CFTR chloride channel mutations cause the lethal and incurable disease cystic fibrosis (CF). CFTR is activated by phosphorylation, and phosphorylated channels exhibit “bursting” behavior—“bursts” of openings separated by short “flickery” closures and flanked by long “interburst” closures—driven by ATP binding/hydrolysis at two nucleotide-binding domains. The human channel (hCFTR) and the distant zebrafish ortholog (zCFTR) display differences both in their gating properties and structures. In phosphorylated ATP-bound hCFTR, the hR117 side chain, conserved across evolution, forms an H-bond that stabilizes the open state. Lack of that bond in the hR117H mutant causes CF. In the phosphorylated ATP-bound zCFTR structure that H-bond is not observable. Here, we show that the zR118H mutation does not affect the function of zCFTR. Instead, we identify an H-bond between the zS109 and zN120 side chains of phosphorylated ATP-bound, but not of unphosphorylated apo-, zCFTR. We investigate the role of that interaction using thermodynamic mutant cycles built on gating parameters determined in inside-out patch clamp recordings. We find that zS109 indeed forms an H-bond with zN120 in the flickery closed state, but not in the open or interburst closed states. Although in hCFTR an isoleucine (hI119) replaces the asparagine, mutation hS108A produces a strong hR117H-like phenotype. Since the effects of the latter two mutations are not additive, we conclude that in hCFTR these two positions interact, and the hS108-hR117 and hR117-hE1124 H-bonds cooperate to stabilize the open state. These findings highlight an example of how the gating mechanism was optimized during CFTR molecular evolution.
Introduction
Cystic fibrosis (CF) is the most common inherited lethal disease in the Caucasian population. It is caused by loss-of-function mutations of the cystic fibrosis transmembrane conductance regulator (CFTR) anion channel, and results in the disruption of transepithelial salt water transport in the lungs, intestines, pancreatic duct, and sweat duct (O’Sullivan and Freedman, 2009). The several hundred identified CF-causing mutations are divided into roughly six classes based on their effects on expression (class I), processing and trafficking (class II), channel gating (class III), anion permeation (class IV), and the amount and stability of the protein in the plasma membrane (classes V–VI; De Boeck and Amaral, 2016). Although CF is currently incurable, recent development of small-molecule drugs that target the CFTR protein marked a milestone, complementing traditional symptomatic treatment. A significant symptomatic improvement is observed for CF patients with class III mutations when treated with the “potentiator” drug Vx-770 (ivacaftor), which overcomes gating defects (Ramsey et al., 2011), and for class II patients when treated with Trikafta, a combination of Vx-770 with two “corrector” drugs that facilitate protein folding (Heijerman et al., 2019; Sutharsan et al., 2022; Middleton et al., 2019; Barry et al., 2021). Since the effects of these small-molecule drugs on different mutants are greatly variable, understanding the underlying molecular background of each CF-causing mutation is of great clinical significance.
CFTR belongs to the superfamily of ATP-binding cassette (ABC) proteins and consists of two transmembrane domains (TMD1/2; Fig. 1, A–C, gray), two cytosolic nucleotide-binding domains (NBD1/2; Fig. 1, A–C, blue and green, respectively), and a cytosolic regulatory (R) domain unique to CFTR (Fig. 1 A, pale magenta), which links the two ABC-typical TMD–NBD halves together (Fig. 1 A; Riordan et al., 1989), and must be phosphorylated to allow channel activity (Cheng et al., 1991).
In phosphorylated channels binding of ATP to the “head” of each NBD promotes the formation of a head-to-tail NBD dimer which occludes two ATP molecules in composite interfacial binding sites (sites 1 and 2), one of which (site 2) possesses ATPase activity. NBD dimerization is coupled to TMD movements that resemble the rearrangement of ABC exporter proteins from an inward-facing (IF) to an outward-facing (OF; Fig. 1, B and C) orientation (Liu et al., 2017; Zhang et al., 2018b; reviewed, e.g., in Csanády et al., 2019). Phosphorylated CFTR channels in the presence of ATP exhibit bursting behavior: bursts of openings (to state O) separated by short “flickery” closures (to state Cf) are flanked by long “interburst” closures (to state Cs). State Cs corresponds to an IF TMD orientation (Bai et al., 2011; Wang et al., 2014) with a closed channel pore and at least partially separated NBDs (Vergani et al., 2005; Tsai et al., 2010). The burst state (B) is a composite state that includes the O and Cf states and exhibits an OF TMD conformation (Bai et al., 2011; Wang et al., 2014) associated with a tight, ATP-occluded NBD dimer (Vergani et al., 2005). For wild-type (WT) CFTR, this highly stable B state is terminated usually by the hydrolysis of ATP at site 2, which leads to the dissociation of the NBD dimer and resets the TMDs to the IF orientation (Csanády et al., 2010). Of note, the currently available cryo-electron microscopy (cryo-EM) structures of IF- and OF-CFTR (PDB accession nos. 5UAK and 6MSM) are not precise representations of the Cs and B states, respectively, as the former was obtained from unphosphorylated CFTR in the absence of ATP, while in the latter structure the external end of the pore is too narrow to accommodate a hydrated chloride ion. Nevertheless, to date, these structures are the best available models for gating-associated conformational changes of the channel pore.
