Article activity feed

  1. Author Response:

    Reviewer #2:

    The authors investigated changes in the unstressed and stressed oligomeric states of the mammalian endoplasmic reticulum (ER) stress sensor, IRE1a. Previous biochemical and microscopy studies in mammalian cells and studies of the related protein Ire1 in yeast, describe an increase in oligomerization of the stress sensor upon treatment of cells with chemical agents that impair the ER protein folding environment. The general view has been that IRE1 in unstressed cells is a monomer and varying degrees of misfolded protein stress stimulate dimerization, activation, and higher order oligomerization. Distinguishing between monomers and dimers, as well as tetramers or other small oligomers is technically challenging, especially for integral membrane proteins. To address this challenge, the authors turned to single particle tracking fluorescence microscopy of Halo-tagged endogenous IRE1. Using a clever combination of random labeling with two fluorescent dyes and oblique angle illumination to visualize single molecules, as well as dimers, the authors surprisingly find that their endogenous IRE1 reporter appears to be dimeric in homeostatic cells. This observation challenges the predominant model in which IRE1 is monomeric in unstressed cells and that even dimerization represents a switch into an active state. The authors claim to detect evidence for higher order oligomers following treatment with stressors. The authors then use a series of IRE1 mutants to identify how oligomerization is regulated and present a new model to reconcile the different models of IRE1 activation in the literature.

    The authors have extensively characterized their novel experimental system in terms of protein expression levels, functionality, and ability to distinguish monomers and dimers. The data are well presented and the authors are clearly familiar with the arguments that have surrounded the IRE1 oligomer question. That the authors observe the characteristic XBP1 mRNA splicing activity in the absence of visible large IRE1 clusters may suggest that the large clusters reported by others may have distinct roles, perhaps in more permissive mRNA cleavage.

    The present study is undermined by two major weaknesses. First, while the authors persuasively demonstrate that they can detect IRE1a dimers, a major claim of the manuscript rests upon detection of tetramers and possibly higher order oligomers. Unfortunately, the authors provide no independent controls to show what tetramer or higher order oligomer data would look like. Thus, the authors can only infer that higher order oligomers are detected, based on modest shifts in the percent of correlated particle trajectories observed in some cells. More robust evidence is needed to make claims of oligomerization. Tools have been developed by others that can induce reversible oligomerization of proteins. Application of these tools would provide powerful controls for tetramers or even higher order oligomers in this study.

    The second, deeper concern, is the discrepancy between the Halo Tag clustering results in this study and studies by this lab and several other labs that report a distinct stress phenotype. In mammalian cells and yeast, IRE1 and Ire1, tagged with different fluorescent proteins or even a small HA peptide epitope tag, undergo quantitative visible formation of puncta or clusters upon treatment with stressors. The small number of bright clusters that form effectively deplete the rest of the ER of IRE1 signal. In the present study, the authors observe no visible change in IRE1-Halo localization in stress cells. The authors do not investigate the cause of this difference. While one might argue that the presence of stress-inducible IRE1 activity is sufficient to argue that the reporter in this study is functional, IRE1 reporters (that do cluster) described in previous studies by the Walter lab and other groups are also demonstrably functional. Does IRE1 normally cluster? Is it cell-type dependent? Tag-dependent? Notably, the Pincus et al. PLoS Biology paper from the Walter lab used two different fluorescent protein tags that do not heterozygously dimerize. Robust colocalization and FRET signals were detected upon treatment of cells with stressors and clustering was subsequently observed. A 2007 Journal of Cell Biology study from Kimata et al. reported clustering in yeast with an Ire1 tagged with an HA epitope peptide. The HA peptide seems unlikely to be prone to any oligomerization propensities that GFP tagged reporters might experience. Importantly, a 2020 PNAS paper from the Walter lab (Belyy et al.) studied clustering of a robustly monomeric mNeonGreen-tagged IRE1 in U2-OS cells and mouse embryonic fibroblasts and this construct readily clustered following stress induction.

