Native CFTR codon bias controls translation rate to balance off-pathway aggregation and channel function by conformational imprinting

Codon usage in the NBD1 α-helical subdomain affects CFTR aggregation propensity

The NBD1 β-sheet core subdomain contains four parallel β-strands (S3, S6, S7, and S8: residues 453–458, 488–491, 568–572, and 599–603, respectively) buried within the hydrophobic interior of the domain (Fig. 1A)^10,17. During NBD1 synthesis, these four β-strands are translated non-contiguously: S3 and S6 are synthesized first, then, after the α-helical subdomain, S7 and S8. Thus, correct β-sheet folding requires insertion of S7 and S8 between the preexisting S3 and S6 following synthesis of the α-helical subdomain. Our previous studies using fluorescence resonance energy transfer (FRET)^18–20 showed that these cotranslational folding events are temporally coupled with translation, facilitating efficient β-sheet formation. Codon usage within residues 525–593, which span the α-helical and β-sheet core subdomains, is enriched in codons predicted to have low translation efficiency (Table S1), designated here as ‘Low-efficiency’ or “Low” codons, due to limited isoacceptor tRNA availability and wobble base-pairing^19,21. Synonymous substitutions with ‘High-efficiency’ (or “High”) codons, defined as those predicted to have high translation efficiency, resulted in increased aggregation propensity of newly synthesized CFTR protein¹⁹. Metabolic pulse-chase labeling experiments revealed that the increased aggregation observed in CFTR with High-efficiency codons, although overall protein processing efficiency remained unchanged, is likely due to misfolded intermediates being more prone to aggregation or to delayed clearance of aggregates, rather than an increase in misfolding during synthesis¹⁹.

To better understand how codon usage influences this critical window of NBD1 cotranslational folding, we divided the 525–593 region into three subregions: segment-a (525–552), segment-b (553–572), and segment-c (573–593) (Fig. 1B–D). We introduced High-codons into each segment, ensuring that the number of synonymous codon substitutions per construct was comparable (18–20 codons) (Table S1). WT and High-codon constructs were exogenously expressed in HEK293 cells, lysed in radioimmunoprecipitation assay (RIPA) buffer, and separated into soluble supernatant and insoluble pellet by centrifugation. Insoluble pellets containing aggregated CFTR were solubilized in 1% (w/v) sodium dodecyl sulfate (SDS) for subsequent analysis as previously described¹⁹. β-actin and co-expressed GFP were used as loading and transfection controls, respectively. Consistent with our previous findings¹⁹, all High-codon constructs showed similar levels of soluble, mature CFTR protein (band C) (Figs. 1B,C and S1), indicating that global processing efficiency was not impaired. By contrast, the amount of insoluble, immature CFTR protein (band B) varied depending on the segment (Figs. 1B,C and S1). Notably, constructs bearing High-codons within segment-b (abc, ab, bc, and b-alone) consistently showed increased aggregation of immature CFTR relative to WT. These findings suggest that codon usage within segment-b includes an important determinant of the aggregation of immature CFTR protein.

To assess the relationship between CFTR codon usage and translation rate, we performed ribosome profiling of WT and High-codons in segment-abc constructs (GEO: GSE305005). Ribosome footprint libraries from mRNA fragments protected by actively translating ribosomes (Fig. S2)^35,36 were prepared without translation inhibitors to avoid any potential bias in ribosome occupancy³⁷. Footprints were mapped onto the CFTR coding sequence (CDS) based on A-site codon identity (see “Methods”) to generate the 15-amino acid moving average of ribosome footprint occupancy across NBD1 (Fig. 1D). When compared to WT, the CFTR construct with High-codons in segment-abc showed a substantial reduction in footprint density within segment-b. As ribosome footprint occupancy is inversely correlated with the relative translation rate (i.e. a slower rate results from increased ribosome dwell time at specific codons), this reduction suggests that High-efficiency codon substitutions elicit a transient increase in translation rate in segment-b. These results support the idea that native codon usage within this region slows translation and that disruption of these decoding kinetics may contribute to the observed increase in immature CFTR aggregation.

