Intronic polyadenylation isoforms in the 5’ part of genes constitute a source of microproteins and are involved in cell response to cisplatin

  1. Alexandre Devaux
  2. Iris Tanaka
  3. Mandy Cadix
  4. Amélie Heneman-Masurel
  5. Sophie Michallet
  6. Quentin Fouilleul
  7. Alina Chakraborty
  8. Céline M. Labbe
  9. Nicolas Fontrodona
  10. Jean-Baptiste Claude
  11. Marc Deloger
  12. Pierre Gestraud
  13. Ludovic Tessier
  14. Hussein Mortada
  15. Sonia Lameiras
  16. Virginie Raynal
  17. Sylvain Baulande
  18. Nicolas Servant
  19. Didier Auboeuf
  20. Béatrice Eymin
  21. Stéphan Vagner
  22. Martin Dutertre

Reviewed by

Read the full article

Listed in

Log in to save this article

Abstract

Transcript isoforms generated by intronic polyadenylation (IPA) are widely regulated in various biological processes and often encode protein isoforms. Microproteins are small proteins translated from small open reading frames (sORFs) in noncoding RNAs and mRNAs, but their production by IPA isoforms is unknown. Using 3’-seq and long-read RNA-seq analyses in lung cancer cells, we show that cisplatin, a DNA-crosslinking anticancer agent, upregulates IPA isoforms relative to full-length mRNAs in long genes. A subset of cisplatin-regulated IPA isoforms are poorly associated with heavy polysomes and terminate upstream of the annotated translation initiation codon of genes. Such IPA isoforms in the PHF20 and PRKAR1B genes are associated with light polysomes, contain Ribo-Seq-supported sORFs in an alternative last exon within the annotated 5’UTR part of genes, and are translated into microproteins. For PRKAR1B , the microprotein was detected by Western blot and immunofluorescence after transfection of a tagged isoform; and siRNA depletion of the endogenous IPA isoform, CRISPR deletion of the IPA site, or CRISPR mutation of the sORF initiation codon led to increased cell survival to cisplatin. Based on Ribo-Seq and mass-spectrometry data sets, we identified 156 genes producing both a canonical protein-coding mRNA and a microprotein-coding 5’UTR-located IPA isoform (coined miP-5’UTR-IPA isoform) regulated by cisplatin. Finally, the regulation of (miP-5’UTR-)IPA versus full-length isoforms by cisplatin involved an inhibition of transcription processivity in a FANCD2 and senataxin-dependent manner. Altogether, these findings reveal the novel paradigm of miP-5’UTR-IPA genes and their role in cancer cell response to a genotoxic agent.

HIGHLIGHTS

  • - Cisplatin increases intronic-polyadenylation versus full-length transcript isoforms in long genes through a FANCD2 and senataxin-dependent decrease of transcription processivity

  • - A subset of cisplatin-regulated intronic-polyadenylation isoforms terminate in the annotated 5’UTR part of genes and encode microproteins, thus we coined them miP-5’UTR-IPA isoforms

  • - The miP-5’UTR-IPA isoform of PRKAR1B impacts cisplatin sensitivity and its effect is mediated by its small ORF

  • - We identify 156 genes producing both a canonical protein-coding mRNA and a microprotein-coding miP-5’UTR-IPA transcript

    • Article activity feed

    5. Review Commons

      Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

      Learn more at Review Commons

      Reply to the reviewers

      1. General Statements

      Regarding significance, we would like to highlight that the main finding and breakthrough of our manuscript is the discovery that intronic polyadenylation (IPA) isoforms are a source of microproteins (indeed, IPA was not known to induce sORF-encoded microproteins). We make the proof of principle of this concept (called miP-5’UTR-IPA) and of its functional relevance for one gene (PRKAR1B).

      A second finding of this study is that IPA (including miP-5’UTR-IPA) isoforms are widely upregulated in cell response to cisplatin, and therefore we show the functional relevance of miP-5’UTR-IPA isoforms in this biological context.

      Regarding the generality of the miP-5’UTR-IPA concept, we provide evidence that many genes generate miP-5’UTR-IPA isoforms, by crossing our 3’-seq data with available Ribo-Seq and mass spectrometry datasets, which were generated without cisplatin treatment. Also, the miP-5’UTR-IPA isoforms of PHF20 and PRKAR1B are detected both in the presence and absence of cisplatin. Thus, the novel concept of microprotein-coding IPA isoforms opens wide perspectives, way beyond cisplatin response.

      2. Description of the planned revisions

      REVIEWER #1

      Evidence, reproducibility and clarity

      Microporteins originating from coding and non-coding transcript are increasingly understood to control various cellular processes. In the present study, the authors investigated whether intronic polyadenylation (IPA) contributes to the formation of transcript isoforms encoding microproteins. Using genotoxic stress by cisplatin as a model in cell cultures, the authors detect abundant IPA. IPA in a subset of such transcripts leads to short 5'UTR transcript isoforms that are poorly associated with heavy polysomes and encode microproteins. For PRKAR1B, they demonstrate the expression of a corresponding microprotein and a function in modulating the cisplatin response. Based on depletion experiments of FANCD2 and STX1, the authors propose that impaired transcription processivity after cisplatin is one mechanism leading to IPA and microprotein production.

      While this is an interesting manuscript, I felt the support for the claimed generalization falls a bit short.

