Making NSCLC Crystal Clear: How Kinase Structures Revolutionized Lung Cancer Treatment
Round 1
Reviewer 1 Report
Here, Groves et al., focuses on subclassification of lung cancer, non-small cell lung cancer (NSCLC), and major oncogenic driver 22 mutations in kinases and how crystallographic structures can be used not only to provide 23 awareness of the function and inhibition of these mutations, but also how these structures can be 24 used in further computational studies aiming at addressing these novel mutations in the field of 25 personalized medicine.Quite nice and interesting to read. I suggest following minor revisions.
- Is it necessary to include all VDW interaction residues in Fig. 3,4,6,7 9?
- Fig.5 is necessary? I feel it is redundant.
Author Response
Review 1
Comments and Suggestions for Authors
Here, Groves et al., focuses on subclassification of lung cancer, non-small cell lung cancer (NSCLC), and major oncogenic driver 22 mutations in kinases and how crystallographic structures can be used not only to provide 23 awareness of the function and inhibition of these mutations, but also how these structures can be 24 used in further computational studies aiming at addressing these novel mutations in the field of 25 personalized medicine.Quite nice and interesting to read. I suggest following minor revisions.
- Is it necessary to include all VDW interaction residues in Fig. 3,4,6,7 9?
- 5 is necessary? I feel it is redundant.
Many thanks to the referee for the positive evaluation of our review. Based on his/her advice we can modify figures 3 4 6 7 and 9 to highlight the most important van de Waals contacts. However, the presented figures are directly output from Discovery Studio Visualiser and we feel deleting specific interaction may be misleading.
Figure 5 is similar to figures 2 (EGFR) and 8 (BRAF) and we can combine these into a single figure that may help to provide a more global overview of the topics discussed.
Reviewer 2 Report
This is a very knowledgeable review about the structure, mechanism and function of four protein targets with a prominent role in development, prognosis and potential cure of Non-Small Cell Lung Cancer (NSCLC). The manuscript contains a wealth of information and a long list of relevant citations and the authors are clearly well versed in the field. Given that kinase research is such an active field, with lots and lots of publications coming out each year, it certainly is a challenge to condense all the information into a comprehensive review.
But while it is commendable how the authors try to be as comprehensive as possible, the text does unfortunately not result in a coherent narrative. The review is entitled “Making NSCLC crystal clear: how kinase structures revolutionized lung cancer treatment”, but one of the general problems is that the revolutionary aspects are merely implied through the novel insights structural biology has provided, but rarely explicitly mentioned or put into perspective. Another problem is that the authors write that they want to specifically highlight the contribution of computational structural biology for our understanding of these proteins, but then most of the data presented refer to crystal structures and biochemical studies etc., whereas computational methods show up only anecdotally. This is a pity, because at least in this reviewer’s opinion, only computational methods will ultimately be able to fulfil the promise of personalised medicine since most experimental methods simply do not scale sufficiently well. Hence, description and discussion of computational methods, as well as their critical assessment and relationship with experimental methods, should be much more elaborate, whereas some of the already well known and often reviewed structural details of protein kinases could be significantly condensed. A minor point, but worth considering: it is not obvious that the RAS family needs to be included here (despite their relevance for cancer research), because it is structurally and mechanistically different to the other three described protein targets. Also, the review would be more readable if subheadings would be introduced and if key residues and structural motifs like DFG-in/out, P-loop etc. would be introduced and explained in the beginning.
A comment on figures:
Figures 2, 5 and 8 are not very informative. Since the review is about protein structure, it would be useful if the structure would be enlarged so that important features become more discernible, whereas primary/ secondary structure could be shown as an inset. The insets A, B and C do not contribute to the understanding of the mechanism. It would also be helpful if additional figures would be included to highlight the molecular details of important mutations and/ or protein-ligand interactions described in the text. The 2D plots shown in figures 3, 4, 6 and 7 are useful for experts in the protein kinase field, but do not make the relationship to the structure understandable for the general reader.
Specific comments:
Line 35: The abstract says lung cancer is responsible for 10 million deaths annually; which number is correct?
Line 41: “Such” at the beginning of the sentence sounds odd.
Line 57: The paragraph about Molecular Modelling (MM) is unclear. The similarities and differences of MM with Molecular Dynamics are not obvious.
Line 77: What does “extensively used to obtain mutants of proteins with already elucidated structures” mean?
Line 97: What exactly is useful about it? It seems easy enough the generate models, but are they useful for generating new hypotheses that can be experimentally tested? It would be helpful to provide an example that illustrates the usefulness.
Line 116: Molecular Dynamics is indeed a powerful method, but MD calculations are neither trivial to set up, run or to analyse. Additionally, the computational requirement are still considerable.
Line 143: This sentence is unclear: Does the incidence depend on ethnicity, life style or gender?
