Tracing the substrate translocation mechanism in P-glycoprotein

Reviewer #1 (Public Review):

Summary:
Here the authors have tethered a Pgp substrate to strategically place cysteine residues in the protein. Notably, the cysteine-linked substrate (ANC-DNPT)- stimulates ATP hydrolyse and so is able to undergo IF to OF transitions. The authors then determined cryo-EM structures of these complexes and MD simulations of bound states. By capturing unforeseen OF conformations with substate they propose that TM1 undergoes local conformational changes that are sufficient to translocate substrates, rather than large bundle movements.

Strengths:
This paper provides the first substrate (ANC-DNPT)- bound conformations of PgP and a new mechanistic model of how substrates are translocated.

Weaknesses:
Although the cross-links stimulate ATP hydrolysis, further controls are needed to convince me that the TM1 conformations observed in the structures are physiologically relevant, since they have been trapped by "large" substrates covalently-tethered by cross-links.

Reviewer #2 (Public Review):

Summary:
The manuscript by Gewering and coworkers is an elegant mechanistic investigation of the mammalian multidrug transporter Pgp. I will not elaborate on the significance of this protein except to point out its clinical involvement in cancer resistance to chemotherapy.

Strengths:
The strengths of the investigation are partly in the combination of sophisticated chemical synthesis, state-of-the-art cryoEM in a well-established biochemical context. What is more exciting is the tackling of a long-standing question in the field: namely how do drugs make their way through the structure to be exported across the membrane? Unfortunately, the field has been stuck in hand waving model based on structures that in the outward-facing conformations are devoid of substrates. The work challenges the dogma that emerged from this hand-waving model and presents an alternative model that appears to be supported by the data.

Weaknesses:
There is much to like about the experimental work here but I am less sanguine on the interpretation. The main idea is to covalently link via disulfide bonds a model tripeptide substrate under different conditions that mimic transport and then image the resulting conformations. The choice of the Pgp cysteine mutants here is critical but also poses questions regarding the interpretation. What seems to be missing, or not reported, is a series of control experiments for further cysteine mutations.

Reviewer #3 (Public Review):

Summary:
The authors employed a new strategy, covalent substrate-labeling, to address the open issue of the substrate transport mechanism by single particle cryogenic-EM. A cyclic peptide (QZ-Ala), which was already used in the past as a substrate for structural purposes, was modified and covalently attached to ABCB1 at strategic positions in the transmembrane domain via Cys-specific coupling chemistry. Overall, four mutations (two per TMD) were generated and functionally analyzed. These residues are in proximity to the QZ-Ala binding site and are labeled by verapamil. Interestingly, two mutants could only be labeled if ATP and Mg2+ were present.

Strengths:
Three of the four mutants were structurally analyzed by single particle cryo-EM with structures in both, the inward- and outward-facing conformation and overall resolutions ranging from 2.6 to 4.3 Å. Applying multi-model analyses allowed for the extraction of additional structures from one data set. These structures formed the basis for a detailed analysis of the substrate translocation pathway. This enabled the researcher to compare the IF and OF states of the same mutant and same substrate, which is pivotal for their conclusions. The mutations 335/978 trap the substrate at different points of the translocation pathway, while 971 located two helical turns away from the first set, trapped the system at a later stage. The described strategy revealed a cascade of conformational changes during substrate transport which focus on TMH1, which is straight in the IF state, but swings out in the OF state. Pivotal for such a change is G72. This residue was mutated to Ala and also structurally analyzed. These structures were supported by MD simulations and functional data. Thus, the new prosed kinking and straightening mechanisms are different from the so far accepted wide-open OF state observed in bacterial transporters. This clearly suggests a different mechanism for ABCB1.

Weaknesses:
I have a couple of minor issues that I have listed in the section recommendations for authors Overall, the manuscript is very well written, sheds new light on the molecular mechanism of substrate translocation by ABCB1, and might even provide a new starting point for inhibitor design. I know that it is unusual, but I like the manuscript in its current version and recommend acceptance in its current form.

Author Response

Reviewer 1 (Public Review):

Weakness: Although the cross-links stimulate ATP hydrolysis, further controls are needed to convince me that the TM1 conformations observed in the structures are physiologically relevant, since they have been trapped by "large" substrates covalently-tethered by crosslinks.