      Our response: The generality of the miP-5’UTR-IPA concept is supported by the large-scale analysis that we presented (Fig. 6): indeed, by crossing our 3’-seq data with Ribo-Seq and MS data (both of which originate from multiple cell types and tissues), we identified 156 genes with cisplatin-regulated miP-5’UTR-IPA isoforms. To strengthen this part and highlight the generality of the miP-5’UTR-IPA concept, we will provide the cell type/ tissue distribution of our set of 156 miP-5’UTR-IPA isoforms, by exploiting available 3’-seq datasets from various cells/tissues. (Please also see major point 1 below.)

      Major:

        • If I see it correctly, the authors mainly refer to existing riboSeq data and evidence from mass spectrometry/proteomics to infer the generality of the mechanism (beyond PRKAR1B). It is important to back this up with further experiments and validate this for the set-up used in this manuscript. This concerns the existence of the microproteins but also the downstream functional impact.* Our response: In our study, we make the proof of principle of miP-5’UTR-IPA (that is, a microprotein-encoding IPA isoform) for the PRKAR1B gene and its sORF#2 (microprotein detection by WB and IF, functional evidence by siRNA and CRISPR of IPA site and sORF initiation codon). If I understand well (also based on minor point 3 below), this reviewer is requesting further evidence of microprotein existence (in addition to Ribo-Seq and mass spectrometry [MS] data) and function, for a second IPA-derived sORF that we study in this manuscript (either PHF20 sORF or PRKAR1B sORF#1). To the best of our knowledge, a proof of principle for a new concept is usually done on a single gene. Nevertheless, for the miP-5’UTR-IPA isoform of PHF20, we already provided evidence for its function by using siRNAs (Fig. 3A-C) and for its translation by polysome profiling (Fig. 4C) in addition to Ribo-Seq and MS evidence (Fig. 4A). The fact that for PHF20 we did not detect the transfected Flag-tagged microprotein in HEK cells could be due to several reasons (as discussed on page 16); __we will __try this approach again with different biological conditions (cell lines, stress) or construct designs (as the sORF context may be important).
      • Also, I wonder is this limited to cisplatin-induced genotoxic stress and the specific cell line used or is this a more global mechanism?*

      Our response: We provided evidence of IPA isoform regulation by cisplatin in two lung cancer cell lines (A549 and H358; Fig. 1A-B) but we agree that our analyses of miP-5’UTR-IPA were mainly done in A549 cells. We will: (i) clarify that we detected the miP-5’UTR-IPA isoforms of PRKAR1B and PHF20 in A549 cells (total cytosol and light polysomes) both in the presence and absence of cisplatin (Fig. 2D, 3A and S3D); (ii) add RT-qPCR validation of their cisplatin regulation in H358 cells; (iii) try to detect the PRKAR1B-encoded microprotein in a second cell line (Fig. 4); (iv) test the impact of PRKAR1B and PHF20 miP-5’UTR-IPA isoforms on cell survival in a second cell line and with a second genotoxic agent; (v) clarify in Fig. 6 that miP-5’UTR-IPA isoforms are regulated by cisplatin in both A549 and H358 cells (our 3’-seq data) and that the Ribo-Seq and MS datasets supporting their translation originate from multiple cell types and tissues without cisplatin treatment; and (vi) provide the cell type/ tissue distribution of our set of 156 miP-5’UTR-IPA isoforms (by exploiting available 3’-seq datasets from various cells/tissues).

      Minor:

        • While the rest of the paper reads well, the abstract could be improved/simplified to increase accessibility* Our response: We will improve the abstract.

      Page 11: Pertaining to Figure 3 and the functional impact: The authors analyze the IPA effect by probing cell viability and cell survival. It would be important to define the effects in further detail, as the mere regulation of cell cycle and/or apoptosis could also result in such outcome (which is then not necessarily a direct cisplatin response). Does this also impact the response to other genotoxic stress (also pertains to the effects studied and shown in Fig. 5)?

      Our response: Because cisplatin effects on cell growth are usually mediated by effects on cell cycle and cell death, we will determine which aspect is impacted by PRKAR1B and PHF20 miP-5’UTR-IPA isoforms, by carrying out FACS analysis of PI/BrdU and Annexin V (both in the presence and absence of cisplatin). As mentioned in major point 2, we will also test the impact of these isoforms on cell survival to a second genotoxic agent.

      Page 12 concerning the microprotein expression: the authors refer to data from other resources to claim that the microproteins are expressed, however they fail to demonstrate this for their setup (at least for 2 out of three they study here). I think this is a weak point as it does not directly support the general claim.

      Our response: Please see major point 1 above.

      Also, I did not understand what the authors intended to demonstrate with the immunoflourescence (Fig. 4E). What should a defined nuclear expression imply versus the diffuse staining throughout after cisplatin? How does this relate to the functional effects?

      Our response: We included in Fig. 4E the observation that the subcellular localization of the PRKAR1B-encoded micropotein is altered in response to cisplatin, because this supports the notion that this micropotein plays a role in cell response to cisplatin. We can remove this data if requested.

      Page 13/Fig. 5E: the different clones of the mATG show very high variability. To my understanding it is difficult to draw a clear conclusion from this heterogeneity.

