PREreview of Full-length direct RNA sequencing uncovers stress-granule dependent RNA decay upon cellular stress

In this manuscript, the authors employed an end-to-end direct RNA sequencing technique called TERA-seq to study RNA decay at the single nucleotide resolution during cellular stress. Their sequencing analyses showed that a subset of RNAs have shortened 5’ ends upon cellular stress conditions that is independent of their 3’ end de-adenylation. They suggested that this 5’ end decay is dependent on XRN1 and G3BP1/2 by showing that 5’ end shortening can be rescued after XRN1 knockdown and ΔΔG3BP1/2, respectively. By comparing with publicly available datasets describing the stress granule transcriptome, they further proposed that stress granule formation is required for 5’ end decay since these shortened RNAs mostly reside in stress granules and absence of stress granules in ΔΔG3BP1/2 cells did not lead to rapid decay of these mRNAs.

Overall, findings from this manuscript are of interest to the field but with the current data in the manuscript, it is difficult to interpret the robustness of the conclusions made by the authors. Some conclusions might be an over-interpretation of the data and would require further experimental validation.

Major points:

1.     A brief look at the genes in Sup. Table 4 suggested that majority of the genes have a length of ~2 kb. For known long RNAs such as MALAT1 (8 kb) and NORAD (5.3 kb), their mean length is also around 1-2 kb, even in the unstressed condition. This is also evident in Supple. Fig. 2a & 2b where the significantly shortened transcripts are longer than non-significantly shortened transcripts. This suggests that either the sequencing technique used here might be biased for a certain transcript length or the transcripts being sequenced are mostly decay intermediates. The authors should clarify this discrepancy to strengthen their conclusions. If the authors are solely sequencing RNA decay intermediates, the work is still of interest but this needs to be made clear and the interpretations clarified.

2.     The authors did not use any orthogonal methods to validate their sequencing results. They could validate some of their top candidates by performing smFISH on both the 5’ end and coding region of the transcript in unstressed and NaAs treatment. Alternatively, they could perform Northern blots to show that transcripts are shortened, and this is specifically due to 5’ end decay.

3.     It is not appropriate to conclude that their data showing an association between G3BP1/2 and RNA decay demonstrates that 5’ end RNA decay is dependent on stress granules (also indicated by the title of the manuscript). Although ΔΔG3BP1/2 cells are deficient in stress granule formation, G3BP1/2 and stress granule function could be independent of each other since stress granule formation can be rescued in ΔΔG3BP1/2 cells, and G3BP proteins can have additional functions. To delineate between G3BP1/2 and stress granule dependence, the authors could test 5’ end shortening in ΔΔG3BP1/2 cells treated with sorbitol which forms stress granules independent of G3BP1/2. Alternatively, they could transfect ΔΔG3BP1/2 cells with FKBP^F36M,2ZnF-G3BP1 constructs (Yang et al., 2020, Cell) that also produces stress granules independent of G3BP1/2 function.

4.     It is difficult to interpret Figure 3e. From Figure 3b & 3c, it suggests that siXRN1 rescued 5’ end shortening during stress compared to unstressed. However, Figure 3e shows that for the significantly shortened genes, siXRN1 not only rescued 5’ end shortening but led to even longer transcripts. If this is the conclusion, the authors should address this observation in the manuscript. Otherwise, it would be better to change the data presentation in Figure 3e to make a clearer and more straightforward point.  

Minor points:

1.     Missing “.” in figure citation for line “stress-induced significantly shortened RNAs (Fig 3e) …”

2.     Wrong citation of sup. fig. in line “Similarly, no difference was observed upon XRN1 silencing (Sup. Fig. 1a) …”

3.     Difficult to see the dashed lines under the solid lines in Fig 4a. Could consider making the solid lines a lighter shade to show that both lines overlapping.  

Competing interests

The authors declare that they have no competing interests.