Structural mechanisms for defective CFTR gating caused by the Q1412X mutation, a severe Class VI pathogenic mutation in cystic fibrosis

Abstract

Key points

• Electrophysiological characterization of Q1412X‐CFTR, a C‐terminal truncation mutation of cystic fibrosis transmembrane conductance regulator (CFTR) associated with the severe form of cystic fibrosis (CF), reveals a gating defect that has not been reported previously.

• Mechanistic investigations of the gating deficit in Q1412X‐CFTR suggest that the reduced open probability in Q1412X‐CFTR is the result of a disruption of the function of the second ATP binding site (or site 2) in the nucleotide binding domains (NBDs).

• Detailed comparisons of several mutations with different degrees of truncation in the C‐terminal region of NBD2 reveal the importance of the last two beta‐strands in NBD2 for maintaining proper gating functions.

• The results of the present study also show that the application of clinically‐approved drugs (VX‐770 and VX‐809) can greatly enhance the function of Q1412X, providing in vitro evidence for a therapeutic strategy employing both reagents for patients bearing Q1412X or similar truncation mutations.

Abstract

Cystic fibrosis (CF) is caused by loss‐of‐function mutations of cystic fibrosis transmembrane conductance regulator (CFTR), a phosphorylation‐activated but ATP‐gated chloride channel. Based on the molecular mechanism of CF pathogenesis, disease‐associated mutations are categorized into six classes. Among them, Class VI, whose members include some of the C‐terminal truncation mutations such as Q1412X, is defined as decreased membrane expression because of a faster turnover rate. In the present study, we characterized the functional properties of Q1412X‐CFTR, a severe‐form premature stop codon mutation. We confirmed previous findings of a ∼90% decrease in membrane expression but found a ∼95% reduction in the open probability (P _o). Detailed kinetic studies support the idea that the gating defect is the result of a dysfunctional ATP‐binding site 2 in the nucleotide binding domains (NBDs). Because the Q1412X mutation results in a deletion of the last two beta‐strands in NBD2 and the whole C‐terminal region, we further characterized truncation mutations with different degrees of deletion in this segment. Mutations that completely or partially remove the C‐terminus of CFTR at the same time as keeping an intact NBD2 (i.e. D1425X and S1455X) assume gating function almost identical to that of wild‐type channels. However, the deletion of the last beta‐strand in the NBD2 (i.e. N1419X) causes gating dysfunction that is milder than that of Q1412X. Thus, normal CFTR gating requires structural integrity of NBD2. Moreover, our observation that clinically‐approved VX‐809 (Lumacaftor, Vertex Pharmaceuticals, Boston, MA, USA) and VX‐770 (Ivacaftor, Vertex Pharmaceuticals, Boston, MA, USA) significantly enhance the overall function of Q1412X‐CFTR provides the conceptual basis for the treatment of patients carrying this mutation.

Key points

• Electrophysiological characterization of Q1412X‐CFTR, a C‐terminal truncation mutation of cystic fibrosis transmembrane conductance regulator (CFTR) associated with the severe form of cystic fibrosis (CF), reveals a gating defect that has not been reported previously.

• Mechanistic investigations of the gating deficit in Q1412X‐CFTR suggest that the reduced open probability in Q1412X‐CFTR is the result of a disruption of the function of the second ATP binding site (or site 2) in the nucleotide binding domains (NBDs).

• Detailed comparisons of several mutations with different degrees of truncation in the C‐terminal region of NBD2 reveal the importance of the last two beta‐strands in NBD2 for maintaining proper gating functions.

• The results of the present study also show that the application of clinically‐approved drugs (VX‐770 and VX‐809) can greatly enhance the function of Q1412X, providing in vitro evidence for a therapeutic strategy employing both reagents for patients bearing Q1412X or similar truncation mutations.

Introduction

Cystic fibrosis (CF) is an autosomal recessive hereditary disease afflicting ∼85,000 patients worldwide (De Boeck & Amaral, 2016), although most patients suffering from this life‐shortening genetic disease are found in Caucasian populations (Riordan et al. 1989; O'Sullivan & Freedman, 2009). The root cause of CF is loss of function of a protein kinase A (PKA)‐activated but ATP‐gated anion channel named cystic fibrosis transmembrane conductance regulator (CFTR) encoded by the cftr gene (Riordan et al. 1989). Recent breakthroughs in solving the cryo‐electronmicroscopy structures of human or zebrafish CFTR reveal the detailed architecture of this physiologically important protein (Zhang & Chen, 2016; Liu et al. 2017; Zhang et al. 2017). As a member of the ATP‐binding cassette (ABC) transporter superfamily, the CFTR protein shares the canonical motifs of two transmembrane domains (TMD1 and 2) constructing the permeation pathway (Bai et al. 2010; Gao & Hwang, 2015; Linsdell, 2016), two nucleotide‐binding domains (NBD1 and NBD2) controlling the opening and closing of the gate by ATP binding and hydrolysis (Gadsby et al. 2006; Jih & Hwang, 2012), and a cytosolic C‐terminal region important for protein interaction and stability (Haardt et al. 1999; Wang et al. 2000; Gentzsch et al. 2002; Duan et al. 2012). What is unique for CFTR is the regulatory (R) domain located in between two TMD‐NBD complexes; it contains multiple serine/threonine residues, phosphorylation of which by PKA is required for channel activation (Ostedgaard et al. 2001).

CFTR plays a critical role in modulating transepithelial water and salt transport in almost every exocrine tissue; hence, the malfunction of CFTR causes a wide range of symptoms including thickened mucous, increased in sweat chloride concentration, severe lung infection, pancreatic deficiency, malnutrition, infertility, and so on (Quinton, 1990; Rowe et al. 2005; Gadsby et al. 2006). To date, ∼2000 mutations in the cftr gene have been reported (http://www.genet.sickkids.on.ca/cftr/StatisticsPage) and more than 200 of them have been well characterized phenotypically (http://www.cftr2.org). Based on the molecular mechanisms of CFTR defects, these mutations are categorized into six classes (Wang et al. 2014): impaired protein synthesis because of faulty translation/transcription (Class I), defective trafficking of the CFTR protein (Class II), gating abnormalities (Class III), decreased single channel conductance (Class IV), diminished protein synthesis (Class V) and reduced cell surface stability (Class VI). The proper classification of pathogenic mutations not only helps researchers understand the molecular basis of CF, but also provides good guidelines for clinicians to develop treatment strategies and evaluate prognosis (e.g. Class I–III are more severe and most of the mutations in Class IV–VI are of a mild form). Currently, several drugs targeting mutant CFTR have been approved for clinical use: Lumacaftor (VX‐809, Vertex Pharmaceuticals, Boston, MA, USA) and Tezacaftor (VX‐661, Vertex Pharmaceuticals, Boston, MA, USA), comprising two CFTR correctors that improve CFTR biogenesis for Class II mutations (Van Goor et al. 2011; Taylor‐Cousar et al. 2017), as well as Ivacaftor (VX‐770, Vertex Pharmaceuticals, Boston, MA, USA), a CFTR potentiator that increases channel function for Class III mutations (Van Goor et al. 2009; Eckford et al. 2012; Jih & Hwang, 2013). However, multiple biochemical defects are usually found in disease‐associated mutations. For example, the most common mutation F508del is considered a prototypic Class II mutation, although the few proteins that successfully reach the cell membrane exhibit gating defects (Class III) and decreased plasma membrane stability (Class VI) (Dalemans et al. 1991; Lukacs et al. 1993; Jih et al. 2011). Hence, the combination of Lumacaftor and Ivacaftor (i.e. Orkambi, Vertex Pharmaceuticals, Boston, MA, USA) and the combination of Tezacaftor and Ivacaftor (i.e. Symdeko, Vertex Pharmaceuticals, Boston, MA, USA) are now approved for the treatment of homozygous F508del patients, aiming to partially rectify these multifaceted deficits (Wainwright et al. 2015; De Boeck & Amaral, 2016; Taylor‐Cousar et al. 2017). Considering the effectiveness of both CFTR potentiators and correctors on a range of mutations (Char et al. 2017; Xue et al. 2017), we suspect that this combination approach could be an effective treatment for patients carrying mutations other than Class II and III.

