Activation of the mitochondrial unfolded protein response regulates the formation of stress granules

  1. Marta Lopez-Nieto
  2. Zhaozhi Sun
  3. Emily Relton
  4. Rahme Safakli
  5. Brian D. Freibaum
  6. J Paul Taylor
  7. Alessia Ruggieri
  8. Ioannis Smyrnias
  9. Nicolas Locker

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

To rapidly adapt to harmful changes to their environment, cells activate the integrated stress response (ISR). This results in an adaptive transcriptional and translational rewiring, and the formation of biomolecular condensates named stress granules (SGs), to resolve stress. In addition to this first line of defence, the mitochondrial unfolded protein response (UPR mt ) activates a specific transcriptional programme to maintain mitochondrial homeostasis. We present evidence that SGs and UPR mt pathways are intertwined and communicate. UPR mt induction results in eIF2α phosphorylation and the initial and transient formation of SGs, which subsequently disassemble. The induction of GADD34 during late UPR mt protects cells from prolonged stress by impairing further assembly of SGs. Furthermore, mitochondrial functions and cellular survival are enhanced during UPR mt activation when SGs are absent, suggesting that UPR mt -induced SGs have an adverse effect on mitochondrial homeostasis. These findings point to a novel crosstalk between SGs and the UPR mt that may contribute to restoring mitochondrial functions under stressful conditions.

Summary statement

We describe a novel crosstalk between the mitochondrial unfolded protein response and the integrated stress response involving stress granules that protects cells from further stress.

Article activity feed

  1. Version published to 10.1101/2023.10.26.564187v2 on bioRxiv
  2. Review Commons

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

    Learn more at Review Commons

    Reply to the reviewers

    Reviewer #1 (Evidence, reproducibility and clarity (Required)): Throughout, the authors claim that there is a cross-talk between UPRmt and SG. This is unsubstantiated and unclear.

    We strongly disagree this comment. Throughout the manuscript, we show how manipulating UPRmt signalling affects SG formation, and how manipulating SG assembly alters mitochondrial functions and UPRmt-associated mitochondrial ouputs. In addition, both other reviewers are supportive of our conclusions.

    Major: Link between UPRmt and stress granules:

    The authors claim a link between the UPRmt and stress granule formation based on the finding that the loss of ATF5 affects the expression of UPRmt markers, but not ISR markers. Yet, the authors actually show that GTPP-induced SGs form in a manner independent of ATF5 (Supp. Fig. 2). Thus, there is no data in the manuscript that substantiates this claim.

    In the revised manuscript, we show that reducing ATF5 level results in defective SG assembly, with SGs displaying small size and more numerous, reflecting a maturation defect (Sup Figure 6B, 6C and 6D). In addition, we show a clear dependence of SGs to PERK activation (see comment below) and a specific increase of the ISR main negative regulator GADD34 (Figure 2A and 2B). Therefore, we disagree with this reviewer's conclusion and provide data supporting a link between UPRmt and SG formation.

    PERK-mediated activation of the ISR. The authors claim that PERK mediates activation of the ISR following GTPP treatment. However, the experiments in Fig. 2E were done 1h after treatment. The authors in Fig. 1C nicely show that SG formation begins at 2h. Thus, it is possible that following a longer GTPP treatment (i.e. >2h) the ISR is activated by different branches; for example, the mitochondrial branch that is mediated by HRI. Thus, the authors should determine which kinase mediates ISR activation at the time point that SG formation is maximal.

    We apologise if the description of the experimental procedure was unclear. These experiments are performed at 2h post GTPP treatment as explained in the text (see line 222) and legend (see lines 715-717, Figure 2 legend), and therefore performed at a time of maximal SG induction. Therefore, the identification of PERK as the driver for eIF2α-P and SG formation is performed at a time point where SG formation is maximal.

    Role of SG-linked decrease in cellular adaptation to stress. The finding that SGs limit mitochondrial respiration is interesting. Presumably this promotes cellular adaptation to mitochondrial stresses. The authors should test whether G3BP1/2 DKO cells are more susceptible to death following longer GTPP treatments.