Line 177: Please add a citation.
Line 182: “Magnesium divalent ion Mg2+” sounds strange
Line 191: What is “Molecular Mechanics Generalized Born Surface Area (MMGBSA) method”?
How does this method specifically help to understand the role of ATP binding and how does this relate to the subsequent sentences?
Line 221: “different from other drugs described later in this paper. Lapatinib” Why is this mentioned?
Line 230: Reference [36] dates back to 2006, hence, it does not seem to be very “recent”.
Line 251: What does this sentence mean, i.e. that MD simulations by sampling experimentally obtained conformations strengthens the reliability of MD methods? And most importantly, how would this be verified?
Line 371: What are “thermodynamic integration” studies? Can you elaborate?
Line 386: Can you put this observation into perspective? Even though the cancer becomes resistant to the drug, the binding affinity for its targets is still high. The way it is written is confusing.
Line 389: Why is there a new paragraph when the text continues with what was described in the previous paragraph. And what does “energetically expensive” mean? It would be helpful to clearer establish how experimental and computational methods complement each other here. There is a “(ref)” at the end of sentence.
Line 435: This is really interesting and deserves more details.
Line 463: Different citation style (superscript).
Line 466: Was the G724S mutation also found in the clinic or is it a mere prediction at this point?
Line 473: Can computational methods help to identify suitable candidates for repurposing?
Line 476: What are the insights? Did it make a difference in the clinic? Are there lessons to be learned?
Line 491: This is very interesting. But what does this concretely mean?
Line 545: The following paragraphs contains a detailed description of the structure of ALK which is mostly based on work referenced in [120] and to a lesser extent in [140] and [142]. In the next approximately 100 lines, the manuscript contains a detailed summary of these three publications, but only once is figure 5 mentioned. It would be helpful if the text would be condensed and key points being illustrated in figure 5.
Line 549: Should it be “it was” instead of “It is” since the structure was published 10 years ago.
Line 566: What is an accelerated MD study and what is the consequence of seamless transition?
Line 571: Please add context and relevance of this observation.
Line 605: Can computational studies really “confirm” experimental results or would “support’ be more appropriate?
Line 615: What does “ALK adopts a DFG-in conformation” refer to? Is it the conformation captured in the crystal structure that the citation refers to?
Line 750: Please elaborate.
Line 752: Please provide more context about “Computational studies are heavily reliant on x-ray crystallography”. The availability of experimentally determined protein structures is an obvious prerequisite, but does the quality of the model, as well as experimental limitations of the method influence computational analysis? Can you elaborate?
Line 779: Can you clarify what “is said to be” means?
Line 798: What does this mean? What did the computational studies actually achieve that led to testable hypothesis?
Line 804: “recently” is used twice in the sentence
Line 811: Why is there twice a reference to the work of Shaw et al in two subsequent paragraphs. Is this intentional?
Line 822: The work of Shaw et al is truly remarkable, but it is not clear how “This case is a prime example of computational biology in the setting of diagnosis and personalized medicine”? Please elaborate.
Line 944: Would “shows” be more appropriate than “suggests”?
Line 1098: Figure 5 label was already used before.
Line 1223: It is easy enough to set up a pipeline, but is this really a useful approach? Can you comment on the scientific and logistic bottlenecks of converting a software pipeline into a reliable source of novel lead compounds.
Line 1239: missing “there”
Line 1255: Although, most of the data presented stems from protein crystallography; the specific contribution of computational methods appeared more anecdotal.
Line 1259: This is not well described and mixes high level messages with details of each methods. Also, it seems of utmost importance to discuss this in more detail.
Author Response
Review 2
This is a very knowledgeable review about the structure, mechanism and function of four protein targets with a prominent role in development, prognosis and potential cure of Non-Small Cell Lung Cancer (NSCLC). The manuscript contains a wealth of information and a long list of relevant citations and the authors are clearly well versed in the field. Given that kinase research is such an active field, with lots and lots of publications coming out each year, it certainly is a challenge to condense all the information into a comprehensive review.
But while it is commendable how the authors try to be as comprehensive as possible, the text does unfortunately not result in a coherent narrative. The review is entitled “Making NSCLC crystal clear: how kinase structures revolutionized lung cancer treatment”, but one of the general problems is that the revolutionary aspects are merely implied through the novel insights structural biology has provided, but rarely explicitly mentioned or put into perspective. Another problem is that the authors write that they want to specifically highlight the contribution of computational structural biology for our understanding of these proteins, but then most of the data presented refer to crystal structures and biochemical studies etc., whereas computational methods show up only anecdotally. This is a pity, because at least in this reviewer’s opinion, only computational methods will ultimately be able to fulfil the promise of personalised medicine since most experimental methods simply do not scale sufficiently well. Hence, description and discussion of computational methods, as well as their critical assessment and relationship with experimental methods, should be much more elaborate, whereas some of the already well known and often reviewed structural details of protein kinases could be significantly condensed. A minor point, but worth considering: it is not obvious that the RAS family needs to be included here (despite their relevance for cancer research), because it is structurally and mechanistically different to the other three described protein targets.
