CFTR function, pathology and pharmacology at single-molecule resolution

The cystic fibrosis transmembrane conductance regulator (CFTR) is an anion channel that regulates salt and fluid homeostasis across epithelial membranes1. Alterations in CFTR cause cystic fibrosis, a fatal disease without a cure2,3. Electrophysiological properties of CFTR have been analysed for decades4–6. The structure of CFTR, determined in two globally distinct conformations, underscores its evolutionary relationship with other ATP-binding cassette transporters. However, direct correlations between the essential functions of CFTR and extant structures are lacking at present. Here we combine ensemble functional measurements, single-molecule fluorescence resonance energy transfer, electrophysiology and kinetic simulations to show that the two nucleotide-binding domains (NBDs) of human CFTR dimerize before channel opening. CFTR exhibits an allosteric gating mechanism in which conformational changes within the NBD-dimerized channel, governed by ATP hydrolysis, regulate chloride conductance. The potentiators ivacaftor and GLPG1837 enhance channel activity by increasing pore opening while NBDs are dimerized. Disease-causing substitutions proximal (G551D) or distal (L927P) to the ATPase site both reduce the efficiency of NBD dimerization. These findings collectively enable the framing of a gating mechanism that informs on the search for more efficacious clinical therapies.

Major comments: 1) The conclusions hinge on quantitative comparisons between smFRET data and electrophysiology data.Some clarifications are needed.The properties of CFTR are compared to the unlabeled CFTRFRET in E1b-g and the fluorophorelabelled version in Fig. E1i-k and E2.For Fig. 1 and 3 it is not clear whether this is the labeled or unlabeled version.The authors should clarify this in text, Methods and legends in a consistent manner.A clear nomenclature would help.If Figs. 1 and 3 use the unlabeled version, authors should justify why.Also, authors should note if the constructs being compared differ in possessing a Cterminal GFP.
In E1i-j the current-voltage relationships suggest that CFTRFRET has higher conductance (authors should indicate how many patches went into this comparison).In addition, missing is a comparison of open probability, which is critical for the interpretation.
2) Even when the reader knows that patch recordings and FRET come from different experiments, placing these side by side or one above the other could give the confusing impression that they were measured simultaneously.This confusion is aggravated by sentences like this in the text (line 206): "The W401A variant, which is capable of binding and hydrolyzing ATP via the consensus site, underwent rapid transitions between NBD-separated and -dimerized states that correlated with channel gating (Figure 2c)."This sentence should be changed (e.g. to "more closely resembled," not correlated).To further clarify this, authors should add to Authors should also define for panel f ratio between probabilities of opening and dimerization which is the numerator.
3) The conclusion of Fig. 3 is that NBD dimerization precedes channel opening.The average rise upon rapid ATP perfusion of the phosphorylated channel of high FRET occupancy (dimerization) in smFRET and of channel opening in patch is faster for dimerization.Quantification was by mono-exponential fits.These should be superimposed on the data.Tau of dimerization is shown as <100 ms.The precise value and SEM should be given.Fig. E7 is supposed to demonstrate that differences in kinetic behavior seen in Fig 3 are not due to differences in solute exchange between the two sets of experiments, but this figure only displays ephys data.What is the equivalent dataset for smFRET, and where is it displayed?The Tau of patch solution exchange in Fig. E7 is 150 ms, slower than the <100 ms value given for smFRET in Fig. 3a, so it is unclear how well the exchange rates actually match.
The authors state: "We conclude … that the mean first passage time between NBD dimerization and channel opening in productive instances is approximately 500 ms."However, their Tau of opening is 490 ms and they begin to see channels opening almost immediately.The delay they seek is the one between dimerization and opening and would be a smaller value.

Minor comments:
1) E1j is mistakenly referred to as E1k in the text (line 161) 2) All electrophysiology traces should have current amplitude scale bar (e.g.missing in Fig. 2c) 3) Figure 1.E1371Q should be shown without ATP as well.4) No mass spec data to demonstrate that phosphorylation is consistent/present across all conditions 5) Figure 2. 2cd ATP concentration and phosphorylation status should be stated in legend for clarity.It is also not so easy to determine by eye the contribution of open probability and dimerization to the coupling ratio by looking at 2e and 2f.It would help to additionally show the dimerization probability for the five different variants.6) Example traces in E1i look different at lower membrane potentials (-90 and -150)--is there a better representative trace or what is the hypothesized reason for this difference?7) Analogous examples of current from WT channels should be shown for comparison in E1b and E1k.8) Line 140 NDB typo 9) A question regarding comparison of electrophysiology and FRET data-are there any integral membrane phosphatases which could contribute to the observed differences between channel opening and NBD dimerization probability?10) The authors should make clear that disease-causing CFTR variants have multiple modes/effects, and should describe the direction in which a pharmacological agent should act (i.e.promote channel opening as opposed to promoting channel closure, etc).On p. 9, the effects of G551D and L927P on channel gating should be described first, before explaining effects on FRET-based readouts.11) Presence of a native lipid membrane or detergent environment may affect the dynamics of the channel.To address this, the authors perform a set of FRET experiments in proteoliposomes, which should more accurately approximate a cell membrane.The FRET dynamics and dependence on ATP look similar to what they see in detergent, bolstering confidence in their experiments.One odd thing is that only a small fraction of channels are ATP sensitive.They attribute this to channel orientation within the bilayer.If the 10xHis tag is located on the NBD side of the protein, it is unclear why any immobilized molecules should have inward-facing NBDs which are inaccessible to ATP. 12) It is not clear if the kinetic model adds much to the discussion, but it also is presented with appropriate caveats.Does the model make a prediction in a way that could be used to test the alternative hypotheses presented in the introduction of the paper?13) Figure 1i shows that a very large percentage of channel openings have a dwell time of at least 10 ms, which could be detected using smFRET.Though it may not be feasible for this study, it would be highly convincing if it were possible to design a FRET sensor of channel pore activity that approximates the electrophysiological readout?This would allow direct comparison within a single controlled system.
Referee #2 (Remarks to the Author): The current manuscript by Levring et al. presents incredibly comprehensive studies of CFTR function as an ATP-gated ion channel; these studies include electrophysiological recordings of CFTR channel activity, real-time monitoring of the dimerization of CFTR's two nucleotide binding domains (NBDs) with smFRET imaging, biochemical measurements of ATP hydrolysis rate, and computer simulations based on a proposed gating model.Here are the main findings and their mechanistic implications: 1.The NBDs are readily dimerized by ATP binding to the two ATP-binding sites in fully phosphorylated CFTR, but the dimerized state (high FRET) is much more stable than the duration of either channel open state or the closed state at a saturating concentration of ATP.This uncoupling between NBD dimerization and channel gating suggests that multiple cycles of gate opening and closing can occur with NBDs remained dimerized (inferred from high FRET), contradicting previous idea of a strict coupling of NBD dimerization to gate opening and hydrolysis-triggered NBDs separation (low FRET) to gate closing (Csanady et al., 2010).
2. By manipulating ATP binding to each site (Y1219A for the consensus site and W401A for the degenerate site), they provided evidence for NBD dimerization by ATP occupancy at either site.While both NBD dimerization and gating were mostly preserved in W401A, Y1219A channels showed robust NBD dimerization but little activity, supporting previous reports that gating of CFTR is mainly through ATP binding to the consensus site (Vergani et al., 2003;Zhou et al., 2006).3.By comparing the rate of NBD dimerization and the rate of channel activation upon sudden addition of ATP, they found that NBD dimerization takes place much faster than gate opening, contradicting previous proposition that NBD dimerization constitutes the rate-limiting step in CFTR gating (Vergani et al., 2005).Although NBD dimerization precedes gate opening, NBD dimerization itself is not sufficient to open the gate because dimerization (high FRET) can be seen in conditions when there is little channel activity (e.g., upon ATP washout, or in the presence of inhibitory concentration of ADP). 4. Supporting the idea that ATP at the consensus site plays a critical role in opening the gate, the authors demonstrated that the disease-causing mutant G551D at the consensus site exhibits defective gating and NBD dimerization.Interestingly, the dimerized NBDs in G551D show FRET between high and low FRETs, suggesting the presence of an intermediate NBD dimer proposed previously (Tsai et al., 2010).Similar but not identical intermediate FRET was seen in another pathogenic mutation L927P (in TM8), which is 50 Å away from the consensus site.This latter observation suggests a long-distance effect of the L927P mutation on NBD dimerization, but the nature of this allosteric effect is unclear.
5. By testing the effects of CFTR potentiators ivacaftor and GLPG1837, they found a dramatic gating effect with minimal and moderate changes of NBD dimerization in G551D and L927P variants respectively.These are interesting observations, but it seems that not much additional insight can be gained (see below).
6.These results were put together to form the basis of a kinetic model, which was further examined through computer simulations.Although the proposed kinetic mechanism recapitulates many of the experimental data, some simulated results do deviate from actual data.
No doubt, these are important studies that reveal many unexpected phenomena, which lay a solid foundation for settling the debates over the coupling mechanism between NBD dimerization and gate opening, as well as for advancing our overall understanding of the structural-functional relationships of CFTR gating.This reviewer applauds the authors for such a remarkable attainment.
Major comments: 1.The two conformations of NBDs-separated versus dimerized-can be resolved by the shift of FRET from 0.25 to 0.49 (Fig. 1).In Fig. 1e, the FRET signal remains stable at 0.49 at 10 mM ATP while the electrophysiological recording shows repeated opening and closing of the WT-CFTR channel.The authors thus concluded that the NBDs remain dimerized throughout multiple rounds of ATP binding/hydrolysis (Fig. 5).Conceptually, the consensus site in the "dimerized NBDs" must open up a space wide enough to allow the exchange of hydrolytic products for the next ATP.Indeed, in ATP washout (and Mg addition) experiments (Fig. 3b, c), transition from high FRET to low FRET was observed immediately upon ATP removal (or Mg addition).As the waiting time is too short for ATP dissociation from the degenerate site, this result suggests that upon ATP hydrolysis at the consensus site, disengagement of the two NBDs does occur.The reason this high to low FRET transition is not seen in the continuous presence of millimolar ATP may be due to rapid rebinding and redimerization of the disengaged NBDs at millimolar ATP.This scenario supports the notion that the state "degenerate site ATP, pore closed" should assume a different dimeric state with a cracked consensus site to allow step 6 (Fig. 5) to take place.The canonical NBD dimer in the cryo-EM structure leaves no room for ATP to diffuse into (or out of) its binding pocket.
2. I wish the authors would have elaborated the rationale behind choosing positions 388 and 1435 for their FRET pair.T388 is at a position without a concrete secondary structure.Wouldn't this positioning result in some uncertainty in FRET measurements as the FRET efficiency is determined by not only the distance but also the relative orientation between the donor and the acceptor?It would be nice if another pair is chosen to replicate some of the key observations described above.
3. The W401A and Y1219A mutations were designed to weaken ATP binding.The authors observed that the NBDs of W401A and Y1219A mutants can still dimerize, and thus concluded that ATP binding at either site is sufficient for NBD dimerization (P7, line 215).Such conclusion is based on the assumption that the degenerate site in W401A (or the consensus site in Y1219A) must be vacant at 3 mM ATP.However, in electrophysiological study, the Y1219G mutation reduces the apparent affinity for ATP to a K1/2 of ~5 mM (Zhou et al., 2006).If we assume that Y1219A affects the affinity similarly, at least a portion of the Y1219A channel population in Fig. 2 would still have ATP bound to the consensus site at 3 mM ATP.Given that the degenerate site has much higher affinity for ATP, it is unlikely that the degenerate site in W401A would be vacant at 3 mM ATP.Thus, the data in Fig. 2 only support that NBDs of W401A and Y1219A can dimerize.The occupancy of ATP at each site cannot be determined, and thus whether a single ATP is enough to trigger dimerization cannot be inferred conclusively.However, the conclusion may not be incorrect once data in Fig. 1j are taken into consideration.The ATP concentration dependence of high to low FRET transition rate suggests the existence of different states of NBD dimer with one or both sites occupied.
4. Some of the kinetic steps in Fig. 5 need justification/explanation: a. Dissociation/binding of ATP in the degenerate site happens when the NBDs are separated (step 7, 8), but dissociation of hydrolytic products and binding of ATP to the consensus site happen when the NBDs are dimerized (step 4, 5, 6).Some intermediate conformations of NBDs need to exist.b.Why step 3 is drawn differently in Fig. 5 and Fig. E11?Also, why did the authors propose only the open state hydrolyzes ATP, but not 2 ATP pore flicker-closed state or 2 ATP pore closed state?c.Is the flicker-closed state the same as the cryo-EM structure of E1371Q-hCFTR?It's been known that many of the flickery events seen at negative membrane potentials are not likely real functional "state".Because of their high voltage dependence (see Fig. E1i.), many of the events are considered blocking of the pore by cytosolic large anions.To get the right kinetic parameter for the true flickerclosed state, one has to analyze data from recordings at positive membrane potentials.d.There is a mistake both in the text and in the figure legend.Steps 6 -8 represent transitions among closed states; therefore, it is not a new gating cycle, which by definition should include a closed event and an open event.
Minor comments: 1.I am not sure I truly understand the meaning of the term "hierarchical" gating mechanism.
2. P2, line 3: Do you mean electrophysiological properties? 3. P5, line 140: It should be "…to a stably NBD-dimerized state." 4. P6, line 157: Previous electrophysiological studies of R117H-CFTR and zebrafish CFTR, as well as the cryo-EM structure of E1371Q all indicate the presence of conductive and non-conductive states with dimerized NBDs.Please more explicitly state the unique insight from these data. 5. P6, line 161: …with ATP concentration (Figure 1j), not 1k.

