Contribution of proline residues in the membrane-spanning domains of cystic fibrosis transmembrane conductance regulator to chloride channel function.

Proline residues located in membrane-spanning domains of transport proteins are thought to play an important structural role. In the cystic fibrosis transmembrane conductance regulator (CFTR), the predicted transmembrane segments contain four prolines: Pro99, Pro205, Pro324, and Pro1021. These residues are conserved across species, and mutations of two (P99L and P205S) are associated with cystic fibrosis. To evaluate the contribution of these prolines to CFTR Cl- channel function, we mutated each residue individually to either alanine or glycine or mutated all four simultaneously to alanine (P-Quad-A). We also constructed the two cystic fibrosis-associated mutations. cAMP agonists stimulated whole cell Cl- currents in HeLa cells expressing the individual constructs that resembled those produced by wild-type CFTR. However, the amount of current was decreased in the rank order: wild-type CFTR = Pro324 > Pro1021 > Pro99 >/= Pro205 mutants. The anion selectivity sequence of the mutants (Br- >/= Cl- > I-) resembled wild-type except for P99L (Br- >/= Cl- = I-). Although the Pro99, Pro324, and Pro1021 mutants produced mature protein, the amount of mature protein was much reduced with the Pro205 mutants, and the P-Quad-A made none. Because the Pro99 constructs produced mature protein but had altered whole cell currents, we investigated their single-channel properties. Mutant channels were regulated like wild-type CFTR; however, single-channel conductance was decreased in the rank order: wild-type CFTR >/= P99G > P99L >/= P99A. These results suggest that proline residues in the transmembrane segments are important for CFTR function, Pro205 is critical for correct protein processing, and Pro99 may contribute either directly or indirectly to the Cl- channel pore.

The cystic fibrosis transmembrane conductance regulator (CFTR) 1 (1) is a regulated Cl Ϫ channel (for reviews see Refs. 2 and 3). Several studies employing site-directed mutagenesis have shown that residues in the two membrane-spanning domains (MSDs) contribute to the formation of the Cl Ϫ channel pore. Mutation of specific basic residues within MSD1 (Lys 95 , Arg 117 , Arg 334 , Lys 335 , and Arg 347 ) altered the channel's anion selectivity, single-channel conductance, and/or gating behavior (4 -6), suggesting that those residues contribute to the channel pore. Akabas and collaborators (7) employed sulfhydryl-specific reagents and cysteine scanning mutagenesis to identify residues in M1 (Gly 91 , Lys 95 , and Gln 98 ) and to suggest that M1 has an ␣-helical structure. In addition, data from McDonough and collaborators (8) suggest that diphenylamine-2-carboxylic acid inhibits the channel by interacting with residues in M6 (Ser 341 ) and M12 (Thr 1134 ).
One approach to identifying additional residues that contribute to the CFTR Cl Ϫ channel pore is to mutate residues that play a key functional role in other ion channels and transporters. Brandl and Deber (9) observed that proline residues are frequently found in the transmembrane segments of ion channels and transporters but not in the transmembrane segments of proteins that have no transport function. Based on that observation, they suggested that proline residues located in the MSDs of transport proteins have a key functional role. However, proline residues are not favored in ␣-helices because the backbone nitrogen is not available for hydrogen bonding and because of steric constraints caused by their ring structure (10). As a result, prolines introduce "kinks" in transmembrane ␣-helices (10). Proline-kinked ␣-helices may pack to form either funnel-or cage-shaped structures, which could form either a channel vestibule or ion binding site(s) (11).
CFTR contains four proline residues in the predicted transmembrane segments: Pro 99 , Pro 205 , Pro 324 , and Pro 1021 (Fig. 1). These residues are conserved across species and mutations of two (P99L and P205S) are associated with cystic fibrosis (CF) (12,13). Because P99L and P205S are associated with a milder clinical phenotype characterized by retention of some pancreatic function (termed pancreatic sufficiency) (2), these mutants may retain some residual Cl Ϫ channel function. Based on the above information, the aims of this study were to determine the contribution of proline residues located within the MSDs of CFTR to Cl Ϫ channel function and to understand how P99L and P205S cause a loss of Cl Ϫ channel function. after transfection for protein expression studies or assayed 8 -30 h after transfection for Cl Ϫ channel function. Mutants are named by their single-letter amino acid code and their position within the CFTR sequence followed by the specific mutation. In P-Quad-A, alanines are substituted for prolines at amino acid positions 99, 205, 324, and 1021.
Electrophysiology-Whole cell and single-channel currents were recorded as described previously (5,19). Experiments were conducted at 34 -36°C. The established sign convention was used throughout. Liquid junction potentials and potentials at the tip of the patch-pipette were measured, and current-voltage (I-V) relationships were corrected for the corresponding offset.
