G_Cl alone. S₂O has a diameter that is very close to the predicted minimal pore diameter (Fig. 3) and yet the conductance in S₂O-containing solutions fell below the baseline conductance carried by the residual Cl. Hence G_S2O3/G_Cl values in WT CFTR reflected a significant degree of pore block by this anion. In confirmation, we found that substitution of S₂O had a greater concentration-dependent impact on macroscopic conductance than did a similar substitution with the nonconductive, nonblocking anion gluconate (Fig. 4).

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Several approaches have been used to identify amino acids in CFTR that contribute to anion selectivity. However, most studies have not been performed in a manner allowing comparison of the effects of mutagenesis at multiple positions in the pore. A single mutation almost always has a measurable effect on at least one parameter of selectivity. To facilitate interpretation, experimental designs that allow one to look for patterns avoid the potential pitfall of generating conclusions based on the effects of mutations at a single position. By studying the effects of a limited number of substitutions made at several positions, one can look for patterns in the aggregate results, which may provide information about the structure of the pore in the WT channel. This study provides proof of concept because we report the effects of two classes of mutations at five positions in the CFTR channel. The magnitude of the effects of alanine or glutamic acid substitution should be largest at positions that contribute most strongly to selectivity. These data allow us to identify a region of strong discrimination between monovalent anions and between monovalent and divalent anions, which appears to reside between T338 and S341 in TM6. Selectivity between monovalent anions was most sensitive to mutation at T338 or S341 depending on the anion. Selectivity between Cl and a divalent anion was most sensitive to mutation at S341. K335 and T1134, near the predicted ends of TM6 and TM12, respectively, were relatively insensitive to mutation. These observations are consistent with a narrowing of the pore from the extracellular end toward a constriction near S341.

Amino acids involved in permeation in CFTR. Several individual amino acids in the TM domains of CFTR have been shown to contribute to permeation properties such as single-channel conductance (21, 24, 33, 36, 37, 41, 42, 47, 49), interaction with pore blockers (20-22, 26, 33, 44, 49, 50), anion or cation selectivity (6, 14), and selectivity between monovalent anions (3, 19, 22-24, 26, 41, 49). These studies have focused on TM6, where several amino acids appear to play important roles in establishing the character of the CFTR pore (for a review, see Ref. 29). It should be pointed out that all studies attributing a functional role to R347 must be reconsidered because substitution of this amino acid with anything other than lysine or histidine grossly disrupts the conformation of the channel (8). K335 in TM6 appears to make only indirect contributions to selectivity between small monovalent anions (3, 26; present study); however, this study shows that K335 does directly contribute to selectivity between Cl and large anions. When the alanine in K335A CFTR was replaced by the bulkier phenylalanine, the effects on anion selectivity were greater (Tables 2 and 4). Glutamic acid substitution here had limited effects on discrimination between anions (Fig. 8). These observations suggest that the pore may be wide at the cross-sectional level of K335 such that anions do not make intimate contact with the walls at this putative end of the pore. K335 may lie near the innermost edge of the outer vestibule and not within a region of strong discrimination between anions. Alanine and phenylalanine mutations at T1134 in TM12 had effects very similar to those of mutations at K335, suggesting that this amino acid also may contribute only indirectly to anion selectivity. T1134 is the first amino acid in TM12 of CFTR for which selectivity data have been described after mutation. In most cases, the effects of equivalent mutations were somewhat smaller at T1134 than at K335; this is consistent with our hypothesis that T1134 projects into the pore at a level slightly more extracellular than does K335.

