Author Response
The authors' responses to the public reviews can be found here
The following is the authors’ response to the most recent recommendations.
Recommendations for the authors:
Reviewer #2 (Recommendations For The Authors):
I appreciate the effort that the authors have put into this revised version of the manuscript. Before going into details, I would suggest that, in the future, the authors include enough information in their response to allow reviewers to follow the changes made. Not simply "Fixed", but instead "we have modified the description of these results and now state on lines XXX to XXX (revised text)".
We greatly apologize, we certainly did not wish to cause more work for the reviewer to find the necessary changes. We will list the line number and our changes in the following response.
The authors' response to my comments was confined to the minor points, with no attention to more important questions regarding speculations about mechanism which were (and still are) presented as factual conclusions. I do not consider the responses adequate.
We responded to each of your comments and where we disagree, we have explained in detail.
With respect to the meaning of "above" and "below" in the context of an intracellular organelle, I think that referring to up and down in a figure is fine, provided that the cytoplasmic and luminal sides are indicated in that figure. I think that labeling to that effect in each figure would be immensely helpful for the reader.
We agree with this point and have updated all the figures to include these labels.
The statement on lines 333-335 about non-competitive inhibition is a bit naïve. The only thing ruled out by this type of inhibition is that substrate and TBZ binding do not share the same binding process, in which case they would compete. It doesn't show that TBZ gets to its binding site from the lumen or from the bilayer, or by any other process that isn't shared with substrate. It also doesn't rule out kinetic effects, such as slow inhibitor dissociation, that result in non-competitive kinetics. Please rewrite this sentence to indicate that one explanation of the non-competitive nature of TBZ inhibition would be that TBZ diffuses into the vesicle and binds from the lumen. It's not the only explanation.
We have changed this sentence lines 334-336 to be more speculative and not include any statement about non-competitive inhibition. Please see, “Studies have proposed that TBZ first enters VMAT2 from the lumenal side, binding to a lumenal-open conformation.”
The revised version integrates the MD simulations into a plausible mechanism for luminal release of substrate. A key element in this mechanism is the protonation of D33, E312 and D399, which allows substrate to leave following water entry into the binding site. The acidic interior of synaptic vesicles should facilitate such protonation, but the fate of those protons needs to be considered. Are any of them predicted to dissociate prior to the return to a cytoplasm-facing conformation? If so, are all 3 released in that conformation? Postulating protonation events at one point in the reaction cycle requires some accounting for those protons - or at least recognition of the problem of reconciling their binding with the known stoichiometry of VMAT.
We completely agree with this point and while we cannot account for all protons with a single structure and simulation of neurotransmitter release, some discussion of the fate of the protons is warranted. We have included a highly speculative statement in the discussion on this point, see lines 462-465, “Given the known transport stoichiometry of two protons per neurotransmitter, we speculate that two protons may dissociate back into the lumen, perhaps driven by the formation of salt bridges between D33 and K138 or R189 and E312 for example in an cytosol-facing state.”
Reviewer #3 (Recommendations For The Authors):
On page 13, line 238, the statement "The protonation states of titratable residues D33, E312, D399, D426, K138 and R189, which are in close proximity to TBZ, also impact its binding stability (Table 4)" is misleading. Table 4 only shows that D426 is charged and what the pKa values are. This should be rephrased to separate out which residues are in close proximity from what is known about how their protonation states affect TBZ stability.
We agree with this statement and have rephrased this on line 290-294 on page 13 to read, “Several titratable residues, including D33, E312, D399, D426, K138, and R189, line the central cavity of VMAT2 and impact TBZ binding stability (Table 4). We found that maintaining an overall neutral charge within the TBZ binding pocket, as observed in system TBZ_1, most effectively preserves the TBZ-bound occluded state of VMAT2. Residues R189 and E312 in particular are within close proximity of TBZ and participate directly in binding.” We note that given the acidic pH of the vesicle lumen (5.5), it is likely all four residues may be protonated to a significant degree in this state.