    We thank the reviewer for this comment. These data are presented in Figure 8, where we show that G3BP1/2 dKO cells are less viable compared to wild-type cells following GTPP treatment for up to 28 hours.

    Minor: Fig. 2C should be moved to supplemental as well as the data indicated the lack of ISR inhibition.

    Figure 2C is now supplementary Figure 3.

    Fig. 3A should have representative images of all conditions from Fig. 3B.

    This has now been included as supplementary Figure 4.

    IFAs in Fig. 3 and 4 are hard to interpret given both DAPI and G3BP1 are in shades of blue. Ideally, insets of a merged panel should show each individual panel.

    We adopted the combination cyan, magenta and clue for our images to make scientific figures accessible to readers with red/green color-blindness. For these figures, G3BP1 is in light cyan and DAPI in dark blue, a colour we adopted previously in three publications (PMID 36965618, PMID 35098996, PMID 31905230), allowing colour blind reader to appreciate the results.

    Reviewer #1 (Significance (Required)): The link between the UPRmt and SGs is interesting and would be an advance. However, the authors put forward data that indicates SGs form in an UPRmt (ATF5)- independent manner. An interesting aspect of this story for which there is data is that SGs limit mitochondrial function. This should be explored further (i.e. although it limits mitochondrial respiration, perhaps SGs protect mitochondria against chronic ISR stress).

    As suggested we now provided an extensive amount of additional data supporting a role in mitochondrial functions, with data demonstrating that the absence of SGs rescues cell viability (Figure 8A and 8B), restoring mitochondrial functions such as respiration, ATP production (Figure 6D, 6E and 6F) or translation (Figure 7A), and reducing the production mitochondrial ROS (Figure 6C) or mitochondrial fragmentation (Figure 6A and 6B).

    Reviewer #2 (Evidence, reproducibility and clarity (Required)): Summary: The article by Lopez-Nieto Jordana et al entitled "Activation of the mitochondrial unfolded protein response regulates the dynamic formation of stress granules" describes the identification of a novel cross talk between the mitochondrial unfolded protein response (UPRmt) and the integrated stress response (ISR) and the contributory role SG regulation plays in mitochondrial function and adaptation to stress. This manuscript presents data highlighting that activation of the UPRmt results in the temporal modulation of SG formation via GADD34 levels and further this analysis by suggesting that these levels of GADD34 may enable cells to be protected from prolonged stress.

    Minor comments: This is a very well written manuscript with beautifully presented data. There are some inconsistencies/typos with the abbreviation GTPP- this needs to be checked within the manuscript but examples are on Lines: 204/206/214/324/328/357.

    This has now been corrected throughout.

    Check reference list for inconsistencies; line 680 reference has no page numbers, line 718 reference has no issue or page numbers

    This has now been corrected, references curated throughout.

    Line 255 - is it correct to say induction here? I think impairment should be used.

    This has now been corrected, see lines 283-284.

    Cell type not mentioned in Fig 2 legend.

    This has now been corrected, see line 707.

    Errors in Fig 4 legend - 4F, G do not exist.

    This has now been corrected, see lines 748-750.

    Major comments: In figure 1- the GTPP treatment only results in 25% of cells showing SGs compared with 80% in Ars treated cells. While the activation of ISR markers by GTPP treatment is convincing (in Figure 2A), What happens to overall protein synthesis levels in these cells? Puromycin incorporation assays would be a useful addition here.

    We now show in Figure 1D that GTPP treatment result in a global reduction in translation, and that cells displaying SGs present with a stronger shut-off when compared with treated cell lacking SGs.

    Fig. 1A - ATF4 upregulation is lower in ATF5 siRNA treated cells - what is % uptake of the siRNA in these cells - also see comment below. If possible, it would be nice to see the re-localisation of ATF5 to the nucleus to confirm the UPRmt activation of this protein

    These are experiments that we had planned to perform, however in our hands none of the commercially available antibodies allowed us to determine with confidence the localisation of ATF5. We have not determined the uptake of ATF5 siRNA but show by qPCR a reduction in ATF5 mRNA levels following siRNA treatment (see Figure 1A).