In unphosphorylated CFTR the pore is closed (IF), and the R domain is wedged between the NBDs (Liu et al., 2017), preventing their dimerization. Upon activation, cAMP-dependent protein kinase (PKA) binds to the R domain (Mihályi et al., 2020), and in the presence of ATP, it phosphorylates multiple serines (Cheng et al., 1991), leading to the release of the R domain from its inhibitory intercalated position (Zhang et al., 2018b).
CFTR is a vertebrate-specific anion channel, the most ancient identified ortholog to date is found in the sea lamprey (Cui et al., 2019). Although all CFTR orthologs show a considerable degree of sequence homology, alterations in channel mechanics of different CFTRs are observable compared with human CFTR (hCFTR), one of the current endpoints of CFTR molecular evolution. The extensively studied zebrafish CFTR (zCFTR) is among the most ancient orthologs and might serve as a relatively early representative of CFTR development. Compared with hCFTR, the open probability (Po) of zCFTR is drastically reduced, resembling that of hCFTR harboring the CF-causing mutation hR117H (Zhang et al., 2018a). The available high-resolution structures of hCFTR and zCFTR, solved by cryo-EM, have shed light on structural differences that provide invaluable insight into understanding the structural development that has occurred during CFTR molecular evolution. While the IF structures of the two orthologs (hCFTR PDB accession no. 5UAK; zCFTR PDB accession no. 5UAR) are closely similar (Zhang and Chen, 2016; Liu et al., 2017), the two OF structures (hCFTR PDB accession no. 6MSM, Fig. 1 B; zCFTR PDB accession no. 5W81, Fig. 1 C) show distinct differences. In OF-hCFTR, the external end of the pore is open (though not wide enough to allow a hydrated chloride ion to permeate), whereas in OF-zCFTR it is tightly closed. Furthermore, the positions of the NBDs at site 1 and the arrangement of the extracellular loops in the two OF structures differ to some extent (Zhang et al., 2017; Zhang et al., 2018b). Understanding how the CFTR gating mechanism was optimized through its structural evolution might provide important insight into CFTR structure and function.
In hCFTR, the hR117 side chain in the first extracellular loop (ECL1) plays a critical role in stabilizing the O state by forming an H-bond with the backbone carbonyl group of hE1124 in ECL6, as suggested by the OF-hCFTR structure (Fig. 1 D) and confirmed by functional studies (Simon and Csanády, 2021). Lack of that stabilizing interaction in the class III mutant hR117H dramatically reduces Po (Sheppard et al., 1993; Yu et al., 2016) and causes CF (Dean et al., 1990; Wilschanski et al., 1995). Interestingly, an H-bond between the corresponding positions (zR118–zD1132) is not observable in the OF-zCFTR structure (Fig. 1 E): the zR118–zD1132 α-carbon distance is too large and the zR118 side chain is resolved only to the β-carbon atom, likely due to increased side-chain flexibility. Based on these structural observations, one possible explanation for the reduced Po of WT zCFTR might be that the stabilizing interaction of zR118 never forms during zCFTR gating, resulting in an hR117H-like phenotype. Alternatively, considering that the arginine is evolutionarily highly conserved, the absence of the stabilizing zR118–zD1132 interaction in the OF-zCFTR structure might be explained by assuming that the latter represents the Cf rather than the O state. To address this functional question, we aimed to test the effect of the zR118H mutation on zCFTR gating using inside-out patch clamp recordings. Furthermore, our purpose was to deconvolve the structural correlates of evolutionary changes in CFTR gating mechanisms.
Materials and methods
Reagents
NaCl (27800.291) and HEPES (441476L) were purchased from VWR Chemicals, KCl (P4504-1KG), MgCl2 (M1028-100ML), NaOH (S8045-1KG), CaCl2 (21115-250ML), NMDG (M2004-500G), and EGTA (03777-10G) were purchased from Sigma-Aldrich, and HCl was purchased from Supleco.
Molecular biology
The zCFTR coding sequence was subcloned into the XmaI and SalI sites of pGEMHE. Mutations were introduced into the hCFTR–E1371S (pGEMHE) and zCFTR–E1372S (pGEMHE) sequences using the QuikChange II XL Kit (#200521; Agilent Technologies) and confirmed by Sanger sequencing (LGC Genomics GmbH). To generate cRNA, the cDNA was linearized (NheI-HF, R3131 S; New England Biolabs) and transcribed in vitro using T7 polymerase (mMessage mMachine T7 Kit, AM1344; Thermo Fisher Scientific), and cRNA was stored at −80°C.
Functional expression of CFTR constructs in Xenopus laevis oocytes
Oocytes were removed by surgery from anesthetized Xenopus laevis, following Institutional Animal Care Committee guidelines, and separated by collagenase treatment (Collagenase type II, 17101015; Gibco). Isolated oocytes were kept at 18°C in a modified frog Ringer’s solution (82 mM NaCl, 2 mM KCl, 1 mM MgCl2, and 5 mM HEPES, pH 7.5 with NaOH) supplemented with 1.8 mM CaCl2 and 50 μg/ml gentamycin (G1397-10 mL; Sigma-Aldrich). Injections of 0.1–10 ng of cRNA, to obtain microscopic or macroscopic currents, were done in a fixed 50-nl volume (Nanoject II, Drummond). Recordings were performed 1–3 d after injection.