Typos:
- luminal is another name for the drug generically known as phenobarbital, lumenal means in the lumen. (This typo seems to have crept into the published literature now too).
Thank you for pointing this out. Indeed, we had considered carefully whether to use ‘lumenal’ or ‘luminal’ in our revised text. In fact, both are used interchangeably throughout the scientific literature and luminal is the more commonly used term. Please also see: https://www.merriam-webster.com/medical/luminal we do agree that there may be confusion because ‘Luminal’ is a trademark of phenobarbital. Therefore, we have changed the text to read ‘lumenal’ throughout.
The following is the authors’ response to the original recommendations.
Reviewer #1 (Recommendations For The Authors):
I congratulate the authors on this study, which I enjoyed reading. Overall, the study reports a novel and exciting new structure for a member of the SLC18 family of vesicular monoamine transporters. Associated MD, binding and transport assays provide support for the hypothesis and firm up the modelled pose for the TBZ drug. The main strengths of the study largely sit with the structure, which, as the authors say, provides additional and essential insights above those available from AF2. The structures also reveal several potentially interesting observations concerning the mechanism of gating and proton-driven transport. The main weakness lies in the limited mutational data and studies into the role of pH in regulating ligand binding. As detailed below, my main comment would be to spend a little extra time expanding the mutational data (perhaps already done during the review?) to enable more evidence-based conclusions to be drawn.
We thank reviewer #1 for their helpful comments and suggestions. We agree that mutational analysis specifically of neurotransmitter transport would strengthen the mechanistic conclusions of the work. We also agree with reviewer #1 and #3 that the role of pH and the protonation state of charged residues was a weakness in the first version of the manuscript. Therefore, we have expanded our mutational and computational data as detailed below and we believe that this has further solidified our findings.
Specific comments & suggestions:
It is an interesting strategy to fuse the mVenus and anti-GFP nanobody to the N-/C-termini. The authors should also include in SI Fig. 1 a full model for the features observed in these maps and deposit this in the PDB.
Great point, we have made a main text panel describing the construct. Figure S1 includes a full description of the construct. The reviewer will note that the PDB entry contains the entire amino acid sequence of the construct and while the GFP and GFP-Nb cannot be well modeled into the density, we have included all of the relevant information for the reader.
Difficult to make out the ligand in Fig. 2b, I would suggest changing the color of the carbon atoms.
Fixed.
It is difficult to make out the side chains in ED Fig. 5d.
This is now its own supplemental figure and is presented larger.
ED Figures are called out of order in the manuscript. For example, in line 143 ED Fig.6 is called before ED Fig. 5d (line 152), and then ED 5d is called before ED 5a. This makes it rather confusing to follow the description, analysis, and data when reading the paper. Although there are other examples. I would suggest trying to order the figure callouts to flow with the narrative of the study.
Agreed. Fixed.
It wasn't clear to me what the result was produced by just imaging the ligand-free chimaera protein. It would be useful to say whether this resulted in low-resolution maps and whether the presence of the TBZ compound was essential for high-resolution structure determination.
The ligand is likely required for structure determination. We have not, however, made such a statement largely because we have yet to determine an apo reconstruction.
The role of E127 and W318 on EL1 in gating the luminal side of the transporter is very intriguing. As the authors suggest, this may represent an atypical gating mechanism for the MFS (line 182). I did wonder if the authors had considered providing more insight into this potentially novel mechanism. Additional experiments would be further mutations of W318 to F, Y, V, and I to see if they can identify a non-dead variant that could be analysed kinetically. They may have more luck with variants of E127, as they suggest this stabilises W318. If these side chains are important for gating and transport regulation, one might expect to see interesting effects on the transport kinetics.
