The CFTR anion channel is a key component of transepithelial salt-water transport in the lung, the pancreas and the intestine, and its mutations cause cystic fibrosis (O’Sullivan and Freedman, 2009). CFTR (ABCC7) is the only known ion channel member of the family of ATP-Binding Cassette (ABC) proteins. Its two ABC-typical halfs, each consisting of a transmembrane domain (TMD1 and 2, Figure 1A, gray) followed by a cytosolic nucleotide-binding domain (NBD1 and 2; Figure 1A, blue and green), are linked by a unique regulatory (R) region (Figure 1A, pale red) which must be phosphorylated by cAMP-dependent protein kinase (PKA) to allow channel activity (Riordan et al., 1989; Cheng et al., 1991).

Figure 1

Download asset Open asset

ABC proteins are present in all kingdoms of life. The 48 human ABC proteins play essential physiological roles by mediating transmembrane transport of a variety of substrates (Dean et al., 2022). Within the ABC family the NBDs are the most highly conserved modules, and consist of a core subdomain that binds ATP (the ‘head’) and an ABC-specific alpha-helical subdomain (the ‘tail’). In all ABC proteins in the presence of ATP the two NBDs form a tight head-to-tail dimer which occludes two ATP molecules (Figure 1B, yellow) at the interface. Both nucleotide-binding sites are flanked on one side by the conserved Walker A and Walker B motifs in the head of one NBD (Figure 1B, red), and on the other side by the conserved ‘ABC signature sequence’ in the tail of the other NBD (Figure 1B, magenta). The ATP-bound NBD dimer is stable, but is disrupted following ATP hydrolysis (Locher, 2016). In a subset of ABC proteins, which includes the entire C subfamily, only one of these two composite ATP-binding sites is catalytically active (Procko et al., 2009). In CFTR the composite site formed by the head of NBD2 and the tail of NBD1 (‘site 2’, Figure 1B, upper site) contains canonical sequence motifs and hydrolyzes ATP with an overall turnover number of ~0.5–1 s⁻¹ (Li et al., 1996; Liu et al., 2017). In contrast, the other site (‘site 1’, Figure 1B, lower site) has accumulated non-canonical substitutions in several key motifs (Figure 1B, pale residues) and is catalytically inactive (Aleksandrov et al., 2002; Basso et al., 2003).

During substrate transport the TMDs of ABC exporters alternate between inward-facing (IF) and outward-facing (OF) conformations. In a phosphorylated CFTR channel opening and closure of the transmembrane ion pore (gating) follows analogous TMD movements (Csanády et al., 2019; Figure 1C). CFTR likely adopts an IF conformation, in which the pore is sealed near its extracellular end, during the long lived (~1 s) state (Figure 1C, left) which corresponds to long ‘interburst’ (IB) closed dwell times observable in single channel current recordings. We refer to these closed states as ‘IF states’. When CFTR adopts the OF conformation the external gate is predominantly open, and a lateral portal which connects the channel pore with the cytosol generates a transmembrane aqueous pathway permeable to anions (Zhang et al., 2018; Figure 1C, right, double arrows). However, functional studies show that the continuity of the transmembrane pore is occasionally disrupted for brief (~10 ms) intervals by a smaller conformational change (not depicted in Figure 1C) likely confined to the external ends of the TMD helices (Zhang et al., 2017; Zhang et al., 2018; Simon and Csanády, 2021). Correspondingly, in single-channel recordings, this ‘OF’ or ‘bursting’ (B) state corresponds to clusters of channel openings separated by brief (flickery) closures (Winter et al., 1994). The OF state is also relatively stable, with a dwell time of hundreds of milliseconds.

In ABC exporters NBD and TMD movements are coupled: formation of the tight ATP-bound NBD dimer flips the TMDs from an IF to an OF conformation, and disruption of the tight NBD dimer following ATP hydrolysis resets the TMDs to the IF conformation (Locher, 2016). Gating of phosphorylated CFTR channels is driven by an analogous unidirectional conformational cycle (Csanády et al., 2019; Figure 1C). In all ABCC proteins, the OF state of the TMDs is coupled to tight dimerization of both NBD catalytic sites (Johnson and Chen, 2018; Zhang et al., 2018; Huang et al., 2023). Correspondingly, functional studies on CFTR showed that site 2 is tightly dimerized in the B state (Vergani et al., 2005; Figure 1C, right). In contrast, in IF structures of ABCC proteins obtained in the absence of ATP (Johnson and Chen, 2017; Liu et al., 2017; Huang et al., 2023), or for unphosphorylated CFTR even in its presence (Levring et al., 2023), the NBDs are seen to widely separate losing all contact across both composite sites. Interestingly, in the asymmetric bacterial ABC protein Tm287–288 some contacts across the degenerate site are retained throughout the entire conformational cycle, even in the IF state, as shown by extensive structural (Hohl et al., 2012; Hohl et al., 2014) and functional (Timachi et al., 2017) studies. Similarly, for phosphorylated CFTR channels functional studies suggested that site 1 does not completely separate throughout the entire ATP-driven gating cycle, even during IB events (Tsai et al., 2010; Szollosi et al., 2011) (see NBDs in Figure 1C, left), a prediction confirmed by recent single-molecule FRET measurements (Levring et al., 2023).

In ABC exporters thermodynamically uphill transport requires unidirectional conformational cycling. The ‘coupling ratio’ (CR), that is, the fraction of initiated cyles that are completed through ATP hydrolysis, depends on the relative rates of the two possible exit pathways from the prehydrolytic OF state. In most ABC transporters the actual values of these rates are hard to directly estimate, but in CFTR direct measurements of conformational dwell times are made feasible by single-channel current recordings, providing estimates for microscopic transition rates. Thus, for a wild-type (WT) channel ATP hydrolysis (rate k₁, Figure 1C) vs. non-hydrolytic NBD dimer dissociation rate (k₋₁, Figure 1C) can be estimated. Because k₁ >> k₋₁, CR = k₁/(k₁ + k₋₁) is near unity. In particular, for WT CFTR the mean burst duration (τ_b) reports the sum of the life times of the prehydrolytic (Figure 1C, state B₁) and posthydrolytic (Figure 1C, state B₂) OF conformations. Because the life time of B₂ is short compared to that of B₁ (k₂ >> k₁; Figure 1C; Gunderson and Kopito, 1995; Csanády et al., 2010), 1/τ_b provides a rough estimate of rate k₁, which is ~4–5 s⁻¹ at room temperature for pre-phosphorylated CFTR channels gating in ATP. In contrast, rate k₋₁ in WT CFTR is difficult to directly assess, and its estimates are based on the burst-state stability of various catalytic site mutants (Figure 1A, red star) in which ATP hydrolysis is disrupted (Figure 1C, red cross) and gating in saturating ATP is thus reduced to reversible IB₁ ↔ B₁ transitions (Figure 1B, red box). Numerous studies in the past have employed multiple different mutations for that purpose, including NBD2 Walker A lysine mutation K1250A (Gunderson and Kopito, 1995; Carson et al., 1995; Zeltwanger et al., 1999; Vergani et al., 2003; Csanády et al., 2006; Csanády et al., 2013; Csanády and Töröcsik, 2014), Walker B aspartate mutation D1370N (Gunderson and Kopito, 1995; Bompadre et al., 2005; Csanády et al., 2010; Sorum et al., 2015; Yeh et al., 2015; Sorum et al., 2017; Simon and Csanády, 2021), and E1371S/Q mutations which eliminate the catalytic base (Vergani et al., 2003; Bompadre et al., 2005; Zhou et al., 2005; Vergani et al., 2005; Zhou et al., 2006; Csanády et al., 2013; Yu et al., 2016). Of note, so far all OF structures of CFTR (PDBID: 6msm, 6o2p, 6o1v, 7svd, 7sv7), as well as OF structures of several other ABC proteins (6bhu, 6cov, 6hbu, 6s7p, 7ekl, 8iza), were obtained from E-to-Q catalytic glutamate mutants.

However, interpretation of such mutant data is hampered by the fact that – even when compared under identical experimental conditions (expression system, temperature, and phosphorylation state) – non-hydrolytic closing rates of the above mutants scatter over a range of ~200-fold, from ~0.0025 s⁻¹ for E1371Q (Vergani et al., 2005; Yu et al., 2016) to ~0.5–1 s⁻¹ for D1370N (Vergani et al., 2003; Yeh et al., 2015). The reason for that large scatter is unknown. One extreme possibility is that ATP hydrolysis is completely abolished only in the slowest closing mutant (E1371Q) which therefore faithfully reports WT B₁-state stability, whereas in the faster closing mutants significant residual hydrolytic activity persists. Such a scenario would call into question conclusions based on the assumption of equilibrium gating in the latter mutants (Sorum et al., 2015; Yeh et al., 2015; Sorum et al., 2017; Simon and Csanády, 2021). The other extreme possibility is that all of the above mutations eliminate ATP hydrolysis, but different mutations differentially affect – either increase or decrease – B₁-state stability. In either case, estimating the true stability of state B₁ (i.e., rate k₋₁) for WT CFTR would substantially promote our understanding of gating energetics. The aim of the present study was to estimate the true rate k₋₁ for a pre-phosphorylated WT CFTR channel gating in ATP, as well as to verify disruption of ATP hydrolysis and evaluate effects on B₁-state stability of the various non-hydrolytic mutations that have been extensively employed as models in the past. The results suggest unique features of the site 2 NBD interface in human CFTR compared to other ABCC family proteins.

