Molecular structures reveal synergistic rescue of Δ508 CFTR by Trikafta modulators

1Cystic fibrosis (CF) is a common genetic disease caused by mutations in the gene that encodes the cystic fibrosis transmembrane conductance regulator (CFTR). CFTR is widely expressed in epithelial cells, regulating salt and fluid homeostasis in a variety of tissues. The absence or dysfunction of CFTR causes health issues including malnutrition, liver disease, recurrent bacterial infection, chronic inflammation, and respiratory failure.

2CFTR belongs to the ATP-binding cassette (ABC) transporter family but functions as an ATP-gated anion channel. It contains an N-terminal interfacial structure called the lasso motif, two transmembrane domains (TMDs) that form an anion conduction pathway, two cytoplasmic nucleotide-binding domains (NBDs) that bind and hydrolyze ATP, and a unique regulatory (R) domain that must be phosphorylated to open the channel. Although more than 300 mutations cause CF, approximately 90% of patients carry at least one copy of Δ508 CFTR, in which a single phenylalanine at position 508 is deleted. This Δ508 mutant exhibits a severe trafficking defect that results in intracellular retention and premature degradation of the channel. Furthermore, the few channels that reach the plasma membrane are unstable and functionally compromised.

3The structure of wild type (WT) CFTR shows that F508 resides on the surface of NBD1, where it makes extensive interactions with the cytosolic region of TM helix 11 and intracellular loop 4. These interactions are critical for both CFTR folding and coupling of ATP-dependent NBD dimerization to pore opening, suggesting that disruption of these interactions may underlie both trafficking and functional defects in Δ508 CFTR. Indeed, a previous crystal structure of an isolated NBD1 containing Δ508 revealed a conformation nearly identical to WT NBD1, supporting the hypothesis that Δ508 primarily affects interdomain assembly. Other studies have shown that lack of F508 causes thermodynamic instability of NBD1, as well as the entire protein. Unfortunately, the intrinsic instability of Δ508 CFTR has hindered efforts to structurally characterize the mutant in the context of the entire CFTR protein.

4Despite these structural obstacles, recently developed CFTR modulators have transformed CF therapy from symptom management to disease correction. These modulators include potentiators that enhance the function of CFTR in the plasma membrane and correctors that increase the presence of CFTR at the cell surface. CFTR correctors are further categorized into three different classes based on their functional redundancy. Currently, four pharmacological molecules are used in CF therapy, either singly or in combination. These include the potentiator ivacaftor (VX-770), the type I correctors lumacaftor (VX-809) and tezacaftor (VX-661), and the type III corrector elexacaftor (VX-445) (Fig. 1A). The most advanced therapy Trikafta (branded as Kaftrio in the EU) is a combination of ivacaftor, tezacaftor, and elexacaftor. The molecular mechanisms of ivacaftor and type I correctors have been well studied. Elexacaftor, however, was discovered only recently and little is known about its mode of action. Interestingly, elexacaftor has been shown to have a dual function, improving CFTR folding as well as ion conductance, but it is unknown whether it potentiates and corrects via the same binding site.

5In this study, we determined cryo-electron microscopy (cyro-EM) structures of Δ508 CFTR and analyzed the molecular effects of lumacaftor (VX-809), tezacaftor (VX-661), and elexacaftor (VX-445), revealing how elexacaftor might potentiate the activity of Δ508 as well as stabilize its structure. We also solved the structure of Δ508 CFTR in the presence of the three modulators comprising the triple therapy, providing a molecular description of how they synergistically rectify Δ508 CFTR to a functional state.

6We sought to determine the molecular structure of Δ508 CFTR in order to investigate the mechanisms underlying its trafficking defect and gating deficiency. Like the WT CFTR, substituting the catalytic residue E1371 with a glutamine was necessary to stabilize an ATP-bound, NBD-dimerized conformation for cryo-EM study. To test whether E1371Q alters the folding of Δ508 CFTR, both with and without pharmacological correctors, we used a gel-shift assay to quantify the relative abundance of the fully glycosylated mature protein relative to the core-glycosylated immature form. In the absence of correctors, Δ508 and Δ508/E1371Q CFTR were predominantly in their immature form (core-glycosylated, lower molecular weight). The addition of correctors increased the abundance of mature forms (fully glycosylated, higher molecular weight) for both variants to a similar extent (Fig. 1B). Confocal microscopy confirmed that Δ508 and Δ508/E1371Q CFTR were retained in the endoplasmic reticulum (ER), and that correctors increased their presence at the plasma membrane (Fig. 1C and fig S1). These data demonstrate that similar to Δ508 CFTR, the double mutant Δ508/E1371Q exhibits folding defects that can be reverted by correctors.

