Write a comment

The manuscript “Full-length direct RNA sequencing uncovers stress-granule dependent RNA decay upon cellular stress" by Dar and colleagues uses end-to-end RNA sequencing by Nanopore to examine the effect of oxidative stress on RNA length. They observe transcript shortening under arsenite stress that could be rescued through XRN1 or G3BP1/2 knockdown. The data presented in this paper has interesting implications for the field of mRNA decay, but the authors need to clarify their methodology and interpretations at times. Some experimental issues impact the robustness of their conclusions.

Major comments:

(1)    As an overall comment of the manuscript, the way in which the data is presented can sometimes be confusing to the reader. For example, I found the bar graphs in 2b and 3e to be unclear. One way to make the data clearer would be to pull a couple of examples of unchanged transcripts and significantly changed transcripts (as in supplemental figure 2) and use those in a paired dot plot or scatter plot. Those same genes can also be highlighted in all the plots so the reader can easily understand how different plots relate to each other.

(2)    The authors can end-to-end sequence transcripts by ligating an adapter to 5’ monophosphates. As there was no mention of an uncapping step in the protocol, this indicates that the end-to-end sequencing occurs only in transcripts that are already uncapped and in the process of decay. Thus, as written, the authors are only comparing the lengths of RNA decay intermediates during their analyses. That detail is not made clear in the text, but it’s an important fact to consider when interpreting the results. Furthermore, it raises some questions about the quality and curation of the library:  What’s the efficiency of the ligation step? What percentage of the entire library becomes ligated? Does the population of adaptor-ligated transcripts increase in the XRN1 knockdown due to an accumulation of uncapped transcripts? What is the expected rate of aborted sequencing in Nanopore? Is it possible to identify what percentage of the library is capped mRNAs? Also, what percent of the library is sequenced? Is there a possible bias towards shorter transcripts being sequenced? There is still a lot of quality information to be obtained with the adaptor ligated library, but it’s important to contextualize it properly.

(3)    The authors do not observe polyA tail shortening prior to 5’ decay. However, the protocol includes a polyA enrichment step. This step would bias the library to the presence of tails and possibly to longer tails. Given this bias, I’m not sure it’s possible to draw robust conclusions about the length of polyA tails without a CNOT knockdown.

(4)    This manuscript has an interesting observation of G3BP1/2 knockdown rescues transcript length. The conclusion that the effect is “dependent on stress granule formation” is a bit premature. To differentiate a G3BP1/2 specific result from a stress granule specific result I recommend an experiment in which stress granules can form in the absence of G3BP, such as with osmotic stress through sorbitol or using a synthetic condenser that is not G3BP (see Yang et al. 2020, Taylor Lab)

Minor comments:

(1)    According to figure 3e, XRN1 knockdown under arsenite stress results in lengthened transcripts. A rescue of the length in knockdown would be logical, but the data as it is presented indicate a dramatic increase in length. I’d like the authors to discuss possible mechanisms to explain this extremely interesting result. In addition, the effects of XRN1 knockdown in an unstressed condition would be important for comparison.

(2)    How is “differential transcript length” defined?

(3)    During stress, only an approximate 10% of bulk mRNA molecules accumulate in stress granules and only 185 genes have more than half their molecules localizing to stress granules. I think this manuscript would be enriched by a discussion on if those results reflect those numbers and if not, through which mechanisms stress granules would have such bulk effects while only directly localizing a small portion of transcripts.

Competing interests

The authors declare that they have no competing interests.