Cellular organization depends on the ability of cells to recruit mRNA and protein molecules to precise subcellular localizations. In eukaryotic cells, mRNA transcripts that encode membrane and secreted proteins are targeted to the endoplasmic reticulum (ER) to facilitate the efficient and often co-translational delivery of their protein products to the ER lumen. mRNA targeting is mediated through the co-translational recognition of an N-terminal signal sequence by the signal-recognition particle (SRP; Walter et al., 1981). SRP-ribosome-nascent chain complexes are recruited to the surface of the ER by the SRP receptor (Gilmore et al., 1982), which channels the nascent polypeptide into the ER lumen through interaction with the Sec61 translocon (Görlich et al., 1992).

The unfolded protein response (UPR) acts as a combination of quality control pathways that monitor the folding status of proteins within the ER lumen and adjust the capacity of the ER’s folding machinery (Walter and Ron, 2011). IRE1α (inositol-requiring enzyme 1 alpha) triggers the most conserved branch of the UPR (Cox et al., 1993; Mori et al., 1993). It is an ER membrane resident stress sensor that is activated by the accumulation of misfolded proteins in the ER lumen and signals ER stress through the non-canonical splicing of X-box binding protein 1 mRNA (XBP1, HAC1 in yeast; Sidrauski and Walter, 1997; Tirasophon et al., 1998; Yoshida et al., 2001).

Processing of unspliced XBP1 (XBP1u) mRNA is initiated upon oligomerization and trans-autophosphorylation of IRE1α (Ali et al., 2011), which leads to the allosteric activation of its cytosolic kinase and RNAse domains (Korennykh et al., 2009). Once activated, IRE1α excises a highly conserved 26 nucleotide intron from the XBP1 coding sequence (Calfon et al., 2002; Yoshida et al., 2001) and the severed exons are rejoined by the tRNA ligase RtcB (Jurkin et al., 2014; Kosmaczewski et al., 2014; Lu et al., 2014). Intron excision causes a translational frameshift in the spliced XBP1 (XBP1s) transcript, which encodes a potent transcription factor that increases the folding capacity of the ER through a broad activation of stress response genes (Acosta-Alvear et al., 2007), including expression of ER-associated degradation factors (Brodsky, 2012). Beyond processing XBP1 mRNA, metazoan IRE1α is able to cleave a variety of mRNAs to initiate their rapid degradation in a pathway known as regulated IRE1-dependent decay (RIDD; Hollien et al., 2009; Hollien and Weissman, 2006). Even though RIDD has been found to play a key role in some pathological conditions, XBP1 splicing stands out as the main physiological output of IRE1 activation (Ishikawa et al., 2017).

To efficiently support rapid responses to ER stress, eukaryotic organisms display different strategies to ensure the timely encounter of IRE1α and its substrate mRNAs. In Saccharomyces cerevisiae, acute ER stress triggers the rapid oligomerization of IRE1 protein into a discrete number of foci (Aragón et al., 2009; Kimata et al., 2007). HAC1 mRNA, the yeast homolog of XBP1, is then recruited into these foci through a bipartite element that is located in the HAC1 3' untranslated region (UTR) while translational repression is imposed by the HAC1 intron itself (Aragón et al., 2009; Rüegsegger et al., 2001; van Anken et al., 2014). This swift targeting of HAC1 mRNA to pre-formed IRE1p clusters is essential to allow a timely response to ER stress and to sustain yeast proteostasis (Pincus et al., 2010).

The activation of metazoan IRE1α has been proposed to follow the same principles that were defined in yeast. Under ER stress, ectopic, fluorescently labeled IRE1α was found to cluster into large dynamic foci, and the kinetics of cluster assembly and disassembly approximately correlated with XBP1 splicing rates (Li et al., 2010). Yet, there is no direct evidence that the formation of large IRE1α clusters is required for splicing. Even though oligomerization of IRE1α has been proven to be the regulatory step that coordinates mRNA cleavage (Korennykh et al., 2009; Li et al., 2010) and the disruption of oligomerization interfaces has been shown to diminish RNAse activity (Karagöz et al., 2017; Sanches et al., 2014), the specific oligomeric state of splicing-competent IRE1α assemblies has not been precisely determined. In addition, only a minor fraction (~5%) of all cellular IRE1α protein concentrates in detectable foci (Belyy et al., 2020) and there is no direct evidence that they are indeed the sites of XBP1 processing at the ER.

In contrast to yeast HAC1, metazoan XBP1 mRNA is recruited to the ER surface through co-translational targeting that involves a peptide signal sequence and not a cis-acting localization element. Specifically, XBP1u transcripts encode a hydrophobic stretch (HR2) located at the C-terminal half of the protein that mimics a secretion signal (Yanagitani et al., 2009). On translation, this hydrophobic stretch is recognized by SRP, which delivers the nascent chain complex to the Sec61 translocon in the ER membrane (Plumb et al., 2015). Recognition of the HR2 peptide is aided by a translational pausing mechanism that has been proposed to stall the translating ribosome through high-affinity interactions with the peptide exit tunnel. This conveys stability to the mRNA-ribosome-nascent chain complex that facilitates its delivery to the ER membrane (Kanda et al., 2016; Yanagitani et al., 2011). Such a co-translational targeting mechanism suggests that IRE1α encounters XBP1u mRNA at the Sec61 translocon, where translating ribosomes would be poised. This notion is supported by the reported interaction of IRE1α with the translocon complex as well as by crosslinking data that find IRE1α in close contact with SRP, ribosomal RNAs (rRNAs), and a subset of ER-targeted mRNAs (Acosta-Alvear et al., 2018; Plumb et al., 2015). However, this model is difficult to reconcile with a situation where IRE1α molecules are recruited into large clusters with complex topologies that are not simply 2D patches in the ER membrane but have also been described to exclude the Sec61 translocon from specific regions within the clusters (Belyy et al., 2020).