We agree with the opinion of this reviewer that the future of personalised medicine will be driven by computational rather than experimental approaches. We have expanded our discussion on the computational methods, as well as their critical assessment and relationship with experimental methods.
Also, the review would be more readable if subheadings would be introduced and if key residues and structural motifs like DFG-in/out, P-loop etc. would be introduced and explained in the beginning.
We agree with the reviewer here and have made an introductory paragraph on the general structural features, and structural motifs.
A comment on figures:
Figures 2, 5 and 8 are not very informative. Since the review is about protein structure, it would be useful if the structure would be enlarged so that important features become more discernible, whereas primary/ secondary structure could be shown as an inset. The insets A, B and C do not contribute to the understanding of the mechanism. It would also be helpful if additional figures would be included to highlight the molecular details of important mutations and/ or protein-ligand interactions described in the text. The 2D plots shown in figures 3, 4, 6 and 7 are useful for experts in the protein kinase field, but do not make the relationship to the structure understandable for the general reader.
As requested by referee 1 we have modified these figures for clarity. We believe that the figure shows important interactions that provide the reader with an overview of the structure of the drugs and interacting amino acids extensively mentioned throughout the text.
Specific comments:
Line 35: The abstract says lung cancer is responsible for 10 million deaths annually; which number is correct? Corrections towards statistics were made based on Bray, F.; Ferlay, J.; Soerjomataram, I.; Siegel, R.L.; Torre, L.A.; Jemal. CA. Cancer J. Clin. 2018, 68, 394–424, doi:10.3322/caac.21492.
Line 41: “Such” at the beginning of the sentence sounds odd. Agreed and fixed
Line 57: The paragraph about Molecular Modelling (MM) is unclear. The similarities and differences of MM with Molecular Dynamics are not obvious. We have modified the text to make the similarities and differences of MM with Molecular Dynamics clearer.
Line 77: What does “extensively used to obtain mutants of proteins with already elucidated structures” mean? The mutational landscape regarding kinase protein is extensive and not all the known mutants have their structure already elucidated. In this case, homology modelling can propose a model using the structure of the wild-type structure of the same kinase, rather than another kinase. Kulman and Bradley provide a great insight of protein structure prediction on their review (Nat Rev Mol Cell Biol 20, 681–697 (2019). https://doi.org/10.1038/s41580-019-0163-x). The text has been modified to reflect this.
Line 97: What exactly is useful about it? It seems easy enough to generate models, but are they useful for generating new hypotheses that can be experimentally tested? It would be helpful to provide an example that illustrates the usefulness. An example has been provided
Line 116: Molecular Dynamics is indeed a powerful method, but MD calculations are neither trivial to set up, run or to analyse. Additionally, the computational requirement are still considerable.
We agree with this statement and have modified the text to clarify this point.
The authors would like to combine answers to review’s observations for line 97 and 116. As flawlessly pointed out by the reviewer, MD simulations are not trivial and despite major advances in making them accessible, there are still limitations to run seamless and comparable simulations on multiple mutations in a representative time scale. To overcome this limitation, either of time or knowledge, multiple articles make use of a homology model to visualize the location of a desired mutation and proximity to either the pocket of interest or or to regulatory structures. As one of the examples present on our paper, a model of mutant L505H on BRAF was obtained and based on its location in a hydrophobic pocket known for accommodating hydrophobic moieties from marketed drugs, the authors suggested a noxious effect of this mutation for the mentioned drugs. The group went with further studies using YUMAC cell lines Choi J, Landrette SF, Wang T, et al. Pigment Cell Melanoma Res. 2014;27(2):253-262. doi:10.1111/pcmr.12197
Line 143: This sentence is unclear: Does the incidence depend on ethnicity, life style or gender? We have clarified this sentence to indicate multiple factor impact upon incidence.
Line 177: Please add a citation. The issue was addressed
Line 182: “Magnesium divalent ion Mg2+” sounds strange. A more appropriate term is used
Line 191: What is “Molecular Mechanics Generalized Born Surface Area (MMGBSA) method”?
How does this method specifically help to understand the role of ATP binding and how does this relate to the subsequent sentences? This has been clarified in the text.
Line 221: “different from other drugs described later in this paper. Lapatinib” Why is this mentioned? This has been clarified in the text.
Line 230: Reference [36] dates back to 2006, hence, it does not seem to be very “recent”.This has been clarified in the text.