Translational pauses occur at specific codons during CFTR synthesis

To examine how individual codons affect translation rate across the CFTR CDS, we assessed codon-level ribosome footprint occupancy through ribosome profiling. We exogenously overexpressed the WT CFTR construct in HEK293 cells to obtain sufficient ribosome footprints for this single-transcript analysis. This approach yielded an average of ~ 10,500 footprints per sample preparation across five independent replicates (range: about 2600–33,000). Mapping of ribosome footprints displayed a characteristic triplet periodicity along the CFTR CDS (Fig. 2A), consistent with transcriptome-wide patterns (Fig. S2), confirming that the data reflect actively translating ribosomes³⁸.

Ribosome occupancy was highly non-uniform across the CFTR CDS, with certain codons exhibiting substantially higher occupancies than others, characteristic of transient translational pauses (Fig. 2B). For this analysis, we arbitrarily defined a translational pause as codons with footprint occupancy greater than tenfold above the transcript-wide average. This criterion identified eleven discrete pause sites within the 1,480-amino acid residue CFTR CDS, nine of which occurred at either Leu(UUA) or Glu(GAA) codons. To further explore codon-specific translational pauses across the CFTR CDS, we visualized ribosome occupancy across all 61 sense codons (Fig. 2C). While UUA and GAA were overrepresented among pause sites, most instances of these codons, 31 out of 37 for UUA and 60 out of 63 for GAA, were not identified as pauses. These findings suggest that while certain codons might preferentially induce a translational pause, sojourns at a given codon are not absolute and are likely influenced by additional factors beyond codon identity^25,32,39.

Transcript-encoded pauses at Leu558 reduce CFTR aggregation

One of the prominent translational pauses across the CFTR CDS occurred at Leu558(UUA), located within segment-abc of NBD1 (Fig. 2B) and the site of a CF-causing CFTR variant (L558S), which transiently disrupts NBD1 decoding kinetics²⁰. To investigate the effect of synonymous codon substitutions on ribosome occupancy within this region, we analyzed the CFTR construct with High-codons in segment-abc, in which 57 out of 69 codons (~ 83%) were substituted with High-efficiency codons (Fig. 3A and Table S1). While most substitutions had minimal impact on occupancy relative to their WT counterparts (Figs. 3A and S3A,B), six residues (Asp537, Leu558, Ala559, Ala561, Lys564, and Thr582) showed statistically significant changes. Among these six residues, Ala561 exhibited increased occupancy, whereas the others showed decreased occupancy, consistent with augmented ribosome decoding at those codons. Notably, the most dramatic change occurred at Leu558, where the UUA(native)-to-CUC(High) codon substitution eliminated the translational pause (Figs. 3A and S3A,B).

To assess the role of this pause, we first introduced a single synonymous UUA-to-CUC codon substitution at Leu558 in the WT construct. Consistent with prior observations (Figs. 1B,C, 2B, and 3A), this substitution eliminated the native pause (Fig. S3C) without affecting CFTR band C processing (Figs. 3B, lanes 1–2; 3C and S4). Although Leu558(CUC) alone caused a modest increase in aggregation of immature CFTR (Figs. 3B, lanes 5–6; 3C and S4), reintroduction of the native UUA codon at Leu558 into the CFTR construct with High-codons in segment-abc partially restored the pause (Fig. S3D) and markedly reduced the aggregation of immature CFTR by this construct (Figs. 3B, lanes 7–8; C and S4). These findings indicate that the native translational pause at Leu558 contributes to preventing aggregation of immature CFTR, but that additional synonymous codons within segment-abc also modulate this effect. Together, these results provide further evidence that transcript-encoded codon bias can cooperate to tune NBD1 folding and reduce early off-pathway aggregation during CFTR biogenesis.

