Reviewer 2's Review (Julian Tanner)

“Major Claims of Paper
In this manuscript, the authors investigate protein-protein interactions of the SARS Coronavirus proteome by yeast two-hybrid analysis, and identify 65 interactions including 6 self interactions. They go on to confirm a number of these interactions by Co-IP. In particular, they demonstrate nsp2, nsp8 and ORF9b have a significantly high number of interactions. Intriguingly, nsp8 is essential whilst nsp2 and ORF9b are not. Furthermore, the authors perform an analysis of the network, and compare to their recent work on the Kaposi sarcoma-associated herpesvirus (KHSV), and attempt to draw conclusions regarding the general connectivity of the SARS proteomic network. Overall, I believe the paper has a high technical standard and has a great deal of interesting and useful data for the community. However, the manuscript perhaps oversteps its claims in the last part of the paper (please see below).

Are the claims properly placed in context of previous literature?
The SARS genome has been well-characterized so that the nsp’s are now well-defined and many functions are known. It is well-described in the introduction how deletion of certain ORFs result in no apparent changes in viral viability, however the introduction would be stronger if known protein-protein interactions of the SARS proteome were discussed in a little more depth to allow the reader to more clearly see the gap in our knowledge that this study successfully fills.

Does the experimental data support the claims?
The authors start their results with the two-hybrid screen results and present their results elegantly in Figure 1. Figure 1, effectively presents the two-hybrid data, but a little confusingly also shows Co-IP in yeast without showing the raw data – my first recommendation for minor revision is thus:

1. The authors should show the Co-IP data in yeast (perhaps in the supplementary data) to support the Co-IP claims made in Figure 1.

In figure 2, CoIP is investigated for nsp2 and nsp8 in 293 cells. There is a significant difference in this CoIP experiment (in mammalian cells), and the CoIP experiments in Figure 1 (in yeast cells) which should be discussed in more detail – this difference is also not really mentioned in discussion.

Figure 2 would be far clearer if split into two separate figures 2a (the left panel), and figure 2b (the right panel). Is nsp6 not expressed – no anti c-myc band? There appears to be further information on these Western blots that is not discussed. For example, a multimeric nsp2 band appears to be present on the anti-HA WB that is not present in the other lanes, further supporting the nsp2-nsp2 interaction. Why are the nsp8 anti-HA WBs so much cleaner than the nsp2anti-HA WBs on left-hand side? Species with highermolecular weight species are discussed in results, but perhaps could be due to posttranslational modifications?

Figure 3 needs significant changes:

2. The raw sequence data does not been to shown in figure 3 (this could either be shifted to supplementary information, or be deleted entirely). The graph of viral growth needs to be much better presented though with proper error bars of repeat experiments and better presentation of data. Perhaps presenting as a bar chart for each time point with errors may be more suitable rather than 5 individual points on a graph with an obscure way of fitting lines to each set.

This is important as perhaps there is a subtle change in growth rate after ORF9b deletion that is only detectable with an accurate experiment. The interactions of ORF9b are described in Supplementary Figure 1, however despite the Western blots being relatively messy, there are no controls for the background bands, unlike previously shown in Figure 2, therefore:

3. There need to be suitable controls on Supplementary Figure 1, similarly to those presented in Figure 2.

The authors then go on to a bioinformatic analysis. This is particularly interesting when compared to KSHV, and figure 4a is a particularly interesting figure demonstrating the viral-viral interactions in a node diagram visually. However, at this point I feel the authors overstep what they realistically can claim with regards to a separate intraviral network. The problem is that the authors directly compare interactions identified from their Y2H screen, to those from a “literature screen” of virus-host interactions, which is unlikely to be exhaustive. Indeed the authors note this point themselves in the penultimate paragraph of the discussion, but nevertheless based on this approach attempt to claim that the SARS network works quite separately from the host network, a result opposite to that which they observed with the KSHV network. It appears to be more likely that this is just due to there being no comprehensive analysis of protein-host interactions, and without that being the case it is difficult to stretch their conclusions this far. I therefore have issues with Figure 4b and Figure 4c as the “literature screen” could well be quite misrepresentative of the real situation. Thus,

4. The authors need to either really justify their comparison of the Y2H screen to the “literature screen”, or need to remove figures 4b and 4c and tone down their speculation regarding SARS having a network separate to the host network, a result opposite from their KSHV conclusion.

The immunofluorescent data is good and useful, but it would be interesting to better put this in context in terms of the interactions identified – is it generally true that interacting partners localize together, and that those that don’t do not localize together? What are the exceptions to this general trend and what might be the explanations?

Furthermore, some important papers are omitted in the discussion. For example, Imbert et al.’s paper showing that nsp9 is a second RDRP should be discussed in context of this paper, and hence the RDRP nsp12 –nsp9 interaction could carry functional significance:

Imbert, I., Guillemot, J.C., Bourhis, J.M., Bussetta, C., Coutard, B., Egloff, M.P.,Ferron, F., Gorbalenya, A.E. & Canard, B. A second, non-canonical RNA-dependent RNA polymerase in SARS coronavirus. Embo J 25, 4933-4942 (2006).

Overall, I feel the paper will be a very good contribution to PloS One and will be
interesting to a fairly broad audience after these points above have been addressed.”

n.b. These are the general comments made by the reviewer when reviewing the originally submitted version of this paper. The manuscript was revised before publication. Specific minor points addressed during revision of the paper are not shown.