    When evaluated against the backdrop of the extensive literature describing the visual behavior of IRE1a in live cells, the absence of stress-induced clustering is both puzzling and disconcerting. Given the focus of this study is to use visual techniques to study IRE1a interactions, the burden of proof is on the authors to resolve this significant discrepancy with the rest of the IRE1a literature. One can easily imagine that incorporation of the majority of the pool of IRE1a into 10-100 clusters could produce very different correlated trajectory behavior. Until the authors can determine why their reporters behave differently from other IRE1a reporters and establish which version accurately reflects physiologic IRE1a behavior, the potential impact of the findings of this manuscript are of unknown value.

    We thank the reviewer for this detailed assessment of our work. We agree that the question of apparent discrepancy in the formation of observable IRE1 clusters between this manuscript and earlier work is important. We have now addressed this issue both in the revised version of the manuscript and in specific point-by-point responses to reviewers’ comments. As a brief summary, we addressed the reviewer’s first concern (lack of controls larger than dimers) by cloning and validating a tetrameric HaloTag construct, the measurements from which were entirely consistent with the model we presented in the original version of the manuscript. To address the reviewer’s second concern, we present several lines of evidence showing that the discrepancy between the formation of microscopically visible IRE1 clusters in earlier studies and the absence of such clusters in the present work almost certainly results from differences in expression levels. First, our IRE1-HaloTag construct is perfectly capable of forming stress- induced clusters, as we show in the new Figure 1 – Figure Supplement 3. Second, we point to a parallel study by Gómez-Puerta et al., who demonstrate that a more “conventional” IRE1-GFP construct does not form visible stress-dependent puncta when it is expressed at a low level comparable to that of untagged IRE1 in HeLa cells, despite being fully active. Third, our earlier work in the 2020 PNAS paper referenced by the reviewer actually showed that even in the overexpression context, IRE1-mNeonGreen only forms visible puncta in just over half of all cells, despite the fact that XBP1 processing is nearly 100% effective in bulk assays. Furthermore, in the same paper we show that, rather than all IRE1 molecules being sequestered in clusters, only a small fraction (~5%) of IRE1-mNeonGreen assembles into large puncta while the remaining 95% of IRE1 stays uniformly distributed throughout the ER. Taken together, we believe that IRE1 does have the propensity to assemble into larger clusters when its expression levels are high (regardless of the tag used), but that these clusters are not strictly required for its activation. We have made significant changes to the discussion section of the manuscript to clarify the above points and directly address the apparent discrepancy between the present work and earlier studies.

    Reviewer #3:

    In this paper, the authors' aim was to test how IRE1's oligomerization state relates to its activation status without relying on ectopic overexpression. The principle underlying the work is a rather simple one, which is that, if the population of IRE1 can be labeled stochastically with either of two different fluorescent probes, then if the protein dimerizes, presuming single molecules can be visualized, correlated migration of a spot of each fluorophore should be observed for some of those dimers. Any correlated migration, maintained for long enough, will by necessity by some sort of dimer or multimer. In principle, if my math is right, the correlation should be 50% of spots of each color, assuming all the molecules are in a dimer, all molecules are labeled with one fluorophore or the other, and the koff of the fluorophores is very low. In practice, the correlation appears closer to 10%, which the authors establish using a control molecule that should not dimerize except by chance, and another for which pseudo-dimerization is enforced due to the two HALO domains used to bind the fluorophores being conjugated to the same molecule in cis. Much of the paper is devoted to establishing the fundamentals of the system. For these experiments, the authors replaced endogenous IRE1 with the HALO-tagged version to generate near-normal expression and show that the IRE1-HALO behaves similarly to endogenous. They also show that correlated migration is observed in the dimer control to a much greater extent than in the monomer.

    Using these findings, they demonstrate, in my mind quite conclusively, that IRE1 exists as a dimer even in the unstimulated state. During ER stress, the authors observe a state that is more highly ordered. Mathematical modeling suggests a transition from predominantly dimers to a mix of dimers and something more highly ordered, with tetramers being the simplest explanation. Satisfyingly, a mutation that breaks the known dimer interface causes the protein to exist solely in monomers, as does deletion of the IRE1 lumenal domain, while disrupting the oligomerization interface keeps the protein as dimers. Mutation or deletion of the kinase and RNase domains does not affect higher order status, suggesting that activation of these domains is not a prerequisite for assembly. It is clear from this that the central claims of the paper, which is that IRE1 exists in a dimer in the basal state and transitions to a higher ordered structure in the activated state, are supported. Moreover, the general approach is likely to be appealing to the study of other molecules activated by multimerization.