Class VI mutations were first defined in 1999 (Haardt et al. 1999). In the original study, these mutations were found in several heterozygous patients with F508del in one allele and a C‐terminal mutation caused by either frameshift mutation (e.g. 4326delTC, 4279insA and 4271delC) or premature stop codons (e.g. S1455X, Q1412X and L1399X) in the second allele. The number of amino acids deleted ranges from 26 (S1455X) to 98 (4279insA), although only patients with more than 70 residues deleted (i.e. more deletion than Q1412X) exhibit severe phenotypes (Haardt et al. 1999). The same study also reported that, for those mutations with more than 70 residues deleted, the biosynthesis, processing and channel function are essentially normal; the major deficit is the 90–95% decrease in amounts of surface proteins because of a faster degradation rate, resulting in a much‐reduced chloride flux to ∼10% of wild‐type. However, this level of functional loss is probably insufficient to account for the severe phenotype associated with mutations such as Q1412X. For example, the missense R117H mutation reduces the overall chloride flux to only ∼5% of wild‐type, yet it causes a mild‐form CF (Sheppard et al. 1993; Yu et al. 2016). We therefore hypothesize that severe‐form Class VI C‐terminal truncation mutations may acquire additional functional deficits, which is an idea more in line with several studies showing that C‐terminal truncations could influence gating (Gentzsch et al. 2002; Ostedgaard et al. 2003).

In the present study, our main focus is the Q1412X mutation, which leads to the deletion of the entire C‐terminus of CFTR and two beta‐strands at the C‐terminal end of NBD2 (Fig. 1) (Gentzsch et al. 2002; Liu et al. 2017). Moreover, among the severe‐form C‐terminal truncation mutations, Q1412X has the shortest deletion and thus may be more amenable to pharmacological modulation of its overall function. We confirmed a previous study (Haardt et al. 1999) reporting that the steady‐state surface expression of Q1412X is dramatically reduced. However, the CFTR corrector VX‐809 improves the expression level by ∼2‐fold, suggesting that Lumacaftor may be beneficial for patients carrying this mutation. Patch clamp studies of Q1412X reveal a ∼20‐fold decrease of the open probability (P _o). Experiments probing the NBD function of CFTR demonstrate that the underlying mechanism for the gating defect in Q1412X is a dysfunction of ATP‐binding site 2. To further explore the structural mechanism for this gating defect, we constructed N1419X, D1425X and S1455X mutations to sequentially recover the truncated portion of Q1412X. Although the truncation mutants D1425X and S1455X, which retain the structural integrity of NBD2, assume almost identical gating behaviours to that for wild‐type (WT)‐CFTR, N1419X channels show mild gating abnormality. Collectively, our results provide the structure/function basis for the gating defects associated with the Q1412X mutation and support the notion that a combination of CFTR correctors and potentiators could be beneficial for patients harbouring severe phenotype Class VI mutations.

Figure 1. Atomic structure of human CFTR highlighting the C‐terminal region of NBD2.

A, high‐resolution cryo‐electronmicroscopy structure of human CFTR (Protein Data Bank: 5UAK). The dashed box encloses NBD2 and the C‐terminal region. Thin black lines mark individual hydrogen bonds between beta‐strands. B, magnified view of the dashed box in (A). The side chains of the residues relevant to the present study are shown as sticks and coloured blue for D1425, green for N1419 and red for Q1412. S1455 is not resolved in this structure. Of note, E1371 (orange) and K1250 (magenta), two residues essential for ATP binding and/or hydrolysis in ATP‐binding site 2 (Hwang & Sheppard, 2009), are at the junction between a beta‐strand and an alpha‐helix (side chains shown in sticks). Also note that K1250, a residue important for both ATP binding and hydrolysis, is located at the N‐terminus of an alpha‐helix that is surrounded on three sides by beta‐sheets, including the one investigated in the present study. This intimate structural relationship between functionally important residues and the whole beta‐sheet network suggests that deletion of the beta‐strands in N1419X and Q1412X may influence the structural and functional integrity of NBD2.

Methods

Cell culture and transfection

Chinese hamster ovary (CHO) cells were grown at 37°C in Dulbecco's modified Eagle's medium (Gibco, Waltham, MA, USA) supplemented with 10% (v/v) fetal bovine serum (Sigma‐Aldrich, St Louis, MO, USA) in a humidified atmosphere of 5% CO₂. Cells were trypsinized and transferred to 35 mm culture dishes 1 day before the transfection. For electrophysiological experiments, CHO cells were co‐transfected with CFTR‐cDNA [pcDNA 3.1 Zeo (+) vector; Invitrogen, Carlsbad, CA, USA] and green fluorescent protein encoding pEGFP‐C3 (Takara Bio Inc., Shiga, Japan) using PolyFect transfection reagent (Qiagen, Valencia, CA, USA). Cells were incubated at 27°C for 2–3 days before microscopic current recording experiments or 3–6 days for macroscopic current recording experiments after transfection. For Western blot studies, CHO cells were transfected with CFTR‐DNA [pcDNA 3.1 Zeo (+) vector; Invitrogen, Carlsbad, CA, USA] using X‐tremeGENE transfection reagent (Roche, Basel, Switzerland). Six hours after transfection, drugs were added to the medium to desired concentrations.

Western blotting

Cells were lysed 18 h post drug treatment using 1 × SDS loading buffer. Cell lysates were sheared by pushing through 18 G needles. Whole cell lysates were separated in 4–20% gradient gels (Bio‐Rad Laboratories, Hercules, CA, USA) and transferred onto nitrocellulose membranes. Non‐specific bindings were minimized with 5% milk in Tris‐buffered saline with Tween‐20 (TBST) buffer (20 mm Tris, 137 mm NaCl, 0.1% Tween 20) at 4˚C overnight. The membranes were then probed with anti‐CFTR monoclonal antibody Ab 596 (dilution 1:3000; CF foundation, Bethesda, MD, USA) and anti‐Vimentin antibody (dilution 1:3000; Santa Cruz Biotechnology, Dallas, TX, USA) at room temperature for 2 h. The membranes were washed with TBST five times and then incubated with anti‐mouse IgG, HRP linked antibody (dilution 1:3000; Cell Signaling Technology, Danvers, MA, USA) at room temperature for 1 h. The membranes were washed three times with TBST and developed with chemiluminescence reagent (Thermo Fisher Scientific, Waltham, MA, USA). The luminescence was detected by Molecular Imager Chemidoc (Bio‐Rad Laboratories). CFTR and vimentin bands were quantitated using Image Lab (Bio‐Rad Laboratories). The intensity of each CFTR band was normalized against that of the vimentin within the same lane. After quantification, the contrast and the gamma value of the images of the gels were adjusted for clearer presentations of the bands. As a result of a large difference in signal intensity between CFTR and vimentin, the CFTR and vimentin bands in the western blots were acquired by applying two different exposure times to the same gel to avoid saturation.

Mutagenesis

CFTR mutants were constructed using QuikChange XL kit (Agilent Technologies, Santa Clara, CA, USA) in accordance with the manufacturer's instructions. The entire sequences of CFTR mutants were verified by DNA sequencing (DNA Core Facility, University of Missouri, MO, USA).

