Potentiation of the cystic fibrosis transmembrane conductance regulator by VX‐770 involves stabilization of the pre‐hydrolytic, O₁ state

Abstract

Background and Purpose

Cystic fibrosis (CF) is a debilitating hereditary disease caused by mutations in the cystic fibrosis transmembrane conductance regulator (CFTR) gene, which encodes an anion channel. Wild type‐CFTR gating is a non‐equilibrium process. After ATP binding, CFTR enters a stable open state (O₁). ATP hydrolysis leads it to a short‐lived post‐hydrolytic open state (O₂), from which channels close. Here, we use mutations to probe the mechanism of VX‐770, the first compound directly targeting the CFTR protein approved for treatment of CF. D1370N and K1250R mutations reduce or abolish catalytic activity, simplifying the gating scheme to an equilibrium (C↔O₁); K464A‐CFTR has a destabilized O₁ state and rarely closes via hydrolysis.

Experimental Approach

Potentiation by VX‐770 was measured using microscopic imaging of HEK293 cells expressing an anion‐sensitive YFP‐CFTR. A simple mathematical model was used to predict fluorescence quenching following extracellular iodide addition and estimate CFTR conductance. Membrane density of CFTR channels was measured in a parallel assay, using CFTR‐pHTomato.

Key Results

VX‐770 strongly potentiated WT‐CFTR, D1370N‐CFTR and K1250R‐CFTR. K464A‐CFTR was also strongly potentiated, regardless of whether it retained catalytic activity or not.

Conclusions and Implications

Similar potentiation of hydrolytic and non‐hydrolytic mutants suggests that VX‐770 increases CFTR open probability mainly by stabilizing pre‐hydrolytic O₁ states with respect to closed states. Potentiation of K464A‐CFTR channels suggests action of VX‐770 did not strongly alter conformational dynamics at site 1. Understanding potentiator mechanism could help develop improved treatment for CF patients. The fluorescence assay presented here is a robust tool for such investigations.

Abbreviations

ΔF_M

change in fluorescence reporting on membrane‐localized CFTR‐pHTomato

τ_trans

time constant of the transient anion conductance

ABC

ATP‐binding cassette

CF

Cystic fibrosis

CFTR

Cystic fibrosis transmembrane conductance regulator

F/F_max

normalized YFP fluorescence

F_pHTomato

weighted average fluorescence obtained in pHTomato assay

G_CFTR

CFTR conductance

G_trans

transient anion conductance

IRES

internal ribosome entry site

MES

2‐(N‐morpholino) ethanesulfonic acid

NBD

nucleotide binding domain

P_o

open probability

SSR

sum of squared residuals

V_M

membrane potential

WT

wild type

YFP

yellow fluorescent protein

Introduction

Mutations in the cystic fibrosis transmembrane conductance regulator (CFTR) gene cause cystic fibrosis (CF), a common lethal hereditary disease (Riordan et al., 1989). http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=707 is a member of the ATP‐binding cassette (http://www.guidetopharmacology.org/GRAC/FamilyDisplayForward?familyId=136) superfamily of transporters, but it is unique in its function as an ion channel, allowing passive anion flow across the plasma membrane.

http://www.guidetopharmacology.org/GRAC/FamilyDisplayForward?familyId=284‐dependent phosphorylation of the unique regulatory (R)‐domain is required for sustained gating (Chang et al., 1993). Once channels are phosphorylated, ATP binding at two cytosolic nucleotide binding domains (NBDs) favours pore opening in the transmembrane domains. Opening is coupled to formation of an NBD dimer, with ATP bound at composite binding sites formed at the NBD interface (Figure 1; Vergani et al., 2003, 2005). The channel remains open until ATP hydrolysis at catalytic site 2 (Csanády et al., 2010) triggers entry into a short‐lived, post‐hydrolytic O₂ state, from which closing occurs rapidly (Gunderson and Kopito, 1995). ATP at site 1 remains bound for many gating cycles (Tsai et al., 2010), although conformational changes associated with gating and linked to partial site 1 opening of the dimer interface have been demonstrated (Csanády et al., 2013; Chaves and Gadsby, 2015).

Figure 1.

CFTR gating mechanism. The channel enters a stable open state (O₁) following formation of a head‐to‐tail NBD dimer, with ATP bound at the two ATP‐binding sites formed at the interface. Upon hydrolysis of the ATP bound at site 2 (lower site), the channel closes, via a short‐lived post‐hydrolytic (O₂) state. Whereas >95% of WT‐CFTR opening events terminate via hydrolysis of ATP, for the mutants, ~100% of D1370N‐CFTR and K1250R‐CFTR and ~80% of K464A‐CFTR opening events terminate via non‐hydrolytic closing. NBD1 = blue, NBD2 = green, TMDs = grey, ATP = yellow, ADP + P_i = orange.

The approval of http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=4342 (Ivacaftor; Vertex Pharmaceuticals, Cambridge, MA, USA) marked a step forward in treatment for CF. VX‐770 was the first drug to target the CFTR channel directly, rather than disease symptoms. It is now approved to treat CF caused by a number of gating mutations (Van Goor et al., 2009, 2014; Yu et al., 2012), and the clinical results have been very positive (Ramsey et al., 2011; Davies et al., 2013).

VX‐770 is a potentiator, a compound which increases the open probability (P_o) of CFTR (Van Goor et al., 2009). At millimolar, physiological, ATP concentrations, it causes a marked prolongation of mean open time, and a modest increase in opening rate, as seen in single channel patch‐clamp experiments (Yu et al., 2012; Jih and Hwang, 2013). VX‐770 increases the mean open time of both wild type (WT)‐CFTR channels, which gate via the non‐equilibrium mechanism outlined above (Vergani et al., 2005; Csanády et al., 2010), and G551D‐CFTR channels, which gate via an equilibrium mechanism retaining little ATP sensitivity (Bompadre et al., 2007; Lin et al., 2014). It has been hypothesized that VX‐770 achieves this mainly via stabilization of the post‐hydrolytic O₂ state (Jih and Hwang, 2013). However, Kopeikin et al. (2014) demonstrate that the closing rate of E1371S‐CFTR, a non‐hydrolytic mutant (Vergani et al., 2003), is slowed by VX‐770. Slowed closure of E1371S‐CFTR channels cannot be caused by a stabilized O₂ state. To further investigate how VX‐770 interacts with open CFTR channels, gating mutants, whose gating kinetics have been extensively characterized, were used to alter open state stability and thus manipulate the proportion of time spent in the O₁ and O₂ states.