Synonymous substitutions affect the conformational properties of newly synthesized NBD1

We asked whether High-codon substitutions might locally alter the conformational properties of NBD1, even though such subtle changes do not significantly impact global CFTR processing efficiency. To test this possibility, we performed immunoprecipitation (IP) assays on newly synthesized NBD1 to assess potential differences in epitope accessibility, using three well-characterized monoclonal CFTR antibodies, 3G11, 7D12, and UNC660. These antibodies recognize defined epitopes in the N-terminal (epitope residues 396–405), α-helical (epitope residues 531–540), and β-sheet core (epitope residues 576–585) subdomains, respectively (Fig. 4A)^40,41.

NBD1 constructs containing WT and either High- or Low-codons in segment-abc were tested. The construct with Low-codons served as a negative control, as it has negligible effects on protein expression, processing, and aggregation¹⁹. HEK293 cells heterogeneously expressing NBD1 (residues 389–673, ~30 kDa) were radiolabeled and subsequently lysed, and the RIPA-soluble lysates were equally divided for IP with each antibody. The IP results revealed that WT NBD1 and that with Low-codons in segment-abc showed similar levels of protein recovery across all three antibodies (Figs. 4B, lanes 1–2, 4–5, and 7–8; C and S5), indicating no major difference in epitope accessibility. By contrast, NBD1 with High-codons in segment-abc exhibited increased recovery with each antibody (Figs. 4B, lanes 3, 6, and 9; C and S5), consistent with elevated expression due to enhanced translation efficiency¹⁹. Interestingly, when the recovery of NBD1 with High-codons in segment-abc was normalized to that of WT NBD1 (3G11: Fig. 4B, lane 3 to 1; 7D12: Fig. 4B, lane 6 to 4; UNC660: Fig. 4B, lane 9 to 7), the recovery with 7D12 was disproportionately higher than with either 3G11 or UNC660 (Fig. 4C), suggesting increased accessibility of the 7D12 epitope in the α-helical subdomain. These results suggest that High-codons in segment-abc alter the cotranslational folding pathway of NBD1, resulting in a local conformational change within the α-helical subdomain.

Conformational properties established during cotranslational folding persist throughout CFTR maturation

We next tested whether the increased 7D12 reactivity observed in the isolated NBD1 domain also occurs in full-length CFTR. CFTR constructs with either WT, High-, or Low-codons in segment-abc were transiently expressed in HEK293 cells, solubilized in RIPA buffer, and subjected to IP using two monoclonal antibodies: 7D12 (specific to NBD1) and M3A7 (specific to NBD2, epitope residues 1365–1390). As the M3A7 epitope resides in NBD2, it served as a control for overall protein expression. Lysates were equally divided for IP with either 7D12 or M3A7, followed by immunoblotting with 3G11 (Figs. 5A and S6A).

Using both 7D12 (Fig. 5A, lanes 1–2) and M3A7 (Fig. 5A, lanes 4–5), CFTR constructs with either WT or Low-codons in segment-abc showed similar recovery of both immature (band B) and mature (band C) CFTR protein (Fig. 5B), consistent with previous results¹⁹. As expected, recovery of all three constructs was comparable with M3A7 (Fig. 5A, lanes 4–6; B), confirming equivalent expression levels. However, the CFTR construct with High-codons in segment-abc showed significantly increased recovery with the 7D12 antibody for both immature (band B) and mature (band C) CFTR protein (Fig. 5A, lane 3; B), indicating that the altered conformation of 7D12 epitope, established during NBD1 synthesis, is preserved throughout CFTR maturation.

To determine whether the codon substitutions directly affect the 7D12 epitope itself (residues 531–540), we examined the CFTR construct with High-codons in segment-bc, which contains WT codons for segment-a (residues 525–552; where the 7D12 epitope resides) and High-codons for the downstream segment-bc (Figs. 1D and 4A). As with the CFTR construct with High-codons in segment-abc (Fig. 5A,B), this CFTR construct showed increased 7D12 reactivity (Figs. 5C,D and S6B), suggesting that the change likely occurs in close temporal proximity as segment-a emerges from the ribosome during translation of segment-bc. These observations imply that synonymous codon changes can affect cotranslational events in a manner that imprints conformational properties retained in the mature CFTR protein.