Here, we have developed a single-molecule imaging approach that visualizes individual XBP1 mRNA transcripts and thus provides an important framework for the investigation of fundamental UPR biology principles and the recruitment of XBP1 mRNA to IRE1α and the ER surface. We image individual XBP1 transcripts that have been recruited for splicing on the ER and show that their recruitment is mediated by a translation-dependent targeting mechanism that involves SRP but functions through a non-canonical signal sequence. We demonstrate that XBP1 mRNAs are mobilized from the ER surface upon induction of ER stress by IRE1α-catalyzed splicing. Using a dual-color live imaging approach, we visualize individual XBP1 mRNA transcripts together with IRE1α-GFP (green fluorescent protein), which only assembles into clusters at increased expression levels and does not stably associate with XBP1 mRNA under splicing inhibition conditions. Instead, when expressed at endogenous levels, IRE1α-GFP simply outlines the ER and cleaves XBP1 mRNA in the absence of cluster formation during ER stress. This finding is further confirmed by a complimentary study that images single IRE1α molecules to characterize their oligomerization dynamics in response to ER stress and also demonstrates that large IRE1α clusters are not required for splicing activity (Belyy et al., 2021).

In order to directly visualize the recruitment of XBP1 mRNA to the ER, we developed a single-molecule imaging approach that takes advantage of the MS2 labeling system to detect individual reporter mRNAs in living cells (Bertrand et al., 1998). We generated an XBP1 wild-type (WT) reporter transcript that comprises the complete Mus musculus open reading frame (ORF) as well as its complete 3'UTR (Figure 1A, red; Calfon et al., 2002; Sugimoto et al., 2015). To enable the detection of single mRNA molecules at high signal-to-noise ratios, we further included 24 MS2 stem-loops in the 3'UTR of all reporter transcripts (Figure 1A) and made use of their specific recognition by fluorescently labeled synonymous tandem MS2 coat proteins (stdMCPs; Bertrand et al., 1998; Wu et al., 2015).

Figure 1 with 2 supplements see all

Download asset Open asset

Figure 1—source data 1

Raw gel images for Figure 1.

https://cdn.elifesciences.org/articles/75580/elife-75580-fig1-data1-v1.zip

Download elife-75580-fig1-data1-v1.zip

To complement the XBP1 WT reporter, we introduced a frameshift mutation downstream of the ER intron (Figure 1A, yellow, HR2 mutant) to prevent synthesis of the HR2 peptide, which has been shown to be essential for non-canonical SRP-mediated translocation of XBP1u mRNA to the ER membrane (Kanda et al., 2016; Yanagitani et al., 2009; Yanagitani et al., 2011). In addition, we employed a previously characterized SRP-recruited reporter (Voigt et al., 2017) to benchmark ER association of XBP1 transcripts against this established reporter construct encoding a secreted Gaussia luciferase protein (Figure 1A, gray).

We next generated HeLa cell lines stably expressing these reporter transcripts under a doxycycline-inducible promoter and from single genomic loci (Weidenfeld et al., 2009). To allow detection of individual mRNA particles as diffraction limited spots in the cytoplasm of living cells, we co-expressed nuclear localization signal-encoding fluorescently labeled NLS-stdMCP-stdHalo fusion proteins, which recruit excess stdMCP to the nucleus and thereby increase the signal-over-noise ratio in the cytoplasm (Voigt et al., 2017).

To confirm that these reporter constructs were indeed splicing competent, we first performed qPCR-based splicing assays (Figure 1B). As expected, we detected an increase in the levels of spliced XBP1 WT mRNA (red), and a transient drop in the levels of unspliced WT mRNA in response to the induction of ER stress with thapsigargin (TG). Using these measurements, we calculated the splicing ratio (spliced/unspliced) as a quantitative readout of splicing efficiency. As expected, splicing ratios increased sharply upon induction of ER stress in WT reporter-expressing cells and were much lower in HR2 mutant-expressing cells (yellow). In agreement with the RNA analysis, we detected increased XBP1s protein levels in response to TG treatment in cells expressing WT reporter transcripts (black triangle, Figure 1C). HR2-mutant cells produced only residual levels of XBP1s in response to TG treatment (white triangle, Figure 1C) while the majority of their XBP1 protein products was still derived from unspliced HR2 mutant transcripts (black triangle, same size as WT XBP1s protein).

We validated that MS2 tagging of ectopic XBP1 mRNA did not compromise its capacity to undergo splicing when ER stress was induced with TG or tunicamycin (TM) and ensured it did not compromise the activation of other UPR signaling mechanisms, such as the one initiated by PERK (Figure 1—figure supplement 1).

To assess XBP1 mRNA mobility and investigate particle dynamics of individual transcripts, we acquired streaming movies at fast frame rates (20 Hz) that detected XBP1 mRNAs as Halo-labeled diffraction limited spots in the cytoplasm of individual HeLa cells (Figure 1D, red). We performed single-particle tracking (SPT) over 100 consecutive frames and used the resulting particle coordinates to determine instantaneous diffusion coefficients (IDCs) as a measure of particle mobility (Berg, 1993; Voigt et al., 2017).

According to current models, XBP1u WT mRNA (but not the HR2 mutant) should be constitutively recruited to the ER surface for IRE1α-mediated splicing during ER stress. To investigate XBP1 mRNA association with the ER, we therefore integrated a fluorescently labeled ER marker protein (Sec61b-SNAP) into the reporter cell lines introduced above (analogous to Belyy et al., 2020). We imaged dual-labeled cells using a fluorescence microscope equipped with two parallel light paths and registered cameras for simultaneous detection of mRNA and ER signal in independent channels (Video 1).

Video 1

Download asset

Next, we quantified the mobility of individual particles with respect to their ER localization, which not only allowed us to visualize the recruitment of individual mRNAs to the ER surface (Figure 1D, upper panels; Video 2) but also enabled us to assess whether a particle is stably associated with the ER (Figure 1D, middle panels; Video 3). To this aim, we segmented the ER signal (Figure 1D, lower panels) and used it to generate distance maps that allowed us to correlate particle coordinates with ER boundaries. In these distance maps, positions on the ER were given positive values, and positions away from the ER were defined as negative. Based on particle trajectories, we determined the localization of individual transcripts throughout the entire image series and calculated cumulative ER localization indices that highlight robust localization phenotypes (Voigt et al., 2017). We combined the diffusion and ER colocalization analysis and employed it to benchmark the behavior of a Gaussia luciferase reporter transcript (Figure 1A) that encodes a secreted protein product and that we have previously shown to be predominantly localized to the ER (Voigt et al., 2017). Next, we performed the same analysis to quantify the mobility and ER association of XBP1 WT and HR2 mutant reporters (Figure 1E).