    Does the dispersal of SGs also correlate with a recovery of protein synthesis- there is still a relatively high level of eIF2alph-P at the 8h (from Figure 2A).

    We have not performed these experiments as we do not believe they would have added depth to our study. It is well accepted that SG disassembly results in mRNA re-entry in polysomes and the restart of translation (PMID: 30664789). SGs disappear a few minutes before translation is resumed.

    In Figure 2A the 30 min treatment of GTPP induces a robust level of eIF2α-P yet SGs are only observed following the induction of ATF4/GADD34 at 2h. Puromycin incorporation assays may also be able to shed light on the lack of SG inductions at this stage. The formation of SGs around the time when ATF4 and GADD34 are induced seems counterintuitive and should be commented on.

    As commented in response to an earlier point, our analysis shows that GTPP result in a global reduction in translation level, the assembly of SGs in a subpopulation of cells (as reported also in the context of many viral infection) may reflect cell-specific differences in the levels of eIF2α kinases and/or differences in reaching the threshold needed for eIF2α phosphorylation to induce SG assembly (as shown in PMID 30674674 and PMID 35319985).

    In line 207-208 you state that "PERK is the main eIF2α kinase responsive to GTTP. Overall, these results suggest that induction of the UPRmt is associated with an early SG assembly and ISR activation through PERK." Does the PERK inhibitor inhibit the formation of SG following GTTP treatment?

    This is now shown in Figures 2E and 2F. Indeed pharmacological inhibition of PERK following GTPP treatment resulted in inhibition of SG assembly.

    Additionally, does GTPP activation of the UPRmt also induce an oxidative stress and therefore activate an additional EIF2AK such as HRI? If so could be the reason you don't get formation of SGs following Ars treatment? Have you considered what would happen if you used the UV stress which activates GCN2 followed by Ars treatment?

    As shown on Figures 2D and 2E, we could not detect contribution from the other eIF2a kinases GCN2 and PKR following GTPP treatment; and Figures 2E, 2F demonstrate that PERK inhibition is sufficient to revert eIF2a phosphorylation and ablate SG induction, as noted in the response to the point above. This strongly suggest that the eIF2a kinase HRI does not contribute to eIF2a signalling, however we do not exclude in the broader sense (beyond eIF2a signalling) an induction of oxidative during UPRmt activation. Furthermore, as shown in Figure 2D, A-92 treatment reduced p-eIF2a levels in response to UV treatment but not those induced by GTPP therefore we can exclude a contribution from GCN2. If we understand correctly, this reviewer asks what would happen if cells were UV-stressed to activate GCN2 followed by oxidative stress with arsenite. This is outside the scope of this manuscript, but based on our previous work showing that mRNA GADD34 mRNA levels act as the molecular memory of the ISR and drives cell adaptation to acute and chronic stress, we would expect that the response to a second pulse of stress would be dampened by the sustained level of GADD34 mRNA induced following the first stress (see PMID 35319985). In these previous studies we already demonstrated that induction of p-eIF2a and SGs by a first acute stress (heat shock or thapsigargin) impairs the induction of p-eIF2a and SGs by a second acute (heat shock or arsenite) or chronic (HCV infection) stress (PMID 35319985, see Figure 6; PMID: 38602876, see Figure 7).

    Overall, this and the response to the previous comment strongly support that PERK activation, and the resulting induction of GADD34, are responsible for SG regulation following GTPP treatment.

    In Figure 3, for the paraquat experiments have you missed the transient induction of SGs by only looking at 48h? You already have GADD34 levels high here so SGs/eIF2α-P levels will already be lowered.

    We have now included additional timepoints, see supplementary Figure 5, showing the absence of SGs at 1, 2, 6 and 24h post paraquat treatment, to complement the 48h treatment previously shown.

    In addition, when analysing GTPP + Ars treatment impact on SG formation (Fig 2B), could the 2 h GTPP + Ars data also be included, as this is the peak time for SG induction by GTPP

    This is now included in Figure 3B.

    In line 211 you refer to the early and late stages of the stress, how have these been defined? It seems that the ability of the UPRmt to be protective to an additional stressor is time dependent- the number of SGs that are present following the additional stress increases from 4-8h. Does this correlate with a decrease in the level of GADD34?