7Next, we purified the ER-retained Δ508/E1371Q CFTR without any pharmacological correctors (fig. S2) and determined its structure in the absence and presence of phosphorylation and ATP (Fig. 1D). In both conditions, the flexibility of NBD1 became apparent in the initial 2D classification steps of data processing (fig. S3). After extensive 3D classification, the structure of the dephosphorylated, ATP-free form was determined at approximately 6 Å resolution (Fig. 1D and fig. S3), revealing an NBD-separated conformation similar to that of WT CFTR. Densities corresponding to the TM helices and NBD2 revealed well-defined secondary structural features. The density for NBD1 was visible but amorphous, with a size and shape consistent with that of NBD1, indicating that Δ508 NBD1 is folded but flexibly attached to the TM helices. This structural observation supports the hypothesis that Δ508 disrupts interdomain assembly. In contrast to WT CFTR, in which the structured R domain inserts into a cytosolic cleft, little density corresponding to the R domain was visible in Δ508 CFTR. As the R domain packs mainly along the surface of NBD1, it is possible that defects in NBD1 assembly also disrupt the correct positioning of the R domain.

8The lack of structural stability in Δ508 CFTR was further pronounced in the phosphorylated, ATP-bound conformation. Previous structures of WT, phosphorylated and ATP-bound CFTR carrying the same E1371Q mutation were determined to resolutions between 2.7 and 3.8 Å. However, the analogous cryo-EM analysis of Δ508 CFTR stalled at approximately 9 Å resolution due to the heterogenous nature of the particles (Fig. 1D and fig. S2). The overall structure is consistent with an NBD-dimerized conformation, but a notable difference is the absence of visible density corresponding to NBD1 (Fig. 1D and fig. S3). Based on these data, we suggest that the R domain disengages upon phosphorylation as it does in the WT protein, permitting the TMDs to come into close contact. However, in the absence of F508, NBD1 is too flexible to support a stable NBD dimer. Because NBD dimerization is coupled to channel gating in CFTR, the inability of the NBDs to dimerize in Δ508 explains its impaired channel function.

9To investigate the conformational changes that CFTR correctors induce, we performed cryo-EM analyses of phosphorylated, ATP-bound Δ508/E1371Q CFTR in four pharmacological conditions: in the presence of lumacaftor (VX-809); elexacaftor (VX-445); a combination of these two correctors; and the triple Trikafta therapy of ivacaftor (VX-770), tezacaftor (VX-661), and elexacaftor (VX-445) (Fig. 2 and fig. S2 and S3). The Δ508/E1371Q CFTR was expressed in the absence of correctors and solubilized from the ER membrane. Correctors were added during protein purification (fig. S2) to reveal their post-translational effects on the structure of Δ508 CFTR without any confounding effects on folding kinetics and other cellular processes involved in Δ508 biogenesis.

10The structure of Δ508 CFTR in the presence of lumacaftor (VX-809) was similar to that in its absence, indicating that post-translational addition of lumacaftor alone is insufficient to correct the structural defects of Δ508 (Fig. 2A). This observation is consistent with the understanding that type I correctors bind to and stabilize TMD1 at an early stage of CFTR biogenesis, preventing its premature degradation and increasing the overall probability of fully folded CFTR being formed.

11In contrast, the type III corrector elexacaftor (VX-445) stabilized NBD1, resulting in a cryo-EM reconstruction of 3.7 Å resolution with ordered NBDs (Fig. 2A, B). The TMDs of elexacaftor-bound Δ508 CFTR closely resemble those of full-length CFTR in the phosphorylated, ATP-bound conformation, but the NBDs are very different (Fig. 2B). Elexacaftor-bound Δ508 has a “cracked-open” NBD dimer in which the catalysis-incompetent (degenerate) site is solvent accessible, and ATP makes contacts exclusively with the NBD1 face of the composite site (Fig. 2B).

