The January 2005 version of the Neuraminidase paper: Reviewers’ comments (in blue) and my replies (in red). This version dealt only with N2 and N9 type neuraminidases.
Note that page, paragraph, Figure and reference numbering do not correspond with the latest version of the paper, but this does not confuse the meaning of the dialogue.
Click to return to http://nativeproteins.blogspot.com
Title: Identification of a native arrangement of subunits in a tetrameric enzyme.
First cover letter: Symmetry relationships arising from the theory of Natural Selection, together with biophysical requirements for a tetrahedral structure, lead to a new quaternary structure for influenza virus neuraminidase. The atomic coordinates of the enzyme subunits in crystals fit very well into the new tetrahedral arrangement. The locations of glycosylation sites also support the tetrahedral structure. These results, and the theory that led to them, link physics, chemistry and biology in a simple description of the process by which living things transform materials. The investigation can be reproduced and extended by any scientist using publicly available data and software. The results give us insight into protein structure and function, the role of water and the physics of a real molecular machine. They provide a new way to use the results of x-ray crystallography to generate meaningful protein structures. This discovery will provide clues to the discovery of the catalytic forms of other enzymes and the functional native conformations of many other proteins.
Reviewer comments
Reviewer #1: Ms. Title: Identification of a native arrangement of subunits in a tetrameric enzyme Corresponding Author: Donald George Vanselow, PhD This manuscript describes a re-arrangement of subunits from the tetrameric enzyme of the influenza virus neuraminidase. Guided by various biophysical requirements and symmetry, the author identified a new quaternary structure for this enzyme. All sections are clearly written with the key selling point, that is, an enzyme must be able to constrain its active site. Overall, the paper is presented in a simple and logical way. I would support its publication in …… if the author could further comment on how this new quaternary structure would impact on design inhibitors against neuraminidase. As neuraminidase is an essential target for development of antiviral agents to treat influenza virus infections, the information regarding its structure is of particular significance to design novel and potent inhibitors against this enzyme or for explanation of the inhibition and/or resistance mechanism of existing drugs.
Reply to #1: I have added a new paragraph 3.11 dealing with the possible implications of this discovery for immunological and enzyme inhibition approaches to treatment of influenza virus infections.
Reviewer #2: I would say that the paper is in need of very serious revision before it is suitable for publication. Dr. Vanselow has proposed that enzymes work by physical constraints placed on their substrates, a theory which he suggests would naturally lead to a preference for homotetrameric forms, arranged tetrahedrally. The crystal structure of neuraminidase shows a single copy of the protein, which forms a tetramer through four-fold crystallographic symmetry. Crystallography is not always able to show unequivocally whether the form observed in the crystal is the form found in solution. It is therefore reasonable to ask if another type of arrangement can be found. Following the author's instructions I recreated his model for PDB 1NN2. Unfortunately I find many steric clashes which cannot be described as "very minor". Reply re “Steric Clashes”. When I described the clashes as very minor, I did so from the point of view of energetic implications. I have rewritten the paragraph and described the clashes as "apparent" to make it clear that the clashes would never occur in reality but would be side-stepped by the affected side-chains moving aside as in the "soft docking" literature. I have rewritten the paragraph to make that clear. Furthermore, I have increased the distance of each subunit from the origin by 0.87 Å and included as supplementary data, a set of rotamers from the Swiss PDBViewer library that eliminate almost all steric clashes. Since the rotamer structures are accessible at virtually no energy cost, the clashes they eliminate can be considered "apparent".
The shape of the fit does indeed seem reasonable to the eye, viewing the whole tetramer, but a calculation of buried surface areas suggests that over 3000 A2 of surface area is buried by the monomer in the crystal structure. This is quite considerable, and the interfaces match beautifully. Reply re “Buried Surface”. The discrepancy between the reviewer's quoted value for the buried surface area and the value reported in my paper was introduced by the reviewer using different software with a different definition of surface area. Using the software recommended by the reviewer I have been able to characterise the buried surface area and the residual crevices using multiple measurements at various probe radii. This is a much more robust technique than the single measurements in my original paper as it is not so sensitive to the closeness of approach of the subunits. These results are shown in a new Figure 3.