    We define early and late to the time points corresponding to induction (early) or disassembly (late) of SGs. Also see lines 227-230.

    In line 254 you state that ATF5 silencing didn't impact the ISR or SG formation? These data suggest that the formation of SGs is not a direct impact of activation of the UPRmt but rather activation of the cellular ISR possibly due to the proteotoxic and/or oxidative stress? Can the authors comment on this?

    We now show in supplementary Figure 6 that reducing the expression of ATF5 results in defects in SG maturation with GTPP treatment resulting in more numerous and smaller SGs. Moreover, it should be noted that HSF1, in addition to ATF5, is a key controller of UPRmt induction and future studies could aimed at dissecting the role of HSF1 in the SG-UPRmt crosstalk (discussed in lines 459-461).

    In Figure 4, If GADD34 was driving the loss of SGs in GTPP treated cells why are SGs not persistent in these KO cells. Please comment on this.

    Two phosphatases are known to catalyse eIF2a-P dephosphorylation, GADD34 and CReP. The current model proposes that GADD34, which is induced following stress, acts in a negative feedback loop to resolve cellular stress. In contrast, CReP is constitutively expressed and controls basal P-eIF2α levels independently from stress levels (PMID 27161320). In recent work, we have shown that when GADD34 expression is silenced, CReP takes over to revert eIF2a -P and therefore disassemble SGs (PMID: 38602876). This work also showed that CreP is stress-induced in the absence of GADD34. Therefore, in Figure 4 we can speculate that the absence of SGs in GTPP treated KO cells is due to the ability of CReP to compensate for the absence of GADD34. In the context of GTPP treatment followed by arsenite, GADD34 is important to increase the threshold at which SGs can form, altering the response to a second pulse of stress.

    In addition, in these GADD34KO cells there should also be a persistent level of eIF2α-P when treated with GTPP and Pq, there is some as evidenced by the quantification but this is not very convincing

    As noted here, we do provide evidence of sustained levels of eIF2a-P in cells treated with GTPP at least, the results of independent experiments (n=3) showing persistent phosphorylation when compared treatment in GADD34 KO relative to WT cells. But as noted in the point above the likely activity of CReP can compensate for the lack GADD34, and therefore dampen the amount of eIF2a phosphorylation observed.

    Fig 4B shows no cells exhibiting SG following 4h GTPP treatment, which does not correlate with other experiments in the original cell line, e.g. supp 2B - please explain. Can GTPP still activate the UPR-mt in this CRISPR control cell line

    GTPP still activates the UPRmt in the CRISPR control cell line has shown by the inhibition of arsenite-induced SGs assembly when cells are pre-treated with GTPP for 4h (Figure 4A). However, we have noted that the timings of the response to GTPP can vary slightly, impacting on the exact SG kinetics, depending on the purity of the drug (synthetised through organic routes by our collaborator Dr Altieri), with the SG peak either at 2 h or at 4 h post-GTPP treatment. Potentially live imaging of SGs in control and GADD34 KO cells would alleviate this caveat, however in the time frame of the rebuttal, further engineering of GADD34 KO and parental lines into G3BP1/2 knock-outs / GFP-G3BP1 knock-ins was not achievable.

    In Figure 5, of the 80% of SG still present in GTPP treated Sil SGs- was size or frequency impacted here too as in Pq treatment?

    These data are now provided, see Figure 5C and in the result section lines 325-329. These show that GTPP treatment resulted in a reduction in average size of silvestrol-induced SGs, from 0.98 μm2 to 0.9 μm2, and increased average number of SGs, from 18 to 22, when compared to non-treated cells. Additionally, we also quantified features of Ars-induced SGs in GTPP-pretreated cells, data provided in Figure 3C and in the result section lines 245-250. The analysis showed that as paraquat, GTPP pre-treatment also impacts size and frequency of arsenite-induced SGs.

    This is just for clarification but If GTPP is a hsp90 inhibitor, is it specific to mitochondrial Hsp90 proteins?

    Indeed GTPP is specific to mitochondrial Hsp90.