12The combination of elexacaftor (VX-445) and lumacaftor (VX-809) had an effect that was greater than the sum of each corrector alone, fully restoring Δ508 to an NBD-dimerized conformation with ATP fully bound to both the consensus and degenerate sites (Fig. 2A, C). Moreover, this structure is essentially identical to that obtained in the presence of Trikafta (Fig. 2A), indicating that ivacaftor (VX-770) does not induce further conformational changes. This is consistent with ivacaftor being a potentiator, not a corrector. Both structures closely resemble that of full-length CFTR, having an overall root-mean-square deviation (r.m.s.d) to the full-length protein of 0.5 Å over 1090 Cα positions (Fig. 2C, 2D). We therefore designated both elexacaftor/lumacaftor and Trikafta-bound Δ508 CFTR structures as having a “corrected” conformation, and selected Trikafta-bound Δ508 for further analysis.

13The corrected Δ508 structure in the presence of Trikafta differs from that of full-length CFTR in the region of the F508 deletion site (Fig. 3A–C). F508 is located in a loop on the surface of NBD1, projecting its aromatic side chain into a hydrophobic pocket at the end of TM helix 11 (Fig. 3B). Deletion of F508 shortens this loop, leaving a crevice at the NBD1/TMD interface. In addition, the helical subdomain of Δ508 NBD1 (residues 500-564) is shifted away from the interface by approximately 2 Å (Fig. 3B). The crevice at the NBD1/TMD interface is partially filled by R1070, whose side chain swings into contact with main chain atoms in NBD1 (Fig. 3C). We also observed a strong spherical density in the Δ508 cavity area, which may be an ion or water molecule.

14Previous work has shown that the V510D mutation can stabilize Δ508 CFTR, likely due to aspartic acid forming a salt bridge with R1070. The structure of Δ508 is compatible with a salt bridge between these residues and thus lends support to this hypothesis (fig. S4A). The structural differences at the NBD1/TMD interface also explain the opposing effects of the R1070W mutation in full-length versus Δ508 CFTR. In full-length CFTR, substituting R1070 with tryptophan inhibits protein folding and leads to CF. This is because the large tryptophan side chain at position 1070 would sterically clash with F508 (fig. S4B). In contrast, the same substitution in the Δ508 background partially restores CFTR folding, likely due to R1070W strengthening the NBD1/TMD interface by filling the space devoid of F508 and forming hydrophobic and hydrophilic contacts with Δ508 NBD1 (fig. S4C).

15In the cryo-EM reconstruction of Trikafta-bound Δ508 CFTR, the densities for ivacaftor (VX-770), tezacaftor (VX-661), and elexacaftor (VX-445) were strong and unambiguous (Fig. 3). Viewed from the extracellular space perpendicular to the membrane, these compounds form a triangular belt encircling the TMDs (Fig. 3A). The potentiator ivacaftor (Fig. 3, purple molecule) binds to a cleft formed by TM helices 4, 5, and 8, approximately halfway through the lipid bilayer, coincident with the TM8 hinge region involved in gating. The molecular details of ivacaftor binding are identical to those in full-length CFTR, indicating that ivacaftor potentiates both full-length and Δ508 CFTR via the same mechanism. Similarly, tezacaftor (Fig. 3, orange molecule) binds to Δ508 by inserting into a hydrophobic pocket in TMD1 (Fig. 3D) in a manner identical to that in the WT protein. As previously discussed, such a penetrating cavity would cause substantial destabilization of the protein in the absence of tezacaftor. Furthermore, TM helices 1, 2, 3, and 6 that form the binding site are predicted to be unstable in the membrane, thus type I correctors would stabilize TMD1 both by filling the cavity and structurally linking together the four unstable helices.

16In contrast, the binding site for elexacaftor (VX-445) has not previously been observed. Densities corresponding to elexacaftor are very similar and well-defined in the three elexacaftor-bound reconstructions (elexacaftor alone, elexacaftor with lumacaftor, and Trikafta), enabling an accurate description of elexacaftor’s binding pose and the chemistry of drug recognition (fig. S5A). Elexacaftor (Fig. 3, yellow molecule) binds to CFTR within the membrane, extending from the center of the lipid bilayer to the edge of the inner leaflet (Fig. 4A). The binding pocket is much shallower than that of type I correctors, as if the drug molecule is patched onto the surface of CFTR. Elexacaftor interacts most extensively with TM helix 11 through electrostatic and van der Waals interactions. It also forms contacts with TM helices 2, 10, and the lasso motif (Fig. 4A, fig. S5B).