The importance of anion binding. One significant difference between our approach and that of some other investigators (e.g., Ref. 23) is that we can ascertain the effects of mutations on G_x/G_Cl as well as on P_x/P_Cl because we made measurements in the presence of Cl and substitute anions in the same cell and we can correct for changes in the activation status of the CFTR channels. Most other studies (e.g., Ref. 19) of selectivity in CFTR have only reported the effects of mutations on V_rev values; single-channel conductance (or G_x/G_Cl) was not reported for most mutants. Anion binding, which affects G_x/G_Cl, is known to be an important determinant of selectivity in CFTR (9, 26). Consistent with this notion, we found that for small anions, G_x/G_Cl values were generally more prone to change on mutation at T338 and S341 than were P_x/P_Cl values (note the differences in ordinate scales for Fig. 6, A vs. B). This was true for both alanine and glutamic acid substitutions at T338 and S341. In contrast, for ions larger than SCN, P_x/P_Cl values were more sensitive to mutations than G_x/G_Cl values. Superficially, this observation appears to be inconsistent with the work by Mansoura et al. (26). However, those investigators only studied the effects of mutations on P_x/P_Cl and G_x/G_Cl of ions as large as I or smaller, and they did not study mutations in TM6 at positions cytoplasmic to K335. Hence, given the aggregate picture, anion binding is clearly important for selectivity between Cl and small anions but less important for selectivity between Cl and large anions.

Distributed determinants of selectivity. It has been suggested by others (19) that the CFTR pore includes a discretely localized selectivity filter, which confers on the CFTR channels most, if not all, aspects of anion selectivity. Alternatively, lyotropic permselectivity may be a characteristic imposed on the CFTR pore by virtue of distributed interactions of different anions with several points along the length of the pore (9). If this were the case, one might expect to find regions of the pore that are highly sensitive to mutation and other regions that are less sensitive; our results suggest that this is the case. One might also expect to find that some anions may be most sensitive to mutation at one position, whereas other anions are more sensitive to mutation at different positions. This also appears to be indicated by our alanine substitution data that show that the relative affinities for NO and Br were affected most by mutations at S341, the relative affinity for SCN was equally affected by mutations at S341 and T338, the relative affinities for I and ClO were affected most by mutations at T338, and the relative affinity for acetate was equally sensitive to mutations at S341 and T338. This does not appear to reflect a dependence on anion size because the results differed for ClO and acetate, which are very similar in radius. Further evidence against a discrete selectivity filter is the greater than sixfold increase in relative affinity for I exhibited by T1134F CFTR. These data suggest that the pore contains features that provide for selectivity between anions at several positions along its length. Furthermore, mutation of a non-pore lining residue could grossly alter pore structure along the full length of the helix, thereby affecting selectivity indirectly (19, 24).

Reconciling the results of alanine and glutamic acid substitutions. The alanine substitution experiments had the greatest impact on anion selectivity at T338 (Fig. 6). In contrast, the effects of glutamic acid substitution were greatest at position S341 (Fig. 8). This discrepancy likely arises from the fact that alanine substitutions represent loss-of-function mutations where the side chain present in the WT protein is replaced by the methyl side chain of alanine, whereas glutamic acid substitutions represent gain-of-function mutations that introduce a point-negative charge. Hence the electrostatic consequence of a glutamic acid substitution at S341 may reach well past this locus, whereas the effects of an alanine substitution may be more localized. Also, it may be true that ions more closely approach a charge placed at S341 than a charge placed at T338 by virtue of the constriction, leading to greater destabilization of anion binding in S341E CFTR. Our results with K335E (outside the region of high discrimination) and S341E (inside the region of high discrimination) show that this interpretation is feasible.

Localizing the narrowest point in the pore. By comparing the effects of a common mutation at multiple positions, as shown in Figs. 6 and 8, we can interpret these data in a structural context that allows us to make predictions about the shape of the pore. It is clear that the largest effects of alanine and glutamic acid substitutions are found at T338 and S341, suggesting that the narrowest region of the pore lies in the vicinity of these two positions. If these amino acids are separated by one turn of the -helix, then selectivity between some ions is very strong over this ~5.4-Å distance. We can localize the narrowest point even further by considering the following scenarios and predictions from their effects on permeation properties.