For whole cell experiments the pipette (intracellular) solution contained (in mM): 120 N-methyl-D-glucamine, 85 aspartic acid, 3 . For whole cell experiments, patch-pipettes had resistances of 2-4 M⍀, and series resistance was 4 -10 M⍀. CFTR Cl Ϫ currents were activated with cAMP agonists (10 M forskolin, 100 M 3-isobutyl-1-methylxanthine, and 500 M 8-(4-chlorophenylthio)-cAMP sodium salt. To determine anion selectivity, 140 mM NaCl in the extracellular (bath) solution was replaced with 140 mM NaBr and 140 mM NaI in the continued presence of cAMP agonists. Permeability ratios, P X /P Cl , where X is Br Ϫ or I Ϫ , were calculated from reversal potential (E rev ) measurements using the Goldman-Hodgkin-Katz equation as described previously (4). Chord conductance was measured as the slope between E rev and E rev plus 25 mV as described previously (4). Conductance ratios, G X /G Cl , where X is Br Ϫ or I Ϫ , are the ratio of chord conductance of Br Ϫ or I Ϫ relative to that of Cl Ϫ . Whole cell currents, filtered at 0.5 kHz, and digitized at 2 kHz were not capacitance-or leakage-subtracted.
Reagents-Recombinant vaccinia virus (vTF7-3) was from American Type Culture Collection (Rockville, MD). cAMP-dependent protein kinase was from Promega (Madison, WI) and [␥-32 P]ATP was from Du-Pont NEN. MgATP, Na 2 ATP, 8-(4-chlorophenylthio)-cAMP sodium salt, forskolin, and 3-isobutyl-1-methylxanthine were from Sigma. All other chemicals were of reagent grade. Statistics-Results are expressed as the means Ϯ S.E. of n observations. To compare mean values, we used Student's t test. Differences were considered statistically significant when the p value was Ͻ0.05.

Expression of Proline Mutants Generates cAMP-activated
Cl Ϫ Currents-To examine the contribution of prolines in the MSDs to CFTR Cl Ϫ channel function, we mutated each proline individually to either alanine or glycine. We chose alanine and glycine because they might be expected to have different effects (10). Alanine has a small side chain and is prevalent in transmembrane ␣-helices. Because it lacks a side chain, glycine has significant conformational freedom and is frequently found in hinge regions. Like proline, glycine is not favored in transmembrane ␣-helices.
We expressed the mutants in HeLa cells and studied their function using the whole cell patch-clamp technique. Cyclic AMP agonists activated whole cell currents in cells expressing each of the individual proline to alanine or glycine mutations and in cells expressing the CF-associated mutations, P99L and P205S. As an example, Fig. 2 shows data from studies of P99A, P99G, and P99L; qualitatively similar results were obtained with the Pro 205 , Pro 324 , and Pro 1021 mutants (data not shown). All of the proline mutants had whole cell properties that resembled those of wild-type CFTR (2,3). Under basal conditions there was little or no whole cell current. Whole cell currents were reversibly activated by cAMP agonists, were time-and voltage-independent ( Fig. 2A), had linear I-V relationships, and were selective for anions over cations (Fig. 2B).
Because the MSDs of CFTR contribute to the Cl Ϫ conducting pore and control anion selectivity and because proline residues may have an important structural role, we speculated that these proline residues might contribute directly or indirectly to the anion selectivity filter. We tested this hypothesis by examining the anion permeability and conductance sequence of cAMP-stimulated whole cell currents. The anion selectivity sequence of whole cell currents in wild-type CFTR is Br Ϫ Ն Cl Ϫ Ͼ I Ϫ (Table I and Ref. 4). Most proline mutants had anion selectivity sequences qualitatively similar to that of wild-type CFTR ( Fig. 3 and Table I). However, the CF-associated mutation P99L had an altered anion selectivity sequence, Br Ϫ Ն Cl Ϫ ϭ I Ϫ (Fig. 3 and Table I). Despite the change in anion selectivity, all of the mutant channels retained their selectivity for anions over Na ϩ (for example see Figs. 2B and 3).
Because many mutations in CFTR have reduced cAMP-stimulated Cl Ϫ currents compared with wild type (5, 23), we measured the change in whole cell Cl Ϫ current that was stimulated by cAMP agonists. Fig. 4 shows that the Pro 99 and Pro 205 mutants generated Ͻ30% of wild-type Cl Ϫ current. Moreover, the CF-associated mutants retained Ͻ15% of wild-type Cl Ϫ current. The amount of Cl Ϫ current generated by the Pro 324 mutants was indistinguishable from that of wild type, whereas the Pro 1021 mutants retained intermediate amounts of Cl Ϫ current (Fig. 4). Uninfected HeLa cells and cells infected with vTF7-3 as a control did not generate cAMP-activated Cl Ϫ currents (Fig. 4). In cells expressing P-Quad-A, no increase in current was observed under either basal or cAMP-stimulated conditions (Fig. 4). This suggests that either P-Quad-A does not form a functional Cl Ϫ channel, or it is severely misprocessed.
