Conformational Changes of the ABC Transporter BmrA Depend on Membrane Curvature

Reviewer #1 (Public review):

Summary:

This study uses single-molecule FRET to analyze the conformational ensemble of an ABC transporter at different temperatures, with different substrate analogs, and under different membrane curvatures (i.e., two populations of vesicles with different radii). The authors combine this data into a general model that describes the influence of membrane curvature on membrane protein conformation.

Strengths:

This interesting and quantitative work uses detailed FRET measurements at two different temperatures and in the presence of substrate and two substrate analogs to tease out the energetic contribution of membrane curvature in the conformational change of an ABC transporter. The mechanistic model distinguishes between equilibrium conditions (non-hydrolyzable ATP analog) and steady-state conditions (ATP analog), and describes the data well. The authors are careful with the experimental measurement of the liposome size distribution and perform appropriate controls to ensure it is maintained throughout the experiment.

Weaknesses:

An important aspect of this paper is the difference in mechanism between inhibitors AMP-PNP (a substrate analog) and vanadate (together with ADP, forms a transition state analog inhibitor). The mechanisms and inhibitory constants/binding affinities of these inhibitors are not very well-supported in the current form of the manuscript, either through citations or through experiments. Related to this, the interpretation of the different curvature response of BmrA in the presence of vanadate vs AMPPNP is not very clear.

Overall, the energetic contribution of the membrane curvature is subtle (less than a kT), so while the principles seem generalizable among membrane proteins, whether these principles impact transport or cell physiology remains to be established.

Reviewer #2 (Public review):

Summary:

Membrane transport proteins function by the alternating access model in which a central substrate binding site is alternately exposed to the soluble phase on either side of the membrane. For many members of the ABC transporter family, the transport cycle involves conformational isomerization between an outward-facing V-shaped conformation and an inward-facing Λ-shaped conformation. In the present manuscript, it is hypothesized that the difference in free energy between these conformational states depends on the radius of curvature of the membrane and hence, that transport activity can be modulated by this parameter.

To test this, BmrA, a multidrug exporter in Bacillus subtilis, was reconstituted into spherical proteoliposomes of different diameters and hence different radii of curvature. By measuring flux through the ATP turnover cycle in an enzymatic assay and conformational isomerization by single-molecule FRET, the authors argue that the activity of BmrA can be experimentally manipulated by altering the radius of curvature of the membrane. Flux through the transport cycle was found to be reduced at high membrane curvature. It is proposed that the potential to modulate transport flux through membrane curvature may allow ABC transporters to act as mechanosensors by analogy to mechanosensitive ion channels such as the Piezo channels and K2P channels.

Although an interesting methodology is established, additional experimentation and analyses would be required to support the major claims of the manuscript.

Strengths:

Mechanosensitivity of proteins is an understudied phenomenon, in part due to a scarcity of methods to study the activity of proteins in response to mechanical stimuli in purified systems. Useful experimental and theoretical frameworks are established to address the hypothesis, which potentially could have implications for a large class of membrane proteins. The tested hypothesis for the mechanosensitivity of the BmrA transporter is intuitive and compelling.

Weaknesses and comments:

(1) Although this study may be considered as a purely biophysical investigation of the sensitivity of an ABC transporter to mechanical perturbation of the membrane, the impact would be strengthened if a physiological rationale for this mode of regulation were discussed. Many factors, including temperature, pH, ionic strength, or membrane potential, are likely to affect flux through the transport cycle to some extent, without justifying describing BmrA as a sensor for changes in any of these. Indeed, a much stronger dependence on temperature than on membrane curvature was measured. It is not clear what radii of curvature BmrA would normally be exposed to, and whether this range of curvatures corresponds to the range at which modulation of transport activity could occur. Similarly, it is not clear what biological condition would involve a substantial change to membrane curvature or tension that would necessitate altered BmrA activity.

