We thank the reviewers for their useful feedback and constructive comments for improvement of our manuscript. We have considered all their suggestions and have addressed these below.
Reviewer 1: Dr. Frank Eisenhaber (Bioinformatics Institute, Singapore)
In an earlier study, the authors concluded that the minor sequence variations distinguishing eEF1A1 and eEF1A2 are confined in structurally limited sections of the two proteins (mostly at what they call back side). Here, they claim that these places harbour most of the PTMs. A few mutations are suggested that might have an effect on differential binding/regulation.
There are several concerns beyond the issue that the finding might be incremental.
First of all, the authors operate at a quite qualitative level. It would be good to consider in number terms how many PTMs are in the sequence variation clusters, how many are just nearby and how many are differentiated between the two proteins. Together with some statistical assessment, this would substantiate the main claim of this work, namely the enrichment of PTMs in the clusters of sequence variation.
Authors’ Response: The reviewer is right and we thank them for making this point. In the revised text we make clear 1) the total number of PTMs present, 2) how many are surface exposed on each face of the protein, and 3) what residues are subject to more than one form of modification. This quantification substantiates our point that PTMs are significantly more likely in locations close to sites of sequence variation, on the “variable face” of the protein, than elsewhere. From the number-counts there is an almost two-fold increase in modifications on the variable face of the protein (19 on the conserved face; 37 on the variable face). The second paragraph of the Findings section now reports and clarifies these results. We have also added in new Additional file 3: “Post-translational modifications (PTMs) in structural proximity to sequence variants between eEF1A1 and eEF1A2” - to enumerate and specify which PTMs are proximal to each variant using a structure-based 5-Angstrom sphere-radius cut-off for each residue. We have also revised Figure 1 (all PTM residues are now labelled and further annotated if a specific residue can be modified in more than one way) that now clearly emphasises the non-random distribution of PTMs.
Second, the PTMs have been measured in varying biological contexts; some might be artifacts or only applicable to specific biological situations. It would be good to have a separate consideration as detailed above for the PTMs that might be considered most reliable and most likely to be biologically significant.
Authors’ Response: We completely agree with the referee, and have stated the limitations in the text. For example we say: “Experimental verification of PTMs discovered by mass spectrometry will be needed using additional complementary techniques. It is also of note that many of the studies have been performed on human tumour or cancer cell line material; it is therefore conceivable that some of these PTMs may be specific to tumours, depending on the relative expression of different modifying enzymes in cancerous and normal cells.” The PhosphoSitePlus database is a manually curated database of good quality, however most site assignments are not linked with corresponding spectra at present. Therefore, in order to place some measure of reliability on the basis of the quality of evidence we have provided an additional layer of annotation in the revised version (also see reply to Reviewer 2 point 4, for discussion of false positive rates). So for example where a specific experiment has confirmed a particular site in the published literature, and/or a number of five or more citations are associated with a specific site from the high-throughput mass spectrometry screens in PhosphoSitePlus, we have now indicated this in the table in a revised Additional file 2. It is interesting that the majority of high-throughput sites have been repeatedly seen in proteomic screens and thus almost certainly represent true modifications (specifically, for Phosphorylation: 22 out of 36; Acetylation: 11/25; Methylation: 5/9; Ubiquitination: 23/25 have five or more citations assigned to them in PhosphoSitePlus). We have also prepared an additional figure to depict where these more reliable PTM sites are located on the surface by means of a comparison with Figure 1. As is readily apparent in this new figure (Additional file 4), when only the more reliable PTMs are considered, there remains an almost two-fold enhancement of post-translationally modified residues on the variable face (conserved face: 16 modified residues; variable face: 31 modified residues).
Third, there are serious issues to which extent PTMs can occur in globular sections. The problem is that protein-modifying enzymes have active site clefts/cavities and the polypeptide stretch of the substrate protein has to get somehow into it (Eisenhaber et al., Current Protein and Peptide Science, 2007, 8, 197). The problem disappears if there are auto-catalysis, non-enzymatic reactions, modifications prior to substrate protein folding, unstructured segments/long, conformationally variable loops or unstable structural parts that readily unfold. It would be necessary to substantiate what mechanism is going on with eEF1A1/2 since this is part of the proof that the PTM seen is biologically significant.
Authors’ Response: Again, the referee is right to make this point. There are however, several PTMs that have been confirmed by site-specific experiments, which are indeed surface-exposed on more structured regions (e.g. S21, K36, K165 on alpha-helices; E374 on beta-strand etc.). The eEF1A1/eEF1A2 models [5] were based upon the co-crystal structure of (~81% identical) yeast eEF1A when bound to eEF1Balpha [28], so it is hard to speculate on what conformational changes occur when eEF1A transitions from free to bound forms. It is indeed possible that protein-modifying enzymes access their substrates in less globular forms of eEF1A. Some evidence for less globular structure is provided in Budkevich et al. [17] who suggested that the eEF1A1 isoform possesses an extended conformation in solution that changes to an essentially more compact conformation upon interaction with tRNAs. In support of such structural rearrangement, Negrutskii et al. [14] discussed tyrosine phosphorylation in the elongation factors from proteomic studies and noted burial of some of these residues. While we can’t experimentally substantiate context-dependent structural rearrangements in the current paper, it is notable that (as yet) there is no evidence that the eEF1A isoforms are bound to eEF1B (and therefore in this structured conformation) when performing their non-canonical roles in other functional pathways. Other protein-interaction dependent PTM may also influence accessibility of specific residues to their modifying enzymes. In sum, at least with respect to this “structured” conformation of the eEF1As bound to eEF1Balpha, it does seem highly unlikely that the observed clustering of PTMs around sites of sequence variability on one face of the protein has occurred by chance. In response, we have updated the main text so that it covers the substance of the above observations.