Reviewer 1 raised concerns about the relatively large size of our covalently attached AAC substrate that would potentially distort TM1 in Pgp. We would like to clarify that AAC has a molecular weight of 462 Da, which, in comparison to many known Pgp substrates ranging from 250 to over 1,000 Da, is not a large compound. For instance, the few other Pgp substrates mentioned in our manuscript all have a comparable or larger size: verapamil, 455 Da; doxorubicin, 544 Da; FK506, 804 Da; valinomycin, 1,111 Da; cyclosporin A, 1,203 Da.

Furthermore, AAC was strategically attached to a site distant from TM1 in the inwardfacing Pgp conformation. After it was exported to the outward-facing state, several TM helices accommodate the compound. The observation that only TM1 exhibited significant conformational changes suggests its potential role in the transport mechanism. This hypothesis is supported by our findings, where a conservative substitution (G72A) in TM1 resulted in a dramatic loss of transport function for various drug substrates and impaired verapamil-stimulated ATPase activity.

Reviewer 1 (Recommendations for the Authors):

I understand the need for an unconventional approach to understanding the translocation pathway. What would help to support this model is to cross-link a much smaller substrate, as the one used is quite large and could potentially distort TM1 in the outward-state when cross-linked.

We thank the reviewer for this recommendation, and we have outlined plans for future experiments involving other substrates, including smaller ones, to further investigate our proposed model. However, it is important to acknowledge that conducting these studies will require a significant amount of effort and resources, which we believe extend beyond the scope of our current manuscript.

In unbiased MD simulations starting from the IF state are there any simulations where the substrate follows the same path as proposed here?

All our MD simulations were performed in the outward-facing state to focus on potential substrate release pathways. Starting MD simulations from the inwardfacing state would introduce complexities in capturing the necessary domain motions and nucleotide binding and hydrolysis required for substrate translocations. Therefore, we opted not to perform MD studies starting from the inward-facing state.

Reviewer 2 (Public Review):

Weakness: There is much to like about the experimental work here but I am less sanguine on the interpretation. The main idea is to covalently link via disulfide bonds a model tripeptide substrate under different conditions that mimic transport and then image the resulting conformations. The choice of the Pgp cysteine mutants here is critical but also poses questions regarding the interpretation. What seems to be missing, or not reported, is a series of control experiments for further cysteine mutations.

Reviewer 2 raised concerns about the interpretation of our results and suggested the need for additional mutant designs to validate our proposed TM1 mechanism. Firstly, we believe that the observed TM1 conformational changes are valid in our cryoEM structures, despite the use of different conditions and several mutants to capture Pgp in the outward-facing state.

Regarding the G72A mutant, we consider it conclusive that this single point mutation in the TM1 has a profound effect. Importantly, the G72A mutant was readily expressed and purifiable as a stable protein. We were able to resolve a high-resolution structure of the G72A mutant (without the substrate), confirming that the protein is not generally destabilized but properly folded.

Above all, we appreciate the Reviewer’s suggestion to explore additional mutations and intend to do so in future studies.

Reviewer 2 (Recommendations for the Authors):

I am sold on the results regarding TM1 conformational changes as they are evident in the cryoEM structures. However, the set of states compared between mutants are not biochemically equivalent: for 335 and 978 they used an ATP-impaired Pgp whereas for 971 they used what appears to be WT, and the conformation was imaged presumably subsequent to ATP hydrolysis and Vanadate trapping. This is significant if the authors were unable to trap the OF in the impaired mutant background and should be highlighted. I have to believe that they tried that condition but I could be wrong.

We acknowledge the point made by the Reviewer about the biochemical equivalence of mutant states and the potential significance of using an ATP-impaired mutant for trapping the outward-facing conformation of 971. We have not yet attempted to use the ATPase-deficient 971C mutant for crosslinking and intend to address this question in future studies.

In our current approach, we used the ATPase-active 971C for two specific reasons:

1. Our biochemistry data, as shown in Fig 1C, indicates that 971C only crosslinks in the presence of ATP hydrolysis conditions. Vanadate trapping was employed to stabilize the outward-facing conformation.

2. Based on our experience, we have observed that the conformations of ATP-bound (mutant) and vanadate-trapped states of an ABC transporter are structurally equivalent at this resolution level of our study (see ref. 21: Hoffmann et al. NATURE 2019).

