Cell stress triggers major changes in the control of gene expression at the transcriptional and post-transcriptional levels. One of the main responses to stress is the blockade of global translation allowing cells to save energy. This process results from inactivating the canonical cap-dependent mechanism of translation initiation (Holcik and Sonenberg, 2005). However, translation of specific mRNAs is maintained or even increased during stress via alternative mechanisms of translation initiation. One of these mechanisms involves internal ribosome entry sites (IRES), structural elements mostly present in the 5’ untranslated regions of specific mRNAs, which drive the internal recruitment of ribosomes onto mRNA and promote cap-independent translation initiation (Godet et al., 2019).

Hypoxia, or the lack of oxygen, is a major stress in pathologies such as cancer and cardiovascular diseases (Pouysségur et al., 2006). In particular, in ischemic heart failure disease, coronary artery branch occlusion exposes cardiac cells to hypoxic conditions. The cell response to hypoxia induces angiogenesis and lymphangiogenesis to reperfuse the stressed tissue with new vessels and allow cell survival (Morfoisse et al., 2014; Pouysségur et al., 2006; Tatin et al., 2017). The well-known response to hypoxia is the transcriptional induction of specific genes under the control of the hypoxia-induced factors 1 and 2 (HIF1, HIF2) (Hu et al., 2003; Koh et al., 2011). However, we have recently reported that most mRNAs coding (lymph)angiogenic growth factors are induced at the translatome level in hypoxic cardiomyocytes (Hantelys et al., 2019). Expression of these factors allows the recovery of functional blood and lymphatic vasculature in ischemic diseases, including myocardial infarction (Tatin et al., 2017; Ylä-Herttuala and Baker, 2017). The mRNAs of the major (lymph)angiogenic growth factors belonging to the fibroblast growth factor (FGF) and vascular endothelial growth factor (VEGF) families all contain IRESs that are activated in early hypoxia (Morfoisse et al., 2014; Hantelys et al., 2019).

IRES-dependent translation is regulated by IRES trans-acting factors (ITAFs) that are in most cases RNA-binding proteins acting as positive or negative regulators. A given ITAF can regulate several IRESs, while a given IRES is often regulated by several ITAFs (Godet et al., 2019), depending on the cell type or physiology. This has led to the concept of IRESome, a multi-partner ribonucleic complex allowing ribosome recruitment onto the mRNA via the IRES.

ITAFs often exhibit several functions in addition to their ability to control translation. Many of them play a role in alternative splicing, transcription, ribosome biogenesis or RNA stability (Godet et al., 2019). Clearly, a large part of ITAFs are nuclear proteins able to shuttle between nucleus and cytoplasm. Previous data have also shown that a nuclear event is important for cellular IRES activity, leading to the hypothesis of IRESome formation in the nucleus (Ainaoui et al., 2015; Semler and Waterman, 2008; Stoneley et al., 2000).

Interestingly, several ITAFs are components of a nuclear body, the paraspeckle, formed in response to stress (Choudhry et al., 2015; Fox et al., 2002). These ITAFs include several hnRNPs, as well as major paraspeckle proteins such as P54^nrb nuclear RNA binding (P54^nrb/NONO) and splicing factor proline and glutamine-rich (SFPQ/PSF). P54^nrb and SFPQ belong to the family of Drosophila melanogaster behavior and human splicing (DBHS) proteins whose third member is the paraspeckle protein C1 (PSPC1). P54^nrb and SFPQ are essential for paraspeckle formation while PSPC1 is not. These three DBHS proteins are known to interact with each other and function in heteroduplexes (Fox et al., 2005; Lee et al., 2015; Passon et al., 2012). In addition, P54^nrb and SFPQ interact with the long non-coding RNA (lncRNA) Neat1 (nuclear enriched abundant transcript 1), that constitutes the skeleton of the paraspeckle (Clemson et al., 2009; Sunwoo et al., 2009). This lncRNA, a paraspeckle essential component, is present as two isoforms Neat1_1 and Neat1_2 whose sizes in mouse are 3.2 and 20.8 kilobases, respectively (Sunwoo et al., 2009). Its transcription is induced during hypoxia by HIF2 and promotes paraspeckle formation (Choudhry et al., 2015). Neat1 is overexpressed in many cancers (Yang et al., 2017). Recently, its induction by hypoxia has been shown in cardiomyocytes where it plays a role in cell survival (Kenneweg et al., 2019).

According to previous reports, paraspeckle is able to control gene expression via the retention of edited mRNAs and transcription factors (Hirose et al., 2014; Imamura et al., 2014; Prasanth et al., 2005). In 2017, Shen et al. have also shown that the paraspeckle might inhibit translation by sequestering p54^nrb and SFPQ which are ITAFs of the MYC IRES (Shen et al., 2017).

In this study, we were interested in finding new ITAFs responsible for activating (lymph)angiogenic factor mRNA IRESs in HL-1 cardiomyocytes, during early hypoxia. We have previously shown that the two paraspeckle proteins p54^nrb and hnRNPM are ITAFs, activators of the FGF1 IRES during myoblast differentiation (Ainaoui et al., 2015). This incited us to investigate the potential role of the paraspeckle and of Neat1 in the control of IRES-dependent translation in hypoxic cardiomyocytes. We show here that Neat1 expression and paraspeckle formation correlate with the activation of the FGF1 IRES during hypoxia, in cardiomyocytes and breast cancer cells. The knock-down of p54^nrb, PSPC1 or Neat1 generates a decrease in FGF1 IRES activity and in endogenous FGF1 expression. Furthermore, our data revealed that IRES-containing mRNA is colocalized with Neat1 in paraspeckle during hypoxia. By quantitative mass spectrometry analysis of the p54^nrb interactome, we identified two additional ITAFs able to control the FGF1 IRES activity: nucleolin and ribosomal protein RPS2. Analysis of IRESs in the knock-down experiments showed that p54^nrb and PSPC1 are activators of several but not all IRESs of (lymph)angiogenic and cardioprotective factor mRNAs whereas Neat1 appears as a strong activator of all the cellular IRESs tested. These data suggest that the paraspeckle, via Neat1 and several protein components would be the site of IRESome assembly in the nucleus. In addition, a polysome PCR array reveals that Neat1 affects the translation of most IRES-containing mRNAs and of several mRNA families involved in hypoxic response, angiogenesis and cardioprotection.

The present data demonstrate a link between the paraspeckle and the control of IRES-dependent translation during hypoxia in mouse cardiomyocytes. We show that three major paraspeckle components regulate IRES-dependent translation: p54^nrb, PSPC1, and Neat1, as well as by two proteins present in the p54^nrb nuclear interactome, nucleolin and RPS2. Neat1 appears as the key to this paraspeckle-related activation of translation in response to hypoxia. This lncRNA is an activator of all cellular IRESs tested, but not of the viral EMCV IRES. More broadly, Neat1 isoforms impact the recruitment into polysomes of most IRES-containing mRNAs and several families of mRNAs involved in the response to hypoxia. The colocalization of IRES-containing mRNA with Neat1 RNA in paraspeckles increased in hypoxia conditions, suggesting that the paraspeckle may be a recruitment platform for IRES-containing mRNAs during stress and that the IRESome could be assembled in the paraspeckle before mRNA export from the nucleus (Figure 9).

Figure 9

Download asset Open asset

It may be noted that the inhibition of IRES activities resulting from ITAF depletion is quite moderate for the different proteins while stronger for the lncRNA Neat1. This cannot be explained only by differences in knock-down efficiency. We hypothesize is that several proteins are present in the IRESome complex and that there may be a certain redundancy between them. Thus, the depletion of a single ITAF would not be sufficient to abolish the IRES activity completely. Also, to explain why paraspeckle ITAFs such as p54^nrb and PSPC1 do not inhibit all the IRESs, we propose that the paraspeckle IRESome protein composition varies depending on the IRES and the hypoxic or normoxic condition, while Neat1 remains the main actor of the process. Several observations suggest that Neat1_2 may be the main isoform involved. However, the knock-down of Neat1-2 isoform with a specific gapmer does not affect IRES activity as much as the knock-down of both Neat1 isoforms (Figure 2—figure supplement 2). We were not successful in knocking down the isoform Neat1_1, as its sequence is entirely contained in Neat1_2. Thus at this stage we conclude that the two isoforms are probably involved. The fluidigm PCR array suggests that they may affect translation differently (Figure 8).

We searched for an ITAF able to regulate a set of IRESs during hypoxia and found the lncRNA Neat1 as a wide activator of IRES-dependent translation. However, our data show that Neat1 also regulates IRES activities both in normoxia and hypoxia. One explanation may be that Neat1 is already expressed in normoxia in HL-1 cells, which are transformed cells despite their cardiomyocyte beating phenotype (Claycomb et al., 1998). Although Neat1 expression and paraspeckle number increase in response to hypoxia, a significant percentage of cells already contain paraspeckles in normoxia, which may explain why IRESs are already active in normoxia. It has been reported that Neat1_2 is not expressed in all tissues in vivo, whereas it is found in all transformed or immortalized cell lines (data not shown) (Nakagawa et al., 2011). In concordance with this observation, previous reports show that cellular IRESs are active in all cultured cell lines while inactive or tissue-specific in mice (Créancier et al., 2000; Créancier et al., 2001). The presence of paraspeckles in normoxia may also reflect the stress due to the transfection procedure, which could interfere with the effect of the hypoxic stress performed in our study. A different approach to obtain Neat1 silencing, such as CRISPR/Cas9 mediated knock-down or knock-out could provide an interesting solution to this issue.

Our data contrast with the study of Shen et al. who showed that Neat1 depletion allows redistributing p54^nrb and SFPQ/PSF onto the MYC mRNA, in correlation with an increase in MYC protein (Shen et al., 2017). Several reasons may explain this lack of concordance. Firstly, different cell lines were used: HL-1 cardiomyocytes and 67NR breast tumor cells in the present study, HeLa and MCF7 tumor cells in the report by Shen et al. The regulation of IRES-dependent translation varies depending on cell lines. Secondly, they worked with human cell lines while our report is focused on mouse cells. In human, MYC expression is different from mouse as the MYC gene contains an additional upstream promoter, P0, which generates a longer transcript with a second IRES (Nanbru et al., 2001). Thirdly, they have not directly analyzed the MYC IRES activity but only the binding of p54^nrb and SFPQ to the MYC endogenous mRNA. Moreover an increase in myc protein expression does not necessarily correspond to increased IRES activity as the MYC mRNA is also translated by the cap-dependent mechanism (Nanbru et al., 1997). Taken together, the two studies are different rather than discordant.

A surprising result has been finding a ribosomal protein, RPS2, in the nuclear p54^nrb interactome. This suggests an extra-ribosomal role of this protein. Its interaction with p54^nrb favors the hypothesis that RPS2 would impact the IRES activity as an IRESome component in the paraspeckle. The presence of nucleolin in the complex also suggests a link of paraspeckle with nucleolus and ribosome biogenesis. Supporting this, PSPC1 was first identified in the nucleolus proteome (Fox et al., 2002). The nuclear binding of specific ribosomal proteins to IRESs might be a mechanism for forming specialized ribosomes.

Neat1 is not the first lncRNA to exhibit an ITAF function. The lncRNA TP53-regulated modulator of p27 (TRMP) has been recently described as an ITAF of the Cdkn1b/p27^kip IRES (Yang et al., 2018). TRMP inhibits the p27^kip IRES activity by competing with the IRES for pyrimidine tract binding protein (PTB) binding and prevents IRES activation mediated by PTB. Also, the lncRNA ARAP-as1 directly interacts with SFPQ, which results in release of PTB and activation of MYC IRES (Zhang et al., 2020). We have not yet deciphered the mechanism of action of Neat1. We propose that the paraspeckle would be a recruitment platform for IRES-containing mRNAs. Neat1, by interacting with p54^nrb and other paraspeckle proteins/ITAFs, would thus allow IRESome formation in the paraspeckle (Figure 9). Is the role of Neat1 exclusively nuclear in the paraspeckle, or is it exported to the cytoplasm with the IRESome complex? Several observations argue for the presence of Neat1 in the cytoplasm: our FISH experiments clearly identify the Neat1-2 isoform in the cytoplasm (Figure 1—figure supplement 2), while a recent report shows that Neat1-1 isoform is released from nucleus to cytoplasm where it suppresses the Wnt signaling in leukemia stem cells and acts as a tumor suppressor in acute myeloid leukemia (Yan et al., 2021). Neat1-2 isoform has been detected in the cytoplasm of hematopoietic cells by other authors. Interestingly, they identified a histone modifier, ASXL1, interacting with p54^nrb/NONO and involved in paraspeckle formation. Mutation of ASXL1 generates Neat1_2 export to the cytoplasm (Yamamoto et al., 2021). Furthermore, the role of cytoplasmic Neat1 in translation is suggested by our previous data showing that Neat1 is present in HL-1 cell polysomes and that this association with polysomes is increased in early hypoxia (Hantelys et al., 2019). The involvement of Neat1 in translation control via a cytoplasmic location is also supported by the presence of the triple helix in the 3’UTR of Neat1_2, whose role in translation activation has been demonstrated (Wilusz et al., 2012).