P6, line 172:
The ADP experiments are interesting in that the low FRET value seems different from that without ATP.Is it reproducible?If it is, doesn't that suggest different degrees of disengagement of NBDs? 7. P7, line 191: The time constant for ligand exchange at the degenerate site is around 30 s (Tsai et al., 2010), which interestingly is very similar to the duration of high FRET upon ATP washout (Fig. 3b).

P7, line 226:
It should be open probability, not opening probability.(In other parts of the text too.) 9. P8, line 237: The changes of FRET upon sudden addition of ATP show heterogeneity.This observation may be related to a recent paper by Yeh et al. (2021).The stated similarity to W401A and double exponential current rising upon addition of ATP (Fig. 1b and E1b) all remind this reviewer of the results published in this paper, which shows different closed channel conformations (perhaps different NBDs conformations) once ATP is removed.10.P8, line 255: Of note, rapid removal of ATP caused a double exponential decay of WT-CFTR currents (Fig. 2B in Lin et al., 2014) with time constants (τ1 < 1 s and τ2 =29 s) similar to these values reported for NBD dimerization relaxation.These shouldn't be just coincident.11.P9, line 277: Not a correct reference and interpretation.The cited ligand exchange experiments lead to two different turn-over rates of ATP and the idea of a "partial dimer."On the other hand, by carrying out different ligand exchange experiments, Jih et al. (2012) propose a re-entry pathway featuring repetitive cycles of ATP turnover in a single opening burst.Both studies done by Hwang and colleagues invoke a partial dimer state where the consensus site is open up to allow replenish of ATP.They do not suggest that ATP turnover can occur in a single canonical NBD-dimerized conformation, where the ligand is occluded.
12. P10, line 312-314: I don't understand what the authors mean by the similar nature in "the similar nature of both mutations' impacts on NBD dimerization and channel opening".G551D and L927P actually exhibit different behaviors in their experiments.In Fig. 4b, the transitions between high and low FRET in G551D is different from those of L927P.Also, GLPG1837 drastically changes L927P's FRET pattern, but barely affects G551D.Moreover, electrophysiological properties of these two mutants and pharmacological effects of CFTR potentiators should be discussed in a more quantitative manner (see comment 18 below) before one can make more definitive conclusions.
14. P11, line 375-377: Previous studies suggest that reagents such as NPPB and high-affinity ATP analogs, act synergistically with Ivacaftor and GLPG1837 on G551D channels by promoting NBD dimerization (Lin et al., 2016;Yeh et al., 2017).The authors may consider incorporating these studies in the Discussion.16. Figure 1e: If I understand correctly, the electrophysiological recordings are from WT-CFTR, not from the CFTRFRET protein.Wouldn't it be a better comparison to use the electrophysiological recording of CFTRFRET protein here?Of note, the CFTRFRET construct altered the majority of the endogenous cysteine residues; this alteration could affect CFTR gating as seen in the literature.In Figure E1, the authors demonstrated their ability to use CFTRFRET proteins in electrophysiological experiments and concluded that its gating kinetics are similar to WT-CFTR (but see comment 19 below).17. Figure 1j: I don't understand the data points at 0 ATP.Do the authors still observe FRET transitions even without ATP, meaning that the NBDs dimerize without ATP?In 1e, FRET always stays at low FRET.
18. Figure 4b: In response to GLPG1837, the current trace of G551D barely changes as if GLPG1837 has little effect.This potentiation effect is far from the reported ~30-fold stimulation shown in Fig. 4g.Please comment on the discrepancy and consider replacing the recording with a more representative one.Similar concerns are raised for the effect of GLPG1837 on L927P.The singlechannel activity in 4b does not change 20-fold as indicated in Fig. 4g.
19. Figure 4f: It is shown that the Po of E1371Q is ~0.7 and increases to ~0.8 with GLPG1837.It is puzzling how E1371Q has a Po of only 0.7.Numerous studies showed that with hydrolysis eliminated, the closing rate of E1371Q (or E1371S) is >100-fold slower than WT, resulting in a Po close to unity (Bompadre et al., 2005;Csanady et al., 2013;Vergani et al., 2003;Yu et al., 2016).Indeed, the E1371Q single channel in Figure 1f seems to be almost always open, though it is hard to determine whether those brief downward deflections are flicker closures or interburst closures with this time scale.It would be helpful to expand an E1371Q single channel trace to allow a better assessment of channel behavior in a quantitative manner.20. Figure E1i: From the raw trace, the Po of the FRET construct is apparently higher than the WT.The open burst time of FRET variant is visibly longer than WT.Is this a consistent finding?Please comment on the discrepancy.E8g: What is the second arrow below the ADP+Pi for?22. Figure E11b, d, and e: In b, the simulated single-channel trace shows a much longer open time than the experiment, while the closed time seems not affected much (or slightly shorter).However, in d and e, the experimental and simulated dwell times match perfectly but the two closed times deviate significantly.Please comment on this discrepancy.

Figure
23. Figure E11h: The experimental ATP turnover rate is 15/min (or 0.25/s), corresponding to one ATP hydrolyzed every four seconds, a much longer time than a gating cycle time of wildtype CFTR (~1 s).If, as proposed, multiple ATP hydrolysis cycles occur, the theoretical ATP hydrolysis rate must be even higher than 1/s.As stated, there is a likelihood that this biochemical measurement is not very accurate because of imperfect protein preparation.Alternatively, the proposed gating model is overly simplified.Either way, I wonder whether the simulation really provides more insights into the gating mechanism of CFTR.Sometimes it may be better to be vaguely right than precisely wrong.I leave this to the authors to decide.potentiators.J Gen Physiol, 149(12), 1105Physiol, 149(12), -1118. https. https The manuscript by Levring et al describes characterization of molecular cycles involved in activation of the CFTR channels.Specifically, single molecule FRET studies done in parallel with single channel electrophysiology enable to characterize the relationship between ATP binding, hydrolysis and channel opening.Furthermore, how disease-causing mutations and therapeutic correctors modulates these events are studied and discussed.
The study is very nicely crafted, the experiments are very well performed, and the conclusions are convincing.
This work brings a clear view on a long-standing issue in the CF field and deserves publication.
Nevertheless, a number of points require clarification.
1/ FRET values CryoEM structures of human CFTR have -so far-captured 2 conformations with the NBDs either separated or dimerized.Structural data on ABC structures suggest more extended conformational diversity with several different substates.The differences in FRET values observed in the present manuscript may be indicative of such diversity.
Therefore, the authors need to clearly assign -when possible-the link between the FRET valued and the identified conformer .
The ATP-free CFTR shows a low FRET state at 0.25.Does this match the cryo-EM structures of the NBD separated state ?Similarly does the 0.49 FRET value match the NBD-dimerized state for the used probes?The authors need to run careful simulations (for examples with FPS, see Kalinin et al) with the appropriate probe models to evaluate the calculated FRET value for each conformation.ATP-bound dephosphorylated CFTR shows a 0.28 FRET value.How does that translate in terms of distance change given the R0 for these probes ?
The corresponding cryo-EM structure showed in Figure E4 does not address thes issues, in great part due to the limited resolution.In fact, the authors only vaguely describe "proximal changes in local structure".What does that mean?In my opinion in its current state this structure does not bring relevant information and should be either improved or removed from the manuscript.
2/ The E1371Q mutant appear to be locked in the high FRET state during the time course of the experiments, at least in the trace shown (100s, Fig1f).However, the distribution on Fig1c seems to show a significant low FRET population (~5-10%), pretty much identical to that seen for the "wt" CFTR (which can hydrolyze ATP and thus should spend more time in the NBD separated).
In that line, it is unclear to me why the authors see no NBD separation in this mutant after phosphatase treatment (Fig E3).One may expect to see some separation over the course of the measurement.
In the same Figure, can the author comment on the difference in ATP concentration dependence of channel gating vs NBD dimerization?At 3mM ADP 80% of the channels are still in dimers but gating is almost gone.
3/ Very little is shown regarding the membrane-embedded measurements.What does the sentence "While the fraction of the molecules responsive to ATP was significantly reduced" refer to ?Only single traces are shown.No statistics ("significantly" ?).