Line 251: What does this sentence mean, i.e. that MD simulations by sampling experimentally obtained conformations strengthens the reliability of MD methods? And most importantly, how would this be verified? The text has been modified for better understanding
Line 371: What are “thermodynamic integration” studies? Can you elaborate? This has been clarified in the text.
Line 386: Can you put this observation into perspective? Even though the cancer becomes resistant to the drug, the binding affinity for its targets is still high. The way it is written is confusing. This has been clarified in the text.
Line 389: Why is there a new paragraph when the text continues with what was described in the previous paragraph. And what does “energetically expensive” mean? It would be helpful to clearer establish how experimental and computational methods complement each other here. This has been clarified in the text.
There is a “(ref)” at the end of sentence. Solved
Line 435: This is really interesting and deserves more details. This section was expanded
Line 463: Different citation style (superscript).Solved
Line 466: Was the G724S mutation also found in the clinic or is it a mere prediction at this point? The mutation G724S was also found on patients after administration of osimertinib, as described by Oztan, A., et al. Lung cancer 111 (2017): 84-87 and Peled, Nir, et al. Journal of Thoracic Oncology 12.7 (2017): e81-e84.
Line 473: Can computational methods help to identify suitable candidates for repurposing? Drug repurposing is already an ongoing approach , as we now discuss in the text, however, there are still bottlenecks in the process socially considering the amount of drugs and diseases to be cross tested. In this scenario, computational tools have flourished as an attempt to accelerate the process, but indeed, experimental confirmations are , and always will be, indispensable.
Line 476: What are the insights? Did it make a difference in the clinic? Are there lessons to be learned? The major focus here lies on showing that repurposing might also be reconsidering drugs previously discarded, especially in face of mutations with unknown effect. The text was improved.
Line 491: This is very interesting. But what does this concretely mean? This has been clarified in the text.
Line 545: The following paragraphs contains a detailed description of the structure of ALK which is mostly based on work referenced in [120] and to a lesser extent in [140] and [142]. In the next approximately 100 lines, the manuscript contains a detailed summary of these three publications, but only once is figure 5 mentioned. It would be helpful if the text would be condensed and key points being illustrated in figure 5.
A section including general structural features of kinases was added and figure 5 was adapted following the reviewers advise.
Line 549: Should it be “it was” instead of “It is” since the structure was published 10 years ago. Solved
Line 566: What is an accelerated MD study and what is the consequence of seamless transition? This has been clarified in the text.
Line 571: Please add context and relevance of this observation. This has been clarified in the text.
Line 605: Can computational studies really “confirm” experimental results or would “support’ be more appropriate? Indeed, supporting is more suitable for the correlation of computational and experimental data
Line 615: What does “ALK adopts a DFG-in conformation” refer to? Is it the conformation captured in the crystal structure that the citation refers to? This has been clarified in the text.
Line 750: Please elaborate. For clarification, the introduction was rewritten to contribute to this section
Line 752: Please provide more context about “Computational studies are heavily reliant on x-ray crystallography”. The availability of experimentally determined protein structures is an obvious prerequisite, but does the quality of the model, as well as experimental limitations of the method influence computational analysis? Can you elaborate? The text was modified to include the relation between computational studies and three dimensional representation of proteins taking into consideration.
Line 779: Can you clarify what “is said to be” means? It was a inadequate use of words, this has been clarified in the text.
Line 798: What does this mean? What did the computational studies actually achieve that led to testable hypothesis? Extra info was added to the text for clarity.
Line 804: “recently” is used twice in the sentence. Solved.
Line 811: Why is there twice a reference to the work of Shaw et al in two subsequent paragraphs. Is this intentional? Due to the extensive work done in the mentioned paper, the authors opted for splitting the major information for clarity.
Line 822: The work of Shaw et al is truly remarkable, but it is not clear how “This case is a prime example of computational biology in the setting of diagnosis and personalized medicine”? Please elaborate. The authors praise for this work is due to the correlation of experimental data from other sources with the modeling approaches, This has been clarified in the text.
Line 944: Would “shows” be more appropriate than “suggests”? Indeed, the paragraph is related to a structure rather than a model.
Line 1098: Figure 5 label was already used before. Figures will be renumbered to adapt to reviewers request.
Line 1223: It is easy enough to set up a pipeline, but is this really a useful approach? Can you comment on the scientific and logistic bottlenecks of converting a software pipeline into a reliable source of novel lead compounds. Indeed, the referee presents a reasonable concern towards software development into a reliable pipeline. However, as mentioned in our text, Gupta and colleagues also implement biophysical and cell assays to further screen the compounds yielded by the computational arm of their presented pipeline
Line 1239: missing “there”.Solved
Line 1255: Although, most of the data presented stems from protein crystallography; the specific contribution of computational methods appeared more anecdotal. Indeed, the reviewer points out the limitations of the previous version of the review. However, with the useful comments from referees and the demonstration of Molecular modelling techniques on drug binding, effect of mutants in the conformational landscape and also on the mechanism of resistance of point mutations the authors hope to supply enough information.