Video 2

Download asset

Video 3

Download asset

The combined data show that a large fraction of XBP1 WT transcripts (Figure 1F, red dots) behaves similar to the secreted Gaussia mRNAs (Figure 1F, gray dots). Many XBP1 WT transcripts exhibit low mobility and colocalize with the ER. However, there is another population of WT reporter tracks not observed for the Gaussia reporter that is more mobile and tends to not localize to the ER. Interestingly, the behavior of this population is exactly matched by the XBP1 HR2 mutant tracks (Figure 1F, yellow dots). These reporter mRNAs seem to have lost their ability to be recruited to the ER surface and exhibit a generally higher degree of mobility that is also apparent upon visual inspection (Figure 1F, Video 4). We employed the correlated diffusion and ER colocalization analysis to quantify the fraction of ER-associated particles per cell. To this aim, we used the clearly ER-associated Gaussia cluster to define cut-offs (D<0.06 μm²s⁻¹ and positive ER localization index, dashed lines in Figure 1F) for identification of XBP1 mRNA particles that showed a similar behavior. Based on these parameters, we found an average (per cell) of 27.4 ± 19.4% (mean ± SD) of all XBP1 WT and 3.1 ± 6.2% of all HR2 mutant transcripts to be associated with the ER (Figure 1G).

Video 4

Download asset

To corroborate the findings from the single-particle imaging approach through an independent method, we performed membrane flotation assays that allow separation of membrane from cytosolic fractions (Figure 1—figure supplement 2A; Mechler and Rabbitts, 1981). As expected, we found that XBP1 WT reporter mRNAs were detected in the membrane fractions to a similar extent as the endogenous XBP1u mRNA. In contrast, XBP1u HR2 mutant mRNA lacked membrane association and behaved like endogenous XBP1s (Figure 1—figure supplement 2B). Upon reconstitution of the original ORF through integration of two additional nucleotides that restore the HR2 reading frame but not the upstream part of the ORF, membrane association was restored (Figure 1—figure supplement 2C, D). Association of XBP1 reporter transcripts with the ER is therefore unambiguously linked to the expression of the HR2 peptide.

Together, these findings indicate that XBP1 WT reporter mRNA is recruited to the ER surface albeit to a lesser extent than canonical secretion-signal encoding Gaussia transcripts. Recruitment depends on the expression of the HR2 peptide, since a reporter mRNA that does not produce HR2 failed to associate with the ER. Our results are consistent with the non-canonical mechanism of XBP1 delivery to the ER and confirm that HR2 expression conveys stable ER association in a co-translational manner.

To test if translation-dependent recruitment of XBP1 transcripts to the ER membrane is necessary to enable mRNA splicing, we generated an XBP1 translation site reporter that would allow us to directly monitor XBP1u translation on the ER (Figure 2A). Specifically, we used a nascent polypeptide imaging approach that relies on the co-expression of a well-folded protein scaffold (spaghetti monster, SM) containing nine GCN4 antigen repeats (Eichenberger et al., 2018; Morisaki et al., 2016; Yan et al., 2016) and the MS2 stem loop array introduced above. To quantify protein synthesis of XBP1u transcripts on the ER, we generated a XBP1u translation reporter construct that contains the GCN4-SM downstream of the UPR intron but in frame with the XBP1u ORF. Upon splicing and excision of the intron by IRE1α, the ORF changes to XBP1s and the GCN4-SM is no longer in frame. Thus, the translation site signal can only be detected prior to mRNA splicing.

Figure 2 with 2 supplements see all

Download asset Open asset

Figure 2—source data 1

Source data containing raw gel images for Figure 2.

https://cdn.elifesciences.org/articles/75580/elife-75580-fig2-data1-v1.zip

Download elife-75580-fig2-data1-v1.zip

Quantitative real-time (RT)-PCR as well as western blot analysis confirmed that this construct was able to undergo splicing upon induction of ER stress (Figure 2B and C). To characterize the translational status of XBP1 mRNA in live imaging experiments, we employed GFP-fused single-chain antibodies (scAB-GFP; Voigt et al., 2017; Yan et al., 2016) that specifically recognize GCN4 peptides and allow for detection of individual translation sites as diffraction-limited spots in the cytoplasm of HeLa cells co-expressing the Sec61b-SNAP ER marker (Figure 2D, Figure 2—figure supplement 1A, Video 5). As a further characterization of this experimental set-up, we performed a similar dual-color live imaging experiment but this time focused on the simultaneous detection of mRNA and translation site signals. To test whether the bright GFP signal corresponded to individual translation sites, we first acquired the NLS-stdMCP-stdHalo mRNA in parallel with the scAB-GFP translation site signal (Figure 2—figure supplement 1B) and then treated the cells with puromycin (PUR) to inhibit translation (Figure 2—figure supplement 1C). Upon PUR treatment, all scAB-GFP spots disappeared, which led us to conclude that they represent active translation sites.

Video 5

Download asset

We proceeded to quantify the degree of ER association observed for XBP1u translation sites in individual cells through the correlated diffusion and ER colocalization analysis introduced above (Figure 2E, Figure 2—figure supplement 1A). Interestingly, this analysis revealed that the majority of XBP1u translation sites colocalized with the ER (53.8 ± 22.1%, mean per cell ± SD) and exhibited a low mobility that is comparable to the behavior of predominantly ER-localized Gaussia transcripts (mean ER association = 57.3 ± 16.8%) but very different from the average degree of ER association assumed by XBP1 WT transcripts (mean ER association = 27.4 ± 19.4%). Thus, we conclude that XBP1u reporters are tethered to the ER surface in a translation-dependent manner.