In ABC proteins the E-to-Q mutation of the catalytic base, a conserved glutamate that follows the Walker B aspartate, abrogates ATP hydrolysis and traps isolated NBD proteins in stable dimer form (Moody et al., 2002; Smith et al., 2002; Janas et al., 2003). For that reason, E-to-Q mutants have been adopted as model systems in a multitude of structural studies, in an attempt to trap the protein in a prehydrolytic state. Indeed, under cryo-EM conditions in the presence of ATP full-length E-to-Q mutant ABC proteins yield OF structures in which ATP is seen intact, confirming lack of hydrolysis (Liu et al., 2017; Johnson and Chen, 2018; Kim and Chen, 2018; Olsen et al., 2020; Song et al., 2021; Huang et al., 2023). In hCFTR a first hint for a possible additional role, apart from simply disrupting ATP hydrolysis, of the engineered Q1371 side chain came from a recent study which compared non-hydrolytic closing rates of CFTR orthologs (Simon and Csanády, 2023). Intriguingly, whereas for hCFTR τ_b is >10-fold longer for the E-to-Q compared to the E-to-S mutant (Figure 2A, B, black vs. dark blue), for the zebrafish ortholog τ_b of the E-to-S and E-to-Q mutants is similar (Figure 2A, B, brown vs. light blue). Careful comparison of the dimer interface structures of OF hCFTR and zCFTR suggests the presence of a non-native H-bond in the human structure which forms between the side-chain amide nitrogen of hQ1371 and the backbone carbonyl group of hG576, located at a distance of ~3 Å in the opposing D-loop of NBD1 (Figure 2E). Such a bond cannot form in WT hCFTR, as the native glutamate side chain cannot act as a H-donor. Mutant cycle analysis indeed confirmed strong energetic coupling between positions 1371 and 576 in hCFTR/E1371Q throughout the bursting state (Figure 3A–C), maintained even during flickery closures (Figure 3D–F). Of note, based on our functional results per se it would remain possible that hQ1371 interacts with some other NBD1 D-loop residue. However, the structure (whose resolution in this region is particularly high) does not support any alternative interaction: the distance between all other D-loop functional groups and either polar group of the hQ1371 side chain is >4.2 Å. Interestingly, formation of that H-bond in the E-to-Q mutant seems a unique feature of hCFTR, since the analogous positions are further apart in OF structures of both zCFTR (4.3 Å) and of other ABCC proteins (e.g., 4.7 Å in bovine MRP1, PDBID: 6bhu).

In several ABC proteins Walker A K-to-A and Walker B D-to-N mutations were shown to decrease the rate of ATP hydrolysis below the limit of detection (Ramjeesingh et al., 1999; Urbatsch et al., 1998; Hrycyna et al., 1999). However, the sensitivities of such enzymatic assays are limited, and for hCFTR ATPase activity of the Walker B mutant D1370N has not been directly measured. Thus, for hCFTR the >10-fold faster closing rate of hK1250A and ~200-fold faster closing rate of hD1370N compared to hE1371Q raised the possibility of residual ATP hydrolysis in the Walker mutants. Here, we evaluated that possibility using combinations of catalytic site mutations. We found that mutation hE1371S does not significantly slow closing rate of either Walker mutant (Figure 5B, C, Figure 5D, E), confirming that all three mutations (hK1250A, hD1370N, and hE1371S) individually abolish ATP hydrolysis. Furthermore, the mutual lack of effect of mutation hE1371S on τ_b of hK1250A (Figure 5B, C, blue vs. gray) and of mutation hK1250A on τ_b of hE1371S CFTR (Figure 5B, C, blue vs. Figure 3A, B, blue) suggests that neither of those two mutations significantly alters the stability of the B₁ state. Thus, τ_b of hK1250A or hE1371S CFTR can be taken as relatively unbiased estimates of the life time of the B₁ state in WT hCFTR, suggesting a true k₋₁ of ~0.03 s⁻¹ (Figure 6, red). Consistent with that notion, mutation hE1371S also failed to significantly alter τ_b of another non-hydrolytic background construct, hD1370N CFTR (Figure 5D, E, blue vs. gray).

Figure 6

Download asset Open asset

In contrast, mutation hD1370N dramatically shortened τ_b of both hE1371S (by ~14-fold, Figure 5D, E, blue vs. Figure 3A, B, blue) and hE1371Q (by ~15-fold, Figure 5D, E, black vs. Figure 3A, B, black) CFTR, suggesting that the shorter τ_b of the hD1370N single mutant compared to other non-hydrolytic mutants is explained by a large destabilizing effect of the hD1370N mutation on B₁-state stability, as opposed to substantial residual ATP-ase activity. That conclusion is consistent with the monotonically decaying distribution of hD1370N burst durations, interpreted to reflect (near-) equilibrium gating of the mutant (Csanády et al., 2010).

Prompted by the discovery of the artificial H-bond in the hE1371Q mutant we sought to dissect its functional effects on B₁-state stability. Mutation hE1371Q indeed greatly increases τ_b when introduced into any other non-hydrolytic background: by ~13-fold in the hE1371S background (Figure 3A, B, black vs. blue), by ~16-fold in the hD1370N background (Figure 5D, E, black vs. gray), but only by ~5-fold in the hK1250A background (Figure 5B, C, black vs. gray). A possible explanation for the latter smaller effect could be that the K1250A mutation causes a slight increase in the hG576-hQ1371 distance, resulting in a small (~0.6 kT) but significant (p = 0.003) decrease in the strength of the non-native H-bond.

The observed shortening or lengthening of non-hydrolytic τ_b by different site 2 mutations is in line with similarly bidirectional effects of different mutations at the site 1 interface documented in earlier studies. Indeed, whereas non-hydrolytic τ_b is decreased ~5- to 10-fold by NBD1 Walker A mutation hK464A (Powe et al., 2002; Vergani et al., 2003; Csanády et al., 2010; Csanády et al., 2013), it is increased ~3-fold by NBD2 signature sequence mutation hH1348A (Szollosi et al., 2011; Csanády et al., 2013) or by N⁶-(2-phenylethyl)-ATP (P-ATP) bound in site 1 (Tsai et al., 2010; Csanády et al., 2013), consistent with the notion that even site 1 undergoes substantial conformational changes coincident with pore opening/closure (Csanády et al., 2013; Mihályi et al., 2016; Sorum et al., 2017; cf., cartoon in Figure 1C). Along those lines, it may seem possible that under conditions when ATP hydrolysis at site 2 is disrupted, pore closure might be initiated by separation of the NBD interface around site 1: if so, that would undermine the above interpretation of the lack of effect of the hK1250A mutation in an hE1371S background (and vice versa). However, our results directly demonstrate that separation around site 2 remains the prevalent pathway for pore closure even in the absence of ATP hydrolysis. If that were not the case then slowing of site 2 separation by the non-natural H bond would not be expected to prolong bursts. Specifically, if in hE1371S or hD1370N channels pore closure predominantly proceeded through dissociation of site 1 then further stabilization of site 2 by introduction of the E1371Q mutation would not be expected to prolong τ_b – in contrast to the ~13- to 16-fold prolongation observed in both backgrounds (Figure 3A, B, black vs. blue; Figure 5D, E, black vs. gray).

Earlier studies documented a clear pattern of larger effects on the rate of opening compared to non-hydrolytic closure (k_open vs. k₋₁, Figure 1C) for mutations introduced into opposing sides of site 2 positions 555 (Sorum et al., 2017) and 1246 (Sorum et al., 2015) in CFTR, suggesting that the inter-NBD contacts across the site 2 interface are already established in the transition state (T^‡) that separates states IB₁ and B₁. The lack of effect on k₋₁ of mutations hK1250A and hE1371S demonstrated here is in line with that suggestion. On the other hand, the ~15-fold increase and decrease in non-hydrolytic τ_b caused by mutations hE1371Q and hD1370N, respectively, clearly indicate that those two side chains (located near the center of the NBD dimer interface between sites 1 and 2; Figure 5—figure supplement 1A) undergo significant rearrangements even during the T^‡ ↔ B₁ gating step. Alignment of the cryo-EM structures of NBD2 in the absence of nucleotide (from unphosphorylated WT hCFTR apo structure 5uak), of NBD2-ATP without contact to NBD1 (from unphosphorylated WT ATP-bound hCFTR structure 8fzq), and of NBD2-ATP tightly dimerized with NBD1 (from phosphorylated ATP-bound hE1371Q structure 6msm) affords some speculation on the possible nature of those movements. For non-native hQ1371 that movement might involve flipping of the mutated side chain, prompted by the vicinity of the hG576 carbonyl oxygen established in state T^‡. Indeed, whereas in the dimerized NBD2 structure the mutant hQ1371 side chain is clearly resolved and points toward the NBD1 D-loop (Figure 2—figure supplement 1), in the ATP bound but de-dimerized NBD2 structure no clear density is observable for the native E1371 side chain, suggesting that it is flexible or perhaps adopts a different rotameric conformation. For the hD1370 side chain the same two structures suggest a substantial rearrangement in the NBD-dimerized state which is not simply due to side-chain flipping: position 1370 is at the boundary between the ‘head’ and ‘tail’ subdomains of NBD2 which in all ABC proteins rotate toward each other upon ATP binding (Karpowich et al., 2001). Comparison of all three structures shows that in hCFTR’s NBD2 that ‘subdomain closure’ is fully completed only in the NBD-dimerized state (Figure 5—figure supplement 1B), reminiscent to earlier findings on the maltose transporter (Orelle et al., 2010). That movement strengthens multiple interactions of the hD1370 side-chain carboxylate, by reducing its distance both to the Mg²⁺ ion (from ~5.0–3.7 Å) and to the hS1251 side chain (from ~4.1–3.0 Å) (Figure 5—figure supplement 1C). If that movement happened during step T^‡ ↔ B₁ then lack of those tightening interactions in hD1370N could explain selective destabilization of the B₁ state relative to T^‡ in the mutant.

In the past, assuming k₋₁ < 0.2 s⁻¹ and k₁ ~ 4 s⁻¹, a lower limit of ~0.95 was provided for the CR of prephosphorylated WT hCFTR gating at 25°C (Csanády et al., 2010). Our present estimate of k₋₁ ~ 0.03 s⁻¹ under the same conditions provides a more accurate value of CR ~0.99. Thus, only 1 out of a ~100 bursts is terminated by non-hydrolytic separation of site 2 (Figure 6). Furthermore, earlier estimates of temperature and PKA dependence of rates k₁ and k₋₁ allow some extrapolations to different experimental conditions. Based on the measured activation enthalpies (ΔH^‡_B1→B2 ~ 70 kJ/mol, ΔH^‡_B1→IB1 ~ 40 kJ/mol; Csanády et al., 2006), k₁ is ~3-fold faster (~12 s⁻¹) and k₋₁ ~ 1.9-fold faster (~0.06 s⁻¹) at 37°C, predicting a CR of ~0.995. On the other hand, the presence of PKA slows both rates by ~twofold (Csanády et al., 2010; Vergani et al., 2003), predicting little effect on the CR.