4/ ATP binding sites
The authors use the W401A and Y1219A mutants originally described by Hwang and colleagues (although mutated into Gly in that study).What do the authors mean by "destabilize ATP binding" ?
It should be noted that the mutations decrease affinity but may not prevent it.For example, W401A is still able to bind ATP (albeit whith lower affinity) and 3mM is quite a large concentration.
Perturbating the ATP binding at the degenerate site may also alter the conformational equilibrium of NBD1 that we recently described (Scholl et al, 2021) On P8 the authors state that for mutant Y1219A "ATP binds only at the degenerate site".This is not proven.
On P7, the authors describe that "the Y1219A variant, which binds ATP mainly at the degenerate site, transitioned between NBD-dimerized and separated states slowly, and very rarely exhibited channel opening" but they only show single traces.They should show distributions.5/ Figure E3K is unclear to me.Is it expected to see such a significant effect at 0.1 and 0.3 mM ADP while 3mM ATP is maintained (and while Fig E3I show predominantly NBD in dimer form)?I assume that the last part of the graph has 0mM ADP (correct?).Why is the rebound much faster than what is observed during the first 2 mins?6/ The coupling ratio seems unclear to me.What if the two would anticorrelate ?The coupling ratio would still be 1 ?7/ P7 the authors state "We infer from these observations that channel opening probability is enhanced by interactions of CFTR with the terminal phosphate moiety of the ATP molecule bound within the consensus site.".This should be readily testable as the terminal phosphate at the consensus site is coordinated by K1250 and T1246.Point mutations should support this statement.

8/ Mutation G551D
The authors show an intermediate FRET value of 0.37.The authors suggest that the mutation introduce a "steric and electrostatic repulsion at the NBD interface" which is very vague.An Asp at position 551 will likely prevent binding of ATP at the consensus site (at least in a NBD dimerize state) because the acidic side chain would directly and closely face the terminal phosphate.This suggests that the NBD1-NBD2 dimer is not formed at the consensus site and that the protein is adopting an intermediate/different conformation.This also suggest that the ATP "stimulation" is due to dimerization at the degenerate site.Therefore the "dimerized dwell time" on Fig E9E may actually not correspond to dimerized state as "expected", ie.where both sites are closed, as observed by Cryo EM.
In addition, the statement on P10 that "Importantly, both the G551D and L927P variants exhibited ATP-dependent FRET response indicative of wild-type ATP binding affinities" is hard to reconciliate with the position of D551 in the mutant Furthermore, the effect of potentiators on G551D are not clear, and not discussed.Ivacaftor was approved to treat G551D patients, we would expect to see an effect.The authors show some individual traces with GLPG1837, but not the approved drug Ivacaftor (why?) from which there seem to be no significant effect.The authors should explore and or discuss this.In contrast, stimulation of the L927P mutation does lead to fully dimerized state which does seem to correlate with increase channel opening (although Fig 4f shows quite modest improvement).

Point-by-point response
We sincerely thank the Reviewers for their time and constructive comments, which have helped to significantly improve the clarity and scholarship of the manuscript.
Reviewer #1 (Remarks to the Author): The Levring et al. paper seeks to determine how dimerization of CFTR NBDs is coupled to ATP binding and hydrolysis, and to pore opening, using a combination of smFRET and electrophysiology.CFTR cryo-EM studies revealed two conformations: one in absence of phosphorylation and ATP with the pore-closed and NBDs separated by and the R domain sterically blocking their dimerization and the other phosphorylated with ATP bound and NBDs dimerized with two ATP molecules bound at their interface.The goal here was to elucidate the relationship between NBD dimerization and gating, understand how ion permeation is coupled to ATP hydrolysis and NBD isomerization and gain insight into the effect of disease-causing mutations and pharmacological modulation for therapeutics.The manuscript presents a detailed molecular explanation for how conformational change and channel opening are coupled and provides insight into modes for potentiator action.The authors do a commendable job of integrating multiple types of biophysical and functional data.The writing is clear (in most places), the experiments are well designed and thoughtfully interpreted.The study contributes significantly to the field.Several issues of data presentation and interpretation should be addressed.Major comments: 1) The conclusions hinge on quantitative comparisons between smFRET data and electrophysiology data.Some clarifications are needed.The properties of CFTR are compared to the unlabeled CFTRFRET in E1b-g and the fluorophore-labelled version in Fig. E1i-k and E2.For Fig. 1 and 3 it is not clear whether this is the labeled or unlabeled version.The authors should clarify this in text, Methods and legends in a consistent manner.A clear nomenclature would help.We revised the text, figure legends, and methods to clearly indicate which CFTR construct was used in every experiment.If Figs. 1 and 3 use the unlabeled version, authors should justify why.Also, authors should note if the constructs being compared differ in possessing a C-terminal GFP.Macroscopic current measurements using inside-out patches (e.g.Extended Data Figure 1b-g) were not labeled with fluorophores because efficient and specific labeling of CFTR cannot be achieved in the heterogeneous cellular membrane.C-terminally GFP-fused CFTR were used in these experiments as it has been shown previously that C-terminal fusion with GFP does not affect CFTR gating (https://bpsbioscience.com/media/wysiwyg/60506_3.pdf).In our revised manuscript, we now provide further evaluations of fluorophore-labelled CFTRFRET that support our conclusion that it recapitulates wild-type CFTR behaviour (see Extended Data Figure 1).Site-specifically labeled CFTRFRET solubilized in detergent (Figure E2a-b) hydrolyzed ATP with a rate nearly identical to that of wild-type CFTR (Figure E1h).Fluorophore-labelled CFTRFRET and wild-type CFTR (without fluorophore labels) reconstituted into synthetic planar lipid bilayers also exhibited similar current-voltage relationships, open probabilities, and responses to GLPG1837 (Figure E1i-m).In E1i-j the current-voltage relationships suggest that CFTRFRET has higher conductance (authors should indicate how many patches went into this comparison).In addition, missing is a comparison of open probability, which is critical for the interpretation.
We have now discussed the higher conductance on page 4 "Single-channel conductance of fluorophore-labelled CFTRFRET was slightly higher (Figure E1i-j), possibly due to the C343S substitution, a residue bordering the pore."This difference does not affect interpretation of singlemolecule FRET measurements, as it does not relate to the dynamics of the nucleotide-binding domains.We also noted that the data points in Extended Data Figure 1j reflect means and standard errors of 3-18 channels reconstituted in synthetic planar lipid bilayers.
Per the Reviewer's suggestion, we determined the open probability of the fluorophore-conjugated CFTRFRET variant channels and show that it is essentially identical to that of the wild-type, unlabeled CFTR (Extended Data Figure 1l).
2) Even when the reader knows that patch recordings and FRET come from different experiments, placing these side by side or one above the other could give the confusing impression that they were measured simultaneously.This confusion is aggravated by sentences like this in the text (line 206): "The W401A variant, which is capable of binding and hydrolyzing ATP via the consensus site, underwent rapid transitions between NBD-separated and -dimerized states that correlated with channel gating (Figure 2c)."This sentence should be changed (e.g. to "more closely resembled," not correlated).To further clarify this, authors should add to Authors should also define for panel f ratio between probabilities of opening and dimerization which is the numerator.
We thank the Reviewer for pointing out this confusion and we have now revised the text to clearly indicate that channel recordings and FRET were two different experiments.For example, on page 7: "Relative to wild-type CFTR, the W401A variant, which is capable of binding and hydrolyzing ATP at the consensus site, underwent comparatively rapid transitions between NBD-separated and -dimerized states that more closely resembled the dynamics of pore opening measured in electrophysiological recordings".And: "By contrast, the Y1219A variant, which binds ATP principally at the degenerate site, slowly transitioned between NBD-dimerized and separated states … single-channel measurements of the Y1219A variant exhibited only sporadic opening events".On page 8: "Here, we separately tracked the time-courses of NBD dimerization and macroscopic current increase upon application of saturating ATP (3 mM)".
As requested, we also added a panel quantifying dimerization probability of the variants (Figure 2f 3) The conclusion of Fig. 3 is that NBD dimerization precedes channel opening.The average rise upon rapid ATP perfusion of the phosphorylated channel of high FRET occupancy (dimerization) in smFRET and of channel opening in patch is faster for dimerization.Quantification was by mono-exponential fits.These should be superimposed on the data.Tau of dimerization is shown as <100 ms.The precise value and SEM should be given.As requested, we now superimposed mono-exponential fits on the individual time-courses as Extended Data Figure 7a-b in the revision.In Figure 3a we reported the means and standard errors of τopening from 42 inside-out membrane patches.We did not specify the relaxation time constant for dimerization as the measured rate may be limited by mixing time and therefore not inform on the biologically relevant transition.The fitted values for τDimerization (~100 ms) are on the same scale as measured rates of mixing (τExchange = 115 ms, see below), indicating the measured dimerization response may be limited by the rate of solvent mixing in the fluorescence microscope.For these reasons, we believe it is appropriate to report the upper bound on the rate of dimerization instead of a precise value.Fig. E7 is supposed to demonstrate that differences in kinetic behavior seen in Fig 3 are not due to differences in solute exchange between the two sets of experiments, but this figure only displays e-phys data.What is the equivalent dataset for smFRET, and where is it displayed?The Tau of patch solution exchange in Fig. E7 is 150 ms, slower than the <100 ms value given for smFRET in Fig. 3a, so it is unclear how well the exchange rates actually match.We performed additional experiments to measure the solute exchange within the microfluidics of our single-molecule imaging microscope (Extended Data Figure 7e).The estimated exchange rate is τExchange = 115 ms.We revised the text on pages 8-9 to include this new data: "The rate of channel opening (τopening = 490 ± 40 ms) (Figure 3a, E7a), was approximately 3-fold slower than the solvent exchange rate of the perfusion system (τexchange ~ 150 ms) (Figure E7c,d).By contrast, the fitted rates for NBD dimerization from FRET measurements (τDimerization ~ 100 ms) were on the same scale as the solvent exchange rate in the fluorescence microscope (τexchange = 115 ms) (Figure E7b,e).Thus, the observed delay in current activation could not be ascribed to differences in rates of mixing of the two experimental methods.We therefore conclude that the observed delay reflects conformational changes within the NBD-dimerized state that precede channel opening…" The authors state: "We conclude … that the mean first passage time between NBD dimerization and channel opening in productive instances is approximately 500 ms."However, their Tau of opening is 490 ms and they begin to see channels opening almost immediately.The delay they seek is the one between dimerization and opening and would be a smaller value.Given the imprecision of our estimate of the rate of the initial dimerization step (τDimerization < 100 ms), we cannot confidently deconvolve the individual sequential rates that lead to the pore open state.However, the Reviewer is correct that the delay time must be corrected relative to τOpening = 490 ms.We modify the text in the revision to include this consideration (page 9): "… the mean first passage time of this process is approximately 400-500 ms." Minor comments: 1) E1j is mistakenly referred to as E1k in the text (line 161) Corrected.
2) All electrophysiology traces should have current amplitude scale bar (e.g.missing in Fig. 2c) Corrected.
3) Figure 1.E1371Q should be shown without ATP as well.
We agree that addition of panels reflecting ATP-free E1371Q CFTR would nicely demonstrate retention of ATP-dependence for this variant.However, due to the slow rate of non-hydrolytic ATP dissociation we have found it difficult to isolate a homogenous population of ATP-free E1371Q CFTR molecules.Incubation with Apyrase for 1 hour at 20 ⁰C was insufficient to produce a uniformly low-FRET (NBD-separated) population (Figure R1).Nonetheless, these experiments confirm that dimerization of E1371Q is ATP-dependent and reversible, and are consistent with electrophysiological recordings.However, as we have been unable to perform 'ATP-free' measurements of E1371Q CFTR at steady state, we believe that addition of such panels to Figure 1 may confuse the reader.
4) No mass spec data to demonstrate that phosphorylation is consistent/present across all conditions.
In electrophysiological measurements protein kinase A (PKA) was applied until current amplitude saturated (see e.g. Figure 1b).This state should be consistently phosphorylated.
In cases where a real-time read-out is not available such as for ATP hydrolysis or single-molecule FRET measurements, CFTR was incubated with PKA and ATP for 30-60 minutes at room temperature.We have experimentally determined that under these conditions the response to PKAphosphorylation saturated within 1-2 minutes (Figure 1d).
Similarly, CFTR was treated with Lambda protein phosphatase for 30 minutes at room temperature to dephosphorylate the R-domain.Under these conditions the response to phosphatase also saturated within 1-2 minutes (Extended Data Figure 3r).
While a minor population of molecules may be refractory to kinase or phosphatase treatment, we would expect for this to be consistent across experiments.Mass spec quantification of the phosphorylation status in each experiment would be a significant undertaking and beyond the scope of the present study.5) Figure 2. 2cd ATP concentration and phosphorylation status should be stated in legend for clarity.It is also not so easy to determine by eye the contribution of open probability and dimerization to the coupling ratio by looking at 2e and 2f.It would help to additionally show the dimerization probability for the five different variants.
We have now specified ATP concentration to be 3 mM in the legend of Figure 2. We also add a panel that shows the dimerization probability for the nucleotide binding site variants (Figure 2f).