Cotranslational imprinting enhances CFTR channel function

During CFTR channel gating, conformational transitions are regulated by ATP-driven NBD1-NBD2 heterodimerization and ATP hydrolysis-catalyzed dissociation of the NBD dimer, which correspond to channel opening and closing, respectively^42,43. Because the 7D12 epitope in NBD1 is located near the NBD dimer interface^10,17 (Fig. 6A), we tested whether the altered 7D12 epitope observed in the CFTR construct with High-codons in segment-abc might affect channel gating at the single-channel level. Functional analyses comparing WT and the CFTR construct with High-codons in segment-abc were independently performed by two research groups with researchers blinded to construct identity. Using different experimental conditions, the two groups studied HEK293 (Fig. 6B–D) and CHO-K1 (Fig. 6E–G) cells heterologously expressing CFTR constructs with the excised inside-out membrane patch configuration of the patch-clamp technique.

Single-channel current amplitudes were indistinguishable between WT and the CFTR construct with High-codons in segment-abc using the different cell types and experimental conditions (Fig. 6C and F), indicating comparable ion conductance (Fig. 6C). However, notable differences were observed in channel gating properties despite the different experimental conditions used by the two research groups, which led to different open probability (P_o) values^44–46. In both HEK293 and CHO-K1 cells, the CFTR construct with High-codons in segment-abc had a noticeably higher P_o than WT in the presence of ATP (3 mM) (Fig. 6D and G). Consistent with this finding, analysis of ATP concentration–P_o relationships revealed that the CFTR construct with High-codons in segment-abc displayed enhanced ATP sensitivity compared with WT CFTR, as reflected by its increased ATP affinity (lower ATP dissociation constant (K_D)) and efficacy (higher maximal open probability (P_o(max))) (WT: K_D = 281 µM, P_o(max) = 0.35, r² = 0.99; High-abc: K_D = 177 µM, P_o(max) = 0.46, r² = 0.94; determined from Michaelis–Menten fits to mean data). At all ATP concentrations tested, the enhanced ATP sensitivity of the CFTR construct with High-codons in segment-abc was manifested by increased frequency and duration of channel openings (Fig. 6E and G). Thus, these results demonstrate that synonymous codon usage in a specific region of NBD1 modulates the conformation of mature CFTR protein to enhance its gating properties.

Plasmid constructs, cell culture, and transient expression

Synonymous codon substitutions were introduced into the mammalian expression plasmids pcDNA3-CFTR and pcDNA3-CFTR NBD1⁴ by PCR overlap extension with synthetic oligonucleotides as previously described¹⁹. All modified regions were verified by Sanger sequencing.

Human embryonic kidney (HEK) 293 cells (RRID:CVCL_0045) and Chinese hamster ovary K1 (CHO-K1) cells (RRID:CVCL_0214) were obtained from the American Type Culture Collection (ATCC; HEK293: Cat# CRL-1573, CHO-K1: Cat# CCL-61) and maintained at 37 °C in a humidified atmosphere of 5% CO₂. HEK293 cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal calf serum, 100 U ml⁻¹ penicillin, and 100 µg ml⁻¹ streptomycin (Life Technologies), and CHO-K1 cells were cultured in Ham’s F-12 medium with the same supplements.

Cells were transiently transfected with plasmid DNA using Lipofectamine (Life Technologies). For immunoblotting, immunoprecipitation, and metabolic labeling experiments, HEK293 cells were harvested 24 h (NBD) or 48 h (CFTR) post-transfection. Transfection efficiency was confirmed by co-transfection with pEGFP-N3 (Clontech) and GFP immunoblotting. For single-channel studies, HEK293 or CHO-K1 cells were co-transfected with CFTR constructs and GFP and 36–60 h later GFP-positive cells were selected for study using the patch-clamp technique.