    We thank the reviewer for this thoughtful and helpful analysis of our work.

    The principal advance of the paper is the technological approach for tracking IRE1 (and, presumably, other molecules whose activity is regulated by dimerization). The approach is quite elegant for that purpose. Its impact in terms of conclusions about IRE1 is perhaps less clear. The authors rationalize their endogenous-replacement approach by describing how their previous efforts and those of others relied on ectopic overexpression of GFP-tagged IRE1. The authors take great pains to claim that the observed multimerization status of the IRE1-HALO constructs is not a function of expression level, which would imply then that expression level alone is not responsible for the previously observed IRE1 oligomeric puncta. It is not clear why exactly the authors' results differ from this group's previous studies on the topic nor where the truth lies, including whether something inherent to the GFP-tagged overexpression approach favors non-physiologic structures, whether the difference is fundamentally one of cell type, or whether multimerization and activation are correlated but not causally related, with multimer-breaking mutations killing IRE1 by some other mechanism.

    The question of reconciling our present data with earlier work (including work from our group) is clearly and understandably a central question for all three reviewers. As we detailed above in our responses to reviewers 1 and 2, we are convinced that the formation of large IRE1 clusters is largely dependent on expression level rather than the differences between fluorescent protein tags and the HaloTag. We added new supplementary figures and substantially revised the text of the manuscript to address this question directly.

    Interpreting the data is also complicated by the fact that, while the authors point out that the percent of correlated trajectories (i.e., the measurement of multimerization state) does not itself correlate with expression level (using trajectories-per-movie as a proxy), the proper conclusion from that lack of correlation is not that variance in expression level does not account for the changes in apparent multimerization status, but instead that it cannot be the only factor. In some sense, the authors are attempting to play the argument both ways, by arguing that expression level matters for IRE1 activation (from previous studies) and that it doesn't (from this study). I think to address this the authors will need to better account, one way or another, for why the findings presented here differ from their previous findings and why these are the more salient (if in fact they are).

    This is a very important point, and we thank the reviewer for raising it. We are not arguing that expression levels do not matter for the formation of oligomers; quite the contrary, as detailed above and in the revised version of the text, we believe that the formation of massive IRE1 oligomers observed in previous studies and in the new Figure 1 – Figure Supplement 3 is mainly a function of elevated concentration. What we do claim is that our approach can reliably pick out oligomeric differences within the relatively narrow range of concentrations used for single-particle tracking experiments in this paper. We are using the very weak truncated CMVd3 promoter in all transient transfection experiments, and we are only analyzing data from cells that have a comparable density of single-molecule spots to the density we observe in endogenously tagged IRE1-HaloTag cells. In fact, the metric of “trajectories per movie” used as a proxy for expression levels in Figure 5 – Figure Supplement 1 is an overestimation of the true variability of expression levels, since each movie only covers a small fraction of each cell’s area and the number of observed molecules varies depending on cell morphology. Practically speaking, all cells that we image have expression levels that are clustered together rather narrowly, roughly within differences of no more than a factor of 3. These levels, in turn, are significantly lower than the expression levels used in earlier papers by our group and others.

    The other somewhat substantial issue is that there is no control for what higher order structures look like. The authors give no sense for the dynamic range of the multimerization assay. I would presume that tetramers would show a higher percentage of correlated trajectories than dimers, and octamers higher still, and that the mathematical model accounts for this theoretical possibility in calculating an average protomer number of 2.7 in the stress condition, but it would be better to see that in practice; at first glance it would seem that engineering a tetrameric and/or higher order control and validating it would be straightforward.

    This is another great point raised by all reviewers. In the revised version of the manuscript, we engineered a new tetrameric control construct (See Figure 2 – Figure Supplement 1), the results from which agree remarkably well with the mathematical model we developed in the original version of the manuscript (see Figure 2 – Figure Supplement 3)

    Lastly, the data analysis lacks statistical justification for its conclusions. I presume given the high number of readings that the observed changes are all statistically significant, but that should be indicated, as in most cases the 95% confidence intervals shown are overlapping.

