Yes, the proper comparison would be an additional transgenic strain expressing a suitably 'inert' protein to thus differentiate between the metabolic demands of

Yes, the proper comparison would be an additional transgenic strain expressing a suitably 'inert' protein to thus differentiate between the metabolic demands of high levels of protein expression vs. some particular effect of the heterologous EPSPS. Ideally, you'd use the same protein with a deletion of an active domain. It is common use multiple deletion lines to identify the causative domain; this sort of mechanistic work is essential for understanding what is actually going on. Moreover, the use of 'near-isogenic' lines is very troubling, as they are rarely that similar. You can do e.g. RNAseq on 'identical' inbreds from different labs and get substantial differences. Seed is passed around so much, mis-labled in the field or out, etc. They certainly do not have an identical inbreed from something that was developed decades ago, not even close. Calling them near-isogenic is, well, let's say 'not that honest'. Very important to note that they have no way to control for this difference in genetic background. Also noteworthy; Michael Antoniou has co-authored papers with e.g. Seralini, notorious for a widely-discredited paper on RoundUp https://en.wikipedia.org/wiki/S%C3%A9ralini_affair https://en.wikipedia.org/wiki/S%C3%A9ralini_affair That certainly does not discredit this work, but it is concerning when an investigator agrees to co-author a manuscript of such poor quality. Finally, having done a lot of work on '-omics' in plants, I've yet to see an experiment that doesn't turn up some differences. Their list of GO terms is not remarkable, and changes of '30 fold' are not nearly as dramatic as they appear. This is well within the range of what I'd expect to see between different similar inbreds, which these lines certainly are.