The model of IRESome formation mediated by Neat1 in the paraspeckle, and the absence of any impact of Neat1 on the picornaviral EMCV IRES activity, are both consistent with previous reports suggesting that the site of mRNA synthesis is crucial for IRES structure and function (Semler and Waterman, 2008). For picornaviruses whose mRNAs are synthesized in the cytoplasm, IRES elements would be able to form an IRESome RNP in the cytoplasm. In contrast, cellular mRNAs (as well as DNA viruses and retroviruses mRNAs) transcribed in the nucleus need a nuclear event (Ainaoui et al., 2015; Stoneley et al., 2000). The present data provide a mechanism for this nuclear history and reveal a new function of the paraspeckle, a nuclear body, in IRESome formation (Figure 9).

A role of Neat1 in ischemic heart has been recently reported showing that Neat1 down-regulation would protect cardiomyocytes from apoptosis by regulating the processing of pri-miR-22 (Gidlöf et al., 2020). Surprisingly, these authors show that hypoxia down-regulates Neat1 expression in cardiomyocytes. This contradicts our data showing that Neat1 is induced by hypoxia. Our data are however in agreement with the rest of the literature showing that Neat1 is induced by hypoxia in tumors, its transcription being activated by HIF-2 (Choudhry et al., 2015). Another study also showed that Neat1 overexpression protects cardiomyocytes against apoptosis by sponging miR125a-5p, resulting in upregulation of the apoptosis repressor gene B-cell lymphoma-2-like 12 (BCL2L12) (Yan et al., 2019). These contradictory reports highlight the complex impact of Neat1 on miRNA-mediated gene regulation.

In the present study, we have uncovered a novel role of Neat1 in the translational control of several families of genes involved in stress response, angiogenesis and cardioprotection, while it does not affect global translation. The increased protein synthesis from mRNAs coding ITAFs favors a wide role of Neat1 and of the paraspeckle in activating IRES-dependent translation. Many of the genes involved in angiogenesis or cardioprotection tested here have not been described as containing an IRES in their mRNAs. We can make the hypothesis that these mRNA families either contain IRESs that have not been identified yet, or are translated by another cap-independent mechanism such as m6A-induced ribosome engagement sites (MIRES) (Prats et al., 2020).

Neat1, as a stress-induced lncRNA, plays a role in many pathologies including cancer and ischemic diseases, thus its central role in the translational control of expression of genes involved in tissue revascularization and cell survival makes it a potential therapeutic target of great interest.

Lentivector plasmid sequences are available on Dryad. https://doi.org/10.5061/dryad.2330r1b and https://doi.org/10.5061/dryad.m0cfxpp75. The MS proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD024067.

1. 1. Froment C

   (2021) ProteomeXchange

   ID PXD024067. Long non-coding RNA Neat1 is a key translational regulator in hypoxia.

   http://proteomecentral.proteomexchange.org/cgi/GetDataset?ID=PXD024067
2. 1. Godet A
   2. Roussel E
   3. David FP
   4. Hantelys F
   5. Morfoisse F
   6. Alves J
   7. Pujol F
   8. Ader I
   9. Bertrand E
   10. Burlet-Schiltz O
   11. Froment C
   12. Henras A
   13. Vitali P
   14. Lacazette E
   15. Tatin F
   16. Garmy-Susini B

   (2022) Dryad Digital Repository

   Long non-coding RNA Neat1 and paraspeckle components are translational regulators in hypoxia.

   https://doi.org/10.5061/dryad.m0cfxpp75

1. 1. Hantelys F
   2. Godet A
   3. David F
   4. Tatin F
   5. Renaud-Gabardos E
   6. Pujol F
   7. Diallo L
   8. Ader I
   9. Ligat L
   10. Henras A
   11. Sato Y
   12. Parini A
   13. Lacazette E
   14. Garmy-Susini B
   15. Prats A

   (2020) Dryad Digital Repository

   Data from: Vasohibin1, a new IRES trans-acting factor for induction of (lymph)angiogenic factors in early hypoxia.

   https://doi.org/10.5061/dryad.2330r1b

1. 1. Ainaoui N
   2. Hantelys F
   3. Renaud-Gabardos E
   4. Bunel M
   5. Lopez F
   6. Pujol F
   7. Planes R
   8. Bahraoui E
   9. Pichereaux C
   10. Burlet-Schiltz O
   11. Parini A
   12. Garmy-Susini B
   13. Prats A-C

   (2015) Promoter-dependent translation controlled by p54nrb and hnrnpm during myoblast differentiation

   PLOS ONE 10:e0136466.

   https://doi.org/10.1371/journal.pone.0136466

   • PubMed
   • Google Scholar
2. 1. Bouyssié D
   2. Dubois M
   3. Nasso S
   4. Gonzalez de Peredo A
   5. Burlet-Schiltz O
   6. Aebersold R
   7. Monsarrat B

   (2015) MzDB: a file format using multiple indexing strategies for the efficient analysis of large LC-MS/MS and SWATH-MS data sets

   Molecular & Cellular Proteomics 14:771–781.

   https://doi.org/10.1074/mcp.O114.039115

   • PubMed
   • Google Scholar
3. 1. Bouyssié D
   2. Hesse A-M
   3. Mouton-Barbosa E
   4. Rompais M
   5. Macron C
   6. Carapito C
   7. Gonzalez de Peredo A
   8. Couté Y
   9. Dupierris V
   10. Burel A
   11. Menetrey J-P
   12. Kalaitzakis A
   13. Poisat J
   14. Romdhani A
   15. Burlet-Schiltz O
   16. Cianférani S
   17. Garin J
   18. Bruley C

   (2020) Proline: an efficient and user-friendly software suite for large-scale proteomics

   Bioinformatics 36:3148–3155.

   https://doi.org/10.1093/bioinformatics/btaa118

   • PubMed
   • Google Scholar
4. 1. Chen J
   2. Guo K
   3. Kastan MB

   (2012) Interactions of nucleolin and ribosomal protein L26 (RPL26) in translational control of human p53 mRNA

   Journal of Biological Chemistry 287:16467–16476.

   https://doi.org/10.1074/jbc.M112.349274

   • PubMed
   • Google Scholar
5. 1. Choudhry H
   2. Albukhari A
   3. Morotti M
   4. Haider S
   5. Moralli D
   6. Smythies J
   7. Schödel J
   8. Green CM
   9. Camps C
   10. Buffa F
   11. Ratcliffe P
   12. Ragoussis J
   13. Harris AL
   14. Mole DR

   (2015) Tumor hypoxia induces nuclear paraspeckle formation through HIF-2α dependent transcriptional activation of NEAT1 leading to cancer cell survival

   Oncogene 34:4482–4490.

   https://doi.org/10.1038/onc.2014.378

   • PubMed
   • Google Scholar
6. 1. Claycomb WC
   2. Lanson NA
   3. Stallworth BS
   4. Egeland DB
   5. Delcarpio JB
   6. Bahinski A
   7. Izzo NJ

   (1998) Hl-1 cells: a cardiac muscle cell line that contracts and retains phenotypic characteristics of the adult cardiomyocyte

   PNAS 95:2979–2984.

   https://doi.org/10.1073/pnas.95.6.2979

   • PubMed
   • Google Scholar
7. 1. Clemson CM
   2. Hutchinson JN
   3. Sara SA
   4. Ensminger AW
   5. Fox AH
   6. Chess A
   7. Lawrence JB

   (2009) An architectural role for a nuclear noncoding RNA: NEAT1 RNA is essential for the structure of paraspeckles

   Molecular Cell 33:717–726.

   https://doi.org/10.1016/j.molcel.2009.01.026

   • PubMed
   • Google Scholar
8. 1. Cobbold LC
   2. Spriggs KA
   3. Haines SJ
   4. Dobbyn HC
   5. Hayes C
   6. de Moor CH
   7. Lilley KS
   8. Bushell M
   9. Willis AE

   (2008) Identification of internal ribosome entry segment (IRES) -trans-acting factors for the Myc family of IRESs

   Molecular and Cellular Biology 28:40–49.

   https://doi.org/10.1128/MCB.01298-07

   • PubMed
   • Google Scholar
9. 1. Créancier L
   2. Morello D
   3. Mercier P
   4. Prats AC

   (2000) Fibroblast growth factor 2 internal ribosome entry site (IRES) activity ex vivo and in transgenic mice reveals a stringent tissue-specific regulation

   The Journal of Cell Biology 150:275–281.

   https://doi.org/10.1083/jcb.150.1.275

   • PubMed
   • Google Scholar
10. 1. Créancier L
    2. Mercier P
    3. Prats AC
    4. Morello D

    (2001) C-Myc internal ribosome entry site activity is developmentally controlled and subjected to a strong translational repression in adult transgenic mice

    Molecular and Cellular Biology 21:1833–1840.

    https://doi.org/10.1128/MCB.21.5.1833-1840.2001

    • PubMed
    • Google Scholar
11. 1. Fox AH
    2. Lam YW
    3. Leung AKL
    4. Lyon CE
    5. Andersen J
    6. Mann M
    7. Lamond AI

    (2002) Paraspeckles: a novel nuclear domain

    Current Biology 12:13–25.

    https://doi.org/10.1016/s0960-9822(01)00632-7

    • PubMed
    • Google Scholar
12. 1. Fox AH
    2. Bond CS
    3. Lamond AI

    (2005) P54Nrb forms a heterodimer with PSP1 that localizes to paraspeckles in an RNA-dependent manner

    Molecular Biology of the Cell 16:5304–5315.

    https://doi.org/10.1091/mbc.e05-06-0587

    • PubMed
    • Google Scholar
13. 1. Gidlöf O
    2. Bader K
    3. Celik S
    4. Grossi M
    5. Nakagawa S
    6. Hirose T
    7. Metzler B
    8. Olde B
    9. Erlinge D

    (2020) Inhibition of the long non-coding RNA NEAT1 protects cardiomyocytes from hypoxia in vitro via decreased pri-mirna processing

    Cell Death & Disease 11:677.

    https://doi.org/10.1038/s41419-020-02854-7

    • PubMed
    • Google Scholar
14. 1. Giraud S
    2. Greco A
    3. Brink M
    4. Diaz JJ
    5. Delafontaine P

    (2001) Translation initiation of the insulin-like growth factor I receptor mrna is mediated by an internal ribosome entry site

    The Journal of Biological Chemistry 276:5668–5675.

    https://doi.org/10.1074/jbc.M005928200

    • PubMed
    • Google Scholar
15. 1. Godet AC
    2. David F
    3. Hantelys F
    4. Tatin F
    5. Lacazette E
    6. Garmy-Susini B
    7. Prats AC

    (2019) Ires trans-acting factors, key actors of the stress response

    International Journal of Molecular Sciences 20:924.

    https://doi.org/10.3390/ijms20040924

    • PubMed
    • Google Scholar
16. 1. Hantelys F
    2. Godet A-C
    3. David F
    4. Tatin F
    5. Renaud-Gabardos E
    6. Pujol F
    7. Diallo LH
    8. Ader I
    9. Ligat L
    10. Henras AK
    11. Sato Y
    12. Parini A
    13. Lacazette E
    14. Garmy-Susini B
    15. Prats A-C

    (2019) Vasohibin1, a new mouse cardiomyocyte IRES trans-acting factor that regulates translation in early hypoxia

    eLife 8:e50094.

    https://doi.org/10.7554/eLife.50094

    • PubMed
    • Google Scholar
17. 1. Hertz MI
    2. Landry DM
    3. Willis AE
    4. Luo G
    5. Thompson SR

    (2013) Ribosomal protein S25 dependency reveals a common mechanism for diverse internal ribosome entry sites and ribosome shunting

    Molecular and Cellular Biology 33:1016–1026.