Immunoblotting and immunoprecipitation

Cells were lysed in 600 μl of ice-cold RIPA buffer [50 mM Tris–HCl/pH 8.0, 150 mM NaCl, 1 mM EDTA, 1% (v/v) Triton X-100, 0.5% (w/v) sodium deoxycholate, 0.1% (w/v) SDS] supplemented with cOmplete protease inhibitors (Roche). Lysates were incubated on ice for 15 min and centrifuged at 16,000 × g for 15 min at 4 °C. The RIPA-insoluble pellet was rinsed with Dulbecco’s phosphate-buffered saline (PBS; Invitrogen), resuspended in 20 μl of 20 mM Tris–HCl/pH 7.4 containing 1% (w/v) SDS, incubated at room temperature for 20 min, and then diluted with 180 μl RIPA buffer.

For immunoprecipitation, RIPA-soluble lysates were incubated overnight at 4 °C with one of the following antibodies: rat anti-CFTR 3G11 (CFTR Folding Consortium), mouse anti-CFTR 7D12 (CFTR Folding Consortium), mouse anti-CFTR UNC660 (CFTR Antibody Distribution Program of the Cystic Fibrosis Foundation), or mouse anti-CFTR antibody M3A7 (Millipore, Cat# MAB3480, RRID:AB_2081244). Immune complexes were captured with Affi-Gel Protein G beads (Bio-Rad) for 2 h at 4 °C, washed three times with RIPA buffer and twice with Tris-buffered saline [20 mM Tris–HCl/pH 7.5, 137 mM NaCl], and then prepared for SDS–polyacrylamide gel electrophoresis (PAGE).

Proteins were separated by SDS-PAGE and transferred to polyvinylidene fluoride (PVDF) membranes (Millipore). Membranes were probed with primary antibodies (rat anti-CFTR 3G11, mouse anti-CFTR M3A7, rabbit anti-β-actin (Cell Signaling Technology, Cat# 4967, RRID:AB_330288), or rabbit anti-GFP (Clontech, Cat# 632592, RRID:AB_10013427), followed by appropriate secondary antibodies. Immunoblots were visualized using the Odyssey Infrared Imaging System (LI-COR) or ChemiDoc XRS+ System (Bio-Rad) and analyzed with the accompanying image analysis software.

Metabolic radiolabeling

HEK293 cells were transiently transfected and incubated for 24 h. Cells were rinsed once with 37 °C PBS, incubated in methionine- and cysteine-free medium for 30 min, and then labeled with 30 μCi Trans³⁵S-label (MP Biomedicals) per well for 15 min. Radiolabeling was terminated by rinsing cells with ice-cold PBS, followed by snap-freezing in liquid nitrogen. Frozen cells were thawed at room temperature and lysed in RIPA buffer for 15 min at 4 °C. Cell debris was removed by centrifugation at 16,000 × g for 15 min at 4 °C, and cleared lysates were subjected to IP. After separation on SDS-PAGE, the radiolabeled bands were visualized and analyzed using the Personal FX phosphor imager and Quantity One software (Bio-Rad).

Ribosome footprint profiling—library preparation

HEK293 cells transiently transfected with CFTR constructs were cultured in DMEM supplemented with 10% fetal calf serum for 24 h without penicillin–streptomycin. Ribosome profiling libraries were prepared as described^58,59 with a few modifications. In brief, cells were rapidly chilled by rinsing with ice-cold PBS, lysed in ice-cold lysis buffer [20 mM Tris–HCl/pH 7.4, 150 mM NaCl, 5 mM MgCl₂, 1 mM dithiothreitol (DTT), 1% (v/v) Triton X-100, and 25 U ml⁻¹ Turbo DNase I] for 10 min on ice, and then triturated ten times through a 26-G needle. After clearance by centrifugation at 16,000 g for 10 min at 4 °C, the lysates were digested with 100 U RNase I (Ambion) per A₂₆₀ lysate for 45 min at room temperature with gentle agitation prior to adding 200 U RiboLock RNase Inhibitor (Thermo Fisher Scientific).