Line 1259: This is not well described and mixes high level messages with details of each methods. Also, it seems of utmost importance to discuss this in more detail.
We agree on the importance and have extended and clarified the text.
Reviewer 3 Report
This is a long and very comprehensive review citing 320 papers. Its strength may lie in the area of drug design and molecular modelling, which are not my fields of expertise. In some other areas, however, the authors seem to not know very well what they are writing about. I consider this a major problem for a serious scientific review. The whole manuscript needs to be corrected with scrutiny and would probably profit from including one or two experienced senior authors to ensure scientific correctness.
The introduction is broad, covering e.g. molecular modeling methods and signal transduction pathways. However, to me the introduction appears superficial. Overall, the text does not excel at explaining the architecture and molecular function of the various kinases (and the GTPase RAS). Likewise, the figure and text explaining intracellular signal transduction pathways are not particularly helpful and partly even wrong.
In areas that I know quite well, I found several mistakes. Some mistakes are so basic that any undergraduate student with some basic knowledge in cell biology should spot them. E.g. in lines 1083 and 1098 RAS proteins are classified as kinases, which is plainly wrong. Although I am not an expert on kinase domains, I found in the main text about the different kinases several obvious mistakes just based on general knowledge of structural biology and protein function. I assume that the text could contain many more scientifically wrong statements that I did not notice.
Major issues
Line 49-51: PKA is a Ser/Thr kinase, not a tyrosine kinase
Line 135-138. I consider this sentence imprecise, misleading or even wrong. To me, it implies phosphorylation of the SOS proline-rich domain by ErbB family members, which does not happen, as these are Ser/Thr phosphorylations. As far as I know, these phosphorylations are not central to the activation of downstream signaling. The sentence also implies an important function of phosphorylation of the GAB1 proline-rich domain. As far as I know, phosphorylation of GAB1 by receptor tyrosine kinases is important, but not in its proline-rich domain.
Fig. 1 The figure is not particularly helpful. Probably any cell biology textbook offers a clearer picture of MAP kinase activation by EGFR. Also, figure and legend do not fit. The legend does not explain the difference between the EGFR and the ErbB receptor and does not mention the EML4-ALK receptor, which is mentioned in the text only much later. The legend refers to activation of the AKT pathway, which is not shown in the figure.
Line 180-182: I doubt (but I am not sure) that the statement is correct. I would except the aspartate side chain, not the Phe, to coordinate the Mg2+.
Here (https://pubs.acs.org/doi/full/10.1021/jm400072p), I found the following statement contradicting the description in this manuscript: “In active kinases, the DFG motif adopts an “in” conformation and the Asp residue is oriented toward bound ATP, which is able to coordinate the magnesium ion bound to the β- and γ- phosphate groups of ATP.(6) In inactive kinases, the conformation of the DFG motif is flipped outward, such that the Asp no longer coordinates the magnesium at the catalytic site.(7)“
Line 354-356: These lines contain an inherent contradiction. They mention both an increased Kd and an increased affinity of G719X mutations for ATP. However, increasing affinity decreases Kd. Checking the cited reference 35, I found that this paper describes for L858R and G719S a 50-fold and 10-fold, respectively, increase over wildtype of activity (kcat) not of affinity. The authors apparently mix up the concepts of affinity, Kd, Km, and activity (kcat). To me, such incorrect citation of data from the literature is not acceptable for a serious review article.
Line 445-448: The authors describe the effect of a leucine to glutamic acid substitution. However, the cited paper (reference 53) only describes the L718Q mutation, not the L718E mutation.
Line 450-452: Same as above. Q is not glutamic acid.
Line 511/513: SOS instead of SRC in SRC/RAS/MEK/ERK1/”
Line 559 GxGXΦG instead of GxGΦG; Otherwise this would also contradict Fig. 5 and the P-loop motif mentioned for EGFR in line 162 of this manuscript
Figure 8 the gray box highlighting the P-loop does not cover the correct residue range
Line 985 – 995 I thought that Sorafenib stabilizes the DFG-out and not the DFG-in conformation. Also the PDB entries 1UWH and 1UWJ should be in the inactive DFG-out state, not in the DFG-in state. The authors themselves classify Sorafenib as type II inhibitor that targets the inactive kinase domain αC-in/ DFG-out kinase state in lines 1013 – 1024.
Line1083: RAS proteins are not kinases
Line 1098 KRAS does not have a kinase domain, because it is a GTPase, but not a kinase
Line1131 Wildtype RAS contains glutamine in position 61 not glycine. Thus it should probably be Q61 and not G61. The authors themselves write this in line 1153 and 1159 of this manuscript.