Reply re The beauty of the match in the crystal. I have written new material into para 3.10 pointing out the ambiguity of interpretation of complementary surfaces on soft materials in a crystal.
The author does not mention the pattern of hydrogen bonding or hydrophobic patches in his model. Reply re “Patterns of hydrogen bonding or hydrophobicity within the interfaces”. I have extended the theory and investigation into the detail of the interfaces in new paragraphs 3.8 and 3.9. Since interactions between charged groups are an order of magnitude stronger than hydrogen bonds and are not ubiquitous, they are much better indicators of the action of Natural Selection than the features suggested by the reviewer. We must be careful not to introduce preconceptions about the nature of a reversible interface in a protein, because this interface may be the first to be described.
To me, the four-in-a-ring model seems rather more convincing. The author has examined his model "visually" (page 4) and by "inspection" (page 5), but this is not enough. Reply re “Comments calling for more than visual inspection of interfaces”. As described above, I have more fully characterised the residual crevices and buried surface area and supplied a set of rotamers from a rotamer library that largely eliminate steric clashes.
A very large buried area means little if it is achieved by forcing two protein molecules unrealistically close together. Reply re “Forcing proteins together to achieve large buried area”. See "Steric clashes" above. Because there are negligible energy costs in bringing the subunits together, there can only be negligible forces involved. What I have done is better described as "Soft Docking". With the new Figure 3 the large buried area can be seen to not depend on the amount of steric clashing.
Other points that the author needs to address are: Page 4. Cys 92 is described as the point of attachment of the stalk or tether to neuraminidase. But Cys 92 forms a disulphide with another cysteine in the protein and is unavailable for attachment to any such tether. My reading of the literature tells me the stalk or tether is the N terminal region of the polypeptide chain. Reply re Cys92. I have included a more extensive discussion of the point of attachment of the tether in paragraph 3.2. The reviewer has not realised that, for such a large protein, the N-terminal region includes up to residue 91, which is, without a doubt, attached to residue 92.
Experiments have been performed to change the length of this chain, without functional effect. This seems to invalidate the author's suggestion that Cys 92 need lie in a plane containing two symmetry axes (page 6). Reply re “Effect of changing chain length”. I have not argued that a particular chain length is required for function. The argument is biological rather than physical. I have argued that a certain point of attachment can simultaneously satisfy two otherwise competing pressures of Natural Selection. One pressure that we can always assume is that if a shorter chain functions just as well as a longer one, then a shorter chain will be preferred as it is quicker to synthesise and requires less energy during reproduction. The competing pressure is presumably that a longer chain functions slightly better. There may be a variety of chain lengths that achieve a reasonable balance between these competing pressures and, indeed, some variation is observed in nature. However, there is only one point of attachment which provides the maximum effective length or the minimum number of amino acids in the chain. I believe, from discussion with another biologist, that this argument is uncontroversial and does not need to be laboured in the text. I have rewritten the argument emphasising the minimisation of number of amino acids so that the reader will more likely recognise this as a biological argument.
The function of neuraminidase is to release the virion from the cell surface. How is it to do this if the active sites are found within a tetrahedral structure? Some fitting is required to show the proposed channel is adequate to allow substrate access. Reply re “Access of substrate to active site”. This question is addressed fully in reference (1) (the Constraint Theory paper) and again in paragraph 3.10. I have rewritten the paragraph to emphasise the dynamic nature of the model. I am mindful of the fact that a number of potential readers do not look upon proteins as soft, flexible and dynamic structures and will have difficulty with the idea of a set of native conformations. However, there is a perceptible shift of opinion towards flexibility and dynamic behaviour in recent years. Because, in nature, this enzyme acts on the terminal group of a very large glycoprotein molecule, there needs to be allowance in its structure for the enzyme to envelop the terminal chain while it is still attached to the external protein. The small channels at the intersections of interfaces seem to allow for that. As suggested by the Reviewer, I have fitted a disaccharide into the narrowest part of the channel to demonstrate this possibility. The coordinates of this molecule are included in my Supplementary Data.