The authors propose a new model for substrate translocation. It is based on three mutants and a number of structures. If the authors were not challenging the current dogma I would not have written the next comment. Considering the impact of the findings, I would have designed a couple more cysteine mutants based on their model. For instance, this pathway has a number of stabilizing interactions, can't they make a mutant that preserves conformational switching but eliminates substrate translocation? I like the G97A mutant result but I am worried that the effect could just be a general destabilization or misfolding as part of the cryoEM particles seem to suggest. The authors advance one interpretation of the disorder observed in this mutant but it could easily be my interpretation.

We thank the reviewer for the suggestion to design additional mutants to further validate our proposed model for substrate translocation. We agree that this would be highly valuable, considering the potential impact of our findings. However, given the time-intensive nature of our approach, we believe that presenting these additional designs in a future study is a reasonable course of action.

Regarding the G72A mutation, we believe that our current data fully supports our model and the role of TM1 in regulating the Pgp activity. Importantly, we would like to emphasize that the G72A mutant was readily expressed and purifiable as a stable protein. Additionally, our cryoEM structural determination of the G72A mutant at high resolution confirmed that the protein is not generally destabilized but properly folded.

There are a couple of troubling methodological questions that I want the authors to address or clarify:

1- In the methods they report that the final sample for cryoEM was prepared on a SEC devoid of detergent. It is obvious that the sample was folded but I was wondering why the detergent was removed? Was that critical for observing these structures with multiple ligands? Did they observe any lipids in their cryoEM?

We avoid detergent in the buffer on final SEC purification. This step is to remove free detergent from the background which helps during cryoEM imaging. Of course, this cannot be done with every detergent but due to the very low CMC of LMNG it is possible. By now, we have verified this method for several other transporters with the same success. While this procedure helps us to obtain better images it is not necessary to obtain specific conformations or ligand bound states, nor does it affect these states or conformations.

In our cryoEM structures , we did observe multiple cholesterol hemisuccinate (CHS) molecules on the outer transmembrane surface of Pgp.

2- Can the authors comment on why labeling was carried out in the presence of ATP? Does it matter if the substrate was added prior to ATP and incubated for a few minutes?

For every dataset, we first added the substrate to be cross-linked and afterwards added the ATP. In the cases of 335C and 978C, labeling was successful before ATP was added, as evidenced by the inward-facing structures with cross-linked substrate.

However, for 971C, cross-linking only occurred after the addition of ATP. We interpret this data to suggest that the 971 site is inaccessible to the substrate in the inward-facing state, and cross-linking can only occur after the transporter transitions to outward-facing state. This is in line with our inward-facing structure which does not show a cross-linked substrate, and our biochemical data shown in Fig 1C, where 971C only crosslinked in the presence of ATP.

3- I am not an expert on MD simulations and I understand that carrying out simulations at higher temperatures used to be a trick to accelerate the process. Is this still necessary? Why didn't the author use approaches such as WESTPA?

Most so-called enhanced sampling methods, including WESTPA, explicitly define a reaction coordinate for the process of interest, usually based on intuition or prior studies. If this coordinate is chosen poorly, enhanced sampling usually fails, either because the sampling becomes inefficient or because the sampling biases the transition pathway (or both). Lacking reliable intuition or prior knowledge on which motions would result in substrate release, we chose temperature to speed up the process. High temperature largely avoids the introduction of an any bias through the definition of a progress coordinate. By contrast, the weighted ensemble method underlying WESTPA is a great method to simulate unbiased dynamics of a process with a known progress coordinate, but unfortunately requires to choose a progress coordinate prior to the simulation and will then mostly sample the process along this progress coordinate, because this is the only direction in which sampling is improved. High temperature MD on the other hand accelerates all processes in the system under study. Indeed, we have now confirmed that the pathway found at high temperature is also feasible at near-ambient conditions.

In new simulations, we have now observed a similar release pathway at T=330 K. As the only difference, the substrate has not fully dissociated from the protein after 2.5 us, with weak interactions persisting at the top part of TM1 from the extracellular side. Importantly, this is a configuration observed also in higher temperature simulations but with much shorter lifetime.

In response, we will include these new findings in the revised manuscript.

4- One way to show that the two substrates binding mode is biochemically relevant is to measure Vmax at different substrate concentrations. One would expect a cooperative transition if that interaction is mechanistically important.

We have measured Vmax as a function of QZ-Ala concentration in a previous report (ref. 24), supporting positive cooperativity for binding to two sites.

Reviewer 3 (Public Review and Recommendations for the Authors):

We thank Reviewer 3 for recommending the acceptance of our manuscript as is. We will address all minor comments from Reviewer 3 in the revised manuscript.