Reduced efficacy of a Src kinase inhibitor in crowded protein solution

The inside of a cell is highly crowded with proteins and other biomolecules. How proteins express their specific functions together with many off-target proteins in crowded cellular environments is largely unknown. Here, we investigate an inhibitor binding with c-Src kinase using atomistic molecular dynamics (MD) simulations in dilute as well as crowded protein solution. The populations of the inhibitor, 4-amino-5-(4-methylphenyl)−7-(t-butyl)pyrazolo[3,4-d]pyrimidine (PP1), in bulk solution and on the surface of c-Src kinase are reduced as the concentration of crowder bovine serum albumins (BSAs) increases. This observation is consistent with the reduced PP1 inhibitor efficacy in experimental c-Src kinase assays in addition with BSAs. The crowded environment changes the major binding pathway of PP1 toward c-Src kinase compared to that in dilute solution. This change is explained based on the population shift mechanism of local conformations near the inhibitor binding site in c-Src kinase.

The manuscript is an important contribution and an exciting piece of science. Since the pioneering computational efforts carried out by the authors teams (M. Feig's and Y. Sugita's), this new work is a beautiful example showing that the modelling of the macromolecular crowding is an essential aspect to consider in order to understand in vivo biological processes. The manuscript is clear, well written and presented. The technical aspects are neat. The in-silico results, obtained with great statistics considering the systems, are complemented by ad hoc experiments that try to reinforce the main conclusion of the investigation. I consider the manuscript publishable in Nature Communication but several aspects need to be addressed prior to pubblication. The points are listed below.
My first aspect concerns the "sequestration" thermodynamics of the inhibitor that is at the origin of the trapping mechanism of the molecule on the surface of the crowders. This interaction limits the independent diffusion of the molecule in the solvent 'phase', producing a strong slowdown that in turn quenches the access to the target protein. It will be important to report/show/stress that within the used force field the solvation free energy of the ligand and the interaction free energy with representative surface patches of the crowder BSAs are not unbalanced artificially. The authors can easily produce these data by independent calculations.
Strictly related to the point mentioned above, the authors should better discuss and account for the discrepancy between the simulated and the experimental conditions. In the experiments, they only add 5 g/L of BSA (which corresponds to a very small volume fraction of ~0.004, right?). So, to observe any strong difference in the bulk concentration of the inhibitor, which they suggest as a main rationale of the decreased inhibitor efficiency, the inhibitor should bind very strongly to BSA. However, in the simulations, they only observe a drop in the bulk concentration of the inhibitor from 80 % in dilute conditions to 30 % at a volume fraction of 0.11. This says that for the much smaller experimental volume fraction, the bulk concentration should be very similar to the dilute case. Therefore, the sequestration would not be sufficient for explaining the experimentally detected orderof-magnitude shift in inhibitor concentration.
On the basis of the sequestration on crowders surfaces and the associated slowdown dynamics of the inhibitor, the authors open a discussion about drug design strategy in order to account for the realistic environmental conditions of the cell interior. Here I have a question. Looking at the experiments reported in Fig. 2 bottom panel it seems that in order to recover the inhibition efficiency observed in dilute condition it suffices to increase the inhibitor concentration (a factor of 10). In drug design and subsequent essays, the concentration is always a free parameter that needs to be tuned to grant efficiency and avoid toxicity. If the author enters in this arena, they should for instance discuss what concentration limits should be avoided. Is there an example that can be reported? Given the great experience cumulated with past simulations (see the authors past works), is there a rule of thumbs that relates slowdown motion & local sequestration and crowder concentration so to estimate the possible shift in drugs concentration needed to recover dilute solution efficiency? Finally, a personal view. Considering that great effort in drug design is spent to refine the chemistry of the active molecule to grant high affinity with target binding sites, and that basically the slowdown/sequestration caused by the environment seems not highly specific, I think very difficult to incorporate these two requirements (highly affinity for a target, poor interactions with general protein surfaces) at once. Have the authors some ideas? A finest analysis of the inhibitor with the crowders, that is not present in the manuscript and should, could provide some suggestions. I think also the authors should add in the introduction a note on the effect of diffusion in molecular recognition under crowding. The competitive effect of local high concentration and slowdown motion reported in the "Discussion" section, can be already anticipated in the introduction.
