|Home | About | Journals | Submit | Contact Us | Français|
Inactivation of the Pseudomonas aeruginosa mutM, mutY, or mutT gene conferred a 2.4-, 17.2-, or 38.1-fold increase in spontaneous mutation frequency, respectively. Importantly, the mutY and mutT strains each displayed a robust H2O2-induced mutation frequency. In addition, the mutM, mutY, and mutT mutations severely sensitized P. aeruginosa to killing by H2O2, suggesting that these gene products act to repair one or more cytotoxic lesions in P. aeruginosa. Nucleotide sequence analysis of a fragment of the rpoB gene from rifampicin resistant mutM-, mutY-, and, mutT-deficient strains was consistent with this conclusion. These findings are discussed in terms of possible roles for mutM, mutY, and mutT in contributing to survival and mutagenesis of P. aeruginosa colonizing the airways of cystic fibrosis patients.
Pseudomonas aeruginosa is an opportunistic human pathogen that is especially proficient at infecting the airways of individuals afflicted with the autosomal recessive disease cystic fibrosis (CF). As a result of inflammation, P. aeruginosa colonizing CF airways is exposed to remarkably high levels of reactive oxygen species (ROS). ROS includes superoxide (O2−), hydrogen peroxide (H2O2), and the extremely reactive hydroxyl radical (HO•). It is hypothesized that ROS-induced DNA damage contributes to mutations that afford an adaptive advantage upon P. aeruginosa inhabiting the thick mucus lining the CF airway (Ciofu, et al., 2005). One of the most common ROS-induced DNA lesion is 7, 8-dihydro-8-oxo-dGuanine (8-oxo-dG or GO). Mutagenesis mediated by 8-oxo-dG results from misinsertion of dA opposite template–8-oxo-dG during DNA replication. If left uncorrected, the misinserted dA is used as template during the subsequent round of replication, leading to a GC→TA transversion mutation (Shibutani, et al., 1991, Maki & Sekiguchi, 1992). In addition, certain error prone DNA polymerases (Pols) are reported to incorporate oxidized precursor deoxynucleotide triphosphates, including 8-oxo-dG, into nascent DNA, contributing to mutations (Yamada, et al., 2006, Maga, et al., 2007). E. coli Pol IV is reported to incorporate 8-oxo-dG opposite template-dA in vitro (Yamada, et al., 2006) and, importantly, P. aeruginosa possesses a Pol IV ortholog (Sanders, et al., 2006). In E. coli, 8-oxo-dG is thought to be identified and efficiently repaired primarily via the base excision repair pathway, which includes the GO system (Michaels & Miller, 1992), the Nei (endonuclease VIII) protein (Asagoshi, et al., 2000, Hazra, et al., 2000), as well as possibly the DNA mismatch repair system (Wyrzykowski & Volkert, 2003).
The E. coli GO repair system is comprised of the mutM, mutY, and mutT genes (Michaels & Miller, 1992). The mutM gene encodes a formamidopyridine-DNA (Fapy) glycosylase that removes 8-oxo-dG inserted opposite dC, mutY encodes an adenine glycosylase that removes dA opposite 8-oxo-dG, while mutT encodes an 8-oxo-dGTP hydrolase that cleaves 8-oxo-dGTP to 8-oxo-dGMP, effectively ‘cleansing’ the dNTP pools. E. coli strains deficient in any one of these genes display an increased spontaneous mutation frequency on the order of 10- to over 100-fold as compared to the wild type control strain (Michaels & Miller, 1992). Based on DNA sequence analysis, mutations observed in the E. coli GO mutants result almost exclusively from GC→TA (Cabrera, et al., 1988, Nghiem, et al., 1988) or AT→CG transversions (Yanofsky, et al., 1966).