Functional expression of CFTR constructs in HEK293T cells
HEK293T cells were cultured at 37°C in 5% CO2 in Dulbecco’s modified Eagle’s medium with 4.5 g/liter glucose (Lonza), supplemented with 10% fetal bovine serum (EuroClone), 2 mM L-glutamine, and 100 units/ml penicillin/streptomycin (Lonza). HEK293T cells transiently expressing CFTR and green fluorescent protein (GFP) were obtained by cotransfection of HEK293T cells with a given CFTR–pcDNA3 construct and GFP–pcDNA3 at a 10:1 ratio (FuGENE HD transfection reagent, Promega). Measurements were performed 1–2 d after transfection.
Excised inside-out patch recording
The patch pipette solution contained 138 mM NMDG, 2 mM MgCl2, and 5 mM HEPES, pH 7.4 with HCl. The bath solution contained 138 mM NMDG, 2 mM MgCl2, 5 mM HEPES, and 0.5 mM EGTA, pH 7.1 with HCl. Following excision into the inside-out configuration patches (each from a different cell to enhance biological variability) were moved into a flow chamber in which the composition of the continuously flowing bath solution could be exchanged with a time constant of <100 ms using electronic valves (ALA-VM8, Ala Scientific Instruments). CFTR channel gating was studied at 25°C, in the presence of 2 mM MgATP (A9187-1G, Sigma-Aldrich), following activation by ∼1 min exposure to 300 nM bovine PKA catalytic subunit (P2645, Sigma-Aldrich; activation time constants were <20 s for each construct). Macroscopic currents were recorded at −20/−40/−80 mV and microscopic currents at −80/−120 mV membrane potential. Currents were amplified and low-pass filtered at 2 kHz (Axopatch 200B; Molecular Devices), digitized at a sampling rate of 10 kHz (Digidata 1550B; Molecular Devices), and recorded to disk (Pclamp 11; Molecular Devices).
Kinetic analysis of electrophysiological data
Relaxation time constants (τb; see Figs. 2 B; 3, B and D; 4 B; 5 E; 6 B; 7 E; and S2 B) were obtained by fitting macroscopic current relaxations to single exponentials using nonlinear least squares (Clampfit 11).
For intraburst kinetic analysis, ATP was briefly added to open prephosphorylated channels and was then removed. Segments of record following the last observed superimposed channel opening were analyzed (highlighted by black horizontal lines in Figs. 2 C, 5 A, 6 D, and 7 A). Mean open times (τopen) and mean flickery closed times (τflicker) were obtained as the simple arithmetic averages of the mean open and closed dwell-time durations, respectively, and Keq|B was calculated as Keq|B = τopen/τflicker (see Figs. 2 D, 5 B, 6 E, and 7 B). Of note, for low-Po constructs, this approach does not eliminate the possibility that two channels are still active during some initial part of the analyzed segment even if they do not open simultaneously. However, such a scenario predicts a sudden irreversible drop in apparent Po at the time point of the final closure of the last-but-one channel somewhere within the analyzed segment. Since we did not observe such trends upon visual inspection, we believe that for each of the studied constructs, the majority of total “last-channel” recording time must have indeed documented the gating of a single active channel, i.e., the potential overestimation of intraburst Po should be minimal.
Mutant cycle analysis
Changes in the strengths of the zS109–zN120, zS109–zR118, hS108–hR117, and hS108–hI119N interactions between various gating states were evaluated using thermodynamic mutant cycle analysis as described previously. Briefly, for an S1↔T‡↔S2 gating step, changes in the relative stabilities of those three states induced by a particular mutation were calculated as follows. The change in the height of a transition-state barrier, such as for step S1→T‡, was calculated as ΔΔG0T-S1 = −kTln(r'12/r12), where k is Boltzmann’s constant, T is absolute temperature, and r12 and r'12 are the rates for the S1→S2 transition in the background construct and in the mutant, respectively. The change in the stability of the two ground states relative to each other was calculated as ΔΔG0S2−S1 = −kTln(K'eq/Keq), in which Keq and K'eq are the equilibrium constants for the S1↔S2 transition in the background construct and in the mutant, respectively. Interaction free energy (ΔΔGint) was defined as the difference between ΔΔG0 values along two parallel sides of a mutant cycle. All ΔΔG values are given as mean ± SEM; SEM values were estimated assuming that rij and Keq are normally distributed random variables using second-order approximations of the exact integrals (Mihályi et al., 2016).
Statistics
All values are given as mean ± SEM, with the numbers of independent samples provided in each figure legend. Data distribution was assumed to be normal, but this was not formally tested. Statistical significances of interaction free energies were calculated using Student’s two-tailed t-test, ΔΔGint is reported significantly different from 0 for P < 0.05.
Sequence alignment
Sequences of CFTR orthologs were obtained using NCBI blast by providing the protein sequence of hCFTR (UniProt accession no. P13569) for searching in the non-redundant protein database. Sequences were manually curated and then aligned using the clustalW algorithm. Sequences to be displayed in the sequence alignment (Fig. S1) were selected based on a simplified phylogenetic tree of CFTR molecular evolution (Infield et al., 2021).