    https://doi.org/10.1128/MCB.00879-12

    • PubMed
    • Google Scholar
18. 1. Hirose T
    2. Virnicchi G
    3. Tanigawa A
    4. Naganuma T
    5. Li R
    6. Kimura H
    7. Yokoi T
    8. Nakagawa S
    9. Bénard M
    10. Fox AH
    11. Pierron G

    (2014) Neat1 long noncoding RNA regulates transcription via protein sequestration within subnuclear bodies

    Molecular Biology of the Cell 25:169–183.

    https://doi.org/10.1091/mbc.E13-09-0558

    • PubMed
    • Google Scholar
19. 1. Holcik M
    2. Sonenberg N

    (2005) Translational control in stress and apoptosis

    Nature Reviews. Molecular Cell Biology 6:318–327.

    https://doi.org/10.1038/nrm1618

    • PubMed
    • Google Scholar
20. 1. Hu CJ
    2. Wang LY
    3. Chodosh LA
    4. Keith B
    5. Simon MC

    (2003) Differential roles of hypoxia-inducible factor 1alpha (HIF-1alpha) and HIF-2alpha in hypoxic gene regulation

    Molecular and Cellular Biology 23:9361–9374.

    https://doi.org/10.1128/MCB.23.24.9361-9374.2003

    • PubMed
    • Google Scholar
21. 1. Huez I
    2. Créancier L
    3. Audigier S
    4. Gensac MC
    5. Prats AC
    6. Prats H

    (1998) Two independent internal ribosome entry sites are involved in translation initiation of vascular endothelial growth factor mrna

    Molecular and Cellular Biology 18:6178–6190.

    https://doi.org/10.1128/MCB.18.11.6178

    • PubMed
    • Google Scholar
22. 1. Imamura K
    2. Imamachi N
    3. Akizuki G
    4. Kumakura M
    5. Kawaguchi A
    6. Nagata K
    7. Kato A
    8. Kawaguchi Y
    9. Sato H
    10. Yoneda M
    11. Kai C
    12. Yada T
    13. Suzuki Y
    14. Yamada T
    15. Ozawa T
    16. Kaneki K
    17. Inoue T
    18. Kobayashi M
    19. Kodama T
    20. Wada Y
    21. Sekimizu K
    22. Akimitsu N

    (2014) Long noncoding RNA NEAT1-dependent SFPQ relocation from promoter region to paraspeckle mediates IL8 expression upon immune stimuli

    Molecular Cell 53:393–406.

    https://doi.org/10.1016/j.molcel.2014.01.009

    • PubMed
    • Google Scholar
23. 1. Kenneweg F
    2. Bang C
    3. Xiao K
    4. Boulanger CM
    5. Loyer X
    6. Mazlan S
    7. Schroen B
    8. Hermans-Beijnsberger S
    9. Foinquinos A
    10. Hirt MN
    11. Eschenhagen T
    12. Funcke S
    13. Stojanovic S
    14. Genschel C
    15. Schimmel K
    16. Just A
    17. Pfanne A
    18. Scherf K
    19. Dehmel S
    20. Raemon-Buettner SM
    21. Fiedler J
    22. Thum T

    (2019) Long noncoding RNA-enriched vesicles secreted by hypoxic cardiomyocytes drive cardiac fibrosis

    Molecular Therapy - Nucleic Acids 18:363–374.

    https://doi.org/10.1016/j.omtn.2019.09.003

    • PubMed
    • Google Scholar
24. 1. Koh MY
    2. Lemos R
    3. Liu X
    4. Powis G

    (2011) The hypoxia-associated factor switches cells from HIF-1α- to HIF-2α-dependent signaling promoting stem cell characteristics, aggressive tumor growth and invasion

    Cancer Research 71:4015–4027.

    https://doi.org/10.1158/0008-5472.CAN-10-4142

    • PubMed
    • Google Scholar
25. 1. Lampe S
    2. Kunze M
    3. Scholz A
    4. Brauß TF
    5. Winslow S
    6. Simm S
    7. Keller M
    8. Heidler J
    9. Wittig I
    10. Brüne B
    11. Schmid T

    (2018) Identification of the TXNIP IRES and characterization of the impact of regulatory IRES trans-acting factors

    Biochimica et Biophysica Acta - Gene Regulatory Mechanisms 1861:147–157.

    https://doi.org/10.1016/j.bbagrm.2018.01.010

    • PubMed
    • Google Scholar
26. 1. Lee M
    2. Sadowska A
    3. Bekere I
    4. Ho D
    5. Gully BS
    6. Lu Y
    7. Iyer KS
    8. Trewhella J
    9. Fox AH
    10. Bond CS

    (2015) The structure of human SFPQ reveals a coiled-coil mediated polymer essential for functional aggregation in gene regulation

    Nucleic Acids Research 43:3826–3840.

    https://doi.org/10.1093/nar/gkv156

    • PubMed
    • Google Scholar
27. 1. Martineau Y
    2. Le Bec C
    3. Monbrun L
    4. Allo V
    5. Chiu IM
    6. Danos O
    7. Moine H
    8. Prats H
    9. Prats AC

    (2004) Internal ribosome entry site structural motifs conserved among mammalian fibroblast growth factor 1 alternatively spliced mRNAs

    Molecular and Cellular Biology 24:7622–7635.

    https://doi.org/10.1128/MCB.24.17.7622-7635.2004

    • PubMed
    • Google Scholar
28. 1. Morfoisse F
    2. Kuchnio A
    3. Frainay C
    4. Gomez-Brouchet A
    5. Delisle M-B
    6. Marzi S
    7. Helfer A-C
    8. Hantelys F
    9. Pujol F
    10. Guillermet-Guibert J
    11. Bousquet C
    12. Dewerchin M
    13. Pyronnet S
    14. Prats A-C
    15. Carmeliet P
    16. Garmy-Susini B

    (2014) Hypoxia induces VEGF-C expression in metastatic tumor cells via a HIF-1α-independent translation-mediated mechanism

    Cell Reports 6:155–167.

    https://doi.org/10.1016/j.celrep.2013.12.011

    • PubMed
    • Google Scholar
29. 1. Morfoisse F
    2. Tatin F
    3. Hantelys F
    4. Adoue A
    5. Helfer A-C
    6. Cassant-Sourdy S
    7. Pujol F
    8. Gomez-Brouchet A
    9. Ligat L
    10. Lopez F
    11. Pyronnet S
    12. Courty J
    13. Guillermet-Guibert J
    14. Marzi S
    15. Schneider RJ
    16. Prats A-C
    17. Garmy-Susini BH

    (2016) Nucleolin promotes heat shock-associated translation of VEGF-D to promote tumor lymphangiogenesis

    Cancer Research 76:4394–4405.

    https://doi.org/10.1158/0008-5472.CAN-15-3140

    • PubMed
    • Google Scholar
30. 1. Naganuma T
    2. Nakagawa S
    3. Tanigawa A
    4. Sasaki YF
    5. Goshima N
    6. Hirose T

    (2012) Alternative 3’-end processing of long noncoding RNA initiates construction of nuclear paraspeckles

    The EMBO Journal 31:4020–4034.

    https://doi.org/10.1038/emboj.2012.251

    • PubMed
    • Google Scholar
31. 1. Nakagawa S
    2. Naganuma T
    3. Shioi G
    4. Hirose T

    (2011) Paraspeckles are subpopulation-specific nuclear bodies that are not essential in mice

    The Journal of Cell Biology 193:31–39.

    https://doi.org/10.1083/jcb.201011110

    • PubMed
    • Google Scholar
32. 1. Nanbru C
    2. Lafon I
    3. Audigier S
    4. Gensac MC
    5. Vagner S
    6. Huez G
    7. Prats AC

    (1997) Alternative translation of the proto-oncogene c-myc by an internal ribosome entry site

    The Journal of Biological Chemistry 272:32061–32066.

    https://doi.org/10.1074/jbc.272.51.32061

    • PubMed
    • Google Scholar
33. 1. Nanbru C
    2. Prats AC
    3. Droogmans L
    4. Defrance P
    5. Huez G
    6. Kruys V

    (2001) Translation of the human c-myc P0 tricistronic mRNA involves two independent internal ribosome entry sites

    Oncogene 20:4270–4280.

    https://doi.org/10.1038/sj.onc.1204548

    • PubMed
    • Google Scholar
34. 1. Passon DM
    2. Lee M
    3. Rackham O
    4. Stanley WA
    5. Sadowska A
    6. Filipovska A
    7. Fox AH
    8. Bond CS

    (2012) Structure of the heterodimer of human nono and paraspeckle protein component 1 and analysis of its role in subnuclear body formation

    PNAS 109:4846–4850.

    https://doi.org/10.1073/pnas.1120792109

    • PubMed
    • Google Scholar
35. 1. Pavao M
    2. Huang YH
    3. Hafer LJ
    4. Moreland RB
    5. Traish AM

    (2001) Immunodetection of nmt55/p54nrb isoforms in human breast cancer

    BMC Cancer 1:15.

    https://doi.org/10.1186/1471-2407-1-15

    • PubMed
    • Google Scholar
36. 1. Peddigari S
    2. Li PWL
    3. Rabe JL
    4. Martin SL

    (2013) HnRNPL and nucleolin bind LINE-1 RNA and function as host factors to modulate retrotransposition

    Nucleic Acids Research 41:575–585.

    https://doi.org/10.1093/nar/gks1075

    • PubMed
    • Google Scholar
37. 1. Pouysségur J
    2. Dayan F
    3. Mazure NM

    (2006) Hypoxia signalling in cancer and approaches to enforce tumour regression

    Nature 441:437–443.

    https://doi.org/10.1038/nature04871

    • PubMed
    • Google Scholar
38. 1. Prasanth KV
    2. Prasanth SG
    3. Xuan Z
    4. Hearn S
    5. Freier SM
    6. Bennett CF
    7. Zhang MQ
    8. Spector DL

    (2005) Regulating gene expression through RNA nuclear retention

    Cell 123:249–263.

    https://doi.org/10.1016/j.cell.2005.08.033

    • PubMed
    • Google Scholar
39. 1. Prats AC
    2. Van den Berghe L
    3. Rayssac A
    4. Ainaoui N
    5. Morfoisse F
    6. Pujol F
    7. Legonidec S
    8. Bikfalvi A
    9. Prats H
    10. Pyronnet S
    11. Garmy-Susini B

    (2013) CXCL4L1–fibstatin cooperation inhibits tumor angiogenesis, lymphangiogenesis and metastasis

    Microvascular Research 89:25–33.

    https://doi.org/10.1016/j.mvr.2013.05.005

    • PubMed
    • Google Scholar
40. 1. Prats AC
    2. David F
    3. Diallo LH
    4. Roussel E
    5. Tatin F
    6. Garmy-Susini B
    7. Lacazette E

    (2020) Circular RNA, the key for translation

    International Journal of Molecular Sciences 21:8591.

    https://doi.org/10.3390/ijms21228591

    • PubMed
    • Google Scholar
41. 1. Semler BL
    2. Waterman ML

    (2008) IRES-mediated pathways to polysomes: nuclear versus cytoplasmic routes

    Trends in Microbiology 16:1–5.

    https://doi.org/10.1016/j.tim.2007.11.001

    • PubMed
    • Google Scholar
42. 1. Sharathchandra A
    2. Lal R
    3. Khan D
    4. Das S

    (2012) Annexin A2 and PSF proteins interact with p53 IRES and regulate translation of p53 mRNA

    RNA Biology 9:1429–1439.

    https://doi.org/10.4161/rna.22707

    • PubMed
    • Google Scholar
43. 1. Shen W
    2. Liang XH
    3. Sun H
    4. De Hoyos CL
    5. Crooke ST

    (2017) Depletion of NEAT1 lncRNA attenuates nucleolar stress by releasing sequestered p54NRB and PSF to facilitate c-myc translation

    PLOS ONE 12:e0173494.

    https://doi.org/10.1371/journal.pone.0173494

    • PubMed
    • Google Scholar
44. 1. Shevchenko A
    2. Wilm M
    3. Vorm O
    4. Mann M

    (1996) Mass spectrometric sequencing of proteins silver-stained polyacrylamide gels

    Analytical Chemistry 68:850–858.

    https://doi.org/10.1021/ac950914h

    • PubMed
    • Google Scholar
45. 1. Shi Y
    2. Yang Y
    3. Hoang B
    4. Bardeleben C
    5. Holmes B
    6. Gera J
    7. Lichtenstein A

