Structural basis for CFTR inhibition by CFTRinh-172

Significance Cystic fibrosis transmembrane conductance regulator (CFTR) inhibitors, including CFTRinh-172, have been developed as therapeutic candidates to treat secretory diarrhea and autosomal dominant polycystic kidney disease. They are also widely used in laboratories to investigate the mechanisms underlying CFTR gating. This study offers a structural understanding of CFTRinh-172's mode of action, elucidating its ability to obstruct ion conduction and modulate channel gating. The molecular description of how CFTRinh-172 interacts with CFTR provides a structural foundation for its potency and efficacy. The observation that CFTR inhibitors and potentiators both interact with transmembrane helix 8 strengthens the notion that this helix serves as an allosteric link between the catalytic site and the channel gate and is therefore a hotspot for pharmacological modulation.

Despite functioning as an ion channel, CFTR belongs to the superfamily of ATP-binding cassette (ABC) transporters.It is composed of two transmembrane domains (TMDs) and two nucleotide-binding domains (NBDs) that are common to all ABC transporters, along with a cytosolic regulatory (R) domain specific to CFTR (22,23).CFTR's activity is regulated at two levels.Phosphorylation by PKA releases the auto-inhibition imposed by the unphosphorylated R domain (24,25).Once phosphorylated, adenosine triphosphate (ATP) binding promotes NBD dimerization and pore opening, whereas ATP hydrolysis leads to pore closure (26).
Substantial effort has been devoted to understanding CFTR in the context of cystic fibrosis.Disease-causing mutations have been extensively characterized, and small-molecule drugs that potentiate gating or correct folding of CFTR have been successfully developed for clinical use (27)(28)(29)(30)(31)(32)(33)(34)(35)(36)(37)(38).By contrast, relatively few studies have addressed CFTR hyperactivation.Although secretory diarrhea is the second leading cause of death in children under 5 worldwide (20,39,40), and ADPKD is the most common inherited cause of end-stage renal disease (19), both conditions lack broadly effective, generalizable pharmacological treatments.Despite the therapeutic potential of CFTR inhibitors and although several small-molecule CFTR inhibitors have been identified (5,6,41,42), their mechanisms and sites of action remain poorly understood.
In this study, we investigate the mechanism of CFTR inh -172, a highly efficacious CFTR inhibitor developed in the Verkman laboratory (5).This inhibitor was shown to block cholera toxin-induced intestinal fluid secretion (5) and suppress cyst growth in animal models of polycystic kidney disease (3).Substitution of pore-lining residues reduced the potency of CFTR inh -172, suggesting that it directly binds to CFTR (43).However, several studies have also suggested that CFTR inh -172 acts as a gating modulator rather than a classical pore-blocker (44,45).Using cryogenic electron microscopy (cryo-EM), singlemolecule fluorescence resonance energy transfer (smFRET), and electrophysiology, we have found that CFTR inh -172 binds within the pore, stabilizing the transmembrane helices in a nonconductive conformation without obstructing NBD dimerization.These findings enable us to propose a mechanism for CFTR inh -172 that reconciles previous observations.

Results
CFTR inh -172 Inhibits Wild-Type (WT) CFTR and the "Locked-Open" CFTR (E1371Q).Previously, the hydrolysis-deficient CFTR (E1371Q) variant has been used to obtain high-resolution structures of CFTR in complex with modulators, leveraging its ability to stabilize the canonical NBD dimer (25,36,37,46,47).To test whether the same variant is suitable to study CFTR inh -172, we characterized the effects of CFTR inh -172 on both WT CFTR and CFTR (E1371Q) (Fig. 1).In excised inside-out membrane patches (Fig. 1 A and B), application of 10 µM CFTR inh -172 reduced the macroscopic current of WT CFTR by 96%.A comparable level of inhibition was observed with CFTR (E1371Q), despite its open dwell time being 1,000-fold longer than that of WT CFTR (48).Our data are consistent with previous work on WT CFTR and two other hydrolysis-deficient variants, D1370N and E1371S (45).