Online supplemental material
Results
Non-hydrolytic zCFTR background construct zE1372S exhibits lower intraburst Po but similar burst duration when compared to human ortholog
In WT CFTR channels, pore closure can happen along two pathways: NBD dimer dissociation may follow hydrolysis of ATP at site 2 or may occur occasionally in the absence of ATP hydrolysis. Thus, mutational effects on open-state stability are difficult to interpret. Mutations of the Walker B glutamate side chain in NBD2 (e.g., hE1371S/zE1372S) prevent ATP hydrolysis. Such “non-hydrolytic” mutations prolong the mean burst duration (τburst) by more than two orders of magnitude and reduce CFTR gating in saturating ATP to reversible Cs↔B transitions (Csanády et al., 2010). For such channels, the slow rate of the non-hydrolytic B→Cs transition can be measured in macroscopic inside-out patch-clamp recordings by activating prephosphorylated channels with exposure to ATP (2 mM for 1 min), followed by sudden ATP removal. Fitting the resulting slow current decay time courses (Fig. 2 A, black and gray traces) with exponentials the obtained time constant reports the average lifetime of the B state under non-hydrolytic conditions (τburst; Table 1 and Fig. 2 B, black and gray bars). Using that approach, we determined τburst for both hE1371S (Fig. 2 A, black trace; Fig. 2 B, black bar) and zE1372S (Fig. 2 A, gray trace; Fig. 2 B, gray bar), the two background constructs used in this study. We found that mean burst duration is slightly, but not significantly (P = 0.071) longer for zE1372S (57 ± 8.3 s, n = 9, biological replicates) than for hE1371S (37 ± 3.7 s, n = 7, biological replicates), consistent with the formation of a stable NBD dimer already in the zebrafish ortholog.
The B state is a compound state, consisting of the O and Cf states. As the burst is prolonged in the hE1371S and zE1372S backgrounds, it is possible to inspect large numbers of openings and flickery closings without any interburst event by observing the last open channel following ATP removal under such conditions gating continues until the first transition to the interburst closed state from which the probability of reopening is extremely low in the absence of ATP (Simon and Csanády., 2021). This allows studying effects on intraburst Po (Po|B) or the intraburst closed–open equilibrium constant (Keq|B = Po|B/[1 − Po|B]). By comparing the gating of last open channels for the hE1371S (Fig. 2 C, black trace) and zE1372S (Fig. 2 C, gray trace) background constructs, a large difference in intraburst Po values is immediately evident: the calculated Keq|B values (Table 1 and Fig. 2 D, black symbol—hE1371S, gray symbol—zE1372S) show a striking ∼54-fold difference between the two orthologs. In summary, under non-hydrolytic conditions, intraburst kinetics of zE1372S resembles that of hR117H, but the B state of zE1372S seems as stable as that of hE1371S.
Different substitutions of the catalytic glutamate differentially affect burst stability in zCFTR and hCFTR
Our finding of a reduced intraburst Po for zE1372S compared with hE1371S, obtained in patches excised from Xenopus oocytes, is in agreement with an earlier report which compared zE1372Q and hE1371Q channels expressed in mammalian cell membranes (Zhang et al., 2018a). On the other hand, the similar burst durations of zE1372S and hE1371S seem to contrast with the findings of the earlier study which reported approximately fourfold shorter τburst for zE1372Q compared with hE1371Q.
To evaluate the possibility that membrane composition might influence the stability of NBD dimers, we performed similar experiments on CFTR channels expressed in HEK293T mammalian cells instead of our Xenopus oocyte model system (Fig. 3 A; brown trace—hE1371S, pale brown trace—zE1372S) and determined τburst (Fig. 3 B, colored bars). However, we found no significant difference either in τburst of hE1371S (P = 0.42) or in τburst of zE1372S (P = 0.091) between the two expression systems.
To evaluate the alternative possibility that the different substitutions used in the two studies underlie the seemingly discrepant results, we constructed hE1371Q and zE1372Q and assessed their mean burst durations in our oocyte-based model system. Surprisingly, replacement of the catalytic glutamate with glutamine resulted in bursts which for hCFTR were >10-fold longer, but for zCFTR not significantly longer (P = 0.092), compared with those of the respective serine mutants (Fig. 3 A vs. 3 C, black and gray traces; Fig. 3 B vs. 3 D, black and gray bars). As a consequence, τburst was approximately sixfold shorter for zE1372Q compared with hE1371Q (Fig. 3, C and D), consistent with the earlier report (Zhang et al., 2018a). Thus, different substitutions of the catalytic glutamate differentially affect burst stability in zCFTR and hCFTR, suggesting important differences in the network of interdomain interactions that stabilize the NBD dimer interface. Whereas identification of the source of these differences is beyond the scope of the present work, these findings underscore the importance of employing a single non-hydrolytic background construct for the evaluation of the energetic effects of second-site mutations. In the present study, serine mutants were chosen as the background for both orthologs.
Mutation R118H in zCFTR produces no functional phenotype
We next examined the effect of the zR118H mutation on the gating of zCFTR (Fig. 4 A, pale blue trace). Surprisingly, unlike the acceleration caused by mutation hR117H in the hCFTR background (hE1371S; Fig. 4 A, black vs. dark blue trace), in the zCFTR background (zE1372S), the zR118H mutation did not result in an hR117H-like acceleration of the current decay (Fig. 4 A, gray vs. pale blue trace). Consistently, analysis of fitted τburst values (Table 1 and Fig. 4 B, colored bars) reported a slight, but statistically not significant (P = 0.27) ∼1.3-fold increase in τburst for zR118H compared to zE1372S, instead of an ∼4-fold decline predicted based on the effect of the analogous mutation on hE1371S. Since at the bandwidth of our recordings the unitary amplitude of last channel currents was smaller than 0.3 pA at −120 mV (0.6 pA for zE1372S), intraburst kinetics could not be examined reliably.