Why is it a problem with the PDB structure that a non-glycosylated site is external? There are many non-glycosylated asparagine residues on proteins which are glycosylated. Reply re “Location of unglycosylated site”. I followed up the reviewer's remarks about the common occurrence of unoccupied glycosylation sites on otherwise glycosylated proteins. It turns out that the current view of experts in the field is still the same as it was when Varghese and Colman (1991) [2] felt compelled to try to explain the non-glycosylation of Asn402. I have included a more recent reference to the current understanding of glycosylation, that a structural explanation is expected for any such unoccupied glycosylation site. The reviewer's criticism is unfounded.
To describe the electron microscopy images as "obviously an artefact" is unsubstantiated. EM has revealed many protein shapes, some decidedly non-planar. The EM results agree with the PDB structure, and cannot be dismissed so lightly. Reply re “Electron microscopy”. The Reviewer has missed the word "latter" in my original manuscript and so I have rewritten this paragraph. The point of the paragraph is to draw attention to the limited ability of these EM images to corroborate the X-ray structure. This is not the same as dismissing them. It is about discouraging undue weight being put upon them, in line with the cautious language used by the original author (Colman [7]). I have expanded this paragraph to make the case more strongly. I have discussed the similarities between the process of adsorption of a protein to an EM grid and the process of adsorption which is integral to protein crystal growth. I have referred to a publication discussing the artefacts from adsorption to an EM grid. I would also draw the reviewer's attention to the hazard of extrapolating from some proteins, which may well retain non-planar shape on an EM grid, to any other protein. Without some underlying scientific principle for guidance, extrapolation is dangerous. It is my view that most of the perceived evidence in support of crystallography being 100% correct is just this kind of unfounded extrapolation. When one looks for evidence supporting any one crystallisation being 100% correct, such as this one, there is very little to find. I have added a new paragraph 3.12 drawing attention to how little is known about the crystallisation step.
Dr Vanselow should consider finding other proteins with which to test his hypothesis. There are many enzymes, very thoroughly described, which are clearly not tetrameric. Can these be fitted to his hypothesis? Reply re “Other proteins”. Neuraminidase was only the first protein structure to be reconstructed. Others will follow. It was particularly suitable because its biochemistry was well known and its structure well studied. I should point out that constraint theory does not specify homotetramers; I chose a homotetramer because of the simplifying effect of its symmetry. The general location of the active site was known. The location of the tether attachment provided a strong clue to the likely orientation and the glycosylation pattern was useful in providing supporting evidence. The steric and electrostatic fit was as good as could be expected if one assumed that the native form had been rearranged by crystallisation. Finally, I have been able to reconcile the immunology data with the proposed native structure. This is included in the new paragraph 3.11. I have added to paragraph 3.12 the fact that the proposed structure was achieved by altering less than 1% of the spatial information in the PDB file (origin, orientation and point group). This work should not be seen as in conflict with that technique but rather as a qualified validation of the technique.
I recommend the CCP4 suite of programs, and in particular DISTANG, for the measurement of clashes in molecular models.
Second cover letter accompanying the revised paper:
Dear Editor
The target audience for this paper is a diverse range of protein scientists ranging from those who see protein molecules as dynamic structures with real mechanical properties through to those who see a protein molecule as a set of precise coordinates defining a rigid body. As this work is grounded in the former view it could be initially unpalatable to some of those holding the latter view. It is my intention to make the paper as accessible as possible to as wide a range of protein scientists as possible. I need to limit the scope for misreading and misrepresentation of the paper. The work described here requires very careful thought if it is to be understood, although I have tried to reduce the line of reasoning to its simplest form.
Editor’s final refusal letter:
Dear Dr. Vanselow:
On behalf of the Executive Editors of ………….., I would like to thank you for submitting the above-named, revised manuscript to ………………...