One extra remark concerns the discussion of the conformational changes induced by the BSAs crowders and the altered distribution of the Y82-G86 distance, a proxy for the conformational state of Tyr82. To me, the text seems light on this. The authors observe a change but I do not find a connection with the presence of the crowders. I think this must be better described/analyzed and a structural effect indicated, or at least quested.
One delicate aspect the manuscript does not address is the homogenous vs heterogenous crowding effect. Since we know that big vs small crowders can have different effects on local packing, stability, and conformational fluctuations, and since the change in probability of Tyr82 is visible even in presence of two BSAs, the authors could, with a small effort, produce a simulation with smaller crowders, or replacing 1 BSA with a different crowder. This would reinforce the manuscript and the results.
Final curiosity. It is intriguing that binding events are visible only in the simulations of the dilute system and at the highest crowding concentration, and not at intermediate values of crowding. See Table 1 in SI. Why? This must be discussed! The paragraph reporting on binding pathway is present twice in the manuscript (pg 9 lines 176-190, and pg 10 191-205). A cut and paste error, I guess. Fig. 2. The volume fraction associated to the experimental condition, BSA 5 mg/ml, should be indicated so to have a link between the top and the bottom panels. Fig. 3 can be improved by labeling explicitly panels c and d, with the "dilute" and "Src8BSA" labels so to ease a reader to spot the differences immediately.
Reviewer #2 (Remarks to the Author): The authors investigated the diffusion of the low molecular weight inhibitor PP1 and its binding to Src kinase in the absence and presence of crowding. Crowding is mimicked by the use of different concentrations of the crowder bovine serum albumin. The authors suggest, based on experiments, that PP1 is less efficient as an inhibitor in the presence of crowders. They also observe that PP1 is less available in solution with increasing crowder concentration, and that crowders may change the pathway for ligand binding due to conformational changes in Tyr82 of Src kinase.
The manuscript is generally clearly presented and mostly well written. The manuscript addresses a significant problem for which molecular simulations can give important insights. The results are original and interesting. However, I have some major concerns as detailed below, some details are missing in the methods, and the conclusions are not fully supported by the results.
Major concerns 1-Low sampling of binding pathways Page 14, line 269: 'In Src8BSA simulations, the PP1 binding pathway is entirely different from that in dilute simulations.'. The authors obtained very few binding events in the simulations, especially in crowded conditions (4 and 2 events in the absence and presence of crowding, respectively). Therefore, they would need to sample more binding events to be able to claim that binding pathways are indeed different in the absence and presence of crowding.
The plot in Fig 2a shows computed values that do not directly relate to the main heading of the figure as a function of crowder protein volume fraction. The corresponding experimental value for the protein volume fraction would be at less than 1% and at a different ratio of kinase to crowder. Thus, this plot has little relation to the experimental data in Fig 2b and I think it would be better to put these two plots into different figures.
Page 8, lines 151-153: 'The 50% inhibition concentration (IC50) of the PP1 considerably increased from 33.80 nM to 290.4 nM in the presence of BSA ( Fig. 2(b)). This validates our hypothesis of a reduced inhibitor efficacy in crowded protein solutions.'. Here, the authors need to point out that the enzyme activity is higher in the presence of BSA and give what the reference was for defining 100% enzyme activity. If the reduced inhibitor efficacy is due to differences in inhibitor-protein binding, one wonders why the same effect is not observed for substrate -indeed, the data suggest the opposite. However, at no point is it stated what the Src substrate measured actually is (l309, p15). This information needs to be added, along with data on the dependence of Src substrate binding on BSA crowding, e.g. from analogous simulations to those for the inhibitor.