WT‐CFTR open bursts are initiated after ATP‐binding at site 2 (lower site; Figure 1) and terminated following hydrolysis of that ATP. The rate of non‐hydrolytic closure (k₋₁, O₁ → C) is slow compared to hydrolysis (k₁, O₁ → O₂), so almost no events close via the backward non‐hydrolytic closing pathway (Figure 1; Csanády et al., 2010), resulting in non‐equilibrium cycling between conformations (central arrow, Figure 1).

Mutations of catalytic residues in site 2 can block the hydrolytic pathway to closure. D1370 and K1250 are the Walker B aspartate and the Walker A lysine, respectively, in NBD2 (Walker et al., 1982). Atomic level determination of the structure of ABC transporters and CFTR (Yuan et al., 2001; Smith et al., 2002; Chen et al., 2003; Zhang et al., 2017) confirmed their role as key catalytic residues within site 2. The corresponding mutations in other ABC transporters have been shown to abolish hydrolytic activity (Urbatsch et al., 1995; Lerner‐Marmarosh et al., 1999; Rai et al., 2006). In CFTR, prolonged burst durations are observed for D1370N‐CFTR and K1250R‐CFTR (Vergani et al., 2003, 2005; Csanády et al., 2010), D1370N‐CFTR does not enter a higher conductance post‐hydrolytic state discernible in specific conditions (Gunderson and Kopito, 1995), and open dwell‐time distributions are consistent with virtually all opening events terminating via the non‐hydrolytic closing pathway (Csanády et al., 2010). The underlying gating cycle is likely to be very similar for K1250R‐CFTR, with both non‐hydrolytic mutants never entering the post‐hydrolytic O₂ state during ATP‐dependent gating.

K464 is the Walker A lysine of NBD1. Despite being part of the degenerate, non‐catalytic site 1, the alanine replacement in K464A‐CFTR reduces the turnover rate of ATP hydrolysis ~10‐fold (Ramjeesingh et al., 1999). Dwell‐time distribution analysis suggests that this is due to the vast majority of K464A‐CFTR opening events closing via the non‐hydrolytic pathway, thus reducing the proportion of openings undergoing hydrolysis (Figure 1; Csanády et al., 2010). This, and other experimental results (Powe Jr et al., 2002; Vergani et al., 2003; Bompadre et al., 2005), suggests that the K464A mutation destabilizes the pre‐hydrolytic open state O₁, with respect to the opening transition state, decreasing the energetic barrier for backward non‐hydrolytic closure. Because VX‐770 slows closure of E1371S‐CFTR channels (Kopeikin et al., 2014) – which is most simply interpreted as due to a stabilization by the drug of the pre‐hydrolytic open state O₁ with respect to the opening transition state – VX‐770 and the K464A mutations appear to affect O₁ stability in opposite ways. To investigate whether the drug and site 1 mutation might be affecting CFTR structure and dynamics via a similar mechanism, we tested the effect of VX‐770 on K464A‐CFTR channels.

We used cell‐based assays exploiting genetically encoded fluorescent proteins tagged to CFTR (Langron et al., 2017), to study how VX‐770 alters CFTR gating. In this system, the CFTR proteins are embedded in mammalian membranes and exposed to a physical and chemical environment, such as physiological cytosolic ionic and ATP concentrations, presence of cytoskeleton and of anchoring proteins, very close to those present in the native epithelial cells. While such fluorescence assays do not have the exquisite sensitivity of electrophysiological techniques allowing detection of gating events at the single channel level, they provide a robust, informative readout of the overall functional changes caused by drugs in populations of CFTR channels.

Methods

Cell culture

HEK‐293 cells (ATCC, Middlesex, UK) were maintained in DMEM, supplemented with 2 mM L‐glutamine, 100 U·mL⁻¹ penicillin and streptomycin, and 10% FBS (Life Technologies, Paisley, UK). For fluorescence imaging, cells were seeded in poly‐D‐lysine‐coated, black walled 96‐well plates (Costar, Fisher Scientific, Leicestershire, UK).

Plasmids and transfections

eGFP‐CFTR in pcDNA3.1 (a gift from Bruce Stanton, Geisel School of Medicine, NH, USA), in which eGFP is tagged to the N‐terminal of CFTR via a 23 amino‐acid linker, was mutated to yellow fluorescent protein (YFP)‐CFTR using site‐directed mutagenesis (introducing F46L, L64F, S65G, V68L, S72A, H148Q, I152L and T203Y, see Langron et al., 2017; Quikchange protocol, Stratagene, CA, USA). pIRES2‐eGFP‐CFTR was a gift from David Gadsby (Rockefeller University, NY, USA). pHTomato (provided by Dr Li, Peking University, China, and Prof. Richard Tsien, NYU School of Medicine, NY, USA) was inserted after position 901 in CFTR, using a primer overlap‐extension strategy. pIRES2‐CFTR‐pHTomato contained eGFP under control of an internal ribosome entry site (IRES), which allows transcription of a single mRNA containing eGFP and CFTR‐pHTomato sequences, but translation of the two as separate proteins (Langron et al., 2017). Point mutations were introduced to pcDNA3.1‐YFP‐CFTR and pIRES2‐CFTR‐pHTomato using site‐directed mutagenesis (Quikchange protocol, Stratagene).