    This is another excellent point. The reviewer is correct that all relevant conclusions are statistically supported by the data, and our analysis code immediately calculates pairwise p- values for every plot using one of several relevant tests. Our preferred test is the permutation test, since it makes no assumptions about the underlying distributions being compared. To avoid cluttering the main plots, we have included tables of pairwise p-values for each plot in the revised version of the manuscript.

    Read the original source
    Was this evaluation helpful?
  2. Evaluation Summary:

    The manuscript reports on a new technical advance in fluorescence microscopy with which changes in the oligomerization of an integral membrane protein can be measured in live cells. The method is used to define the initial steps during activation of the IRE1 signaling arm of the unfolded protein response, leading to the discovery that IRE1 exists as a stable dimer in the absence of stress - which is in contrast to inferences from prior work. In response to stress, the protein assembles into a higher-order oligomer (likely a tetramer), an event that is mediated by the IRE lumenal domain and serves as a prelude to autophosphorylation. While the work will be widely noticed and excitedly discussed in the community, a reconciliation between the different results obtained in this study and in prior work, some of which was reported previously by the same lab, is currently lacking.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. The reviewers remained anonymous to the authors.)

    Read the original source
    Was this evaluation helpful?
  3. Reviewer #1 (Public Review):

    In this manuscript, the authors sought to define the early events associated with activation of the ER stress-responsive membrane protein IRE1. Towards that aim, they used CRISPR to integrate a HALO tag into the genomic locus of IRE1 at the C-terminus of the protein. The authors then adapted a single molecule fluorescence microscopy approach where the HALO tag is liganded with two different fluorophores to define the oligomeric state of membrane proteins in cellular models. They validated this approach using ER membrane proteins containing defined number of HALO tags (single or double) and imaged with oblique angle illumination microscopy to confirm their ability to detect effect monomer and dimers of these tags. Using this approach with IRE1, they showed that in the absence of stress, there is a high fraction of apparent IRE1 dimers in the membrane. In response to ER stress, this oligomer size (calculated by correlated trajectories) increased, suggesting that ER stress promotes IRE1 oligomerization, eventually returning to dimers at longer treatment times. Intriguingly, using the ER stressor thapsigargin, the authors indicate that oligomerization precedes auto-phosphorylation of IRE1, suggesting that oligomerization is a key step in the activation of this enzyme. Extending this, the authors then transition to an overexpression model where they incorporate IRE1 constructs containing mutant that disrupt specific parts of the protein or prevent dimeric or oligomeric interactions to probe their importance in this early oligomerization observed in response to ER stress. This demonstrated that the oligomerization was primarily dictated by the ER luminal domain and involved two distinct interfaces specifically required for IRE1 dimer formation (in the absence of stress) and oligomer formation (following ER stress). Ultimately, with these results, the authors propose a model whereby IRE1 exists primarily as a autophosphorylation-deficient, back-to-back dimer that upon ER stress oligomerizes to a phosphorylation competent oligomer that allow autophosphorylations and IRE1 activation.

    Overall this is an interesting approach and study to define early stages of IRE1 activation. Notably, it reveals a different model of these early stages of IRE1 activation than those previously reported by this group and others using GFP-tagged IRE1 overexpression constructs (something that was enabled by the integration of HALO tags into the genomic locus). The experiments are well performed and the data appear to all be interpreted correctly, although there are a few remaining questions that should still be addressed.

    Read the original source
    Was this evaluation helpful?
  4. Reviewer #2 (Public Review):