Next, we purified CFTR and reconstituted it into a synthetic planar lipid bilayer.Upon activation by PKA, single-channel currents were measured in the presence of 3 mM ATP with or without 10 µM CFTR inh -172 (Fig. 1C).CFTR inh -172 reduced the open probability of WT CFTR from 0.21 ± 0.05 (mean and SE) to 0.007 ± 0.003, whereas that of CFTR (E1371Q) decreased from 0.79 ± 0.03 to 0.0011 ± 0.0004 (Fig. 1D).Previous studies had reported conflicting evidence regarding the effect of CFTR inh -172 on open dwell time (44,45).Here, we observed both a large effect on closed dwell time and a fivefold reduction in mean open dwell time, from 487 ± 92 ms to 109 ± 14 ms for WT CFTR (Fig. 1E), consistent with findings from Kopeikin and colleagues (45).
Taken together, both macroscopic and single-channel current measurements indicate that CFTR inh -172 inhibits CFTR (E1371Q) as effectively as it does WT CFTR, suggesting a comparable mechanism of inhibition.
CFTR inh -172 Binds within the Pore.To identify the binding site of CFTR inh -172, we determined the cryo-EM structure of inhibitor-bound CFTR using phosphorylated CFTR (E1371Q) (Fig. 2 and SI Appendix, Fig. S1).The structure was determined to an overall resolution of 2.7 Å.In the presence of CFTR inh -172 and ATP, CFTR (E1371Q) adopts an NBD-dimerized, poreclosed conformation distinct from any previously observed structures (Fig. 2A).The density for CFTR inh -172, as strong as the protein main chain atoms, was observed inside the pore, at a position corresponding to the membrane outer leaflet (Fig. 2A).The chemical structure of CFTR inh -172 can be divided into three rings (Fig. 2B): a central thiazolidine ring (ring B) with a (3-trifluoromethyl)phenyl substitution at position 3 (ring A) and a (4-carboxyphenyl)methylene substitution at position 5 (ring C).The density shows well-defined features corresponding to the trifluoromethyl phenyl and the heavy sulfur atoms in the thiazolidine ring (Fig. 2B).The density for ring C is not as welldefined, indicating that this moiety may be mobile.
CFTR inh -172 binds within the CFTR pore and interacts with both TMDs (Fig. 2A).The three rings of the compound form an elongated shape wedged between transmembrane helices (TMs) 1, 6, 8, 9, and 12. Whereas ring A and ring B are completely buried, ring C is exposed to the aqueous ion-conduction pathway (Fig. 2A).With this structure, we can now better understand previous structure-activity-relationship (SAR) analyses (5,49).Specifically, these studies showed that the trifluoromethyl group (CF 3 ) is essential for inhibition and that the addition of polar substituents or removal of CF 3 on ring A diminished activity (49).The highest potency is achieved when CF 3 is at position 3 of ring A, as in CFTR inh -172 (5,49).In the structure, the trifluoromethyl group fits snugly into a hydrophobic cavity, establishing van der Waals interactions with five hydrophobic residues on TMs 6, 8, 9, and 12 (Fig. 2 C and D).If the CF 3 substitution were at positions 2 or 4, many of these interactions would be lost, resulting in lower inhibitory potency, as observed (49).The SAR studies further demonstrate the significance of negative or polar substitutions on ring C (49).Indeed, we observe that this ring is positioned within a spacious, solvent-exposed cavity surrounded by numerous charged and polar residues, including K95 on TM1 and N1138, S1141, and T1142 on TM12 (Fig. 2 C and D).The carboxy group on CFTR inh -172 forms a salt bridge with K95 (Fig. 2 C and D).Consistent with this observation, esterification or amidation of the carboxy group in CFTR inh -172 resulted in inactive compounds (49), presumably due to the loss of this interaction.The reciprocal change on CFTR, substitution of K95 with alanine, resulted in a nearly sevenfold decrease in potency, increasing the half maximal inhibitory concentration (IC 50 ) of CFTR inh -172 from 0.6 ± 0.1 µM to 3.5 ± 0.9 µM (Fig. 2E).Perturbation of the binding site by T1142I substitution increased the IC 50 to 5.3 ± 3.1 µM (Fig. 2E).The effect of the T1142I substitution is most likely due to steric hindrance, as the side chain of T1142 is close to the methylene group at position 2 of ring B (Fig. 2 C and D).The reciprocal modification on the inhibitor, adding a methyl at this position, was shown to increase the IC 50 to 8 µM (49).By contrast, alanine substitution of a non-interacting residue, S1141, had no effect on inhibitory potency (Fig. 2E and SI Appendix, Fig. S2).