Lipofectamine transfection was used for all experiments. Cells plated in 96‐well plates were transiently transfected with the appropriate plasmid using Lipofectamine 2000 (Life Technologies), according to the manufacturer's instructions. Following transfection, cell plates were returned to 37°C for 24 h. Chronic VX‐809 treatment, where indicated, was started 24 h after transfection and lasted 24 h.

The N‐terminal YFP tag has been shown to only mildly alter function compared to the untagged channel (Langron et al., 2017), and others have shown that the GFP tag on the N‐terminal, which we have used as a template for generating the YFP, does not significantly affect trafficking or function (Moyer et al., 1998; Vais et al., 2004). The pHTomato tag – inserted within the fourth extracellular loop – is also at a site known to be permissive for tag insertion in CFTR (Howard et al., 1995; Phuan et al., 2014; Veit et al., 2014; Hildebrandt et al., 2015).

YFP‐CFTR fluorescence imaging

All imaging was carried out using ImageXpress (ImageXpress Micro XLS, Molecular Devices, San Jose, CA, USA), an image‐acquisition system equipped with wide‐field inverted fluorescence microscope, CMOS camera and fluidics robotics; 96‐well cell plates were held in an environmental chamber, at 37°C. YFP was imaged using a 20× objective and excitation/emission filters 472 ± 30 and 520 ± 35 nm. For each plate, the laser intensity and exposure were optimized to achieve the highest possible fluorescence while avoiding both photobleaching and saturation (illumination intensity 100–150/225 c.d., and exposure 0.1–0.2 s). Images were taken at a frequency of 0.5 Hz.

To avoid systematic errors and achieve a degree of randomization, conditions were not always tested in the same order. Blinding of these experiments was not feasible, given the resources available. However, the automated acquisition and rigorous analysis applied to the data effectively prevents subjective bias.

Before imaging, cells were washed twice with 100 μL standard buffer (140 mM NaCl, 4.7 mM KCl, 1.2 mM MgCl₂, 5 mM HEPES, 2.5 mM CaCl₂, 11 mM glucose, pH 7.4). During imaging, after 20 s of baseline fluorescence acquisition, CFTR was activated in the absence of extracellular iodide (I⁻), by addition of 50 μL standard buffer containing activating compounds. For steady‐state experiments, after a further 230 s, 50 μL extracellular I⁻ was added (as standard buffer with 140 mM NaCl replaced with 400 mM NaI; resulting in final [I⁻] of 100 mM). For experiments carried out to measure the rate of CFTR activation (see Figure 3), the second addition occurred after a pre‐incubation time of variable length (0–320 s). Activating compounds were included in the second addition so as not to alter final extracellular concentrations.

Figure 3.

YFP‐CFTR protocol optimization for steady state measurements. (A) WT‐CFTR fluorescence time course during incubation with 0 or 10 μM forskolin (black bar) and 100 mM extracellular I⁻ (red bar). The length of pre‐incubation (blue arrow) was altered to investigate activation time course in HEK293 cells. (B) Rate of I⁻ entry following incubation with 1 or 10 μM forskolin as a function of pre‐incubation time, for WT‐CFTR. 1 μM τ = 68.5 ± 15 s, 10 μM τ = 91.7 ± 15 s, n = 2–10, from three plates (for n values corresponding to each individual point, see Supporting Information Data S6). Dashed line indicates time chosen for subsequent experiments, 230 s, beyond which CFTR activation has reached a steady state. (C) Rate of I⁻ entry following pre‐incubation with 10 μM forskolin, for non‐hydrolytic mutants of YFP‐CFTR. D1370N‐CFTR τ = 64.9 ± 9 s, K1250R‐CFTR τ = 57.1 ± 8 s, n = 2–6, from two plates (see Supporting Information Data S6). (D) Rate of I⁻ entry following pre‐incubation with 10 μM forskolin for K464A‐CFTR, τ = 105.1 ± 27 s, n = 3–6, from two plates (see Supporting Information Data S6).

YFP‐CFTR data analysis

Images were analysed using ImageJ (http://rsbweb.nih.gov/ij/). For each well, data were exported as a stack, with each time point represented by an image in the stack. Fluorescent areas corresponding to transfected cells were selected, before addition of I⁻, using a threshold. Fluorescence was normalized to this maximal value, to allow comparison of fluorescence values between images, despite the variation in transfection efficiency and in other factors influencing absolute fluorescence values. CFTR activation was quantified by one of two methods, depending on the experiment type.

Quantification of non‐stationary CFTR activity

For experiments used to measure the time course of CFTR activation (see Figure 3), gating was quantified using the maximal rate of I⁻ influx, as described previously (Langron et al., 2017). Briefly, anion binding to YFP abolishes fluorescence in our system, so normalized fluorescence quantifies the proportion of unbound chromophores, and can be described by an equation including a Hill–Langmuir component:

$$F/F_{\mathit{max}}=1-\frac{{\left[I^{-}\right]}_{\mathit{\text{in}}}}{K_I+{\left[I^{-}\right]}_{\mathit{\text{in}}}}$$ (1)

where $F/F_{\mathit{max}}$ is the normalized fluorescence, K _I is the K _D for I⁻ binding to YFP (1.9 mM; Galietta et al., 2001a) and [I⁻]_in is the concentration of intracellular I⁻.

Equation [(1)] can be rearranged to express [I⁻]_in as a function of $F/F_{\mathit{max}}$:

$${\left[I^{-}\right]}_{\mathit{\text{in}}}=K_I\frac{1-F/F_{\mathit{max}}}{F/F_{\mathit{max}}}$$ (2)

The maximal rate of I⁻ entry into cells $\left(\frac{\mathrm{d}{\left[{\mathrm{I}}^{-}\right]}_{\text{in}}}{\mathit{dt}},\mathrm{mM}{\mathrm{s}}^{-1}\right)$, immediately upon extracellular addition of I⁻, was used to quantify CFTR activity.