    In the last results section the authors suggest that G3BP1/2 KO cells unable to assemble SGs present with improved mitochondrial function during stress. Firstly, is the UPRmt activated in these KO cells? Could the increased activity just be a consequence of the cells not being able to sense the stress and adapt? Are these cells able to recover from the GTPP stress to the same extent as the wt? Do they die at later timepoints? If you inhibited the disassembly of SGs using DYRK3 inhibitors would you decrease mitochondrial activity?

    The figure below confirms the upregulation of UPRmt genes mRNA levels after GTPP treatment in U2OS G3BP1/2 dKO (rebuttal Figure 1). We did not include this in the main manuscript given it is figure heavy already and this did not add depth to our results. Our extensive additional analysis shows that cells unable to assemble SGs present with multiple restored mitochondrial functions following UPRmt induction, including increased ATP production (Fig 6D), and respiration (FIG 6E, 6F), reduced mitochondrial ROS level (Fig 6C) and fragmentation (Fig 6A, 6B). These all support a model in which SG assembled following UPRmt induction contribute to impaired mitochondrial function and that their inhibition/disassembly is necessary to restore mitochondrial homeostasis.

    Rebuttal Figure 1: RT-qPCR analysis of the UPRmt and ISR markers DNAJA3, HSPD1, CHOP and ATF4 mRNA levels in U2OS cells treated with GTPP for up to 6 h. Results shown representative of n=3, normalised to RPL9 mRNA and shown relative to DMSO.

    Reviewer #2 (Significance (Required)): Significance: This is an interesting and clearly important observation providing mechanistic insight into the role SGs may play in the cells control of mitochondrial function during stress. The functional role of SGs in disease and stress is still widely unknown and this manuscript therefore sheds light on how the cell may use SGs to modulate and adapt to mitochondrial stress. This is an exciting area of research that will be applicable to a large audience as SGs are implicated in a wide range of diseases. While the data is significant there are currently a number of important experiments required to strengthen the current observational analysis. Below are some minor and major comments linked to the manuscript.

    We thank the reviewer for highlighting the importance of our work in an 'exciting area of research'.

    Reviewer #3 (Evidence, reproducibility and clarity (Required)): As it stands, this study will be suited for a specialized cell biology journal. In order to be published in a journal of a broader readership, the authors would need to address two major points:

    1. Mitochondrial dysfunction affects cellular function in many ways. Reduced levels of ATP, oxidative stress by increased ROS levels and mitochondrial precursor proteins that challenge proteostasis in the cytosol are just three major consequences of mitochondrial defects. Arguably, for the generation of stress granules, it will be important which of these consequences of mitochondrial dysfunction are prevalent. Since mitochondrial dysfunction is an ill-defined umbrella term, this study would be stronger if the authors could link stress granule formation to the specific molecular defects that arise from specific inhibition of mitochondrial functions.

    We agree with this reviewer that mitochondrial dysfunction can take many shapes and therefore to address their comment we have now performed an extensive amount of additional experiments probing various aspects of mitochondrial functions. In addition to the data previously included we can now show to that inhibition of SG formation during UPRmt induction result in increased cell viability (Figure 8A-B), restoring mitochondrial functions such as respiration, ATP production (Figure 6C-F) or translation (Figure 7A), and reduce mitochondrial ROS (Figure 6C) or fragmentation (Figure 6A-B). These all support a model in which SGs assembled following UPRmt induction contribute to impaired mitochondrial function and that their inhibition/disassembly is necessary to restore mitochondrial homeostasis.

    1. Also stress granules are an umbrella term. Different treatments will presumably change the spectrum of transcripts that are sequestered in these granules. As mitochondrial defects remodel the transcription and translation of mitochondrial precursor proteins, the study would benefit from a comprehensive analysis of the spectrum of transcripts that are contained in granules induced by GTPP and sodium arsenite, respectively.

    Previous studies, including our own, have demonstrated that indeed different stress (or infections) can result in the assembly of compositionally distinct SGs (or SG-like foci) that sequester specific subset of mRNAs or proteins. These studies are based on affinity purification or proximity ligation approaches followed by multi-omics analysis of SG components by RNA-seq and mass spectrometry. While we agree with this reviewer that determining the composition of UPRmt-induced SGs could help understand their function, we believe these studies are outside the scope of the current manuscript, and this would instead form the basis of subsequent study and manuscript.