Electrophysiological recordings

For the patch clamp experiment, glass chips carrying the transfected cells were transferred to a chamber on the stage of an inverted microscopic (IX‐51; Olympus, Tokyo, Japan). The pipettes used for patch clamp experiment were made from borosilicate glass capillaries (Kimble & Chase, Vineland, NJ, USA) using a two‐step vertical micropipette puller (PP‐81; Narishige, Tokyo, Japan) and polished with a home‐made microforge to a resistance of 2–4 MΩ in the bath solution. Membrane patches were excised into an inside–out configuration after the seal resistance reached >40 GΩ. After the excision, the pipette was placed to the outlet of a three‐barrel perfusion system and perfused with 25 IU of PKA and 2 mm ATP until the current reached a steady‐state. To maintain the phosphorylation level, 25 IU of PKA was added to all other ATP‐containing solutions (except for as shown in Fig. 4 A where PKA was not added in solutions with ATP). All the solution changes were performed using a fast solution change system (SF‐77B; Warner Instruments, Hamden, CT, USA) with a dead‐time of ∼30 ms (Tsai et al. 2009).

Figure 4. ATP responsiveness of Q1412X‐CFTR.

A, real‐time recording of macroscopic Q1412X current showing an initial biphasic current response to PKA, ATP and VX‐770, and a slow decrease of the current upon ATP removal (highlighted in the dashed box). B, normalized current relaxation enlarged from the dashed box in (A). The current decay can be fitted with a single exponential function (white line) with a relaxation time constant of 30.8 s in this recording and 34.5 ± 2.8 s for five membrane patches averaged. Note a residual current of 33% of the original current is seen upon the completion of current decay for this specific patch. Overall, the proportion of ATP‐sensitive current is 62 ± 5% (n = 5). C, representative real‐time recording of Q1412X‐CFTR at three different ATP concentrations. D, compared to the current in 2 mm ATP, there is no significant difference in current magnitude in 5 mm or 50 μm ATP (n = 4). n.s., not significant.

All of the electrophysiological data were recorded at room temperature with a patch clamp amplifier (EPC10; HEKA Elektronik, Lambrecht/Pfalz, Germany). Data were filtered online at 100 Hz with an eight‐pole Bessel filter (LPF‐9; Warner Instruments) and digitized to a computer at a sampling rate of 500 Hz. The membrane potential (V _m) was held at –50 mV during the entire experiment, unless noted otherwise.

Chemicals and solutions

For experiments using excised inside–out membrane patches, the pipette solution contained (in mm): 140 NMDG‐Cl, 5 CaCl₂, 2 MgCl₂ and 10 Hepes (pH was adjusted to 7.4 with NMDG). Before excision, cells were perfused with a bath solution containing (in mm): 145 NaCl, 5 KCl, 1 CaCl₂, 2 MgCl₂, 5 glucose, 5 Hepes and 20 sucrose (pH was adjusted to 7.4 with NaOH). After the excision, cells were bathed in a perfusion solution containing (in mm): 150 NMDG‐Cl, 1 CaCl₂, 2 MgCl₂, 10 EGTA and 8 Tris (pH was adjusted to 7.4 with NMDG). For the experiment using nitrate as charge carrier, NMDG‐Cl was replaced with NMDG‐NO₃ for the perfusion solution.

MgATP and PKA were purchased from Sigma‐Aldrich (St Louis, MO, USA). MgATP was dissolved into a 500 mm stock solution and stored at –20˚C, the [MgATP] was 2 mm in all experiments, unless indicated otherwise. CFTR_inh‐172, as provided by Dr Robert Bridges (Rosalind Franklin University, Chicago, IL, USA) who directs the Chemical Compound Distribution Program sponsored by Cystic Fibrosis Foundation Therapeutics (Bethesda, MD, USA), was stored at –70˚C as a 5 mM stock solution in DMSO. VX‐770, purchased from Selleckchem (Houston, TX, USA), was made into a 100 μM stock solution in DMSO and stored at –70˚C. VX‐809 was purchased from Aobious Inc. (Gloucester, MA, USA) and stored at –70˚C as a 10 mm stock solution in DMSO. Reagents used in electrophysiological experiments were freshly thawed and diluted to the desired concentration with the perfusion solution for excised inside–out membrane patches and the pH was adjusted to 7.4 with NMDG. After each experiment with VX‐770, all of the devices in contact with the solution containing VX‐770 were washed repeatedly using 50% DMSO to prevent contamination of the residue as described previously (Jih & Hwang, 2013).

Statistical analysis

For the patch clamp experiments, the steady‐state mean current was measured with Igor Pro (Wavemetrics, Lake Oswego, OR, USA) and the relaxation time constant was estimated by fitting the current decay upon ATP washout with built‐in exponential functions. Single‐channel kinetic analysis was performed using software developed by Csanady (2000). Although the software is capable of analysing recordings containing up to eight openings, we only analysed the kinetics for membrane patches containing fewer than four opening steps. Furthermore, although the software has been successfully utilized for WT‐CFTR channels (Yeh et al. 2015) and some mutants (Yu et al. 2016), for channels with low open probability (P _o) such as Q1412X, it is difficult to estimate the exact number of channels in the membrane patch. By boosting the P _o with CFTR potentiators such as VX‐770 and NO₃ ⁻, we mitigated this technical difficulty to some extent, although the measured P _o should be considered as an apparent open probability (or the maximal P _o).

For the N306D/Q1412X experiments, the single‐channel amplitudes of the O₁ and O₂ states were measured by fitting the all‐points histograms with the built‐in multi‐Gaussian function (Igor Pro). The categorization of the gating pattern was conducted using the same method as that employed in our previous study (Jih et al. 2012; Zhang & Hwang, 2017). The cut‐off duration for identifying a state was set at 8 ms, which represents five data points at the 500 Hz sampling rate to exclude non‐CFTR current spikes or noises that are ∼4 ms or three data points.

Student's t tests assuming equal variance were conducted for comparisons between two mean values. To analyse differences between more than two groups, one‐way ANOVA was we used, followed by Tukey's range test. P < 0.05 was considered statistically significant. All error bars represent the SEM and the data are shown as the mean ± SEM, with n indicating the number of experiments.

Protein structure visualization

The structure of human CFTR (Protein Data Bank code: 5UAK) was visualized using UCSF Chimera (Pettersen et al. 2004). The hydrogen bond pairs were identified using the software's built‐in find hydrogen‐bond function.

Results

The fundamental mechanism underlying the Class VI mutation is the decrease in surface expression as a result of an accelerated protein turnover rate (Haardt et al. 1999). Hence, we first examined the protein expression of several constructs with different degrees of C‐terminal truncation using western blotting and investigated whether the expression can be enhanced by CFTR corrector VX‐809. Both core‐glycosylated (Band B, MW: ∼ 150 kDa) and complex‐glycosylated (Band C, MW: ∼ 170 kDa) forms of WT and all the mutations can be detected by anti‐CFTR antibody Ab 596, which recognizes residues 1204–1211 of CFTR (Tosoni et al. 2013), indicating that the CFTR proteins under study are synthesized, folded and trafficked to the cell membrane (Fig. 2 A). Densitometric analysis of the western blot data shows that, compared to WT, only the Q1412X mutation significantly decreases the amount of Band C proteins (7.6 ± 2.3% of WT, n = 4, P < 0.01) (Fig. 2 A). Although the expression level of S1455X, D1425X and N1419X appears to be lower than that of WT control, the differences do not reach a statistically significant level perhaps because of large variation in our assay. This large variation may be a result of variations in transfection efficiency. Nonetheless, these results are in line with previous studies suggesting that only truncations equal to or more severe than Q1412X influence protein expression (Mickle et al. 1998; Haardt et al. 1999). Although CFTR correctors such as VX‐809 have been utilized mainly to partially rectify the trafficking defects associated with Class II mutations (Van Goor et al. 2011), Fig. 2 B shows that incubation of cells with 3 μm VX‐809 enhanced the expression of all the constructs tested. For Q1412X, VX‐809 increased the intensity of band C by 224 ± 27% (n = 4) compared to the one treated with DMSO, suggesting that VX‐809 could be beneficial for patients carrying Class VI mutations.

Figure 2. Surface expression of C‐terminal truncation mutations can be enhanced by VX‐809.