    (2016) Therapeutic potential of targeting IRES-dependent c-myc translation in multiple myeloma cells during ER stress

    Oncogene 35:1015–1024.

    https://doi.org/10.1038/onc.2015.156

    • PubMed
    • Google Scholar
46. 1. Shi Z
    2. Fujii K
    3. Kovary KM
    4. Genuth NR
    5. Röst HL
    6. Teruel MN
    7. Barna M

    (2017) Heterogeneous ribosomes preferentially translate distinct subpools of mRNAs genome-wide

    Molecular Cell 67:71–83.

    https://doi.org/10.1016/j.molcel.2017.05.021

    • PubMed
    • Google Scholar
47. 1. Stoneley M
    2. Subkhankulova T
    3. Le Quesne JP
    4. Coldwell MJ
    5. Jopling CL
    6. Belsham GJ
    7. Willis AE

    (2000) Analysis of the c-myc IRES; a potential role for cell-type specific trans-acting factors and the nuclear compartment

    Nucleic Acids Research 28:687–694.

    https://doi.org/10.1093/nar/28.3.687

    • PubMed
    • Google Scholar
48. 1. Sunwoo H
    2. Dinger ME
    3. Wilusz JE
    4. Amaral PP
    5. Mattick JS
    6. Spector DL

    (2009) Men epsilon/beta nuclear-retained non-coding RNAs are up-regulated upon muscle differentiation and are essential components of paraspeckles

    Genome Research 19:347–359.

    https://doi.org/10.1101/gr.087775.108

    • PubMed
    • Google Scholar
49. 1. Takagi M
    2. Absalon MJ
    3. McLure KG
    4. Kastan MB

    (2005) Regulation of p53 translation and induction after DNA damage by ribosomal protein L26 and nucleolin

    Cell 123:49–63.

    https://doi.org/10.1016/j.cell.2005.07.034

    • PubMed
    • Google Scholar
50. 1. Tatin F
    2. Renaud-Gabardos E
    3. Godet AC
    4. Hantelys F
    5. Pujol F
    6. Morfoisse F
    7. Calise D
    8. Viars F
    9. Valet P
    10. Masri B
    11. Prats AC
    12. Garmy-Susini B

    (2017) Apelin modulates pathological remodeling of lymphatic endothelium after myocardial infarction

    JCI Insight 2:e93887.

    https://doi.org/10.1172/jci.insight.93887

    • PubMed
    • Google Scholar
51. 1. Tsanov N
    2. Samacoits A
    3. Chouaib R
    4. Traboulsi AM
    5. Gostan T
    6. Weber C
    7. Zimmer C
    8. Zibara K
    9. Walter T
    10. Peter M
    11. Bertrand E
    12. Mueller F

    (2016) SmiFISH and FISH-quant-a flexible single RNA detection approach with super-resolution capability

    Nucleic Acids Research 44:e165.

    https://doi.org/10.1093/nar/gkw784

    • PubMed
    • Google Scholar
52. 1. Vagner S
    2. Gensac MC
    3. Maret A
    4. Bayard F
    5. Amalric F
    6. Prats H
    7. Prats AC

    (1995) Alternative translation of human fibroblast growth factor 2 mRNA occurs by internal entry of ribosomes

    Molecular and Cellular Biology 15:35–44.

    https://doi.org/10.1128/MCB.15.1.35

    • PubMed
    • Google Scholar
53. 1. Wilusz JE
    2. JnBaptiste CK
    3. Lu LY
    4. Kuhn CD
    5. Joshua-Tor L
    6. Sharp PA

    (2012) A triple helix stabilizes the 3’ ends of long noncoding RNAs that lack poly (a) tails

    Genes & Development 26:2392–2407.

    https://doi.org/10.1101/gad.204438.112

    • PubMed
    • Google Scholar
54. 1. Yamamoto K
    2. Goyama S
    3. Asada S
    4. Fujino T
    5. Yonezawa T
    6. Sato N
    7. Takeda R
    8. Tsuchiya A
    9. Fukuyama T
    10. Tanaka Y
    11. Yokoyama A
    12. Toya H
    13. Kon A
    14. Nannya Y
    15. Onoguchi-Mizutani R
    16. Nakagawa S
    17. Hirose T
    18. Ogawa S
    19. Akimitsu N
    20. Kitamura T

    (2021) A histone modifier, ASXL1, interacts with nono and is involved in paraspeckle formation in hematopoietic cells

    Cell Reports 36:109576.

    https://doi.org/10.1016/j.celrep.2021.109576

    • PubMed
    • Google Scholar
55. 1. Yan H
    2. Liang H
    3. Liu L
    4. Chen D
    5. Zhang Q

    (2019) Long noncoding RNA NEAT1 sponges mir‑125a‑5p to suppress cardiomyocyte apoptosis via BCL2L12

    Molecular Medicine Reports 19:4468–4474.

    https://doi.org/10.3892/mmr.2019.10095

    • PubMed
    • Google Scholar
56. 1. Yan H
    2. Wang Z
    3. Sun Y
    4. Hu L
    5. Bu P

    (2021) Cytoplasmic NEAT1 suppresses AML stem cell self-renewal and leukemogenesis through inactivation of wnt signaling

    Advanced Science 8:e2100914.

    https://doi.org/10.1002/advs.202100914

    • PubMed
    • Google Scholar
57. 1. Yang C
    2. Li Z
    3. Li Y
    4. Xu R
    5. Wang Y
    6. Tian Y
    7. Chen W

    (2017) Long non-coding RNA NEAT1 overexpression is associated with poor prognosis in cancer patients: a systematic review and meta-analysis

    Oncotarget 8:2672–2680.

    https://doi.org/10.18632/oncotarget.13737

    • PubMed
    • Google Scholar
58. 1. Yang Y
    2. Wang C
    3. Zhao K
    4. Zhang G
    5. Wang D
    6. Mei Y

    (2018) TRMP, a p53-inducible long noncoding RNA, regulates G1/S cell cycle progression by modulating IRES-dependent p27 translation

    Cell Death & Disease 9:886.

    https://doi.org/10.1038/s41419-018-0884-3

    • PubMed
    • Google Scholar
59. 1. Ylä-Herttuala S
    2. Baker AH

    (2017) Cardiovascular gene therapy: past, present, and future

    Molecular Therapy 25:1095–1106.

    https://doi.org/10.1016/j.ymthe.2017.03.027

    • PubMed
    • Google Scholar
60. 1. Zhang Y
    2. Wu D
    3. Wang D

    (2020) Long non-coding RNA ARAP1-AS1 promotes tumorigenesis and metastasis through facilitating proto-oncogene c-myc translation via dissociating PSF/PTB dimer in cervical cancer

    Cancer Medicine 9:1855–1866.

    https://doi.org/10.1002/cam4.2860

    • PubMed
    • Google Scholar

Decision letter after peer review:

Thank you for submitting your article "Long non-coding RNA Neat1 is a key translational regulator in hypoxia" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and James Manley as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

The paper reports highly novel data that connect nuclear paraspeckles and lncRNA to IRES activity. However, many of the experiments are incomplete, and some of the conclusions are unjustified. Technical problems. such as missing loading controls must be rectified. The full reviewers' reports are included. They include a long list of criticisms and constructive suggestions for improvements.

Essential revisions:

1. Depletion of NEAT1 by control gapmers already exerts stress on the cardiomyocytes. Can the authors be sure that observed changes in IRES activities were not caused by higher levels of cell stress? Is eIF2alpha phosphorylated in the gapmer experiments (Figure 2)? If this is the case, it could explain why only a few differences were noted in Figure 4D.

Relevant to this point, you silence the long non-coding RNA, but transfection per se causes stress to cell. It seems that a different approach instead of silencing should be used since stress response is the question that needs to be answered. Perhaps a stable cell line or CRISPR KO can be used instead? Otherwise, the authors need to explain why the cell stress induced by transfection is not relevant here.

2. While NEAT1 abundance has an effect on IRES, the NEAT-ITAFs do not. The authors should check the effects of DDX17, Eno3, and Hspa2 that were enriched in the cytoplasm during hypoxia (Figure 4E).

3. Depletion of NEAT1 or NEAT1-ITAFs has always more effect on endogenous FGF1 protein abundance than on IRES activity. Is the half-life of FGF1 affected during the siRNA and gapmer treatments?

4. You need to provide a more direct cause-effect relationship between Neat1 and the proteins claimed to be ITAFs for the FGF1 IRES activity.

5. Validate the idea that paraspeckles are platforms for IRESome formation as this would imply the IRES needs to transit through the paraspeckle to gain full activity.

6. Validate some of the MS-IP data where 2013 proteins interacti with p54nrb.

7. There is an issue with the timing of the response, inasmuch as the amount of Neat1 induction of the FGF IRES and paraspeckle activity does not seem to align. Therefore, how can the conclusion be drawn that Neat1 is absolutely crucial?

8. Only one cell line was used in these studies. The authors make the point in the Discussion that IRESs can behave differently in different cell lines, which raises the question regarding how it can be known that Neat1 is generally important.

9. In the hypoxia studies, typically the rate of canonical protein synthesis is significantly reduced, mTOR is inactivated and eIF2α is phosphorylated. Does this occur in HL-1 cardiomyocytes under their conditions of early hypoxia and if so is mTOR inhibition or eIF2 phosphorylation necessary for the effects that they see? In other words, are the novel data presented connected to established effects of hypoxia on protein synthesis?

10. According to the data in Figure 1 and supplemental Figure 1, there is a better correlation with the formation of paraspeckles and greater expression under hypoxia of the longer isoform, NEAT1_2. You should explain the reason for focusing on just NEAT1, the shorter isoform, on the effect of the regulation of FGF1 IRES activity in Figure 2?

11. The mRNAs encoding the major (lymph)angiogenic growth factors are fibroblast growth factor (FGF) and vascular endothelial growth factor (VEGF). You should evaluate the effect of Neat1 on the regulation of more than just one IRES. The same experiments shown in Figures 1, 2, and 3 should be included for VEGF, especially since the results presently cannot be extrapolated to other IRES elements, and because recent studies have shown that FGF can act synergistically with VEGF.

Reviewer #1 (Recommendations for the authors):

The manuscript investigates an interesting topic regarding IRES regulation by the Neat1 lncRNA and several ITAFs. General conclusion regarding paraspeckles being sites of IRESome formation appear premature as they are supported by correlative information. It is difficult to conclude direct cause-effect relationships on a quick-response regulatory mechanism such as translation through the use of gene knockdowns. The reason being that it takes several days for appropriate knockdown levels of RNA or protein to be reached – making it difficult to eliminate the possibility that secondary/pleiotropic effects are responsible for, or significantly contributing to, the observed physiological effect.

I think experiments are required that (i) provide a more direct cause-effect relationship between Neat1 and the proteins claimed to be ITAFs for the FGF1 IRES activity, (ii) validate the idea that paraspeckles are platforms for IRESome formation as this would imply the IRES needs to transit through the paraspeckle to gain full activity, and (iii) validate the MS-IP data where 2013 proteins interacting with p54nrb.

Specific:

1. Section entitled: "FGF1 IRES activation during hypoxia correlates with paraspeckle formation and with Neat1 induction."

– Measurements from the RLuc/FLuc reporter are made 4, 8, 24 after induction of hypoxia. Please show individual RLuc and FLuc measurements, not the ratios. Is eIF2-α phosphorylated under the experimental conditions? Under hypoxic conditions, are the same isoforms of Neat1 produced to the same ratios? Are new isoforms seen? A Northern blot would provide insight into whether the architecture of this lncRNA changes.

– Last sentence of this paragraph reads "This correlation fits better with Neat1_2 than with the total Neat1, suggesting a link between expression of the long Neat1 isoform and IRES activation." Can correlation of Neat1 expression and IRES activation be used to imply a link between the two?

2. Section entitled: "LncRNA Neat1 knock-down drastically affects the FGF1 IRES activity and endogenous FGF1 expression."

The authors deplete Neat1 using gap-mers, I could not find how long a treatment was used to achieve the 3-fold knockdown (I'm assuming at least 2 days). The long time period in these experiments likely makes it difficult to conclude that reductions in Neat are directly responsible for the effects on FGF1 IRES activity or FGF1 protein levels. How selective is the reported effect for the FGF1 5' leader? Hence, I think it is premature to conclude that "Neat1 regulates FGF1 expression, and acts as an ITAF of the FGF1 IRES" (last sentence of paragraph) since an ITAF would imply a direct interaction (which is not shown in this MS).