The structure also offers a molecular explanation for previous data showing that substitution of R347 with alanine decreases the potency of CFTR inh -172 by over 30-fold and that R347D substitution nearly eliminated its inhibitory effect (43).Although R347 does not interact with the inhibitor directly, it forms a salt bridge with D924, creating a surface against which the inhibitor is tightly packed (Fig. 2C).Substitutions at position 347 are likely to modify the structure of the binding site, consequently reducing the inhibitory activity.
CFTR inh -172 Stabilizes a Closed Conformation of CFTR.Previous structural studies of human CFTR have revealed the conformational changes required for pore opening (Fig. 3A).Dephosphorylated CFTR exhibits an NBD-separated conformation with an inner vestibule open to the cytosol but the pore closed off to the extracellular space (Fig. 3A) (50).In the presence of ATP, phosphorylated CFTR (E1371Q) forms an NBD-dimerized conformation (49), in which the pore is open and a dehydrated chloride ion is bound at the selectivity filter near the extracellular entrance (51) (see accompanying paper) (Fig. 3A).A comparison of these two structures reveals that phosphorylation and ATP binding cause the NBDs and TMDs to move toward the central axis essentially as rigid bodies.However, local conformational changes of TMs 8 and 12 are also critical for CFTR gating (25).
In the presence of CFTR inh -172, CFTR adopts a conformation distinct from either structure (Fig. 3A).The NBDs form a dimer similar to that observed in the uninhibited structures of CFTR (E1371Q) (25,46).The TMDs undergo global rigid-body movements toward each other, but TMs 1, 8, and 12 are positioned differently (Fig. 3B).Local structural superposition shows that the extracellular segment of TM8 is stabilized in a conformation intermediate between the NBD-separated anddimerized conformations (Fig. 3C).Furthermore, the anion selectivity filter (51) (see accompanying paper) collapses, as TM1 undergoes a ~5° rotation that places the side chain of I106 at the position occupied by the chloride ion in the uninhibited structure (Fig. 3 D and E).This repositioning of the TMs also leads to a complete closure of the lateral exit between TMs 1 and 6 that connects the selectivity filter to the extracellular space (Fig. 3E).
CFTR inh -172 Allosterically Inhibits ATP Turnover.Our structural analysis clearly shows that binding of CFTR inh -172 leads to pore closure by trapping the TM helices in a non-conductive state (Fig. 3).As CFTR gating is coupled to ATP hydrolysis, we tested whether CFTR inh -172 changes the ATP turnover rate (Fig. 4A).We found that the presence of 10 µM CFTR inh -172 decreased saturating ATP turnover (k cat ) by approximately fourfold, from 22.0 ± 2.2 to 5.2 ± 1.0 ATP/protein/min (Fig. 4A).The Michaelis-Menten constant (K m ) for ATP was not significantly changed (Fig. 4A).
To assess the mechanism by which ATP hydrolysis is inhibited by CFTR inh -172, we used a recently established smFRET assay, which reports on the conformational state of CFTR's NBDs (38).In this assay, position 388 in NBD1 and position 1435 in NBD2 were labeled with donor and acceptor fluorophores (Fig. 4B).Conformational isomerizations of individual CFTR molecules were monitored as transitions between a low FRET efficiency (0.25 ± 0.01) NBD-separated state and a high FRET efficiency (0.49 ± 0.02) NBD-dimerized state.As we have previously reported (38), at a saturating (3 mM) ATP concentration, WT CFTR predominantly occupies NBD-dimerized conformations with brief excursions to the NBD-separated state (Fig. 4C).The presence of 10 µM CFTR inh -172 did not significantly affect the probability of NBD dimerization or the dwell time of the NBD-dimerized state for WT CFTR (Fig. 4 D and E).CFTR inh -172 also did not affect the conformational dynamics of CFTR (E1371Q), which remained predominantly NBD-dimerized (Fig. 4 C-E).These data indicate that CFTR inh -172 does not prevent NBD dimerization but rather slows progression through the gating cycle while the NBDs are dimerized.