A, western blot of WT and C‐terminal mutations investigated in the present study. Core‐glycosylated Band B (MW: ∼150 kDa) and complex‐glycosylated Band C (MW: ∼170 kDa) were probed with anti‐CFTR antibody Ab 596. The same blot was probed with an antibody against the cytoskeletal protein vimentin as an internal control for protein loading. The intensity of Band C in each construct is first normalized against that of vimentin, then the normalized intensity of each truncation mutation was compared with that of WT to obtain the relative percentage of band intensity shown in the lower panel. The expression of Band C of Q1412X (n = 4) is significantly less than WT. N1419X (n = 4), D1425X (n = 4) and S1455X (n = 4) show no significant difference from WT. Asterisks indicate statistical significance compared to WT. B, upper: effects of VX‐809 on the protein expression of all constructs investigated. The intensity of Band C in each construct is first normalized against that of vimentin, then the normalized intensity of the one treated with VX‐809 was compared with those treated with DMSO to obtain the relative percentage of band intensity shown below. The numbers of experiments for individual construct are four for WT, four for Q1412X, five for N1419X, five for D1425X and three for S1455X. Asterisks indicate statistical significance compared to DMSO. ^* P < 0.05; ^** P < 0.01; ^*** P < 0.001.

Our results confirm the previous observation of Haardt et al. (1999) indicating that the surface expression level of Q1412X is ∼10% of WT. However, as described in the Introduction, this level of reduction of CFTR proteins in the cell membrane may be insufficient to cause a severe form of CF. Because the Q1412X mutation eliminates the last two beta‐strands in NBD2, the crucial domain for normal ATP‐dependent CFTR gating (Hwang & Kirk, 2013), we suspected that this mutation might affect CFTR gating. To investigate the gating characteristics of Q1412X, we tested the effects of two CFTR potentiators, VX‐770 (Van Goor et al. 2009; Jih & Hwang, 2013) and nitrate (Yeh et al. 2015), on macroscopic Q1412X‐CFTR currents in excised inside–out membrane patches. These two potentiators facilitate gate opening via two independent binding sites, such that the P _o can be greatly increased by their synergistic effect when applied together (Yeh et al. 2015). Figure 3 A shows a representative real‐time recording of Q1412X macroscopic currents. The Q1412X channels were first phosphorylated with ATP and PKA in a chloride‐based bath solution. After almost 30 min of exposure to PKA and ATP, the overall current was very small, although a ∼3‐fold (3.6 ± 0.5, n = 6) increase of the macroscopic current was observed when the perfusion solution was switched to a nitrate‐based one. This effect can be reversed when the cytoplasmic anion is switched back to chloride. As shown in Fig. 3 B, the application of 200 nm VX‐770 in the chloride bath resulted in a 9.6 ± 1.1‐fold (n = 6) increase, whereas the combined effect of VX‐770 and nitrate was 28.0 ± 4.9‐fold (n = 6). By definition, the magnitude of the macroscopic current is the product of the single‐channel current, the number of channels in the membrane patch and P _o. In the same patch, the number of channels activated is presumably constant, such that the fold increase in macroscopic current can be considered as the increase in P _o after correcting for the differences in the single‐channel conductance of CFTR between chloride and nitrate (Yeh et al. 2015). Thus, a potentiation of P _o by 28‐fold by VX‐770 and nitrate indicates that the maximally possible P _o in the absence of potentiators must be less than 1/28 or 0.036 (assuming the P _o in VX‐770 and nitrate is unity), which is more than 10‐fold lower than that of WT‐CFTR (Yeh et al. 2015).

Figure 3. Effect of gating modulators on Q1412X.

A, representative macroscopic recording of Q1412X current under different potentiators. The majority of the current can be inhibited by 10 μm CFTR_inh‐172 (Inh‐172). Dashed line indicates the baseline (same to all figures). B, summary of fold‐increases of macroscopic Q1412X‐CFTR currents by nitrate, VX‐770 and a combination of nitrate and VX‐770 (n = 6). C, single channel recording of Q1412X with ATP in chloride bath (upper) or ATP + VX‐770 in nitrate bath (lower). D, summary of single‐channel kinetic parameters of Q1412X‐CFTR (n = 4). The membrane potential was held at –50 mV. Downward deflections represent channel opening. ^* P < 0.05; ^** P < 0.01; ^*** P < 0.001.

To more accurately estimate the P _o, we resorted to membrane patches that yield microscopic currents even in the presence of VX‐770 and nitrate. Figure 3 C shows continuous single‐channel current traces of Q1412X in ATP or ATP plus VX‐770 and nitrate. In the presence of ATP alone, only several openings interrupted by very long closed events were seen in more than 10 min of recording, making it almost impossible to accurately estimate the number of channels in the patch. However, in the same patch, a robust activity can be observed in the presence of VX‐770 and nitrate, allowing detailed microscopic kinetic analysis. As summarized in Fig. 3 D, in four separate patches, the P _o in ATP alone is 0.08 ± 0.03 (n = 4) with a mean open time (τ_o) of 3.04 ± 1.15 s (n = 4) and a mean closed time (τ_c) of 26.0 ± 6.8 s (n = 4). However, because of the low P _o, which renders the analysis more vulnerable to errors, these kinetic parameters can only be considered as rough estimations. By contrast, after potentiation by VX‐770 and nitrate, the channel activity was dramatically increased so that we can be more certain about the number of channels in the membrane patch. In four different patches with only one simultaneous opening step in the presence of VX‐770 and nitrate, we obtained a P_o of 0.74 ± 0.03 with a τ_o of 12.5 ± 3.0 s and a τ_c of 4.0 ± 0.9 s. Once this more accurate apparent P _o is attained, we can then back‐calculate the P _o of Q1412X in the presence of ATP alone by dividing the P _o in the presence of potentiators by the fold‐increase of macroscopic currents with potentiators: 0.74/28 or 0.026, which is ∼6% of WT (P _o ∼0.45). The discrepancy in P _o between direct measurements based on microscopic kinetic analysis and back‐calculation emphasizes the importance of increasing the channel activity and applying macroscopic measurements for P _o estimation for low‐P _o channels. Because the P _o in the absence of VX‐770 is extremely low, the number of observable opening and closing events is low. Thus, microscopic analysis, which relies on the abundance of gating events, becomes problematic.

Regardless of the difference in P _o between macroscopic and microscopic analyses, the Q1412X mutation does exhibit a severe gating defect featuring a much‐prolonged τ_c. Interestingly, however, the τ_o of ∼3 s is ∼10‐fold larger than that of WT‐CFTR (Yeh et al. 2015). Because the opening and closing of WT‐CFTR are controlled, respectively, by NBD dimerization and separation of the dimer triggered by ATP hydrolysis (Gadsby et al. 2006; Jih & Hwang, 2012), we hypothesized that the low P _o of Q1412X is the result of a malfunction of the gating machinery NBDs. To test the functional integrity of NBDs, we first examined whether Q1412X responds to ATP at all. Because of a much‐reduced number of channels N, and an extremely low P _o of Q1412X‐CFTR, almost all membrane patches yielded microscopic currents in the absence of potentiators. Because the current fluctuations could easily obliterate changes of CFTR activity upon ATP removal under this condition, to quantitatively assess current changes upon ATP removal, we carried out the experiments in the presence of 200 nm VX‐770. Figure 4 A shows a real‐time recording from a patch where Q1412X‐CFTR currents are activated by ATP, PKA and VX‐770. Interestingly, the observed currents follow a biphasic time course, which is consistently seen when the experiments were carried out in the presence but not in the absence of VX‐770 (Fig. 3 A).