3. Section entitled "Paraspeckle proteins P54nrb and PSCP1, but not SFPQ, are ITAFs of the FGF1 IRES." The authors assess three major components of paraspeckles (the SFPQ, p54nrb and PSPC1 proteins) for a potential role in IRES activity. Since knockdowns are used to reduce protein levels and this takes several days to achieve, how can effects on IRES activity or FGF1 protein levels be directly link to the protein activity targeted for knockdown versus a secondary effect arising due to target knockdown? How selective is the reported effect for the FGF1 5' leader? Hence, I think its difficult to conclude that "The ability of three paraspeckle components, Neat1, p54nrb and PSPC1 to regulate the FGF1 IRES activity led us to the hypothesis that the paraspeckle might be involved in the control of IRES-dependent translation. ". Can direct interaction be shown and can this, in turn, be linked to a translational response.

4. Section entitled "P54nrb interactome in normoxic and in hypoxic cardiomyocytes". The authors write "The moderate effect of p54nrb or PSPC1 depletion on FGF1 IRES activity suggested that other proteins are involved". Maybe the moderate effect is due to the poor knockdown achieved with the siRNAs (Figure 3 – Suppl Figure 1; 50% for p54nrb and 25% for PSPC1)? The IP performed with HA-tagged p54nrb pulled down 2013 proteins if I understand this section correctly. Can the authors explain why so many proteins are being pulled down? The supplemental table provided with all the raw data is not reader friendly and the column labeling difficult to interpret for someone like me who is not familiar with seeing data in this format. The identification of some ITAFs in this dataset is taken as evidence that these ITAFs interact with p54nrb in paraspeckles. What independent validation has been undertaken to show that these are associated with p54nrb? How many proteins that are not ITAFs are pulled down by p54nrb and enriched when comparing hypoxic versus normoxic extracts? In Figure 4D and 4E, the identity of all proteins enriched under the conditions shown should be shown in the upper rectangles.

5. Section entitled "Neat1 is the key activator of (lymph)angiogenic and cardioprotective factor mRNA IRESs." In this section, the authors test the response of other IRESes to knockdown of p54, PSPC1, and Neat1 and depending on the response, stratify the IRESes into different regulons based on IRESome composition. Whereas, the experiments do allow the authors to classify the response of the different IRESes, I don't think it provides insight into IRESome composition since the experiments to not show that any of the three tested molecules directly associate with the IRES under study.

Reviewer #2 (Recommendations for the authors):

1. There is an issue with timing of the response, the amount of Neat1 induction of the FGF IRES and paraspeckle activity that do not seem to align. Therefore, how can the conclusion be drawn that Neat1 is absolutely crucial.

2. The authors silence the long non-coding RNA but their transfection per se causes stress to cell. It seems that a different approach instead of silencing should be used since stress response is the question that needs to be answered. Perhaps a stable cell line or CRISPR KO can be used instead? Otherwise, the authors need to explain why the cell stress induced by transfection is not relevant here.

3. Only one cell line was used in these studies. The authors make the point in the Discussion that IRESs can behave differently in different cell lines, which raises the question regarding how it can be known that Neat1 is generally important.

4. How did the authors confirm HL-1 cardiomyocytes were indeed under hypoxia? Which is the hypoxia control? This is not apparent.

5. In the hypoxia studies, typically the rate of canonical protein synthesis is significantly reduced, mTOR is inactivated and eIF2α is phosphorylated. Does this occur in HL-1 cardiomyocytes under their conditions of early hypoxia and if so is mTOR inhibition or eIF2 phosphorylation necessary for the effects that they see? In other words, are the novel data presented connected to established effects of hypoxia on protein synthesis?

6. According to their data (Figure 1 and supplemental Figure 1), there is a better correlation with the formation of paraspeckles and greater expression under hypoxia of the longer isoform, NEAT1_2. Can the authors explain the reason for focusing on just NEAT1, the shorter isoform, on the effect of the regulation of FGF1 IRES activity in Figure 2?

7. The mRNAs encoding the major (lymph)angiogenic growth factors are fibroblast growth factor (FGF) and vascular endothelial growth factor (VEGF). The authors should evaluate the effect of Neat1 on the regulation of more than just one IRES. The same experiments shown in Figures 1, 2, and 3 should be included for VEGF, especially since the results presently cannot be extrapolated to other IRES elements, and because recent studies have shown that FGF can act synergistically with VEGF.

8. If possible, the authors should provide immunoprecipitation and immunoblot analysis validation of their proteomic data.

9. Is the interaction between nucleolin and ribosomal protein Rps2 with p54 direct or indirect? The graphical abstract shows a direct interaction but there is no confirming evidence for this. Also, does this potential interaction only occur under hypoxia?

10. Figures 1-3 need a protein loading control, perhaps actin.

11. Can the authors show that levels of FGF1 protein after SFPQ knockdown are unchanged?

Reviewer #3 (Recommendations for the authors):

My main concern is that depletion of NEAT1 by control gapmers already exerts stress on the cardiomyocytes. Can the authors be sure that observed changes in IRES activities were not caused by a higher levels of cell stress? Is eIF2alpha phosphorylated in the gapmer experiments (Figure 2)? If this is the case, it could explain why only few differences were noted in Figure 4D.

1. While NEAT1 abundance has an effect on IRES, the NEAT-ITAFs do not. The authors should check effects of DDX17, Eno3 and Hspa2 that were enriched in the cytoplasm during hypoxia (Figure 4E).

2. All figures displaying FISH results are impossible to inspect. They need to be bigger.

3. Depletion of NEAT1 or NEAT1-ITAFs has always more effect on endogenous FGF1 protein abundance that on IRES activity. Is the half-life of FGF1 affected during the siRNA and gapmer treatments?

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Long non-coding RNA Neat1 and paraspeckle components are translational regulators in hypoxia" for further consideration by eLife. Your revised article has been evaluated by James Manley (Senior Editor), and the Reviewing Editor and reviewers of the original manuscript.

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

As detailed by reviewer #1 there are omissions in the text of the authors' responses listed in the rebuttal letter. Most importantly, since many of the results are correlative the conclusions are overstated. As requested by reviewer # 1, since the authors do not show a direct association between Neat1 and the ITAFs investigated herein, the authors ought to change or remove Figure 4A which claims that Neat1 and p54nrb interact with the FGF1 IRES.

Reviewer #1 (Recommendations for the authors):

The authors have provided new information which addresses some of my previous concerns. However, my main concern remains in that there is much correlative data from which direct cause-effect relationships are inferred.

The authors indicate that there was a decrease in the number of paraspeckles in cells in which Neat1 was knocked down, as well as a 2-fold reduction in IRES activity. However, in new data presented, the authors show that eIF2alpha is phosphorylated upon Neat1_2 KO (Figure 2, Suppl 4) and in their rebuttal write that this could explain the small difference of IRES activity between normoxia and hypoxia generated by the KO. I, therefore, don't understand how the authors conclude on line 232 of the MS that the Neat1 knockdown results suggest "that Neat1 regulates FGF1 mRNA translation, directly or indirectly". The same comments holds true for the data regarding changes in the polysomal mRNA context following Neat1 knockdown (Figure 8). Are these due to elevated eIF2 α phosphorylation and not necessarily due to the absence of Neat1

The authors use smiFISH to probe the localization of Neat 1 and FGF1 IRES and find a ~ 2-fold increase in overlapping signals for the FGF1 IRES and Neat1 in hypoxic versus normal cells. (Figure 3B). I'm not sure how this led to the "hypothesis that the paraspeckle might be involved in the control of IRES-dependent translation." For this conclusion to be made, wouldn't one have to show altered IRES activity after versus before having encountered the Neat1 speckles?

Since the authors do not show a direct association between Neat1 and the ITAFs investigated herein, I would ask that they change or remove Figure 4A which appears to show Neat1 and p54nrb interacting with the FGF1 IRES.

Essential revisions:

1. Depletion of NEAT1 by control gapmers already exerts stress on the cardiomyocytes. Can the authors be sure that observed changes in IRES activities were not caused by higher levels of cell stress?

As written in page 5, 1^st paragraph, in the experiments with gapmer treatment the number of paraspeckles was already high in normoxia (almost 5 foci per cell). This may suggest suggesting that cells were already stressed by the gapmer treatment, before being submitted to hypoxia. Alternatively this higher number may result from the higher sensitivity of the smiFISH method compared to FISH in Figure 1.

Is eIF2alpha phosphorylated in the gapmer experiments (Figure 2)? If this is the case, it could explain why only a few differences were noted in Figure 4D.

eIF2a phosphorylation has been analyzed after gapmer treatment. (Figure 2, figure supplement 4). Our hypothesis to explain the effect of Neat1 depletion in normoxia is indeed that cells are stressed by the gapmer treatment. We performed the experiment (Figure 2, figure supplement 4): eIF2a phosphorylation slightly increases upon Neat1_2 gapmer treatment, but not using the control gapmer. This could explain the small difference of IRES activity between normoxia and hypoxia generated by the knock-down.

Relevant to this point, you silence the long non-coding RNA, but transfection per se causes stress to cell. It seems that a different approach instead of silencing should be used since stress response is the question that needs to be answered. Perhaps a stable cell line or CRISPR KO can be used instead?

Indeed the stress induced by transfection is an important limitation and it would have been interesting but this experiment. We tried to generate Neat1 CRISPR-based induction and silencing using lentivectors (with Cas9-VP160 and Cas-9-Krab, respectively, but we were unfortunately not successful). One gRNA was not sufficient, we obtain very low titers of the lentivector and in addition it was impossible in our hands to construct a lentivector coding several guides with the mCas9.

Otherwise, the authors need to explain why the cell stress induced by transfection is not relevant here.

I agree with the reviewer that the stress generated by transient transfection is a limitation of our approach. However the gapmer was the best tool to deplete Neat1, a nuclear ncRNA. As you can see in the Figure 1—figure supplement 3, we constructed lentivectors coding artificial miRNAs against Neat1 to deplete Neat1 in 67NR cells. This was done before the experiments with HL-1. We gave up this approach as it was very difficult to have an efficient knock-down. Gapmer provided better data. The stress due to the gapmer transfection can explain why we have no difference between normoxia and hypoxia, whereas we should have less effect in normoxia. Nevertheless the effect of Neat1 depletion on the activities of all IRESs except for EMCV IRES, remains a very convincing result by showing the role of Neat1 in IRES-dependent translation.

2. While NEAT1 abundance has an effect on IRES, the NEAT-ITAFs do not.

As mentioned in the text, the depletion of paraspeckle proteins had different effects depending on the IRES and we make the hypothesis that IRESs could be regrouped in different “regulons” in normoxia and in hypoxia (Figure 7D). We propose that IRESome composition varies for each IRES.

The authors should check the effects of DDX17, Eno3, and Hspa2 that were enriched in the cytoplasm during hypoxia (Figure 4E).

The knock-down of DDX17 has been performed. Unexpectedly, DDX17 knock-down has a slight positive effect on IRES activity in normoxia. We do not wish to publish these data here because the role of helicases is being addressed in our laboratory and will constitute another publication. However we include these data in the present letter.

Author response image 1

Download asset Open asset

3. Depletion of NEAT1 or NEAT1-ITAFs has always more effect on endogenous FGF1 protein abundance than on IRES activity. Is the half-life of FGF1 affected during the siRNA and gapmer treatments?

The half-life of FGF1 has been analyzed and does not vary upon gapmer treatment (Figure 2—figure supplement 5-6). It is superior to 24 hr either without treatment, or with gapmer treatment, or with siRNA treatment (control and target-specific). A hypothesis to explain a more important effect on endogenous FGF1 mRNA than bicistronic mRNA translation may be that the IRES of the endogenous FGF1 mRNA is recruited more efficiently to the paraspeckle than the IRES present in the bicistronic mRNA, whose structure may be altered due to the presence of the upstream cistron. However, if one looks at the Western quantification (normalized to total proteins), the difference is not so big: upon Neat1 depletion endogenous FGF1 protein is 34%, while IRES activity is 50% (Figure 2C and 2D). Upon p54 or PSPC1 depletion, endogenous FGF1 is 61% or 82% while IRES activity is 70% or 80%, respectively (Figure 4E and 4F).

4. You need to provide a more direct cause-effect relationship between Neat1 and the proteins claimed to be ITAFs for the FGF1 IRES activity.