3-Incomplete description of computational methods
The setup and parameters for simulations are well described, but there is no information (in the manuscript or in the supporting information) about how the analysis of these simulations was performed (calculation of spatial distribution functions, free energy landscapes, and mean square displacements; definition of the bound state). This should be added.
4-Page 2, lines 44-45: 'Protein functions in a living cell could be examined using atomistic MD simulations with realistic cellular environments.'. This is an overstatement, compared to what is presented in the paper and must be removed or edited. The interaction with one artificial crowder, BSA, is studied.  9-Page 5, line 105: 'The Cα root mean square deviations (RMSDs) from the X-ray structure are observed around 2-3 Å in all the simulations, indicating that the kinase remains in the active conformation'. A small RMSD does not indicate whether the kinase is in the active conformation or not. The authors should examine the residues in the DFG motif to make this statement. are slightly smaller than that in the absence of BSA (~2.6 Å) (Supplementary Fig. 4(a)). e in the inhibitor-unbound state are also not affected by the presence of the BSAs (Supplementary Fig. 4(b))'. Something is missing in this sentence. Figure S8 The authors could compare the binding sites of PP1 on BSA with the binding sites of other drugs (see, for instance, PDB 4JK4).
13-Consider swapping Figure 2A with Figure S10, which makes clear that, in the presence of many proteins, the probability of PP1 being bound to a protein is proportional to the SASA of the protein.
14- Table S6 Explain the meaning of the terms in the equation. 15-Page 9: 'This binding pathway is different from that observed in Src8BSA: PP1 first reaches the hinge region of the kinase and then intrudes into the canonical binding site.' This is true for 1/2 cases in figure S13. Please rewrite it.
19-Page 11-12, section 'Conformatioal shifts of a Tyr sidechain upon crowding'. The conclusions about Tyr82 are interesting, but I would expect that both paths (hinge region and Gloop) would be observed in crowded conditions. This is probably a sampling problem, as only 2 binding events were observed in crowded conditions. The authors should comment about it in the discussion. 20-What is the conformation of Tyr82 in the apo form of kinase? It seems it is Tyr-out in figure S18 (in which the residues should be labelled). Does the Tyr-in conformation appear in any crystal structure? Could the Tyr-in conformation be a force field artifact? 21-Page 3 of supporting information, section 'Definition of reaction coordinates for free energy landscapes'. The definition of the five anchor atoms (L0, L1, P1, P2, and P3) is very confusing. Does it change according to trajectory or frame? It would be easier to give the identity of each anchor atom with the definition.
Reviewer #3 (Remarks to the Author): In their manuscript (NCOMMS-20-35226) Sugita, Feig, and co-workers use extensive modeling and analyses, and experimental measurements to demonstrate that crowding reduces efficacy of a kinase inhibitor. The study is based on extensive and well-designed set of all-atom simulations of systems containing c-Src kinase, small inhibitor (PP1), and bovine serum albumin (BSA), which is present at varying concentrations and serves as a crowder. The study is detailed and findings are well supported.
The all-atom molecular dynamics trajectories captured spontaneous PP1 binding into the active site of c-Src kinase and the authors were able to show that crowding changes the binding pathways as in the presence of BSA PP1 concentration is reduced near the kinase. It is insightful that the authors were able to provide mechanistic explanation of the importance of excluding volume and of weak nonspecific PP1-protein interactions. Importantly, the computational findings are supported by experimental binding data.
These findings have implications for future studies of drug interactions in native-like environments and will be of great interest to the readers of Nature Communications. The manuscript should be published mostly as is; I have found only a few places in the text which require edits related only to formatting: Line 106 "e in the inhibitor-unbound" has missing text. Line 175: The paragraph is repeated starting from the Line 190.