We suspected the Q1412X channels might undergo some kind of rundown process after patch excision, although this occurred happened relatively slowly. Hence, the current decay only became discernable when the phosphorylation‐dependent activation was accelerated by VX‐770 (Fig. 4 A). Because of this time‐dependent current decay, we chose to start the experimental protocol after the current decrease reached a steady‐state. Before removing the ATP, we first switched to a solution with only ATP to ensure that the phosphorylation process was completed and there was minimal dephosphorylation of CFTR by membrane‐associated phosphatases. The removal of ATP then induced a slow decay of the steady‐state current by 62 ± 5% (n = 5) with a time constant of 34.5 ± 2.8 s (n = 5). The currents could be mostly recovered by adding back ATP and the addition of PKA did not increase the current (Fig. 4 A). This slow, single exponential decay of macroscopic Q1412X‐CFTR currents upon ATP removal is very different from the behaviour of WT reported by our laboratory previously (Lin et al. 2014). The current relaxation of WT upon the removal of ATP in the presence of VX‐770 can be fitted with a double exponential function composed of a major fast phase with a time constant <1 s and a minor slow phase with a time constant of ∼30 s (Lin et al. 2014). This biphasic relaxation was proposed to reflect the dissociation of ATP or its hydrolytic products from two distinct binding sites with different turn‐over rates (Lin et al. 2014). The slow monophasic decay of macroscopic Q1412X‐CFTR currents in the presence of VX‐770 suggests the existence of at least one high‐affinity ATP binding site in this mutant. Of note, the relaxation time constant of ∼35 s is >2‐fold higher than the mean open time, suggesting that this current decay may not entirely reflect closing of the channel. Part of the current decay should also represent a slow decrease of the activity in the absence of ATP (also known as ATP independent gating).

We next examined the response of Q1412X to two more ATP concentrations (5 and 50 μm) in addition to the conventional 2 mm ATP. For WT‐CFTR with a K _1/2 of ∼90 μm (Zhou et al. 2006), 2 and 5 mm ATP are both saturating concentrations for gating, whereas 50 μm ATP would generate ∼40% of the maximal current amplitude. As shown in Fig. 4 C and D, the current amplitude of Q1412X remained static in all three concentrations (Fig. 4 D). However, the current was reduced slowly upon washout of ATP in the same membrane patch (Fig. 4 C). These results thus support the conclusion that the Q1412X mutation does not completely abolish ATP‐dependent gating. However, the much‐prolonged open time in the presence or absence of ATP suggests that ATP hydrolysis by NBD2 may be affected (see below).

Several observations described above suggest defective function of CFTR's catalysis‐competent site 2, the ATP binding site constructed by the head subdomain of NBD2 and the tail subdomain of NBD1. First, the C‐terminal end of NBD2 is deleted in Q1412X. Second, the opening rate (the reciprocal of the mean closed time), which in WT‐CFTR is mainly controlled by NBD dimerization upon ATP binding to site 2, is extremely low even in the presence of VX‐770. Third, the closing rate (the reciprocal of the mean open time), which is normally controlled by ATP hydrolysis at site 2 in WT channels, is also slow. To further assess the role of ATP hydrolysis in Q1412X channel gating, we introduced the E1371S mutation, which abolishes ATPase activity by replacing the catalytic glutamate with a serine (Vergani et al. 2003), into the Q1412X background and examined whether the gating behaviour is altered. For WT‐CFTR, the E1371S mutation significantly increases the P _o to almost unity by prolonging the open time to tens of seconds (Bompadre et al. 2005). We reasoned that, if the gate closure of Q1412X is also mainly triggered by ATP hydrolysis, a similar ‘locked‐open’ phenomenon may be observed in E1371S/Q1412X. However, microscopic recording of Q1412X and E1371S/Q1412X mutation both showed a low P _o (Fig. 5). More importantly, the already prolonged τ_o seen with the Q1412X mutant (3.2 ± 1.1 s, n = 4) is not significantly altered by the additional mutation of E1371S (3.9 ± 1.1 s, n = 3).

Figure 5. Lack of effects on the mean open time of Q1412X by the E1371S mutation.

A, representative microscopic recordings of Q1412X (upper) and E1371S/Q1412X (lower) at 2 mm ATP. Note the exceedingly long closed events before each opening. B, mean open times for Q1412X (n = 4) and E1371S/Q1412X (n = 3) channels. n.s., not significant.

The results shown in Fig. 5 suggest that the hydrolysis process is significantly slowed, if not abolished completely, by the Q1412X mutation. To further probe the role of ATP hydrolysis with respect to modulating the gating cycle Q1412X‐CFTR, we resorted to another approach established previously. As first demonstrated by Gunderson and Kopito (1995), WT‐CFTR exhibits two conductance levels in a burst of channel opening with a smaller conductance O₁ and a larger conductance level O₂ distinguished after heavy filtering of single‐channel records. A predominant and preferred gating pattern of C → O₁ → O₂ → C transition over C → O₂ → O₁ → C in the presence of ATP infers a violation of equilibrium, supporting the idea of input of free energy from ATP hydrolysis to drive the O₁ → O₂ transition. Later, Ishihara & Welsh (1997) found that this observed ‘O₁/O₂ phenotype’ is caused by a different sensitivity of the two states to the pore blocker MOPS⁻ used as a buffer by Gunderson and Kopito (1995). Subsequently, we observed similar gating behaviours by altering the distribution of charges in the CFTR pore using the mutations R352C/Q or N306D/E (Jih et al. 2012; Zhang & Hwang, 2017).

We introduced the N306D mutation into the Q1412X background and studied the conductance phenotypes in the presence or absence of ATP. The O₁/O₂ conductance phenotypes observed in the N306D mutation are voltage independent (Zhang & Hwang, 2017); therefore, to obtain a clearer opening burst without the interference of voltage‐dependent flickery block (Zhou et al. 2001), the membrane potential was held at +50 mV. Two conductance levels can be clearly differentiated in all‐points histograms: a single‐channel amplitude of 0.30 ± 0.01 pA (n = 4) for the smaller O₁ state and 0.41 ± 0.02 pA (n = 3) for the larger O₂ state (Fig. 6), with values almost identical to those measured in the N306D single mutant (Zhang and Hwang, 2017). The pattern of the opening bursts for N306D/Q1412X double mutant (summarized in Table 1), determined as described in the Methods, is dominated by the C → O₁ → C transition in the presence of ATP (93% of all gating events), in contrast to the predominance of the C → O₁ → O₂ → C transition seen with the N306D single mutant (Zhang & Hwang, 2017). As the Q1412X channel opens mainly to the ‘pre‐hydrolytic’ open state O₁ and closes thereafter, we proposed that Q1412X‐CFTR closes mainly through a non‐hydrolytic pathway, or its gate closes faster than the ATP hydrolysis rate. Interestingly, when comparing the gating patterns in the presence or absence of ATP, we found that the opening bursts in the presence of ATP are visibly longer than those in the absence of ATP (Fig. 6). Quantitative analysis revealed that the averaged τ_o of the O₁ state with ATP is 1.3 ± 0.4 s (n = 4), which is significantly higher than 0.30 ± 0.05 s (n = 4) without ATP. Hence, ATP stabilizes the O₁ open state of Q1412X‐CFTR probably by forming an NBD dimer with two ATP bound at the dimer interface.

Figure 6. The gating patterns of N306D/Q1412X.

A, representative real‐time microscopic recording of N306D/Q1412X in the presence of 2 mm ATP (left) and the corresponding all‐points histogram (grey). Black line marks a multi‐Gaussian fitting of the histogram to allow measurements of the single‐channel amplitude (right): 0.30 ± 0.01 pA (n = 4). Although the O₂ state could be observed in the presence of ATP, the events were too rare to be seen in the all‐points histogram (for sample events, see Fig. 10). B, recording of N306D/Q1412X in the absence of ATP (left) and the corresponding all‐points histogram (right). Two distinct opening levels could be readily noted both in the raw trace and in the histogram. The smaller O₁ is of a similar amplitude to the one measured in (A). The larger O₂ state assumes an amplitude of 0.41 ± 0.02 pA (n = 3). C, enlarged opening bursts for those indicated in (A) and (B). Event a shows a long O₁ state in the presence of ATP. Events b to d were gathered from openings without ATP; b, C → O₁ → C; c, C → O₂ → C; and d, C → O₁ → O₂ → C (Table 1). The membrane potential was held at +50 mV. Upward deflection represents channel opening.

Table 1.