This is a fantastic suggestion. We have done this, and we think that the reviewer will find the results to be quite interesting. Some VMAT2 sequences have an R or an H at position 318 while VPAT has an F at the equivalent position. We have made these mutants including the E127A mutant and analyzed them using TBZ binding and transport experiments. Interestingly the W318R, H, and F mutants preserve activity in varying degrees with the R mutant closely resembling wild type. W318A has no transport activity. Only the W318F mutant retains some TBZ binding. The E127A mutant also has little transport activity but nearly wild type like TBZ binding which we believe suggests a role for this residue also in stabilizing W318.
The authors identify an interesting polar network, which is described in detail and shown in Fig. 2d. However, the authors present no experimental data to shed further mechanistic insight into how these side chains contribute to monoamine transport or ligand binding. Additional experiments that would be helpful here might include repeating the binding and competition assays shown in Fig. 1c under different pH conditions for the WT and different mutations of this polar network. At present, this section of the manuscript is very descriptive without providing much novel insight into the mechanism of VMAT transport. I did wonder whether a similar analysis of pH effects on DTBZ binding might also provide insight into the role of E312 and the role of protons in the mechanism.
Thank you, we have addressed this point in several different ways. The first is that many of these residues have already been characterized in several earlier studies, see refs 31, 32, and 42 and we have incorporated this into our discussion where appropriate. With respect to E312, the reviewers’ comments are again very appropriate. We have addressed this using computational experiments exploring the protonation status of E312 and other residues as well as TBZ. Our simulations and Propka calculations clearly show that E312 must be protonated and TBZ must be deprotonated to maintain TBZ binding. We have also extended these computational studies toward understanding the protonation status of residues which orchestrate dopamine binding and release.
The authors then describe the binding pose for TBZ. This section also provides some biochemical characterisation of the binding site, in the form of the binding assay introduced in Fig. 1. However, the insights are again somewhat reduced as the mutants were chosen to show reduced binding. Could the authors return to this assay and try more conservative mutations of the key side chains to illuminate more detail? For example, does an R189K mutant still show binding but not transport? Similarly, what properties does an E312D have? The authors speculate that K138 might play a role in coupling ligand binding/transport to the protonation, possibly through an interaction with D426 and D33 (line 236). Given the presence of D33 in the polar network described previously, I was left wondering how this might occur. I feel that some of the experiments with pH and conservative mutants might shed some light on this important aspect. Please label the data points in Fig. 3d.
Indeed, alanine mutants at these positions while valuable do not provide the level of detailed insight into mechanism that we also would have liked to obtain. Thus, we have made more conservative and targeted mutants like the R189K mutant and various mutants at N34 for example and tested them in both transport and binding assays. We have also made a mutant at K138 and found that it is not transport competent or able to bind TBZ to a significant degree. With respect to labels and color codes, we have made the color codes consistent between the bar graphs and the curves. We have also labeled the data points in the figure legends.
The manuscript currently doesn't present a hypothesis for how TBZ induces the 'dead-end' complex compared to physiological ligands. Does the MD shed any light on this aspect of the study? If the authors place the physiological ligand in the same location as the TBZ and run the simulation for 500ns, what do they observe? 100ns is also a very short time window. I appreciate the comment about N34 in line 303, but is this really the answer? It would be very interesting to provide more evidence on this important aspect of VMAT pharmacology.
MD with a natural ligand (dopamine) provides substantial insight into why TBZ is a dead-end complex. Since water cannot penetrate into the binding site in the TBZ bound complex, this does not allow for substantial luminal release. In contrast, simulations conducted in the presence of DA bound to the occluded VMAT2 show the propensity of that structure to accommodate an influx of water molecules that promote the release of DA to the lumen. The new results are illustrated in Figure 5 (main text) as well as supplemental figure 8 panels d-h. The new simulations further emphasized the importance of the protonation state of acidic residues near the substrate-binding pocket.
Reviewer #2 (Recommendations For The Authors):
Line 68, "both sides of the membrane" -> "alternately to either side of the membrane".
Fixed. Thanks.