To describe the time course of CFTR activation (e.g. Figure 3B–D), the rate of I⁻ entry, as a function of pre‐incubation time, was fit to a single exponential rise to maximum equation:

$$\mathrm{y}=\mathrm{a}\left(1-{\mathrm{e}}^{-\frac{\mathrm{t}}{\mathrm{\tau }}}\right)$$

where a is the amplitude of the curve, t is the pre‐incubation time and τ is the time constant of the exponential rise.

Quantification of CFTR activity at steady state

For steady‐state experiments, a simple mathematical model was used to fit fluorescence quenching measurements and estimate CFTR conductance (see Supporting Information Data S1). Previously, a similar model was used to validate assumptions made when quantifying CFTR activity monitored during activation (Langron et al., 2017). In the current work, CFTR activation was first allowed to reach steady‐state in the absence of I⁻. Then, the changes in fluorescence values sampled in the 40 s following I⁻ addition were fitted to the proportion of anion‐free YFP chromophore predicted by the model (both normalized to the values read at the time point before I⁻ addition). Estimates for four free parameters are obtained: CFTR conductance at steady‐state (G_CFTR), membrane potential at steady‐state (V_M), and conductance (G_trans) and time constant (τ_trans) of a transient, non‐CFTR anion conductance. It was necessary to introduce the transient component to adequately describe the quenching time course. Endogenous anion permeabilities of the HEK293 cells, possibly triggered by a small (<10 mV) hyperpolarization predicted by the model upon I⁻ addition, are likely to underlie this component, because it was present in the absence of CFTR activation (average values of negative controls: G_trans: 13.2 ± 1.6 nS; τ_trans: 4.9 ± 0.4 s, n = 19), its magnitude had no dependence on mutation and was not correlated to CFTR conductance (r = 0.16, P > 0.05, Pearson's product moment correlation test). Furthermore, an anion conductance with a small amplitude, falling to zero within ~15 s of I⁻ addition, was also detected in cells expressing G551D‐CFTR (Supporting Information Data S2), that is expressing a YFP‐CFTR variant with a cellular distribution similar to WT‐CFTR, but with negligible CFTR‐associated conductance (Bompadre et al., 2007).

When CFTR conductance was low, the quenching time course allowed reliable estimation of all four free parameters, including G_trans and τ_trans. However, when CFTR conductance was high, due to the fast quenching rate, the time course did not provide enough information to uniquely identify all four parameters. Therefore, all quenching curves were fit in two ways: (1) with all four parameters estimated by fitting (Figure 2, left panels, with G_trans constrained to <25 nS, based on the distribution of estimates from negative controls) and (2) with G_trans and τ_trans fixed to the average obtained from negative controls, and only G_CFTR and V_M unconstrained (Figure 2, right panels).

Figure 2.

WT‐CFTR quenching curves fitted using mathematical model to obtain estimates of CFTR conductance and V_M. Experimental data are shown as normalized fluorescence values (yellow circles). Black lines show time course of anion occupancy of the YFP chromophore predicted by the model: proportion of I⁻‐bound YFP (dotted line, p_RI), proportion of Cl⁻‐bound YFP (dashed line, p_RCl) and proportion of unoccupied YFP (i.e. fluorescent YFP, solid line, p_R/p_R(t0)). The latter proportion, like fluorescence, is normalized to the time point immediately preceding I⁻ addition. (A) 100 nM forskolin quenching curve, and (B) 10 μM forskolin quenching curve. Left panels: fitted using method (1), in which G_trans and τ_trans are free parameters. Right panels: fitted using method (2), in which G_trans and τ_trans are constrained to the mean values obtained when CFTR is not activated (13.2 nS and 4.9 s respectively).

The sums of squared residuals (SSRs) obtained with the two alternative fits were then compared. If the fit obtained using method (1) was substantially better than that obtained using method (2) (greater than fivefold improvement in SSR), G_CFTR and V_M were estimated using method (1). This was most common in conditions in which CFTR conductance was low (Figure 2A).

If the fit obtained using method (1) was not substantially better (less than fivefold improvement in SSR), it was assumed that the parameters characterizing the transient component had the values measured for the negative controls, and G_CFTR and V_M were estimated using method (2) (Figure 2B, also see Supporting Information Data S1). Preliminary studies had identified conditions in which quenching curves reached a steady state $F/F_{\mathit{max}}$ within 2 s after I⁻ addition (corresponding to a G_CFTR greater than ~200 nS), that is quenching time course was too fast to be followed with our current image‐acquisition frequency (0.5 Hz). To avoid hitting this upper limit in our assay's dynamic range, a low forskolin concentration was used for basal activation.

Simulations and fits were run using MATLAB software, and the code is available upon request.

Quantitative description of concentration–response curves

Concentration–response curves were fitted using the Hill equation, with the n _H, the Hill coefficient, fixed to 1:

$$\mathrm{y}={\mathrm{y}}_0+\frac{\mathrm{a}\left[\mathrm{L}\right]}{{\mathrm{EC}}_{50}+\left[\mathrm{L}\right]}$$

where EC₅₀ is the half maximal effective concentration, y ₀ is the response measured in the absence of ligand, L is ligand and a is the amplitude of the curve.

CFTR‐pHTomato fluorescence imaging

Before imaging, cells were incubated with Hoechst nuclear stain for 20 min at 37°C. Cells were then washed twice with 100 μL standard buffer (as above). During imaging, extracellular pH was changed using the addition of 50 μL pH 6 buffer (as standard buffer, with 5 mM HEPES replaced with 10 mM 2‐(Nmorpholino) ethanesulfonic acid (MES): final [MES] 3.3 mM, final extracellular value approximately pH 6.5), followed by 50 μL pH 9 buffer (as standard buffer, with 5 mM HEPES replaced with 100 mM Tris: final [Tris] 25 mM, extracellular approximately pH 8.8). The change in fluorescence upon pH change (from pH 6.5 to pH 8.8) was used to quantify membrane‐localized CFTR.

pHTomato images were taken at a frequency of 0.5 Hz, using excitation/emission filters 531 ± 20 and 592 ± 20 nm. Single eGFP and Hoechst nuclear stain images were also acquired for each well, using excitation/emission filters 472 ± 30 and 520 ± 35 nm, and 377 ± 25 and 447 ± 30 nm respectively.