Summary of the gating patterns of N306D/Q1412X

	O1–O2	O1	O2	O2–O1	(O1–O2)n
N306D/Q1412X						Total
2 mm ATP	2	235	4	2	2	253
	(1%)	(93%)	(2%)	(1%)	(1%)	(100%)
0 ATP	39	163	29	9	9	271
	(14%)	(60%)	(11%)	(3%)	(3%)	(100%)

Five categories of gating topologies are illustrated. The number and percentage of each gating topology is shown. The gating patterns in 2 mm ATP were gathered from four different membrane patches with a total length of 1490 s; the gating patterns without ATP were collected from three membrane patches with a total duration of 2435 s.

There are two noteworthy issues regarding the averaged τ_o of the O₁ state with ATP. First, the decreased τ_o compared to that of Q1412X is probably a result of the introduction of the N306D mutation, which has been shown to exhibit a shortened τ_o (∼100 ms vs. ∼300 ms in WT) (Yeh et al. 2015; Zhang & Hwang, 2017). Second, the averaged τ_o of ∼3 s for Q1412X in the presence of ATP is still shorter than other hydrolysis‐deficient mutations such as E1371S (∼100 s), indicating that the truncation still destabilizes the open state. As shown in Fig. 1, the Q1412 residue is located close to the end of NBD2 and the Q1412X mutation results in a deletion of two terminal beta‐strands of NBD2 and the entire C‐terminus. Thus, it is reasonable to speculate that removing the last two beta‐strands in NBD2 destabilizes the NBD dimer and hence shortens the open time. The significance of these observations on the gating mechanism is elaborated further in the Discussion.

It has been shown that complete removal of the C‐terminus up to residue 1425 does not influence CFTR gating (Rich et al. 1993; Gentzsch et al. 2002), such that we hypothesized that the detrimental effect of Q1412X on gating is mainly a result of the deletion of the two beta‐strands. To further dissect the functional roles of these segments, we made two constructs that sequentially restore the truncated portion; these include N1419X (one beta‐strand added) and D1425X (two beta‐strands added). We also investigated the behaviour of S1455X, which is a disease‐associated mutation with extremely mild clinical manifestation (Mickle et al. 1998), aiming to determine whether partially truncation of the C‐terminal region influence channel function.

We reasoned that, if the observed gating anomalies in Q1412X are indeed the result of a structural perturbation in NBD2, these three truncation constructs should exhibit different gating behaviours, with S1455X and D1425X being relatively intact, whereas the deletion at N1419 would display a reduced function. We first examined their responsiveness to ATP (Fig. 7 A). Probably because of an abundance of cell surface proteins, once phosphorylated, these three CFTR variants yield macroscopic currents with ATP even in the absence of VX‐770. By contrast to Q1412X, the currents from all three constructs decay rapidly upon removal of ATP, suggesting that the closing of these channels is mediated by ATP hydrolysis similar to WT‐CFTR. In addition, although Q1412X‐CFTR currents remain unchanged upon switching from 2 mm to 50 μm ATP, using the same protocol, the currents of all three mutants decrease (Fig. 7 A). The fraction of currents in 50 μm ATP relative to those in 2 mm ATP, 0.50 ± 0.03 (n = 3) for S1455X or 0.40 ± 0.04 (n = 4) for D1425X, is similar to WT (Zhou et al. 2006), although it is significantly larger than that for N1419X (0.10 ± 0.05, n = 4) (Fig. 7 B). Thus, deleting the last 26 amino acids (S1455X) or the entire C‐terminal region excluding NBD2 (D1425X) does not dramatically influence channel gating, whereas truncating one (N1419X) or two (Q1412X) beta‐strands at the C‐terminal end of NBD2 incrementally impairs CFTR gating in terms of measured open probability.

Figure 7. ATP sensitivity of S1455X, D1425X and N1419X.

A, representative macroscopic recordings of S1455X, D1425X and N1419X in response to 2 mm, 50 μm or 0 ATP. B, summary of the fraction of currents in 50 μm ATP relative to those in 2 mm ATP for S1455X (n = 3), D1425X (n = 4) and N1419X (n = 4). Of note, for the N1419X trace shown in (A), the current in 2 mm ATP after 50 μm ATP did not reach the original level. This phenomenon, although not always seen, is probably the result of some rundown processes. To correct any errors introduced by this time‐dependent decrease of macroscopic currents, we bracketed all the experiments and used the average current in 2 mm before and after 50 μm ATP as the control. ^*** P < 0.001.

The distinct single‐channel gating behaviours of these truncation mutants further support the conclusion that the last two beta‐strands in the NBD2 of CFTR are essential structural components for sustaining normal channel gating. Figure 8 shows single‐channel traces of phosphorylated S1455X, D1425X and N1419X in the presence or absence of VX‐770. Compared to the open probability of WT‐CFTR that we reported previously (Yeh et al. 2015), the steady‐state P _o for S1455X (0.50 ± 0.01, n = 4) or D1425X (0.45 ± 0.03, n = 4) is very similar to that of WT‐CFTR, whereas the P _o of N1419X (0.30 ± 0.02, n = 3) is significantly smaller (cf. P _o = ∼0.03 for Q1412X in Fig. 3). However, the application of VX‐770 significantly increases the P _o of all three mutants, mainly by shortening the closed time and prolonging the open time as reported previously for WT‐CFTR (Yeh et al. 2015). These macroscopic and microscopic data lead to several conclusions. First, although being a disease‐associated mutation, S1455X shows no gating defect, suggesting that the malfunction may be a result of other biochemical deficits, as demonstrated previously (Moyer et al. 1999). Second, complete deletion of the C‐terminal region (i.e. D1425X) does not dramatically alter the gating behaviour and NBD function. Third, as the deletion impinges on the C‐terminal beta‐strands of NBD2, the gating function deteriorates.

Figure 8. Single‐channel kinetic of S1455X, D1425X and N1419X.

A, real‐time current recording of phosphorylation‐activated S1455X in 2 mm ATP or 2 mm ATP plus 200 nm VX‐770 (left). Open probability and kinetic parameters are shown on the right, with the solid box representing the behaviour in 2 mm ATP and dashed box for that in 2 mm ATP and 200 nm VX‐770. The data were gathered from four experiments. B, same protocol and analysis as in (A) for phosphorylation‐activated D1425X (n = 4). C, protocol and analysis as in (A) for phosphorylation‐activated N1419X (n = 3). Asterisks indicate statistical significance compared to ATP. ^* P < 0.05; ^** P < 0.01; ^*** P < 0.001.

Discussion

In the present study, we first confirmed a previous study reporting that the C‐terminal truncation mutation Q1412X decreases the surface expression level of CFTR (Haardt et al. 1999) and demonstrated that VX‐809 could, to a varying degree, increase expression for all of the truncation mutants tested, including Q1412X, N1419X, D1425X and S1455X. By contrast to the previous study (Haardt et al. 1999) suggesting a lack of gating defect with Q1412X‐CFTR, our data show that the P _o of Q1412X is ∼0.08 using direct measurement or 0.026 by back‐calculation, which is ∼6–18% of that of WT‐CFTR (P_o ∼0.45). Our experimental data, together with the finding that the Q1412X mutation removes two critical beta‐strands at the C‐terminus of NBD2, suggest a dysfunctional NBD2 including an impairment to hydrolyse ATP as the plausible mechanism underlying the abnormal gating in Q1412X. By characterizing C‐terminal truncation mutants with various degrees of deletion in these two beta‐strands, we conclude that structural integrity even at the very end of NBD2 is critical for the functional integrity in CFTR gating.