      Our response: The statistical analysis shows a significant difference between the mATG and Control groups (p Page 15 on the mechanism: SETX has been demonstrated to control poly(A) site choice (PMID: 21700224, 32976578). However the quantitative role of SETX in poly(A) site choice regulation (compared to other regulators) seems to be rather marginal and not strictly unidirectional, i.e after SETX depletion also longer transcript isoforms can be detected (PMID: 32976578). How does this relate to the proposed mechanism of SETX-dependent processivity? Interestingly, from PMID: 32976578 it also appears that PRKAR1A has a 5'UTR poly(A) site that is regulated in a SETX-dependent manner.

      Our response: We will add in the discussion statements that (i) the role of SETX in cisplatin regulation of IPA:LE isoform ratio and processivity might be different from its role in APA regulation in the absence of genotoxic treatment (citing PMID 32976578; keeping in mind that we did not compare them side by side on a genome-wide scale) and (ii) PRKAR1A seems to have a 5'UTR poly(A) site regulated by SETX in TREND-DB (PMID 32976578).

      • Page 16, discussion first paragraph. While refs 1-4 are nice reviews that could be quoted here a study that appeared later represents the most comprehensive analysis to date covering the different facets from transcription to RNA processing and the resulting impact on poly(A) site choice (PMID: 30552333).*

      Our response: We will cite PMID 30552333 and 32976578 as resources of APA regulation by various regulators of gene expression (keeping in mind, however, that for most factors these studies do not exclude indirect effects).

      Significance

      This could be a very significant report, provided the generality of the claims and mechanistic insigths are further strengthend.

      Overall it targets a rather specialized readership. This could be improved by simplifing the abstract, additional experimental evidence for the generality of the proposed mechanism, and a stringent rewording of the main text drawing a clear line, omitting unnecessary details and focussing on the novel findings.

      Our response: Please see our responses above. In addition, we will reword the main text where necessary.

      REVIEWER #2

      Evidence, reproducibility and clarity

      Summary:

      *In this manuscript, Devaux et al. report that the anti-cancer drug cisplatin upregulates intronic polyadenylation (IPA) isoforms in non-small cell lung cancer cell lines. Their finding was based on 3' end sequencing and long-read sequencing. Through polysome profiling they confirmed that many of the IPA isoforms are translated, despite being inefficient in most cases. *

      Our response: There is some misunderstanding here. We will clarify in the text that inefficient association with heavy polysomes is observed for a minority (not the majority) of IPA isoforms. For this, in Fig. 2B and S2A, we will add the information that for the majority of IPA sites, the IPA:LE ratio is not significantly different (neither up or down) between total cytosol and heavy polysomes.

      They validated functions of IPA isoforms from two genes, PHF20 and PRKAR1B, in cell survival upon cisplatin treatment, based on an array of methods, including siRNA knockdown, CRISPR knockout of IPA polyA site, and CRISPR mutation of the start codon. They further found that FANCD2 and Senataxin can regulate cisplatin-mediated IPA activation. The authors advocate a new paradigm of expression of IPA-encoding microproteins in cisplatin-treated cells.

      Our response: We would like to point out that our data indicate that cisplatin upregulation of the IPA:LE isoform ratio is mediated at least in part by an inhibition of transcription processivity (explaining the decrease of LE isoforms), and that we do not claim an ‘IPA activation’ (that is, enhanced used of IPA sites) by cisplatin. This remark is also related to major point 1 below.

      Major comments:

        • While the phenomenon of IPA isoform upregulation by Cisplatin is quite convincing, the underlying mechanism is largely elusive. The authors indicated processivity as a potential mechanism and the effects of FACD2 and Senataxin appear in line with this hypothesis. However, they cannot rule out other possibilities based on the data presented in the manuscript. For example, it is not clear if the elongation rate of Pol II (distinct from its processivity) or nuclear RNA degradation is affected by cisplatin, which could also lead to increased expression of IPA isoforms. In addition, enhanced 3' end processing activity has been previously shown to activate IPA sites. Therefore, the underlying mechanism is mostly speculative.* Our response: As explained on page 14, the reason why we focused on transcription processivity is that the cisplatin-induced upregulation of the IPA:LE isoform ratio was enriched in long genes and was accompanied by a decrease of LE isoform levels. Importantly, our data (e.g., for the PHF20 and PRKAR1B genes) indicate that the cisplatin-induced decrease of processivity explains –at least in part– the selective decrease of the LE but not IPA isoform levels and therefore the increase of the IPA:LE isoform ratio; we will clarify this in the manuscript (on pages 14 and 15). Our data also show that cisplatin effects on both processivity and IPA:LE isoform ratio are dependent on FANCD2 and SETX. We agree with the reviewer that we cannot exclude that IPA:LE isoform ratio upregulation by cisplatin might also be mediated in part by additional mechanisms (e.g., ‘factors involved in cleavage/ polyadenylation, splicing, transcription elongation and termination, and epigenetic marks’, as mentioned in the discussion on page 16) and we will add nuclear RNA degradation to the list of potential factors. However, we want to emphasize that the role of processivity is not speculative.

      *The authors used the polysome:cytosolic ratio to indicate translational efficiency. However, because *the CDS size affects the number of ribosomes per mRNA, the translational efficiency should be based on polysome:cytosolic ratio normalized to CDS size. Ideally, the authors should calculate number of ribosome per transcript based on monosome, light polysome and heavy polysome.

      Our response: We cannot normalize ribosome number by CDS size because (i) heavy polysomes are not a precise number of ribosomes and (ii) sORFs are not annotated as CDS.