CFTR‐pHTomato data analysis

Analysis was carried out as previously described (Langron et al., 2017). In brief, regions corresponding to transfected cells were selected based on eGFP fluorescence, and for each selected region, the mean eGFP fluorescence (F _green) was used to normalize the mean pHTomato fluorescence (F _red), to allow for differences in transfection efficiency. A weighted average was then obtained, with each region weighted by cell count:

$${\mathrm{F}}_{\text{pHTomato}}=\frac{\sum \left(\frac{{\mathrm{F}}_{\mathrm{red}}}{{\mathrm{F}}_{\text{green}}}\text{cell count}\right)}{\sum \left(\text{cell count}\right)}$$

The change in F _pHTomato upon changing the extracellular buffer from pH 6.5 to pH 8.8 was used to quantify membrane‐localized CFTR (ΔF _M). ΔF _M measurements were normalized to within‐plate WT‐CFTR controls, to allow comparison of fluorescence changes between plates.

Data and statistical analysis

All average plots including error bars represent mean ± SEM. No pattern or abnormality was noticed, when considering individual measurements, that is not obvious in the data as presented. Comparisons were made using a non‐parametric Kruskal–Wallis one‐way ANOVA, followed, when a difference between groups was identified (F achieved P < 0.05), by a post hoc Dunn's test (for comparison to a control group). Significance indicates P < 0.05. All statistical analysis was carried out using SigmaPlot v11 (Systat Software). The data and statistical analysis comply with the recommendations on experimental design and analysis in pharmacology (Curtis et al., 2015), and its update (Curtis et al., 2018).

Materials

VX‐770 and VX‐809 were purchased from Selleck Chemicals (Munich, Germany). All other chemicals were purchased from Sigma‐Aldrich (Dorset, UK).

Nomenclature of targets and ligands

All measurements presented in this paper were obtained from HEK293 cells expressing CFTR fusions with fluorescent proteins. In order not to overly complicate nomenclature, only the mutation is mentioned when referring to a CFTR version, omitting the fused fluorescent protein (e.g. K1250R‐CFTR, rather than K1250R‐YFP‐CFTR), unless the focus of the sentence/paragraph is on the assay, rather than on the CFTR version (YFP‐CFTR assay).

Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY (Harding et al., 2018), and are permanently archived in the Concise Guide to PHARMACOLOGY 2017/18 (Alexander et al., 2017a,b).

Results

YFP‐CFTR protocol optimization for steady‐state measurements

CFTR activity was measured using a YFP‐CFTR fusion protein (Langron et al., 2017), in which halide‐sensitive YFP (H148Q/I152L; Galietta et al., 2001a) is tagged to the N‐terminal of CFTR. A higher affinity for I⁻ compared to Cl⁻ (1.9 mM K_I vs. 85 mM K_Cl; Galietta et al., 2001a) allows the use of Cl⁻/I⁻ exchange protocols to investigate CFTR ion channel function (see Galietta et al., 2001b).

In most of this paper, YFP‐CFTR, expressed in HEK293 cells, is first fully activated by http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=5190 in the absence of extracellular I⁻. Forskolin, by activating http://www.guidetopharmacology.org/GRAC/FamilyDisplayForward?familyId=257, results in phosphorylation and activation of YFP‐CFTR. Upon addition of extracellular I⁻, I⁻ enters the cell via open anion channels and, as the cytosolic [I⁻] increases, the proportion of YFP chromophores in which the anion‐binding site is unoccupied falls, causing a fall in YFP fluorescence (Figure 2). This protocol, by bypassing the activation time course, minimizes the number of free parameters required to describe the YFP quenching time course (cf. Langron et al., 2017).

However, we first needed to determine the length of the forskolin pre‐incubation required for CFTR to reach steady‐state activation. Therefore, the pre‐incubation time interval (Figure 3A, blue arrow) was varied from 0 to 320 s, after which 100 mM I⁻ was added extracellularly (Figure 3A, red bar).

In the presence of 1 and 10 μM forskolin, WT‐CFTR reached maximal activation after 150–200 s (Figure 3B). To confirm that the D1370N, K1250R and K464A mutations did not greatly alter the rate of CFTR activation, we constructed comparable activation curves for each genotype (Figure 3C,D) and found that the activation time constants were not significantly altered by the mutations (Figure 3B–D, see also Supporting Information Data S3). To limit the length of experiments, while also ensuring CFTR function was measured close to steady‐state, 230 s was chosen as a pre‐incubation time for all further experiments (Figure 3B–D, dashed line).

Membrane exposure of CFTR

CFTR conductance in the YFP‐CFTR assay depends on the number of channels at the membrane, as well as on gating and permeation properties of individual channels. To deconvolve mutation effects on one or the other of these quantities, we require information on how the mutation in question affects plasma membrane density, when CFTR is expressed in HEK293 cells. For instance, we see that the maximal I⁻‐entry rate for K464A‐CFTR in 10 μM forskolin is dramatically reduced compared to that for WT‐CFTR (Figure 3D vs. 3B). How much of this is due to a defect in trafficking (Thibodeau et al., 2010), and how much to a defect in gating (Vergani et al., 2003; Csanády et al., 2010)?

To answer such questions, membrane exposure of CFTR was measured using the CFTR‐pHTomato fusion protein (Langron et al., 2017). In this probe, pHTomato (Li and Tsien, 2012) an acid‐sensitive red fluorescent protein, is tagged to the fourth extracellular loop of CFTR (known to be an insertion‐permissive site; Howard et al., 1995; Hildebrandt et al., 2015). Only for CFTR‐pHTomato molecules located at the plasma membrane is pHTomato exposed to the extracellular buffer. Raising the extracellular pH (from pH 6.5 to pH 8.8, which corresponds to a change in pHTomato normalized fluorescence intensity approximately from 0.1 to 0.9 (Li and Tsien, 2012), increases fluorescence. This fluorescence change is used to quantify the amount of CFTR on the membrane.