Expression of the P. aeruginosa mutM (PA5147), mutY (PA0357), or mutT (PA4400) orthologs from a plasmid complements the mutator phenotype of the corresponding mutant E. coli strains (Oliver, et al., 2002). These results were interpreted to suggest that P. aeruginosa possesses a GO system similar to that described for E. coli (Oliver, et al., 2002). In striking contrast to E. coli, inactivation of mutT in growing P. putida conferred only a 2-fold effect on spontaneous mutation frequency (Saumaa, et al., 2007). The authors suggested that the mutT strain might fail to display a robust mutator phenotype due to the presence of multiple putative mutT-like homologs in P. putida (Saumaa, et al., 2007). Like P. putida, P. aeruginosa also possesses at least 12 mutT-like homologues based on its annotated genome (http://www.pseudomonas.com). Moreover, based on its annotated genome (http://www.pseudomonas.com), P. aeruginosa appears to lack a nei ortholog, which together with the GO system plays an important role in repairing oxidized DNA lesions in E. coli (Friedberg, et al., 2006). Thus, the extent to which the GO system acts to prevent mutations in P. aeruginosa is presently unclear. The purpose of the work discussed in this report was to define the phenotypes for P. aeruginosa strains deficient in mutM, mutY, or mutT function. A complimentary study by Morero and Argaraña (Morero & Argarana, 2009) was published while this report was under review.
Strains were routinely grown in Luria-Bertani (LB) medium. When necessary, rifampicin (Rif) was used at 100 µg/ml, and tetracycline (Tet) was used at 60 µg/ml. Strains MPAO1 (wild type parent), MPA39575 (mutY::ISphoA/hah), MPA16280 (mutM::ISlacZ/hah), MPA35623 (mutT::ISphoA/hah), and MPA52426 (mutT::ISphoA/hah) were obtained from Dr. Michael Jacobs of the University of Washington Genome Center (Jacobs, et al., 2003). The transposon insertion in each strain was confirmed by diagnostic PCR using a protocol provided by Dr. Jacobs. Briefly, diagnostic PCR reactions were performed using primers homologous to the 5’- or 3’-end of the mutY, mutM, or mutT genes paired with a primer homologous to the transposon inserted within the gene of interest (specific primers are described below). In addition, separate PCR reactions were performed to verify the absence of an intact copy of each respective gene. For these reactions, we used primers specific to the 5’- and 3-ends of the mutY, mutM, and mutT genes, respectively. The following primers were used: PAmutMF (5’-end of mutM); 5’-GAT CAC CAT ATG CCC GAA CTA CCC GAA G-3’; PAmutMR (3’-end of mutM); 5’-GCT CAT AAG CTT TCA GCG TTG GCA G-3’; PAmutYF (5’-end of mutY), 5’-GAT CTA CAT ATG ACA CCT GAA GGC TTC AAC-3’; PAmutYR (3’-end of mutY), 5’-GAT CAT TAA TAA GCT TTC ACG CGG CCG TGC-3’; PAmutTF (5’-end of mutT); 5’-GAG CAC CAT ATG AAA CGA GTA CAT GTC GCC GCG-3’; PAmutTR (3’-end of mutT), 5’-GAT TAT AAG CTT TCA AAG GCC GCC CGG CCA G-3’; hah minus 138 (specific to the ISphoA/hah transposon), 5’-CGG GTG CAG TAA TAT CGC CCT-3’; and lacZ (specific to the ISlacZ/hah transposon), 5’-GGG TAA CGC CAG GGT TTT CC-3’.
Single colonies of P. aeruginosa were used to inoculate five ml of LB (for MPAO1), or LB supplemented with Tet (for mutT, mutY, or mutM strains). Following growth at 37°C for ~16 h, appropriate dilutions of each culture were spread onto LB plates with or without Rif. Spontaneous mutation frequencies were calculated by dividing the median number of RifR cells present in each culture by the total number of viable cells in the same culture. The Mann-Whitney Test was used to determine whether observed differences between strains were statistically significant (Hall & Henderson-Begg, 2006).
Strains were grown in LB medium supplemented with Tet to an OD595 of ~0.5. Cells from one ml of culture were collected by centrifugation, resuspended in sterile 0.8% NaCl, and immediately treated with the indicated concentration of H2O2 (Sigma-Aldrich). After 15 min, H2O2 was removed by collecting the cells by centrifugation and washing them with one ml of sterile 0.8% NaCl. For determination of H2O2-induced killing, appropriate dilutions of washed cells were immediately spread onto LB plates, followed by incubation overnight at 37°C. For determination of H2O2-induced mutation frequency, 4.5 ml of fresh LB was inoculated with 0.5 ml of washed cells, and cultures were grown for 16 h at 37°C. Appropriate dilutions of each culture were plated onto LB plates with or without Rif. H2O2-induced mutation frequency was calculated by subtracting the spontaneous frequency of RifR from that observed following H2O2-treatment.