As the translational frameshift induced by IRE1α-mediated splicing should abolish translation of the GCN4 repeats, we assessed the fraction of translating XBP1u transcripts in response to induction of ER stress. We treated the cells with TM and quantified the degree of colocalization for XBP1u mRNA and translation site spots. To maximize detection efficiency and more accurately estimate particle numbers per cell, we performed a combined single-molecule fluorescence in-situ hybridization (smFISH) and immunofluorescence (IF) experiment in fixed cells (Figure 2F, Figure 2—figure supplement 1D). Specifically, we used smFISH probes against the 5'-end of the M. musculus XBP1 ORF and an anti-GFP antibody that allowed for detection of the scAB-GFP labeled nascent polypeptides by IF. We confirmed that XBP1 mRNA and scAB-GFP translation site spots corresponded to single translation sites that only colocalized when expressed from the same mRNA transcripts (Figure 2—figure supplement 2).

Next, we investigated how ER stress affects the association of XBP1 mRNA with the ER and set out to determine how XBP1u molecules encounter IRE1α. In order to distinguish between the behavior of unspliced and spliced mRNA transcripts, we generated a XBP1 reporter variant with point mutations in the 5' and 3' splice sites of the UPR intron that maintain its stem-loop structure but render the substrate cleavage incompetent (unspliceable, dark blue, Figure 3A, Figure 3—figure supplement 1A, B; Calfon et al., 2002; Gonzalez et al., 1999). In addition, we also generated a variant lacking the intron and constitutively expressing the XBP1s protein (spliced, light blue, Figure 3A, Figure 3—figure supplement 1A, B). We performed dual-color live imaging experiments (Figure 3B, Videos 6 and 7) and quantified reporter mobility and their degree of colocalization with the ER through correlated diffusion and ER colocalization analysis under non-stress conditions (Figure 3—figure supplement 1C, D). To further characterize particle behavior and control against potential higher-order oligomeric assemblies of reporter transcripts, we assessed the mean spot intensities detected for unspliceable and spliced reporter transcripts. They were comparable to the mean intensities of XBP1 WT and HR2 mutant spots that were collected in the same imaging experiment (Figure 3—figure supplement 1E) and exhibited a single defined peak confirming that the spot signal originated from single mRNA transcripts that did not associate in larger oligomeric assemblies.

Figure 3 with 1 supplement see all

Download asset Open asset

Video 6

Download asset

Video 7

Download asset

As expected, unspliceable reporter transcripts often exhibited a lower mobility (Figure 3—figure supplement 1C) and associated with the ER to a degree that is comparable to WT transcripts (Figure 3C) while spliced reporter mRNAs tended to diffuse at higher mobilities (Figure 3—figure supplement 1D) and displayed a lower degree of ER association (Figure 3C) comparable to cytoplasmic protein-encoding mRNAs (Voigt et al., 2017).

To determine the effect of ER stress on the ER association of WT, unspliceable and spliced reporter transcripts, we induced ER stress with TM (5 µg/ml) at least 3 hours before the start of the imaging session (Video 8). In addition, and to specifically assess the involvement of IRE1α in such association, we performed the same imaging experiments including 50 µM 4µ8C, a small molecule inhibitor that blocks substrate access to the active site of the IRE1α RNase domain and thereby selectively inactivates XBP1 cleavage (Video 9; Cross et al., 2012). As anticipated, ER stress-induced processing of WT reporters caused a strong decrease of their mean ER association from 27.4 ± 19.4% (mean ± SD) in the untreated condition to 10.1 ± 9.3% under TM treatment (Figure 3C, red). This result supports the idea that, upon completion of the splicing reaction, WT mRNAs are released from the ER membrane and behave like intron-free transcripts (Figure 3C, light blue) in the absence of ER stress (10.0 ± 9.1%). ER stress-induced mobilization of spliced WT reporter transcripts was a genuine consequence of IRE1α catalysis, since addition of 4µ8C to the TM condition restored ER association of WT reporters back to 33.2 ± 15.6% (Figure 3C, red). Taken together, these findings suggest that IRE1α-mediated splicing drives the release of translationally active, ER-tethered mRNAs.

Video 8

Download asset

Video 9

Download asset

In line with this notion, unspliceable reporter transcripts (Figure 3C, dark blue) not only associate with the ER to a high level (32.4 ± 13.5%) but also fail to show a similar reduction in ER association upon treatment with TM (23.7 ± 18.4%) and TM + 4µ8C (20.3 ± 12.6%) (Figure 3C). The same holds true for the spliced reporter construct. Since it does not encode the HR2 peptide and can therefore not be delivered to the translocon, it only associates with the ER to a limited extent (10.0 ± 9.1%). Upon induction of ER stress, its ER association rate is further reduced and similar to the unspliceable reporter, we do not observe significant changes in ER association between ER stress conditions in the absence (4.8 ± 5.0%) and presence (4.6 ± 4.4%) of 4µ8C.

We noticed that, for unspliceable and spliced reporters, ER stress caused a slight reduction of ER association when compared to untreated conditions. This effect may result from the general inhibition of cellular translation initiation triggered by the eIF2α kinase PERK, that promotes the UPR branch of the integrated stress response (Pakos-Zebrucka et al., 2016). It is plausible that the slightly reduced levels of ER association under ER stress conditions are due to the decreased recruitment of translating mRNPs to the ER surface, affecting all mRNAs to a limited extent (Voigt et al., 2017). This effect is not observed for the XBP1 WT reporter, where conversion of unspliced into spliced molecules is the major driver of mobilization from the ER. In summary, these experiments demonstrate that IRE1α activity is not required for ER association of XBP1 reporter mRNAs but suggest that IRE1α-mediated catalysis might contribute to the release of spliced mRNA molecules to the cytosol.