(2) The size distributions of vesicles were estimated by cryoEM. However, grid blotting leaves a very thin layer of vitreous ice that could sterically exclude large vesicles, leading to a systematic underestimation of the vesicle size distribution.

(3) The relative difference in ATP turnover rates for BmrA in small versus large vesicles is modest (~2-fold) and could arise from different success rates of functional reconstitution with the different protocols.

(4) The conformational state of the NBDs of BmrA was measured by smFRET imaging. Several aspects of these investigations could be improved or clarified. Firstly, the inclusion and exclusion criteria for individual molecules should be more quantitatively described in the methods. Secondly, errors were estimated by bootstrapping. Given the small differences in state occupancies between conditions, true replicates and statistical tests would better establish confidence in their significance. Thirdly, it is concerning that very few convincing dynamic transitions between states were observed. This may in part be due to fast photobleaching compared to the rate of isomerization, but this could be overcome by reducing the imaging frequency and illumination power. Alternatively, several labs have established the ability to exchange solution during imaging to thereby monitor the change in FRET distribution as a ligand is delivered or removed. Visualizing dynamic and reversible responses to ligands would greatly bolster confidence in the condition-dependent changes in FRET distributions. Such pre-steady state experiments would also allow direct comparison of the kinetics of isomerization from the inward-facing to the outward-facing conformation on delivery of ATP between small and large vesicles.

(5) A key observation is that BmrA was more prone to isomerize ATP- or AMP-PNP-dependently to the outward-facing conformations in large vesicles. Surprisingly, the same was not observed with vanadate-trapping, although the sensitivity of state occupancy to membrane curvature would be predicted to be greatest when state occupancies of both inward- and outward-facing states are close to 50%. It is argued that this was due to irreversibility of vanadate-trapping, but both vanadate and AMP-PNP should work fully reversibly on ABC transporters (see e.g. PMID: 7512348 for vanadate). Further, if trapping were fully irreversible, a quantitative shift to the outward-facing condition would be predicted.

Reviewer #3 (Public review):

Summary:

The manuscript explores the dependence of ABC transporter activity on membrane curvature. The underlying concept being analysed here is whether membrane mechanics can regulate the conformation of the protein and thereby its activity.

Strengths:

The protein of choice here is BmrA, a bacterial transmembrane ABC transporter. This protein was previously found to exhibit two states: open conformation with Nucleotide Binding domains (NBDs) separated from each other and an ATP-bound closed conformation with dimerised NBDs. The protein was purified and reconstituted into liposomes of varying diameters, largely categorised as Small vesicles (SV) and Large vesicles (LV). The authors find that the activity of the protein is reduced with the changing curvature of the membrane vesicles used to make the proteoliposomes. This could be modulated by making vesicles at different temperatures, LV at high and SV at lower temperature (4 {degree sign}C), following which they perform biochemical measurement of activity or smFRET experiments at HT or RT. They use well-characterized single-molecule FRET-based measurements to assess the change in conformation of the protein during the ATPase cycle. They find that a significant fraction of the protein is in an open (inactive) conformation in vesicles of higher curvature (SVs) at a given temperature. The authors develop a simple yet elegant theoretical model based on the energy of protein configuration states and their coupling to membrane energetics (bending rigidity) and curvature to explain these findings. The model provides a parameter-free fit that predicts the open/closed state distributions as well as the ATPase activity differences between SV and LV. Using experimentally determined values of the protein conicity, the authors to extract reasonable values of membrane rigidity, consistent with available literature.

The data and theoretical model together convincingly support the claim that membrane mechanics via local curvature modulation may bias membrane protein conformation states and thereby modify the activity of membrane proteins. This is an important and general conclusion that the authors also elaborate on in their discussion.

Weaknesses:

The authors say that the protein activity is irreversibly inhibited by orthovanadate, but 50% of the proteins are still in open conformation, while being accessible to the analogue (Table 2). It is unclear what this means in the context of activity vs. conformation.