Outward-facing zCFTR structure suggests alternative stabilizing interaction within ECL1
As the bursting state in zE1372S is as stable as that of hE1371S (Fig. 4 B, black vs. gray bar) but τburst is not shortened by mutation zR118H, an alternative stabilizing interaction—which does not involve the zR118 side chain—must be present in zCFTR. Therefore, we examined the extracellular region of the OF-zCFTR structure in greater depth, searching for an interaction between ECL1 and ECL6, which is absent in all other CFTR structures (IF-zCFTR, IF/OF-hCFTR). Interestingly, we discovered an H-bond between the side chains of zS109 and zN120 (Fig. 4 D), which might potentially serve to stabilize ECL1. This interaction is absent in the IF-zCFTR structure (Fig. 4 F), as well as in all hCFTR structures (Fig. 4 C and Fig. 3 E), in which the serine is conserved (hS108) but the asparagine is replaced by an isoleucine (hI119). Moreover, for the latter position, a clear evolutionary trend is observable based on sequence alignment (Fig. S1): a polar amino acid present in more ancient organisms (e.g., E in sea lamprey, N/Q in most fishes) is replaced by a hydrophobic residue in later orthologs (e.g., I in mammals). However, because of the limited resolution of these extracellular regions in both the electrondensity maps (EMD-9230 and EMD-8782) and the typically poor cryo-EM signal of charged side chains in general, these modeled side-chain interactions must be interpreted with caution, unless their existence is functionally verified. We thus aimed to test the validity and functional relevance of the putative H-bond between positions z109 and z120.
Intriguingly, in the OF-hCFTR structure (Fig. 4 C), the modeled backbone carbonyl group of hS108 forms an H-bond with the δ-amino group of the hR117 side chain, suggesting a functional coupling between the two positions (hS108–hR117). Thus, we also aimed to verify the functional role of position hS108 in the human ortholog.
The zS109–zN120 interaction stabilizes the flickery closed state in zCFTR
The functional impact of a mutation at a certain position of a protein depends on the whole interactome of that specific residue, as one amino acid often takes part in multiple interactions, all of which are perturbed upon mutagenesis. To isolate and quantify the energetic effect of altering a single specific interaction between two target residues, thermodynamic mutant cycle analysis must be applied (Fersht, 2002; Vergani et al., 2005). For that purpose, a thermodynamic cycle is formed by the background construct, two single-mutant constructs, each perturbing one participant of the selected interaction, and a double mutant containing both mutations at the selected positions. If the interaction of the target residues changes between two gating (ground/transition) states (S1, S2), then the impact on the relative stability of state S2 (ΔΔG°S2-S1) caused by perturbing one target residue depends on the nature of the side chain present at the other target position; thus, the energetic effects of each single mutation are not additive in the double mutant. By contrast, additive effects of the single mutations signal that the interaction is not present or does not change between the S1 and S2 gating states. The change in interaction energy of the target interaction between the two gating states (ΔΔGint (S1 → S2)) is calculated as the difference of ΔΔG°S2-S1 values along two parallel sides of the mutant cycle (Fig. 5 C).
To study how the zS109–zN120 interaction changes during intraburst gating, we extracted the intraburst equilibrium constant (Fig. 5 B, colored symbols) from records of last open channels (Fig. 5 A, colored traces) for zE1372S, and for the zN120A, zS109A, and zS109A–N120A mutants constructed in the same background. The zN120A mutation increased Keq|B by approximately sevenfold when introduced into the background construct (Table 1 and Fig. 5, A and B, gray vs. magenta), but not when introduced into the zS109A background (Fig. 5, A and B, orange vs. green). The calculated change in interaction energy between positions zS109 and zN120 during step Cf → O is +2.7 ± 0.46 kT (Fig. 5 C, purple; k, Boltzmann's constant; T, temperature). This value is similar to the average energy of an H-bond (∼2–3 kT), and its positive signature is consistent with a stabilizing interaction selectively present in state Cf, confirming the assumed H-bond between the target positions. As the S109–N120 interaction is present only in the Cf state and is absent in the O state, the phosphorylated OF-zCFTR structure must represent the Cf rather than the O state.
Next, we investigated how the strength of the zS109–zN120 interaction changes while the channel proceeds from the bursting state (B) to the transition state for interburst closure (T‡) by comparing τburst of our background construct (zE1372S), single mutants zS109A and zN120A, and the corresponding double mutant (Fig. 5 D, colored traces; Fig. 5 E, colored bars; and Table 1). Both single mutations decreased τburst, consistent with the overall destabilization of the bursting state by the mutations, but these effects were largely additive, e.g., the zS109A mutation decreased τburst by approximately threefold, regardless of whether it was introduced into the zCFTR background (Fig. 5, D and E, gray vs. orange) or into zN120A (Fig. 5, D and E, magenta vs. green). Similar changes in ΔΔG°T‡-B along two parallel sides of the mutant cycle (Fig. 5 F, numbers on vertical arrows) resulted in a ΔΔGint value close to 0 for the B → T‡ gating step (Fig. 5 F, purple). This suggests that the interaction is either not present or static between the two states. As the zE1372S background construct spends the majority of the burst in the Cf state (Fig. 5, A and B, gray) in which the target interaction is formed, it is assumed that the interaction is maintained in the T‡ state (but is likely absent in the Cs state as it is not observed in the IF-zCFTR structure).