    The authors investigated changes in the unstressed and stressed oligomeric states of the mammalian endoplasmic reticulum (ER) stress sensor, IRE1a. Previous biochemical and microscopy studies in mammalian cells and studies of the related protein Ire1 in yeast, describe an increase in oligomerization of the stress sensor upon treatment of cells with chemical agents that impair the ER protein folding environment. The general view has been that IRE1 in unstressed cells is a monomer and varying degrees of misfolded protein stress stimulate dimerization, activation, and higher order oligomerization. Distinguishing between monomers and dimers, as well as tetramers or other small oligomers is technically challenging, especially for integral membrane proteins. To address this challenge, the authors turned to single particle tracking fluorescence microscopy of Halo-tagged endogenous IRE1. Using a clever combination of random labeling with two fluorescent dyes and oblique angle illumination to visualize single molecules, as well as dimers, the authors surprisingly find that their endogenous IRE1 reporter appears to be dimeric in homeostatic cells. This observation challenges the predominant model in which IRE1 is monomeric in unstressed cells and that even dimerization represents a switch into an active state. The authors claim to detect evidence for higher order oligomers following treatment with stressors. The authors then use a series of IRE1 mutants to identify how oligomerization is regulated and present a new model to reconcile the different models of IRE1 activation in the literature.

    The authors have extensively characterized their novel experimental system in terms of protein expression levels, functionality, and ability to distinguish monomers and dimers. The data are well presented and the authors are clearly familiar with the arguments that have surrounded the IRE1 oligomer question. That the authors observe the characteristic XBP1 mRNA splicing activity in the absence of visible large IRE1 clusters may suggest that the large clusters reported by others may have distinct roles, perhaps in more permissive mRNA cleavage.

    The present study is undermined by two major weaknesses. First, while the authors persuasively demonstrate that they can detect IRE1a dimers, a major claim of the manuscript rests upon detection of tetramers and possibly higher order oligomers. Unfortunately, the authors provide no independent controls to show what tetramer or higher order oligomer data would look like. Thus, the authors can only infer that higher order oligomers are detected, based on modest shifts in the percent of correlated particle trajectories observed in some cells. More robust evidence is needed to make claims of oligomerization. Tools have been developed by others that can induce reversible oligomerization of proteins. Application of these tools would provide powerful controls for tetramers or even higher order oligomers in this study.

    The second, deeper concern, is the discrepancy between the Halo Tag clustering results in this study and studies by this lab and several other labs that report a distinct stress phenotype. In mammalian cells and yeast, IRE1 and Ire1, tagged with different fluorescent proteins or even a small HA peptide epitope tag, undergo quantitative visible formation of puncta or clusters upon treatment with stressors. The small number of bright clusters that form effectively deplete the rest of the ER of IRE1 signal. In the present study, the authors observe no visible change in IRE1-Halo localization in stress cells. The authors do not investigate the cause of this difference. While one might argue that the presence of stress-inducible IRE1 activity is sufficient to argue that the reporter in this study is functional, IRE1 reporters (that do cluster) described in previous studies by the Walter lab and other groups are also demonstrably functional. Does IRE1 normally cluster? Is it cell-type dependent? Tag-dependent? Notably, the Pincus et al. PLoS Biology paper from the Walter lab used two different fluorescent protein tags that do not heterozygously dimerize. Robust colocalization and FRET signals were detected upon treatment of cells with stressors and clustering was subsequently observed. A 2007 Journal of Cell Biology study from Kimata et al. reported clustering in yeast with an Ire1 tagged with an HA epitope peptide. The HA peptide seems unlikely to be prone to any oligomerization propensities that GFP tagged reporters might experience. Importantly, a 2020 PNAS paper from the Walter lab (Belyy et al.) studied clustering of a robustly monomeric mNeonGreen-tagged IRE1 in U2-OS cells and mouse embryonic fibroblasts and this construct readily clustered following stress induction.

    When evaluated against the backdrop of the extensive literature describing the visual behavior of IRE1a in live cells, the absence of stress-induced clustering is both puzzling and disconcerting. Given the focus of this study is to use visual techniques to study IRE1a interactions, the burden of proof is on the authors to resolve this significant discrepancy with the rest of the IRE1a literature. One can easily imagine that incorporation of the majority of the pool of IRE1a into 10-100 clusters could produce very different correlated trajectory behavior. Until the authors can determine why their reporters behave differently from other IRE1a reporters and establish which version accurately reflects physiologic IRE1a behavior, the potential impact of the findings of this manuscript are of unknown value.