The difference in the fraction of proteins in closed conformation is quite similar between LV and SV treated with AMP-PNP at 20 {degree sign}C (Figure 2B), and it is not clear if the difference is significant. The presence of a much higher FRET tail in the plots of smFRET experiment in SVs at 20 {degree sign}C or 33 {degree sign}C in the apo conformation of the protein (Figure 3A-B) is cause of some concern since one would not expect BmrA to access the closed states more frequently in the Apo conformation especially when incorporated in the SV. This is because the subtraction of the higher fraction of closed states in the Apo conformation contributes directly to enhancing the bias between the closed states in SV versus LV membrane bilayers.

Author response:

Global answer about the ATP analogs (concerns the 3 reviewers)

We use ATP-Vanadate essentially for detecting the FRET efficiency for the closed state. But these data are not included in our theoretical model. Thus, even if the comments of the reviewers on the observation of a non-negligible fraction of proteins in the open state in the presence of ATP-vanadate are justified, this has no consequence on our conclusions on the effect of curvature on BmrA on the conformational changes with ATP or AMP-PNP.

We agree with the comments of the reviewers that the binding of vanadate is not irreversible, but the reported lifetime of the closed state is very long compared to our experimental conditions (see (Urbatsch et al. JBC (1995)) on PgP).

Nevertheless, we will perform new experiments independent of ATP analogs using the E504A BmrA mutant. It has been shown structurally and enzymatically to bind and not hydrolyze ATP and to be 100% in a closed conformation at 5 mM ATP (A. Gobet et al., Nat. Commun. 16, 1745 (2025)). It will clear up all doubts about our experiments.

We will also add new references:

I. L. Urbatsch, B. Sankaran, J. Weber, A. E. Senior, J. Biol. Chem. 270, 19383 (1995)

T. Baukrowitz, T.-C. Hwang, A. C. Nairn, D. C. Gadsby, Neuron 12, 473 (1994)

A. Gobet et al., Nat. Commun. 16, 1745 (2025)

Y. Liu, M. Liao, Sci. Adv. 11, eadv9721 (2025) (on the effect of vanadate and temperature on a plant ABC)

Public Reviews:

Reviewer #1 (Public review):

(1) An important aspect of this paper is the difference in mechanism between inhibitors AMP-PNP (a substrate analog) and vanadate (together with ADP, forms a transition state analog inhibitor). The mechanisms and inhibitory constants/binding affinities of these inhibitors are not very well-supported in the current form of the manuscript, either through citations or through experiments. Related to this, the interpretation of the different curvature response of BmrA in the presence of vanadate vs AMPPNP is not very clear.

See the global answer about ATP-analogs (above)

(2) Overall, the energetic contribution of the membrane curvature is subtle (less than a kT), so while the principles seem generalizable among membrane proteins, whether these principles impact transport or cell physiology remains to be established.

This is correct that the effect is limited to high curvature in the case of BmrA. Our theoretical model allows predictions for different protein parameters. The effect is particularly dependent on the protein size and on protein conicity, which can vary over a wide range. We show that larger proteins, such as piezo 1 are in principle expected to display a much stronger curvature dependence than BmrA. But testing our predictions on other proteins and on their physiological function is indeed an exciting perspective but beyond the objective of the current manuscript.

Reviewer #2 (Public review):

(1) Although this study may be considered as a purely biophysical investigation of the sensitivity of an ABC transporter to mechanical perturbation of the membrane, the impact would be strengthened if a physiological rationale for this mode of regulation were discussed. Many factors, including temperature, pH, ionic strength, or membrane potential, are likely to affect flux through the transport cycle to some extent, without justifying describing BmrA as a sensor for changes in any of these. Indeed, a much stronger dependence on temperature than on membrane curvature was measured. It is not clear what radii of curvature BmrA would normally be exposed to, and whether this range of curvatures corresponds to the range at which modulation of transport activity could occur. Similarly, it is not clear what biological condition would involve a substantial change to membrane curvature or tension that would necessitate altered BmrA activity.