Transmembrane proteins in intracellular organelles present unique issues of nomenclature. I suggest the authors refer to cytoplasmic and luminal faces of the protein (not intracellular or extracellular (line 124)) and adhere to these names to avoid confusion. This creates problems for loops called IL and EL, but they could be defined on first use.
We agree with this point and had initially gone with the conventional definitions used in the literature. We have now changed this throughout the text to be luminal and cytosolic.
Lines 135-6, are these residue numbers correct? The pdb file lists 126 as Asp and 333 as Ala.
Thank you. This is fixed.
ED Fig. 6 is not clear. A higher-resolution figure is needed.
We have updated this figure and hope that the reviewer will find it to be much clearer.
Lines 158-9, Is there any data to support effects on dynamics or folding? If not, please indicate that this is speculation.
Fixed.
Line 174, Should "I315" be "L315"?
Fixed.
Line 179, Please indicate what is meant by "inner" and "below" (also lines 183 and 258).
We have added Figure calls here where needed.
Line 192, S197 is listed as part of polar network 1, but not discussed further. Is it actually involved, or just in the neighborhood?
It is part of the network, but we did not discuss in further detail because we do not have data indicating its precise function and thus have left this as a description.
Line 199, E312, and N388 are fairly distant from each other. Do you want to clarify why they represent a network?
While they are not within hydrogen bonding distance, we nevertheless include them as part of the same network because they may come into closer proximity in a different conformational state.
Line 206, Protonation of all 3? VMAT2 doesn't transport 3 protons per cycle. Please clarify.
We believe that these residues may be protonated, but they may not necessarily all be involved in proton transport.
Line 219, Do you mean the aspartate unique to DAT, NET, and SERT? This is Gly in all the amino acid transporters in the NSS family. Please be specific.
Fixed. Thank you.
Line 224, "mutation of E312 to Q" or "mutation of Glu312 to Gln".
Fixed. Thank you.
Fig. 3d, Normally, one would expect full saturation curves for each mutant. How can a reader distinguish between low affinity or a decrease in the number of binding sites? Would full binding curves be prohibitive for the mutants because of the cost or availability of the ligand? These points should be addressed. A couple of the curves are not visible. Would an expanded scale inset show them more clearly? Also, would it be possible to include chemical structures for all ligands discussed?
Many if not most of these mutants bind TBZ with such low affinity that it is not possible to measure a full saturation curve either because of ligand availability (radioactive ligand concentration is only in µM) or due to technical issues with being able to measure such low affinity binding. We have changed the presentation of the curves and have split the gating and binding site mutants into their own figures. We feel this improves the readability of these curves. We have also included a table with the respective Kd values determined for each of the mutants where possible.
Line 235, The distances are long for a direct interaction between K138 and the TBZ methoxy groups. The unusual distances should be mentioned if an interaction is being proposed.
We do not think that K138 is directly involved in TBZ binding, however this was written in a confusing way and has been now changed.
Line 243, Please give a quantitative estimate of the affinity difference. "modestly" is vague.
It is an approximately 2-fold difference. Fixed in the text.
Line 248, 150 nM is, at best, a Kd, not an affinity.
Agreed, this is changed.
Reviewer #3 (Recommendations For The Authors):
The (3 x ~100ns-long) molecular dynamics simulations provided suggest some instability of the pose identified by cryo-EM. While it is not unreasonable that ligands shift around and adopt multiple conformations within a single binding site (in a reversible manner), the present results do raise questions about the assumptions made when starting the simulations, in particular (1) the protonation states of charged residues in the TBZ binding sites; (2) the parameters used for tetrabenazine; (3) the conformations of acidic side chains that are notoriously difficult to resolve in cryoEM maps; and (4) any contributions of the truncated regions truncated in the simulated structure, namely the cysteine cross-linked loop and the terminal domains. The authors should examine and/or discuss these contributions before attributing mechanistic insights into the newly observed binding orientation.
In order to estimate the effects of protonation states on TBZ binding, we now added three new systems with altered protonation on TBZ and binding pocket lining residues (see Table 3 in the revised vision); and for each system, we performed multiple MD runs to address the question and concerns raised by reviewer.