Reviewer #1 (Remarks to the Author):
The manuscript is an important contribution and an exciting piece of science. Since the pioneering computational efforts carried out by the manuscript is an important contribution and an exciting piece of science. Since the pioneering computational efforts carried out by the authors teams (M. Molecular mechanics-generalized Born surface area (MM-GBSA) analysis on the binding affinity of PP1 to either c-Src kinase or BSA was performed to address the concern. We confirmed the interactions of PP1 with c-Src kinase and with BSA are overall "well balanced", i.e. the calculated binding affinities to different surface regions are of similar magnitude between the two proteins and between different surface patches. The canonical binding site for PP1 in c-Src kinase is found at the lowest binding free energy. There are several similarly or even slightly more favorable binding sites on BSA as well, but it is in fact the point of this study that off-target ligand binding may significantly distract from on-target binding events under crowded conditions. The binding free energies at different sites are shown in Supplementary Fig. 9. In addition, following sentences were added at the end of the page 7 in the revised manuscript:  We additionally carried out a new experiment increasing BSA concentration up to 100 mg/ml (Fig.   3). This concentration corresponds to the concentration in the Src2BSA simulation. With the high BSA concentration, the inhibition of PP1 becomes much milder. On page 9 in the revised manuscript, we added the sentences: "We also performed the same inhibition assays at the higher concentration of BSA (100 mg/ml).
In the experiments, the IC50 of PP1 further increased to 12.2 µM, indicating even weaker inhibition by PP1 in the presence of higher concentrations of BSA." The experimental conditions of the old and new experiments are the same except for the concentration of DMSO in the buffer solutions (0.2% and 1% in the old and new experiments).
We explained these differences between the first and the second experiments on page 17: "The second in vitro inhibition assay was performed in the same conditions except for the concentrations of BSA (100 mg/ml) and DMSO (1.0%)." We believe that our work further supports the qualitative assertion that a consequence of cellular crowding is an effectively lower inhibitor concentration due to reduced activity and diffusion, but it may require further efforts to arrive at the more quantitative 'rule of thumb' that the reviewer is asking for. One issue is the large time-scale gap between simulations and experiments that limits exact quantitative predictions. To address this challenge, we have to accelerate biological events in simulations at higher crowder concentrations in order to be able to study ligand binding events.
We also generally consider higher macromolecule concentrations in the simulations than experiments to reduce computational costs, which again is limiting the ability to make exact quantitative predictions until now. First, we added sentences about the crowding effect of diffusion in Introduction, on page 3 of revised manuscript, according to the suggestion of the reviewer: "Diffusion of a ligand toward a target protein is generally reduced, while the transition-state stabilization and/or encounter-complex formation are facilitated, as the protein concentration increases.
These two effects make it difficult to understand the overall effect of crowding on enzymes 4 .
As the reviewer pointed out, it is not straightforward to satisfy the two requirements (high affinity for a target and poor interactions with general protein surfaces) in the current drug discovery strategy. More quantitative ligand-binding calculations, for instance, free-energy calculations, may help to look for such a drug candidate. Since we have already mentioned such a future work on page 15 in the original manuscript, we did not change the revised manuscript for this comment.
"In future work, binding free energy calculations under crowded conditions could quantitatively address the relation between the effective inhibitor concentration and its efficacy." The reviewer is also asking for a more extensive analysis of the inhibitor with crowders. This is an excellent point but in a qualitative sense we expect to find well-known patterns of interaction (shape complementarity, electrostatic hot spots) which, we believe may not add much here. We did however, also in response to the second reviewer, compare binding of PP1 to BSA with reported binding poses of other compounds. The finding is that they overlap to some degree.
It would be of greater value to carry out a thorough analysis of how exactly inhibitor binding to the crowders differs from binding to the kinase and binding to BSA. We think that this is a good question for future work and in some sense separate from the crowding simulations presented here because it can be addressed more efficiently by comparing inhibitor binding to a single molecule of Src-kinase vs. a single molecule of BSA. Such a study should also consider a wider range of other 'crowder' proteins in order to be able to come to more general conclusions. We believe that such work is clearly out of the scope of the present study.