      The functions of PHF20 and PRKAR1B IPA isoforms are based on knockdown or knockout mutations. Because of its gain-of-function property, overexpression of the isoforms in cisplatin-treated cells would be necessary to definitively confirm their funcitons.

      Our response: For PRKAR1B sORF#2, we ____will carry out overexpression of the sORF microprotein in A549 cells and CRISPR clones and analyze its effects on cell growth and cisplatin survival. We have appropriate constructs for this.

      Minor:

        • Fig. 1H, the numbers of IPA and LE transcripts should be provided. The statistical significance for the difference should also be included.* Our response: The numbers of IPA and LE transcripts were provided in Fig. S1I and we will provide the statistical significance (which is good), as requested.

      Fig. 1I, the image should be accompanied with fold difference as indicated in the text. Some statistics for difference between vehicle only and CisPt only is necessary.

      Our response: We will indicate the fold differences and provide the statistical significance, as requested.

      • Fig. 6, the authors did data mining of ribo-seq data and mass-spec data and identified 156 genes whose IPA isoforms have potentials of protein expression. The enriched GO terms for the 156 IPA genes are different than the overall IPA isoforms shown in Fig. 1C. Does this mean some genes, like those in DNA damage stimulus, produce IPA isoforms with different consequences, such as to inhibit their expression? *

      Our response: We think it is difficult to compare the enriched GO terms between overall IPA and miP-5’UTR-IPA. Indeed, differences could be due in part to trivial reasons (e.g., different number of genes in the lists). As suggested by this reviewer, it could be that for some gene sets enriched in particular functions, IPA may serve to downregulate the expression level of the full-length (canonical) mRNA. We discuss that this may be the case for the PRIM2 gene involved in DNA replication (page 17), but expanding on this would be speculative. Likewise, IPA isoforms encoding carboxy-terminal isoforms of canonical proteins, or IPA isoforms with a noncoding function (like ASCC3 or SPUD), might be enriched in particular gene functions, but again this idea is speculative and it goes beyond the scope of our manuscript.

      In addition, the authors need to use ribo-seq and mass spec data as a validation tool for their polysome profiling data to indicate the reliability of using polysome data to call protein expression.

      Our response: This comment seems to concern those IPA isoforms that are abundant in heavy polysomes. We do not wish to validate protein production from such isoforms, because they are not the focus of our study.

      Significance

      The significance of this work is its novelty in reporting IPA isoform activation by cisplatin. More importantly, some IPA isoform give rise to microproteins that have functional roles in cell survival upon cisplatin treatment.

      Our response: We would like to highlight that the main finding of our manuscript is the discovery that IPA isoforms are a source of microproteins. Cisplatin response is the biological context in which we did the study, and therefore our functional and mechanistic analyses.

      REVIEWER #3

      Evidence, reproducibility and clarity

      Devaux et al. report how cisplatin treatment changes the abundance of mRNA isoforms, favoring the expression of short transcripts originating from intronic polyadenylation (IPA) events relative to the expression of the corresponding mRNA isoform that includes the last annotated exon (LE). To detect IPA events the authors performed 3' end sequencing of polyadenylated mRNAs, long-read sequencing and conventional total RNA sequencing experiments in control and cisplatin treated cells. Analysis of the 3' end sequencing data revealed numerous genes showing an increase in the IPA:LE ratio upon cisplatin treatment, whereas few events with a decreased IPA:LE ratio were detected. Many of the identified events could be corroborated by the long-read sequencing data, sequencing of total RNA, and an existing polyA database. Furthermore, the authors validate IPA:LE ratios for a few selected genes using quantitative PCR. Subsequently, the authors continue to analyze if IPA isoforms are translated with a specific focus on IPA isoforms that do not contain any parts of the LE isoform coding sequence but terminated transcription in what is annotated as 5' untranslated region (UTR). These experiments show that IPA isoforms (including 5' UTR-IPAs) are translated but frequently associated with fewer ribosomes than the corresponding LE isoform. For two selected 5' UTR-IPA isoforms the authors identified potential small open reading frames (sORFs) that could give rise to microproteins with a potential function during cisplatin treatment. siRNA experiments targeting either the 5' UTR-IPA or the LE mRNA isoform of selected genes identified a small but significant differential effect on cell viability upon cisplatin treatment. Similar results were obtained when the endogenous IPA locus was deleted or the start codon of the potential sORF was mutated. Finally, the authors shed some light onto the molecular mechanisms of how cisplatin affects the IPA:LE ratio by decreasing transcription processivity.

      *This is an interesting manuscript suggesting a link between IPA, sORFs and cancer treatment. The manuscript offers valuable datasets as a resource for the research community. While the authors generally present a well-analyzed and validated dataset supporting their claims, some aspects require further evidence or clearer presentation for robustness and reader comprehension. In addition, the manuscript would benefit from improving data visualization and we have several suggestions (see below) on how to make the representation of the data in the figures more appealing to the reader. We encourage the authors to reconsider several of their bar plots and instead plot their data on a continuous axis, e.g. using a scatter plot (fold change versus FDR) instead of a bar chart that can only represent up/down total numbers. *

      Our response: Please see our responses below.