Minor issues
Formatting of e.g. “et al., “in vitro” or the call out of figures is not uniform
Line 157/158: “a five stranded β-sheet” instead of “five stranded β-sheets”
Line 166 “structure” instead of “sequence”
Line 167: “P-loop” instead of “G-loop”?
Line 225 Src instead of Scr
Line 229: plotted instead of plotter
Line 340. A full stop is missing after osimertinib; plotted instead of plotter
Line 394 and 400: (ref) indicates that references might still be missing here.
Line 463: wrong format for reference 69.
Line 530: EGFR or Kirsten instead of EGFR of Kirsten? Otherwise I do not understand this sentence.
Line 575 P-loop instead of G-loop. The authors seem to use P-loop in the text and G-loop in their figure legends. They do not explain G-loop. Both terms can be used synonymously, but I recommend to consistently use only one abbreviation or to explain both.
Line 676 plotted instead of plotter
Line 717 plotted instead of plotter
Line 804 two times “recently”
Line 858 figure 8 instead of figure 4
Legend to Figure 8 P-loop instead of G-loop
Line 997 plotted instead of plotter
Line 1098 Figure 10 instead of Figure 5
Author Response
Reviewer 3
This is a long and very comprehensive review citing 320 papers. Its strength may lie in the area of drug design and molecular modelling, which are not my fields of expertise. In some other areas, however, the authors seem to not know very well what they are writing about. I consider this a major problem for a serious scientific review. The whole manuscript needs to be corrected with scrutiny and would probably profit from including one or two experienced senior authors to ensure scientific correctness.
We would like to thank the referee for the comment and his/her detailed reading. We have addressed the comments raised by the referee as indicated below as well as carefully re-examining the manuscrlpt’s contents. As the referee points out this is a long and very comprehensive review and we appreciate the effort involved from the 3 referees in identifying the errors in our initial submission. We believe that the comments from all referees have resulted in a significantly improved review.
The introduction is broad, covering e.g. molecular modeling methods and signal transduction pathways. However, to me the introduction appears superficial. Overall, the text does not excel at explaining the architecture and molecular function of the various kinases (and the GTPase RAS). Likewise, the figure and text explaining intracellular signal transduction pathways are not particularly helpful and partly even wrong.
In areas that I know quite well, I found several mistakes. Some mistakes are so basic that any undergraduate student with some basic knowledge in cell biology should spot them. E.g. in lines 1083 and 1098 RAS proteins are classified as kinases, which is plainly wrong.
We apologies for these two instances and have corrected this and other misidentifications of KRAS as a kinase, but we do state on line 1065 “The RAS family is a membrane-bound GTPase family“. However, this is an egregious error and we have carefully checked the KRAS section.
Although I am not an expert on kinase domains, I found in the main text about the different kinases several obvious mistakes just based on general knowledge of structural biology and protein function. I assume that the text could contain many more scientifically wrong statements that I did not notice.
We have corrected the mistakes raised by the 3 referees and checked the manuscript for accuracy.
Major issues
Line 49-51: PKA is a Ser/Thr kinase, not a tyrosine kinase The authors recognize and apologize by the misuse of the kinase classification.
Line 135-138. I consider this sentence imprecise, misleading or even wrong. To me, it implies phosphorylation of the SOS proline-rich domain by ErbB family members, which does not happen, as these are Ser/Thr phosphorylations. As far as I know, these phosphorylations are not central to the activation of downstream signaling. The sentence also implies an important function of phosphorylation of the GAB1 proline-rich domain. As far as I know, phosphorylation of GAB1 by receptor tyrosine kinases is important, but not in its proline-rich domain. The authors apologize if the sentence was unclear. The main goal was to explain the cascade effects of TRK activation and we have now modified the text to clarify.
Fig. 1 The figure is not particularly helpful. Probably any cell biology textbook offers a clearer picture of MAP kinase activation by EGFR. Also, figure and legend do not fit. The legend does not explain the difference between the EGFR and the ErbB receptor and does not mention the EML4-ALK receptor, which is mentioned in the text only much later. The legend refers to activation of the AKT pathway, which is not shown in the figure. Figure was adapted to fit referee’s comments
Line 180-182: I doubt (but I am not sure) that the statement is correct. I would except the aspartate side chain, not the Phe, to coordinate the Mg2+. This has been corrected in the text.
Here (https://pubs.acs.org/doi/full/10.1021/jm400072p), I found the following statement contradicting the description in this manuscript: “In active kinases, the DFG motif adopts an “in” conformation and the Asp residue is oriented toward bound ATP, which is able to coordinate the magnesium ion bound to the β- and γ- phosphate groups of ATP.(6) In inactive kinases, the conformation of the DFG motif is flipped outward, such that the Asp no longer coordinates the magnesium at the catalytic site.(7)“ This has been corrected in the text.