The direct interaction of at least p54 and SFPQ with Neat1 has been established previously (for example see: Yamazaki et al. (2018). Functional Domains of NEAT1 Architectural lncRNA Induce Paraspeckle Assembly through Phase Separation. Molecular cell, 70(6), 1038–1053.e7. https://doi.org/10.1016/j.molcel.2018.05.019). The three DBHS proteins (p54, SFPQ and PSPC1) are known to function in heteroduplex (possible in vitro but not physiologically significant).

Furthermore, we show by smi-FISH in this revised version that Neat1 interacts with the IRES-containing mRNA (Figure 3).

As regards the other proteins, it is not necessary that interact directly with Neat1 to function as ITAFs, but we can propose that they are in the RNP complex assembled to the IRES in the paraspeckle (as they interact with p54^nrb).

5. Validate the idea that paraspeckles are platforms for IRESome formation as this would imply the IRES needs to transit through the paraspeckle to gain full activity.

To validate the idea that the paraspeckle is a platform for IRESome assembly we performed an experiment of smiFISH (single molecule inexpensive FISH). This method using unlabeled primary probes and a fluorescently labelled secondary detector oligonucleotide allows to detect single RNA molecules (Tsanov et al. Nucleic Acids Res. 2016 Dec 15; 44(22): e165. 10.1093/nar/gkw784). The results show that the bicistronic mRNA containing the IRES is significantly co-localized with Neat1, which indicates its recruitment in the paraspeckle (Figure 3). Furthermore, the presence of Neat1 in the cytoplasm, observed in our data (Figure 1, Supplement 2, and Figure 3), as well as its presence in hypoxic polysomes shown in our previous paper (Hantelys et al., eLife 2019), support the hypothesis of an export of this complex (including Neat) out of the nucleus.

6. Validate some of the MS-IP data where 2013 proteins interacti with p54nrb.

The interaction of nucleolin with p54^nrb has been validated by co-immunoprecipitation. (Figure 5—figure supplement 2).

A list of proteins present in the paraspeckle complex with p54nrb has been reported previously, in addition to the main paraspeckle proteins SFPQ and PSPC1 known to interact with p54 (Both of them are found in our MS-IP)(Figure 5—figure supplement 1, and MS data in Supplement file 7). Among 40 paraspeckle proteins listed in previous reports (Naganuma et al., EMBO J, 2012 and Yamamoto et al 2021), 22 have been found interacting with p54nrb in our SM-IP (supplementary file 7). Furthermore, among these 22 proteins, 6 have been described as ITAFs: FUS, hnRNPA1, hnRNPK, hnRNPM, p54nrb/NONO, SFPQ/PSF (Godet et al., IJMS 2018). This analysis has been added page 6, 1^st paragraph, and (Table 1). In our view it is significant to have found more than half of the main paraspeckle proteins in our MS-IP.

7. There is an issue with the timing of the response, inasmuch as the amount of Neat1 induction of the FGF IRES and paraspeckle activity does not seem to align. Therefore, how can the conclusion be drawn that Neat1 is absolutely crucial?

We agree that the correlation between IRES activation and Neat1 induction is not concordant (Figure 1), but it is fully concordant with Neat1_2: the IRES shows a peak of activation at 4 hr, whereas Neat1_2 induction and its presence in paraspeckles also exhibit a peak at 4 hr (observed Figure 1 with both RT qPCR and FISH). This correlation favors the hypothesis of an important role for Neat1_2, although our data with Neat1_2 gapmer had only a small impact on IRES activity (Figure 2 figure supplement 3).

8. Only one cell line was used in these studies. The authors make the point in the Discussion that IRESs can behave differently in different cell lines, which raises the question regarding how it can be known that Neat1 is generally important.

The experiments have been performed with another cell line, 67NR (mouse breast tumor Figure 1—figure supplement 3). Data show that this cell line is more resistant to hypoxia: Neat1 is not induced whereas Neat1_2 is strongly induced after 24 hr of hypoxia. This time, 24 hr, also corresponds to the IRES activation. Furthermore, in 67NR, we performed Neat1 and Neat1_2 knock-down: the data show that the role of Neat1 in the control of IRES-dependent translation is not limited to HL-1 cells.

9. In the hypoxia studies, typically the rate of canonical protein synthesis is significantly reduced, mTOR is inactivated and eIF2α is phosphorylated. Does this occur in HL-1 cardiomyocytes under their conditions of early hypoxia and if so is mTOR inhibition or eIF2 phosphorylation necessary for the effects that they see? In other words, are the novel data presented connected to established effects of hypoxia on protein synthesis?

4E-BP1 and eIF2a phosphorylation have been analyzed in these conditions of early hypoxia in HL-1 and published in our previous paper in eLife (Hantelys et al. 2019, https://elifesciences.org/articles/50094 Figure 2 and Figure 2 Supplement 4). Clearly, no change in 4E-BP phosphorylation is observed but eIF2a is phosphorylated in these conditions. We also showed in the paper by Hantelys et al. that global protein synthesis is decreased (P/M ratio, Figure 2).

10. According to the data in Figure 1 and supplemental Figure 1, there is a better correlation with the formation of paraspeckles and greater expression under hypoxia of the longer isoform, NEAT1_2. You should explain the reason for focusing on just NEAT1, the shorter isoform, on the effect of the regulation of FGF1 IRES activity in Figure 2?

The knock-down of Neat1_2 has been performed and shows a decrease of FGF1 IRES activity (Figure 2figure supplement 3). Unfortunately, it is not possible to knock-down only Neat1-1 as its sequence is entirely comprised in that of Neat1_2 (Figure 1E). These data, in addition to the results provided by fluidigm (Figure 8) and knock-down in 67NR cells (Figure 1—figure supplement 3) indicate that Neat1_2 is involved in the regulation. However, the knock-down of both isoforms (Neat1) has a more important effect on IRES activity, suggesting that both isoforms play a role. Due to the data with Neat1_2 gapmer (Figure 2—figure supplement 3) we have moderated the conclusion about the key role of Neat1_2 that we were not able to demonstrate.

11. The mRNAs encoding the major (lymph)angiogenic growth factors are fibroblast growth factor (FGF) and vascular endothelial growth factor (VEGF). You should evaluate the effect of Neat1 on the regulation of more than just one IRES. The same experiments shown in Figures 1, 2, and 3 should be included for VEGF, especially since the results presently cannot be extrapolated to other IRES elements, and because recent studies have shown that FGF can act synergistically with VEGF.

IRES activation in early hypoxia shown in Figure 1 has been published for several other IRESs of the FGF and VEGF families, in the same cell model and with the same bicistronic lentivectors, in our previous paper (Hantelys et al., eLife 2019).

Neat1, p54^nrb and PSPC1 knock-down shown in Figure 2 and 4 for FGF1 IRES have been assessed with the IRESs of FGF2, VEGFA, VEGFC, VEGFD, IGF1R, c-myc and EMCV in Figure 7 (Figure 6 in the first submission). Data show that Neat1 knock-down affects all IRESs except for EMCV, while knock-down of p54^nrb or PSPC1 only affects groups of IRESs.

Reviewer #1 (Recommendations for the authors):

The manuscript investigates an interesting topic regarding IRES regulation by the Neat1 lncRNA and several ITAFs. General conclusion regarding paraspeckles being sites of IRESome formation appear premature as they are supported by correlative information. It is difficult to conclude direct cause-effect relationships on a quick-response regulatory mechanism such as translation through the use of gene knockdowns. The reason being that it takes several days for appropriate knockdown levels of RNA or protein to be reached – making it difficult to eliminate the possibility that secondary/pleiotropic effects are responsible for, or significantly contributing to, the observed physiological effect.

I think experiments are required that (i) provide a more direct cause-effect relationship between Neat1 and the proteins claimed to be ITAFs for the FGF1 IRES activity, (ii) validate the idea that paraspeckles are platforms for IRESome formation as this would imply the IRES needs to transit through the paraspeckle to gain full activity, and (iii) validate the MS-IP data where 2013 proteins interacting with p54nrb.

To validate the idea that the paraspeckle is a platform for IRESome we performed an experiment of smiFISH (see point 5 above). The results show that the bicistronic mRNA containing the IRES is significantly co-localized with Neat1, which indicates its recruitment in the paraspeckle, supporting the hypothesis that paraspeckles are platforms of IRESome formation (Figure 3).

Also, the interaction of nucleolin with p54^nrb has been validated by co-immunoprecipitation (Figure 5—figure supplement 2).

A list of proteins present in the paraspeckle complex with p54^nrb has been reported previously, in addition to the main paraspeckle proteins SFPQ and PSPC1 known to interact with p54 (Both of them are found in our MS-IP) (Figure 5—figure supplement 1, and MS data in Supplement file 7). Among 40 paraspeckle proteins listed in previous reports (Naganuma et al., EMBO J, 2012 and Yamamoto et al. 2021), 22 have been found interacting with p54^nrb in our SM-IP (supplementary file 7). Furthermore, among these 22 proteins, 6 have been described as ITAFs: FUS, hnRNPA1, hnRNPK, hnRNPM, p54^nrb/NONO, SFPQ/PSF (Godet et al., IJMS 2018). This analysis has been added page 12 and Table 1. In our view it is significant to have found more than half of the main paraspeckle proteins in our MSIP.

Specific:

1. Section entitled: "FGF1 IRES activation during hypoxia correlates with paraspeckle formation and with Neat1 induction."

– Measurements from the RLuc/FLuc reporter are made 4, 8, 24 after induction of hypoxia. Please show individual RLuc and FLuc measurements, not the ratios. Is eIF2-α phosphorylated under the experimental conditions? Under hypoxic conditions, are the same isoforms of Neat1 produced to the same ratios? Are new isoforms seen? A Northern blot would provide insight into whether the architecture of this lncRNA changes.

Indeed the table with LucR and LucF activities of Figure 1 data had been forgotten in the first submission. It is a mistake. The LucR and LucF values are now presented in Figure 1 Supplement 1 and new Supplementary file 1. eIF2a is already phosphorylated after 4 hours of hypoxia (published in our previous paper Hantelys et al., 2019, Figure 2 Supplement 4). Interestingly, LucR activity increases at 4 hr, and decreases only after 24 hr of hypoxia. This suggests an effect of the IRES on translation of the first cistron, as observed previously (Conte et al., PLoS One. 2008 Aug 27;3(8):e3078. doi: 10.1371/journal.pone.0003078.)

The experiments presented in Figure 1 were done in parallel to that published in 2019.

Regarding Neat1 isoforms, we can see in Figure 1D that the global Neat1 (two isoforms) induction is different from that of Neat 1-2 isoform. It is impossible to detect Neat1-1 shorter isoform separately by RT PCR as its sequence is contained in that of Neat1_2. The paraspeckle quantification confirms that difference in Figure 1F-K (FISH probing either Neat1 or Neat1_2): Neat1_2 shows a peak at 4 hr of hypoxia (as for RTqPCR), whereas for Neat1 reaches a plateau at 4 hr. This indicates a variation of the isoform ratio.

– Last sentence of this paragraph reads "This correlation fits better with Neat1_2 than with the total Neat1, suggesting a link between expression of the long Neat1 isoform and IRES activation." Can correlation of Neat1 expression and IRES activation be used to imply a link between the two?

We agree that a correlation is not a link but this observation incited us to look for a link, as p54 had been identified as an ITAF in our previous paper (Ainaoui et al., PlOS ONE 2015). The sentence has been changed (page 4, penultimate paragraph).

2. Section entitled: "LncRNA Neat1 knock-down drastically affects the FGF1 IRES activity and endogenous FGF1 expression."

The authors deplete Neat1 using gap-mers, I could not find how long a treatment was used to achieve the 3-fold knockdown (I'm assuming at least 2 days). The long time period in these experiments likely makes it difficult to conclude that reductions in Neat are directly responsible for the effects on FGF1 IRES activity or FGF1 protein levels. How selective is the reported effect for the FGF1 5' leader? Hence, I think it is premature to conclude that "Neat1 regulates FGF1 expression, and acts as an ITAF of the FGF1 IRES" (last sentence of paragraph) since an ITAF would imply a direct interaction (which is not shown in this MS).