A strong native H-bond between extracellular loops 1 and 6 of hCFTR (between residues R117 and E1124) was shown to be present only in the open state, but disrupted during both IB and flickery closures (Simon and Csanády, 2021). In contrast, we show here that the non-native hQ1371–hG576 H-bond is maintained during flickery closures, and disrupted only during IB closures (Figure 3). These findings are consistent with the notion that flickery closures of hE1371Q CFTR involve local movements confined to the external ends of the TMD helices but no disruption of the tight NBD dimer, whereas IB closures represent a global rearrangement during which site 2 of the NBD dimer opens up to allow nucleotide exchange. The precise extent of that separation is presently unknown: it is large enough to disrupt the site 2 interfacial H-bonds (Figure 5—figure supplement 1A) formed between the side chains of hT1246 and hR555 in WT (or K1250R) hCFTR (Vergani et al., 2005) and between residues hQ1371 and hG576 in E1371Q hCFTR (Figures 2E and 3A–C), but too small to be detected by a pair of FRET sensors engineered into position h388 of NBD1 and position h1435 of NBD2 (Levring et al., 2023). Precisely gauging the extent of those motions will require a cryo-EM structure of the active IB state of phosphorylated hCFTR in the presence of ATP.

All data generated or analyzed during this study are included in the manuscript and the figures.

1. 1. Aleksandrov L
   2. Aleksandrov AA
   3. Chang XB
   4. Riordan JR

   (2002) THe first nucleotide binding domain of cystic fibrosis transmembrane conductance regulator is a site of stable nucleotide interaction, whereas the second is a site of rapid turnover

   The Journal of Biological Chemistry 277:15419–15425.

   https://doi.org/10.1074/jbc.M111713200

   • PubMed
   • Google Scholar
2. 1. Basso C
   2. Vergani P
   3. Nairn AC
   4. Gadsby DC

   (2003) Prolonged nonhydrolytic interaction of nucleotide with CFTR’s NH2-terminal nucleotide binding domain and its role in channel gating

   The Journal of General Physiology 122:333–348.

   https://doi.org/10.1085/jgp.200308798

   • PubMed
   • Google Scholar
3. 1. Bompadre SG
   2. Cho JH
   3. Wang X
   4. Zou X
   5. Sohma Y
   6. Li M
   7. Hwang TC

   (2005) CFTR gating II: Effects of nucleotide binding on the stability of open states

   The Journal of General Physiology 125:377–394.

   https://doi.org/10.1085/jgp.200409228

   • PubMed
   • Google Scholar
4. 1. Cai Z
   2. Scott-Ward TS
   3. Sheppard DN

   (2003) Voltage-dependent gating of the cystic fibrosis transmembrane conductance regulator Cl- channel

   The Journal of General Physiology 122:605–620.

   https://doi.org/10.1085/jgp.200308921

   • PubMed
   • Google Scholar
5. 1. Carson MR
   2. Travis SM
   3. Welsh MJ

   (1995) The two nucleotide-binding domains of cystic fibrosis transmembrane conductance regulator (CFTR) have distinct functions in controlling channel activity

   The Journal of Biological Chemistry 270:1711–1717.

   https://doi.org/10.1074/jbc.270.4.1711

   • PubMed
   • Google Scholar
6. 1. Cheng SH
   2. Rich DP
   3. Marshall J
   4. Gregory RJ
   5. Welsh MJ
   6. Smith AE

   (1991) Phosphorylation of the R domain by cAMP-dependent protein kinase regulates the CFTR chloride channel

   Cell 66:1027–1036.

   https://doi.org/10.1016/0092-8674(91)90446-6

   • PubMed
   • Google Scholar
7. 1. Csanády L
   2. Nairn AC
   3. Gadsby DC

   (2006) Thermodynamics of CFTR channel gating: a spreading conformational change initiates an irreversible gating cycle

   The Journal of General Physiology 128:523–533.

   https://doi.org/10.1085/jgp.200609558

   • PubMed
   • Google Scholar
8. 1. Csanády L
   2. Vergani P
   3. Gadsby DC

   (2010) Strict coupling between CFTR’s catalytic cycle and gating of its Cl- ion pore revealed by distributions of open channel burst durations

   PNAS 107:1241–1246.

   https://doi.org/10.1073/pnas.0911061107

   • PubMed
   • Google Scholar
9. 1. Csanády L
   2. Mihályi C
   3. Szollosi A
   4. Töröcsik B
   5. Vergani P

   (2013) Conformational changes in the catalytically inactive nucleotide-binding site of CFTR

   The Journal of General Physiology 142:61–73.

   https://doi.org/10.1085/jgp.201210954

   • PubMed
   • Google Scholar
10. 1. Csanády L
    2. Töröcsik B

    (2014) Catalyst-like modulation of transition states for CFTR channel opening and closing: new stimulation strategy exploits nonequilibrium gating

    The Journal of General Physiology 143:269–287.

    https://doi.org/10.1085/jgp.201311089

    • PubMed
    • Google Scholar
11. 1. Csanády L
    2. Vergani P
    3. Gadsby DC

    (2019) Structure, gating, and regulation of the cftr anion channel

    Physiological Reviews 99:707–738.

    https://doi.org/10.1152/physrev.00007.2018

    • PubMed
    • Google Scholar
12. 1. Dean M
    2. Moitra K
    3. Allikmets R

    (2022) The human ATP-binding cassette (ABC) transporter superfamily

    Human Mutation 43:1162–1182.

    https://doi.org/10.1002/humu.24418

    • PubMed
    • Google Scholar
13. 1. Gunderson KL
    2. Kopito RR

    (1995) Conformational states of CFTR associated with channel gating: the role ATP binding and hydrolysis

    Cell 82:231–239.

    https://doi.org/10.1016/0092-8674(95)90310-0

    • PubMed
    • Google Scholar
14. 1. Hohl M
    2. Briand C
    3. Grütter MG
    4. Seeger MA

    (2012) Crystal structure of a heterodimeric ABC transporter in its inward-facing conformation

    Nature Structural & Molecular Biology 19:395–402.

    https://doi.org/10.1038/nsmb.2267

    • PubMed
    • Google Scholar
15. 1. Hohl M
    2. Hürlimann LM
    3. Böhm S
    4. Schöppe J
    5. Grütter MG
    6. Bordignon E
    7. Seeger MA

    (2014) Structural basis for allosteric cross-talk between the asymmetric nucleotide binding sites of a heterodimeric ABC exporter

    PNAS 111:11025–11030.

    https://doi.org/10.1073/pnas.1400485111

    • PubMed
    • Google Scholar
16. 1. Hrycyna CA
    2. Ramachandra M
    3. Germann UA
    4. Cheng PW
    5. Pastan I
    6. Gottesman MM

    (1999) Both ATP sites of human P-glycoprotein are essential but not symmetric

    Biochemistry 38:13887–13899.

    https://doi.org/10.1021/bi991115m

    • PubMed
    • Google Scholar
17. 1. Huang Y
    2. Xue C
    3. Wang L
    4. Bu R
    5. Mu J
    6. Wang Y
    7. Liu Z

    (2023) Structural basis for substrate and inhibitor recognition of human multidrug transporter MRP4

    Communications Biology 6:549.

    https://doi.org/10.1038/s42003-023-04935-7

    • PubMed
    • Google Scholar
18. 1. Janas E
    2. Hofacker M
    3. Chen M
    4. Gompf S
    5. van der Does C
    6. Tampé R

    (2003) The ATP hydrolysis cycle of the nucleotide-binding domain of the mitochondrial ATP-binding cassette transporter Mdl1p

    The Journal of Biological Chemistry 278:26862–26869.

    https://doi.org/10.1074/jbc.M301227200

    • PubMed
    • Google Scholar
19. 1. Johnson ZL
    2. Chen J

    (2017) Structural basis of substrate recognition by the multidrug resistance protein MRP1

    Cell 168:1075–1085.

    https://doi.org/10.1016/j.cell.2017.01.041

    • PubMed
    • Google Scholar
20. 1. Johnson ZL
    2. Chen J

    (2018) ATP Binding enables substrate release from multidrug resistance protein 1

    Cell 172:81–89.

    https://doi.org/10.1016/j.cell.2017.12.005

    • PubMed
    • Google Scholar
21. 1. Kaczmarek LK
    2. Jennings KR
    3. Strumwasser F
    4. Nairn AC
    5. Walter U
    6. Wilson FD
    7. Greengard P

    (1980) Microinjection of catalytic subunit of cyclic AMP-dependent protein kinase enhances calcium action potentials of bag cell neurons in cell culture

    PNAS 77:7487–7491.

    https://doi.org/10.1073/pnas.77.12.7487

    • PubMed
    • Google Scholar
22. 1. Karpowich N
    2. Martsinkevich O
    3. Millen L
    4. Yuan YR
    5. Dai PL
    6. MacVey K
    7. Thomas PJ
    8. Hunt JF

    (2001) Crystal structures of the MJ1267 ATP binding cassette reveal an induced-fit effect at the ATPase active site of an ABC transporter

    Structure 9:571–586.

    https://doi.org/10.1016/s0969-2126(01)00617-7

    • PubMed
    • Google Scholar
23. 1. Kim Y
    2. Chen J

    (2018) Molecular structure of human P-glycoprotein in the ATP-bound, outward-facing conformation

    Science 359:915–919.

    https://doi.org/10.1126/science.aar7389

    • PubMed
    • Google Scholar
24. 1. Levring J
    2. Terry DS
    3. Kilic Z
    4. Fitzgerald G
    5. Blanchard SC
    6. Chen J

    (2023) Author Correction: CFTR function, pathology and pharmacology at single-molecule resolution

    Nature 617:E11.

    https://doi.org/10.1038/s41586-023-06115-3

    • PubMed
    • Google Scholar
25. 1. Li C
    2. Ramjeesingh M
    3. Wang W
    4. Garami E
    5. Hewryk M
    6. Lee D
    7. Rommens JM
    8. Galley K
    9. Bear CE

    (1996) ATPase activity of the cystic fibrosis transmembrane conductance regulator

    The Journal of Biological Chemistry 271:28463–28468.

    https://doi.org/10.1074/jbc.271.45.28463

    • PubMed
    • Google Scholar
26. 1. Liu F
    2. Zhang Z
    3. Csanády L
    4. Gadsby DC
    5. Chen J

    (2017) Molecular structure of the Human CFTR Ion Channel

    Cell 169:85–95.

    https://doi.org/10.1016/j.cell.2017.02.024

    • PubMed
    • Google Scholar
27. 1. Locher KP

    (2016) Mechanistic diversity in ATP-binding cassette (ABC) transporters

    Nature Structural & Molecular Biology 23:487–493.