Of note, whereas single mutation zN120A shifted the Cf – O equilibrium toward the O state (Fig. 5 B, gray vs. magenta), as expected for disruption of an interaction that stabilizes the Cf state, single mutation zS109A rather decreased Keq|B (Fig. 5 B, gray vs. orange). This suggests that besides its interaction with zN120, zS109 is also involved in a strong stabilizing interaction, which is formed in the O state, with an unknown residue different from zN120. Disruption of both of these interactions in the zS109A mutant would explain on one hand the lack of a net effect on Keq|B and, on the other hand, the much larger effect on τburst (Fig. 5 E, magenta vs. orange) of the zS109A single mutation.
In hCFTR, hS108 cooperates with hR117 to stabilize the O state
Although the hydrophilic residue corresponding to zN120 is replaced by hydrophobic isoleucine in higher eukaryotes, the serine corresponding to zS109 is conserved throughout evolution (Fig. S1). Moreover, in the OF-hCFTR structure, a potential H-bond is observable between the backbone carbonyl group of the latter position (hS108) and the δ-amino group of hR117 (Fig. 4 C). We, therefore, examined if residue hS108 plays any role in hCFTR gating. To attempt to perturb the position of its backbone carbonyl group, we truncated the side chain of hS108, which is involved in stabilizing the terminal portion of transmembrane helix 1 (TM1).
We first studied macroscopic current relaxations upon ATP removal (Fig. 6 A) to extract mean burst durations (Table 1 and Fig. 6 B). Interestingly, mutation hS108A decreased τburst approximately threefold compared with the hE1371S construct (Fig. 6, A and B, black vs. orange). As mutation hR117H affected τburst in a similar way (Fig. 6, A and B, black vs. blue), we constructed the double mutant to address additivity between the two positions. Introducing mutation hS108A into an hR117H background increased τburst by approximately twofold (Fig. 6, A and B, blue vs. green), revealing strong coupling between positions hS108 and hR117. The extent of this coupling, quantified by the thermodynamic mutant cycle, amounts to ΔΔGint (B → T‡) = 1.9 ± 0.20 kT (Fig. 6 C, purple), confirming the existence of the potential H-bond in the bursting state.
Next, we studied the change in hS108–hR117 interaction energy between Cf and O states by quantifying mutational effects on the intraburst gating equilibrium (Fig. 6 E) extracted from records of last open channels (Table 1 and Fig. 6 D). Introduction of mutation hS108A decreased Keq|B dramatically, by ∼70-fold, compared with the background (Fig. 6, D and E, black vs. orange), again mimicking the hR117H phenotype (Fig. 6, D and E, black vs. blue). To complete the thermodynamic mutant cycle, we studied the effect of the hS108A mutation also in the hR117H background and found that it resulted in only an approximately fourfold decrease in Keq|B (Fig. 6, D and E, blue vs. green). The magnitude of the calculated ΔΔGint (Cf → O) = −2.9 ± 0.47 kT (Fig. 6 F; purple) is comparable with the strength of an H-bond, and its negative signature implies that within the B state, the H-bond between positions hS108 and hR117 is present in the O state but not in the Cf state.
We also examined whether a potential link might exist also in zCFTR between zS109 and zR118. For that purpose, mutation zR118H was introduced into the zS109A construct (Fig. S2 A, green trace) and τburst was determined as described above (Table 1 and Fig. S2 B, green bar). Although the two single mutations zS109A and zR118H exerted inverse effects on τburst of the zE1372S background (Table 1 and Fig. S2 B, blue and orange vs. gray), their effects in the double mutant proved additive (Fig. S2 A, green vs. gray). The calculated ΔΔGint value was 0.43 ± 0.26 kT (Fig. S2 C, purple), which is not significantly different from 0 (P = 0.14). Therefore, no coupling is present between the zR118 and zS109 residues in the zebrafish ortholog.
In hCFTR, the ancient interaction which stabilizes the Cf state can be restored
Although at the position corresponding to zN120 replacement of the polar residue (N/Q/E) by isoleucine happened relatively early in CFTR evolution, we were curious whether such an interaction can be restored in later orthologs. Therefore, we introduced the hI119N mutation into our hCFTR background (hE1371S) and examined its potential interaction with hS108 by deducing the Keq|B (Table 1 and Fig. 7 B, colored symbols) from last open channel recordings (Fig. 7 A, colored traces).
The hI119N mutation decreased Keq|B by ∼17-fold when introduced into the background construct (Fig. 7, A and B, black vs. magenta), but only by ∼3.5-fold when introduced into the hS108A background (Fig. 7, A and B, orange vs. green), suggesting non-additive effects of the two mutations. To correctly interpret these data, it has to be considered that the postulated H-bond may exist only in the hI119N mutant, which contains both the hS108 and the hN119 target side chain. Thus, following conventional mutant cycle formalism, this construct should be considered the “background” (Fig. 7 C, top left corner), whereas constructs hE1371S (only hS108 present) and hS108A-I119N (only hN119 present) should be considered “single mutants” (Fig. 7 C, top right and bottom left corners), and construct hS108A (neither hS108 nor hN119 present) the “double mutant” (Fig. 7 C, bottom right). The calculated change in interaction energy between positions hS108 and hN119 of the hI119N construct during step Cf → O is +1.6 ± 0.38 kT (Fig. 7 C, purple). That value is comparable with the average energy of an H-bond (∼2–3 kT), and its positive signature is consistent with a stabilizing interaction present in state Cf, similar to the zS109–zN120 interaction (Fig. 5, A–C). Thus, the H-bond between positions h108 and h119, lost over evolution, is restored in the hI119N mutant.