      *Main points: *

        • We disagree with one of the data interpretations concerning the high polysome (HP) versus total cytosolic polysomes (cytosol) localized IPA and LE mRNA isoforms in the paragraph "A subset of IPA isoforms are depleted in heavy polysomes and terminate in the annotated 5'UTR part of genes". Preferential IPA isoform localization to cytosol versus HP in comparison to the LE isoform does not mean that the IPA isoform translation efficiency is lower than that of the LE isoform. It just reflects the fact that IPA isoform coding sequence is considerably shorter than the coding sequence of the LE isoform (and thus can accommodate fewer ribosomes!). The authors mention that point later in the text but it should already be made clear at this point in the manuscript. They should make sure not to confuse translation efficiency (ribosome density across an open reading frame) and open reading frame length. * Our response: We will modify the text of this section (pages 10-11). We __will __state that ‘the HP:cytosol ratio is usually considered as a proxy for translation efficiency’ and we will only make conclusions in terms of ‘HP:cytosol ratio’ or ‘HP recruitment efficiency’, instead of ‘translation efficiency’ (we had used this term in a few sentences for the sake of simplicity). Please note that these changes will not alter the main conclusion of this part, because both the title (‘a subset of IPA isoforms are depleted in heavy polysomes and terminate in the annotated 5'UTR part of genes’) and the end of this section (page 11), as well as the legend of Fig. 2, were already written in such terms. Thus, in this section, we do not need to discuss ORF length (and we cannot, because sORFs are not annotated as CDS and we introduce sORFs only two sections later [Fig. 4]).
      • In Figure 5, the authors claim that the "cisplatin survival phenotype of the PRKAR1B 5'UTR-IPA isoform is attributable to its small ORF#2". This is an interesting phenotype but the authors only present a WST1 assay to support these claims. Given that it is an important Figure in their manuscript and links the observations made earlier to cisplatin-induced survival, it would be critical to bolster these claims with additional data, e.g. AnnexinV/PI staining and flow cytometry to distinguish changes in cisplatin-induced apoptosis from proliferation.*

      Our response: We will make the requested experiments with FACS analysis of Annexin V and PI/BrdU to distinguish changes in cisplatin-induced apoptosis from proliferation (cell cycle).

      • Along the same line, it would be important to test the overexpression of the sORF microprotein upon cisplatin treatment. Changes in the mRNA sequence (such as the AUG mutation) could potentially also alter the mRNA structure. It would therefore be critical to show that the sORF microprotein is indeed responsible for the changes in cisplatin-induced viability (for instance by expression of a sORF::P2A::GFP construct). *

      Our response: As requested, we will test whether overexpression of the sORF microprotein can rescue the cisplatin survival phenotype of our PRKAR1B IPA and ATG mutants. We have appropriate constructs for this.

      • Figure 5C: Please show the Western blot of PRKAR1B and GAPDH and not just the quantification. There is plenty of space in Figure 5. *

      Our response: We will show the Western blots for PRKAR1B and GAPDH.

      • In the following, we list suggestions to improve different figures where the data could be more adequately presented:*

      - Figure 1A and B: We suggest representing the data in a scatter plot log fold change on the x-axis and FDR on the y-axis. The authors decided for an FDR cutoff of 10%. This is quite high. Why did the authors decide for this cutoff? How many genes would be identified with a more stringent cutoff (1% for example)? Please list the corresponding FDR values in TableS4.

      Our response: We have never seen in the literature 3’-seq (or related) data of IPA:LE ratio regulation plotted as a scatter plot with log fold change on the x-axis and FDR on the y-axis. Instead, we propose to provide scatter plots with IPA fold change on the x-axis and LE fold change on the y-axis, as in many previous studies. We were not very stringent on the FDR or adjusted p values, in order to reduce the rate of false negatives, because we then cross our lists of regulated IPAs in different compartments (e.g., cytosol and heavy polysomes; Fig. 2C). We provided adjusted p values in Table S4; with an adjusted p value of 1%, we observe 1986 upregulated IPA sites and 33 downregulated ones.

      *-Figure 1C: There are many ways to visualize fold change, p value and number of genes of a GO term analysis. The authors could choose one of the common ways to represent such data instead of just showing raw numbers in a table. *

      Our response: We like showing GO terms as tables, but we can provide a figure if necessary.

      -Figure 1E-G: Add to the figure that PRIM2 was assayed. It is only written in the figure legend.

      Our response: We will write ‘PRIM2’ in the figure.

      *-Figure 2A and B: Same suggestion as for Figure 1A and B, a scatter plot log fold change on the x-axis and FDR on the y-axis would visualize the data much better. *

      Our response: Same response as for Fig 1A-B above.

      -Figure S1B: Where does the number of 2118 cisplatin regulated genes come from? It was not described anywhere else. Should it not be 1987 regulated genes?

      Our response: We will clarify that 2118 is the union of genes with cisplatin upregulated IPA:LE ratio in H358 and/or A549 cells.

      -Figure S1H: Typo in the y-axis.

      Our response: This typo will be corrected, thanks.

      -Figure S2A: Same suggestion as for Figure 1A and B, a scatter plot log fold change on the x-axis and FDR on the y-axis.

      Our response: Same response as for Fig 1A-B above.

      -Figure S3C: If possible, show the plotted digital data of the polysome curves.

      Our response: We do not have digital data for the polysome curves, just the printed graph shown at the bottom of the figure.

      • Data availability: The provided UCSC genome browser link unfortunately does not load the data bam files. Please fix.*

      Our response: We will fix this upon submission to journal.

      *Minor points: *

      • Please check the text for typos, e.g. page 8: artefacts instead of artifacts. *

      Our response: We will check for typos.