In combination, our findings support a model where IRE1α-mediated splicing is instrumental for the mobilization of XBP1 transcripts that are anchored to the ER in a translation-dependent manner. Based on this hypothesis, we sought to further investigate and visualize the sites of XBP1 processing on the ER membrane. To this aim, we developed an approach that allowed us to detect IRE1α in the reporter transcript-expressing HeLa cell lines introduced above and that was complementary to the single-molecule imaging approach, which detects endogenously tagged IRE1α protein molecules and is published in parallel to this study (Belyy et al., 2021). We knocked out the endogenous IRE1α using CRISPR/Cas9 and reconstituted its expression with a GFP-tagged IRE1α protein (Figure 4A). Identical to the previously published design of a splicing competent IRE1α-GFP construct, we introduced a GFP moiety in between the lumenal and kinase/RNase domains on the cytoplasmic site of the transmembrane protein (Belyy et al., 2020; Li et al., 2010).

Figure 4 with 1 supplement see all

Download asset Open asset

Figure 4—source data 1

Raw gel images for Figure 4.

https://cdn.elifesciences.org/articles/75580/elife-75580-fig4-data1-v1.zip

Download elife-75580-fig4-data1-v1.zip

IRE1α has been shown to form large oligomeric assemblies and microscopically visible clusters upon induction of ER stress in a number of studies (Belyy et al., 2020; Kimata et al., 2007; Li et al., 2010; Tran et al., 2021). Yet, the physiological relevance of these clusters remains unclear. To determine if the extent of IRE1α-GFP expression could artificially affect IRE1α clustering, we generated lentiviral constructs that induced different IRE1α-GFP expression levels. Specifically, we took advantage of the previously characterized Emi1 5'UTR that has been shown to downregulate translation approximately 40-fold (Yan et al., 2016). Since this 5'UTR was derived from the cell cycle protein Emi1, we termed the construct Emi1-IRE1α-GFP. For comparison, we also generated an IRE1α-GFP expression construct that was lacking the Emi1 5'UTR and expressed the IRE1α-GFP at higher levels.

As anticipated, western blot analysis confirmed that reconstituted IRE1α at low (Emi1-IRE1α-GFP) as well as at high levels (IRE1α-GFP) restored the functionality of IRE1α in KO cells, albeit to different extents (Figure 4B). While Emi1-IRE1α-GFP levels were similar to those of endogenous IRE1α, IRE1α-GFP expression was approximately 10-fold higher (Figure 4C). In both cell lines, ectopic IRE1α-GFP expression rescued XBP1 mRNA splicing under ER stress conditions, as determined by quantitative RT-PCR (Figure 4D) and by western blot detection of the resulting XBP1s protein (Figure 4B). At the same time, we noticed that GFP tagging slightly reduced XBP1 splicing levels both in the endogenous XBP1 mRNA and the WT reporter transcript (Figure 4—figure supplement 1A-C and Figure 4D). Other than that, IRE1α-GFP expressed from the Emi1 promoter supported transcription of XBP1-regulated genes, like ErdJ4, and did not significantly affect other signaling branches (Figure 4—figure supplement 1D). In line with previous reports (Li et al., 2010), strong overexpression of IRE1α-GFP triggered XBP1 mRNA splicing (and XBP1s synthesis) even in the absence of ER stress, underscoring the importance of adequate IRE1α expression levels for fine-tuning of the UPR.

In order to investigate if IRE1α clusters were the sites of XBP1 mRNA splicing on the ER, we imaged IRE1α-GFP in the HeLa cell lines stably expressing XBP1 reporter transcripts along with NLS-stdMCP-stdHalo and Sec61b-SNAP introduced above (Figure 4E). In agreement with earlier reports (Li et al., 2010) as well as a parallel study (Belyy et al., 2021), we detected IRE1α-GFP foci (defined as a ≥fivefold enrichment of GFP signal over background) in 21.89 ± 7.31% of cells expressing high levels of the fusion protein already after relatively short induction of ER stress with TM (5 µg/ml) for 2 hours (Figure 4F). Surprisingly, this was not the case for cells expressing Emi1-IRE1α-GFP at low levels, where we were unable to detect IRE1α-GFP clusters even after prolonged exposure to TM (5 µg/ml) for up to 7 hours. To make sure that we were not missing IRE1α clusters due to imaging conditions optimized for detection of fast-moving mRNA particles (e.g. short 50 ms exposure times), we acquired IRE1α-GFP signal from the same cells in the presence and absence of ER stress but this time using longer exposures (2000 ms) and maximum laser intensities. Under such conditions, we were able to detect IRE1α-GFP signal, which exhibited a characteristic ER-like distribution pattern but no IRE1α clusters (Figure 4E, right panel).

This observation suggested that IRE1α clusters are not necessary for the production of XBP1s, which we were able to detect in the absence of cluster formation (Figure 4D). To ensure that we were not missing a potential function of the previously observed IRE1α foci, we proceeded to image XBP1 WT mRNA recruitment to these oligomeric assemblies at high temporal and spatial resolution (Figure 4F, Video 10). Interestingly, we did not find XBP1 WT transcripts accumulating in IRE1α-GFP clusters even after prolonged TM treatment (5 µg/ml for up to 4 hours) and inhibition of IRE1α cleavage activity. XBP1 particles freely diffuse around IRE1α-GFP foci and only very rarely colocalize with the IRE1α-GFP signal (Video 11). This observation was true for both XBP1 WT (in the presence of 4µ8C) as well as unspliceable reporter transcripts (Video 12).

Video 10

Download asset

Video 11

Download asset

Video 12

Download asset

Taken together, our data indicate that Emi1-IRE1α-GFP supports splicing in the absence of foci formation. This suggests that XBP1 mRNA is spliced by lower oligomeric assemblies of IRE1α molecules, which can easily contact ER-associated ribosome-mRNPs, while IRE1α foci or large oligomeric clusters are not the sites of XBP1 mRNA processing during the UPR.

In this study, we present a single-molecule imaging approach that allows visualization of individual XBP1 transcripts and use it to investigate the recruitment of XBP1 mRNA to ER-localized IRE1α, which is a fundamental step of the XBP1 splicing mechanism. Based on previous yeast studies and the visualization of overexpressed IRE1α, splicing of XBP1 mRNA has been suggested to take place in large clusters of IRE1α oligomers that form during the UPR and could function as ER stress response centers (Li et al., 2010). Our findings challenge this view and suggest a different model for mammalian cells, where IRE1α might be recruited to the XBP1 mRNA and not vice versa.