The D1370N and K1250R mutations did not significantly change membrane exposure compared to WT‐CFTR (Figure 4A). The K464A‐CFTR trafficking defect (Thibodeau et al., 2010), however, is severe in this system, resulting in only ~10% K464A‐CFTR reaching the membrane compared with WT‐CFTR (Figure 4B). Treatment with the corrector compound http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=7481, for 24 h, partly corrects trafficking (membrane exposure reaching ~35% of WT‐CFTR; Figure 4B).

Figure 4.

Membrane exposure of CFTR proteins, expressed as % WT‐CFTR exposure. (A) WT‐CFTR (n = 59, from 11 plates), D1370N‐CFTR (n = 32 from six plates) and K1250R‐CFTR (n = 30 from six plates). (B) WT‐CFTR (n = 59 from 12 plates) and K464A‐CFTR ± 24 h 10 μM VX‐809 (n = 42 from eight plates in absence of VX‐809, and n = 35 from seven plates in presence of VX‐809). (C) WT‐CFTR (n = 41 from eight plates) and K464A/K1250R‐CFTR ± 24 h 10 μM VX‐809 (n = 33 from six plates in absence of VX‐809, and n = 18 from four plates in presence of VX‐809). *P < 0.05, significantly different as indicated.

A very similar trafficking defect, which is partly corrected by VX‐809, is also found for the K464A/K1250R double mutant (Figure 4C). In all further experiments, K464A‐CFTR and K464A/K1250R‐CFTR were treated with 10 μM VX‐809 for 24 h before imaging, to increase the magnitude of CFTR‐mediated conductance, allowing more accurate functional studies.

Potentiation by VX‐770

The conductance of CFTR at steady state (estimated by fitting the YFP quenching time course following I⁻ addition to the change in unoccupied chromophores predicted by a simple mathematical model, see Methods and Supporting Information Data S1) was used to compare CFTR function in different genotypes and conditions.

CFTR conductance was estimated after activation with increasing concentrations of forskolin, and concentration–response curves were constructed (Figure 5) to characterize constructs and identify a suitable forskolin concentration at which to study potentiation. Maximal CFTR conductance is smaller for constructs including the K464A mutation (Figure 5C–D vs. Figure 5A–B), probably reflecting reduced membrane exposure (Figure 4) and known gating impairment (Carson et al., 1995; Vergani et al., 2003; Csanády et al., 2010). Fitting also provides estimates of membrane potential, V_M, once steady‐state activation of CFTR is reached. As expected, forskolin caused a concentration‐dependent depolarization of cells expressing CFTR, producing concentration–response curves (Supporting Information Data S4) comparable to those shown in Figure 5, although estimates for V_M appear less well defined (e.g. more sensitive to initial parameter setting), than those for G_CFTR. WT‐CFTR, D1370N‐CFTR, K1250R‐CFTR and K464A‐CFTR G_CFTR curves are all similarly positioned on the forskolin concentration axis (EC₅₀ values in the low micromolar range; no significant difference), suggesting similar sensitivity to phosphorylation. The K464A/K1250R‐CFTR G_CFTR curve is shifted slightly but significantly compared to that of the WT‐CFTR, suggesting a possible slight increase in sensitivity to cAMP/phosphorylation.

Figure 5.

Forskolin concentration–response curves. CFTR conductance estimated using model. (A) WT‐CFTR: EC₅₀ = 3 ± 1.2 μM. (B) D1370N‐CFTR: EC₅₀ = 1.5 ± 0.6 μM. K1250R‐CFTR: EC₅₀ = 2.2 ± 1.3 μM. (C) VX‐809‐corrected K464A‐CFTR: EC₅₀ = 1 ± 0.8 μM. (D) VX‐809‐corrected K464A/K1250R‐CFTR: EC₅₀ 0.33 ± 0.2 μM. Each concentration–response curve is constructed from four experiments. For the n value of each concentration‐point measurement, see Supporting Information Data S6.

A concentration of 300 nM forskolin (between EC₁₀ and EC₄₀ for all genotypes) was chosen to provide a low level of phosphorylation, low enough for fully potentiated conditions to remain within the dynamic range of our assay. For K464A/K1250R‐CFTR (for which 300 nM forskolin is ~EC₄₀ on the G_CFTR‐[forskolin] curve; Figure 5D), a lower forskolin concentration of 100 nM (~EC₂₀) was also used.

YFP‐CFTR quenching was therefore measured at increasing concentrations of VX‐770 at these low phosphorylation levels, and fitting yielded data points to construct [VX‐770]‐G_CFTR (Figure 6) curves (see also [VX‐770]‐V_M curves, Supporting Information Data S5). VX‐770 strongly potentiates WT‐CFTR and the two non‐hydrolytic mutants (Figure 6A–B, Supporting Information Data S5A–B and Figure 7). Cells expressing K464A‐CFTR were treated for 24 h with 10 μM VX‐809, to increase membrane exposure (see Figure 4B). Although maximal quenching rates (and therefore estimated maximal G_CFTR, Figure 6C) are lower compared to WT‐CFTR, K464A G_CFTR is also low in basal conditions, such that VX‐770 still induces large potentiation (Figures 6C and 7, see also Supporting Information Data S5C). Sensitivity to VX‐770 was not altered by K1250R, D1370N or K464A mutations (P = 0.4).

Figure 6.

VX‐770 concentration–response curves. CFTR conductance estimated using model. All in the presence of 300 nM forskolin, except (D). (A) WT‐CFTR: EC₅₀ = 343 ± 276 nM. (B) D1370N‐CFTR: EC₅₀ = 263 ± 128 nM. K1250R‐CFTR: EC₅₀ = 330 ± 181 nM. (C) VX‐809‐corrected K464A‐CFTR: EC₅₀ = 241 ± 179 nM. (D) VX‐809‐corrected K464A/K1250R‐CFTR in the presence of 300 nM forskolin (fsk): EC₅₀ 145 ± 94 nM, or 100 nM forskolin: EC₅₀ 510 ± 329 nM. Each concentration–response curve is constructed from six experiments (or five in the case of WT‐CFTR and D1370N‐CFTR). For the n value of each concentration, see Supporting Information Data S6.