      Significance

      The manuscript describes an interesting link between intronic polyadenylation, sORFs and cancer treatment and will be of interest to the gene expression regulation and RNA communities. As a relatively unknown mechanism to induce sORF-encoded microproteins, the study could lead to follow-up studies tackling intronic polyadenylation and their role in sORF expression.

      Our response: We would like to highlight that IPA was not previously known to induce sORF-encoded microproteins.

      While the authors generally present a well-analyzed and validated dataset, the link between sORF function and cisplatin response will require additional experiments to strengthen the sORF's impact for cellular survival.

      Our response: Please see our responses to main points 2 and 3 above.

      3. Description of the revisions that have already been incorporated in the transferred manuscript

      None.

      4. Description of analyses that authors prefer not to carry out

      REVIEWER #2

      *Major point #2: **The authors used the polysome:cytosolic ratio to indicate translational efficiency. However, because *the CDS size affects the number of ribosomes per mRNA, the translational efficiency should be based on polysome:cytosolic ratio normalized to CDS size. Ideally, the authors should calculate number of ribosome per transcript based on monosome, light polysome and heavy polysome.

      Our response: We cannot normalize ribosome number by CDS size because (i) heavy polysomes are not a precise number of ribosomes and (ii) sORFs are not annotated as CDS.

      *Minor point #3: Fig. 6, the authors did data mining of ribo-seq data and mass-spec data and identified 156 genes whose IPA isoforms have potentials of protein expression. The enriched GO terms for the 156 IPA genes are different than the overall IPA isoforms shown in Fig. 1C. Does this mean some genes, like those in DNA damage stimulus, produce IPA isoforms with different consequences, such as to inhibit their expression? *

      Our response: We think it is difficult to compare the enriched GO terms between overall IPA and miP-5’UTR-IPA. Indeed, differences could be due in part to trivial reasons (e.g., different number of genes in the lists). As suggested by this reviewer, it could be that for some gene sets enriched in particular functions, IPA may serve to downregulate the expression level of the full-length (canonical) mRNA. We discuss that this may be the case for the PRIM2 gene involved in DNA replication (page 17), but expanding on this would be speculative. Likewise, IPA isoforms encoding carboxy-terminal isoforms of canonical proteins, or IPA isoforms with a noncoding function (like ASCC3 or SPUD), might be enriched in particular gene functions, but again this idea is speculative and it goes beyond the scope of our manuscript.

      In addition, the authors need to use ribo-seq and mass spec data as a validation tool for their polysome profiling data to indicate the reliability of using polysome data to call protein expression.

      Our response: This comment seems to concern those IPA isoforms that are abundant in heavy polysomes. We do not wish to validate protein production from such isoforms, because they are not the focus of our study.

      REVIEWER #3

      -Figure S3C: If possible, show the plotted digital data of the polysome curves.

      Our response: We do not have digital data for the polysome curves, just the printed graph shown at the bottom of the figure.

    6. Review Commons

      Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

      Learn more at Review Commons

      Referee #3

      Evidence, reproducibility and clarity

      Devaux et al. report how cisplatin treatment changes the abundance of mRNA isoforms, favoring the expression of short transcripts originating from intronic polyadenylation (IPA) events relative to the expression of the corresponding mRNA isoform that includes the last annotated exon (LE). To detect IPA events the authors performed 3' end sequencing of polyadenylated mRNAs, long-read sequencing and conventional total RNA sequencing experiments in control and cisplatin treated cells. Analysis of the 3' end sequencing data revealed numerous genes showing an increase in the IPA:LE ratio upon cisplatin treatment, whereas few events with a decreased IPA:LE ratio were detected. Many of the identified events could be corroborated by the long-read sequencing data, sequencing of total RNA, and an existing polyA database. Furthermore, the authors validate IPA:LE ratios for a few selected genes using quantitative PCR. Subsequently, the authors continue to analyze if IPA isoforms are translated with a specific focus on IPA isoforms that do not contain any parts of the LE isoform coding sequence but terminated transcription in what is annotated as 5' untranslated region (UTR). These experiments show that IPA isoforms (including 5' UTR-IPAs) are translated but frequently associated with fewer ribosomes than the corresponding LE isoform. For two selected 5' UTR-IPA isoforms the authors identified potential small open reading frames (sORFs) that could give rise to microproteins with a potential function during cisplatin treatment. siRNA experiments targeting either the 5' UTR-IPA or the LE mRNA isoform of selected genes identified a small but significant differential effect on cell viability upon cisplatin treatment. Similar results were obtained when the endogenous IPA locus was deleted or the start codon of the potential sORF was mutated. Finally, the authors shed some light onto the molecular mechanisms of how cisplatin affects the IPA:LE ratio by decreasing transcription processivity.

      This is an interesting manuscript suggesting a link between IPA, sORFs and cancer treatment. The manuscript offers valuable datasets as a resource for the research community. While the authors generally present a well-analyzed and validated dataset supporting their claims, some aspects require further evidence or clearer presentation for robustness and reader comprehension. In addition, the manuscript would benefit from improving data visualization and we have several suggestions (see below) on how to make the representation of the data in the figures more appealing to the reader. We encourage the authors to reconsider several of their bar plots and instead plot their data on a continuous axis, e.g. using a scatter plot (fold change versus FDR) instead of a bar chart that can only represent up/down total numbers.