(5) One extra remark concerns the discussion of the conformational changes induced by the BSAs crowders and the altered distribution of the Y82-G86 distance, a proxy for the conformational state of Tyr82. To me, the text seems light on this. The authors observe a change but I do not find a connection with the presence of the crowders. I think this must be better described/analyzed and a structural effect indicated, or at least quested.
We have tried to figure out the connection between the Tyr82 conformations and the crowding conditions. The Ca atom contacts between c-Src kinase and BSA in each simulation were examined for TYR-in and TYR-out conformations, individually (Supplementary Figure 24).
However, no significant differences appeared between the two conformations. This suggests the effect of crowders on the local protein conformation could be more complicated, and we have to leave this question for future work. To mention this point, we added the following sentence to the Discussion on Page 15.
"To examine the difference of c-Src kinase-BSA interactions between TYR-in and TYR-out conformations, we calculated the average number of the Ca atom contacts between BSA and c-Src kinase for the two conformations individually (Supplementary Figure 24). However, no significant differences between the two conformations is observed, suggesting that the crowder effects on the local protein conformation could be more complicated."  (2011)). However, to simulate different crowding systems is in fact not a small effort for us even if just one BSA in the present systems is replaced by a different crowder. We note that the simulations presented here are based on rather extensive computational resources including simulations on the special-purpose Anton2 supercomputer where our access to resources is very limited. While we would like to address this question, it will need to remain a topic of future work.
In the revised manuscript on page 16, we just added the following sentences added in Discussion.
"The effects of different crowder types, for instance, the size of crowders 6 on protein structures and stability were investigated experimentally. The intracellular environments indeed consist of heterogeneous proteins and other biomolecules, as we simulated in the previous simulation study 15 . A further investigation using heterogenous crowders along the present study would lead to a deeper understanding of how the local conformational changes in more realistic cellular environments alter the binding mechanism." This is an interesting suggestion about further discussion in the paper. As shown in Fig. 5, the dilute system emphasizes the Tyr-in conformation, while the highest crowding concentration increases the population of Tyr-out. One hypothesis is that these two extreme cases increase the ligand binding possibilities through different pathways. If we can perform our MD simulations much longer, more binding events through either one of the pathways would be observed in Src2BSA and Src4BSA systems, since the conformational changes seem to happen due to the population shifts. However, it is currently difficult using our available computer resources. In this study, we already carried out extensive conformational samplings using very significant computer time on supercomputers as well as the MD-specific supercomputer, Anton2. We added the discussion pointed out by the reviewer on page 15.
"In the study, we observed binding events only in the dilute and Src8BSA systems, which are considered as two extreme conditions. These two conditions may increase the possibility of binding processes on different pathways. Since the conformational changes between Tyr-in and Tyr-out happen as population shifts (Fig. 5), longer MD simulations in Src2BSA and Src4BSA could increase the possibilities of ligand-binding processes through one of the two pathways."

(8)
The paragraph reporting on binding pathway is present twice in the manuscript (pg 9 lines 176-190, and pg 10 191-205). A cut and paste error, I guess.
We are sorry for our mistake. We removed the paragraph in the revised manuscript. (10) Fig. 3 can be improved by labeling explicitly panels c and d, with the "dilute" and "Src8BSA" labels so to ease a reader to spot the differences immediately.
The figure 4 in the revised manuscript was modified accordingly.

Reviewer #2 (Remarks to the Author):
The authors investigated the diffusion of the low molecular weight inhibitor PP1 and its binding  Figure 17). Thus, we can claim that the crowder proteins alter the binding pathways. Following sentences were added in the subsection "Ligand-binding pathways in crowded environments" (on Page 11).