    https://doi.org/10.1038/nsmb.3216

    • PubMed
    • Google Scholar
28. 1. Mihályi C
    2. Töröcsik B
    3. Csanády L

    (2016) Obligate coupling of CFTR pore opening to tight nucleotide-binding domain dimerization

    eLife 5:e18164.

    https://doi.org/10.7554/eLife.18164

    • PubMed
    • Google Scholar
29. 1. Moody JE
    2. Millen L
    3. Binns D
    4. Hunt JF
    5. Thomas PJ

    (2002) Cooperative, ATP-dependent association of the nucleotide binding cassettes during the catalytic cycle of ATP-binding cassette transporters

    The Journal of Biological Chemistry 277:21111–21114.

    https://doi.org/10.1074/jbc.C200228200

    • PubMed
    • Google Scholar
30. 1. Olsen JA
    2. Alam A
    3. Kowal J
    4. Stieger B
    5. Locher KP

    (2020) Structure of the human lipid exporter ABCB4 in a lipid environment

    Nature Structural & Molecular Biology 27:62–70.

    https://doi.org/10.1038/s41594-019-0354-3

    • PubMed
    • Google Scholar
31. 1. Orelle C
    2. Alvarez FJD
    3. Oldham ML
    4. Orelle A
    5. Wiley TE
    6. Chen J
    7. Davidson AL

    (2010) Dynamics of alpha-helical subdomain rotation in the intact maltose ATP-binding cassette transporter

    PNAS 107:20293–20298.

    https://doi.org/10.1073/pnas.1006544107

    • PubMed
    • Google Scholar
32. 1. O’Sullivan BP
    2. Freedman SD

    (2009) Cystic fibrosis

    Lancet 373:1891–1904.

    https://doi.org/10.1016/S0140-6736(09)60327-5

    • PubMed
    • Google Scholar
33. 1. Powe AC
    2. Al-Nakkash L
    3. Li M
    4. Hwang TC

    (2002) Mutation of Walker-A lysine 464 in cystic fibrosis transmembrane conductance regulator reveals functional interaction between its nucleotide-binding domains

    The Journal of Physiology 539:333–346.

    https://doi.org/10.1113/jphysiol.2001.013162

    • PubMed
    • Google Scholar
34. 1. Procko E
    2. O’Mara ML
    3. Bennett WFD
    4. Tieleman DP
    5. Gaudet R

    (2009) The mechanism of ABC transporters: general lessons from structural and functional studies of an antigenic peptide transporter

    FASEB Journal 23:1287–1302.

    https://doi.org/10.1096/fj.08-121855

    • PubMed
    • Google Scholar
35. 1. Rai V
    2. Gaur M
    3. Shukla S
    4. Shukla S
    5. Ambudkar SV
    6. Komath SS
    7. Prasad R

    (2006) Conserved Asp327 of walker B motif in the N-terminal nucleotide binding domain (NBD-1) of Cdr1p of Candida albicans has acquired a new role in ATP hydrolysis

    Biochemistry 45:14726–14739.

    https://doi.org/10.1021/bi061535t

    • PubMed
    • Google Scholar
36. 1. Ramjeesingh M
    2. Li C
    3. Garami E
    4. Huan LJ
    5. Galley K
    6. Wang Y
    7. Bear CE

    (1999) Walker mutations reveal loose relationship between catalytic and channel-gating activities of purified CFTR (cystic fibrosis transmembrane conductance regulator)

    Biochemistry 38:1463–1468.

    https://doi.org/10.1021/bi982243y

    • PubMed
    • Google Scholar
37. 1. Riordan JR
    2. Rommens JM
    3. Kerem B-S
    4. Alon N
    5. Rozmahel R
    6. Grzelczak Z
    7. Zielenski J
    8. Lok S
    9. Plavsic N
    10. Chou J-L
    11. Drumm ML
    12. Iannuzzi MC
    13. Collins FS
    14. Tsui L-C

    (1989) Identification of the Cystic Fibrosis Gene: cloning and characterization of complementary DNA

    Science 245:1066–1073.

    https://doi.org/10.1126/science.2475911

    • Google Scholar
38. 1. Simon MA
    2. Csanády L

    (2021) Molecular pathology of the R117H cystic fibrosis mutation is explained by loss of a hydrogen bond

    eLife 10:e74693.

    https://doi.org/10.7554/eLife.74693

    • PubMed
    • Google Scholar
39. 1. Simon MA
    2. Csanády L

    (2023) Optimization of CFTR gating through the evolution of its extracellular loops

    The Journal of General Physiology 155:e202213264.

    https://doi.org/10.1085/jgp.202213264

    • PubMed
    • Google Scholar
40. 1. Smith PC
    2. Karpowich N
    3. Millen L
    4. Moody JE
    5. Rosen J
    6. Thomas PJ
    7. Hunt JF

    (2002) ATP binding to the motor domain from an ABC transporter drives formation of a nucleotide sandwich dimer

    Molecular Cell 10:139–149.

    https://doi.org/10.1016/s1097-2765(02)00576-2

    • PubMed
    • Google Scholar
41. 1. Song G
    2. Zhang S
    3. Tian M
    4. Zhang L
    5. Guo R
    6. Zhuo W
    7. Yang M

    (2021) Molecular insights into the human ABCB6 transporter

    Cell Discovery 7:55.

    https://doi.org/10.1038/s41421-021-00284-z

    • PubMed
    • Google Scholar
42. 1. Sorum B
    2. Czégé D
    3. Csanády L

    (2015) Timing of CFTR pore opening and structure of its transition state

    Cell 163:724–733.

    https://doi.org/10.1016/j.cell.2015.09.052

    • PubMed
    • Google Scholar
43. 1. Sorum B
    2. Töröcsik B
    3. Csanády L

    (2017) Asymmetry of movements in CFTR’s two ATP sites during pore opening serves their distinct functions

    eLife 6:e29013.

    https://doi.org/10.7554/eLife.29013

    • PubMed
    • Google Scholar
44. 1. Szollosi A
    2. Muallem DR
    3. Csanády L
    4. Vergani P

    (2011) Mutant cycles at CFTR’s non-canonical ATP-binding site support little interface separation during gating

    The Journal of General Physiology 137:549–562.

    https://doi.org/10.1085/jgp.201110608

    • PubMed
    • Google Scholar
45. 1. Timachi MH
    2. Hutter CA
    3. Hohl M
    4. Assafa T
    5. Böhm S
    6. Mittal A
    7. Seeger MA
    8. Bordignon E

    (2017) Exploring conformational equilibria of a heterodimeric ABC transporter

    eLife 6:e20236.

    https://doi.org/10.7554/eLife.20236

    • PubMed
    • Google Scholar
46. 1. Tsai MF
    2. Li M
    3. Hwang TC

    (2010) Stable ATP binding mediated by a partial NBD dimer of the CFTR chloride channel

    The Journal of General Physiology 135:399–414.

    https://doi.org/10.1085/jgp.201010399

    • PubMed
    • Google Scholar
47. 1. Urbatsch IL
    2. Beaudet L
    3. Carrier I
    4. Gros P

    (1998) Mutations in either Nucleotide-Binding site of P-glycoprotein (Mdr3) prevent vanadate trapping of Nucleotide at both sites

    Biochemistry 37:4592–4602.

    https://doi.org/10.1021/bi9728001

    • Google Scholar
48. 1. Vergani P
    2. Nairn AC
    3. Gadsby DC

    (2003) On the mechanism of MgATP-dependent gating of CFTR Cl- channels

    The Journal of General Physiology 121:17–36.

    https://doi.org/10.1085/jgp.20028673

    • PubMed
    • Google Scholar
49. 1. Vergani P
    2. Lockless SW
    3. Nairn AC
    4. Gadsby DC

    (2005) CFTR channel opening by ATP-driven tight dimerization of its nucleotide-binding domains

    Nature 433:876–880.

    https://doi.org/10.1038/nature03313

    • PubMed
    • Google Scholar
50. 1. Winter MC
    2. Sheppard DN
    3. Carson MR
    4. Welsh MJ

    (1994) Effect of ATP concentration on CFTR Cl- channels: a kinetic analysis of channel regulation

    Biophysical Journal 66:1398–1403.

    https://doi.org/10.1016/S0006-3495(94)80930-0

    • Google Scholar
51. 1. Yeh HI
    2. Yeh JT
    3. Hwang TC

    (2015) Modulation of CFTR gating by permeant ions

    The Journal of General Physiology 145:47–60.

    https://doi.org/10.1085/jgp.201411272

    • PubMed
    • Google Scholar
52. 1. Yu YC
    2. Sohma Y
    3. Hwang TC

    (2016) On the mechanism of gating defects caused by the R117H mutation in cystic fibrosis transmembrane conductance regulator

    The Journal of Physiology 594:3227–3244.

    https://doi.org/10.1113/JP271723

    • PubMed
    • Google Scholar
53. 1. Zeltwanger S
    2. Wang F
    3. Wang GT
    4. Gillis KD
    5. Hwang TC

    (1999) Gating of cystic fibrosis transmembrane conductance regulator chloride channels by adenosine triphosphate hydrolysis: quantitative analysis of a cyclic gating scheme

    The Journal of General Physiology 113:541–554.

    https://doi.org/10.1085/jgp.113.4.541

    • PubMed
    • Google Scholar
54. 1. Zhang Z
    2. Liu F
    3. Chen J

    (2017) Conformational Changes of CFTR upon Phosphorylation and ATP Binding

    Cell 170:483–491.

    https://doi.org/10.1016/j.cell.2017.06.041

    • PubMed
    • Google Scholar
55. 1. Zhang Z
    2. Liu F
    3. Chen J

    (2018) Molecular structure of the ATP-bound, phosphorylated human CFTR

    PNAS 115:12757–12762.

    https://doi.org/10.1073/pnas.1815287115

    • PubMed
    • Google Scholar
56. 1. Zhou Z
    2. Wang X
    3. Li M
    4. Sohma Y
    5. Zou X
    6. Hwang TC

    (2005) High affinity ATP/ADP analogues as new tools for studying CFTR gating

    The Journal of Physiology 569:447–457.

    https://doi.org/10.1113/jphysiol.2005.095083

    • PubMed
    • Google Scholar
57. 1. Zhou Z
    2. Wang X
    3. Liu HY
    4. Zou X
    5. Li M
    6. Hwang TC

    (2006) The two ATP binding sites of cystic fibrosis transmembrane conductance regulator (CFTR) play distinct roles in gating kinetics and energetics

    The Journal of General Physiology 128:413–422.

    https://doi.org/10.1085/jgp.200609622

    • PubMed
    • Google Scholar

Decision letter after peer review:

Thank you for submitting your article "Estimating the true stability of the prehydrolytic outward-facing state in an ABC protein" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, including Andres Jara-Oseguera as Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Merritt Maduke as the Senior Editor.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission. All three reviewers agreed that the work presented is of high quality, and that the findings are relevant and interesting. However, reviewers also considered that some of the assumptions made throughout the manuscript and the limitations of the data should be more clearly acknowledged and discussed. Also, strong support for the stabilizing effect of the non-native hydrogen bond described requires some additional experimental work. These points that need to be addressed are described in the list of Essential Revisions below.