Next, we examined macroscopic current relaxations upon removing ATP (Fig. 7 D, colored traces) to deduce τburst values (Table 1 and Fig. 7 E, colored bars). Mutation hI119N increased τburst slightly, by ∼1.5-fold, when introduced into the hE1371S background construct (Fig. 7, D and E, black vs. magenta), but by 2.5-fold when introduced into construct hS108A (Fig. 7, D and E, orange vs. green). The change in interaction energy quantified by the thermodynamic mutant cycle is −0.64 ± 0.26 kT (Fig. 7 F, purple), which is small, but significantly different from 0 (P = 0.042). The small but negative ΔΔG reports a small increase in the strength of the H-bond in the hI119N channel as it transits from the B state to the T‡ state. As during a burst the H-bond forms only in state Cf and hI119N channels spend ∼50% of the burst-time in that state (Fig. 7 B, magenta symbol), the time-average of the interaction energy within a burst amounts to approximately half of that of an H-bond. Thus, the apparent further small stabilization of the interaction in state T‡ suggests that the H-bond is formed in the T‡ state. In contrast, based on the modeling of the hI119N mutant by Missense3D, no interaction is present between the target positions in the IF state, thus, the hS108–hN119 interaction is likely to be absent in state Cs. These data together suggest that in the hI119N construct, the restored H-bond between hS108 and hN119 stabilizes the Cf state, but is absent in the O and Cs states, similar to the zS109–zN120 interaction natively present in the zebrafish ortholog.
Discussion
The NBD-dimerized OF conformation of the CFTR channel corresponds to the bursting state within which the ion-conducting pore alternates between O and Cf states. For human CFTR, the bursting state is highly stable: when ATP hydrolysis is prevented τburst is increased by >100-fold, from ∼0.2 s (Csanády et al., 2010) to ∼37 s (Fig. 3 B, black) or ∼470 s (Fig. 3 D, black), depending on the catalytic site mutation employed. The mutation-specific variations in non-hydrolytic burst duration clearly indicate that, in addition to disrupting ATP hydrolysis, different catalytic site mutations differentially affect the stability of the prehydrolytic NBD dimer. Nevertheless, regardless of which mutant most closely resembles the prehydrolytic state of a WT channel, the >100-fold difference in hydrolytic and non-hydrolytic closing rates indicates that in WT hCFTR the vast majority of bursts is terminated by ATP hydrolysis (cf. Csanády et al., 2010). In that regard, the bursting state itself is comparably stable in the zebrafish ortholog. Given the reported τburst of ∼0.7 s of WT zCFTR (Zhang et al., 2018a), the observed ∼100-fold longer τburst of non-hydrolytic mutants (Fig. 2, B and D, gray) reports that the gating (burst–interburst) cycle is strictly coupled to ATP hydrolysis already in the zebrafish ortholog. On the other hand, for human CFTR, the bursting state is also highly conductive: within a burst, the pore is open for ∼95% of the time (intraburst Po is ∼0.95; cf., Fig. 2 D, black). In contrast, the intraburst Po of zCFTR is much lower (∼0.29, Fig. 2 D, gray; cf. Zhang et al., 2018a). Thus, from the point of view of ion conduction, zCFTR is a “less efficient” ion channel, which was greatly optimized in the course of evolution.
Cryo-EM structures of hCFTR and zCFTR are highly similar in both IF and OF conformations but show marked differences in the arrangement of extracellular loops (Zhang et al., 2017; Zhang et al., 2018b), a region known to have a strong impact on intraburst Po by contributing interactions that stabilize the NBD-dimerized OF conformation (Cui et al., 2014; Infield et al., 2016; Yu et al., 2016). Inspection of these structures provided a good starting point for identifying functionally relevant state-dependent interactions that differ for zCFTR and hCFTR (Fig. 4, C–F) and are formed between conserved or systematically evolving positions (Fig. S1). However, because the resolution of the cryo-EM maps is poor in this region, the relevance of these suggested interactions must be validated by functional studies. In particular, an earlier functional study showed that in hCFTR, a strong H-bond between the side chain of hR117 in ECL1- and ECL6-residue hE1124 (Fig. 1 D), selectively present in the O state, is required both for the stable bursting state and the high intraburst Po characteristic of hCFTR: mutations of hR117 dramatically impair channel gating and cause CF (Simon and Csanády., 2021). Here we show that despite its high conservation throughout evolution (Fig. S1), the ECL1 arginine does not play such a stabilizing role in the ancient zCFTR ortholog, as its mutation does not produce a phenotype (Fig. 4, A and B, gray vs. light blue). Instead, based on the OF-zCFTR structure, we identify here zN120 and zS109, located at the extracellular ends of TM1 and TM2, respectively, and show that they play a pivotal role in burst stabilization: truncation of their side chains shorten τburst (Fig. 5, D and E, gray vs. magenta and orange). Interestingly, however, intraburst Po is oppositely affected by these mutations (Fig. 5, A and B, gray vs. magenta or orange), suggesting that the zN120 and zS109 side chains are differentially involved in stabilizing the O and the Cf state. We further show here that these effects involve state-dependent interactions of zS109 with two distinct residues (cartoon in Fig. 8, upper panel). An H-bond with zN120, located at the extracellular end of TM2, is selectively formed in state Cf (Fig. 5 C). As that bond is visible in the OF-zCFTR structure, we conclude that the latter must represent a Cf state. In contrast, identification of the O state interaction partner of zS109 (question mark in Fig. 8, upper panel) will require an open zCFTR structure.