As originally proposed by Haardt et al. (1999), the major molecular defect of the Class VI mutation is a decreased number of the channels resulted from high turnover rates of CFTR variants in the plasma membrane. Subsequent studies suggested that the underlying mechanism for membrane instability is an increased sensitivity to proteasome‐dependent degradation in post‐Golgi compartments for truncation mutants such as Q1412X (Benharouga et al. 2001). In line with these reports, only Q1412X in our western blot experiment shows a significantly reduced Band C proteins, whereas surface expression of N1419X and other less‐truncated mutations remains relatively normal. Indeed, Gentzsch and Riordan (2001) found that a highly‐conserved hydrophobic patch FLVI (amino acids 1413–1416) is crucial for the stability of mature CFTR proteins in the cell membrane. The deletion of this very segment, as well as the replacement of all amino acids within the segment with alanine, decreases the amount of mature CFTR. This sequence was also proposed to be the binding site for the intermediate filament protein keratin 18, which, when binded to CFTR, can prolong the half‐life of CFTR in the cell membrane (Duan et al. 2012). Regrading the other truncation mutations investigated in the present study, although, statistically, their surface expression was not significantly lower than WT, the apparent decrease in band B and C may be a result of the nonsense‐mediated decay of the mRNA, a well‐known mechanism responsible for the reduced amount of mRNA and protein expression associated with premature stop codon mutations such as G542X and W1282X (Linde et al. 2007). Further studies are needed to investigate whether this mechanism can account for our results and whether different mutations may present varied vulnerability to this mechanism.

Based on the notion that a ∼90% reduction of protein expression itself may not be sufficient to inflict severe CF phenotype, we characterized the gating properties of Q1412X by taking advantage of numerous tools accumulated over the years. Our previous work (Yeh et al. 2015) showing a synergistic action between gating potentiators VX‐770 and nitrate is especially useful in this respect because, for a mutation with a very low P _o, the chance to observe simultaneous openings of all the functional channels in the system is exceedingly small, resulting in an inevitable underestimation of the channel number and overestimation of the P _o. By pushing the P _o to a much higher level, we can more accurately estimate the number of functional channels recorded and, consequently, a more accurate P_o value of 0.026 is attained for Q1412X‐CFTR. Taking into consideration a 10‐fold change in the surface expression, we conclude that, relative to WT channels, the overall reduction of chloride flux caused by the Q1412X mutation could be as much as ∼200‐fold, which more probably explains the severe phenotype seen in patients harbouring this mutation. Note the chloride flux of the severe‐form mutation G551D is suggested to be <1% of WT because of a >100‐fold reduction in P _o (Bompadre et al. 2007; Char et al. 2014; but see also Van Goor et al. 2009).

One interesting and unexpected observation that we made for Q1412X‐CFTR is a biphasic current change upon the activation by PKA and VX‐770 (Fig. 4). Because the experiments were carried out in the continuous presence of PKA and ATP, the decay phase may be a result of thermo‐instability of the truncated protein, similar to that found in rescued F508del‐CFTR (Meng et al. 2017) rather than dephosphorylation. Of note, the rescued F508del (i.e. F508del proteins that have been successfully expressed on the plasma membrane) is also categorized in the Class VI mutation as a result of the decreased half‐life on the membrane (Lukacs et al. 1993; Sharma et al. 2001). The stability of F508del proteins can be enhanced by small molecules such as VX‐809 or by revertant mutations (Loo et al. 2010; Wang et al. 2011; Meng et al. 2017). Whether the mechanism underlying the instability of F508del is shared by Q1412X‐CFTR is beyond the scope of the present study and awaits further investigation.

To tackle the mechanism of the gating defect in Q1412X‐CFTR, we conducted a series of experiments to examine the NBD function. Interestingly, although Q1412X‐CFTR currents are mostly ATP‐dependent, the current relaxation upon ATP removal is extremely slow, as if the ATP can remain bound very tightly in NBDs. This latter observation, together with the observation that eliminating ATP hydrolysis with the E1371S mutation did not affect the already prolonged mean open time of ∼3 s (vs. ∼0.4 s for WT‐CFTR), suggests that ATP hydrolysis in site 2 is impaired in Q1412X, a proposition also supported by studies using the N306D mutation. By introducing mutations that have been shown to lower the binding affinity for ATP in site 1 (W401G) or site 2 (Y1219G), we found that, unexpectedly, the ATP sensitivity is abolished in both mutations (Fig. 9). The reasons underlying this phenomenon are unclear, although one possible explanation is that ATP‐elicited increase of Q1412X‐CFTR currents depends on intact binding of ATP to both ATP‐binding sites, and the monotonic current decay after washing out ATP reflects ATP dissociations from either site. Taken together, these observations suggest a simple scenario: the Q1412X mutation, although it may not eliminate ATP binding at site 2, drastically decreases the rate of NBD dimerization and hence reduces the opening rate of the channel; however, once the channel is open, two ATP molecules are bound in NBD dimer interface and a slower ATP hydrolysis rate in site 2 explains why the channel can remain open for seconds.

Figure 9. ATP responsiveness of W401G/Q1412X and Y1219G/Q1412X.

Real‐time recordings of W401G/Q1412X (A) and Y1219G/Q1412X (B) upon the addition of PKA and ATP, and the removal of ATP. Neither mutation shows significant changes in current amplitude upon the removal of ATP (cf. Fig. 4). The ratio of the current with ATP vs. without ATP is 105 ± 1% (n = 3) for W401G/Q1412X and 111 ± 6% (n = 3) for Y1219G/Q1412X. Of note, the current decay upon phosphorylation‐dependent activation in the presence of VX‐770 is again seen with these double mutants.

To further clarify the role of ATP hydrolysis in the gating cycle of Q1412X, we introduced the N306D mutation into Q1412X, which allows direct visualization of the gating transition in WT‐CFTR (Zhang & Hwang, 2017). As a result of the input of free energy harvested from ATP hydrolysis, the N306D mutation in the WT background shows a predominant and preferred transition of C → O₁ → O₂ → C over C → O₂ → O₁ → C in the presence of ATP. For Q1412X gating in the presence of ATP (Fig. 8 A), the dominance of C → O₁ → C (93% of events) with the absence of a preferred transition of C → O₁ → O₂ → C is consistent with the notion that the hydrolysis process is defective in Q1412X. In those prevalent gating events, the channel opens into a long opening bursts (1.3 ± 0.4 s; cf. Fig. 6) with a smaller single‐channel amplitude, supporting the proposition that the observed open state is equivalent to the stabilized O₁ state seen in N306D trapped in the pre‐hydrolytic open state (Jih & Hwang, 2012; Jih et al. 2012; Zhang & Hwang, 2017). It follows that this O₁ state in N306D/Q1412X‐CFTR should assume an open channel conformation in the transmembrane pore of CFTR with a dimerized NBD. That this very state possesses a hydrolysis‐impaired NBD dimer with two ATP molecules sandwiched in the NBD dimer interface is also consistent with the observation that the O₁ state observed in the presence of ATP is more stable than the O₁ state in the absence of ATP (0.30 ± 0.05 s lifetime). The example traces of events other than C → O₁ → C in the presence of ATP are shown in Fig. 10. Note these rare events display an open time in the range of hundreds of milliseconds resembling events in the absence of ATP (see below). Thus, these events may also represent gate opening without ATP bound in NBDs. Interestingly, we indeed noted that, in the absence of ATP, many more opening events contain transitions between O₁ and O₂ (Table 1) and the percentage of events of C → O₁ → O₂ → C is higher than that of C → O₂ → O₁ → C, an observation apparently contradicting the prediction of equilibrium gating. Because of the extremely low open probability of the Q1412X mutation, it is technically difficult to gather opening events in the absence of ATP; hence, the apparent difference between these two transitions could result from sampling errors.

Figure 10. Sample events other than C → O₁ → C in the presence of ATP.

Events (A) to (D) were categorized as C → O₁ → O₂ →C, C → O₂ → C, C → O₂ → C and C → (O₁ → O₂)ⁿ → C, respectively.