Line 354-356: These lines contain an inherent contradiction. They mention both an increased Kd and an increased affinity of G719X mutations for ATP. However, increasing affinity decreases Kd. Checking the cited reference 35, I found that this paper describes for L858R and G719S a 50-fold and 10-fold, respectively, increase over wildtype of activity (kcat) not of affinity. The authors apparently mix up the concepts of affinity, Kd, Km, and activity (kcat). To me, such incorrect citation of data from the literature is not acceptable for a serious review article. We have carefully reviewed this section to be both clearer and consistent with the cited work.
Line 445-448: The authors describe the effect of a leucine to glutamic acid substitution. However, the cited paper (reference 53) only describes the L718Q mutation, not the L718E mutation. Line 450-452: Same as above. Q is not glutamic acid. This has been corrected in the text.
Line 511/513: SOS instead of SRC in SRC/RAS/MEK/ERK1/”Issue is address on the text
Line 559 GxGXΦG instead of GxGΦG; Otherwise this would also contradict Fig. 5 and the P-loop motif mentioned for EGFR in line 162 of this manuscriptI This has been corrected in the text.
Figure 8 the gray box highlighting the P-loop does not cover the correct residue rangeThis has been corrected in the text.
Line 985 – 995 I thought that Sorafenib stabilizes the DFG-out and not the DFG-in conformation. Also the PDB entries 1UWH and 1UWJ should be in the inactive DFG-out state, not in the DFG-in state. The authors themselves classify Sorafenib as type II inhibitor that targets the inactive kinase domain αC-in/ DFG-out kinase state in lines 1013 – 1024. This has been corrected in the text.
Line1083: RAS proteins are not kinases The authors recognize and apologize for the misuse of the words kinase domain.
Line 1098 KRAS does not have a kinase domain, because it is a GTPase, but not a kinase. The authors recognize and apologize for the misuse of the words kinase domain.
Line1131 Wildtype RAS contains glutamine in position 61 not glycine. Thus it should probably be Q61 and not G61. The authors themselves write this in line 1153 and 1159 of this manuscript. Indeed, from the uniprot code mentioned on the text it is a Q61. This has been corrected in the text.
Minor issues
Formatting of e.g. “et al., “in vitro” or the call out of figures is not uniform. Solved
Line 157/158: “a five stranded β-sheet” instead of “five stranded β-sheets” Solved
Line 166 “structure” instead of “sequence”Solved
Line 167: “P-loop” instead of “G-loop”? As mentioned further in this review, all the G-loop references were modified into P-loop
Line 225 Src instead of Scr Solved
Line 229: plotted instead of plotter Solved
Line 340. A full stop is missing after osimertinib; plotted instead of plotter Solved
Line 394 and 400: (ref) indicates that references might still be missing here.Solved
Line 463: wrong format for reference 69. Solved
Line 530: EGFR or Kirsten instead of EGFR of Kirsten? Otherwise I do not understand this sentence. Solved
Line 575 P-loop instead of G-loop. The authors seem to use P-loop in the text and G-loop in their figure legends. They do not explain G-loop. Both terms can be used synonymously, but I recommend to consistently use only one abbreviation or to explain both.Solved
Line 676 plotted instead of plotterSolved
Line 717 plotted instead of plotter Solved
Line 804 two times “recently” Solved
Line 858 figure 8 instead of figure 4 The final numbering was modified to adapt to reviewers suggestions
Legend to Figure 8 P-loop instead of G-loopSolved
Line 997 plotted instead of plotter. Solved
Line 1098 Figure 10 instead of Figure 5. The final numbering was modified to adapt to reviewers suggestions
Round 2
Reviewer 2 Report
The authors took many of the suggestion raised by the reviewers on board and spent a significant effort on reworking of the text. The overall structure of the manuscript has improved and there is now a clearer emphasis on the contribution of computational structural biology. The manuscript still makes for dense reading, but this obviously depends on personal preference. However, there are still many small mistakes like missing characters, articles etc. (see examples in list below, but which is not exhaustive) and the text would therefore benefit from another round of careful reading. Also, some sentences could be rephrased to make them easier to understand.
Line 71: duplication: “homology modelling homology model”
Line 76: “model” – plural instead?
Line 77: “model” – plural instead?
Line 116: “has” instead of “have”?
Line 116: What specifically has improved? Speed or accuracy?