    Read the original source
    Was this evaluation helpful?
  5. Reviewer #3 (Public Review):

    In this paper, the authors' aim was to test how IRE1's oligomerization state relates to its activation status without relying on ectopic overexpression. The principle underlying the work is a rather simple one, which is that, if the population of IRE1 can be labeled stochastically with either of two different fluorescent probes, then if the protein dimerizes, presuming single molecules can be visualized, correlated migration of a spot of each fluorophore should be observed for some of those dimers. Any correlated migration, maintained for long enough, will by necessity by some sort of dimer or multimer. In principle, if my math is right, the correlation should be 50% of spots of each color, assuming all the molecules are in a dimer, all molecules are labeled with one fluorophore or the other, and the koff of the fluorophores is very low. In practice, the correlation appears closer to 10%, which the authors establish using a control molecule that should not dimerize except by chance, and another for which pseudo-dimerization is enforced due to the two HALO domains used to bind the fluorophores being conjugated to the same molecule in cis. Much of the paper is devoted to establishing the fundamentals of the system. For these experiments, the authors replaced endogenous IRE1 with the HALO-tagged version to generate near-normal expression and show that the IRE1-HALO behaves similarly to endogenous. They also show that correlated migration is observed in the dimer control to a much greater extent than in the monomer.

    Using these findings, they demonstrate, in my mind quite conclusively, that IRE1 exists as a dimer even in the unstimulated state. During ER stress, the authors observe a state that is more highly ordered. Mathematical modeling suggests a transition from predominantly dimers to a mix of dimers and something more highly ordered, with tetramers being the simplest explanation. Satisfyingly, a mutation that breaks the known dimer interface causes the protein to exist solely in monomers, as does deletion of the IRE1 lumenal domain, while disrupting the oligomerization interface keeps the protein as dimers. Mutation or deletion of the kinase and RNase domains does not affect higher order status, suggesting that activation of these domains is not a prerequisite for assembly. It is clear from this that the central claims of the paper, which is that IRE1 exists in a dimer in the basal state and transitions to a higher ordered structure in the activated state, are supported. Moreover, the general approach is likely to be appealing to the study of other molecules activated by multimerization.

    The principal advance of the paper is the technological approach for tracking IRE1 (and, presumably, other molecules whose activity is regulated by dimerization). The approach is quite elegant for that purpose. Its impact in terms of conclusions about IRE1 is perhaps less clear. The authors rationalize their endogenous-replacement approach by describing how their previous efforts and those of others relied on ectopic overexpression of GFP-tagged IRE1. The authors take great pains to claim that the observed multimerization status of the IRE1-HALO constructs is not a function of expression level, which would imply then that expression level alone is not responsible for the previously observed IRE1 oligomeric puncta. It is not clear why exactly the authors' results differ from this group's previous studies on the topic nor where the truth lies, including whether something inherent to the GFP-tagged overexpression approach favors non-physiologic structures, whether the difference is fundamentally one of cell type, or whether multimerization and activation are correlated but not causally related, with multimer-breaking mutations killing IRE1 by some other mechanism. Interpreting the data is also complicated by the fact that, while the authors point out that the percent of correlated trajectories (i.e., the measurement of multimerization state) does not itself correlate with expression level (using trajectories-per-movie as a proxy), the proper conclusion from that lack of correlation is not that variance in expression level does not account for the changes in apparent multimerization status, but instead that it cannot be the only factor. In some sense, the authors are attempting to play the argument both ways, by arguing that expression level matters for IRE1 activation (from previous studies) and that it doesn't (from this study). I think to address this the authors will need to better account, one way or another, for why the findings presented here differ from their previous findings and why these are the more salient (if in fact they are).

    The other somewhat substantial issue is that there is no control for what higher order structures look like. The authors give no sense for the dynamic range of the multimerization assay. I would presume that tetramers would show a higher percentage of correlated trajectories than dimers, and octamers higher still, and that the mathematical model accounts for this theoretical possibility in calculating an average protomer number of 2.7 in the stress condition, but it would be better to see that in practice; at first glance it would seem that engineering a tetrameric and/or higher order control and validating it would be straightforward.

    Lastly, the data analysis lacks statistical justification for its conclusions. I presume given the high number of readings that the observed changes are all statistically significant, but that should be indicated, as in most cases the 95% confidence intervals shown are overlapping.

    Read the original source
    Was this evaluation helpful?