The times of gapmer treatment (48 hr for Neat 1 and 72 hr for Neat1_2 knock-down) are indicated in the “Cell transfection” section of Material and methods. However it has been added in the figure legends (Figure 2 and Figure 2 supplement 1). It is indeed premature to conclude that Neat1 is an ITAF at this stage. However the definition of ITAF does not mean that there is a direct interaction with the IRES: by example the ncRNA TMRP acts as a negative IRES trans-acting factor by competing for PTB binding and preventing it to bind to the IRES (Godet et al., IJMS 2019, Yang et al., Cell Death Diff 2018). By this way TRMP regulates p27^kip expression. In our view the IRESome is not limited to the protein directly linked to the mRNA but to a protein (or RNP) complex. Many ncRNAs are considered as regulators of gene expression without binding directly to mRNA or gene promoters: lcnRNAs and circRNAs are sponges for miRNAs and RBPs and by this mechanism they regulate gene expression. In our case Neat1, by participating in forming the IRESome by binding different RBPs present in the paraspeckle, acts as a positive regulator of gene expression. The smiFISH data showing the co-localization of Neat1 with the IRES-containing mRNA (Figure 3). The role of Neat1 in translational control is also supported by its export from the nucleus observed in Figure 1F-K, Figure 1—figure supplement 2 and Figure 3, and its presence in hypoxic polysomes shown in our previous paper (Hantelys et al., 2019, Figure 2).

3. Section entitled "Paraspeckle proteins P54nrb and PSCP1, but not SFPQ, are ITAFs of the FGF1 IRES." The authors assess three major components of paraspeckles (the SFPQ, p54nrb and PSPC1 proteins) for a potential role in IRES activity. Since knockdowns are used to reduce protein levels and this takes several days to achieve, how can effects on IRES activity or FGF1 protein levels be directly link to the protein activity targeted for knockdown versus a secondary effect arising due to target knockdown? How selective is the reported effect for the FGF1 5' leader? Hence, I think its difficult to conclude that "The ability of three paraspeckle components, Neat1, p54nrb and PSPC1 to regulate the FGF1 IRES activity led us to the hypothesis that the paraspeckle might be involved in the control of IRES-dependent translation. ". Can direct interaction be shown and can this, in turn, be linked to a translational response.

The direct interaction of at least p54 and SFPQ with Neat1 has been established previously, and the three DBHS proteins (p54, SFPQ and PSPC1) are known to function in heteroduplex (for example see: Yamazaki et al. (2018). Functional Domains of NEAT1 Architectural lncRNA Induce Paraspeckle Assembly through Phase Separation. Molecular cell, 70(6), 1038–1053.e7. https://doi.org/10.1016/j.molcel.2018.05.019). The interaction of Neat1 with the IRES-containing mRNA has been shown by smi-FISH (Figure 3). At this stage we cannot conclude whether the IRES interacts with Neat1 or with one or several of the paraspeckle proteins, but we can conclude that they are all in the same nuclear body. In a recent paper, it has been established that Neat1 interacts directly with several mRNA targets (Jacq et al., RNA Biol. 2021), thus we cannot exclude the hypothesis of a direct RNA-RNA interaction of the IRES with Neat1.

The selectivity for FGF1 leader (and other growth factor leaders) is shown in Figure 7: The EMCV IRES is not regulated by either of these paraspeckle components. The set of cellular IRESs assessed are all regulated by Neat1, and selectively by the paraspeckle proteins p54 and PSPC1.

4. Section entitled "P54nrb interactome in normoxic and in hypoxic cardiomyocytes". The authors write "The moderate effect of p54nrb or PSPC1 depletion on FGF1 IRES activity suggested that other proteins are involved". Maybe the moderate effect is due to the poor knockdown achieved with the siRNAs (Figure 3 – Suppl Figure 1; 50% for p54nrb and 25% for PSPC1)? The IP performed with HA-tagged p54nrb pulled down 2013 proteins if I understand this section correctly. Can the authors explain why so many proteins are being pulled down? The supplemental table provided with all the raw data is not reader friendly and the column labeling difficult to interpret for someone like me who is not familiar with seeing data in this format. The identification of some ITAFs in this dataset is taken as evidence that these ITAFs interact with p54nrb in paraspeckles. What independent validation has been undertaken to show that these are associated with p54nrb? How many proteins that are not ITAFs are pulled down by p54nrb and enriched when comparing hypoxic versus normoxic extracts? In Figure 4D and 4E, the identity of all proteins enriched under the conditions shown should be shown in the upper rectangles.

We agree that the moderate effect is due to the poor knock-down. HL1 are beating cardiomyocytes (see Video1), not efficiently transfected (and for that reason we have transduced them with lentivectors expressing the bicistronic reporter constructs). This has been changed in the text, page 6 2^nd paragraph. It was not expected to pull down so many proteins (Figure 5). This experiment has been done in independent quadruplicates. A possible explanation is that the paraspeckle is a huge RNP: a recent paper showed that 4268 RNAs are targeted to paraspeckle (Jacq et al., RNA Biol 2021). A single paraspeckle contains about 50 molecules of Neat 1-2 (Hirose WIR RNA 2019) binding multiple proteins. Many proteins are components of paraspeckle but the proteins interacting with Neat1 are mainly the core proteins p54 and SFPQ, thus is expected that p54 may form a big complex with many associated proteins. In addition, p54 can form complexes outside the paraspeckle (role in splicing).

Among 40 paraspeckle proteins listed in previous reports (Naganuma et al., EMBO J, 2012 and Yamamoto et al. 2021), 22 have been found interacting with p54^nrb in our SM-IP (supplementary file 7). Furthermore, among these 22 proteins, 6 have been described as ITAFs: FUS, hnRNPA1, hnRNPK, hnRNPM, p54^nrb/NONO, SFPQ/PSF (Godet et al., IJMS 2018). This analysis has been added page 6 and Table 1. In our view it is significant to have found more than half of the main paraspeckle proteins in our MS-IP. The identity of all enriched proteins has been shown in Figure 5D and 5E.

5. Section entitled "Neat1 is the key activator of (lymph)angiogenic and cardioprotective factor mRNA IRESs." In this section, the authors test the response of other IRESes to knockdown of p54, PSPC1, and Neat1 and depending on the response, stratify the IRESes into different regulons based on IRESome composition. Whereas, the experiments do allow the authors to classify the response of the different IRESes, I don't think it provides insight into IRESome composition since the experiments to not show that any of the three tested molecules directly associate with the IRES under study.

We agree we have not definitely proven about IRESome composition by our data suggest it anyway, as we show that p54 interacts with PSPC1 and nucleolin (Figure 5), and we also show the colocalization of IRES-containing mRNA with Neat1 in the paraspeckle (Figure 3). Our data, and the direct interactions shown between PSP proteins and Neat1 in previous reports provide a set of arguments to support our proposition.

Reviewer #2 (Recommendations for the authors):

1. There is an issue with timing of the response, the amount of Neat1 induction of the FGF IRES and paraspeckle activity that do not seem to align. Therefore, how can the conclusion be drawn that Neat1 is absolutely crucial.

The amount of Neat1 (Neat1-1 + Neat1_2) does not align indeed, but the amount of Neat1_2 shows a peak detected by RT PCR as well by FISH (Figure 1), which completely aligns with the FGF1 IRES peak of activity at 4h. In addition, this alignment is also observed in another cell line, the 67NR breast tumor murine cell line (Figure 1—figure supplement 3). Data show that this cell line is more resistant to hypoxia: Neat1 is not induced whereas Neat1_2 is strongly induced after 24h of hypoxia. This time, 24 hr, also corresponds to the IRES activation. Furthermore, in 67NR, we performed Neat1 and Neat1_2 knockdown, which generated IRES downregulation.

2. The authors silence the long non-coding RNA but their transfection per se causes stress to cell. It seems that a different approach instead of silencing should be used since stress response is the question that needs to be answered. Perhaps a stable cell line or CRISPR KO can be used instead? Otherwise, the authors need to explain why the cell stress induced by transfection is not relevant here.

We tried several knock-down approaches in HL-1 without success (siRNA, phosphorothiate oligonucleotides, shMir…). The gapmer was the best way to obtain a significant knock-down in these cells. We agree that CRISPR KO would be a nice approach and we also thought about it but it seemed us quite long to develop, as our experiments have already been delayed by the confinement and stock ruptures related to COVID19. We tried however to generate Neat1 CRISPR-based induction and silencing using lentivectors (with Cas9-VP160 and Cas-9-Krab, respectively, but we were unfortunately not successful). One gRNA was not sufficient, we obtain very low titers of the lentivector and in addition it was impossible in our hands to construct a lentivector coding several guides with the mCas9.

3. Only one cell line was used in these studies. The authors make the point in the Discussion that IRESs can behave differently in different cell lines, which raises the question regarding how it can be known that Neat1 is generally important.

As mentioned above, the experiments have been performed in another cell line, the 67NR breast tumor murine cell line (Figure 1 figure supplement 3). Data show that this cell line is more resistant to hypoxia: Neat1 is not induced whereas Neat1_2 is strongly induced after 24 hr of hypoxia. This time, 24 hr, also corresponds to the IRES activation. Furthermore, in 67NR, we performed Neat1 and Neat1_2 knockdown, which generated IRES downregulation. These data show that the role of Neat1 in IRESdependent translation, is not limited to HL-1 cells.

4. How did the authors confirm HL-1 cardiomyocytes were indeed under hypoxia? Which is the hypoxia control? This is not apparent.

The hypoxia conditions had been set up in our previous paper (Hantelys et al., eLife 2019). The 0₂ pressure is measured in the incubator. Then expression a HIF-1 target, VEGFA, were checked by RT PCR. Figure 1 in Hantelys et al. shows that the transcriptional response occurs at 8 hours. eIF2a phosphorylation was measured by capillary Western (Hantelys Figure 2), showing that phosphorylation appears at 4 hr, before the transcriptional response via HIF. The experiments presented in Figure 1 of the present paper were done in parallel to that published in 2019.

5. In the hypoxia studies, typically the rate of canonical protein synthesis is significantly reduced, mTOR is inactivated and eIF2α is phosphorylated. Does this occur in HL-1 cardiomyocytes under their conditions of early hypoxia and if so is mTOR inhibition or eIF2 phosphorylation necessary for the effects that they see? In other words, are the novel data presented connected to established effects of hypoxia on protein synthesis?

4E-BP1 and eIF2a phosphorylation have been analyzed in these conditions of early hypoxia in HL-1 and published in our previous paper in eLife (Hantelys et al. 2019, https://elifesciences.org/articles/50094 Figure 2 and this paper Figure 2—figure supplement 4). Clearly, no change in 4E-BP phosphorylation is observed but eIF2a is phosphorylated in these conditions. We also showed in that paper that global protein synthesis is decreased (P/M ratio, Figure 2).

6. According to their data (Figure 1 and supplemental Figure 1), there is a better correlation with the formation of paraspeckles and greater expression under hypoxia of the longer isoform, NEAT1_2. Can the authors explain the reason for focusing on just NEAT1, the shorter isoform, on the effect of the regulation of FGF1 IRES activity in Figure 2?

The knock-down of Neat1_2 has been performed and shows a decrease of FGF1 IRES activity (Figure 2figure supplement 3). Unfortunately, it is not possible to knock-down only Neat1-1 as its sequence is entirely comprised in that of Neat1_2 (Figure 1E). These data, in addition to the results provided by fluidigm (Figure 8) and knock-down in 67NR cells (Figure 1—figure supplement 3) indicate that Neat1_2 is involved in the regulation. However the knock-down of both isoforms (Neat1) has a more important effect on IRES activity, suggesting that both isoforms play a role. Due to the data with Neat1_2 gapmer (Figure 2—figure supplement 3) we have moderated the conclusion about the key role of Neat1_2 that we were not able to demonstrate.

Please note that Neat1 corresponds to Neat1-1 + Neat1_2.

7. The mRNAs encoding the major (lymph)angiogenic growth factors are fibroblast growth factor (FGF) and vascular endothelial growth factor (VEGF). The authors should evaluate the effect of Neat1 on the regulation of more than just one IRES. The same experiments shown in Figures 1, 2, and 3 should be included for VEGF, especially since the results presently cannot be extrapolated to other IRES elements, and because recent studies have shown that FGF can act synergistically with VEGF.

It has been done in Figure 7: IRESs of FGF2, VEGFA, -C, D, IGF1R, c-myc and EMCV have been tested. Neat-1 appears as a key regulator for all IRESs except for EMCV IRES.

8. If possible, the authors should provide immunoprecipitation and immunoblot analysis validation of their proteomic data.

The interaction of p54 with nucleolin has been validated by co-immunoprecipitation. (Figure 5—figure supplement 2).

9. Is the interaction between nucleolin and ribosomal protein Rps2 with p54 direct or indirect? The graphical abstract shows a direct interaction but there is no confirming evidence for this. Also, does this potential interaction only occur under hypoxia?