Spontaneous and H2O2-induced (25 mM H2O2) RifR clones were selected as described above. One hundred independent RifR clones of each strain were co-inoculated into LB medium supplemented with both Tet and Rif, and grown for ~16 h at 37°C. Genomic DNA was isolated from each pool of clones using a kit from Promega, as per the manufacturer’s recommendations. A 250 bp fragment of the rpoB gene (PA4270) encompassing amino acid residues 499–582 was PCR amplified from each DNA sample using barcoded primers (see below). More than 90% of all identified rpoB mutations that confer RifR in E. coli map to this region (Garibyan, et al., 2003). PCR reactions contained 50 ng purified genomic DNA, 1.25 U Pfu-Ultra thermostable DNA polymerase (Stratagene), 250 nM each dNTP, 10% DMSO, and 100 ng of each primer. Reactions were denatured at 94°C for 10 min, then amplified for 35 cycles at 94°C for 30 sec, 55°C for 30 sec, and 68°C for 30 sec. For each reaction, a single Primer A (A1-A6) was paired with Primer B1: Primer A1, 5’-GCC TCC CTC GCG CCA TCA GAG CGC GCC AAG CCG GTG GCT GCC-3’; Primer A2, 5’-GCC TCC CTC GCG CCA TCA GAT CCC GCC AAG CCG GTG GCT GCC-3’; Primer A3, 5’-GCC TCC CTC GCG CCA TCA GGC TGC GCC AAG CCG GTG GCT GCC-3’; Primer A4, 5’-GCC TCC CTC GCG CCA TCA GAG AGC GCC AAG CCG GTG GCT GCC-3’; Primer A5, 5’-GCC TCC CTC GCG CCA TCA GTG CAC GCC AAG CCG GTG GCT GCC-3’; Primer A6, 5’-GCC TCC CTC GCG CCA TCA GAT GCC GCC AAG CCG GTG GCT GCC-3’; Primer B1, 5’-GCC TTG CCA GCC CGC TCA GGG TCG CCA GGG AGT TGA TCA GAC C-3’. Sequences in bold correspond to the conserved region to which the amplicon sequencing primers annealed, underlined sequences correspond to the unique barcode present in each Primer A variant, and sequences in italics correspond to P. aeruginosa rpoB coding sequence.
Amplicons were purified with AMPure Beads (Agencourt, Beverly, MA) using the standard the FLX-454 Amplicon DNA Library protocol. Amplicon quality was assessed using a BioAnalyzer DNA 1000 LabChip (Agilent, Santa Clara, CA) to ensure that the product was of the expected size and free of primer dimers and other extraneous products. The amplicons were quantified by fluorometry using the Quant-iT PicoGreen dsDNA Assay Kit (Invitrogen, Carlsbad, CA) and a TBS-380 Fluorometer (Turner BioSystems, Sunnyvale, CA) to ensure that the samples were pooled in equal-molar ratios, resulting in an equal representation of all samples in the amplification and sequencing reactions. Amplicons were diluted to a concentration of 2 × 105 molecules/µL and then pooled together in equal volumes of 100µL, producing a 600uL pooled amplicon library at a concentration of 2 × 105 molecules/µL.
A FLX-454 emulsion titration was performed to determine the optimal quantity of the pooled library to use during the amplification step (emulsion PCR) to maximize the number of high quality reads possible from the sequencing run. The amplification step for the sequencing run was performed using FLX-454 Emulsion PCR Kit II, the amplified products were loaded onto a FLX-454 2-Region 70X75 Pico Titer Plate and run on the FLX-454 Instrument using a Long Read 70X75 Sequencing Kit.
Initial image and signal processing was performed directly on the FLX-454 instrument. The 4 nucleotide tag was used to separate DNA sequences into six individual pools of corresponding fasta (.fna) and quality (.qual) files. The tags were removed and sequences aligned to rpoB by BLASTALL to identify polymorphisms within the P. aeruginosa rpoB gene.