      Main points:

      1. We disagree with one of the data interpretations concerning the high polysome (HP) versus total cytosolic polysomes (cytosol) localized IPA and LE mRNA isoforms in the paragraph "A subset of IPA isoforms are depleted in heavy polysomes and terminate in the annotated 5'UTR part of genes". Preferential IPA isoform localization to cytosol versus HP in comparison to the LE isoform does not mean that the IPA isoform translation efficiency is lower than that of the LE isoform. It just reflects the fact that IPA isoform coding sequence is considerably shorter than the coding sequence of the LE isoform (and thus can accommodate fewer ribosomes!). The authors mention that point later in the text but it should already be made clear at this point in the manuscript. They should make sure not to confuse translation efficiency (ribosome density across an open reading frame) and open reading frame length.

      2. In Figure 5, the authors claim that the "cisplatin survival phenotype of the PRKAR1B 5'UTR-IPA isoform is attributable to its small ORF#2". This is an interesting phenotype but the authors only present a WST1 assay to support these claims. Given that it is an important Figure in their manuscript and links the observations made earlier to cisplatin-induced survival, it would be critical to bolster these claims with additional data, e.g. AnnexinV/PI staining and flow cytometry to distinguish changes in cisplatin-induced apoptosis from proliferation.

      3. Along the same line, it would be important to test the overexpression of the sORF microprotein upon cisplatin treatment. Changes in the mRNA sequence (such as the AUG mutation) could potentially also alter the mRNA structure. It would therefore be critical to show that the sORF microprotein is indeed responsible for the changes in cisplatin-induced viability (for instance by expression of a sORF::P2A::GFP construct).

      4. Figure 5C: Please show the Western blot of PRKAR1B and GAPDH and not just the quantification. There is plenty of space in Figure 5.

      5. In the following, we list suggestions to improve different figures where the data could be more adequately presented:

        • Figure 1A and B: We suggest representing the data in a scatter plot log fold change on the x-axis and FDR on the y-axis. The authors decided for an FDR cutoff of 10%. This is quite high. Why did the authors decide for this cutoff? How many genes would be identified with a more stringent cutoff (1% for example)? Please list the corresponding FDR values in TableS4.
        • Figure 1C: There are many ways to visualize fold change, p value and number of genes of a GO term analysis. The authors could choose one of the common ways to represent such data instead of just showing raw numbers in a table.
        • Figure 1E-G: Add to the figure that PRIM2 was assayed. It is only written in the figure legend.
        • Figure 2A and B: Same suggestion as for Figure 1A and B, a scatter plot log fold change on the x-axis and FDR on the y-axis would visualize the data much better.
        • Figure S1B: Where does the number of 2118 cisplatin regulated genes come from? It was not described anywhere else. Should it not be 1987 regulated genes?
        • Figure S1H: Typo in the y-axis.
        • Figure S2A: Same suggestion as for Figure 1A and B, a scatter plot log fold change on the x-axis and FDR on the y-axis.
        • Figure S3C: If possible, show the plotted digital data of the polysome curves.

      6. Data availability: The provided UCSC genome browser link unfortunately does not load the data bam files. Please fix.

      Minor points:

      1. Please check the text for typos, e.g. page 8: artefacts instead of artifacts.

      Significance

      The manuscript describes an interesting link between intronic polyadenylation, sORFs and cancer treatment and will be of interest to the gene expression regulation and RNA communities. As a relatively unknown mechanism to induce sORF-encoded microproteins, the study could lead to follow-up studies tackling intronic polyadenylation and their role in sORF expression.

      While the authors generally present a well-analyzed and validated dataset, the link between sORF function and cisplatin response will require additional experiments to strengthen the sORF's impact for cellular survival.

    7. Review Commons

      Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

      Learn more at Review Commons

      Referee #2

      Evidence, reproducibility and clarity

      Summary:

      In this manuscript, Devaux et al. report that the anti-cancer drug cisplatin upregulates intronic polyadenylation (IPA) isoforms in non-small cell lung cancer cell lines. Their finding was based on 3' end sequencing and long-read sequencing. Through polysome profiling they confirmed that many of the IPA isoforms are translated, despite being inefficient in most cases. They validated functions of IPA isoforms from two genes, PHF20 and PRKAR1B, in cell survival upon cisplatin treatment, based on an array of methods, including siRNA knockdown, CRISPR knockout of IPA polyA site, and CRISPR mutation of the start codon. They further found that FANCD2 and Senataxin can regulate cisplatin-mediated IPA activation. The authors advocate a new paradigm of expression of IPA-encoding microproteins in cisplatin-treated cells.

      Major comments:

      • While the phenomenon of IPA isoform upregulation by Cisplatin is quite convincing, the underlying mechanism is largely elusive. The authors indicated processivity as a potential mechanism and the effects of FACD2 and Senataxin appear in line with this hypothesis. However, they cannot rule out other possibilities based on the data presented in the manuscript. For example, it is not clear if the elongation rate of Pol II (distinct from its processivity) or nuclear RNA degradation is affected by cisplatin, which could also lead to increased expression of IPA isoforms. In addition, enhanced 3' end processing activity has been previously shown to activate IPA sites. Therefore, the underlying mechanism is mostly speculative.
      • The authors used the polysome:cytosolic ratio to indicate translational efficiency. However, because the CDS size affects the number of ribosomes per mRNA, the translational efficiency should be based on polysome:cytosolic ratio normalized to CDS size. Ideally, the authors should calculate number of ribosome per transcript based on monosome, light polysome and heavy polysome.
      • The functions of PHF20 and PRKAR1B IPA isoforms are based on knockdown or knockout mutations. Because of its gain-of-function property, overexpression of the isoforms in cisplatin-treated cells would be necessary to definitively confirm their funcitons.

