Review for "The β3-integrin endothelial adhesome regulates microtubule dependent cell migration"

Completed on 20 Jun 2017 by Jonathan Humphries . Sourced from http://biorxiv.org/content/early/2017/06/03/145839.

Login to endorse this review.


Comments to author

Author response is in blue.

Dear Authors,
Thanks for posting this very interesting collection of MS datasets and experimental observations that arose from your data. I have a number of suggestions and comments that I hope you feel will help with the analysis and presentation of the data. My comments mainly focus on the proteomics side of things.
1. The title is misleading as far as I can tell you have not defined the beta 3 integrin endothelial adhesome by MS-based proteomics. From the supplementary data files it is apparent that the adhesion complexes isolated comprise other integrins including a substantial amount of integrin a5b1. I understand that you provide additional evidence for a role of b3 integrin in MT stability but this equally could be due to the shift in balance between the proteins recruited to a5b1 and aVb3.
2. I think you are underselling the interest in what the FN-induced (a5b1 and aVb3) adhesome comprises in endothelial cells. You perform a nice subtractive proteomic approach to enable you to define the FN-enriched proteins (compared to PLL) but you don't tell us what these adhesome components are, or provide any comparison with other proteomic datasets or the Geiger literature-curated adhesome. This sort of analysis would help the field to understand the context-based composition of adhesion complexes (similarities and differences). I was left with the question 'what is the endothelial adhesome?'.
3. I was concerned by the approach you have taken to you proteomic analysis of the isolated complexes. Whilst on the face of it you have used good informatic tools (maxquant / perseus) I note that you performed the MS analysis from 3 pooled adhesion complex isolations. This will only permit a measure of the technical variability in the LC-MS/MS and not give you any idea of the variability in the biology i.e between adhesion complex isolations. Maybe there is a good statistical justification for this approach but it needs to be provided.
4. Also your MS dataset lists of proteins contain proteins from identifications with only 1 unique peptide. In my experience these '1-hit wonders' are a major source of variation in the quantitative values from MS outputs. Again please justify the inclusion of such identifications.
5. Please provide more details in the methods of the isolation of adhesion complexes and MS set up.
6. You have used one EC-derived cell type for all the experiments. Are your findings observed in other EC's such as non-transformed primary ECs?
7. In figure 1 it would be nice to see more IF staining of some cannonical adhesion proteins such as integrins (aV, a5, b3, b1) or paxillin / vinculin. You also note in the methods that you check the quality of the isolations by western blotting before MS analysis - could you provide any of this data to supplement the silver stains? It is reassuring to show blots for components you expect to see and don't expect to see in adhesion complexes before MS analysis.
8. Figure 2e why choose to blot for hspa1a? Are you saying this is an adhesion component? If the idea was to highlight equal recruitment to adhesion complexes why not blot for talin or vinculin?

I hope you find these comments useful. Thanks for sharing your data.
Regards,
Jon Humphries
WTCCMR, The University of Manchester, UK



Hi Jon, I'm the first author in this paper.

Thanks very much for your input, it will be very helpful. I have already been addressing similar suggestions in my thesis at the moment ahead of my viva but not all have made it into the paper so far.

Whilst we take the time to put together an updated version based on yours and other's comments, I thought I would quickly explain how we are going to incorporate your suggestions in the meantime:

1. I agree there is substantial a5b1 in our adhesomes due to the adhesion to fibronectin. We wanted to allow cells to create genuine adhesions before carrying out our enrichment process, rather than using an avb3 specific ligand to pull out only avb3 specific adhesion proteins. This also gave us an insight as to what changes would occur in the absence of beta 3 in endothelial cells, which lead to the investigations on MT stability. One thing I notice is that we forgot to indicate statistical significance in the supplementary files. We will specifically allude to this in the next version, as the reader can then identify all of the proteins on the left hand side of figure 2D. These proteins are the ones lost from endothelial adhesions upon deletion of beta 3 and so represent the beta 3 adhesome. Our intention to highlight this got lost in the editing of the original manuscript.