Normally, revisions are reviewed by one or more or the original reviewers. However, the original two reviewers of your manuscript were not available to review the revision, and so we obtained a third reviewer who is also an expert in the field.
The reviewer's comments are enclosed below. As you can see, the reviewer raises substantial objections and recommends refusal. After careful consideration, I uphold the recommendation. This decision must be considered final.
I regret that I could not be more positive on this occasion.
Yours sincerely,
Executive Editor
Please see the Reviewer's comments
Reviewer #3: I did not referee the first draft of this manuscript, so my comments will address both the manuscript as a whole and the author's revisions to meet the criticisms of the first two referees (Referees 1 and 2).
This paper proposes a new structure for influenza virus neuraminidase based upon some rather crude docking experiments plus the notion that the reason for many enzymes to exist in tetrameric form is that compressive stress adds to the rigidity of the molecule and enables catalysis. The proposed neuraminidase structure is quite different from that observed for all influenza virus structures determined to date, both by X-ray crystallography and by electron microscopy. Neuraminidase activity per se does not require a tetrameric association of the enzyme (the bacterial enzymes are monomers [1] and the paramyxoviral enzymes are loosely-associated dimers [2]).[My comment: In the expanded version of the paper available on this BLOG, this claim that monomers and dimers exist is shown to be an X-ray crystallography observation and therefore cannot be used to support X-ray crystallography] However in the context of influenza virus there appears to be a good reason for the four-fold symmetric arrangement observed in the crystal and EM structures - namely that of orienting the enzyme towards the infected cell wall in order that it may undertake its function of assisting the release of progeny via removal of receptor.
Thus, given that the author's proposed structure is contrary to observation, I consider that the burden of proof regarding the correctness of the proposed structure lies with the manuscript author. [My comment: “burden of proof” is a legal term that has no place in science, in which nothing can be proven. Our burden is to keep an open mind and search for models and explanations with the widest possible application, consistency with other areas of science, and simplicity]. The entire docking experiment appears to be done manually. Protein-protein docking is not a trivial task [3], even when guided by prior information. Simple orientation by the author of the neuraminidase monomer with its propellor axis (defined how?) along the (1,1,1) axis and then "minor adjustments" (process not stated) to "optimise the three interfaces" (optimization protocols not defined) do not suffice to lend credibility to the results. There are no control docking experiments to indicate that this mode of docking is any better than any other. Why for example, does the author not consider other modes of tetrahedral association, e.g. more open forms that have each monomer interacting with only two rather than three neighbours. [My comment: the aim was to find forms consistent with Constraint Theory]. An example of this would be the tetrameric N-acteyl neuraminate lyase [4] . Or does the author also consider this latter structure and its homologues (DHDPS, etc.) artefactual? [My comment: Of course, it could be.]
Whilst the above comment is a criticism of methodology, I am also critical of the arguments that the author raises to support his conclusions. I am unconvinced by the argument that the complementarity of the subunit-subunit interfaces observed in the crystal structures are simply a consequence of packing forces. It is well known that crystal contacts themselves (in this case being tetramer-tetramer) are in general rather tenuous and occlude relatively little buried surface area compared to genuine subunit-subunit interfaces [5]. [My comment: Again this is X-ray crystallography data that cannot be used to support X-ray crystallography].This implies that it is unlikely that crystal packing forces that are on the one hand not strong enough to generate well-packed interfaces at the tetramer-tetramer level are able on the other to generate exquisitely [My comment: All crystals are exquisite] well-packed and extensive interfaces at the subunit-subunit level. The author does not provide quantitative assessment of the interface packing density and shape complementarity [8] or compare these to crystal structure. I am thus not ready to dismiss the observed neuraminidase subunit-sununit surfaces as an artefact of crystallization, especially when they have been observed repeatedly. [My comment: Repeatability is a prerequisite for crystal formation.]