Essential revisions:

1) The authors hypothesized that in hCFTR an artificial H-bond between the side-chain of glutamine at position 1371 (i.e., in E1371Q mutant) and the backbone carbonyl at G576 of the D-loop stabilizes the NBD dimer. Such H-bond is absent in E1372Q zCFTR. The authors employed mutant cycle analysis on the G576Δ-E1371S mutation pair to demonstrate an energetic coupling between the hG576 and hE1371Q. However, how deletion of G576 might alter the local structure is unpredictable. The result does not directly address the discrepancy between zCFTR and hCFTR, either. The D-loop is highly conserved across species with a consensus sequence PFGYLD (residue 574-579 in hCFTR), but in zCFTR the analogous sequence is PFTHLD. The backbone carbonyl oxygen could therefore be harder to access in zCFTR.

Below we provide three alternative suggestions to provide additional support for the interpretation of the discrepancies between human and zebrafish (and other ABC transporters) regarding the presence of a stabilizing H bond involving a catalytic glutamine and the D-loop. The authors can also decide to take a different strategy to provide one more piece of evidence for this point.

– Introducing mutation in the D-loop of zCFTR to match the sequence of hCFTR (and vice versa). The authors' hypothesis would predict that zCFTR with hCFTR's D-loop sequence should recapitulate hCFTR's phenotype: the E-to-Q mutation on the catalytic glutamate would further lengthen the burst duration compared to the E-to-S mutation.

– The mutant cycle could be further applied to the zebrafish, by measuring the closing rate of a zebrafish construct containing the zE1372Q mutation together with a deletion on the D-loop. If D-loop deletion does not give a significant burst shortening in the zE1372Q background, it would support the idea that the H-bond in the hE1371Q-CFTR is unique.

– Obtain analysis of molecular dynamics simulations that shows that there is a hydrogen bond in the human simulations and not the zebrafish. Several papers have been published in which extensive simulations were carried out.

doi.org/10.1007/s00018-019-03211-4

doi.org/10.1016/j.bpj.2018.03.003

doi.org/10.1007/s00018-022-04621-7

Most studies focused their analysis on the transmembrane domains, but results will also include information on the NBDs. It might be possible for Prof Csanády to reach out to some of these computational groups and ask them to investigate the possible presence and persistence of a hydrogen bond between the glutamine and the NBD1 D-loop, in 6MSM or in 5W81.

2) Structurally, hG576, is better aligned to zT575. These two residues are almost completely overlapping in the backbone. There is a clear difference between the two pdb coordinates in the glutamine conformation, with the amide and carbonyl group positions swapped (Figure 2E-F): hQ1371 points the amide and zQ1372 points the oxygen towards the D loop, such that the zQ1372 amide group ends up closer to the backbone oxygen of zH576. The zQ1372 side chain is also shifted (~0.1 nm) closer to the γ phosphate, away from site 1, and further from the switch histidine (zH1403). Finally, site 1 is more open in zCFTR.

However, there are some uncertainties in modelling atomic coordinates from electron density maps. Negatively charged moieties often do not show up in these maps, for unknown reasons. In the electron density map for hCFTR (Figure Suppl. 1), part of the glutamine side chain (apparently the amide group) is missing. The electron density for both the amide and the carbonyl group is missing in the map for 5W81. How confident are we that the different pdb coordinates (hQ1371 vs. zQ1372) reflect a true experimental difference, with the glutamines in the bursting states taking on a different ensemble of conformations, such that only the human CFTR is stabilized by the hydrogen bond? Would it perhaps be possible that some other interaction involving the glutamine and the D-loop is also involved? This needs to be discussed.

3) The authors speculated that the reason for D1370N's relatively fast closing rate compared to other non-hydrolytic mutants is the loss of interaction between Mg²⁺ and the negatively charged aspartate. However, this reasoning fails to explain why non-hydrolytic closure of wildtype CFTR in the absence of Mg²⁺ (e.g., Levring et al. 2023 Extended Data Figure 7g) is even slower than the non-hydrolytic closure of D1370N CFTR opened by MgATP, where at least the Mg²⁺ is present. The authors should caution the readers that so far no definitive experimental evidence can explain the destabilizing effect of D1370N.

4) Based on the results that the double mutant E1371S/K1250A hCFTR has a similar burst duration as single mutant E1371S and K1250A, the authors made a strong claim that both mutations completely abolish ATP hydrolysis. Similar reasoning was applied to D1370N. The limitations in such interpretations should be discussed. The authors made the assumption that the termination of a burst is solely controlled by site 2 (Figure 1C). However, when hydrolysis is significantly diminished, binding of ATP in site 2 is very stable, and thus dissociation of ATP from site 2 versus site 1 becomes hard to distinguish. Whether all hydrolysis-deficient mutants share the same open-to-close transition by releasing ATP from site 2 but retaining ATP in site 1 is still a question. As the authors have elaborated in the text, it is known that mutations in the degenerate site 1 can affect non-hydrolytic closing. When mutations are introduced to site 2, they might as well result in allosteric effects on the stability of ATP binding in site 1, which could subsequently alter the channel's closing rate. The authors might want to make the readers aware of the complicated relationship between channel closure and CFTR's two ATP binding sites.

5) It is known that non-hydrolytic closing rate of CFTR is phosphorylation dependent, which the authors briefly mentioned in the Discussion. Vergani et al. (2003) documented that τburst of K1250A and D1370N in PKA is ~80 s and ~4 s respectively, but both are reduced by roughly twofold when PKA was removed. In this study the burst durations of K1250A (~30 s, Figure 4C) and D1370N (~2 s, Figure 4E) indicate that these channels are not strongly phosphorylated. Similarly, the τburst of E1371S in PKA is over 100 s (Bompadre et al. 2005), significantly longer than that in the current study. Although it is unclear how a different degree of R domain phosphorylation affects non-hydrolytic closing, the fact that it does again suggests that the simplified scheme used as the base for data interpretation may have its limitation. The Discussion would benefit from a more cautionary note on the oversimplification of the IB1↔B1 transition, and clarify that channels are not strongly phosphorylated in the current experimental condition.

6) The τburst of E1371Q CFTR is over 400 second while the τburst of K1250A-E1371Q double mutant is shortened to ~200 second (Figure 3B, black vs Figure 4C, black). The K1250A-E1371S CFTR also seems to have a shorter τburst than E1371S CFTR (Figure 4C, blue vs Figure 3B, blue). Although the effect of the K1250A mutation on shortening τburst of E1371Q and E1371S CFTR is not as dramatic as the D1370N mutation, the authors might want to clearly state if there is indeed a significant difference and address how K1250A mutation has such destabilizing effect, and provide additional statistical analysis.

7) page 10 line 214: "If both single mutations failed to completely abolish k1 then a further slowing in k1, i.e., a further prolongation of taub, would be expected in the double mutant K1250A/E1371S."

This prediction assumes that the effects of the mutations on hydrolysis are independent. One could envisage scenarios in which approximately the same residual hydrolytic activity would remain in K1250A, E1371S and K1250A/E1371S, if the two mutations worked in a similar way to slow hydrolysis down. Admittedly this is extremely unlikely, as we know the Walker A lysine and Walker B glutamate play different roles in catalysis, as the authors explain just above. But since this is a crucial point in reaching one of the main conclusions of the paper, it might be worth developing the argument of the redundancy of the interaction more extensively.

8) It is unclear if the rates for different mutations were measured at -20 or -80 mV. This should be described, and if different voltages were used, a discussion should be provided for how the different voltages are accounted for in the energetic calculations from the rates.

9) A better description should be provided for how the decay currents were normalized.

10) Page 9, line 174 and following are very hard to follow. Possible modification: "The coupling energy (DDGint(B1→T‡); Figure 3C, purple) is a measure of the discrepancy in the effects caused by perturbation of one target position depending on whether the other target position is mutated or intact (Figure 3C, parallel arrows). DDGint(B1→T‡) is thus given by the difference between DDG 0T++ -B1 values on parallel sides of the cycle (Figure 3C, numbers on parallel arrows). DDGint(B1→T‡) thus quantifies the change in the strength of the G576-Q1371 interaction in the E1371Q background construct while the channel proceeds from the B1 state to state T‡."

The latter statement is true only if simplifying assumptions are met. Among these, it is assumed that in each single mutant and in the double mutant the interaction between the target sites is completely abolished. This assumption is, strictly speaking, met in the deletion mutants. However, the serine could well form new contacts with carbonyl oxygens in the D-loop that depend on the glycine deletion. The statement should be qualified and/or the assumptions should be stated.

Non-essential revisions suggested by reviewers:

1) Including the traces for single mutants E1371Q and E1371S on Figure 4B-E would be helpful.

2) The different structures in Figure 4 Suppl. 1 are hard to distinguish, so it would be helpful to use more contrasting colors. It would also be valuable to include the loop for G576 for both structures on panel B.

3) Line 300-306: It would be helpful to provide a better clarification for the meaning of "bidirectional effects" – Does "bidirectional" here just mean both slowing down (as H1348A or P-ATP; E1371Q) or speeding up (as K464A; D1370N on k-1) of k-1, non-hydrolytic closure?

Also, the authors mentioned that non-hydrolytic closure is accelerated by K464A but slowed by H1348A and P-ATP, but it is unclear how perturbations in site 1 are related to the current scheme in this study, in which the IB1↔B1 transition considers only the separation of site 2. Some elaboration might be needed.

4) Figure 3D, E1371S: the scale for the traces appears different than from the other mutants. Is this due to increased overall noise in that recording?