    Reviewer #3 (Significance (Required)): The study is interesting but descriptive. It confirms previous observations. The advance in mechanistic insights is limited. Nevertheless, the study is technically sound and of interest for a specialized readership. As it stands, the study might be published in a specialized journal. In order to be of general interest for a large and general readership, the authors will have to provide much more mechanistic and molecular insight, which will require at least another six months of work.

    We have now produced an extensive additional body of work to answer specific comments made by all three reviewers, bolstering our hypothesis, and delving deeper into the impact of SG assembly on mitochondrial functions.

  3. Review Commons

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

    Learn more at Review Commons

    Referee #3

    Evidence, reproducibility and clarity

    Mitochondrial dysfunction induces a complex remodeling of gene expression. One defined branch in this response is known as the mitochondrial unfolded protein response (UPRmt). The transcription factor ATF5 acts as a key mediator of the mammalian UPRmt signaling. Moreover, mitochondrial defects also mute protein synthesis and trigger the integrated stress response (ISR). ISR is a well-characterized anti-stress routine characterized by eIF2alpha phosphorylation. The induction of cytosolic stress granules is one hallmark of ISR. In the present study, the authors observe that induction of UPRmt by inhibition of the mitochondrial HSP90 chaperone induces cytosolic stress granules. This is not unexpected given the well-established UPRmt/ISR and ISR/stress granules links. Still, the study is technically sound and extends our understanding of the effects of mitochondrial problems on the reactions in the cytosol.

    The authors compare two different inhibitors of mitochondrial functions: gamitrinib-triphenylphosphonium (GTPP) which interferes with HSP90 and whose effect on extra-mitochondrial proteostasis was well characterized by studies from Wade Harper and Christian Munch (Sutandy et al. 2023 Nature; Munch and Harper 2016 Nature); and paraquat which induces the generation of superoxide radicals from the respiratory chain. They found considerable differences of these two drugs in respect to stress granule formation which is consistent with previous observations. GTPP induces the accumulation of mitochondrial precursor proteins in the cytosol, which induces UPRmt. Defects in respiration however do not necessarily block mitochondrial protein import.

    In general, this is an interesting study that confirms previous observations. The molecular and mechanistic insights are limited and the authors neither identified the cascade of events that triggers stress granule formation upon HSP90 inhibition, nor did they analyze the transcripts that are sequestered by the cytosolic stress granules. Nevertheless, despite its rather descriptive nature, the study will be of interest for researchers studying the consequences of mitochondrial dysfunction.

    As it stands, this study will be suited for a specialized cell biology journal. In order to be published in a journal of a broader readership, the authors would need to address two major points:

    1. Mitochondrial dysfunction affects cellular function in many ways. Reduced levels of ATP, oxidative stress by increased ROS levels and mitochondrial precursor proteins that challenge proteostasis in the cytosol are just three major consequences of mitochondrial defects. Arguably, for the generation of stress granules, it will be important which of these consequences of mitochondrial dysfunction are prevalent. Since mitochondrial dysfunction is an ill-defined umbrella term, this study would be stronger if the authors could link stress granule formation to the specific molecular defects that arise from specific inhibition of mitochondrial functions.
    2. Also stress granules are an umbrella term. Different treatments will presumably change the spectrum of transcripts that are sequestered in these granules. As mitochondrial defects remodel the transcription and translation of mitochondrial precursor proteins, the study would benefit from a comprehensive analysis of the spectrum of transcripts that are contained in granules induced by GTPP and sodium arsenite, respectively.

    Significance

    The study is interesting but descriptive. It confirms previous observations. The advance in mechanistic insights is limited.

    Nevertheless, the study is technically sound and of interest for a specialized readership. As it stands, the study might be published in a specialized journal. In order to be of general interest for a large and general readership, the authors will have to provide much more mechanistic and molecular insight, which will require at least another six months of work.