In terms of the electron microscopy data, whilst I readily admit that adsorption can lead to artefact, the overwhelming body of data is in favour of the four-fold tetrameric shape. The author has made no reference at all to the Fab-complexed structures that have been observed both in crystal structures and in 2-D crystals observed by electron microscopy and in isolation of electron microscope grids [6,7]. Again, these support the four-fold symmetric structures observed in the 3D crystals. Note that the 2-D crystals are formed prior to adsorption onto EM-grids and that symmetry considerations would severely restrict the ability of the lattice to re-arrange, simply via adsorption forces, from the tetrameric shape proposed by the author to that which is actually observed. [My comment: If one refers back to the original paper describing these EM images (Tulloch et al, 1986, J Mol Biol 190, 215-225) one finds little or no evidence of four-fold symmetry in the individual tetramers. Rather, that assumption has been made based on the X-ray crystal structure. Again a circular argument has occurred, this time not so obviously.]
Whilst the author cites the lack of glycosylation at Asn 402 and the instability of the E119G mutant as being consistent with his model, these arguments are not in themselves strong enough to overcome the weakness in the methodology adopted by the author. [My comment: The methodology is not the basis of my argument; the observations are. I acknowledge that a shotgun approach to gathering data has become more commonplace since computer use was introduced and that many people think that data gathered that way has more credibility]. The author should also consider whether or not his model places the hemabsorbing site of the N9 neuraminidase in a surface exposed position [9]. [My comment: Constraint theory would suggest that the hemabsorbing site should be buried, which it is.]
I thus am opposed to the publication of this work. If the author is intent on pursuing this idea, then I would suggest that he team up with experimentalists and perform small-angle X-ray scattering experiments at a synchrotron and use that technique to determine the shape of the molecule in solution. [My comment: A brief search of such studies suggests to me that the technique is not yet able to determine the shape of a molecule except very approximately. I would like to be directed towards the “state of the art”].
[1] Proc Natl Acad Sci U S A. 1993 Nov 1;90(21):9852-6.
[2] J Mol Biol. 2004 Jan 30;335(5):1343-57.
[3] Protein Eng. 2001 Feb;14(2):105-13.
[4] Structure. 1994 May 15;2(5):361-9.
[5] Protein Sci. 1997 Oct;6(10):2261-3
[6] Ultramicroscopy. 1991 Feb;35(2):131-43.
[7] Nature. 1987 Mar 26-Apr 1;326(6111):358-63.
[8] J Mol Biol. 1993 Dec 20;234(4):946-50.
[9] Proc Natl Acad Sci U S A. 1997 Oct 28;94(22):11808-12.
My comment: The above criticisms were very useful in continuing the work. Most importantly, extension of the work to related neuraminidases was crucial and achievable. Also it was a simple matter to define the propeller axis for future work.
The January 2007 version of the Neuraminidase paper: Reviewers’ comments (in blue) and my responses (in red). This is the latest version.
Title: Tetrahedral structures in the neuraminidases with
special reference to influenza neuraminidase.
Cover letter: Virtually the same as that accompanying the 2005 version.
Editor’s refusal letter:
Dear Dr. Vanselow,
Your manuscript entitled "Tetrahedral structures in the neuraminidases with special reference to influenza neuraminidase" has been reviewed. As you can see from the comments made by the referees, both disagree with the hypothesis put forward in the paper. Consequently, unfortunately, your paper is not acceptable for publication in …………... Please visit http://.......................to view the reviewers' comments.
Thank you very much for giving us the opportunity to review this work. Although the outcome was not favorable, I trust you will find the referees' suggestions of value as you continue your efforts in this area.
Sincerely,
Editor
My comment: I am uncertain whether the Editor chose carefully the phrase “disagree with the hypothesis” or intended to say that the results did not support the hypothesis. The tenor of the Reviewer Comments below certainly suggests that the reviewers disagreed with the hypothesis from the start and therefore did not consider the results or whether the hypothesis was supported. To disagree with the hypothesis is not a scientifically rational position to take.