5) The authors use IF, inward-facing, and OF, outward-facing, to refer to states in the kinetic scheme likely to adopt these conformations. Page 2, line 51 "The IF conformation, in which the pore is sealed near its extracellular end (Figure 1C, left), is a long lived state (~1 s) and corresponds to long "interburst" (IB) closed dwell times observable in single channel current recordings." I think it might be better to clearly separate structures and states at the start.

Possible alteration: "CFTR likely adopts an IF conformation, in which the pore is sealed near its extracellular end, during the long lived (~1 s) state (Figure 1C, left) which corresponds to long "interburst" (IB) closed dwell times observable in single channel current recordings. We refer to these closed states as "IF states"".

Then later: "When CFTR adopts the OF conformation the external gate is predominantly open, and a lateral portal which connects the channel pore with the cytosol generates a transmembrane aqueous pathway permeable to anions ((Zhang et al., 2018); Figure 1C, right, double arrows). However, functional studies show that the continuity of the transmembrane pore is occasionally disrupted for brief (~10 ms) intervals by a smaller conformational change (not depicted in Figure 1C) likely confined to the external ends of the TMD helices (Zhang et al., 2017; Zhang et al., 2018; Simon and Csanády, 2021).

Correspondingly, in single-channel recordings, this "OF" or "bursting" (B) state corresponds to clusters of channel openings separated by brief ("flickery") closures (Winter et al., 1994). The OF state is also relatively stable, with a dwell time of hundreds of milliseconds."

6) Page 4, line 80: the kinetic scheme used to describe CFTR (Figure 1) is used to illustrate ATP transporter ATPase cycles. Because the scheme has emerged from studies on CFTR gating, and not ABC exporter transport cycles (although the cycle might be similar), I would not refer to specific rates in the figure when introducing exporters. Possible alteration:

"The "coupling ratio" (CR), i.e., the fraction of initiated cycles that are completed through ATP hydrolysis, depends on the relative rates of the two possible exit pathways from the prehydrolytic state. In most ABC transporters the actual values of these rates are hard to directly estimate, but in CFTR direct measurements of conformational dwell times are made feasible by single-channel current recordings, providing estimates for microscopic transition rates. Thus, for a wild-type (WT) channel, we can estimate ATP hydrolysis rate (k1, Figure 1C) vs. non-hydrolytic NBD dimer dissociation rate (k-1, Figure 1C). Because typically k1>>k-1, CR=k1/(k1+k-1) is near unity."

7) Page 14, line 281 (and page 10 line 217): "the mutual lack of effect of mutation E1371S on taub of K1250A (Figure 4B-C, blue vs. gray) and of mutation K1250A on taub of E1371S CFTR" Why is introducing E1371S in K1250A described separately from introducing K1250A in E1371S? Since E1371S, K1250A, and K1250A/E1371S all have a similar taub it seems redundant to discuss the sequence in which the mutations are introduced. Why do the authors feel it is useful to present the results this way?

Reviewer #1 (Recommendations for the authors):

1. It is unclear if the rates for different mutations were measured at -20 or -80 mV. This should be described, and if different voltages were used, a discussion should be provided for how the different voltages are accounted for in the energetic calculations from the rates.

2. A better description should be provided for how the decay currents were normalized.

3. There is a confusing sentence on page 10, lines 217-218: "However, introducing the E1371S into the K1250A background did not significantly alter taub […]. Similarly, the introduction of mutation K1250A into the E1371S background failed to prolong taub." The two sentences say the same thing, because they refer to the same double mutant construct?

4. I suggest including the traces for single mutants E1371Q and E1371S in Figure 4B-E.

5. I suggest coloring the different structures in Figure 4 Suppl. 1 using more different colors – it is harder to see the differences with all shown in green. I would also suggest including the loop for G576 for both structures on panel B.

6. Although not necessary for publication, it would have been interesting and provided even stronger support for conclusions if the authors had performed the double mutant cycle analysis involving E1371Q and G576Δ in the background of one of the other non-hydrolytic mutations that have closing rates like WT, and potentially also the zebrafish channel, to show that the deletion of the residue at G576 has no effect on closure rate.

Reviewer #2 (Recommendations for the authors):

1. Line 300-306: What does "the bidirectional effects" mean for the studies on K464A, H1348A and P-ATP? The authors mentioned that non-hydrolytic closure is accelerated by K464A but slowed by H1348A and P-ATP, but it is unclear how perturbations in site 1 are related to the current scheme in this study, in which the IB1↔B1 transition considers only the separation of site 2. Some elaboration might be needed.

2. Figure 3D, E1371S: The scale is likely wrong. The single-channel amplitude, open-channel noise, and baseline noise are larger than the other three recordings. Please double check.

3. Figure 4-Figure Suppl. 1C: The color gradient of the three conformations could be increased to improve clarity. Currently the distinction between the three is somewhat hard to discern.

Reviewer #3 (Recommendations for the authors):

A. Suggestions for possible additional analysis.

While the electrophysiological study of the E to Q mutants is quite convincing (human and zebrafish E to Q mutants are clearly different, and there is some coupling with D-loop residues in hCFTR), I have some uncertainty regarding the interpretation offered.

Structurally, hG576, is better aligned to zT575. These two residues are almost completely overlapping in the backbone. There is a clear difference between the two pdb coordinates in the glutamine conformation, with the amide and carbonyl group positions swapped (Figure 2E-F): hQ1371 points the amide and zQ1372 points the oxygen towards the D loop, such that the zQ1372 amide group ends up closer to the backbone oxygen of zH576. The zQ1372 side chain is also shifted (~0.1 nm) closer to the γ phosphate, away from site 1, and further from the switch histidine (zH1403). Finally, site 1 is more open in zCFTR.

However, there are some uncertainties in modelling atomic coordinates from electron density maps. Negatively charged moieties often do not show up in these maps, for unknown reasons. In the electron density map for hCFTR (Figure Suppl. 1), part of the glutamine side chain (apparently the amide group) is missing. The electron density for both the amide and the carbonyl group are missing in the map for 5W81. How confident are we that the different pdb coordinates (hQ1371 vs. zQ1372) reflect a true experimental difference, with the glutamines in the bursting states taking on a different ensemble of conformations, such that only the human CFTR is stabilized by the hydrogen bond? Would it perhaps be possible that some other interaction involving the glutamine and the D-loop is also involved?

To strengthen the hypothesis of the different hydrogen bonding pattern the authors propose, one possible option is to look at molecular dynamics simulations. Several papers have been published in which extensive simulations were carried out.

doi.org/10.1007/s00018-019-03211-4

doi.org/10.1016/j.bpj.2018.03.003

doi.org/10.1007/s00018-022-04621-7

Most studies focused their analysis on the transmembrane domains, but results will also include information on the NBDs. It might be possible for Prof Csanády to reach out to some of these computational groups and ask them to investigate the possible presence and persistence of a hydrogen bond between the glutamine and the NBD1 D-loop, in 6MSM or in 5W81. I think some of these authors would be more than willing to collaborate.

B. General comment

This reader is fully convinced that the non-equilibrium gating model presented here realistically approximates CFTR gating. However, I think some readers might appreciate a more cautious presentation, avoiding all statements that might be read as giving for granted a given interpretation, and transparently and clearly describing the evidence underlying the most important assumptions. For instance, a bit more space in the introduction could be devoted to summarizing the evidence presented in reference Csanády et al., 2010, strongly supporting the kinetic scheme presented in Figure 1.

Essential revisions:

1) The authors hypothesized that in hCFTR an artificial H-bond between the side-chain of glutamine at position 1371 (i.e., in E1371Q mutant) and the backbone carbonyl at G576 of the D-loop stabilizes the NBD dimer. Such H-bond is absent in E1372Q zCFTR. The authors employed mutant cycle analysis on the G576Δ-E1371S mutation pair to demonstrate an energetic coupling between the hG576 and hE1371Q. However, how deletion of G576 might alter the local structure is unpredictable.

We note that the modest effect of the G576 deletion observed in the E1371S background (Figure 3) makes it unlikely that the deletion should result in a major structural rearrangement.

The result does not directly address the discrepancy between zCFTR and hCFTR, either. The D-loop is highly conserved across species with a consensus sequence PFGYLD (residue 574-579 in hCFTR), but in zCFTR the analogous sequence is PFTHLD. The backbone carbonyl oxygen could therefore be harder to access in zCFTR.

Below we provide three alternative suggestions to provide additional support for the interpretation of the discrepancies between human and zebrafish (and other ABC transporters) regarding the presence of a stabilizing H bond involving a catalytic glutamine and the D-loop. The authors can also decide to take a different strategy to provide one more piece of evidence for this point.

– Introducing mutation in the D-loop of zCFTR to match the sequence of hCFTR (and vice versa). The authors' hypothesis would predict that zCFTR with hCFTR's D-loop sequence should recapitulate hCFTR's phenotype: the E-to-Q mutation on the catalytic glutamate would further lengthen the burst duration compared to the E-to-S mutation.

We do not think that such a prediction follows from our hypothesis. The suggested non-native H-bond in hCFTR involves a backbone carbonyl group in the NBD1 D-loop. The particular side chain patterns of the human and zebrafish D-loops have evolved in the context of the entire protein, and have been optimized to match the particular structures that surround them. It seems therefore highly unlikely that isolated replacement of the two divergent residues in zCFTR by those present in hCFTR should convert the conformation of the entire NBD interface region of zCFTR to precisely match that of hCFTR (or vice versa). E.g., our novel experiments on zCFTR NBD1 D-loop deletions (new Figure 4) indicate an important additional structural role of the zT575 side chain in zCFTR, which is clearly absent in hCFTR as the corresponding residue is a glycine (hG576).

– The mutant cycle could be further applied to the zebrafish, by measuring the closing rate of a zebrafish construct containing the zE1372Q mutation together with a deletion on the D-loop. If D-loop deletion does not give a significant burst shortening in the zE1372Q background, it would support the idea that the H-bond in the hE1371Q-CFTR is unique.

We have adopted this strategy and performed mutant cycle analyses in the zE1372Q background. We note upfront that the question relevant to our hypothesis is not whether an NBD1 D-loop deletion shortens non-hydrolytic bursts (D-loop residues might be involved also in other interactions that stabilize the burst!), but whether the effects of an NBD1 D-loop deletion depend on the presence of a glutamine at position zE1372.