In order to determine the contribution of mutM, mutY, or mutT function in preventing mutations in P. aeruginosa, we analyzed isogenic derivatives of strain MPAO1 bearing transposon insertions within these genes (Jacobs, et al., 2003). Since the translated MutM and MutY proteins are 71% and 67% similar to their E. coli orthologs, respectively, we selected a single transposon insertion for each (see Table 1). In contrast, the P. aeruginosa mutT gene is predicted to encode a 315 residue bifunctional protein: the deduced N-terminal 127-amino acids of MutT are 37% identical to the 129-amino acid E. coli MutT protein, and are reportedly capable of complementing an E. coli mutT strain (Oliver, et al., 2002), while the C-terminal domain shows strong similarity to thiamine monophosphate synthase based on a cluster of orthologous groups prediction (www.pseudomonas.com). We therefore selected two different mutT strains for phenotypic analysis. One bears an insertion within the C-terminal thiamine monophosphate synthase-like domain at residue 298, while the other bears the transposon at residue 19 within the N-terminal 8-oxo-dG hydrolase domain (see Table 1).
We first measured the relative contributions of mutM, mutY, or mutT function to spontaneous mutagenesis in exponentially growing P. aeruginosa using a RifR assay. RifR results exclusively from base substitution mutations within the essential rpoB gene, which encodes the β subunit of RNA polymerase (Jin & Gross, 1988). As summarized in Table 1, the mutM strain displayed a 2.4-fold higher frequency of RifR compared to the wild-type MPAO1 strain, while the mutY strain displayed a 17.2-fold increase. The mutT strain bearing the transposon within its C-terminal thiamine monophosphate synthase-like domain was indistinguishable from the wild type control, while the mutT strain bearing the transposon within its N-terminal 8-oxo-dGTP hydrolase domain displayed a 38.1-fold higher frequency of RifR as compared to MPAO1 (Table 1). This finding is consistent with results from Oliver and colleagues indicating that the N-terminal domain of the P. aeruginosa mutT ortholog was both required and sufficient for complementation of the spontaneous mutator phenotype of the E. coli mutT::Km strain (Oliver, et al., 2002). Furthermore, this result indicates that the 12 putative mutT-like genes identified in the P. aeruginosa genome are unable to fully complement loss of the canonical MutT ortholog (PA4400) with respect to spontaneous mutagenesis. However, it remains to be determined whether loss of any one or more of the putative mutT-like genes influences mutation frequency.
Loss of mutM or mutY function in E. coli leads to an increase in the frequency of spontaneous GC→TA transversions (Shibutani, et al., 1991, Friedberg, et al., 2006), while loss of mutT results in an elevated frequency of AT→CG transversions (Maki & Sekiguchi, 1992, Friedberg, et al., 2006). In order to determine whether the P. aeruginosa mutant strains behaved similarly, we analyzed the nucleotide sequence of the region of rpoB encoding residues 499–582 from 100 independent RifR isolates for each of the three mutant strains. More than 90% of the known base substitution mutations conferring RifR in E. coli map to this same region (Garibyan, et al., 2003). Importantly, each of the mutations that we identified in the P. aeruginosa rpoB gene map to residues that are conserved in E. coli, and known to confer RifR when mutated (Garibyan, et al., 2003). As summarized in Table 2, three different types of base substitutions were identified in the mutM strain, including CG→TA (57.1%) and AT→GC (17.4%) transitions, and AT→TA (25.5%) transversions. The lack of GC→TA transversions may be due to the fact that the spontaneous mutation frequency was elevated only ~2.4-fold in the mutM strain compared to the wild type strain (Table 1). 98.7% of the mutations identified in the mutY strain were GC→TA transversions, consistent with a role for P. aeruginosa MutY protein in removing dA paired with 8-oxo-dG (Table 2). AT→GC transitions accounted for the remaining 1.3% of the observed mutations. Finally, spontaneous RifR in the mutT strain resulted solely from GC→AT transitions (Table 2). It is possible that one or more of the 12 putative mutT-like gene products act to prevent AT→CG transversions. Nevertheless, results summarized in Tables 1 and and22 are consistent with the idea that MutM (PA5147), MutY (PA0357), and MutT (PA4400), as well as possibly one or more mutT-like gene products comprise a GO-like system in P. aeruginosa.