These observations lead us to consider a possible mechanism for inhibition of ATP hydrolysis.Our recent study showed that conformational changes within NBD-dimerized CFTR governed by ATP turnover are required for chloride conductance (38).Potentiators Ivacaftor and GLPG1837 enhance channel activity by increasing pore opening while the NBDs are dimerized.Additionally, the potentiators increase ATP turnover (38).In comparing the structure of CFTR (E1371Q) bound to GLPG1837 with that bound to CFTR inh -172, we observe that the CFTR inh -172 site is located along the pore-lining side of the TM8 hinge region, in direct juxtaposition to the potentiator binding site (Fig. 4F).As TM8 links ATP hydrolysis and pore opening (36,52), we propose that CFTR inh -172 inhibits ATP turnover via an allosteric mechanism involving TM8, similar in nature but opposite in effect to that of the potentiators.However, the exact step of the ATP hydrolysis cycle perturbed by CFTR inh -172 remains unclear.It is possible that CFTR inh -172 reduces the hydrolysis rate by slowing post-hydrolytic nucleotide exchange or promoting occupancy of a state that is not transited by WT CFTR during physiological gating.

Discussion
Typically, ion channel inhibitors are classified into two categories: pore blockers that bind within the ion conduction pathway to occlude the pore and gating inhibitors that impair channel opening by stabilizing the closed state (53).However, CFTR inh -172 presents a perplexing case, as it has been shown to interact with residues within the pore while also impairing gating (43,44).Through cryo-EM and smFRET analyses, we now have a structural understanding of CFTR inh -172 inhibition that reconciles earlier findings.The binding site of CFTR inh -172 is located within the pore, nestled in a cavity lined by R347, a residue whose substitution significantly reduces the potency of CFTR inh -172 (43).Local conformational changes of TMs 1, 8, and 12 in the CFTR inh -172-bound structure cause a collapse of the chloride selectivity filter and the extracellular exit.Based on kinetic analysis, Hwang and colleagues had predicted that CFTR inh -172 induces conformational change in CFTR (45).The nature of this change is now revealed at a molecular level.Additionally, we found that CFTR inh -172 inhibits ATP hydrolysis through an allosteric mechanism that we hypothesize is similar to that of potentiators Ivacaftor and GLPG1837 but with an opposite functional effect.These observations corroborate the hypothesis that conformational shifts in TM8 link ATP hydrolysis at the NBDs with the state of the pore.Electrophysiological measurements in our laboratory (Fig. 1) and in other studies (45) demonstrate that CFTR inh -172 inhibits WT CFTR and several hydrolysis-deficient variants to similar extents.Although it is theoretically possible that this molecule inhibits WT CFTR and each variant through different mechanisms, it is more likely that the mechanism of action is similar.Indeed, the structure determined with CFTR (E1371Q) is entirely consistent with earlier SAR studies (5,49).Substitutions at the structurally identified binding site made in the WT CFTR background reduced the potency of CFTR inh -172 (43) (Fig. 2E).Additionally, smFRET studies reveal that CFTR inh -172 does not affect NBD isomerization in WT CFTR or CFTR (E1371Q) (Fig. 4 C and D).These data strongly suggest that the mode of action revealed in this study represents a general mechanism for CFTR inh -172.However, as E1371Q substitution stabilizes the NBDs in a canonical dimerized conformation, it is possible that CFTR inh -172 induces local changes at the ATPase site that are obscured in our structure.Further studies will be pursued to identify structural re-arrangements of the WT channel within the NBD-dimerized state, as these play a key role in coupling ATP hydrolysis to channel gating.
Finally, the congruence between structural and functional data not only offers intellectual satisfaction but also opens up new avenues for enhancing the potency and specificity of CFTR inh -172.Specifically, CFTR interacts with ring C of CFTR inh -172 primarily through the K95 salt-bridge and an edge-to-face πstacking interaction with W1145.It is possible that analogs of CFTR inh -172, with modifications on ring C that establish additional interactions with nearby polar residues on CFTR will have enhanced potency and specificity.