Ribosome protected mRNA fragments were isolated by ultracentrifugation through a 1 M sucrose cushion in polysome buffer [20 mM Tris–HCl/pH 7.4, 150 mM NaCl, 7.5 mM MgCl₂, 0.5 mM DTT, 20 U ml⁻¹ RiboLock RNase Inhibitor] at 70,000 rpm for 2 h at 4 °C using a TLA-110 rotor (Beckman Coulter). Ribosome pellets were extracted using TRIzol (Invitrogen), and RNA fragments were separated by electrophoresis on a 12% (w/v) denaturing polyacrylamide gel with 8 M urea. RNA fragments with sizes ranging from 26 to 34 nt were manually excised from the SYBR Gold (Invitrogen)-stained gels and purified.

3′-Oligonucleotide adaptor ligation, reverse transcription, circularization, and rRNA depletion using biotinylated oligonucleotides were performed as described previously^58,59. Libraries were barcoded using indexing primers for each sample during PCR amplification. Barcoded libraries were then pooled with 3% (v/v) PhiX (Illumina) and sequenced on a NextSeq 500 platform (Illumina) as per the manufacturer’s protocol to yield ~ 30–60 million reads per sample.

Ribosome footprint profiling—data analysis

Sequencing reads were first trimmed to remove 3′ adaptors and mapped to human rRNA sequences (RNA5S1, RefSeq: NR_023363; RNA5-8SN5, RefSeq: NR_003285; RNA18SN5, RefSeq: NR_003286 and RNA28SN5, RefSeq: NR_003287) using HISAT (RRID:SCR_015530)⁶⁰ to eliminate rRNA contaminant reads. The remaining reads were aligned to the CFTR construct sequence and the longest transcript variant of each human gene in GRCh38, excluding CFTR. A maximum of two mismatches at the 5′ end of reads was allowed. The 5′ end nucleotide from each read was annotated and mapped onto each transcript, and all multi-mapped reads were discarded.

Ribosome footprints with lengths between 28 to 30 nt were utilized for triplet periodicity analysis and CFTR ribosome profiling. For CFTR ribosome profiling, only footprints mapped onto the first base of a codon (Frame-0) in the CDS of CFTR constructs were used. Three nucleotide sequences positioned 16th-18th from the 5′ end of each footprint were defined as a codon occupying the ribosomal A-site (A-site codon)^35,61 and the A-site position of each footprint was used to infer the position of the ribosome on the CFTR transcript. Processing of sequencing data was performed on the Galaxy platform (RRID:SCR_006281)⁶².

Electrophysiology

[Blind study 1, Yonsei University] Glass coverslips with transfected HEK293 cells were transferred to a chamber on the stage of an inverted microscope (IX71, Olympus). Patch pipettes were fabricated from borosilicate glass capillaries (TW150F-4 from World Precision Instruments) and fire polished to achieve electrode resistances of 15 – 20 MΩ for single-channel recording. The bath (intracellular) solution contained (in mM): 148 N-methyl-d-glucamine (NMDG)-Cl, 1 MgCl₂, 5 EGTA, and 10 HEPES (pH 7.4). The pipette (extracellular) solution contained (in mM): 146 NMDG-Cl, 1 MgCl₂, 1 CaCl₂, 10 D-(+)-Glucose, and 10 HEPES (pH 7.4). For single-channel studies of CFTR using inside-out membrane patches, membrane potential was clamped at − 60 mV except when acquiring data for current–voltage (i–V) relationships when membrane potential was stepped from − 80 to + 80 mV in 20 mV increments. CFTR channels were activated by adding 3 mM MgATP and 10 U/ml PKA catalytic subunit (Promega) to the bath solution after patch excision. All experiments were performed at room temperature (20–23 °C). Command potential and data acquisition were controlled by pClamp 10.6 software (RRID:SCR_011323) (Molecular Devices). CFTR currents were recorded using an Axopatch 200B patch-clamp amplifier (Molecular Devices), filtered with a lowpass Bessel filter at 1 kHz and digitized by a Digidata 1550B (Molecular Devices). The gap-free data acquisition protocol was conducted with the pClamp software, Axoscope, at a sampling rate of 10 kHz. CFTR recordings were further filtered at a 100 Hz cutoff frequency for data analysis and presentation. For data analysis, single-channel current amplitude (i) was measured either manually using cursors to distinguish the open and closed channel levels or by constructing all-points histograms and fitting these with Gaussian functions. P_o was determined by generating events lists using a half-amplitude crossing criterion and analyzing them using Clampfit 10.6 software (Molecular Devices).