In line with comment (2) below, we have constructed the mutant cycle in two different ways, based on deletion of either zT575 (new Figure 4A-C) or zH576 (new Figure 4D-F) in the zCFTR NBD1 D-loop. The results unequivocally demonstrate that the effects of either D-loop deletion are independent of whether a serine or a glutamine side chain is present at position z1372. Thus, ΔΔG_int for the zQ1372-D-loop interaction is not significantly different from zero (-0.60±0.28 kT (p=0.073) and 0.34±0.22 kT (p=0.17), respectively, based on the two models), in full support of our hypothesis.

Interestingly, in zCFTR both D-loop deletions strongly destabilize the burst by disrupting some other stabilizing interaction (possibly a H-bond between the side chain of zT575 and the backbone of zA1375; Figure 4 – Figure Suppl. 1). Further investigating this phenomenon is beyond the scope of the current study.

– Obtain analysis of molecular dynamics simulations that shows that there is a hydrogen bond in the human simulations and not the zebrafish.

In line with our expertise we have chosen the experimental approach discussed above.

2) Structurally, hG576, is better aligned to zT575. These two residues are almost completely overlapping in the backbone. There is a clear difference between the two pdb coordinates in the glutamine conformation, with the amide and carbonyl group positions swapped (Figure 2E-F): hQ1371 points the amide and zQ1372 points the oxygen towards the D loop, such that the zQ1372 amide group ends up closer to the backbone oxygen of zH576. The zQ1372 side chain is also shifted (~0.1 nm) closer to the γ phosphate, away from site 1, and further from the switch histidine (zH1403). Finally, site 1 is more open in zCFTR.

However, there are some uncertainties in modelling atomic coordinates from electron density maps. Negatively charged moieties often do not show up in these maps, for unknown reasons. In the electron density map for hCFTR (Figure Suppl. 1), part of the glutamine side chain (apparently the amide group) is missing. The electron density for both the amide and the carbonyl group is missing in the map for 5W81. How confident are we that the different pdb coordinates (hQ1371 vs. zQ1372) reflect a true experimental difference,

This is always a valid concern, as the pdb coordinates are just those of a model. That is exactly why functional studies are needed to validate/disprove atomic details of model structures. Our functional results confirm that the positioning of the Q1371 side chain in the hCFTR model (6msm) is correct. We further note that in the zCFTR structure (5w81) the amide nitrogen of the zQ1372 side chain would be too far (3.72 Å) to interact with the zT575 backbone even if the positions of the zQ1372 amide and carbonyl groups were swapped (Figure 2F, dark purple dotted line). This is now discussed (lines 145-150), and our new experiments address both possible assignments (see response to comment 1), above.

Would it perhaps be possible that some other interaction involving the glutamine and the D-loop is also involved? This needs to be discussed.

Based on our functional results per se such a possibility can indeed not be excluded. However, the structure (whose resolution in this region is particularly high) does not support any alternative interaction: the distance between all other D-loop functional groups and either polar group of the hQ1371 side chain is >4.2 Å. This is now discussed in the text (lines 298-302).

3) The authors speculated that the reason for D1370N's relatively fast closing rate compared to other non-hydrolytic mutants is the loss of interaction between Mg²⁺ and the negatively charged aspartate. However, this reasoning fails to explain why non-hydrolytic closure of wildtype CFTR in the absence of Mg²⁺ (e.g., Levring et al. 2023 Extended Data Figure 7g) is even slower than the non-hydrolytic closure of D1370N CFTR opened by MgATP, where at least the Mg²⁺ is present. The authors should caution the readers that so far no definitive experimental evidence can explain the destabilizing effect of D1370N.

The cited study (Levring et al., 2023) does not present any data in conflict with our speculation, as neither the burst length of WT CFTR in the absence of Mg²⁺ nor that of D1370N were quantified. To our knowledge the only published study in which the effect of Mg²⁺ removal on burst length was quantified is that by Dousmanis and Gadsby (2002) on CFTR channels in guinea pig heart. Although absolute values of gating rates of those channels are not comparable to those of human epithelial CFTR, interestingly, guinea pig CFTR channels locked open by Mg²⁺ removal closed ~15-fold faster than those locked open by a non-hydrolyzable ATP analog.

Nevertheless, we do agree with the Reviewer that altered Mg²⁺ binding is not necessarily the sole reason for the shortened bursts of D1370N hCFTR. Indeed in the OF structure of WT hCFTR the D1370 side chain also strongly interacts with S1251, and that interaction is also incrementally strengthened as NBD2 proceeds through stepwise subdomain closure (5uak → 8fzq → 6msm). We have reworded this speculative proposal into a more general form (lines 378-382).

4) Based on the results that the double mutant E1371S/K1250A hCFTR has a similar burst duration as single mutant E1371S and K1250A, the authors made a strong claim that both mutations completely abolish ATP hydrolysis. Similar reasoning was applied to D1370N. The limitations in such interpretations should be discussed. The authors made the assumption that the termination of a burst is solely controlled by site 2 (Figure 1C). However, when hydrolysis is significantly diminished, binding of ATP in site 2 is very stable, and thus dissociation of ATP from site 2 versus site 1 becomes hard to distinguish. Whether all hydrolysis-deficient mutants share the same open-to-close transition by releasing ATP from site 2 but retaining ATP in site 1 is still a question.

We surmise that the Reviewer meant separation of the NBD interface around site 2 vs. site 1, not dissociation of ATP from site 2 versus site 1. (It is well established that pore gating is coupled to NBD interface movements rather than to nucleotide binding/unbinding events.)

In principle it might indeed happen that, under conditions when ATP hydrolysis at site 2 is disrupted, the NBD interface separates first around site 1. However, our results directly demonstrate that separation around site 2 remains the prevalent pathway for pore closure even in the absence of ATP hydrolysis. If that were not the case then further slowing of site 2 separation by the non-natural H bond would not be expected to prolong bursts. Specifically, if in hE1371S or hD1370N channels pore closure predominantly proceeded through dissociation of site 1 then further stabilization of site 2 by introduction of the E1371Q mutation would not be expected to prolong τ_b – in contrast to the ~13-16-fold prolongation observed in both backgrounds (Figure 3A-B, black vs. blue; Figure 5D-E, black vs. gray). This is now discussed in the text (lines 344-353).

As the authors have elaborated in the text, it is known that mutations in the degenerate site 1 can affect non-hydrolytic closing. When mutations are introduced to site 2, they might as well result in allosteric effects on the stability of ATP binding in site 1, which could subsequently alter the channel's closing rate. The authors might want to make the readers aware of the complicated relationship between channel closure and CFTR's two ATP binding sites.

Potential allosteric effects of site 2 mutations on site 1 stability cannot be ruled out in general. However, as explained above, our results strongly suggest that closing rate remains controlled by the separation of the site 2 interface, even in the mutants.

5) It is known that non-hydrolytic closing rate of CFTR is phosphorylation dependent, which the authors briefly mentioned in the Discussion.

We disagree. To put it more precisely, what is known is that non-hydrolytic closing rate depends on the presence or absence of PKA, i.e., that rate – just as the microscopic rates of all other gating steps – changes essentially instantaneously upon PKA removal in inside-out patches. On the other hand, there is strong evidence that the abrupt change in CFTR gating pattern upon PKA removal does not reflect dephosphorylation, but rather, loss of a direct effect of bound PKA on CFTR gating (Mihályi et al., 2020; PNAS 117: 21740–6).

Vergani et al. (2003) documented that τburst of K1250A and D1370N in PKA is ~80 s and ~4 s respectively, but both are reduced by roughly twofold when PKA was removed. In this study the burst durations of K1250A (~30 s, Figure 4C) and D1370N (~2 s, Figure 4E) indicate that these channels are not strongly phosphorylated.

We disagree. Just as in the study by Vergani et al. (2003), the patches in the current study were exposed to a high concentration (300 nM) of bovine PKA until the currents had reached steady state, and – as explained above – changes in gating following PKA removal are not caused by dephosphorylation (Mihályi et al., 2020). Thus, the channels in the current study are fully phosphorylated, and the differences in rates between those studies likely reflect modulation by PKA which remains bound until the channels close.

Similarly, the τburst of E1371S in PKA is over 100 s (Bompadre et al. 2005), significantly longer than that in the current study. Although it is unclear how a different degree of R domain phosphorylation affects non-hydrolytic closing, the fact that it does again suggests that the simplified scheme used as the base for data interpretation may have its limitation. The Discussion would benefit from a more cautionary note on the oversimplification of the IB1↔B1 transition, and clarify that channels are not strongly phosphorylated in the current experimental condition.

The fact that bound PKA modulates CFTR gating rates is briefly mentioned in the Discussion (lines 390-392), but the mechanism by which that happens is irrelevant to the scientific questions addressed here, and is beyond the scope of the present study. We agree that all kinetic schemes are "oversimplifications" of the systems they represent, as the microscopic rate constants typically depend on a multitude of factors (including temperature, pH, etc). Correspondingly, the legend to Figure 1C states that the gating scheme is "schematic", and that not all gating states are depicted. Nevertheless, we believe that it is sufficiently detailed to facilitate mechanistic interpretation of the observed phenomena.

6) The τburst of E1371Q CFTR is over 400 second while the τburst of K1250A-E1371Q double mutant is shortened to ~200 second (Figure 3B, black vs Figure 4C, black). The K1250A-E1371S CFTR also seems to have a shorter τburst than E1371S CFTR (Figure 4C, blue vs Figure 3B, blue). Although the effect of the K1250A mutation on shortening τburst of E1371Q and E1371S CFTR is not as dramatic as the D1370N mutation, the authors might want to clearly state if there is indeed a significant difference and address how K1250A mutation has such destabilizing effect, and provide additional statistical analysis.

In the E1371S background introduction of the K1250A mutation decreased τ_burst only by ~28%, which failed to reach significance (p=0.052). The exact number and the significance value are now provided in the text (line 251).

On the other hand, the ~2-fold shortening of τ_burst in the E1371Q background is indeed a significant effect (p=0.003). Another way to express the same fact is to state (see lines 329-332) that the increase in τ_burst caused by the E1371Q mutation is only ~5-fold in the K1250A background, as opposed to 13-16-fold in the hE1371S or hD1370N backgrounds. A possible explanation could be that the K1250A mutation causes a slight increase in the hG576-hQ1371 distance, resulting in a small (~0.6 kT) but significant (p=0.003) decrease in the strength of the non-native H-bond. This is now discussed in the text, and the significance value is provided (lines 332-335).

7) page 10 line 214: "If both single mutations failed to completely abolish k1 then a further slowing in k1, i.e., a further prolongation of taub, would be expected in the double mutant K1250A/E1371S."