Reviewers 1 and 2 both stressed that we showed that activity and conformational changes are mechanosensitive, not that the function of the protein is to be a mechanosensor. This will be corrected.

Regarding the physiological relevance of the mechanosensitivity of BmrA, we have addressed this point in the manuscript (bottom of page 10 and top of page 11). This discussion was positively appreciated by Reviewer #3. We stress that we have used BmrA as a model system, but considering our results and the theoretical model, we can predict the parameters that are relevant for future studies on the sensitivity of other transmembrane proteins to membrane mechanical properties. And, as stated by the reviewer, "mechanosensitivity of proteins is an understudied phenomenon".

(2) The size distributions of vesicles were estimated by cryoEM. However, grid blotting leaves a very thin layer of vitreous ice that could sterically exclude large vesicles, leading to a systematic underestimation of the vesicle size distribution.

We used Lacey carbon grids with large mesh size ranges for our cryoEM images, and we blot on the backside, precisely to measure the largest size range accessible to cryoEM. In our hands, this was not the case when using Quantifoil or C-Flat grids with uniform hole sizes and a large fraction of carbon where the vesicles adhere. With our grids, we are able to image vesicles from 20 to 200 nm diameter and the precision on the diameter is high, but the statistics might not be as good as with DLS or other diffusion-based methods. DLS is an indirect method (as compared to cryoEM) to measure vesicle size distribution, that may overestimate the fraction of large objects and underestimate the small ones. We will perform DLS experiments for comparison purpose.

(3) The relative difference in ATP turnover rates for BmrA in small versus large vesicles is modest (~2-fold) and could arise from different success rates of functional reconstitution with the different protocols.

The ATPase activity is sensitive to several parameters. We thus carefully characterized our reconstituted samples, including ATPase activity, yield of incorporation and orientation of proteins that are often reported. In addition, we showed by cryo-EM the unilamellarity of the proteoliposomes and their stability during the experiments, which were never reported. The ATPase activity of our samples reconstituted in liposomes at 20 ° and at 4°C are high, among the highest reported for BmrA, and less sensitive to errors as compared to the low activities in micelles of detergent.

We would also like to stress that with our protocol, we have prepared the same batch of lipid/protein mixture that we have split it 2 for the reconstitution at 4°C and 20°C conversely. Both preparations contain the same amount of detergent. The only difference is that we include more BioBeads for the preparation at 4°C to account for the difference of absorption of the detergent on the beads at low temperature (D. Lévy, A. Bluzat, M. Seigneuret, J.L. Rigaud Biochim. Biophys. Acta. 179 (1990)), but we also showed that the proteins do not adsorb on the BioBeads (J.-L. Rigaud, B. Pitard, D. Levy, Biochim. Biophys. Acta 1231, 223 (1995)). In addition, the activity of the protein at 37°C is high and comparable to those reported in the literature (E. Steinfels et al., Biochemistry 43, 7491 (2004)., W. Mi et al., Nature 549, 233 (2017).), which speaks for a good functional reconstitution. Finally, our results are consistent between the smFRET where we have only one protein maximum per vesicle and the activity measurements where the amount of protein is higher.

We also performed reconstitution from molar LPR= 1:13600 to 1:1700 and found the same activity per protein, confirming that the proteins are functional, independently of their surface fraction. We will add these data in the revision.

Altogether, these data suggest that we correctly estimate the rate of functional reconstitution in our experiments.

Nevertheless, we will design additional experiments to further compare the activity of the proteins before and after reconstitution.