  4. Review Commons

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

    Learn more at Review Commons

    Referee #2

    Evidence, reproducibility and clarity

    Summary:

    The article by Lopez-Nieto Jordana et al entitled "Activation of the mitochondrial unfolded protein response regulates the dynamic formation of stress granules" describes the identification of a novel cross talk between the mitochondrial unfolded protein response (UPRmt) and the integrated stress response (ISR) and the contributory role SG regulation plays in mitochondrial function and adaptation to stress. This manuscript presents data highlighting that activation of the UPRmt results in the temporal modulation of SG formation via GADD34 levels and further this analysis by suggesting that these levels of GADD34 may enable cells to be protected from prolonged stress.

    Minor comments:

    This is a very well written manuscript with beautifully presented data.

    There are some inconsistencies/typos with the abbreviation GTPP- this needs to be checked within the manuscript but examples are on Lines: 204/206/214/324/328/357.

    Check reference list for inconsistencies; line 680 reference has no page numbers, line 718 reference has no issue or page numbers

    Line 255 - is it correct to say induction here? I think impairment should be used.

    Cell type not mentioned in Fig 2 legend.

    Errors in Fig 4 legend - 4F, G do not exist.

    Major comments:

    In figure 1- the GTPP treatment only results in 25% of cells showing SGs compared with 80% in Ars treated cells. While the activation of ISR markers by GTPP treatment is convincing (in Figure 2A), What happens to overall protein synthesis levels in these cells? Puromycin incorporation assays would be a useful addition here.

    Fig. 1A - ATF4 upregulation is lower in ATF5 siRNA treated cells - what is % uptake of the siRNA in these cells - also see comment below.

    If possible it would be nice to see the re-localisation of ATF5 to the nucleus to confirm the UPRmt activation of this protein oes the dispersal of SGs also correlate with a recovery of protein synthesis- there is still a relatively high level of eIF2alph-P at the 8h (from Figure 2A).

    In Figure 2A the 30 min treatment of GTPP induces a robust level of eIF2α-P yet SGs are only observed following the induction of ATF4/GADD34 at 2h. Puromycin incorporation assays may also be able to shed light on the lack of SG inductions at this stage. The formation of SGs around the time when ATF4 and GADD34 are induced seems counterintuitive and should be commented on.

    In line 207-208 you state that "PERK is the main eIF2α kinase responsive to GTTP.Overall, these results suggest that induction of the UPRmt is associated with an early SG assembly and ISR activation through PERK." Does the PERK inhibitor inhibit the formation of SG following GTTP treatment?

    Additionally, does GTPP activation of the UPRmt also induce an oxidative stress and therefore activate an additional EIF2AK such as HRI? If so could be the reason you don't get formation of SGs following Ars treatment? Have you considered what would happen if you used the UV stress which activates GCN2 followed by Ars treatment?

    In Figure 3, for the paraquat experiments have you missed the transient induction of SGs by only looking at 48h? You already have GADD34 levels high here so SGs/eIF2α-P levels will already be lowered.

    In addition, when analysing GTPP + Ars treatment impact on SG formation (Fig 2B), could the 2 h GTPP + Ars data also be included, as this is the peak time for SG induction by GTPP

    In line 211 you refer to the early and late stages of the stress, how have these been defined? It seems that the ability of the UPRmt to be protective to an additional stressor is time dependent- the number of SGs that are present following the additional stress increases from 4-8h. Does this correlate with a decrease in the level of GADD34?

    In line 254 you state that ATF5 silencing didn't impact the ISR or SG formation? These data suggest that the formation of SGs is not a direct impact of activation of the UPRmt but rather activation of the cellular ISR possibly due to the proteotoxic and/or oxidative stress? Can the authors comment on this?

    In Figure 4, If GADD34 was driving the loss of SGs in GTPP treated cells why are SGs not persistent in these KO cells. Please comment on this.

    In addition, in these GADD34KO cells there should also be a persistent level of eIF2α-P when treated with GTPP and Pq, there is some as evidenced by the quantitation but this is not very convincing/

    Fig 4B shows no cells exhibiting SG following 4h GTPP treatment, which does not correlate with other experiments in the original cell line, e.g. supp 2B - please explain. Can GTPP still activate the UPR-mt in this CRISPR control cell line

    In Figure 5, of the 80% of SG still present in GTPP treated Sil SGs- was size or frequency impacted here too as in Pq treatment? This is just for clarification but If GTPP is a hsp90 inhibitor, is it specific to mitochondrial Hsp90 proteins?