Nevertheless, the presence of clear and more frequent transitions between O₁ and O₂ in the absence of ATP in N306D/Q1412X (Table 1) offers some structure/function insights for our understanding of CFTR gating. As reported previously for mutants exhibiting this O₁/O₂ phenotype (Zhang & Hwang, 2017), the smaller O₁ state has a longer resident time than the larger O₂ state in N306D/Q1412X. The single‐channel amplitudes of these two open states observed in the absence of ATP for N306D/Q1412X are no different from those seen in the presence of ATP or those reported for N306D single mutant (Zhang & Hwang, 2017). Because the conductance level is mostly determined by the structure of the pore, these conductance similarities for both open states recorded in the presence and absence of ATP suggest a similar pore conformation for a particular open state despite expected differences in the status of NBDs between experimental conditions with or without ATP. If we accept the idea that the smaller O₁ state acquires an NBD dimer, it follows that the longer bursting time of the O₁ state seen in the presence of ATP (1.3 s vs. 0.3 s in the absence of ATP) simply reflects a more stable ATP‐bound NBD dimer than a dimer without ATP. Because the transition from O₁ to O₂ must represent a conformational change in the pore embedded in TMDs, we can then consider what might happen to NBDs during this transition. Specifically, is the NBD dimer separated in the O₂ state? Assuming that NBDs remain in a dimeric form in the O₂ state, this means that conformational changes in TMDs can occur without an obligatory motion in NBDs. On the other hand, as proposed previously (Jih & Hwang, 2012; Jih et al. 2012), the conformational changes in TMDs upon transition from the O₁ to O₂ state may be accompanied by a partial separation of the NBD dimer. Either way, the molecular coupling between TMDs and NBDs is not obligatory (Hwang et al. 2018). In the former case, the pore architecture can change without separation of NBD dimer. In the latter case, the NBDs can undergo partial separation, whereas the pore remains open. Of note, if the gate can remain open when NBD dimer is separated, the gate should also be able to close when NBDs are in a dimeric form; the latest cryo‐electromicroscopy structure of zebrafish CFTR indeed shows a non‐conductive conformation with a dimerized NBD (Zhang et al. 2017).

Because the gating function of truncation mutants D1425X and S1455X is fairly normal, whereas N1419X and Q1412X show different degrees of gating abnormalities, we conclude that the C‐terminus region outside of NBD2 is not critical for normal CFTR gating, although the structural integrity of NBD2 even to the last beta‐strand is important. This conclusion echoes research probing the role in gating control of the C‐terminus region (Rich et al. 1993; Gentzsch et al. 2002). However, Ostedgaard et al. (2003) demonstrated that the deletion of an acidic cluster in the C‐terminus influences gating. This discrepancy may be a result of the possibility that the C‐terminal region has both inhibitory and stimulatory effects in different segments. Regarding the functional perturbation of the mutations with deletions over the NBD2, the enlarged structure in Fig. 1 B reveals the structural significance of a beta‐sheet network that consists of three separate beta‐sheets surrounding an alpha‐helix and constitutes a major portion of the head subdomain of a canonical NBD: the molecular scaffold constructed by these beta‐strands may be essential for placing Y1219, K1250 and E1371 at the right position for proper ATP binding and hydrolysis because they are all located in the connecting peptides between these beta‐strands or at the junction between a beta‐strand and an alpha‐helix. Taking the P‐loop (phosphate‐binding loop, also known as Walker A motif) for example, this is a crucial motif for ATP binding and hydrolysis in CFTR and is well‐conserved across ABC protein superfamily (Wilkens, 2015; Hwang et al. 2018). The P‐loop is usually flanked by a beta‐strand and an alpha‐helix, which are formed by amino acids R1239–L1243 and G1249–R1259, respectively (Liu et al. 2017). The beta‐strand R1239–L1243 is the first of the three beta‐strands at the C‐terminus of NBD2 including Q1412–E1417 and K1420–Y1424. Therefore, the Q1412X mutation probbaly disturbs the function of the P‐loop and hence influences ATP binding and hydrolysis at site 2 (Fig. 1). This mechanistic insight also highlights the importance of interpreting functional characteristics in a structural context.

We therefore speculate that gating dysfunction seen in N1419X and Q1412X is the result of different degrees of structural perturbations of these beta‐strands because of loss of the hydrogen‐bonds that stabilize this network. This hypothesis further predicts that the Class VI mutations with more truncation than Q1412X (e.g. L1399X and I1383X) will have even worse gating function, a proposition that could be tested in the future. Of note, N1419X and Q1412X probably retain some function of NBD2 as described above. The structural perturbations by these two mutations may not be so severe as to eliminate NBD dimerization, although our data do support the notion that the NBD dimer in Q1412X‐CFTR may be destabilized. Together with an impairment of ATP hydrolysis, which prolongs the open time, the Q1412X‐CFTR assumes a P _o of 0.026. By contrast, the role of site 2 in gating is diminished in the G551D mutation, where NBD dimerization is prevented by an electrorepulsion between bound ATP at NBD2 and the introduced aspartate at the signature sequence of NBD1 (Lin et al. 2014), resulting in a further 10‐fold reduction of P _o. Thus, depending on the exact nature of structural changes by mutations, a spectrum of functional consequences emerges.

In addition to the mechanistic insights elaborated above, our results also bear clinical implications. To date, therapeutic strategies for treating patients with the Class VI mutation are focused on developing ‘membrane stabilizers’ to enhance the stability of mutant CFTR. To our knowledge, only one compound, N91115, has been evaluated in phase II clinical trials (Donaldson et al. 2017), which unfortunately turned out to be unsuccessful (https://www.cff.org/Trials/Pipeline/details/116/Cavosonstat-N91115). Our results show promising functional rectification using the Food and Drug Administration approved drugs Ivacaftor (VX‐770) and Lumacaftor (VX‐809). More importantly, as reported by Veit et al. (2014), Q1412X, unlike F508del, is insensitive to the abrogating effect of VX‐770 on VX‐809, suggesting that coapplication of these two drugs is expected to exert a synergistic effect on promoting transepithelial chloride transport. Based on our in vitro data presented in the present study, we calculate that a maximal chloride transport of up to 10% of WT can be achieved when Q1412X is treated with Lumacaftor and Ivacaftor. This level of activity can be accomplished by Ivacaftor alone in the case of G551D, comprising a severe phenotype mutation with pure gating defects (Lin et al. 2014). Considering the beneficial clinical outcomes seen in G551D patients taking Ivacaftor, we are optimistic that CF patients carrying the Q1412X mutation could benefit from combinational therapy with Orkambi. Indeed, Ivacaftor has been approved for several premature stop codon mutations such as E831X (https://investors.vrtx.com/news-releases/news-release-details/fda-approves-kalydecor-ivacaftor-more-600-people-ages-2-and). In summary, the present study highlights the importance of characterizing the functional properties of every mutation, no matter what class it currently belongs to. By interpreting the functional characteristics in the context of the full‐length atomic structure of the CFTR protein (Liu et al. 2017) and surveying the response to commonly used clinical compounds, not only can a better understanding of the underlying structural mechanisms for the mutation‐induced CFTR dysfunction be attained, but also individualized and precise clinical treatments can be devised.

Additional information

Competing interests

The authors declare that they have no competing interests.

Author contributions

TCH designed and supervised the project. TCH, YCY and JTY were involved in writing and revising the manuscript. JTY and YCY performed the experiments. JTY analysed the data and wrote the initial draft. All authors approved the final version of the manuscript submitted for publication.

Funding

The work is supported by NIHR01DK55835 and a grant from the Cystic Fibrosis Foundation (Hwang15G0). JTY is a recipient of a scholarship from The Taipei Veterans General Hospital – National Yang‐ Ming University Excellent Physician Scientists Cultivation Program in Taiwan (no. 102‐Y‐A‐001).

Biography

Jiunn‐Tyng Yeh is a PhD student in the Interdisciplinary Neuroscience Program at the University of Missouri‐Columbia, USA. He studied medicine at the National Yang‐Ming University in Taiwan and came to the USA with the support of The Taipei Veterans General Hospital – National Yang‐Ming University Excellent Physician Scientists Cultivation Program. His research focuses on using the electrophysiological method to understand the function and dysfunction of CFTR.