Analysis of the Processing of Proline Mutants-To determine why the proline mutants tended to generate less Cl Ϫ current, we studied the production of mature CFTR in HeLa cells by analyzing the glycosylation status of CFTR protein (14). Wild- type CFTR produces substantial amounts of mature, fully glycosylated protein (band C), whereas the CF-associated mutant ⌬F508 produces little mature protein. Fig. 5A shows that like wild-type CFTR, the Pro 99 , Pro 324 , and Pro 1021 mutants produced the band C form of CFTR. However, band C production was much reduced with the Pro 205 mutants and the P-Quad-A mutant. Fig. 5B shows the relative amount of CFTR present in the mature form at 16 h after transfection. Production of mature protein was reduced in the rank order: wild-type CFTR Ͼ Pro 1021 mutants ϭ Pro 324 mutants Ͼ Pro 99 mutants Ͼ Pro 205 mutants Ն P-Quad-A ϭ ⌬F508.
Single-channel Properties of the Pro 99 Mutants-Because the Pro 99 mutants produced some mature protein, yet had altered whole cell properties, we examined their single-channel properties using excised, inside-out membrane patches. Like wildtype CFTR, mutant Cl Ϫ channels were reversibly activated by phosphorylation with cAMP-dependent protein kinase and required intracellular MgATP to open (Fig. 6A) (2, 3). However, the single-channel current amplitudes of the Pro 99 mutants were decreased compared with wild-type CFTR (Fig. 6B). Al-  were significantly decreased at 4.66 Ϯ 0.25 pS (n ϭ 5, p Ͻ 0.0001) and 4.97 Ϯ 0.24 pS (n ϭ 5, p Ͻ 0.0001), respectively. We could not determine the reversal potential of the mutant channels because of their small current amplitudes. Because the single-channel current amplitudes of P99A and P99L were much reduced and because in most cases the patches of membrane contained large numbers of channels, we could not accurately measure single-channel kinetics. However, visual inspection suggested that the gating behavior of the Pro 99 mutants was not dramatically different from wildtype CFTR with bursts of activity containing brief flickery closures separated by longer closures between bursts (Fig. 6B). DISCUSSION Our data indicate that proline residues located in the predicted transmembrane segments of the MSDs are important for CFTR Cl Ϫ channel function. Pro 205 , which lies in the middle of a putative ␣-helix, is critical for correct protein processing. Pro 99 , which is near the external surface of CFTR, may contribute either directly or indirectly to the Cl Ϫ channel pore.
Proline residues in the transmembrane segments could af- CFTR was immunoprecipitated, phosphorylated, and analyzed by SDSpolyacrylamide gel electrophoresis. Bands B and C are indicated by the arrows. B, relative amounts of total CFTR present in the mature form (band C/band B). The relative amount of band C production by ⌬F508 is indicated by the dashed line. The radioactivity in gels of immunoprecipitated and phosphorylated wild-type and mutant CFTR was quantitated as described under "Experimental Procedures." Similar results were obtained in two separate experiments. fect structure in at least two ways. First, proline residues can form cis peptide bonds, and the energy barrier to cis-trans isomerization about the peptide bond preceding proline is reduced compared with other residues (10). Cis-trans isomerization of peptide bonds could produce conformational changes and thereby regulate channel activity (9). However, because in CFTR the proline mutants showed no dramatic changes in gating, cis-trans isomerization of peptide bonds involving prolines in the MSDs may not be an important determinant of CFTR gating behavior.
Second, as mentioned above, when proline occurs within the interior of a transmembrane ␣-helix, it disrupts the pattern of hydrogen bonding causing the ␣-helix to kink (10). The packing together of proline-kinked ␣-helices may contribute to the for-mation of either a channel vestibule or ion binding site(s) (11). For example, the pore of melittin is built from four prolinekinked ␣-helices (24).
Effect of Proline Mutations on CFTR Biosynthesis-None of the proline mutants were processed as efficiently as wild-type CFTR. We saw no clear correlation between the amount of band C produced by mutation of individual prolines and the propensity of the amino acids to occur in ␣-helices (A Ն L Ͼ G Ն P; Ref. 10). Pro 324 and Pro 1021 appear to be less critical for correct protein processing because other amino acids could be accommodated at these positions. In contrast, Pro 205 , which is predicted to lie in the middle of M3, is crucial for correct protein folding and processing. All Pro 205 mutants tested showed significant processing defects. Thus, Pro 205 mutants may pro- foundly disrupt the normal association and packing of transmembrane segments that would disrupt the processing of CFTR and its delivery to the cell membrane. The defect in processing explains why the P-Quad-A failed to generate an increase in whole cell current under either basal or cAMPstimulated conditions.