Regarding the protonation states: Propka3.0 was used to determine the protonation states, finding that E312 and D399 should be protonated. If I am not mistaken, this version of ProPka cannot account for non-protein ligands (https://github.com/jensengroup/propka). Given their proximity to the binding site, these protonation states will be critical factors for the stability of the simulations. The authors could test their assumption by repeating the calculations with Propka 3.1 or higher, to establish sensitivity to the ligand. Beyond this, showing the resultant hydrogen bond networks will help to reassure the reader that the dynamics in the lumenal gates do not arise from an artifact.
We thank the reviewer for suggestion of using higher version of Propka. We used the most recent Propka3.5 and carried out protonation calculations in the presence and absence of TBZ. The new calculations are presented in Table 4 and SI Figure 8c of the revised version.
It should be possible to assess whether waters penetrate the ligand binding site during the simulations if that is of concern.
We now added the number of waters within the ligand binding pockets for all MD simulations we performed, which are presented in Table 3 and Table 5 of the revised version.
Finally, I didn't fully understand the conclusion based on the simulations and the "binding affinity" calculations: do they imply that the pose identified in the EM map is not stable? What is the value of the binding affinity histogram?
We apologize for this confusion. For each MD snapshot, we calculated TBZ binding affinity using PRODIGY-LIG (Vangone et al., Bioinformatics 2019), which is a contact-based tool for computing ligand binding affinity. The binding affinity histogram shown in the original submission was the histogram of those binding affinities calculated for MD snapshots. In the revision, we replaced binding affinity histogram by time evolution of binding affinity changes (SI Fig 6c in the revision). The simulations confirmed that the pose identified in the EM map is stable, with a flattened binding affinity of -9.4 ± 0.3 kcal/mol in all three runs.
Recommendations regarding writing/presentation:
The authors use active tense terminology in attributing forces to elements of structure (cinching, packing tightly, locking). While appealing and commonplace in structural biology, this style frequently overstates the understanding obtained from a static structure and can give a rather misleading picture, so I encourage rephrasing.
We appreciate this point; the use of these words is not meant to overstate or provide a misleading picture but rather to aid the reader in mechanistic understanding of the proposed processes.
I would also recommend replacing the terms "above" and "below" for identifying aspects of the structure; the protein's location in the vesicular membrane makes these terms particularly difficult to follow.
These terms refer specifically to the Figures themselves which we have always oriented with the luminal side at the top of the page and the cytosolic on the bottom. We have indicated in Figure 1 the orientation of VMAT2. The Figures are the point of reference which we refer to, and the ‘above’ and ‘below’ terms have been used to assist the reader to make the manuscript easier for a more casual or non-expert reader to follow.
Minor corrections:
- the legend in Figure 2 lacks details, e.g. how many simulation frames are shown, how were the electrostatic maps calculated?
We revised Figure 2 and moved simulation frames to SI figure 6e. A total of 503 simulation frames are shown.
- how were the TBZ RMSDs calculated? using all atoms or just the non-hydrogen atoms?
For TBZ RMSDs, we used non-hydrogen atoms. This information is presented in the Methods section.
MD simulation snapshots and input files can be provided via zenodo or another website.
We will upload snapshots and input files to Zenodo upon acceptance of the manuscript.
Reviewing editor specific points:
Specific points
L.97: Remove "readily available"
Fixed.
L.99: The authors are not measuring competition binding. It is well known that reserpine and substrates inhibit TBZ binding only at concentrations 100 times higher than their respective KD and KM values. It is, therefore, surprising that the authors use this isotherm and refrain from commenting on the significance of the finding. Moreover, the presentation of results as "Normalized Counts" does not provide any information about the fraction of VMAT molecules binding the ligand. At least, the authors should provide the specific activity of the ligand, and the number of moles bound per mole of protein should be calculated.