Quality of written English: Acceptable.
Reviewer 2: Dr. Ramanathan Sowdhamini (Tata Institute of Fundamental Research, India)
This manuscript by Soares and Abbott report the comparison of post-translational modification (PTM) data of eEF1A isoforms. The two isoforms under analysis, eEF1A1 and eEF1A2, exhibit differences in tissue localisation and affinity for GTP, despite sharing 91% identity in amino acid sequence. These functional differences have been addressed by the authors in the context of the observed differences in patterns of PTM of the two proteins. I would recommend publication of this manuscript in Biology Direct, after the following points have been addressed:
1. Page 4 onwards: terms like ‘front’ or ‘back’ of the protein sound too colloquial.
Authors’ Response: We have revised all text and figure calls in the manuscript to refer to the previous ‘front’ side of the protein as the ‘conserved face’ and the ‘back’ side as the ‘variable face’ to highlight the eEF1A isoform specific location of the sequence-variation in context of this study.
2. Even in cases where PTM sites are conserved between eEF1A1 and eEF1A2, the molecular players responsible for PTM could be dramatically different. The authors need to consider this point.
Authors’ Response: We thank the reviewer for mentioning this point. Indeed, it is very likely that the protein-modifying enzymes involved could differ between isoforms, even for sites that are identical in sequence. This may be influenced by minor differences in the neighbouring structural landscape of surface charge and hydrophobicity proximal to the modified amino acid residue. We have introduced this point in the penultimate paragraph of the Findings discussion.
3. It will be interesting to compare such PTM motifs of homologues - both closely related and distantly related - to provide a dimension of the role of PTM in the context of evolutionary dynamics.
Authors’ Response: The referee is right. The issue with extending the evolutionary analysis is one of available data. Given the high similarity of the two isoforms within a species, it is important to have some form of additional evidence in order to assign sequences as unequivocally orthologous to either eEF1A1 or eEF1A2. The best evidence comes from expression analysis; if within a species there are two apparent eEF1A sequences and one is ubiquitously expressed but the other expressed only in brain and muscle, it is possible to have more confidence in their identity as eEF1A1 and eEF1A2 orthologues, respectively. Unfortunately such evidence is almost entirely lacking for many species. Our Additional file 1 displayed a multiple sequence alignment of various eEF1A1 (10 species) and eEF1A2 (9 species) orthologues from vertebrates. The modified residue is strictly conserved in almost every case. Taking into account the reviewer’s suggestion we have now added into the alignment the more divergent sequences of zebra fish and also of non-vertebrate, but well characterised, yeast eEF1A. This comparison illustrates clear alterations in post-translational modifications between yeast compared with other eukaryotic vertebrates – 18 sites in total. These are indicated in the revised Additional file 1. The presence of the two isoforms in vertebrates creates the potential for greater complexity than is seen in yeast: Saccharomyces cerevisiae has two genes encoding eEF1A [29] and Schizosaccharomyces pombe has three [30], but the encoded proteins are identical within a given species. We thank the reviewer for this suggestion which clearly points to gain of PTMs among vertebrates across evolution. We have summarised this point in the main text along with the revised Additional file 1 and associated legend.
4. Page 6 - It will be important to identify PTMs recorded for other proteins to note if there maybe false positives.
Authors’ Response: Unless comprehensive complementary experimental validation of high-throughput mass spectrometry studies are undertaken for the two eEF1A isoforms and other proteins it will be difficult to estimate false positive rates. Comparing examples of other proteins as a means of assessing the likelihood of false positives, depends on how that particular protein has been researched. For example, for the comparatively better-studied tumour suppressor protein, p53, a wide-range of PTMs have been confirmed by site-specific methods, which when compared to the mass spectrometry (MS) screens in PhosphoSitePlus (http://www.phosphosite.org/proteinAction.do?id=465&showAllSites=true; accessed 28th October, 2013) show that 95 modified positions exist in total across species, of which 80 were confirmed by site-specific methods and 53 were seen by MS; of the 53 observed by MS, only 15 have yet to be confirmed by a site-specific method. This indicates, at least in this example, a very low false-positive rate; bearing in mind that the rest of the MS sites are yet to be verified independently. The p53 polypeptide is smaller than eEF1As (p53, 393aa cf. 462/463aa for eEF1A1/eEF1A2) but has a greater number of known and experimentally verified PTMs. One of us recently published a study on two highly similar paralogous proteins - NDE1 and NDEL1 - that possess similar structures [31] but are differentially regulated post-translationally [32]. In that case, nine out of ten sites verified by non-mass spectrometry experiments in the primary literature for NDE1 and NDEL1 were seen in the MS assignments in the PhosphoSitePlus database. This suggests a high true positive rate; additionally, >50% of sites assigned only using the MS/high-throughput proteomics criteria had five or more citations corresponding to each site in the database, indicating a high potential for other sites to be true positives too. There were also instances where experimentally confirmed sites had less than five associated MS citations.
As mentioned in the main text and in our response to Reviewer 1, because a lot of the data is based upon publicly available high-throughput mass spectrometry data, the assignments are probabilistic by nature and need to be further confirmed experimentally by complementary techniques such as site-directed mutagenesis, phospho-specific antibodies and dominant-negative constructs. Furthermore, future work should aim to ascertain the specific kinases and other modifying molecular players. Nonetheless, our structure-based mapping of PTMs in context of sequence variation of the two eEF1A isoforms and their putative binding sites form the baseline from which future studies will continue to inform on the regulation of these proteins.
Quality of written English: Acceptable.