By immunoprecipitation we show that p54 interacts with nucleolin Figure 5—figure supplement 2. Figure 5D shows that this interaction increases in hypoxia.

10. Figures 1-3 need a protein loading control, perhaps actin.

Capillary Western (Jess) does not need a loading control : quantification is obtained from peak area, than normalized to total proteins loaded on each capillary. It is much more precise than classical Western blots.

11. Can the authors show that levels of FGF1 protein after SFPQ knockdown are unchanged?

Unfortunately we did not measure endogenous FGF1 because the IRES activity did not change.

Reviewer #3 (Recommendations for the authors):

My main concern is that depletion of NEAT1 by control gapmers already exerts stress on the cardiomyocytes. Can the authors be sure that observed changes in IRES activities were not caused by a higher levels of cell stress? Is eIF2alpha phosphorylated in the gapmer experiments (Figure 2)? If this is the case, it could explain why only few differences were noted in Figure 4D.

As written in page 8, 1^st paragraph, in the experiments with gapmer treatment the number of paraspeckles was already high in normoxia (almost 5 foci per cell), suggesting that cells were already stressed by the gapmer treatment, before being submitted to hypoxia. Our hypothesis to explain the effect of Neat1 depletion in normoxia is indeed that cells are stressed by the gapmer treatment. Regarding eIF2a, we performed the experiment (Figure 2—figure supplement 4): eIF2a is phosphorylated upon Neat1_2 gapmer treatment, but not using the control gapmer. This could explain the small difference of IRES activity between normoxia and hypoxia generated by the knock-down.

1. While NEAT1 abundance has an effect on IRES, the NEAT-ITAFs do not. The authors should check effects of DDX17, Eno3 and Hspa2 that were enriched in the cytoplasm during hypoxia (Figure 4E).

We chose to test only the knock-down of DDX17 (due to the difficulty to obtain the HL-1 culture medium with the COVID19 pandemic). As mentioned above, we observed an upregulation of the LucF/LucR ratio. We prefer no to publish this in the present paper because another study is being performed on the

role of several helicases and these data will be in our next publication.

2. All figures displaying FISH results are impossible to inspect. They need to be bigger.

We have improved the presentation of the FISH results.

3. Depletion of NEAT1 or NEAT1-ITAFs has always more effect on endogenous FGF1 protein abundance that on IRES activity. Is the half-life of FGF1 affected during the siRNA and gapmer treatments?

The half-life of FGF1 has been analyzed and does not vary upon gapmer treatment not with siRNA treatment (Figure 2—figure supplements 5-6 and Figure 4—figure supplements 3-4). It superior to 24 hr either without treatment, or with gapmer treatment, or with siRNA treatment (control and target-specific). A hypothesis to explain a more important effect on endogenous FGF1 mRNA than bicistronic mRNA translation may be that the IRES of the endogenous FGF1 mRNA is recruited more efficiently to the paraspeckle than the IRES present in the bicistronic mRNA, whose structure may be altered due to the presence of the upstream cistron. However, if one looks at the Western quantification (normalized to total proteins), the difference is not so big: upon Neat1 depletion endogenous FGF1 protein is 34%, while IRES activity is 50% (Figure 2C and 2D). Upon p54 or PSPC1 depletion, endogenous FGF1 is 61% or 82% while IRES activity is 70% or 80%, respectively (Figure 4E and 4F).

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

As detailed by referees #1 and 2 there are omissions in the text of the authors' responses listed in the rebuttal letter. Most importantly, since many of the results are correlative the conclusions are overstated. As requested by referee # 1, since the authors do not show a direct association between Neat1 and the ITAFs investigated herein, the authors ought to change or remove Figure 4A which claims that Neat1 and p54nrb interact with the FGF1 IRES.

Reviewer #1 (Recommendations for the authors):

The authors have provided new information which addresses some of my previous concerns. However, my main concern remains in that there is much correlative data from which direct cause-effect relationships are inferred.

We thank you for your detailed analysis and we agree that we often present correlative data, because we were not able to demonstrate the direct cause-effect, but we provide a set of convincing arguments to strengthen our hypothesis. We have modified our conclusions when requested.

The authors indicate that there was a decrease in the number of paraspeckles in cells in which Neat1 was knocked down, as well as a 2-fold reduction in IRES activity. However, in new data presented, the authors show that eIF2alpha is phosphorylated upon Neat1_2 KO (Figure 2, Suppl 4) and in their rebuttal write that this could explain the small difference of IRES activity between normoxia and hypoxia generated by the KO. I, therefore, don't understand how the authors conclude on line 232 of the MS that the Neat1 knockdown results suggest "that Neat1 regulates FGF1 mRNA translation, directly or indirectly". The same comments holds true for the data regarding changes in the polysomal mRNA context following Neat1 knockdown (Figure 8). Are these due to elevated eIF2 α phosphorylation and not necessarily due to the absence of Neat1

This comment is indeed justified. We have deeply analyzed this issue from our data and we can provide a clear answer (added in the text page 5):

The IRES activity corresponds to the ratio LucF/LucR and provides a value normalized to LucR activity. If we look to the Renilla luciferase activity alone, reflecting the first cistron expression, we see that the slight increase in eIF2a phosphorylation is not sufficient to block global translation. We have added a small table in Supplementary file 3, page 2, showing the ratios of LucR activities with Neat1 gapmer versus control gapmer: it shows that LucR expression is not silenced by Neat1 depletion. Furthermore, we have shown in our previous report that the FGF1 IRES activity (as well as all IRES activities tested in our studies) increases in hypoxia in conditions of strong eIF2a phosphorylation (Hantelys et al., 2019). We have also mentioned this observation in page 4. This argument is the most important demonstration that the change in IRES activity is not due to eIF2a phosphorylation. The paper by Hantelys et al., 2019, published in eLife was performed in exactly the same conditions, with the same cells!.

The conclusion has been modified as following page 5:

“All these arguments indicate that the significant decrease of FGF1 IRES activity and of endogenous FGF1 expression observed in Figure 2 cannot result from eIF2a phosphorylation or decrease in FGF1 half-life, and probably results from Neat1 depletion. This suggests that Neat1 might regulate FGF1 mRNA translation, directly or indirectly.”

The text page 8 commenting Figure 8 has also been modified as following:

“As eIF2a phosphorylation was slightly increased in these conditions (Figure 2, figure supplement 4), we cannot completely rule out that it could affect the expression of certain mRNAs, despite the absence of inhibition of global translation. However, the insensitivity of many IRESs to eIF2a phosphorylation shown previously suggests that the present data result from an effect of Neat1, particularly on translation of IRES-containing mRNAs, while the two isoforms may have distinct effects (Hantelys et al., 2019).”

The authors use smiFISH to probe the localization of Neat 1 and FGF1 IRES and find a ~ 2-fold increase in overlapping signals for the FGF1 IRES and Neat1 in hypoxic versus normal cells. (Figure 3B). I'm not sure how this led to the "hypothesis that the paraspeckle might be involved in the control of IRES-dependent translation." For this conclusion to be made, wouldn't one have to show altered IRES activity after versus before having encountered the Neat1 speckles?

The conclusion of the paragraph commenting Figure 3 page 5 is:

“These data suggested that the IRES-containing mRNA is recruited into paraspeckles during hypoxia. This colocalization study, coupled to the functional study showing the involvement of all the major paraspeckle component in the IRES-dependent translation led us to the final hypothesis of our paper of a paraspeckle involvement. It is presented as a hypothesis and not as a conclusion.”

Since the authors do not show a direct association between Neat1 and the ITAFs investigated herein, I would ask that they change or remove Figure 4A which appears to show Neat1 and p54nrb interacting with the FGF1 IRES.

Done: the IRES-containing mRNA has been removed from the drawing.

Author details

1. Anne-Claire Godet

   UMR 1297-I2MC, Inserm, Université de Toulouse, Toulouse, France

   Contribution

   Conceptualization, Formal analysis, Validation, Investigation, Methodology, Writing - original draft, Writing - review and editing

   Contributed equally with

   Emilie Roussel

   Competing interests

   No competing interests declared

2. Emilie Roussel

   UMR 1297-I2MC, Inserm, Université de Toulouse, Toulouse, France

   Contribution

   Conceptualization, Formal analysis, Investigation, Methodology, Writing - review and editing

   Contributed equally with

   Anne-Claire Godet

   Competing interests

   No competing interests declared

3. Florian David

   UMR 1297-I2MC, Inserm, Université de Toulouse, Toulouse, France

   Contribution

   Conceptualization, Data curation, Formal analysis, Methodology

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-9842-1548

4. Fransky Hantelys

   UMR 1297-I2MC, Inserm, Université de Toulouse, Toulouse, France

   Contribution

   Formal analysis, Investigation, Methodology, Writing - review and editing

   Competing interests

   No competing interests declared

5. Florent Morfoisse

   UMR 1297-I2MC, Inserm, Université de Toulouse, Toulouse, France

   Contribution

   Conceptualization, Formal analysis, Methodology, Writing - review and editing

   Competing interests

   No competing interests declared

6. Joffrey Alves

   UMR 1297-I2MC, Inserm, Université de Toulouse, Toulouse, France

   Contribution

   Formal analysis, Investigation, Methodology

   Competing interests

   No competing interests declared

7. Françoise Pujol

   UMR 1297-I2MC, Inserm, Université de Toulouse, Toulouse, France

   Contribution

   Investigation, Methodology

   Competing interests

   No competing interests declared

8. Isabelle Ader

   UMR 1301-RESTORE, Inserm, CNRS 5070, Etablissement Français du Sang-Occitanie (EFS), National Veterinary School of Toulouse (ENVT), Université de Toulouse, Toulouse, France

   Contribution

   Conceptualization, Formal analysis, Supervision, Writing - review and editing

   Competing interests

   No competing interests declared

9. Edouard Bertrand

   UMR5535 CNRS-IGMM, Université de Montpellier, Montpellier, France

   Contribution

   Conceptualization, Supervision, Methodology, Writing - review and editing

   Competing interests

   No competing interests declared

10. Odile Burlet-Schiltz

    Institut de Pharmacologie et Biologie Structurale (IPBS), Université de Toulouse, CNRS, Toulouse, France

    Contribution

    Formal analysis, Methodology, Writing - review and editing

    Competing interests

    No competing interests declared

11. Carine Froment

    Institut de Pharmacologie et Biologie Structurale (IPBS), Université de Toulouse, CNRS, Toulouse, France

    Contribution

    Conceptualization, Data curation, Formal analysis, Methodology, Writing - review and editing

    Competing interests

    No competing interests declared

    "This ORCID iD identifies the author of this article:" 0000-0003-3688-5560

12. Anthony K Henras

    Molecular, Cellular and Developmental Biology Unit (MCD), Centre de Biologie Intégrative (CBI), Université de Toulouse, Toulouse, France

    Contribution

    Conceptualization, Formal analysis, Methodology, Writing - review and editing

    Competing interests

    No competing interests declared

13. Patrice Vitali

    Molecular, Cellular and Developmental Biology Unit (MCD), Centre de Biologie Intégrative (CBI), Université de Toulouse, Toulouse, France

    Contribution

    Conceptualization, Supervision, Methodology

    Competing interests

    No competing interests declared

14. Eric Lacazette

    UMR 1297-I2MC, Inserm, Université de Toulouse, Toulouse, France

    Contribution

    Conceptualization, Supervision, Methodology, Writing - review and editing

    Competing interests

    No competing interests declared

15. Florence Tatin

    UMR 1297-I2MC, Inserm, Université de Toulouse, Toulouse, France

    Contribution

    Formal analysis, Investigation, Methodology, Writing - review and editing

    Competing interests

    No competing interests declared

16. Barbara Garmy-Susini

    UMR 1297-I2MC, Inserm, Université de Toulouse, Toulouse, France

    Contribution

    Conceptualization, Formal analysis, Supervision, Funding acquisition, Writing - review and editing

    Competing interests

    No competing interests declared

17. Anne-Catherine Prats

    UMR 1297-I2MC, Inserm, Université de Toulouse, Toulouse, France

    Contribution

    Conceptualization, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Methodology, Writing - original draft, Project administration, Writing - review and editing

    For correspondence

    anne-catherine.prats@inserm.fr

    Competing interests

    No competing interests declared

    "This ORCID iD identifies the author of this article:" 0000-0002-5282-3776

• 1,268

  Page views

• 159

  Downloads

• 5

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.