The point was not to infer any details about the conformations that TBZ and reserpine bind but merely to point out that both constructs have a similar behavior with respect to their Ki for reserpine. We have added a sentence to say that reserpine binding stabilizes cytoplasmic-open so the reader is aware of the significance of this competition experiment.
L.102: The characterization of serotonin transport activity needs to be more satisfactory. The Km in rVMAT2 is 100-200 nM, so why are the experiments done at 1 and 10 micromolar? Is the Km of this construct very different? The results provided (counts per minute at the steady state) need to give more information.
The Km of human VMAT2 varies somewhat according to the source but has generally been reported to be between 0.6 to 1.4 µM for serotonin according to these references.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3019297/
https://www.cell.com/cell/pdf/0092-8674(92)90425-C.pdf
https://www.pnas.org/doi/abs/10.1073/pnas.93.10.5166
Fig 1B could be more informative. I suggest adding a cartoon model with TMs labeled, similar to ED Fig6a.
This panel is to aid the reader in accessing the overall map quality and thus we do not wish to add additional labels/fits which would distract from that point. Instead, we have added overall views of the model in Figs 2,3.
L.179: The authors claim that the inner gate is located "below" (whatever this could mean) the TBZ ligand. In L.214, they claim that TBZ adopts a pose.....just "below" the location of the luminal gating residues. Please clarify and use appropriate terminology.
This refers to the position of these residues in the Figures themselves. We have added figure calls where appropriate here.
Fig. 4: The cartoon could be more informative.
We have added more information to the mechanism cartoon which is now Figure 6. This incorporates some of our new data and we believe it will be more informative.
L. 213: The paragraph describes residues involved in TBZ binding. Mutagenesis is used to validate the structural information. However, the results (ED fig. 5B) must be corrected for protein expression levels. In the Methods section, the authors state (L.444), "Mutants were evaluated similarly from cell lysates of transfected cells." Without normalization of protein expression levels, the results are meaningless even if they agree with predictions.
In fact, we have normalized the concentrations of protein in our binding experiments. This was noted in the methods section. And to account for these differences, experiments were conducted using 2.5 nM of VMAT2 protein as assessed by FSEC.
L.220: The referral to ED Fig.7 is not appropriate here. The figure shows docking-predicted poses of dopamine and serotonin.
Figure call has been changed.
L.226: The referral to Fig. 3b needs to be corrected. The figure shows TBZ and not the neurotransmitter.
This has been corrected.
L. 337: "The neurotransmitter substrate is bound at the central site." What do the authors mean in this cartoon? Do they have evidence for this? Tetrabenazine is not a substrate.
This cartoon drawing is meant to illustrate the elements of structure. Similar drawings are presented throughout the literature such as here: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5940252/ Figure 3 and here: https://pubs.acs.org/doi/10.1021/acs.chemrev.0c00983 Figure 2.
The same compound is mentioned with different names: 3H-dihydrotetrabenazine and 3H-labeled DTBZ.
Fixed.
ED fig 1d is illegible.
The high-resolution figure is completely legible. We will provide this to the journal upon publication.
Figure 2d: A side view would be more visual.
We have updated this figure and believe that it is much easier to understand now.
L. 179: The inner gate is located 'below' the TBZ ligand
Please see above response, this refers to the figures themselves. The figures are our point of reference.
L. 213-215: Tetrabenazine binding site just 'below' the location of the luminal gating residues.
See above.
Throughout the paper, results are given as cpm or counts. The reader can only estimate the magnitude of the binding/transport by knowing the specific activity of the radiolabel. I recommend switching to nano/picomoles or supplying enough information to understand what the given cpm values could mean.
Binding experiments were done using scintillation proximity assays and therefore converting the CPMs to values in pmol of bound ligand is simply not possible. For the transport experiments (now Fig 1d) the point was to show that the wild type was similar in activity to the chimera. In our new transport experiments we have presented for the mutants, many experiments were combined together and therefore, we have normalized the counts to the relative activity of wild type VMAT2.