Direct visualization of the recruitment of XBP1 mRNAs to the ER surface using single-molecule imaging revealed that XBP1 molecules become ER-associated in an HR2-dependent manner that is consistent with the targeting model proposed previously (Kanda et al., 2016; Yanagitani et al., 2009; Yanagitani et al., 2011). Furthermore, assessment of the translational status of single mRNA particles demonstrated that their ER association depends on interactions with the ribosome-nascent chain complex. Co-translational membrane tethering therefore immobilizes XBP1 transcripts on the ER surface and hints at a substrate recruitment mechanism where IRE1α diffuses through the ER membrane until it encounters XBP1u mRNAs at the Sec61 translocon. Direct interactions that have been reported for IRE1α and the translocon, SRP, as well as RNAs (Acosta-Alvear et al., 2018; Plumb et al., 2015) further increase the affinity of the interaction and underline the potential significance of such a recruitment mechanism.

Upon induction of ER stress, XBP1 transcripts are spliced and released from the ER surface. However, even though we can derive from our data that ER association correlates with IRE1α cleavage activity, we did not find XBP1 mRNAs colocalizing with IRE1α clusters. Moreover, large, microscopy-visible clusters were only detected when IRE1α-GFP was overexpressed at high levels. Thus, IRE1α foci are not the primary sites of XBP1 splicing. Instead, our findings support a model where ER-associated XBP1 transcripts are processed by small IRE1α oligomers that could dynamically assemble throughout the ER membrane.

These observations are in good agreement with an earlier study, in which the pharmacological activation of IRE1α with the flavonol luteolin promoted strong splicing of XBP1 in the absence of IRE1α clustering (Ricci et al., 2019). In addition, and directly related to our work, a parallel study shows that endogenously tagged IRE1α also fails to assemble into large clusters upon induction of ER stress (Belyy et al., 2021). In this work, the authors characterize IRE1α oligomerization during ER stress and find that the resting pool of IRE1α in the ER membrane is dimeric, while in response to stress transient IRE1α tetramers are assembled as the functional subunits that are required for trans-autophosphorylation and XBP1 splicing. Most likely, such a dynamic equilibrium between dimers and small oligomers allows cells to build timely, fine-tuned responses to local or transient perturbations in ER protein folding.

In combination, our findings suggest a novel mechanism for XBP1 recruitment to functional, dynamic IRE1α assemblies that continuously patrol the ER membrane to encounter substrates that are targeted there.

Following a different strategy, yeast IRE1p foci arrange the recruitment of unspliced HAC1 mRNA to the ER membrane and efficiently localize the mRNA for splicing (Aragón et al., 2009; van Anken et al., 2014). Given the strong conservation of most UPR principles, upon visualization of large IRE1α clusters in human cells (Li et al., 2010; Tran et al., 2021) it was plausible to speculate that polarization of IRE1α might build splicing centers. However, these studies mostly relied on overexpression of ectopic IRE1α, which likely contributed to the perception that clusters were required for XBP1 splicing and explains the discrepancies between this and previous reports.

Our findings shed light on a fundamental step of the XBP1 splicing mechanism, which is the recruitment of IRE1α to ER-localized XBP1 transcripts. Yet, several open questions remain:

1. Are XBP1 mRNAs that are tethered to the ER surface as part of ribosome-nascent chain complexes continuously translated? Or is translation stably stalled while the mRNA remains poised for recruitment by IRE1α? And if so, how is translation resumed? And how relevant is the translational status of XBP1 mRNA for splicing?

2. Does IRE1α also bind XBP1 transcripts in the absence of the ribosome/translocon interaction? Since we observe a low degree of splicing for HR2-mutant reporters, we speculate that translation-independent recruitment of XBP1 transcripts and recognition through IRE1α might also be possible.

3. How does IRE1α discriminate between its distinct substrates? Beyond XBP1 splicing, IRE1α processes a broad range of substrates including RIDD mRNAs (Hollien et al., 2009; Hollien and Weissman, 2006) and a recently described, larger group of mRNAs that are processed through an unanticipated mode of cleavage with looser specificity (RIDDLE) (Le Thomas et al., 2021). Most of these mRNA substrates encode signal sequence-containing proteins and are delivered to the ER by SRP. While we know that activation of the IRE1α RNAse domain requires IRE1α dimerization/oligomerization as well as trans-autophosphorylation, the specific role and substrate specificity of the distinct assemblies remain unclear. It is tempting to speculate that the interplay of different oligomeric IRE1α assemblies with the translocon/ribosome/SRP environment may define the code for selective processing of distinct IRE1α substrates, avoiding the detrimental cleavage that might result from unrestrained RNA degradation.

In summary, our data have allowed us to uncover unanticipated features of one of the key steps of UPR initiation, the encounter of XBP1 mRNA with IRE1α to undergo splicing. Additional studies will be needed to further dissect the underlying mechanisms behind the regulation of IRE1α activity in homeostasis and disease.

All imaging data generated in this study have been uploaded to Zenodo.