    In the last results section the authors suggest that G3BP1/2 KO cells unable to assemble SGs present with improved mitochondrial function during stress. Firstly, is the UPRmt activated in these KO cells? Could the increased activity just be a consequence of the cells not being able to sense the stress and adapt? Are these cells able to recover from the GTPP stress to the same extent as the wt? Do they die at later timepoints?

    If you inhibited the disassembly of SGs using DYRK3 inhibitors would you decrease mitochondrial activity?

    Significance

    This is an interesting and clearly important observation providing mechanistic insight into the role SGs may play in the cells control of mitochondrial function during stress. The functional role of SGs in disease and stress is still widely unknown and this manuscript therefore sheds light on how the cell may use SGs to modulate and adapt to mitochondrial stress. This is an exciting area of research that will be applicable to a large audience as SGs are implicated in a wide range of diseases. While the data is significant there are currently a number of important experiments required to strengthen the current observational analysis. Below are some minor and major comments linked to the manuscript.

  5. Review Commons

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

    Learn more at Review Commons

    Referee #1

    Evidence, reproducibility and clarity

    The integrated stress response is activated by kinases that sense diverse stresses including viral infection and ER-linked stress and phosphorylate eIF2. This leads to the inhibition of translation initiation, the disassembly of polysomes, and the phase separation of mRNas and RNA binding proteins into stress granules (SG). Here, the authors show that treatment with GTPP, a previously established activator of the UPRmt, activates the ISR and induces the formation of stress granules. Following induction of the ISR, cells become more resistant to SG formation. The authors pinpoint this resistance to GADD34 dephosphorylation of eIF2a. Finally, the authors show that SGs limit mitochondrial respiration. These findings demonstrate the importance of putting the breaks on the ISR. Throughout, the authors claim that there is a cross-talk between UPRmt and SG. This is unsubstantiated and unclear.

    Major:

    Link between UPRmt and stress granules:

    The authors claim a link between the UPRmt and stress granule formation based on the finding that the loss of ATF5 affects the expression of UPRmt markers, but not ISR markers. Yet, the authors actually show that GTPP-induced SGs form in a manner independent of ATF5 (Supp. Fig. 2). Thus, there is no data in the manuscript that substantiates this claim.

    PERK-mediated activation of the ISR.

    The authors claim that PERK mediates activation of the ISR following GTPP treatment. However, the experiments in Fig. 2E were done 1h after treatment. The authors in Fig. 1C nicely show that SG formation begins at 2h. Thus, it is possible that following a longer GTPP treatment (ie. >2h) the ISR is activated by different branches; for example the mitochondrial branch that is mediated by HRI. Thus, the authors should determine which kinase mediates ISR activation at the time point that SG formation is maximal.

    Role of SG-linked decrease in cellular adaptation to stress.

    The finding that SGs limit mitochondrial respiration is interesting. Presumably this promotes cellular adaptation to mitochondrial stresses. The authors should test whether G3BP1/2 DKO cells are more susceptible to death following longer GTPP treatments.

    Minor:

    Fig. 2C should be moved to supplemental as well as the data indicated the lack of ISR inhibition.

    Fig. 3A should have representative images of all conditions from Fig. 3B.

    IFAs in Fig. 3 and 4 are hard to interpret given both DAPI and G3BP1 are in shades of blue. Ideally, insets of a merged panel should show each individual panel.

    Significance

    The link between the UPRmt and SGs is interesting and would be an advance. However, the authors put forward data that indicates SGs form in an UPRmt (ATF5)- independent manner. An interesting aspect of this story for which there is data is that SGs limit mitochondrial function. This should be explored further (i.e. although it limits mitochondrial respiration, perhaps SGs protect mitochondria against chronic ISR stress).

  6. Version published to 10.1101/2023.10.26.564187v1 on bioRxiv