"To examine if the difference happened to be observed, we performed additional 30 simulations (each for 20 ns) in Src8BSA starting from E and Er states. While the simulations from Er state lead to three binding events, no binding was observed in the simulation from E state. Similar to the binding trajectories shown in Fig. 4(d) and Supplementary Fig. 16, the binding trajectories from Er state fit to 'restraint' FEL ( Supplementary Fig. 17). Thus, it can be safely stated that the binding pathways are altered in the presence of crowders." (2)

Experimental results
The plot in Fig 2a shows  Based on the reviewers' suggestion, we split the original Fig. 2 into the new Fig. 2 (for simulation) and Fig. 3 (for experimental data) in the revised manuscript. Also, as responded to the reviewer 1, we carried out new experiments to collect data at increased BSA concentrations (Fig. 3).
Page 8, lines 151-153: 'The 50% inhibition concentration (IC50) of the PP1 considerably increased from 33.80 nM to 290.4 nM in the presence of BSA (Fig. 2(b)). This validates our hypothesis of a reduced inhibitor efficacy in crowded protein solutions.'.

Here, the authors need to point out that the enzyme activity is higher in the presence of BSA and
give what the reference was for defining 100% enzyme activity. If the reduced inhibitor efficacy is due to differences in inhibitor-protein binding, one wonders why the same effect is not observed for substrate; indeed, the data suggest the opposite. However, at no point is it stated what the Src substrate measured actually is (l309, p15). This information needs to be added, along with data on the dependence of Src substrate binding on BSA crowding, e.g. from analogous simulations to those for the inhibitor.
In the current simulation study, we did not include Src substrate, while, in the experiments shown in Fig. 3, we examined the enzymatic activity of c-Src kinase with ATP and Src substrate in the presence or absence of BSA as crowders. Although the dependence of Src substrate binding on BSA crowding is interesting theoretically and experimentally, analogous simulations to those for the inhibitors are greater effort for us than the current ones. We thank the reviewer to suggest this interesting future research and would like to try it in our future works.

(3) Incomplete description of computational methods
The setup and parameters for simulations are well described, but there is no information (in the manuscript or in the supporting information) about how the analysis of these simulations was performed (calculation of spatial distribution functions, free energy landscapes, and mean square displacements; definition of the bound state). This should be added.
We added the details of computational methods in the supplementary information as "Analysis Details" (Page 3 and 4). This is an overstatement, compared to what is presented in the paper and must be removed or edited. The interaction with one artificial crowder, BSA, is studied.
We follow the reviewers' opinion. The scope of our paper is indeed limited by what we can achieve with current computer methodology. In the revised manuscript, we removed the sentence.  We added the information on BSA concentrations in g/L for each system in Supplementary   Fig. 4(a)).
e in the inhibitor-unbound state are also not affected by the presence of the BSAs ( Supplementary   Fig. 4(b))'.
Something is missing in this sentence.
The sentence was modified as follows (Page 5, line 106).
"The Ca root mean square fluctuations (RMSFs) of the kinase in the inhibitor-unbound state are also not affected by the presence of the BSAs … (abridged) …" (11) Figure S8 The authors could compare the binding sites of PP1 on BSA with the binding sites of other drugs (see, for instance, PDB 4JK4).
We add the figure of the crystal structure 4JK4 and the following sentence was added (Page 7, line 143).
"The binding sites of PP1 on BSA resemble those of other compounds found in the crystal structure (PDB ID: 4JK4), reflecting the general feature of BSA as a vehicle of small molecules in the circulatory system (Supplementary Fig. 7(b)) 56 " (12) Page 8, line 132: 'indepdent assays'; independent.
The mistake was corrected.
(13) Consider swapping Figure 2A with Figure S10, which makes clear that, in the presence of many proteins, the probability of PP1 being bound to a protein is proportional to the SASA of the protein.
The figures were swapped.
(14) Table S6 Explain the meaning of the terms in the equation.
The explanation was added in "Analysis Details" in supporting information as "Residence time correlation functions of PP1" (Page 5).