[Blind study 2, University of Bristol] CFTR Cl⁻ channels were recorded in excised inside-out membrane patches using an Axopatch 200B patch-clamp amplifier and pCLAMP software (version 10) (both from Molecular Devices)⁶³. The pipette (extracellular) solution contained (mM): 140 NMDG, 140 aspartic acid, 5 CaCl₂, 2 MgSO₄ and 10 N-tris[hydroxymethyl]methyl-2-aminoethanesulfonic acid (TES), adjusted to pH 7.3 with Tris ([Cl⁻], 10 mM). The bath (intracellular) solution contained (mM): 140 NMDG, 3 MgCl₂, 1 CsEGTA and 10 TES, adjusted to pH 7.3 with HCl ([Cl⁻], 147 mM; free [Ca²⁺], < 10^–8 M) and was maintained at 37 °C. After excision of inside-out membrane patches, the catalytic subunit of protein kinase A (PKA; 75 nM) and ATP (1 mM) were added to the intracellular solution within 2 min of patch excision to activate CFTR Cl⁻ channels. To minimize channel rundown, PKA and ATP were added to all intracellular solutions and voltage was clamped at − 50 mV. To investigate ATP concentration–response relationships, data were acquired from membrane patches where at least 3 ATP concentrations were tested and the reference concentration (1 mM ATP) studied at least twice. In this study, membrane patches contained ≤ 5 active channels. To determine channel number, the maximum number of simultaneous channel openings observed during an experiment was determined⁶⁴. To minimize errors, experimental conditions were used that robustly stimulate channel activity and recordings were verified to be of sufficient length to ascertain the correct number of channels. Single channel records were filtered with an eight-pole Bessel filter (model F-900C/9L8L, Frequency Devices Inc.) at a corner frequency (f_c) of 500 Hz and digitized using a DigiData 1320A interface (Molecular Devices) and pCLAMP software at a sampling rate of 5 kHz. To measure i, Gaussian distributions were fit to current amplitude histograms. To determine P_o, lists of open- and closed-times were created using a half-amplitude crossing criterion for event detection and dwell-time histograms constructed and fitted with one or more component exponential functions using the maximum likelihood method; transitions < 1 ms were excluded from the analysis (eight-pole Bessel filter rise time (T_10-90) ~ 0.73 ms at f_c = 500 Hz). For the purpose of illustration, single-channel records were filtered at 500 Hz and digitized at 5 kHz before file size compression by fivefold data reduction.

PKA purified from bovine heart was purchased from Calbiochem (now Merck Life Science UK Ltd.). All other chemicals were of reagent grade and supplied by Sigma-Aldrich Ltd. (now Merck Life Science UK Ltd.). Stock solutions of ATP, dissolved in intracellular solution, were prepared fresh before each experiment. Immediately before use, stock solutions of ATP were diluted to final concentrations and, where necessary, the pH of the intracellular solution was readjusted to pH 7.3 to avoid pH-dependent changes in CFTR function⁴⁴.

Statistics

Data recording and analyses were randomized with electrophysiological studies additionally blinded. Results are expressed as means ± SEM of n observations, but some group sizes were unequal due to technical difficulties with the acquisition of single-channel data. To test for differences between two groups of data, Student’s unpaired t-test was used. Tests were performed with either Prism 9 (RRID:SCR_002798; GraphPad Software) or SigmaPlot (version 13.0; RRID:SCR_003210; Systat Software Inc.). Differences were considered statistically significant when P < 0.05. To avoid pseudo-replication, all experiments were repeated at different times.

Competing interests

The authors declare no competing interests.