6) Example traces in
E1i look different at lower membrane potentials (-90 and -150)--is there a better representative trace or what is the hypothesized reason for this difference?
We have updated the single-channel traces of the current-voltage relationship in Extended Data Figure 1i to better reflect open probabilities of the CFTRFRET variant.Channel open probabilities were determined using purified CFTR reconstituted into synthetic lipid bilayers.NBD dimerization probability was determined by FRET measurements performed with purified protein.Neither system contains cellular phosphatases.
10) The authors should make clear that disease-causing CFTR variants have multiple modes/effects, and should describe the direction in which a pharmacological agent should act (i.e.promote channel opening as opposed to promoting channel closure, etc).On p. 9, the effects of G551D and L927P on channel gating should be described first, before explaining effects on FRETbased readouts.
As suggested, the channel gating and ATPase activities of the G551D and L927P mutants are now described before the FRET data.The effects of the potentiators are now specified as to "promote channel opening" (pages 10-11).11) Presence of a native lipid membrane or detergent environment may affect the dynamics of the channel.To address this, the authors perform a set of FRET experiments in proteoliposomes, which should more accurately approximate a cell membrane.The FRET dynamics and dependence on ATP look similar to what they see in detergent, bolstering confidence in their experiments.One odd thing is that only a small fraction of channels are ATP sensitive.They attribute this to channel orientation within the bilayer.If the 10xHis tag is located on the NBD side of the protein, it is unclear why any immobilized molecules should have inward-facing NBDs which are inaccessible to ATP.
Immobilization of proteoliposome-reconstituted CFTR through the C-terminal His-tag (Figure R2a) was designed to select CFTR molecules with the 'NBDs-out' orientation, as the Reviewer correctly specifies.This orientation is necessary for sensitivity to externally applied membrane impermeable reagents.
In practice, proteoliposome experiments with CFTR proved challenging for a number of reasons.First, although CFTR was reconstituted at a low protein-to-lipid ratio, aiming to produce vesicles with zero to one CFTR molecules incorporated, the sample may also contain vesicles containing more than one CFTR molecule.In these cases vesicles may be immobilized via a non-fluorophore conjugated CFTR molecule with the 'NBDs-out' orientation, while FRET is being measured by a passenger molecule with the 'NBDs-in' orientation.The proteoliposome reconstitution procedure may also negatively impact CFTR folding and function so as to produce inactive protein within the bilayer.
We also found that nearly half of our proteoliposomes tethered within microfluidic chambers were unable to be released by imidazole treatment (Figure R2b).By contrast, near-quantitative release from the surface with imidazole was demonstrated with the detergent-solubilized sample (Extended Data Figure 2g).We have previously observed that anionic lipids (e.g.POPS) may associate with Ni 2+ -coated surfaces and the POPS content of our vesicles may contribute to this non-specific association issue.The extent to which such issues relate to losses in CFTR activity in this context is not presently known.
Approximately half of the immobilized and phosphorylated CFTR channels displayed sensitivity to externally applied ATP (Figure R2c).Consistent with a permanent loss of CFTR activity associated with the proteoliposeome reconstitution or immobilization process, the responsive and non-responsive molecules did not dynamically interconvert within the time-scale of imaging.
These complications were a component of decision to focus our quantitative analysis on digitoninsolubilized CFTR.12) It is not clear if the kinetic model adds much to the discussion, but it also is presented with appropriate caveats.Does the model make a prediction in a way that could be used to test the alternative hypotheses presented in the introduction of the paper?
We acknowledge that the model is simplified and therefore cannot recapitulate all experimental observables.We believe having a kinetic model is useful to tie all experimental observables together and to guide future experiments for the following reasons: Establishing the basic topology of the CFTR gating mechanism is a fundamental goal as it provides an essential framework for kinetic modeling, the context for understanding the impact of mutations and drug action and a physical guide for establishing and testing hypotheses about function.The prevailing kinetic frameworks available to the field only included states that could be directly discerned from electrophysiological data.Parallel assessment of smFRET dynamics and electrophysiology demanded additional considerations and constraints to the available kinetic information and thus a new kinetic model had to be established that parsimoniously recapitulates the major kinetic and structure features of CFTR function revealed by both methodologies as well as the estimated ATP hydrolysis rate based on ensemble measurements.Its inclusion ties all available information together into a single kinetic framework with a defined topology and is thus an essential message of our investigation.Kinetic models of function are also critical to scientific advancement because they provide a critical framework for others to challenge and test through new experiments.For instance, our findings suggest that the kinetic model presented explains the majority of the data obtained, but that additional complexities associated with ATP binding and hydrolysis in the consensus site alone should likely be considered in future work.The kinetic framework presented also points to the need for mechanistic clarifications regarding "flicker" open/closed states that have been evidenced in the field for decades.Prior investigations have suggested that such processes reflect "flicker closing" events from otherwise open channel states that reflect clogging of the pore by solutes.In other words, these events are kinetically "off-pathway".While we cannot yet be conclusive in this regard, the kinetic framework proposed leaves open the possibility that "flicker" events may instead reflect "flicker open" events that are "on-pathway" barrier crossing attempts prior to ATP hydrolysis.Both sets of considerations, together with the identification of rate-limiting intramolecular conformational changes within CFTR that govern gating kinetics that is defined by the kinetic framework, will be vital for the field to consider in future investigations.13) Figure 1i shows that a very large percentage of channel openings have a dwell time of at least 10 ms, which could be detected using smFRET.Though it may not be feasible for this study, it would be highly convincing if it were possible to design a FRET sensor of channel pore activity that approximates the electrophysiological readout?This would allow direct comparison within a single controlled system.
We share the Reviewer's ambition to directly correlate the conformational transition of CFTR with a functional readout of pore opening.Making such experiments feasible is certainly an exciting direction to pursue in the future.