(4) The conformational state of the NBDs of BmrA was measured by smFRET imaging. Several aspects of these investigations could be improved or clarified. Firstly, the inclusion and exclusion criteria for individual molecules should be more quantitatively described in the methods. Secondly, errors were estimated by bootstrapping. Given the small differences in state occupancies between conditions, true replicates and statistical tests would better establish confidence in their significance. Thirdly, it is concerning that very few convincing dynamic transitions between states were observed. This may in part be due to fast photobleaching compared to the rate of isomerization, but this could be overcome by reducing the imaging frequency and illumination power. Alternatively, several labs have established the ability to exchange solution during imaging to thereby monitor the change in FRET distribution as a ligand is delivered or removed. Visualizing dynamic and reversible responses to ligands would greatly bolster confidence in the condition-dependent changes in FRET distributions. Such pre-steady state experiments would also allow direct comparison of the kinetics of isomerization from the inward-facing to the outward-facing conformation on delivery of ATP between small and large vesicles.

(a) We will better detail the inclusion and exclusion criteria.

(b) For the smFRET, we have performed N=3 true replicates. We will add statistical tests on our graphs.

(c) We will detail more how we have optimized our illumination protocol, considering the signal to noise ratio and the photobleaching. Practically, we cannot add ATP to our sealed observation chamber on our TIRF system to detect dynamical changes on our immobilized liposomes. The experiment suggested by the reviewer would imply to build a flow chamber to exchange the medium around immobilized liposomes, compatible with TIRF microscopy. This is an excellent idea, which has been achieved only recently (S. N. Lefebvre, M. Nijland, I. Maslov, D. J. Slotboom, Nat. Commun. 16, 4448 (2025)). It will require a full new study to optimize both the flow chamber and the dyes to track the smFRET changes over long periods of time.

Nevertheless, we would like to stress that our objective is not to study the dynamics of the conformational changes, and that we expect it to be slow for BmrA, even at 33°C.

(5) A key observation is that BmrA was more prone to isomerize ATP- or AMP-PNP-dependently to the outward-facing conformations in large vesicles. Surprisingly, the same was not observed with vanadate-trapping, although the sensitivity of state occupancy to membrane curvature would be predicted to be greatest when state occupancies of both inward- and outward-facing states are close to 50%. It is argued that this was due to irreversibility of vanadate-trapping, but both vanadate and AMP-PNP should work fully reversibly on ABC transporters (see e.g. PMID: 7512348 for vanadate). Further, if trapping were fully irreversible, a quantitative shift to the outward-facing condition would be predicted.

See the global answer about ATP-analogs (above)

Reviewer #3 (Public review):

(1) The authors say that the protein activity is irreversibly inhibited by orthovanadate, but 50% of the proteins are still in open conformation, while being accessible to the analogue (Table 2). It is unclear what this means in the context of activity vs. conformation.

See the global answer about ATP-analogs (above)

(2) The difference in the fraction of proteins in closed conformation is quite similar between LV and SV treated with AMP-PNP at 20 {degree sign}C (Figure 2B), and it is not clear if the difference is significant. The presence of a much higher FRET tail in the plots of smFRET experiment in SVs at 20 {degree sign}C or 33 {degree sign}C in the apo conformation of the protein (Figure 3A-B) is cause of some concern since one would not expect BmrA to access the closed states more frequently in the Apo conformation especially when incorporated in the SV. This is because the subtraction of the higher fraction of closed states in the Apo conformation contributes directly to enhancing the bias between the closed states in SV versus LV membrane bilayers.

We have consistently observed, both at 20°C and at 33°C, a fraction of proteins with a high FRET signal in our measurements, higher in SV (about 15% and 17%) than in LV (about 10% and 6%). We have quantified the fraction of proteins with NBDs facing inside the liposomes (page 5), 20% in LV and 23.85% in SV. Considering the inverted curvature of the membrane, this orientation could favor the closed conformation, even in the absence of ATP, more for SV than LV. The fraction with inverted orientation could explain our higher fraction of high FRET signal in SV.

Moreover, for part of it, it can be due to a fraction of proteins with a non-specific labeling that would produce a higher FRET signal. We will add data with Cys-less mutants showing that less than 4% are labeled.