1. 1. Gómez-Puerta S
   2. Ferrero R
   3. Hochstoeger T
   4. Zubiri I
   5. Chao J A
   6. Aragón T
   7. Voigt F

   (2022) Zenodo

   Live imaging data of XBP1 mRNA under control conditions.

   https://doi.org/10.5281/zenodo.6559201
2. 1. Gómez-Puerta S
   2. Ferrero R
   3. Hochstoeger T
   4. Zubiri I
   5. Chao J A
   6. Aragón T
   7. Voigt F

   (2022) Zenodo

   Live imaging data of XBP1 mRNA under ER stress conditions.

   https://doi.org/10.5281/zenodo.6557623
3. 1. Gómez-Puerta S
   2. Ferrero R
   3. Hochstoeger T
   4. Zubiri I
   5. Chao J A
   6. Aragón T
   7. Voigt F

   (2022) Zenodo

   Live imaging data of XBP1 mRNA under ER stress and IRE1a inhibition.

   https://doi.org/10.5281/zenodo.6559058
4. 1. Gómez-Puerta S
   2. Ferrero R
   3. Hochstoeger T
   4. Zubiri I
   5. Chao J A
   6. Aragón T
   7. Voigt F

   (2022) Zenodo

   Live cell imaging data of IRE1a-GFP foci formation.

   https://doi.org/10.5281/zenodo.6559125
5. 1. Gómez-Puerta S
   2. Ferrero R
   3. Hochstoeger T
   4. Zubiri I
   5. Chao J A
   6. Aragón T
   7. Voigt F

   (2022) Zenodo

   Fixed cell imaging data of XBP1 mRNA.

   https://doi.org/10.5281/zenodo.6559109

1. 1. Acosta-Alvear D
   2. Zhou Y
   3. Blais A
   4. Tsikitis M
   5. Lents NH
   6. Arias C
   7. Lennon CJ
   8. Kluger Y
   9. Dynlacht BD

   (2007) XBP1 controls diverse cell type- and condition-specific transcriptional regulatory networks

   Molecular Cell 27:53–66.

   https://doi.org/10.1016/j.molcel.2007.06.011

   • PubMed
   • Google Scholar
2. 1. Acosta-Alvear D
   2. Karagöz GE
   3. Fröhlich F
   4. Li H
   5. Walther TC
   6. Walter P

   (2018) The unfolded protein response and endoplasmic reticulum protein targeting machineries converge on the stress sensor IRE1

   eLife 7:e43036.

   https://doi.org/10.7554/eLife.43036

   • PubMed
   • Google Scholar
3. 1. Ali MMU
   2. Bagratuni T
   3. Davenport EL
   4. Nowak PR
   5. Silva-Santisteban MC
   6. Hardcastle A
   7. McAndrews C
   8. Rowlands MG
   9. Morgan GJ
   10. Aherne W
   11. Collins I
   12. Davies FE
   13. Pearl LH

   (2011) Structure of the Ire1 autophosphorylation complex and implications for the unfolded protein response

   The EMBO Journal 30:894–905.

   https://doi.org/10.1038/emboj.2011.18

   • PubMed
   • Google Scholar
4. 1. Aragón T
   2. van Anken E
   3. Pincus D
   4. Serafimova IM
   5. Korennykh AV
   6. Rubio CA
   7. Walter P

   (2009) Messenger RNA targeting to endoplasmic reticulum stress signalling sites

   Nature 457:736–740.

   https://doi.org/10.1038/nature07641

   • PubMed
   • Google Scholar
5. 1. Bakunts A
   2. Orsi A
   3. Vitale M
   4. Cattaneo A
   5. Lari F
   6. Tadè L
   7. Sitia R
   8. Raimondi A
   9. Bachi A
   10. van Anken E

   (2017) Ratiometric sensing of BiP-client versus BiP levels by the unfolded protein response determines its signaling amplitude

   eLife 6:e27518.

   https://doi.org/10.7554/eLife.27518

   • PubMed
   • Google Scholar
6. 1. Belyy V
   2. Tran NH
   3. Walter P

   (2020) Quantitative microscopy reveals dynamics and fate of clustered IRE1α

   PNAS 117:1533–1542.

   https://doi.org/10.1073/pnas.1915311117

   • PubMed
   • Google Scholar
7. Preprint

   1. Belyy V
   2. Zuazo-Gaztelu I
   3. Alamban A
   4. Ashkenazi A
   5. Walter P

   (2021) Endoplasmic Reticulum Stress Activates Human IRE1α through Reversible Assembly of Inactive Dimers into Small Oligomers

   bioRxiv.

   https://doi.org/10.1101/2021.09.29.462487

   • Google Scholar
8. 1. Berg HC

   (1993) Berg: Diffusion: Microscopic theory in “Random Google scholar

   Random Walk Models in Biology 5:813–824.

   https://doi.org/10.1098/rsif.2008.0014

   • Google Scholar
9. 1. Berg S
   2. Kutra D
   3. Kroeger T
   4. Straehle CN
   5. Kausler BX
   6. Haubold C
   7. Schiegg M
   8. Ales J
   9. Beier T
   10. Rudy M
   11. Eren K
   12. Cervantes JI
   13. Xu B
   14. Beuttenmueller F
   15. Wolny A
   16. Zhang C
   17. Koethe U
   18. Hamprecht FA
   19. Kreshuk A

   (2019) ilastik: interactive machine learning for (bio)image analysis

   Nature Methods 16:1226–1232.

   https://doi.org/10.1038/s41592-019-0582-9

   • PubMed
   • Google Scholar
10. 1. Berthold MR
    2. Cebron N
    3. Dill F
    4. Gabriel TR
    5. Kötter T
    6. Meinl T
    7. Ohl P
    8. Thiel K
    9. Wiswedel B

    (2009) KNIME - the Konstanz information miner: Version 2.0 and beyond

    ACM SIGKDD Explorations Newsletter 11:26–31.

    https://doi.org/10.1145/1656274.1656280

    • Google Scholar
11. 1. Bertrand E
    2. Chartrand P
    3. Schaefer M
    4. Shenoy SM
    5. Singer RH
    6. Long RM

    (1998) Localization of ASH1 mRNA particles in living yeast

    Molecular Cell 2:437–445.

    https://doi.org/10.1016/s1097-2765(00)80143-4

    • PubMed
    • Google Scholar
12. 1. Brodsky JL

    (2012) Cleaning up: ER-associated degradation to the rescue

    Cell 151:1163–1167.

    https://doi.org/10.1016/j.cell.2012.11.012

    • PubMed
    • Google Scholar
13. 1. Calfon M
    2. Zeng H
    3. Urano F
    4. Till JH
    5. Hubbard SR
    6. Harding HP
    7. Clark SG
    8. Ron D

    (2002) IRE1 couples endoplasmic reticulum load to secretory capacity by processing the XBP-1 mRNA

    Nature 415:92–96.