2. Thank you for this suggestion. In the supplementary file 1 we labelled all of the proteins with the clusters generated from the subtractive analysis of the FN adhesome seen in figure 1C to the reader to see these adhesome proteins. We will add an additional category to supplementary file 1 indicating which proteins are in the FN adhesome – currently labelled as clusters D, E and F.

As for comparison with other adhesomes, we wanted to carry out such analysis but were unsure as to which datasets to compare to. Our enrichment techniques are based on Schiller et al 2011 (doi 10.1038/embor.2011.5) but more and more adhesomes are being published using other methods leaving us unsure. The literature-curated adhesome is a good suggestion, thank you.

3. Our main interest was to investigate the difference in adhesome compensation on FN vs PLL / with or without VEGF / with or without beta 3 manipulation rather than assess variability between isolations, given that this was an established method by Schiller et al for fibroblasts.

Variability between isolations was a very big concern for us and was actually the single biggest issue in this investigation that consumed the most time and effort by far but, as is often the case, this kind of troubleshooting doesn’t come across in print. We decided to reduce variability as much as possible biochemically . We knew from previously published adhesomes what kind of proteins to expect and not what to expect in adhesion enrichments, hence we performed countless enrichments and analysed by western blotting to perfect the technique and reduce variability. Once we had honed our technique we were able to use silver staining to quality control any samples before pooling and sending for mass spectrometry. Pooling we hoped would take into account the biological variability and allow us to focus purely on the conditions mentioned earlier. We can certainly provide this and further justification in the final version of the manuscript.

4. We left these proteins in our supplementary data tables to give a more complete picture of our MaxQuant outputs, to allow others to do their own statistical analyses if desired on our datasets. We included the number of unique peptides tables so readers can make their own judgements about these proteins and exclude them if desired.
This is also why we haven’t yet approached a mass spectrometry repository to deposit our data. These are the first attempts of our lab to produce and share such information and we were waiting for feedback like yours first.

Important mechanistic proteins in the paper such as Anxa2 and Rcc2 were not 1-hit wonders and tubulin was represented by multiple monomers. We also validated the presence of tubulin in the beta 3 depleted adhesome by western blot before committing ourselves to a larger investigation.

5. We had already carried out and published this technique for endothelial cells in Ellison et al 2015 (DOI 10.1242/dmm.019927) in order to place Nrp1 in the adhesome of endothelial cells but did not give any more details at the time. We referenced this publication but can go into more details here instead as it is a much bigger part of this investigation than it was in Ellison et al.

6. For label-free quantification we were forced to use ECs that can be sufficiently expanded to generate enough adhesome protein (in some cases we needed close to a billion cells per experiment). We needed too many cells to justify the number of animals we would need to do these types of studies from primary cells. We take the point though and could certainly make more of a point in the discussion about the potential limitations of the study. Perhaps illustrating the similarities to other adhesomes (as you suggest above) also partially addresses this issue.

7. Yes we can provide the western blots for this data as there were a lot of them! We didn’t think they would be that interesting but if you think they strengthen the argument we can. Generating the IF staining you suggested wouldn’t be too difficult either.

8. We used Hspa1a (aka Hsp70) as a convenient loading control, which according to published adhesomes was present in adhesion complexes. We didn’t mention that we had used the silver staining quality control method on these samples as well, as this left just enough material for a western blot for proteins with good antibodies such as tubulin and Hspa1a. We will make sure this is clear in the paper.

Talin or Vinculin would have been a good idea, thanks for the suggestion. I wish we had thought of that at the time! I think because of the silver staining quality control it would have shown the same thing as Hspa1a but I can always re-probe the original blots or generate new samples.

Thanks again for your feedback.

Samuel Atkinson
Biomedical Research Centre
School of Biological Sciences
University of East Anglia
Norwich