Reviewer #2 (Remarks to the Author):
The current manuscript by Levring et al. presents incredibly comprehensive studies of CFTR function as an ATP-gated ion channel; these studies include electrophysiological recordings of CFTR channel activity, real-time monitoring of the dimerization of CFTR's two nucleotide binding domains (NBDs) with smFRET imaging, biochemical measurements of ATP hydrolysis rate, and computer simulations based on a proposed gating model.Here are the main findings and their mechanistic implications: 1.The NBDs are readily dimerized by ATP binding to the two ATP-binding sites in fully phosphorylated CFTR, but the dimerized state (high FRET) is much more stable than the duration of either channel open state or the closed state at a saturating concentration of ATP.This uncoupling between NBD dimerization and channel gating suggests that multiple cycles of gate opening and closing can occur with NBDs remained dimerized (inferred from high FRET), contradicting previous idea of a strict coupling of NBD dimerization to gate opening and hydrolysis-triggered NBDs separation (low FRET) to gate closing (Csanady et al., 2010).2. By manipulating ATP binding to each site (Y1219A for the consensus site and W401A for the degenerate site), they provided evidence for NBD dimerization by ATP occupancy at either site.While both NBD dimerization and gating were mostly preserved in W401A, Y1219A channels showed robust NBD dimerization but little activity, supporting previous reports that gating of CFTR is mainly through ATP binding to the consensus site (Vergani et al., 2003;Zhou et al., 2006).3.By comparing the rate of NBD dimerization and the rate of channel activation upon sudden addition of ATP, they found that NBD dimerization takes place much faster than gate opening, contradicting previous proposition that NBD dimerization constitutes the rate-limiting step in CFTR gating (Vergani et al., 2005).Although NBD dimerization precedes gate opening, NBD dimerization itself is not sufficient to open the gate because dimerization (high FRET) can be seen in conditions when there is little channel activity (e.g., upon ATP washout, or in the presence of inhibitory concentration of ADP). 4. Supporting the idea that ATP at the consensus site plays a critical role in opening the gate, the authors demonstrated that the disease-causing mutant G551D at the consensus site exhibits defective gating and NBD dimerization.Interestingly, the dimerized NBDs in G551D show FRET between high and low FRETs, suggesting the presence of an intermediate NBD dimer proposed previously (Tsai et al., 2010).Similar but not identical intermediate FRET was seen in another pathogenic mutation L927P (in TM8), which is 50 Å away from the consensus site.This latter observation suggests a long-distance effect of the L927P mutation on NBD dimerization, but the nature of this allosteric effect is unclear.5.By testing the effects of CFTR potentiators ivacaftor and GLPG1837, they found a dramatic gating effect with minimal and moderate changes of NBD dimerization in G551D and L927P variants respectively.These are interesting observations, but it seems that not much additional insight can be gained (see below).
6.These results were put together to form the basis of a kinetic model, which was further examined through computer simulations.Although the proposed kinetic mechanism recapitulates many of the experimental data, some simulated results do deviate from actual data.No doubt, these are important studies that reveal many unexpected phenomena, which lay a solid foundation for settling the debates over the coupling mechanism between NBD dimerization and gate opening, as well as for advancing our overall understanding of the structural-functional relationships of CFTR gating.This reviewer applauds the authors for such a remarkable attainment.We thank this Reviewer for their concise, thorough, and positive summary of our investigations.Major comments: 1.The two conformations of NBDs-separated versus dimerized-can be resolved by the shift of FRET from 0.25 to 0.49 (Fig. 1).In Fig. 1e, the FRET signal remains stable at 0.49 at 10 mM ATP while the electrophysiological recording shows repeated opening and closing of the WT-CFTR channel.The authors thus concluded that the NBDs remain dimerized throughout multiple rounds of ATP binding/hydrolysis (Fig. 5).Conceptually, the consensus site in the "dimerized NBDs" must open up a space wide enough to allow the exchange of hydrolytic products for the next ATP.Indeed, in ATP washout (and Mg addition) experiments (Fig. 3b, c), transition from high FRET to low FRET was observed immediately upon ATP removal (or Mg addition).As the waiting time is too short for ATP dissociation from the degenerate site, this result suggests that upon ATP hydrolysis at the consensus site, disengagement of the two NBDs does occur.The reason this high to low FRET transition is not seen in the continuous presence of millimolar ATP may be due to rapid rebinding and re-dimerization of the disengaged NBDs at millimolar ATP.This scenario supports the notion that the state "degenerate site ATP, pore closed" should assume a different dimeric state with a cracked consensus site to allow step 6 (Fig. 5) to take place.The canonical NBD dimer in the cryo-EM structure leaves no room for ATP to diffuse into (or out of) its binding pocket.The CFTR gating cycle involves hydrolytic turnover and nucleotide exchange in the consensus ATP binding site.We agree with the Reviewer that nucleotide exchange cannot occur from the conformation adopted by phosphorylated and ATP-bound E1371Q CFTR (PDB 6MSM).In this state the ATP is occluded from solvent.The absence of observable FRET dynamics in the ATP saturated state and the monotonic increase of the dwell time of the high FRET state with ATP concentration supports that ATP may rebind without complete separation of the NBDs.We posit that small conformational changes occur after ATP hydrolysis that are sufficient to allow inorganic phosphate and ADP release while the NBD's remain ostensibly dimerized.As noted in our discussion (page 12): "The structure of the NBD-dimerized CFTR 14 suggests that only small changes at the consensus site, such as disrupting the hydrogen bond between R555 and T1246 15 , would be sufficient for inorganic phosphate, ADP release and ATP rebinding."Given the absence of additional structural information on such processes, we may only conclude that the magnitude and rate of such changes are not detectable with the specific structural perspective that CFTR folding and biochemistry enabled us to achieve.
To clarify this point we revised the text on page 6: "These findings suggest that FRET-silent processes occur within the NBD-dimerized conformation that trigger channel opening and closure and that only subtle rearrangements at the dimer interface are required for nucleotide exchange." The legend for Figure 5 was also revised to clarify the need for conformational change to allow for nucleotide exchange.See also our response for point 4a. 2. I wish the authors would have elaborated the rationale behind choosing positions 388 and 1435 for their FRET pair.T388 is at a position without a concrete secondary structure.Wouldn't this positioning result in some uncertainty in FRET measurements as the FRET efficiency is determined by not only the distance but also the relative orientation between the donor and the acceptor?It would be nice if another pair is chosen to replicate some of the key observations described above.As CFTR has intrinsic folding/stability issues, the identification of structurally informative and functionally tolerable sites of labelling was a substantial undertaking.These efforts were not elaborated in the present manuscript due to space constraints and because it was noncritical for the conclusions drawn in the manuscript.We briefly summarize the selection of labelling sites below: Based on extant structures of dephosphorylated and ATP-free wild-type CFTR (PDB 5UAK) and phosphorylated and ATP-bound E1371Q CFTR (PDB 6MSM) we measured the separation of alpha-carbons for all structurally resolved residue pairs in CFTR.Using the Förster radius of the LD555/LD655 FRET pair, we predicted FRET efficiencies in both conformational states, as well as ΔFRET upon conformational isomerization (Figure R3).From residue pairs with favourable predicted FRET efficiencies, we selected a subset of residues with solvent-exposed sidechains to ensure efficient labelling with maleimide-conjugated fluorophores.Residues mutated in cystic fibrosis patients or reported to affect folding or gating in the literature were excluded.
A panel (40) of cysteine-substituted CFTR variants were tested for whether they expressed, folded, site-specifically labelled with maleimide-conjugated fluorophores, and retained key functional properties of the wild-type channel.Several (5) pairs were experimentally tested for interpretable FRET responses.This led to the final choice of labelling positions 388 and 1435.While T388 is not involved in secondary structure, the density corresponding to the backbone of this residue is well-defined in maps of both the ATP-free and dephosphorylated CFTR, as well as the ATP-bound and phosphorylated E1371Q CFTR.Consistent with this, the 388/1435 pair exhibited tight distributions of FRET efficiencies in conditions corresponding to those used for structure determination.3. The W401A and Y1219A mutations were designed to weaken ATP binding.The authors observed that the NBDs of W401A and Y1219A mutants can still dimerize, and thus concluded that ATP binding at either site is sufficient for NBD dimerization (P7, line 215).Such conclusion is based on the assumption that the degenerate site in W401A (or the consensus site in Y1219A) must be vacant at 3 mM ATP.However, in electrophysiological study, the Y1219G mutation reduces the apparent affinity for ATP to a K1/2 of ~5 mM (Zhou et al., 2006).If we assume that Y1219A affects the affinity similarly, at least a portion of the Y1219A channel population in Fig. 2 would still have ATP bound to the consensus site at 3 mM ATP.Given that the degenerate site has much higher affinity for ATP, it is unlikely that the degenerate site in W401A would be vacant at 3 mM ATP.Thus, the data in Fig. 2 only support that NBDs of W401A and Y1219A can dimerize.The occupancy of ATP at each site cannot be determined, and thus whether a single ATP is enough to trigger dimerization cannot be inferred conclusively.However, the conclusion may not be incorrect once data in Fig. 1j are taken into consideration.The ATP concentration dependence of high to low FRET transition rate suggests the existence of different states of NBD dimer with one or both sites occupied.We agree with the Reviewer that the mutations may not completely abolish ATP binding and we revised the text as that the mutations were designed "to reduce the affinity for ATP" and "the Y1219A variant, which binds ATP principally at the degenerate site" (page 7).However, the fraction of such a dual ATP-occupied mutants is small and its presence will not alter the conclusions we draw, for the following reasons: 1.The bulk ATPase activity of the Y1219A variant at 3 mM ATP is reduced to only 15 % of the wild-type CFTR (Figure 2b).This indicates that the Y1219A substitution nearly abolished ATP binding at the consensus site.2. The open probability of the Y1219A variant is reduced to nearly zero, also suggesting that the substitution nearly abolished ATP binding (Figure 2f). 3. The NBD-dimerization probability of the Y1219A variant is approximately 50 % of the wild-type level, consistent with ATP binding being perturbed.Yet the relative reduction in dimerization probability is much less than the effects on ATP hydrolysis and pore opening.These observations are consistent with the degenerate site ATP supporting dimerization events, that in the absence ATP occupancy in the consensus site do not lead to pore opening or ATP hydrolysis (Figure 2e). 4. Most importantly, the FRET and gating dynamics of the mutants are completely different from that of wild-type CFTR, indicating the mutated site is mostly vacant under the experimental conditions (Figure 2c). 5.As the Reviewer points out, data in Figure 1i lends further support that NBD may dimerize with one or both sites occupied.4. Some of the kinetic steps in Fig. 5 need justification/explanation: a. Dissociation/binding of ATP in the degenerate site happens when the NBDs are separated (step 7, 8), but dissociation of hydrolytic products and binding of ATP to the consensus site happen when the NBDs are dimerized (step 4, 5, 6).Some intermediate conformations of NBDs need to exist.We agree with the Reviewer that intermediate conformations of NBD separation must exist.Because we cannot explicitly measure transitions of CFTR to intermediate states or estimate the lifetimes of inferred intermediates we have not included them in our simplified kinetic scheme in Figure 5.
To clarify this point we revised the legend: "ATP rebinding may occur with subtle rearrangement at the dimer interface (step 6) or with complete NBD-separation (step 8) to initiate a new gating cycle."b.Why step 3 is drawn differently in Fig. 5 and Fig. E11?Also, why did the authors propose only the open state hydrolyzes ATP, but not 2 ATP pore flicker-closed state or 2 ATP pore closed state?
In Figure 5 we deliberately draw step 3 from bracketed channel conformations as we do not have evidence allowing us to determine which of '2 ATP pore flicker-closed' and '2 ATP pore open' states are catalytically competent.We are also unaware of literature that might distinguish between models allowing hydrolysis from both or either states.One might speculate that flicker closures are local to the pore and therefore do not affect hydrolysis.As the rates of exchange between these states is much greater than the rates into or out of these states, this ambiguity has negligible impact on our kinetic simulations as direct connectivity to either state yielded indistinguishable results.For the simulations presented, we explicitly specified the connectivity of the underlying model by allowing only hydrolysis from the on-pathway, '2 ATP pore open' state.We altered the depiction of the kinetic scheme in Extended Data Figure 11