Although zCFTR cannot be considered a direct predecessor of hCFTR, a clear evolutionary tendency can be observed for the residue corresponding to zN120: polar residues present in more ancient organisms were eventually replaced by isoleucine in later species. Based on sequence alignment (Fig. S1), a zS109–zN120-like interaction might have been present not only in bony fishes but already in the sea lamprey, a jawless vertebrate in which the ECL1 arginine did not yet appear, suggesting that it might represent an early stabilizing strategy. To address the timing of the evolutionary switch which eventually replaced that interaction with an H-bond between ECL6 and the ECL1 arginine, multiple CFTR orthologs would have to be studied both functionally and structurally, which is beyond the scope of the present study. Interestingly, the ancient polar interaction must have been lost also in the course of the evolution of the Chondrichthyes branch, as extant cartilaginous fishes (e.g., sharks) also lack a polar residue at the position corresponding to zN120 (Fig. S1). Interestingly, the ancient polar interaction can be restored in the human ortholog by replacing the isoleucine with asparagine (hI119N, Fig. 7).
Remarkably, despite the eventual loss of the zS109–zN120-like polar interaction, the serine has remained preserved throughout evolution (Fig. S1), and we show here that in hCFTR its presence is required for the open-state stabilization provided by the hR117 side chain (Fig. 6). Although the hS108–hR117 H-bond suggested by the OF-hCFTR structure forms between the hR117 side chain and the backbone carbonyl group of hS108 (Fig. 4 C), the structure also suggests that the serine side chain is important for the positioning of its backbone by stabilizing the last turn of the TM1 helix—this would explain the strong functional effect of the hS108A side chain truncation (Fig. 6). It is presently unclear whether the hS108–hR117 interaction is required for proper positioning of the arginine side chain to facilitate its interaction with ECL6-residue hE1124 or for extending the mechanical coupling from ECL6 to the C-terminal end of TM1, but it is clear that the simultaneous presence of both interactions is required for open-state stabilization in hCFTR (Fig. 8, lower panel).
Mutations of hR117 are the cause of CF for thousands of people. Interestingly, despite the disastrous impacts on CFTR gating of the hS108A mutation (Fig. 6), variations of hS108 are not registered in the CF database. There are several possible reasons to explain the lack of CF patients with such a mutation. First, serine is encoded by six codons, and any replacement of the third base in the native codon of hS108 (TCT) is silent. Second, although the same is true for hR117, the replacement of hS108 by other polar amino acids, such as threonine or asparagine, might preserve the native orientation of its backbone, whereas any replacement of the hR117 side chain will necessarily disrupt the ECL1–ECL6 H-bond, destabilizing the O state. Third, it is also possible that mutations of the serine residue are embryonic lethal.
The bursting gating pattern of CFTR might be explained by two linear three-state schemes: Cs – Cf – O or Cs – O – Cf. To date, no data exist that would allow a distinction between these two possibilities. In particular, the dependences of microscopic reaction rates on ATP concentration (Winter et al., 1994), pH (Chen et al., 2017), or voltage (Cai et al., 2003), as well as the effects of disrupting an interaction that selectively stabilizes the O state (Simon and Csanády., 2021), are equally compatible with either model. Although in principle a distinction might be offered by mutations that either disrupt an interaction selectively present in the Cf state (cf., zN120A) or create a Cf-specific interaction (cf., hI119N), evaluation of the effects of these two mutations by fitting either scheme did not provide any strong bias. However, a slight preference for the Cs – Cf – O scheme (Fig. 8) is supported by the state dependence of the target interaction (z109–z120 native, h108–h119 in hI119N). Based on the mutant cycles (Fig. 5 and Fig. 7), the H-bond forms in states Cf and T‡ for both orthologs. Whereas in the Cs – Cf – O scheme those two states are directly linked, in the Cs – O – Cf scheme they are separated.
In conclusion, we have pinpointed specific structural differences in ECL1 that are responsible for the functional differences between two distant CFTR orthologs. Our findings provide first insights into how the CFTR gating mechanism was optimized in the course of CFTR molecular evolution.
Acknowledgments
Joseph A. Mindell served as editor.
The zCFTR coding sequence was a generous gift of T.C. Hwang, University of Missouri-Columbia, Columbia, MO, USA.
This work was supported by EU Horizon 2020 Research and Innovation Program grant 739593, MTA Lendület grant LP2017-14/2017, and Cystic Fibrosis Foundation Research Grant CSANAD21G0 to L. Csanády. M.A. Simon received support from the ÚNKP-21–3-II-SE-19 New National Excellence Program of the Ministry for Innovation and Technology from the source of the National Research, Development and Innovation Fund.
The authors declare no competing financial interests.
Author contributions: M.A. Simon: Conceptualization, Formal analysis, Funding acquisition, Investigation, Software, Visualization, Writing – original draft. L. Csanády: Data curation, Funding acquisition, Project administration, Methodology, Software, Supervision, Validation, Writing – review & editing.