Evidence against the dual-pore hypothesis. Yue et al. (46) recently proposed that the amino-terminal half of CFTR forms one Cl conducting pore and the carboxy-terminal half forms a separate pore with distinct anion selectivity and conductance. The amino-terminal pore was suggested to contribute to the main conductance state, whereas the carboxy-terminal pore contributed only to a subconductance state. If this were the case, mutations at comparable positions in TM6 and TM12 should not have similar effects on selectivity in macroscopic currents. However, our data show that mutations K335A and T1134A had nearly identical effects on selectivity patterns between large and small anions, as if these amino acids occupy nearly homologous positions in TM6 and TM12. This similarity of effects argues strongly that TM6 and TM12 contribute to the same pore, in contradiction of the one-channel, dual-pore model. Other observations against the dual-pore hypothesis include the ability to transfer the DPC binding site from TM6 to TM12, which resulted in affinity and voltage dependence for block by DPC that was very similar to that of WT CFTR (33). This would not be the expected result if the second-membrane-spanning domain only contributed to the subconductance state as the dual-pore model proposes.

Structural predictions for the pore of CFTR. CFTR, like other anion channels, does not exhibit strong selectivity. It may be generally true that channels with pores lined by -helices are poorly selective (CFTR, -aminobutyric acid type A receptor, and nicotinic ACh receptor channels), whereas channels lined by combinations of -helices and -strands, such as the prototypical voltage-gated K⁺ channel (11), which exhibit 100-fold discrimination between ions that differ in radius by <0.5 Å (16), are much more selective. Because the lyotropic selectivity pattern in CFTR suggests that the barriers to permeation reflect ion-water interactions rather than ion-channel interactions, whereas anion binding appears to play a critical role in anion selectivity in this channel, studies that identify anion binding sites in CFTR may be very informative. We have suggested that S341 contributes to a Cl binding site (33); T338 may contribute to another. These observations suggest that hydroxylated residues may play very important roles in coordinating anions in the permeation pathway of the CFTR channel. We originally proposed this concept based on studies of ligand-gated cation and anion channels (33). We are now fortunate to be able to interpret our results in CFTR in relation to the recently published crystal structure of halorhodopsin, the light-driven anion pump of bacteria (17). Only five TMs in halorhodopsin are required to contain the Cl and a protonated Schiff base of retinal, which serves as a cofactor. There is no evidence of contributions to Cl binding from -strands or carbonyl dipoles. The only amino acid in halorhodopsin that directly interacts with Cl in the binding pocket is a serine. In fact, the interaction between the hydroxyl group in this serine, Cl, and two waters of hydration contributes 8.9 of the 21.6 kcal/mol of the stabilization energy, far greater than the contribution by any other component. These observations in halorhodopsin add credibility to our proposal that hydroxylated amino acids make important contributions to anion binding and selectivity in the pore of CFTR.

Limitations of this study and future directions. A few limitations of this study are obvious. First, as always, the effects of any one mutation may not be specific but may reflect a structural change experienced at some distance from the site of mutation (1, 4). However, there was no evidence of gross structural changes because the Cl currents in oocytes retained their dependence on activation by cAMP-dependent pathways, the currents were time independent, and most mutations did not affect all selectivity patterns. Second, because the results of all mutations described in this study were presented as P_x/P_Cl, G_x/G_Cl, or relative affinity, it is impossible to determine directly whether the mutations affect changes in the interaction of Cl with the pore. To answer this question, we will have to perform single-channel experiments in WT CFTR and in those variants that exhibited large changes in selectivity. Finally, although we have separately described the effects of mutations on P_x/P_Cl and G_x/G_Cl, there is some interdependence of these two parameters. The use of the Goldman-Hodgkin-Katz equation to estimate P_x/P_Cl implicitly assumes that the behavior of one ion is independent of the behavior of another ion or of the occupancy of the channel by any ion. This assumption fails for CFTR and for any other channel with multiple ion-binding sites (42) because tight binding of one anion would be expected to increase the mean occupancy of the binding sites in the pore and, thereby, alter the permeability of another anion. However, this approach is intuitively useful and has been widely used for assessing selectivity in a broad range of channels (9). Furthermore, changes in relative affinity should not be impacted by this interdependence if the concentration of substitute anions is held constant, as was the case in our experiments.