The results reported in this paper may be biased by the reuse of the same control samples for comparing multiple conditions (e.g. RES and SUS vs. CON in Fig. 1e,f). This issue--identified by Karl Pearson in 1897!--is explained in a slightly different context in a recent preprint by Nicol et al.

Comparisons of differential expression between groups should be performed by using an appropriate statistical model with an interaction term, as I discussed here.