Reviewer comments
Reviewer 1
The author presents no credible evidence to support the claim that enzymes favor certain tetrahedral arrangements. The known structures in the PDB do not support the idea, so the author has undertaken a model building exercise to argue that neuraminidase uses such an arrangement. The proposed arrangement differs from the quaternary structural arrangements reported in earlier experimental crystal structures. The author argues that ignoring the symmetry information in the PDB amounts to only overlooking 1% of the information. This makes no sense. It also ignores decades of evidence that the crystal structures of enzymes generally give reliable views of their quaternary structures.
My comment: The reviewer has not addressed the results and arguments in the paper, but merely quotes the traditional view that the paper sets out to criticise.
Reviewer 2
The manuscript submitted by Dr. Don Vanselow “Tetrahedral structures in the neuraminidases with special reference to influenza neuraminidase” presents the author’s attempt to construct quaternary structures for neuraminidases based on D2 symmetry instead of the C4 symmetry observed in the crystal structure of influenza virus neuraminidase. The author then proceeds to interpret the disposition of active sites, sites of carbohydrate attachment and a few other biochemical attributes on the basis of the proposed model. The author claims that the structures constructed by him are “native” in terms of intersubunit packing and are adequate to account for the observed biochemical properties of neuraminidases. However, the model constructed and its interpretation confronts the vast body of knowledge gained in the course of over forty years of research on protein structure and function.
My comment: This reviewer’s position is also dominated by the sense of tradition. In fact, there is very little “knowledge”, outside of crystallography, about the structure and function of the neuraminidases and it is dealt with in this paper. “Knowledge” or tradition based on other proteins may have little relevance to these proteins. However, it is true that the present results for neuraminidases would undermine confidence in our “knowledge” of other proteins. This is a good thing.
Protein quaternary structures observed in the crystalline state are robust. This is supported by the countless observations of agreement between radius of gyration calculated from the quaternary structure observed in the crystal and the radius of gyration obtained by dynamic light scattering measurements in solution. The agreement in the quaternary structures of homologous proteins isolated from different organisms that crystallize in different crystal forms provides innumerable instances of the robustness of protein quaternary structure. Frequently, a given protein crystallizes in several forms. In these forms also, the quaternary structure is mostly invariant.
My comment: Again the reviewer resorts to discussion of other proteins. The language used is rhetorical and the extent of support for crystal structures is exaggerated. It is doubtful that a radius of gyration measurement could distinguish between alternative compact structures, and since no alternative structures have been proposed until now, I doubt such comparisons have been done. The lack of robustness of protein structure under crystallization conditions is probably the main reason that crystallization is an art that fails more often than it succeeds. (How’s that for rhetoric?)
The author has commented that the electron micrographs might show a square arrangement of protein subunits despite their oligomeric state being D2 in solution due to the forces operative during specimen preparation. However, even negatively stained electron micrographs were sufficiently reliable to allow visualization of icosahedral structures of viruses. The images reconstructed from electron micrographs have been vindicated by later crystal structure determinations. My comment: I have no argument with viruses showing similar structures by EM and X-ray crystallography. Similarly, the topology of protein subunits of ribosomal particles deduced by immune electron microscopy and chemical cross linking, are in good agreement. These studies were carried out decades before the determination of the three-dimensional structure of ribosomal particles by single crystal methods. My comment: I have no argument with ribosomes showing similar structures by EM and cross-linking. Currently, cryo electron microscopy is revealing “native” structures of biomolecular complexes and there is no hard evidence to believe that these would be different in the solution state. My comment: I can’t argue against this belief. I would urge cryo-electron-microscopists to consider tetrahedral structures as well as any others suggested by crystal structures when interpreting their 2-dimensional images.
For these reasons, it is not possible to accept the model proposed by the author. Therefore all the discussion the author has made on the “reasonableness” of the model with D2 instead of C4 symmetry becomes meaningless.
My comment: The reviewer cannot accept the model because of prior beliefs based on other proteins. The reviewer has therefore avoided commenting on the present results and conclusions.
Click to return to http://nativeproteins.blogspot.com