Mutagenesis.CFTR variants were generated using the SPRINP mutagenesis method (SI Appendix, Table S2) (54).Briefly, mutagenic primers were designed to be complementary to the template plasmid except for the mutated bases and to be 15 to 45 nucleotides in length.Plasmid containing CFTR cDNA was amplified in separate reactions containing forward or reverse primer.The singleprimer products of these reactions were combined and denatured at 95 °C for 5 min and gradually cooled to 37 °C over the next 5 min.The sample was then digested by DpnI for 4 h.Then, 5 µL of sample was added to 50 µL of competent XL2Blue cells for transformation and incubated on ice for 30 min.The bacteria were then heat-shocked at 42 °C for 45 s and allowed to recover on ice for 2 min.Following that, 200 µL of warmed SOC (Invitrogen) was then added directly to the cells, and the mixture was allowed to shake at 225 RPM in a 37 °C incubator for 30 min.Then, 200 µL of this mixture was spread on LB/ampicillin plates and left to incubate at 37 °C overnight.Random colonies were then picked and expanded in LB/ampicillin.Plasmid DNA was then purified (QIAGEN Plasmid Kit) and sequenced (Genewiz).
Recordings were carried out using the inside-out patch configuration with local perfusion at the patch.Recording pipettes were pulled from borosilicate glass (outer diameter 1.5 mm, inner diameter 0.86 mm, Sutter) to 1.5 to 3.0 MΩ resistance.Currents were recorded at 25 °C using an Axopatch 200B amplifier, a Digidata 1550 digitizer, and the pClamp software suite (Molecular Devices).Membrane potential was clamped at −30 mV.Current traces reflect inward currents with inverted signatures.Recordings were low-pass filtered at 1 kHz and digitized at 20 kHz.
For all measurements, CFTR was activated by exposure to PKA (Sigma-Aldrich) and 3 mM ATP. Displayed recordings were low-pass filtered at 100 Hz.Data were analyzed using Clampfit, GraphPad Prism, and OriginPro.
Protein Expression and Purification.CFTR constructs were expressed and purified as previously described (55,56).Bacmids encoding human CFTR fused to a C-terminal PreScission Protease-cleavable GFP tag were generated in Escherichia coli DH10Bac cells (Invitrogen).Recombinant baculovirus was produced and amplified in Sf9 cells.HEK293S GnTl − suspension cells, at a density of 2.0 to 3.0 × 106 cells/mL, were infected with 10% (v/v) P3 or P4 baculovirus.Protein expression was induced by addition of 10 mM (final concentration) sodium butyrate 12 h after infection.The cells were cultured at 30 °C for an additional 48 h and then harvested by centrifugation.
EM Data Acquisition and Processing.Immediately following size-exclusion chromatography, the CFTR (E1371Q) sample was concentrated to 5 mg/mL (32 µM) and incubated with 8 mM ATP, 10 mM MgCl 2 , and 100 µM CFTR inh -172 on ice for 30 min.Three mM fluorinated Fos-choline-8 was added to the samples directly before application onto glow-discharged Quantifoil R0.6/1 300 mesh Cu grids.Samples were then vitrified using a Vitrobot Mark IV (Field Electron and Ion Company, FEI).
Following initial round of 3-dimensional classification, the best class was refined and further classified into four classes and processed until no further improvement was observed (SI Appendix, Fig. S1A).Despite having slightly different nominal resolutions, the top three classes essentially represent the same conformation of CFTR (SI Appendix, Fig. S1B), with the density for the inhibitor best defined in the highest resolution map (compare SI Appendix, Fig. S1B with Fig. 2B).
Model Building and Refinement.Initial protein models were built by fitting the published structure of the NBD-dimerized CFTR (E1371Q) (PDB: 6MSM) into the cryo-EM map using Coot (60).The model was then adjusted based on the cryo-EM density.CFTR inh -172 was built into the density and refined in PHENIX (61) using restraints generated by the Global Phasing web server (grade.globalphasing.org).MolProbity (62) was used for geometry validation.