17Studies of ivacaftor (VX-770) binding to CFTR have shown that hydrogen bonds (H-bonds) are critical for drug recognition as they stabilize polar groups in the low dielectric environment of the membrane. To test if this principle applies to elexacaftor (VX-445) binding, we substituted R1102 with an alanine to eliminate the formation of an H-bond and a salt bridge (Fig. 4A), and directly measured binding using the scintillation proximity assay (SPA) (Fig. 4B). Specific binding of elexacaftor to full-length CFTR increased as a function of drug concentration. In contrast, the interaction of elexacaftor with the R1102A mutant was barely detectable. We also evaluated the contribution of R1102 to the restoration of Δ508 folding by elexacaftor using the gel-shift assay (Fig. 4C). The R1102A mutation abolished correction by elexacaftor, but not tezacaftor, which binds to a pocket distant from R1102. Finally, to test if R1102 also contributes to the potentiation activity of elexacaftor, we measured macroscopic current in inside-out membrane patches containing phosphorylated CFTR (Fig. 4D). We consistently observed stronger potentiation by the S enantiomer in both WT and Δ508 CFTR, in agreement with the higher efficacy of the S enantiomer observed in prior work. However, R1102A CFTR did not respond to either R- or S- enantiomer, indicating that both enantiomers bind to the site identified in the structures. A control potentiator, GLPG1837, increased macroscopic current in all CFTR variants (Fig. 4D). Together, these data suggest that elexacaftor achieves both correction and potentiation via the same binding site.

18The structures of Δ508 CFTR reveal that the absence of F508 disrupts the ability of NBD1 to assemble with the TM helices, which leads to intracellular retention and degradation of the protein. Furthermore, the inability to form a stable NBD dimer following phosphorylation and ATP binding results in a dysfunctional channel, even if Δ508 CFTR reaches the plasma membrane. Pharmacological correctors used in clinic – even added during protein purification– can partially or fully restore the NBD-dimerized conformation. These correctors can improve the folding of WT CFTR as well as disease-causing mutations. Although it is theoretically possible that correctors salvage the different mutant and WT forms of CFTR by entirely different mechanisms, it is likely that their mechanism of action is the same in all cases. Indeed, lumacaftor (VX-809) binds to Δ508 CFTR and WT CFTR at the same site and by interacting with the same residues. Likewise, ivacaftor (VX-770) interacts with WT and Δ508 CFTR in an identical manner, indicating that its mechanism of potentiation is the same for both WT and mutant CFTR.

19The dual-function modulator elexacaftor (VX-445) engages TM helices 10-11 and the lasso motif, all of which are important for CFTR folding and function. Mutations in the lasso motif can cause intracellular retention or abnormal gating, and some lead to CF. TM10-11 are the “domain swapped” helices of TMD2 that extend into the cytosol and interact with NBD1. This interface is not only important for protein assembly, but also critical for transmitting conformational changes of the NBDs to the TMDs to control ion permeation. Although the details of how elexacaftor potentiates CFTR await incisive electrophysiology measurements, our structural observations indicate that elexacaftor stabilizes TM 10-11, thereby strengthens the TMD/NBD1 interface, which is particularly vulnerable to disease-causing mutations.

20Recently Braakman and colleagues demonstrated that CFTR folding occurs in a stepwise manner. It is likely that the type I corrector tezacaftor binds at an early stage of CFTR biogenesis to stabilize the N-terminal TMD1 and the type III corrector elexacaftor binds at a later stage when TMDs assemble to form a protease-resistant form. Together, these two correctors prevent premature degradation in the ER. Once CFTR reaches the plasma membrane, the presence of elexacaftor strengthens allosteric communication between ATP-bound NBD dimers and the channel gate, thereby increasing ion conductance. Channel activity is further enhanced by ivacaftor, which stabilize the open configuration of the pore. In this manner, the three modulators in the triple therapy act synergistically to improve the folding and function of CFTR.