Figure 7.

Summary of potentiation by VX‐770. All constructs were potentiated by VX‐770 in the presence of 300 nM forskolin. K464A/K1250R‐CFTR was tested in both 100 and 300 nM forskolin (fsk). Bars represent mean estimated CFTR conductance in the presence or absence of saturating VX‐770. n = 10 for forskolin in the absence of VX‐770 (n = 8 for WT‐CFTR and n = 9 for D1370N‐CFTR), and n = 6 for forskolin in the presence of VX‐770 (n = 4 for WT‐CFTR and n = 5 for D1370N‐CFTR).

To eliminate effects due to the interaction of VX‐770 with the rarely encountered post‐hydrolytic O₂ conformation of K464A‐CFTR, the K1250R mutation was also introduced in a K464A‐CFTR background. For both 100 and 300 nM forskolin concentrations (see above and Figure 5D), changes in K464A/K1250R‐CFTR activity due to VX‐770 action are comparable to those seen for K464A‐CFTR (Figure 6C–D, Supporting Information Data S5C–D and Figure 7).

Discussion

VX‐770 has been shown to potentiate a very large number of mutant versions of CFTR (Yu et al., 2012; Van Goor et al., 2014), and evidence is accumulating on the clinical benefits provided to patients (Accurso et al., 2010; Bessonova et al., 2018). However, the effects of many of these mutations on CFTR gating have not been characterized in detail, and even less is known about how VX‐770 corrects their gating defect. In this study, we used mutants that have been studied extensively with electrophysiological techniques, as a tool to investigate VX‐770 mechanism. Such mutants, showing relatively minor alterations of P_o (less than threefold change), are thought to visit open channel states structurally equivalent to those visited by WT‐CFTR, but to progress from pre‐hydrolytic O₁ to post‐hydrolytic O₂ states rarely, or not at all. Quantitatively comparing the magnitude of VX‐770 potentiation of WT and of a number of such mutants, while minimizing uncontrolled variables, allows us to make general inferences on the mechanism of action of VX‐770 on CFTR, likely to be valid for many gating mutants. A better mechanistic understanding of how VX‐770 affects CFTR function could guide the development of improved genotype‐specific CFTR‐targeting drugs in the future.

VX‐770 potentiation involves stabilization of the pre‐hydrolytic open state O₁

To investigate VX‐770's mechanism of action, we used two non‐hydrolytic CFTR mutants: D1370N‐CFTR and K1250R‐CFTR. These mutations are likely to impair catalytic activity at site 2, and electrophysiological evidence is consistent with channels toggling, at saturating [ATP], between ATP‐bound closed, C, and pre‐hydrolytic O₁ states, without ever entering the O₂ state (Gunderson and Kopito, 1995; Vergani et al., 2003, 2005; Csanády et al., 2010). Due to the similar membrane exposure (Figure 4) and single‐channel conductance of WT‐CFTR and non‐hydrolytic CFTR mutants, similar G_CFTR values reflect similar P_o values. Our G_CFTR measurements (Figures 3B–C, 5A–B) are largely consistent with what is known for these mutants obtained from patch‐clamp experiments (Vergani et al., 2003, 2005; Bompadre et al., 2005; Csanády et al., 2010).

Because in non‐hydrolytic mutants opening and closing are forward and backwards transitions along the same kinetic pathway (C₁↔O₁), the only mechanism by which VX‐770 can increase P_o in these mutants is via stabilization of the O₁ state with respect to C. Merely stabilizing the opening transition state would cause an increase in opening rate with a concurrent increase in closing rate and therefore result in no net change in P_o.

Our finding that WT‐CFTR and non‐hydrolytic mutants were similarly potentiated by VX‐770 (~10‐fold, Figure 7) suggests that in WT‐CFTR too VX‐770 potentiates gating mainly by slowing the closing rate. This is most simply explained by a selective stabilization of the O₁ state by VX‐770. Whereas O₁ stabilization slows closure in non‐hydrolytic mutants by slowing step O₁→C₁ (rate k₋₁; Figure 1), in WT‐CFTR it does so by slowing step O₁→O₂ (rate k₁; Figure 1). In the absence of substantial effects on the stabilities of the transition states for these two steps, selective stabilization of the O₁ ground state will similarly slow closure of WT and non‐hydrolytic mutants, while leaving opening rate unaffected (consistent with Jih and Hwang, 2013 at millimolar [ATP]). However, more complex scenarios are also possible.

The similarity of the effects of VX‐770 on the CFTR versions tested here strongly suggests that stabilization of the post‐hydrolytic O₂ state (Jih and Hwang, 2013), which are likely never to be visited by non‐hydrolytic mutants gating in millimolar ATP, and visited rarely by K464A‐CFTR (Csanády et al., 2010), is not fundamental for VX‐770 action. Burst duration distributions of D1370N‐CFTR mutants show a monotonic decay, with no evidence of a negative fractional amplitude exponential component, consistent with virtually all bursts closing non‐hydrolytically. In contrast, burst duration distributions of K464A mutants include a significant negative exponential component, a distinctive paucity of events accumulating in the briefest burst duration bins, reflecting the rare bursts terminating through hydrolysis (Csanády et al., 2010). Biochemical studies measured a residual ATPase activity in K464A‐CFTR of approximately 10% of that of WT‐CFTR (Ramjeesingh et al., 1999), suggesting that any minute residual ATPase activity in D1370N‐CFTR is well below this level. To result in equal VX‐770 potentiation on all CFTR versions tested here, an action of VX‐770 on O₂ would require the magnitude of stabilization of these post‐hydrolytic events to be roughly inversely related to their differing frequency, resulting in a similar overall potentiation of WT and mutants. This interpretation seems much less likely than a simple, similar action on O₁.