accordingly. [REDACTED]
To our best knowledge, there is no evidence to support that ATP hydrolysis occurs without pore opening.In our kinetic scheme, we aim only to outline the dominant gating path, deliberately ignoring rare events, even those that we know to occur.c.Is the flicker-closed state the same as the cryo-EM structure of E1371Q-hCFTR?It's been known that many of the flickery events seen at negative membrane potentials are not likely real functional "state".Because of their high voltage dependence (see Fig. E1i.), many of the events are considered blocking of the pore by cytosolic large anions.To get the right kinetic parameter for the true flicker-closed state, one has to analyze data from recordings at positive membrane potentials.The molecular nature of flicker-closure events, the origin of their voltage-dependence, and whether the structure of human E1371Q CFTR (PDB 6MSM) reflects a flicker-closed state has yet to be elucidated to our knowledge.As our kinetic scheme allows ambiguity in this regard, we believe that detailed discussions on this specific point lie beyond the scope of the current study.For the [REDACTED] purpose of our gating model and simulations, we aim only to recapitulate the empirical observation of flicker-closure events during the pre-hydrolytic period of the open burst.d.There is a mistake both in the text and in the figure legend.Steps 6 -8 represent transitions among closed states; therefore, it is not a new gating cycle, which by definition should include a closed event and an open event.We thank the Reviewer for pointing this out.We mean to state that ADP dissociation (step 5) leads to a dynamic intermediate that may isomerize (step 7) and to which ATP may bind to NBDseparated (step 8) or NBD-dimerized states (step 6).A new gating cycle is initiated after ATP binding (steps 1 and 2).We do not consider steps 6-8 a separate gating cycle.We have revised the text as (page 13): "Dissociation of ADP (Step 5) results in a dynamic intermediate to which ATP can rebind (steps 6-8) thereby initiating another gating cycle."and the legend of Figure 5 as "ADP dissociation (step 5) leads to a dynamically isomerizing intermediate (step 7).ATP rebinding may occur with subtle rearrangement at the dimer interface (step 6) or with complete NBD-separation (step 8) to initiate a new gating cycle."Minor comments: 1.I am not sure I truly understand the meaning of the term "hierarchical" gating mechanism.We use this word to emphasize our finding that NBD dimerization does not necessarily lead to pore opening.Pore opening requires conformational changes within the NBD-dimerized channel, governed by ATP hydrolysis.2. P2, line 3: Do you mean electrophysiological properties?Corrected.3. P5, line 140: It should be "…to a stably NBD-dimerized state."Corrected.4. P6, line 157: Previous electrophysiological studies of R117H-CFTR and zebrafish CFTR, as well as the cryo-EM structure of E1371Q all indicate the presence of conductive and nonconductive states with dimerized NBDs.Please more explicitly state the unique insight from these data.
As the Reviewer has pointed out earlier, whether NBD dimerization is strictly coupled to gate opening has been a long-debated topic.There are reports in the literature supporting both arguments.Here we address this question by directly comparing the kinetics of NBD dimerization with channel gating.Furthermore, our finding that CFTR spends most of the time in the NBDdimerized state is original.5. P6, line 161: …with ATP concentration (Figure 1j), not 1k.
Corrected.6. P6, line 172: The ADP experiments are interesting in that the low FRET value seems different from that without ATP.Is it reproducible?If it is, doesn't that suggest different degrees of disengagement of NBDs?
The data is reproducible from three independent experiments.To investigate if this reflects global or local conformational changes of the NBDs would require cryo-EM analyses that are beyond the scope of this study.7. P7, line 191: The time constant for ligand exchange at the degenerate site is around 30 s (Tsai et al., 2010), which interestingly is very similar to the duration of high FRET upon ATP washout (Fig. 3b).
Please see our response to point 10 below.8. P7, line 226: It should be open probability, not opening probability.(In other parts of the text too.)Corrected here and throughout the text.9. P8, line 237: The changes of FRET upon sudden addition of ATP show heterogeneity.This observation may be related to a recent paper by Yeh et al. (2021).The stated similarity to W401A and double exponential current rising upon addition of ATP (Fig. 1b and E1b) all remind this reviewer of the results published in this paper, which shows different closed channel conformations (perhaps different NBDs conformations) once ATP is removed.We thank the Reviewer for pointing this out.The different configurations of closed states described by Yeh et al. (2021) are clearly evident in our ATP washout experiments (Figure 3b) where molecules first release nucleotide from the consensus site, but retain ATP in the degenerate site which supports dimerization (Extended Data Figure 8g).Reversible rundown may correspond to release of degenerate site ATP leading to stably separated NBDs.We have now added the following sentence on page 9: "Dissociation of ATP from both sites likely leads to the reversible rundown of CFTR currents that occurs after prolonged exposure to nucleotide-free solutions 30 " The observed heterogeneity in the ATP injection experiment (Figure 3a) may have a different origin.At time of ATP injection CFTR molecules had been in ATP-free buffer for a prolonged period and both ATP binding sites should be vacant.To us, the simplest origin of heterogeneity arises from stochastic and sequential ATP binding to the individual binding sites (Figure R5).That is, the population is homogenous at time of injection.CFTR molecules that bind ATP in the consensus site first, may dimerize, hydrolyze ATP, release products, and relax back to the starting point.Repeated cycles of this may occur.Upon ATP binding in the degenerate site the NBDs dimerize stably.Eventually, every molecule reaches steady state with both binding sites simultaneously occupied.10.P8, line 255: Of note, rapid removal of ATP caused a double exponential decay of WT-CFTR currents (Fig. 2B in Lin et al., 2014) with time constants (τ1 < 1 s and τ2 =29 s) similar to these values reported for NBD dimerization relaxation.These shouldn't be just coincident.We agree.The apparent correlation between the biexponential time constants for relaxation of pore opening and dimerization suggests a common underlying molecular mechanism determining both transitions.
To us, the simplest explanation is that CFTR relaxes in two steps associated with release of ATP from either site.Hydrolysis-catalyzed release from the consensus site is fast.Dissociation from the degenerate site is slow.The intermediate with ATP in the degenerate site alone behaves like the Y1219A CFTR variant, exhibiting high dimerization probability but low channel open probability.This interpretation is in line with conclusions made by Lin et al (2014).
The biphasic current relaxation upon nucleotide exchange reported by Tsai et al., 2010 (Reviewer comment 7) is also consistent with this model.Further, we note that the reduced rate of the slow exchange component reported by Tsai et al (τ2 = 51 s) compared to the slow decay component measured in washout experiments (τ2 ≈ 20-30 s) is consistent with our observations.During the second phase of washout, CFTR dynamically isomerizes between dimerized and separated conformations (Figure 3b).When NBDs separate, both the consensus and degenerate site are able to readily release nucleotide.In the ligand exchange experiment continuous presence of nucleotide stabilizes the dimerized conformation thereby reducing the rate of exchange at the degenerate site.
We revised the text to reference the correlation with the cited manuscripts (page 9): "Parallel FRET experiments showed that the time course of NBD-separation is biphasic, with time constants of 1.6 seconds and 20 seconds, respectively (Figure 3b).These rates correlate with the double exponential time constants reported for CFTR current decay and ligand exchange 18,29 .This apparent correlation suggests a common underlying molecular mechanism determining both transitions."11.P9, line 277: Not a correct reference and interpretation.The cited ligand exchange experiments lead to two different turn-over rates of ATP and the idea of a "partial dimer."On the other hand, by carrying out different ligand exchange experiments, Jih et al. (2012) propose a re-entry pathway featuring repetitive cycles of ATP turnover in a single opening burst.Both studies done by Hwang and colleagues invoke a partial dimer state where the consensus site is open up to allow replenish of ATP.They do not suggest that ATP turnover can occur in a single canonical NBD-dimerized conformation, where the ligand is occluded.We thank the Reviewer for pointing out the differences in the models.The shared concept of the present inferences and the cited manuscript by Tsai et al., is that during wild-type CFTR gating at steady state, the NBDs do not fully separate for each hydrolysis event and open burst.Rather they remain associated through continuous ATP binding in the degenerate site.Nonetheless, nucleotide exchange at the consensus site necessarily requires some degree of separation (see also discussion above).Tsai et al. propose the formation of a partial dimer.By contrast, we do not detect changes in FRET efficiency between repeated gating cycles and therefore infer that the changes required for nucleotide exchange must be subtle in magnitude and/or transient in time.We revise the text to reflect this (pages 9-10): "These findings suggest that the probability of ATP rebinding exceeds that of complete NBD separation when ATP concentration is greater than 100 µM.This concept, consistent with ligand exchange experiments 18 , suggests that repetitive cycles of ATP turnover can occur in an ostensibly NBD-dimerized conformation with only subtle changes at the consensus site required for nucleotide exchange.In cellular settings, repetitive gating cycles are therefore expected to persist until the finite rate of NBD separation at cellular ATP concentrations allows the dephosphorylated R-domain to reinsert, terminating CFTR gating."12. P10, line 312-314: I don't understand what the authors mean by the similar nature in "the similar nature of both mutations' impacts on NBD dimerization and channel opening".G551D and L927P actually exhibit different behaviors in their experiments.In Fig. 4b, the transitions between high and low FRET in G551D is different from those of L927P.Also, GLPG1837 drastically changes L927P's FRET pattern, but barely affects G551D.Moreover, electrophysiological properties of these two mutants and pharmacological effects of CFTR potentiators should be discussed in a more quantitative manner (see comment 18 below) before one can make more definitive conclusions.This section has been extensively revised based on this comment and Reviewer 1's comments.Thank you.13.P11, line 347: It should be noted that both VX-770 and GLPG1837 work on a CFTR variant with its NBD2 completely removed (Yeh et al., 2017;Yeh et al., 2015).These drugs also exert dramatical gating effects on G551D whose NBDs may not assume a canonical NBD dimer state (Lin et al., 2014;Yeh et al., 2017) (Fig. 4 in current study).
Thank you for pointing this out.We have revised the text to include this information.On page 12: "These data lead to the conclusion that the major effect of Ivacaftor or GLPG1837 is not to support transition from NBD-separated to dimerized conformations.Rather, these potentiators principally operate by promoting pore opening in CFTR when NBDs are already dimerized.In other words, potentiators affect the coupling efficiency between NBD dimerization and channel opening, possibly by stabilizing the TMDs in the pore open configuration 43 .This effect also manifests in variants unable to form a canonical NBD dimer, such as G551D and a variant devoid of the entire NBD2 40 ."14.P11, line 375-377: Previous studies suggest that reagents such as NPPB and high-affinity ATP analogs, act synergistically with Ivacaftor and GLPG1837 on G551D channels by promoting NBD dimerization (Lin et al., 2016;Yeh et al., 2017).The authors may consider incorporating these studies in the Discussion.
We included a discussion of the investigational compound NPPB in the revision (page 12): "The investigational compound 5-nitro-2-(3-phenylpropylamino) benzoate (NPPB) was proposed to stimulate pore opening by such a mechanism 44 ."15. Figure 1 Spelling has been unified as "dimerization".16. Figure 1e: If I understand correctly, the electrophysiological recordings are from WT-CFTR, not from the CFTRFRET protein.Wouldn't it be a better comparison to use the electrophysiological recording of CFTRFRET protein here?Of note, the CFTRFRET construct altered the majority of the endogenous cysteine residues; this alteration could affect CFTR gating as seen in the literature.In Figure E1, the authors demonstrated their ability to use CFTRFRET proteins in electrophysiological experiments and concluded that its gating kinetics are similar to WT-CFTR (but see comment 19 below).
Yes, Figure 1e displays electrophysiological recordings of wild-type CFTR.The comparison is justified by measurements made in Extended Data Figure 1, which demonstrate that the CFTRFRET and wild-type variants have similar time-courses of current relaxation in response to rapid ATP application or withdrawal.Dependences on ATP and PKA, and sensitivity to GLPG1837 are also similar.Finally, ATP hydrolysis, which reflects flux through the gating cycle is also comparable between the two variants.
Consistent with literature reports (e.g.Cui et al. 2006 andMense et al. 2006), we found that substituting all cysteines in CFTR to serine or alanine caused expression defects that precluded purification.We therefore conducted an extensive screen of cysteine variants to isolate residues causing this unwanted effect.Residues C524 and C590 are buried within NBD1 and could be retained without a great extent of non-specific fluorophore labelling.C76L and C592M substitutions were found to not affect expression.Collectively, these changes to the cysteine-less CFTR variant allowed purification.
While cysteine substitutions may cause subtle changes to gating, our observations seem incompatible with large gating defects that would alter the major conclusions drawn in this manuscript.
To further substantiate that key gating properties are retained by the CFTRFRET variant we add a comparison of open probabilities and open dwell times for the wild-type CFTR and the fluorophore-conjugated CFTRFRET to the revision (Extended Data Figure 1l,m).Cui, L., et al.The role of cystic fibrosis transmembrane conductance regulator phenylalanine 508 side chain in ion channel gating.J Physiol 572, 347-358 (2006).
Mense, M., et al.In vivo phosphorylation of CFTR promotes formation of a nucleotide-binding domain heterodimer.EMBO J 25, 4728-4739 (2006).17. Figure 1j: I don't understand the data points at 0 ATP.Do the authors still observe FRET transitions even without ATP, meaning that the NBDs dimerize without ATP?In 1e, FRET always stays at low FRET.NBD-dimerization in the absence of ATP does occur, albeit infrequently (see e.g. Figure 1e -end of the no ATP trace and Extended Data Figure 10).We also observe infrequent pore opening in the absence of ATP (Extended Data Figure 10) consistent with literature reports (e.g.Wang et al. 2010 andMihályi et al. 2016).Both dimerization and pore opening also occur with Apyrase, which was added to remove potential ATP contamination.
Our working model is that CFTR gating, like other stochastic processes in biology, is facilitated by ATP binding and NBD dimerization, but can occur independently with very low efficiency and duration with finite probability.Wang, W., et al.ATP-independent CFTR channel gating and allosteric modulation by phosphorylation.Proc Natl Acad Sci USA 107, 3888-3893 (2010).
Mihályi, C., Töröcsik, B., & Csanády, L. Obligate coupling of CFTR pore opening to tight nucleotide-binding domain dimerization.eLife 5, e18164 (2016).18. Figure 4b: In response to GLPG1837, the current trace of G551D barely changes as if GLPG1837 has little effect.This potentiation effect is far from the reported ~30-fold stimulation shown in Fig. 4g.Please comment on the discrepancy and consider replacing the recording with a more representative one.Similar concerns are raised for the effect of GLPG1837 on L927P.The single-channel activity in 4b does not change 20-fold as indicated in Fig. 4g.The open probability of G551D is very low and gating is slow.Therefore, in any given 60 second window we are likely to see zero or one open events.Neither is truly representative.We have updated the current trace in Figure 4b to better represent the relative current stimulation with GLPG1837.
The window of time presented for the L927P trace is also updated in Figure 4b of the revision.The traces reflect the relative potentiation presented in Figure 4g.Because of the very short open dwell time and only modestly affected closed dwell time of unpotentiated L927P CFTR, the open probability may be perceived greater in sample traces than it is.19. Figure 4f: It is shown that the Po of E1371Q is ~0.7 and increases to ~0.8 with GLPG1837.It is puzzling how E1371Q has a Po of only 0.7.Numerous studies showed that with hydrolysis eliminated, the closing rate of E1371Q (or E1371S) is >100-fold slower than WT, resulting in a Po close to unity (Bompadre et al., 2005;Csanady et al., 2013;Vergani et al., 2003;Yu et al., 2016).Indeed, the E1371Q single channel in Figure 1f seems to be almost always open, though it is hard to determine whether those brief downward deflections are flicker closures or interburst closures with this time scale.It would be helpful to expand an E1371Q single channel trace to allow a better assessment of channel behavior in a quantitative manner.The deflections are flicker-closureshowever, the duration of flickers make them hard to distinguish from inter-burst closures.Consistent with literature, we determine very slow current relaxation for E1371Q upon ATP withdrawal in excised inside-out patches: τclosure = 427 ± 91 s (mean and standard error, n = 6 patches) (Extended Data Figure 8a).Thus, the vast majority of the observed closures do not involve release of ATP from the consensus site.We add an expanded view of the E1371Q single-channel trace to Figure 1e of the revision.20. Figure E1i: From the raw trace, the Po of the FRET construct is apparently higher than the WT.The open burst time of FRET variant is visibly longer than WT.Is this a consistent finding?Please comment on the discrepancy.We have updated the single-channel traces of the current-voltage relationship in Extended Data Figure 1i to better reflect the open probability of the CFTRFRET variant.We also add a comparison of open probabilities and open dwell times for the phosphorylated wild-type and fluorophoreconjugated CFTRFRET variant channels (Extended Data Figure 1l-m).We do not find a statistically significant increase in open probability for the CFTRFRET variant.21. Figure E8g: What is the second arrow below the ADP+Pi for?The multiple arrows aimed to illustrate that several molecular events occur within this step.The second arrow is removed in the revision to avoid confusion.22. Figure E11b, d, and e: In b, the simulated single-channel trace shows a much longer open time than the experiment, while the closed time seems not affected much (or slightly shorter).However, in d and e, the experimental and simulated dwell times match perfectly but the two closed times deviate significantly.Please comment on this discrepancy.The open burst dwell times of the experimental and simulated traces in Extended Data Figure 11b match, as is also reflected in the quantification in Extended Data Figure 11d.However, as we do not have an accurate estimate of flicker-closure rates, this may give rise to the different appearances of the open burst in experimental and simulated traces.Due to the limited bandwidth of our experiments, we ignore flicker-closure events in idealization of experimental single-channel recordings.The dwell time histogram therefore reflect the life-time of the open burst.To best mimic the experimental analysis, the dwell times of the open burst were also extracted from simulated traces, again ignoring flicker-closure events.In our revised manuscript we clarify these points in the legend of Extended Data Figure 11.Yeh, H. I., Yeh, J. T., & Hwang, T. C. (2015).Modulation of CFTR gating by permeant ions.J Gen Physiol, 145 (1), 47-60.https://doi.org/10.1085/jgp.201411272Yeh, H. I., Yu, Y. C., Kuo, P. L., Tsai, C. K., Huang, H. T., & Hwang, T. C. (2021).Functional stability of CFTR depends on tight binding of ATP at its degenerate ATP-binding site.J Physiol, 599(20), 4625-4642.https://doi.org/10.1113/JP281933Yu, Y. C., Sohma, Y., & Hwang, T. C. (2016).On the mechanism of gating defects caused by the R117H mutation in cystic fibrosis transmembrane conductance regulator.J Physiol, 594(12), 3227-3244.https://doi.org/10.1113/jp271723Zhou, Z., Wang, X., Liu, H. Y., Zou, X., Li, M., & Hwang, T. C. (2006).The two ATP binding sites of cystic fibrosis transmembrane conductance regulator (CFTR) play distinct roles in gating kinetics and energetics.J Gen Physiol, 128(4), 413-422.https://doi.org/10.1085/jgp.200609622