This prediction assumes that the effects of the mutations on hydrolysis are independent. One could envisage scenarios in which approximately the same residual hydrolytic activity would remain in K1250A, E1371S and K1250A/E1371S, if the two mutations worked in a similar way to slow hydrolysis down. Admittedly this is extremely unlikely, as we know the Walker A lysine and Walker B glutamate play different roles in catalysis, as the authors explain just above. But since this is a crucial point in reaching one of the main conclusions of the paper, it might be worth developing the argument of the redundancy of the interaction more extensively.

Thank you, argument added (lines 245-246).

8) It is unclear if the rates for different mutations were measured at -20 or -80 mV. This should be described, and if different voltages were used, a discussion should be provided for how the different voltages are accounted for in the energetic calculations from the rates.

CFTR channel gating is largely voltage-independent: P_o differs by only ~15% between -100 and +100 mV (Cai et al., 2003), and non-hydrolytic closing rates at -120 and +60 mV are similar (Csanády et al., 2014). Thus, the voltage is irrelevant for the calculation of closing rates which simply reflect the inverses of the fitted time constants (see line 461). In the present study non-hydrolytic closing rates were obtained between -20 and -80 mV; -80 mV was used for patches with smaller numbers of channels, to facilitate intraburst kinetic analysis of last-channel segments; -20 to -40 mV was used when recording large currents, to increase seal stability. This is now explained more clearly in Materials and methods (lines 435-438).

9) A better description should be provided for how the decay currents were normalized.

In each figure legend we have replaced the sentence "current amplitudes are shown rescaled by their steady-state values" with the more detailed statement "current amplitudes are shown normalized by their steady-state values in ATP (i.e., just before ATP removal)".

10) Page 9, line 174 and following are very hard to follow. Possible modification: "The coupling energy (DDGint(B1→T‡); Figure 3C, purple) is a measure of the discrepancy in the effects caused by perturbation of one target position depending on whether the other target position is mutated or intact (Figure 3C, parallel arrows). DDGint(B1→T‡) is thus given by the difference between DDG 0T++ -B1 values on parallel sides of the cycle (Figure 3C, numbers on parallel arrows). DDGint(B1→T‡) thus quantifies the change in the strength of the G576-Q1371 interaction in the E1371Q background construct while the channel proceeds from the B1 state to state T‡."

Corrected as suggested, thank you (lines 178-184).

The latter statement is true only if simplifying assumptions are met. Among these, it is assumed that in each single mutant and in the double mutant the interaction between the target sites is completely abolished. This assumption is, strictly speaking, met in the deletion mutants. However, the serine could well form new contacts with carbonyl oxygens in the D-loop that depend on the glycine deletion. The statement should be qualified and/or the assumptions should be stated.

This is highly unlikely. The only D-loop backbone carbonyl oxygen within 4.2 Å of the hQ1371 amino group is that of residue hG576, and the hE1371S mutation is expected to further increase the distance between the D-loop backbone and the h1371 side chain by ~2 Å. Thus, formation of alternative interactions between the D-loop and the hS1371 side chain is precluded. We now discuss the validity of these assumptions in Materials and methods (lines 465-472).

Non-essential revisions suggested by reviewers:

1) Including the traces for single mutants E1371Q and E1371S on Figure 4B-E would be helpful.

We have decided not to add these: at the time scale required to plot the E1371Q trace the other traces would all become squashed on top of each other, especially the traces in panel D.

2) The different structures in Figure 4 Suppl. 1 are hard to distinguish, so it would be helpful to use more contrasting colors. It would also be valuable to include the loop for G576 for both structures on panel B.

This panel focuses on the environment of the D1370 side chain. Adding the D-loop would make the panel overcrowded and distract the attention from its message.

3) Line 300-306: It would be helpful to provide a better clarification for the meaning of "bidirectional effects" – Does "bidirectional" here just mean both slowing down (as H1348A or P-ATP; E1371Q) or speeding up (as K464A; D1370N on k-1) of k-1, non-hydrolytic closure?

Yes, that is correct. This sentence now reads:

"The observed shortening or lengthening of non-hydrolytic τ_b by different site-2 mutations" (line 336).

Also, the authors mentioned that non-hydrolytic closure is accelerated by K464A but slowed by H1348A and P-ATP, but it is unclear how perturbations in site 1 are related to the current scheme in this study, in which the IB1↔B1 transition considers only the separation of site 2. Some elaboration might be needed.

This is a misunderstanding. The fact that site 1 does not completely separate during pore opening/closure does not mean that it remains entirely static. Indeed, the observed effects on non-hydrolytic τ_b of various perturbations at the site-1 interface clearly demonstrate that even site 1 undergoes substantial conformational changes coincident with pore opening/closure, as has been elaborated in multiple earlier studies (Csanády et al., 2013; Mihályi et al., 2016; Sorum et al., 2017). Correspondingly, the cartoon in Figure 1C was drawn to illustrate some movement even in site 1. We have added a clarifying sentence (lines 342-344).

4) Figure 3D, E1371S: the scale for the traces appears different than from the other mutants. Is this due to increased overall noise in that recording?

Yes, the apparent difference is due to larger noise, the scaling is identical. No changes in conductance were observed for the E1371S construct.

5) The authors use IF, inward-facing, and OF, outward-facing, to refer to states in the kinetic scheme likely to adopt these conformations. Page 2, line 51 "The IF conformation, in which the pore is sealed near its extracellular end (Figure 1C, left), is a long lived state (~1 s) and corresponds to long "interburst" (IB) closed dwell times observable in single channel current recordings." I think it might be better to clearly separate structures and states at the start.

Possible alteration: "CFTR likely adopts an IF conformation, in which the pore is sealed near its extracellular end, during the long lived (~1 s) state (Figure 1C, left) which corresponds to long "interburst" (IB) closed dwell times observable in single channel current recordings. We refer to these closed states as "IF states"".

Thank you, corrected as suggested (lines 51-54).

Then later: "When CFTR adopts the OF conformation the external gate is predominantly open, and a lateral portal which connects the channel pore with the cytosol generates a transmembrane aqueous pathway permeable to anions ((Zhang et al., 2018); Figure 1C, right, double arrows). However, functional studies show that the continuity of the transmembrane pore is occasionally disrupted for brief (~10 ms) intervals by a smaller conformational change (not depicted in Figure 1C) likely confined to the external ends of the TMD helices (Zhang et al., 2017; Zhang et al., 2018; Simon and Csanády, 2021).

Correspondingly, in single-channel recordings, this "OF" or "bursting" (B) state corresponds to clusters of channel openings separated by brief ("flickery") closures (Winter et al., 1994). The OF state is also relatively stable, with a dwell time of hundreds of milliseconds."

Thank you, corrected as suggested (lines 54-62).

6) Page 4, line 80: the kinetic scheme used to describe CFTR (Figure 1) is used to illustrate ATP transporter ATPase cycles. Because the scheme has emerged from studies on CFTR gating, and not ABC exporter transport cycles (although the cycle might be similar), I would not refer to specific rates in the figure when introducing exporters. Possible alteration:

"The "coupling ratio" (CR), i.e., the fraction of initiated cycles that are completed through ATP hydrolysis, depends on the relative rates of the two possible exit pathways from the prehydrolytic state. In most ABC transporters the actual values of these rates are hard to directly estimate, but in CFTR direct measurements of conformational dwell times are made feasible by single-channel current recordings, providing estimates for microscopic transition rates. Thus, for a wild-type (WT) channel, we can estimate ATP hydrolysis rate (k1, Figure 1C) vs. non-hydrolytic NBD dimer dissociation rate (k-1, Figure 1C). Because typically k1>>k-1, CR=k1/(k1+k-1) is near unity."

Thank you, corrected as suggested (lines 85-87).

7) Page 14, line 281 (and page 10 line 217): "the mutual lack of effect of mutation E1371S on taub of K1250A (Figure 4B-C, blue vs. gray) and of mutation K1250A on taub of E1371S CFTR" Why is introducing E1371S in K1250A described separately from introducing K1250A in E1371S? Since E1371S, K1250A, and K1250A/E1371S all have a similar taub it seems redundant to discuss the sequence in which the mutations are introduced. Why do the authors feel it is useful to present the results this way?

We discuss the three mutants in this sequence for didactic reasons. In principle, the effects of the two mutations might be asymmetric. E.g., if both mutations disrupted ATP hydrolysis, but only mutation K1250A affected NBD dimer stability, then introducing mutation K1250A into an E1371S background would affect τ_b, whereas introducing mutation E1371S into a K1250A background would not. The mutual lack of (significant) effect thus shows that neither mutation (significantly) alters NBD dimer stability. We have reworded this sentence to read:

"When viewed from a different angle, mutation hK1250A also fails to prolong τ_b in the hE1371S background".

Author details

1. Márton A Simon

   1. Department of Biochemistry, Semmelweis University, Budapest, Hungary
   2. HCEMM-SE Molecular Channelopathies Research Group, Budapest, Hungary
   3. HUN-REN-SE Ion Channel Research Group, Budapest, Hungary

   Contribution

   Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Visualization, Writing – review and editing

   Competing interests

   No competing interests declared

2. Iordan Iordanov

   1. Department of Biochemistry, Semmelweis University, Budapest, Hungary
   2. HCEMM-SE Molecular Channelopathies Research Group, Budapest, Hungary
   3. HUN-REN-SE Ion Channel Research Group, Budapest, Hungary

   Contribution

   Resources

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-8251-5857

3. Andras Szollosi

   1. Department of Biochemistry, Semmelweis University, Budapest, Hungary
   2. HCEMM-SE Molecular Channelopathies Research Group, Budapest, Hungary
   3. HUN-REN-SE Ion Channel Research Group, Budapest, Hungary

   Contribution

   Resources

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-5570-4609

4. László Csanády

   1. Department of Biochemistry, Semmelweis University, Budapest, Hungary
   2. HCEMM-SE Molecular Channelopathies Research Group, Budapest, Hungary
   3. HUN-REN-SE Ion Channel Research Group, Budapest, Hungary

   Contribution

   Conceptualization, Data curation, Supervision, Funding acquisition, Validation, Writing - original draft, Project administration, Writing – review and editing

   For correspondence

   csanady.laszlo@med.semmelweis-univ.hu

   Competing interests

   Reviewing editor, eLife

   "This ORCID iD identifies the author of this article:" 0000-0002-6547-5889

• 206

  Page views

• 56

  Downloads

• 0

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.