(15) Page 9: 'This binding pathway is different from that observed in Src8BSA: PP1 first reaches the hinge region of the kinase and then intrudes into the canonical binding site.' This is true for 1/2 cases in figure S13. Please rewrite it.
The sentence was modified as follows (Page 10, line 195).
"PP1 first makes a contact with either the N-terminal side of G-loop or hinge region ( Supplementary Fig. 13) and then intrudes into the canonical binding sites." (16) Pages 9-10, lines [176][177][178][179][180][181][182][183][184][185][186][187][188][189][190] The argument to be made is that the simulation without restraints is equivalent to the simulation without crowder, while the simulation with restraints on kinase is equivalent to the simulation in the presence of crowder, which limits the motions of kinase, but this is not really apparent in the paragraph. Please rewrite it.
The following sentences were added on Page 11 of the revised manuscript.
"The binding trajectories in dilute solution fit to the major pathway of 'free' FEL ( Fig. 4(c)) and those in Src8BSA fit to the major pathway of 'restraint' FEL ( Fig. 4(d)). The major difference in two pathways is the interaction at an initial encounter state: PP1 interacts with the G-loop in dilute solution (E) ( Supplementary Fig. 12 Figure 17). Thus, we can claim that the crowder proteins alter the binding pathways. Following sentences were added in the subsection "Ligand-binding pathways in crowded environments" (Results on Page 11).
"To examine if the difference happened to be observed, we performed additional 30 simulations (20 ns each) in Src8BSA starting from E and Er states. While the simulations from Er state lead to three binding events, no binding was observed in the simulation from E state. Similar to the binding trajectories shown in Fig. 4(d) and Supplementary Fig. 16, the binding trajectories from Er state fit to 'restraint' FEL ( Supplementary Fig. 17). Thus, it can be safely stated that the binding pathways are altered in the presence of crowders." In addition, as the reviewer pointed out, since the both TYR-in/out conformations are present in all the systems (Fig. 5), the binding events could occur from both the pathways in all the systems if we extend the simulation length. Hence, we added the following sentences in Discussion (Page 14, line 285).
"Since the conformational changes between Tyr-in and Tyr-out happen as population shifts (Fig.   5), longer MD simulations in Src2BSA and Src4BSA could increase the possibilities of ligandbinding processes through one of the two pathways. " What is the conformation of Tyr82 in the apo form of kinase? It seems it is Tyr-out in figure S18 (in which the residues should be labelled).
Does the Tyr-in conformation appear in any crystal structure? Could the Tyr-in conformation be a force field artifact?
In the revision, we additionally performed 1 µs simulations for dilute and Src8BSA with different force field, CHARMM36m. Similar to the original results using AMBER forcefield, the population shift from TYR-in to TYR-out conformations is observed upon the crowding. As for the apo form crystal structures, TYR-out conformation appears in most cases, but several kinases have TYR-in conformation (for instance, protein kinase B/Akt). Following sentences were added.
('Conformational shifts of a Tyr sidechain upon crowding' (Results), on Page 13) "The additional 1 µs-simulations using a different force field, the CHARMM36m forcefield 58 , support that the observed population shift between TYR-in and TYR-out is universal ( Supplementary Fig. 18)." The atoms used for defining anchor atoms are determined from the crystal structure, and hence the anchor atoms do not change according to trajectory. We gave the identity of each anchor atom in the supporting information ('Definition of reaction coordinates for free energy landscapes' in Analysis Details, Page 3)

Reviewer #3 (Remarks to the Author):
In their manuscript   (1) Line 106 "in the inhibitor-unbound" has missing text.
The sentence was modified as follows (Page 5, line 106).
"The Ca root mean square fluctuations (RMSFs) of the kinase in the inhibitor-unbound state are also not affected by the presence of the BSAs … (abridged) …" (2) Line 175: The paragraph is repeated starting from the Line 190.
The paragraph was removed.