Reviewer #3 (Remarks to the Author):
The manuscript by Levring et al describes characterization of molecular cycles involved in activation of the CFTR channels.Specifically, single molecule FRET studies done in parallel with single channel electrophysiology enable to characterize the relationship between ATP binding, hydrolysis and channel opening.Furthermore, how disease-causing mutations and therapeutic correctors modulates these events are studied and discussed.The study is very nicely crafted, the experiments are very well performed, and the conclusions are convincing.This work brings a clear view on a long-standing issue in the CF field and deserves publication.We thank this Reviewer for these supportive summary comments.Nevertheless, a number of points require clarification.1/ FRET values CryoEM structures of human CFTR have -so far-captured 2 conformations with the NBDs either separated or dimerized.Structural data on ABC structures suggest more extended conformational diversity with several different substates.The differences in FRET values observed in the present manuscript may be indicative of such diversity.Therefore, the authors need to clearly assign -when possible-the link between the FRET valued and the identified conformer.The ATP-free CFTR shows a low FRET state at 0.25.Does this match the cryo-EM structures of the NBD separated state ?Similarly does the 0.49 FRET value match the NBD-dimerized state for the used probes?The authors need to run careful simulations (for examples with FPS, see Kalinin et al) with the appropriate probe models to evaluate the calculated FRET value for each conformation.
As requested, we performed all-atom molecular dynamics simulations of CFTR (PDB codes 5UAK and 6MSM) with the LD555 and LD655 dyes attached at their sites of labeling for comparison with the FRET experiments.As previously described, we utilized a simplified potential that maintains native contacts in the original structure while allowing the fluorophores to explore all possible positions and conformations (Noel et al., 2016 andGirodat et al., 2020).Using this approach, we can confirm that our FRET efficiencies are consistent with the existing CryoEM structures of the two conformations (Table R1).For reference, we also included the distances obtained from FPS as suggested by the Reviewer (Kalinin et al., 2012 andSindbert et al., 2011), which show larger differences, which we attribute to the simplified nature of the ball-and-stick model used by the FPS program.ATP-bound dephosphorylated CFTR shows a 0.28 FRET value.How does that translate in terms of distance change given the R0 for these probes ?Regarding the ATP-bound dephosphorylated state (0.28 FRET efficiency), this corresponds to an inter-dye distance of 72.6 Å, which is a ~2 Å decrease from the apo state.
2/ The E1371Q mutant appear to be locked in the high FRET state during the time course of the experiments, at least in the trace shown (100s, Fig1f).However, the distribution on Fig1c seems to show a significant low FRET population (~5-10%), pretty much identical to that seen for the "wt" CFTR (which can hydrolyze ATP and thus should spend more time in the NBD separated).The fraction of E1371Q molecules that does not occupy the high FRET state reflects two populations: 1) Infrequent NBD-separation events from the high FRET state (Figure R6a); 2) A small fraction of nonresponsive molecules.The latter population does not respond to stimuli like ATP addition and might reflect molecules that have denatured during handling, surface tethering, and/or the small fraction of non-specifically fluorophore-labelled molecules (Extended Data Figure 2a).
In that line, it is unclear to me why the authors see no NBD separation in this mutant after phosphatase treatment (Fig E3).One may expect to see some separation over the course of the measurement.
The NBD separation of E1371Q in response to phosphatase does occur, but exceedingly slow.The zoomed-in view of Extended Data Figure 3r reveals a subtle inflection in the dimerized state occupancy of E1371Q CFTR upon phosphatase injection (Figure R6b).The response is too slow to accurately estimate a time-constant in our smFRET setup.
Our experiments suggest that autoinhibitory re-engagement of the R domain requires prior NBDseparation.NBD-separation is limited by the rate of hydrolytic or non-hydrolytic ATP dissociation from wild-type and E1371Q CFTR variants, respectively.Our best estimate of non-hydrolytic ATP dissociation is from E1371Q current relaxation upon ATP withdrawal measured in inside-out excised patches (Extended Data Figure 8a).We have measured the time constant of this reaction to be τclosure = 427 ± 91 s (mean and standard error, n = 6).A similar value was reported by David Gadsby and colleagues (Vergani et al. 2005): τclosure = 476 s.
In the continued presence of ATP, non-hydrolytic ATP dissociation from dephosphorylated E1371Q CFTR may lead to either R domain autoinhibition or ATP rebinding.Thus, 450 seconds should be a lower bound on the predicted time-constant of relaxation.Altogether, this is consistent with the slow rate of NBD separation measured in Extended Data Figure 3r.Due to space-constraints we did not describe the origin of the asymmetry between ADP-mediated inhibition of pore opening and NBD-dimerization.Our working model (Figure R7) posits that ADP competition for the consensus site is sufficient to prevent opening of the pore, but is insufficient to cause separation of NBDs.Thus, at intermediate ADP concentrations CFTR channels are trapped in a configuration equivalent to the post-hydrolytic state which is also transited during normal gating (Figure 5).At higher concentration, ADP competes for both nucleotide-binding sites to inhibit both NBD-dimerization and pore opening.What does the sentence "While the fraction of the molecules responsive to ATP was significantly reduced" refer to ?Only single traces are shown.No statistics ("significantly" ?).
We refer to our response to Reviewer 1 point 11.The text was revised to remove the word "significantly".4/ ATP binding sites The authors use the W401A and Y1219A mutants originally described by Hwang and colleagues (although mutated into Gly in that study).What do the authors mean by "destabilize ATP binding" ?It should be noted that the mutations decrease affinity but may not prevent it.For example, W401A is still able to bind ATP (albeit whith lower affinity) and 3mM is quite a large concentration.Perturbating the ATP binding at the degenerate site may also alter the conformational equilibrium of NBD1 that we recently described (Scholl et al, 2021).On P8 the authors state that for mutant Y1219A "ATP binds only at the degenerate site".This is not proven.We agree with the Reviewer that the mutations may not completely abolish ATP binding and we revised the text as the mutations used were designed "to reduce the affinity for ATP" and "the Y1219A variant, which binds ATP principally at the degenerate site" (page 7).However, the fraction of such a dual ATP-occupied mutants is small and its presence will not alter the conclusions we draw.Please see our response to Reviewer 2, point 3. On P7, the authors describe that "the Y1219A variant, which binds ATP mainly at the degenerate site, transitioned between NBD-dimerized and separated states slowly, and very rarely exhibited channel opening" but they only show single traces.They should show distributions.As suggested, the distributions of NBD dimerization and channel opening are now included as Figure 2e and 2f I assume that the last part of the graph has 0mM ADP (correct?).
Yes, at the end of the current trace in Extended Data Figure 3k there is no ADP.
Why is the rebound much faster than what is observed during the first 2 mins?
We think that this is related to the 'reversible rundown' effect described by Hwang and colleagues (Yeh et al. 2021).Hwang and colleagues describe a slow component for the current activation after ATP application, that appears after extended intervals without ATP.The effect is ascribed to release of ATP from the degenerate ATP binding site.Our model for competitive inhibition described in Figure R7 predicts that the ADP-dependent current inhibition observed in Extended Data Figure 3k would not involve ATP dissociation from the degenerate site, and thus that no reversible rundown should occur.Therefore, a slow component of activation is evident upon initial ATP application, but not upon ADP withdrawal.The definition of coupling ratio is clarified in the revised legend of Figure 2 as: "open probability divided by dimerization probability."In interpreting the coupling ratio it is important to note that NBD dimerization and pore opening correlateat least for wild-type CFTR.In Figures 1d,g we demonstrate correlation between current amplitude and NBD-dimerization probability in the timecourses of activation and in the ATP dose-dependences for wild-type CFTR.The coupling ratio is then the scalar that relates open probability and dimerization probability.
Whilst not essential for the interpretations made in the manuscript, the ratio may be most easily understood by making a simplifying assumption: NBD-dimerization is required for pore opening.In support of this assumption, we note that across all variants and conditions that we have tested, the probability of dimerization exceeds the probability of opening.Under this assumption, the coupling ratio reflects: What is the probability of pore opening, given that the NBDs have dimerized?We find that ATP occupancy in the consensus site is crucial for the coupling, disease mutations weaken coupling, and potentiators partially rescue coupling.7/ P7 the authors state "We infer from these observations that channel opening probability is enhanced by interactions of CFTR with the terminal phosphate moiety of the ATP molecule bound within the consensus site.".This should be readily testable as the terminal phosphate at the consensus site is coordinated by K1250 and T1246.Point mutations should support this statement.
Effects of substituting the Walker A lysine of the consensus site (K1250) have been reported in the literature.Consistent with our inferences and analogous to the E1371Q variant studied in the present manuscript, substitution of K1250 results in CFTR variants that bind ATP but lose hydrolysis activity, thereby trapping CFTR in dramatically prolonged open bursts (Gunderson andKopito, 1995 andZeltwanger et al. 1999).Like E1371Q, these variants likely mimic the prehydrolytic state of wild-type CFTR bound to two ATP molecules.
While T1246 coordinates the terminal phosphate moiety of the consensus site ATP, the side-chain is also engaged in an important hydrogen bond with R555 across the NBD interface.Disruption of the NBD dimer interface was demonstrated in the literature by T1246N substitution (Vergani et al. 2005).Deconvolving the dual effects of substitution at this site is likely to be difficult.