Synthetic planar lipid bilayers were made from a lipid mixture containing DOPE, POPC, and POPS at a 2:1:1 (w/w/w) ratio.Proteoliposomes containing PKA-phosphorylated CFTR were fused with the bilayers.Currents were recorded at 25 °C in a symmetric buffer containing 150 mM NaCl, 2 mM MgCl 2 , 20 mM HEPES (pH 7.2 with NaOH), and 3 mM ATP. Voltage was clamped at −150 mV with an Axopatch 200B amplifier (Molecular Devices).Currents were low-pass filtered at 1 kHz, digitized at 20 kHz with a Digidata 1440A digitizer and recorded using the pCLAMP software suite (Molecular devices).Recordings were further low-pass filtered at 100 Hz.Data were analyzed with Clampfit, GraphPad Prism, and OriginPro.A lower bound of 10 ms was put on event duration in idealization of channel recordings to exclude flicker-closures from the dwell time analysis.Reported open dwell times thus reflect open burst dwell times.
Imaging was carried out with a custom-built wide-field, prism-based total internal reflection fluorescence microscope.Donor (LD555) fluorophores were excited with an evanescent wave generated using a 532-nm laser (Opus, Laser Quantum).Fluorescence emitted from donor (LD555) and acceptor (LD655) fluorophores was collected with a 1.27 NA 60× water-immersion objective (Nikon), spectrally resolved using a T635lpxr dichroic (Chroma), and imaged onto two Fusion sCMOS cameras (Hamamatsu).The integration period of imaging was 100 ms.
Analysis of fluorescence data was performed using the SPARTAN analysis software in MATLAB (65).Single-molecule FRET trajectories were calculated as E FRET = I A /(I A + I D ), where I A and I D are the emitted acceptor and donor fluorescence intensities, respectively.The following pre-established criteria were applied to select FRET trajectories for analysis: Single-step donor photobleaching; a signalto-noise ratio above 8; fewer than 4 donor-blinking events; FRET efficiency below 0.8; and FRET efficiency above baseline for at least 50 frames.The segmental k-means algorithm (66) was used to idealize trajectories to a model containing two non-zero-FRET states with FRET efficiencies of 0.25 and 0.48.Data were further analyzed with OriginPro.Probabilities of high FRET occupancy are likely slightly underestimated as a small fraction of molecules were nonresponsive to ATP exposure.This population likely reflects denatured molecules.These molecules cannot be excluded in an unbiased manner based on their photophysical properties and are therefore included in our analysis.
Data, Materials, and Software Availability.The cryo-EM map of CFTR inh -172bound CFTR has been deposited in the Electron Microscopy Data Bank under the accession code EMD-42101 (67).The corresponding atomic model has been deposited in the Protein Data Bank under accession code 8UBR (68).All other data and information are available in the main text or SI Appendix.

Fig. 2 .
Fig. 2. CFTR inh -172 makes specific interactions with the CFTR inner vestibule.(A, Left) Structure of the CFTR (E1371Q)/CFTR inh -172 complex.(Right) A view of the structure looking down the long axis of the CFTR pore with CFTR inh -172 modeled as orange sticks surrounded by cryo-EM density.(B, Left) Structure of CFTR inh -172 with the three functional rings delineated by gray rectangles.(Right) Structure of CFTR inh -172 modeled into its binding site within the inner vestibule surrounded by cryo-EM density.(C) Interacting residues within 4.5 Å of CFTR inh -172.(Left) The hydrophobic pocket of the CFTR inh -172 binding site.(Right) The solvent-exposed pocket of the CFTR inh -172 binding site.CFTR inh -172/K95 and R347/D924 salt bridges are shown as black dashed lines.(D) Schematic drawing of CFTR-inhibitor interactions.All residues within 4.5 Å of the inhibitor are depicted.Residues substituted in inside-out path-clamp electrophysiology are indicated with colored circles.(E) Example macroscopic current traces showing titration of CFTR inh -172 onto WT, K95A, or T1142I CFTR in inside-out excised patches.CFTR was fully phosphorylated by PKA in the presence of 3 mM ATP before CFTR inh -172 titration.(Right) Dose-response curves for CFTR inh -172 binding site variants.The mean current in the presence of 3 mM ATP alone was used to normalize current elicited at each CFTR inh -172 concentration.Dose-response curves were fitted with the Hill equation to estimate IC 50 values for each variant.Hill coefficients were fixed to 1.Each data point represents mean and SE determined from three to five patches.