P. aeruginosa colonizes the airways of CF patients (Gomez & Prince, 2007). Since neutrophil levels are reported to be ~1,500-fold higher in the infected CF lung than in uninfected individuals (Konstan, et al., 1994), we were interested in determining whether GO function contributed to survival of P. aeruginosa following exposure to ROS. We first measured sensitivity of the different P. aeruginosa strains to elevated levels of O2. For this experiment, we bubbled 100% O2 through LB medium using sterile tubing fed into vented culture tubes housed in a 37°C shaking water bath. These tubes were inoculated through the vent, and growth in the presence of a continuous supply of 100% O2 was monitored by changes in optical density at 595 nm. As shown in Figure 1A, growth of the mutM, mutY, and mutT strains was indistinguishable from that of the wild type MPAO1 parent. We next asked whether GO function contributed to survival of P. aeruginosa following exposure to H2O2. As shown in Figure 1B, the wild type MPAO1 strain displayed a dose-dependent H2O2 sensitivity. Importantly, the mutM and mutY strains were severely sensitive to H2O2, displaying a more than 1,000-fold decrease in survival relative to the wild type MPAO1 strain following exposure to 300 mM H2O2 for 15 min. The mutT strain was ~5- to ~10-fold more sensitive to H2O2 than the mutM or mutY strains, and ~10,000-fold more sensitive than the wild type control after treatment with 300 mM H2O2 (Figure 1B). These results indicate that the mutM, mutY, and mutT gene products play a pivotal role in promoting survival of P. aeruginosa following exposure to ROS. In addition, since 8-oxo-dG is promutagenic, these findings suggest that the P. aeruginosa GO system acts to repair one or more cytotoxic lesions.
In light of their severe H2O2-sensitivity, we next asked whether the P. aeruginosa GO mutants displayed an elevated H2O2-induced mutator phenotype as compared to the wild type MPAO1 strain. As shown in Figure 1C, the wild type MPAO1 strain displayed a dose-dependent H2O2-induced mutator phenotype. For example, exposure of strain MPAO1 to 100 mM H2O2 for 15 min resulted in a ~6-fold increase in the frequency of RifR relative to the untreated control. An even more pronounced H2O2-induced mutator phenotype was observed for the GO mutants. Although the mutM strain mirrored MPAO1 at H2O2 concentrations up to and including 25 mM, it displayed a ~3-fold increase in the frequency of RifR following exposure to higher H2O2 concentrations (Figure 1C). The mutY strain displayed a ~5-fold increase in the frequency of RifR compared to the wild-type strain at H2O2 concentrations equal to or greater than 2.5 mM (Figure 1C). Finally, the mutT strain had the highest frequency of RifR of all GO mutants examined, irrespective of the H2O2 concentration at all concentrations examined. Following exposure to 100 mM H2O2, the mutT strain displayed a more than 12-fold increase in the frequency of RifR compared to the MPAO1 control (Figure 1C). Taken together, results summarized in Figure 1 highlight the critically important role played by MutM, MutY, and MutT in protecting P. aeruginosa against ROS-induced killing and mutagenesis.
In order to distinguish between roles for MutM, MutY, and/or MutT in repairing one or more cytotoxic lesion, and the possibility that 8-oxo-dG is cytotoxic to P. aeruginosa, we sequenced 100 independent H2O2-induced RifR isolates from each mutant. If 8-oxo-dG is cytotoxic to P. aeruginosa, then most mutations conferring RifR should correspond to GC→TA and AT→CG transversions. Alternatively, if these proteins act to repair H2O2-induced lesions in addition to 8-oxo-dG, then other types of mutations should be observed. Analysis of the types and locations of mutations conferring RifR yielded two important conclusions. First, in every case, the position of a significant fraction of the total mutations identified following treatment with H2O2 differed from those observed under spontaneous conditions (Table 2). Second, the types of mutations following H2O2 treatment differed from those observed spontaneously. For example, following H2O2 treatment of the mutM strain, the overwhelming majority of the identified mutations differed from those observed under spontaneous conditions. 90.5% were CG→GC transversions, while 5.3% were CG→AT transversions (Table 2). The remaining 4.2% represented AT→TA tranversions, which were observed at a higher frequency (25.5%) under spontaneous conditions (Table 2). The CG→AT transversions were consistent with a role for MutY in repairing 8-oxo-dG lesions, while the other types of mutations suggest that MutY may recognize one or more additional lesion(s). Similar results were observed with the mutY and mutT strains. 89.9% of the mutations identified in the mutY strain were CG→AT transversions, consistent with 8-oxo-dG, while 6.8% were GC→AT transitions, and 3.2% were CG→GC transversions and may result from other types of lesions (Table 2). 27.3% of the mutations identified in the mutT strain were AT→CG transversions, consistent with the hypothesized role of MutT in cleansing the nucleotide pools of 8-oxo-dGTP, while 3.5% were GC→TA transversions, also consistent with a role for 8-oxo-dG (Table 2). The remaining mutations corresponded to either GC→AT (63.1%) or AT→GC (6.1%) transitions, which may result from lesions other than 8-oxo-dG.