8/ Mutation G551D
The authors show an intermediate FRET value of 0.37.The authors suggest that the mutation introduce a "steric and electrostatic repulsion at the NBD interface" which is very vague.An Asp at position 551 will likely prevent binding of ATP at the consensus site (at least in a NBD dimerize state) because the acidic side chain would directly and closely face the terminal phosphate.This suggests that the NBD1-NBD2 dimer is not formed at the consensus site and that the protein is adopting an intermediate/different conformation.This also suggest that the ATP "stimulation" is due to dimerization at the degenerate site.Therefore the "dimerized dwell time" on Fig E9E may actually not correspond to dimerized state as "expected", ie.where both sites are closed, as observed by Cryo EM.We agree and revised the manuscript.Page 10: "The high FRET, NBD-dimerized conformation is likely to be different from that of E1371Q CFTR previously observed by cryo-EM, evident by a lower coupling ratio (Figure 4e) and a shorter life-time (Extended Data Figure 9e,f).In agreement with these data, a recent cryo-EM study showed that the G551D variant adopts conformations in between the extremes observed for fully NBD-separated (PDB 5UAK) anddimerized (PDB 6MSM) states 35   bioRxiv 2022.10.10.510913;doi: https://doi.org/10.1101/2022.10.10.510913 (2022).
In addition, the statement on P10 that "Importantly, both the G551D and L927P variants exhibited ATP-dependent FRET response indicative of wild-type ATP binding affinities" is hard to reconciliate with the position of D551 in the mutant The FRET response of G551D refers to formation of the intermediate conformation, in which ATP binds in the NBD partially-separated conformation, where residue 551 likely does not contribute to binding.We revised the manuscript (pages 10-11): "For both G551D and L927P variants, FRET transitions exhibited ATP-dependence indicative of wild-type ATP binding affinities (Figure E9ik).Their functional defects are caused by deficits in ATP effecting formation of a tight NBD dimer and in the coupling of the allosteric processes within NBD-dimerized CFTR that give rise to channel opening (Figure 4e)."Furthermore, the effect of potentiators on G551D are not clear, and not discussed.
Ivacaftor was approved to treat G551D patients, we would expect to see an effect.The authors show some individual traces with GLPG1837, but not the approved drug Ivacaftor (why?) from which there seem to be no significant effect.The authors should explore and or discuss this.In contrast, stimulation of the L927P mutation does lead to fully dimerized state which does seem to correlate with increase channel opening (although Fig 4f shows quite modest improvement).
GLPG1837 and Ivacaftor both have large effects on G551D pore opening and share a common binding site (Liu et al. 2019), and mechanism of action (Yeh, et al. 2017).While Ivacaftor is clinically approved, electrophysiologists (including us) favor GLPG1837 because Ivacaftor is very hydrophobic, making it difficult to determine its effective concentration, which has led to some inconsistent reporting for this compound (Csanády and Töröcsik, 2019).As Ivacaftor and GLPG1837 share the same mechanism of action, lessons learned from one compound can be applied to the other.
Our smFRET measurements show that GLPG1837 does not rescue the defect of G551D dimerization.Consistently, Hunt and colleagues (Wang et al. 2022) described a lack of an effect with Ivacaftor in their structural studies of a CFTR variant harboring the G551D substitution.
While GLPG1837 does promote formation of the dimer for the L927P CFTR variant, the key observation is that the relative stimulation of open probability greatly exceeds the relative stimulation of dimerization probability (Figure 4f,g).We conclude that the major effect of potentiators is not to support transition from the separated to dimerized conformation.Rather, potentiator binding acts allosterically on molecules that have already formed an NBD dimer by favoring conformational changes that couple dimerization to pore opening.
As suggested, we have revised the text to incorporate these discussions (page 11): "Consistent with previous reports 40-43 , we observed that both potentiators induced marked increases of channel open probabilities (Figures 4b,f,E10a).By comparison their effects on NBD dimerization were much smaller for all CFTR variants tested (Figures 4b,f, E10a).For example, GLPG1837 increased the open probability of the G551D variant by more than 30-fold, while the change in NBD dimerization was marginal (Figure 4b).This observation, together with the recent cryo-EM study of the G551D CFTR variant in the presence of Ivacaftor 35 , demonstrates that neither Ivacaftor nor GLPG1837 promotes NBD dimerization.Similarly, for the L927P variant, the relative stimulation of open probability greatly exceeded the relative stimulation of dimerization probability (Figure 4f,g Fig 2 a panel showing dimerization probability and compare open and dimerized dwell times the way they did in WT channels in Fig 1.

15. Figure 1 :
I suggest the authors unify the spelling of "dimerization."It is dimerization in the text but dimerisation in Fig 1 and Fig 4.
Fig 2 a panel showing dimerization probability and compare open and dimerized dwell times the way they did in WT channels in Fig 1.
), and a panel comparing the open and NBD-dimerized dwell times for the W401A CFTR variant (Figure 2d).The Y1219A variant has an open probably of nearly zero and gates slowly, preventing us to quantify its open dwell time.The definition of coupling ratio is now clarified in the Figure 2 legend as: "open probability divided by dimerization probability."

Figure
Figure R1 | Nucleotide-free E1371Q CFTR a. Contour plots of phosphorylated E1371Q CFTR.CFTR was incubated with 3 mM ATP (left), treated with Apyrase for 1 hour at 20 ⁰C in nucleotide-free buffer (middle), and 3 mM ATP was then reintroduced (right).b-c.Example single-molecule traces for E1371Q CFTR in ATP-free buffer (b) and upon rapid delivery of 3 mM ATP (c).
7) Analogous examples of current from WT channels should be shown for comparison in E1b and E1k.Example current traces of wild-type CFTR have been added to Extended Data Figure1b and 1k.8) Line 140 NDB typoCorrected.9) A question regarding comparison of electrophysiology and FRET data-are there any integral membrane phosphatases which could contribute to the observed differences between channel opening and NBD dimerization probability?

Figure
Figure R2 | Proteoliposome-reconstituted CFTR a. Schematic drawing of the immobilization strategy for proteoliposome-reconstituted CFTR.b.Quantification of the specificity of His-tag-dependent immobilization.c.Contour plots showing the response of proteoliposome-reconstituted CFTR molecules to rapid application of 3 mM ATP. Proteoliposomes were kinase-treated before the experiments.

Figure R3 |
Figure R3 | Residue selection.Heat map of predicted changes in FRET efficiencies upon conformational isomerization for all possible pairs of labelling in CFTR.

Figure R5 |
Figure R5 | Sequential ATP binding to consensus and degenerate binding sites in the pre-steady state.Schematic of the events occurring after rapid ATP application to PKA-phosphorylated CFTR.If ATP is bound at the consensus site first, CFTR enters Gating cycle A.The NBDs of CFTR dimerize, the pore opens, ATP is hydrolyzed, and catalytic products are released.Repeated cycles may occur.Every passage through Gating cycle A involves complete NBD dimerization and separation.Eventually, CFTR molecules will bind ATP in both sites simultaneously, and gate at steady state (Gating cycle B).Gating cycle B involves only subtle rearrangements at the NBD interface.
: I suggest the authors unify the spelling of "dimerization."It is dimerization in the text but dimerisation in Fig 1 and Fig 4.

Figure
Figure R6 | E1371Q CFTR NBD-separation.a. Example single-molecule trace showing a rare NBD separation event (indicated by an asterisk) for E1371Q CFTR.CFTR was phosphorylated and in the presence of 3 mM ATP. Top panel shows donor (green) and acceptor (red) fluorescence intensities.Bottom panel shows FRET.b.Depopulation from the High FRET state after λ phosphatase injection (at the dashed line) for E1371Q CFTR.The plot is a zoomed-in view of Extended Data Figure 3r in the manuscript.

Figure R7 |
Figure R7 | Competitive inhibition by ADP.At lower concentrations, ADP competes for ATP binding in the consensus site to inhibit pore opening but not NBD dimerization.Higher ADP concentrations compete for ATP binding in the degenerate site to inhibit NBD dimerization.
. 5/ Figure E3K is unclear to me.Is it expected to see such a significant effect at 0.1 and 0.3 mM ADP while 3mM ATP is maintained (and while Fig E3I show predominantly NBD in dimer form)?Based on the literature, we do expect to observe ADP inhibition under those experimental conditions.The dose-dependence of competitive ADP inhibition described in Figure E3K is consistent with measurements by Welsh and colleagues (Randak and Welsh 2005).Randak, C. O. & Welsh, M. J. ADP inhibits function of the ABC transporter cystic fibrosis transmembrane conductance regulator via its adenylate kinase activity.Proc Natl Acad Sci USA 102, 2216-2220 (2005).
." Wang C., et al.Mechanism of dual pharmacological correction and potentiation of human CFTR.