      Minor:

      Fig. 1H, the numbers of IPA and LE transcripts should be provided. The statistical significance for the difference should also be included.

      Fig. 1I, the image should be accompanied with fold difference as indicated in the text. Some statistics for difference between vehicle only and CisPt only is necessary.

      Fig. 6, the authors did data mining of ribo-seq data and mass-spec data and identified 156 genes whose IPA isoforms have potentials of protein expression. The enriched GO terms for the 156 IPA genes are different than the overall IPA isoforms shown in Fig. 1C. Does this mean some genes, like those in DNA damage stimulus, produce IPA isoforms with different consequences, such as to inhibit their expression? In addition, the authors need to use ribo-seq and mass spec data as a validation tool for their polysome profiling data to indicate the reliability of using polysome data to call protein expression.

      Significance

      The significance of this work is its novelty in reporting IPA isoform activation by cisplatin. More importantly, some IPA isoform give rise to microproteins that have functional roles in cell survival upon cisplatin treatment.

    8. Review Commons

      Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

      Learn more at Review Commons

      Referee #1

      Evidence, reproducibility and clarity

      Microporteins originating from coding and non-coding transcript are increasingly understood to control various cellular processes. In the present study, the authors investigated whether intronic polyadenylation (IPA) contributes to the formation of transcript isoforms encoding microproteins. Using genotoxic stress by cisplatin as a model in cell cultures, the authors detect abundant IPA. IPA in a subset of such transcripts leads to short 5'UTR transcript isoforms that are poorly associated with heavy polysomes and encode microproteins. For PRKAR1B, they demonstrate the expression of a corresponding microprotein and a function in modulating the cisplatin response. Based on depletion experiments of FANCD2 and STX1, the authors propose that impaired transcription processivity after cisplatin is one mechanism leading to IPA and microprotein production.

      While this is an interesting manuscript, I felt the support for the claimed generalization falls a bit short.

      Major:

      If I see it correctly, the authors mainly refer to existing riboSeq data and evidence from mass spectrometry/proteomics to infer the generality of the mechanism (beyond PRKAR1B). It is important to back this up with further experiments and validate this for the set-up used in this manuscript. This concerns the existence of the microproteins but also the downstream functional impact.

      Also, I wonder is this limited to cisplatin-induced genotoxic stress and the specific cell line used or is this a more global mechanism?

      Minor:

      • While the rest of the paper reads well, the abstract could be improved/simplified to increase accessibility
      • Page 11: Pertaining to Figure 3 and the functional impact: The authors analyze the IPA effect by probing cell viability and cell survival. It would be important to define the effects in further detail, as the mere regulation of cell cycle and/or apoptosis could also result in such outcome (which is then not necessarily a direct cisplatin response). Does this also impact the response to other genotoxic stress (also pertains to the effects studied and shown in Fig. 5)?
      • Page 11: Pertaining to Figure 3 and the functional impact: The authors analyze the IPA effect by probing cell viability and cell survival. It would be important to define the effects in further detail, as the mere regulation of cell cycle and/or apoptosis could also result in such outcome (which is then not necessarily a direct cisplatin response). Does this also impact the response to other genotoxic stress (also pertains to the effects studied and shown in Fig. 5)?
      • Page 12 concerning the microprotein expression: the authors refer to data from other resources to claim that the microproteins are expressed, however they fail to demonstrate this for their setup (at least for 2 out of three they study here). I think this is a weak point as it does not directly support the general claim.
      • Also, I did not understand what the authors intended to demonstrate with the immunoflourescence (Fig. 4E). What should a defined nuclear expression imply versus the diffuse staining throughout after cisplatin? How does this relate to the functional effects?
      • Page 13/Fig. 5E: the different clones of the mATG show very high variability. To my understanding it is difficult to draw a clear conclusion from this heterogeneity.
      • Page 15 on the mechanism: SETX has been demonstrated to control poly(A) site choice (PMID: 21700224, 32976578). However the quantitative role of SETX in poly(A) site choice regulation (compared to other regulators) seems to be rather marginal and not strictly unidirectional, i.e after SETX depletion also longer transcript isoforms can be detected (PMID: 32976578). How does this relate to the proposed mechanism of SETX-dependent processivity? Interestingly, from PMID: 32976578 it also appears that PRKAR1A has a 5'UTR poly(A) site that is regulated in a SETX-dependent manner.
      • Page 16, discussion first paragraph. While refs 1-4 are nice reviews that could be quoted here a study that appeared later represents the most comprehensive analysis to date covering the different facets from transcription to RNA processing and the resulting impact on poly(A) site choice (PMID: 30552333).

      Significance

      This could be a very significant report, provided the generality of the claims and mechanistic insigths are further strengthend.

      Overall it targets a rather specialized readership. This could be improved by simplifing the abstract, additional experimental evidence for the generality of the proposed mechanism, and a stringent rewording of the main text drawing a clear line, omitting unnecessary details and focussing on the novel findings.

    9. Version published to 10.1101/2023.12.05.569446v1 on bioRxiv