Line 130: missing “and”: “…threonine tyrosine…”
Line 159: remove “present”: “…αC-helix is present orientated toward the ATP binding pocket…”
Line 393: “and the” twice
Line 393: remove “the: “…performed by the Tamirat et al…”
Line 395 add “to”: “…favored due stabilization…”
Line 469: It is not clear what this sentence means: “…loss the double mutant diminished affinity for gefitinib…”
Line 486: Can you clarify what this means: “even in different spatial coordinates”
Line 487 add “d”: “…Park an…”
Line 583 add “to”: “…confers resistance gefitinib…”
Line 730 Which hydrogen atom: “The hydrogen atom of C1097…”
Line 822: add “by”: “…screening of Novartis…’
Line 929: add space between words: “calculationsrevealed”
Line 933: “…, including the van der Waals, …” sounds odd
Line 944: double twice? “…the L11198F single or double L1198F double mutant…”
Line 945: The sentence starting with “The discordance of…” sounds quite convoluted, please try to rephrase
Line 965: remove underscore: “reveal_ed”
Line 1089: “resides”; should this be residue instead?
Line 1221: add “of”: “… a plethora…”
Line 1304: residues
Line 1344: “into tephosphate-binding pocket”
Line 1411 add “of”: “…in the study…”
Line 1420: “phenomenon” 
- plural instead?
Author Response
Referee 2
Line 71: duplication: “homology modelling homology model”
Line 76: “model” – plural instead?
Line 77: “model” – plural instead?
Line 116: “has” instead of “have”?
The comments above were addressed in the text
Line 116: What specifically has improved? Speed or accuracy?
In both, in terms of accuracy, newer force field are able to simplify chemical systems over millisecond instead of the smaller, but more common, nanosecond range.
Line 130: missing “and”: “…threonine tyrosine…”
Line 159: remove “present”: “…αC-helix is present orientated toward the ATP binding pocket…”
Line 393: “and the” twice
Line 393: remove “the: “…performed by the Tamirat et al…”
Line 395 add “to”: “…favored due stabilization…”
The comments above were addressed in the text
Line 469: It is not clear what this sentence means: “…loss the double mutant diminished affinity for gefitinib…”
Indeed, the phrase was structurally incorrect, the word loss was incorrectly placed. This has been addressed in the text.
Line 486: Can you clarify what this means: “even in different spatial coordinates”
During MD simulations the protein of interest has free movement, in long simulation is not uncommon to see large distances identified when comparing spatial coordinates. To overcome such problems, there is an alignment of the final pose of the protein with its initial, To be sure that any modifications on its structure are due to conformational changes rather than just artificial movement during simulations.
Line 487 add “d”: “…Park an…”
Line 583 add “to”: “…confers resistance gefitinib…”
The comments above were addressed in the text
Line 730 Which hydrogen atom: “The hydrogen atom of C1097…”
The backbone of C1097 serves as hydrogen bond donor to side chain hydroxyl group of Y1278.
Line 822: add “by”: “…screening of Novartis…’
Line 929: add space between words: “calculationsrevealed”
Line 933: “…, including the van der Waals, …” sounds odd
Line 944: double twice? “…the L11198F single or double L1198F double mutant…”
Line 945: The sentence starting with “The discordance of…” sounds quite convoluted, please try to rephrase
Line 965: remove underscore: “reveal_ed”
Line 1089: “resides”; should this be residue instead?
Line 1221: add “of”: “… a plethora…”
Line 1304: residues
Line 1344: “into tephosphate-binding pocket”
Line 1411 add “of”: “…in the study…”
Line 1420: “phenomenon” 
- plural instead
Our thanks again for careful reading. All the above changes have been made in the resubmitted manuscript.
Reviewer 3 Report
I checked the changes that the authors made in response to my comments. All my points have been addressed adequately. Overall, I think that the authors substantially improved their manuscript. I still found two point that should be addressed before publication
- In Fig. 1D a part of the C-lobe seems to be coloured in yellow, which is used for the N-lobe. I think that, roughly speaking, everything below the ATP should be blue. If true, this should be corrected. In addition, I would find it useful to highlight in the structure also other sequence elements mentioned in the text and shown in Fig. 1A-C, i.e. the alphaC helix, the hinge region, the DFG motif and the A-loop.
- Line 656: This should probably also be “SOS/RAS/ERK1/2” instead of “SRC/RAS/ERK1/2”
Author Response
Referee 3
I checked the changes that the authors made in response to my comments. All my points have been addressed adequately. Overall, I think that the authors substantially improved their manuscript. I still found two point that should be addressed before publication
- In Fig. 1D a part of the C-lobe seems to be coloured in yellow, which is used for the N-lobe. I think that, roughly speaking, everything below the ATP should be blue. If true, this should be corrected. In addition, I would find it useful to highlight in the structure also other sequence elements mentioned in the text and shown in Fig. 1A-C, i.e. the alphaC helix, the hinge region, the DFG motif and the A-loop.
- Line 656: This should probably also be “SOS/RAS/ERK1/2” instead of “SRC/RAS/ERK1/2”
We appreciate the referee’s comments and implemented the mentioned modifications, with emphasis on the suitable additions to figure 1.