Several reports in the literature describe altered mutation frequencies among clinical isolates of various pathogenic bacteria, including P. aeruginosa isolated from CF airways [(Oliver, et al., 2000, Gutierrez, et al., 2004, Ciofu, et al., 2005, Macia, et al., 2005) reviewed in (Hall & Henderson-Begg, 2006)]. In the case of P. aeruginosa, these strains typically display mutation frequencies anywhere between lower than that observed for common laboratory isolates of the wild type PAO1 strain, to more than 1,000-fold higher than PAO1. In these types of studies, spontaneous mutation frequencies are typically measured using a standard fluctuation assay similar to that described in Table 1. Strains displaying a 20-fold or greater increase in spontaneous mutation frequency are generally referred to as ‘hypermutable’ (Oliver, et al., 2000). However, the phenotypes of the P. aeruginosa mutM and mutY mutants described in this report illustrate that it is also important to consider induced mutation frequency when screening for hypermutable clones: based on its observed spontaneous mutation frequency (Table 1), a mutT mutant could theoretically be identified. Thus, previous studies that focused solely on identifying hypermutators based on spontaneous mutation frequency alone may have missed an important population of clones capable of displaying a DNA-damage induced hypermutable mutator phenotype in vivo. This idea is particularly relevant to P. aeruginosa, which colonizes and persists within the highly ROS-rich CF airways. Indeed, it has been previously suggested that DNA damage-induced mutagenesis may contribute importantly to pathoadaptation (Metzgar & Wills, 2000, McKenzie & Rosenberg, 2001, Ciofu, et al., 2005).
Our findings that loss of mutM, mutY, or mutT function significantly enhances both H2O2 sensitivity as well as the frequency of H2O2-induced mutagenesis raises the question of whether the increased mutation frequency of these strains compensates for their enhanced sensitivity in the context of CF airways. Although mutM-, mutY-, and mutT-deficient strains would likely be impaired for colonization of CF airways, due to their remarkable sensitivity to H2O2 (Figure 1B), their enhanced mutation frequency may help to compensate for their severe sensitivity by increasing the probability of acquiring one or more adaptive mutations. The adaptive value of these GO deficiencies may be higher within certain airway niches. For example, the elevated spontaneous mutation frequency of mutY- and mutT-deficient strains may provide an adaptive advantage within biofilms, which would protect them from H2O2, thus enabling their contribution to genetic variation and adaptation, without the associated cost of their increased H2O2 sensitivity. However, the fact that these mutants, particularly mutT, are not prevalent among hypermutable clones isolated from CF patients (Oliver, et al., 2000, Gutierrez, et al., 2004, Ciofu, et al., 2005, Macia, et al., 2005) implies that their H2O2 sensitivity may outweigh the adaptive value of their enhanced mutation frequency. A more definitive answer to this important question may be obtained using murine models of airway infection to study the colonization and airway persistence phenotypes of these and related mutant strains. Finally, we have previously proposed that error-prone DNA replication by Pol I, PolC, and Pol IV likewise contribute to mutations that may confer an adaptive advantage upon P. aeruginosa inhabiting CF airways (Sanders, et al., 2006). It is important to determine whether mutations induced by these Pols result from their ability to bypass ROS-induced lesions, and/or their ability to incorporate oxidized nucleotide triphosphate precursors into nascent DNA, particularly in mutM, mutY, or mutT-deficient strains.
The authors would like to thank Dr. Michael Jacobs of the University of Washington Genome Center for generously providing P. aeruginosa strains, Dr. Steven Gill, Dr. Anthony Campagnari, and Jennifer Jamison of The University at Buffalo, SUNY, and the Center of Excellence in Bioinformatics and Life Sciences, for assistance with the amplicon sequencing, the three expert reviewers for their insightful comments, as well as the members of our laboratory for numerous helpful discussions. This work was supported by Pilot and Feasibility grant SUTTON07I0 from the Cystic Fibrosis Foundation and Public Service Health grant GM066094, awarded to MDS.