VX‐770 potentiation and site 1 conformational dynamics

It should be noted that cells expressing K464A‐CFTR and K464A/K1250R‐CFTR were treated with VX‐809, to increase membrane expression and allow functional studies. It cannot be ruled out that VX‐809 altered the response to VX‐770 in these CFTR variants. However, the similarity of the effects of VX‐770 on these and on other (non VX‐809‐treated) variants is consistent with there being little drug–drug interaction for the acute responses measured here.

Recent results suggest that conformational changes around the ATP bound at site 1, occurring relatively late along the reaction coordinate for channel opening, contribute substantially to the stabilization of the pre‐hydrolytic open state O₁, preventing the reversal of the opening transition and therefore assuring that WT‐CFTR gating is coupled to the hydrolytic cycle at the NBDs (Sorum et al., 2015). The K464A mutation, removing a lysine side chain crucial for interactions with ATP at site 1, itself destabilizes the pre‐hydrolytic open state O₁, with respect to the transition state for opening (Powe Jr et al., 2002; Vergani et al., 2003; Bompadre et al., 2005; Csanády et al., 2010), most likely by interfering with these conformational changes at site 1. The fact that VX‐770 potentiation is not greatly affected by the K464A mutation (~10‐fold potentiation, as seen in WT‐CFTR, Figure 7) suggests that the effects of the mutation and of the drug are additive, that is that VX‐770 binding does not stabilize O₁ by changing the protein dynamics at site 1. Other conformational changes occurring during the opening of the channel, while the protein relaxes from the strained transition state to the open ground state (O₁) must be altered by binding of VX‐770 to the protein. The extremely hydrophobic physicochemical properties of the drug argue for drug binding in the transmembrane domains, possibly on areas exposed to the lipid bilayer (Jih and Hwang, 2013). More studies are required to further characterize the molecular determinants of VX‐770's action.

VX‐770 substantially increased the opening rate of F508del‐CFTR (Kopeikin et al., 2014). Like K464A, the F508del mutation also destabilizes the pre‐hydrolytic O₁ state (Jih et al., 2011). However, gating of F508del‐CFTR is likely to be largely hydrolytic, as demonstrated by the strong potentiation by 5‐nitro‐2‐(3‐phenylpropylamino)benzoate (NPPB) (Csanády and Töröcsik, 2014 but also see Lin et al., 2016) and by almost 100‐fold slowing of F508del‐CFTR closing rate by the E1371S mutation (Kopeikin et al., 2014). The apparent discrepancy between the effect of VX‐770 on WT‐CFTR and the mutants investigated in this paper and its effects on F508del‐CFTR might be related to the very severe defect in opening measured in the latter mutant (Miki et al., 2010; Cai et al., 2015). While the opening transition of WT, K1250R, D1370N and K464A mutants (for which measured opening rates at saturating ATP concentrations vary less than threefold: Vergani et al., 2003; Vergani et al., 2005; Csanády et al., 2010) is likely to occur via similar structural conformational changes, the energetic landscape visited during opening by F508del‐CFTR channels might be quite different, with bound VX‐770 providing a substantially facilitated pathway.

Mutations do not affect VX‐770 apparent affinity for CFTR

None of the mutations studied here caused any significant shift in [VX‐770]‐G_CFTR concentration–response curves. Because mutation effects on P_o are minor, this result suggests that the mutations do not differentially affect VX‐770‐bound and unbound channels.

Conclusions

In conclusion, VX‐770 is an effective potentiator used to treat CF. As well as potentiating WT‐CFTR, which closes via a hydrolytic pathway, VX‐770 potentiates mutants which gate at equilibrium (Van Goor et al., 2009; Eckford et al., 2012; Yeh et al., 2015). Our data, demonstrating similar potentiation of a number of gating mutants in which hydrolysis is impaired to varying degrees, are consistent with VX‐770 potentiating CFTR by stabilizing the pre‐hydrolytic O₁ state.

As structural information on the closed (Zhang and Chen, 2016; Liu et al., 2017; Zhang et al., 2017) and open CFTR channel becomes available, a better understanding of the conformational changes altered by VX‐770 and other drugs will be important for the development of improved, and genotype‐specific, treatment for CF patients. Our assay – with which researchers can rapidly obtain quantitative information on how drugs alter CFTR ion channel function for many CFTR variants, in near‐native cellular conditions – provides a robust and adaptable tool for further investigations.

Author contributions

Experiments were conceived and designed by E.L. and P.V. E.L. carried out the molecular biology and ran the fluorescence assay acquisition and image analysis. S.P. implemented the mathematical model in the MATLAB environment. Manuscript was written by E.L. and P.V. All authors read and commented on the final draft of the manuscript.

Conflict of interest

The authors declare no conflicts of interest

Declaration of transparency and scientific rigour

This http://onlinelibrary.wiley.com/doi/10.1111/bph.13405/abstract acknowledges that this paper adheres to the principles for transparent reporting and scientific rigour of preclinical research recommended by funding agencies, publishers and other organisations engaged with supporting research.

Supporting information

Data S1 Simple mathematical model allows estimation of G_CFTR and V_M parameters from quenching curves.

Data S2 Transient anion conductance is not CFTR‐dependent.

Data S3 Non‐stationary assay.

Data S4 Forskolin concentration‐response curves, quantified with V_M estimated using model.

Data S5 VX‐770 concentration‐response curves, quantified with V_M estimated using model.

Data S6 n values for individual points on time course activation curves and concentration‐response curves.

Langron, E. , Prins, S. , and Vergani, P. (2018) Potentiation of the cystic fibrosis transmembrane conductance regulator by VX‐770 involves stabilization of the pre‐hydrolytic, O₁ state. British Journal of Pharmacology, 175: 3990–4002. 10.1111/bph.14475.