Prior to this study, all CF-associated mutations known to manifest a processing defect were located within the nucleotide-binding domains (14,23,25). The finding that a mutation in a transmembrane sequence can disrupt processing has a precedent in the T cell antigen receptor and viral glycoproteins (26,27). Moreover, in the related transport protein, human P-glycoprotein, mutations in M7, including P709 located at the intracellular end of M7, disrupt protein processing (28,29).
Effect of Proline Mutations on Cl Ϫ Channel Function-The whole cell properties of the individual proline mutants resembled those of wild-type CFTR. Thus, although the proline mutations affected biosynthesis, they did not produce sufficient disruption of structure to abolish Cl Ϫ channel function. In this regard they are similar to the CF mutations A455E and P574H that disrupt processing but generate channels that retain significant activity (23). Nevertheless, some changes in the properties of the mutant channels were observed. When Pro 99 was mutated to leucine, the channel lost its ability to discriminate between Cl Ϫ and I Ϫ . Interestingly, substitution of alanine and glycine at Pro 99 did not alter anion selectivity. We speculate that leucine, with its bulky side chain, may directly or indirectly disrupt the conformation of an anion binding site located near Pro 99 and alter anion selectivity. In contrast, alanine and glycine with their shorter side chains may be more easily accommodated at this position and do not significantly affect anion selectivity.
Measurements of single-channel conductance also suggest that Pro 99 contributes directly or indirectly to the formation of the Cl Ϫ channel pore. Substitution of alanine, glycine, and leucine at Pro 99 decreased single-channel conductance in the rank order: wild-type CFTR Ն P99G Ͼ P99L Ն P99A. Interestingly, the propensity for these amino acids to occur in ␣-helices follows the reverse order (A Ն L Ͼ G Ն P; Ref. 10). Based on these observations, we speculate that Pro 99 kinks M1 to form part of the channel structure. Glycine, which is found in hinge regions and like proline is not favored in ␣-helices, can substitute for proline at this residue because the single-channel conductance of P99G does not differ from wild type. In contrast, we speculate that alanine and leucine, residues that are prevalent in ␣-helices, eliminate the kink in M1 caused by Pro 99 , thereby reducing the access of permeant ions into the channel pore and decreasing single-channel conductance. Because P99C did not react with sulfhydryl-specific reagents, Akabas and collaborators concluded that Pro 99 does not line the channel pore (7). Our data and their data are compatible if Pro 99 contributes to pore architecture without itself lining the pore where it would be accessible to the hydrophilic sulfhydryl blockers.
Implications for Cystic Fibrosis-P99L and P205S are CF mutations located in MSD1 that are associated with a milder (pancreatic sufficiency) clinical phenotype (12,13). Our studies of the processing and function of P99L and P205S explain why these mutants generate less Cl Ϫ current than wild-type CFTR. Loss of Cl Ϫ channel function caused by P205S was predominantly a result of defective protein processing; whereas that caused by P99L was a consequence of both defective protein processing and altered Cl Ϫ channel function. Our results suggest that these mutations are associated with a milder (pancreatic sufficiency) clinical phenotype because a small amount of mutant protein is processed correctly and generates cAMP-activated CFTR Cl Ϫ currents. When we have studied the ⌬F508 mutant (associated with a severe clinical phenotype) under similar conditions we found no Cl Ϫ current. The mutant ⌬F508 is defectively processed in both native epithelia and heterologous cells (14,22). This suggests that the defective processing of the P205S mutant observed in HeLa cells likely accounts for the loss of Cl Ϫ channel function in patients bearing this mutation. We previously speculated that pharmacological therapies designed to increase the activity of mutant channels in the plasma membrane might be useful for treating patients bearing these mutations associated with a milder clinical phenotype (5). Potential pharmacological therapies may include phosphatase inhibitors such as bromotetramisole and novel Cl Ϫ channel openers such as the substituted benzimidazolone NS004 that have been demonstrated to activate mutant Cl Ϫ channels in recombinant cells (30,31).
The present results complement and extend our previous study of mild CF mutants located in MSD1 (R117H, R334W, and R347P). We showed that these mutants form Cl Ϫ channels with altered permeation properties but are processed normally (5). The mechanism of dysfunction of P205S resembles that of the nucleotide-binding domain 1 pancreatic sufficiency mutants A455E and P574H, which are misprocessed (23). Interestingly, P99L forms a Cl Ϫ channel with altered pore properties and is also misprocessed. Thus, these results demonstrate that the mechanisms by which CF mutations produce defective Cl Ϫ channels are complex and that it is not possible to predict the mechanism of dysfunction of CFTR based solely on the site of mutation.