    https://doi.org/10.1038/415092a

    • PubMed
    • Google Scholar
14. 1. Cox JS
    2. Shamu CE
    3. Walter P

    (1993) Transcriptional induction of genes encoding endoplasmic reticulum resident proteins requires a transmembrane protein kinase

    Cell 73:1197–1206.

    https://doi.org/10.1016/0092-8674(93)90648-a

    • PubMed
    • Google Scholar
15. 1. Crameri A
    2. Whitehorn EA
    3. Tate E
    4. Stemmer WP

    (1996) Improved green fluorescent protein by molecular evolution using DNA shuffling

    Nature Biotechnology 14:315–319.

    https://doi.org/10.1038/nbt0396-315

    • PubMed
    • Google Scholar
16. 1. Cross BCS
    2. Bond PJ
    3. Sadowski PG
    4. Jha BK
    5. Zak J
    6. Goodman JM
    7. Silverman RH
    8. Neubert TA
    9. Baxendale IR
    10. Ron D
    11. Harding HP

    (2012) The molecular basis for selective inhibition of unconventional mRNA splicing by an IRE1-binding small molecule

    PNAS 109:E869–E878.

    https://doi.org/10.1073/pnas.1115623109

    • PubMed
    • Google Scholar
17. 1. Dave P
    2. Griesbach E
    3. Roth G
    4. Mateju D
    5. Chao JA

    (2021) Single-Molecule Imaging Reveals the Coupling of Translation and mRNA Decay

    SSRN Electronic Journal 1:5982.

    https://doi.org/10.2139/ssrn.3885982

    • Google Scholar
18. 1. Eichenberger BT
    2. Griesbach E
    3. Chao JA

    (2018) Imaging Translation Dynamics of Single mRNA Molecules in Live Cells

    Methods of Molecular Biology 1:385–404.

    https://doi.org/10.1007/978-1-4939-7213-5_26

    • Google Scholar
19. 1. Gilmore R
    2. Blobel G
    3. Walter P

    (1982) Protein translocation across the endoplasmic reticulum. I. Detection in the microsomal membrane of a receptor for the signal recognition particle

    The Journal of Cell Biology 95:463–469.

    https://doi.org/10.1083/jcb.95.2.463

    • PubMed
    • Google Scholar
20. Software

    1. Gómez-Puerta S
    2. Ferrero R
    3. Hochstoeger T
    4. Zubiri I
    5. Chao JA
    6. Aragón T
    7. Voigt F

    (2022a) Live imaging data of XBP1 mRNA under control conditions

    Zenodo.

    https://doi.org/10.5281/zenodo.6559201
21. Software

    1. Gómez-Puerta S
    2. Ferrero R
    3. Hochstoeger T
    4. Zubiri I
    5. Chao JA
    6. Aragón T
    7. Voigt F

    (2022b) Live imaging data of XBP1 mRNA under ER stress conditions

    Zenodo.

    https://doi.org/10.5281/zenodo.6557623
22. Software

    1. Gómez-Puerta S
    2. Ferrero R
    3. Hochstoeger T
    4. Zubiri I
    5. Chao JA
    6. Aragón T
    7. Voigt F

    (2022c) Live imaging data of XBP1 mRNA under ER stress and IRE1a inhibition

    Zenodo.

    https://doi.org/10.5281/zenodo.6559058
23. Software

    1. Gómez-Puerta S
    2. Ferrero R
    3. Hochstoeger T
    4. Zubiri I
    5. Chao JA
    6. Aragón T
    7. Voigt F

    (2022d) Dataset Open AccessLive cell imaging data of IRE1a-GFP foci formation

    Zenodo.

    https://doi.org/10.5281/zenodo.6559109
24. Software

    1. Gómez-Puerta S
    2. Ferrero R
    3. Hochstoeger T
    4. Zubiri I
    5. Chao JA
    6. Aragón T
    7. Voigt F

    (2022e) Fixed cell imaging data of XBP1 mRNA

    Zenodo.

    https://doi.org/10.5281/zenodo.6559125
25. 1. Gonzalez TN
    2. Sidrauski C
    3. Dörfler S
    4. Walter P

    (1999) Mechanism of non-spliceosomal mRNA splicing in the unfolded protein response pathway

    The EMBO Journal 18:3119–3132.

    https://doi.org/10.1093/emboj/18.11.3119

    • PubMed
    • Google Scholar
26. 1. Görlich D
    2. Prehn S
    3. Hartmann E
    4. Kalies KU
    5. Rapoport TA

    (1992) A mammalian homolog of SEC61p and SECYp is associated with ribosomes and nascent polypeptides during translocation

    Cell 71:489–503.

    https://doi.org/10.1016/0092-8674(92)90517-g

    • PubMed
    • Google Scholar
27. 1. Grimm JB
    2. English BP
    3. Chen J
    4. Slaughter JP
    5. Zhang Z
    6. Revyakin A
    7. Patel R
    8. Macklin JJ
    9. Normanno D
    10. Singer RH
    11. Lionnet T
    12. Lavis LD

    (2015) A general method to improve fluorophores for live-cell and single-molecule microscopy

    Nature Methods 12:244–250.

    https://doi.org/10.1038/nmeth.3256

    • PubMed
    • Google Scholar
28. 1. Hollien J
    2. Weissman JS

    (2006) Decay of endoplasmic reticulum-localized mRNAs during the unfolded protein response

    Science (New York, N.Y.) 313:104–107.

    https://doi.org/10.1126/science.1129631

    • PubMed
    • Google Scholar
29. 1. Hollien J
    2. Lin JH
    3. Li H
    4. Stevens N
    5. Walter P
    6. Weissman JS

    (2009) Regulated Ire1-dependent decay of messenger RNAs in mammalian cells

    The Journal of Cell Biology 186:323–331.

    https://doi.org/10.1083/jcb.200903014

    • PubMed
    • Google Scholar
30. 1. Hooks KB
    2. Griffiths-Jones S

    (2011) Conserved RNA structures in the non-canonical Hac1/Xbp1 intron

    RNA Biology 8:552–556.

    https://doi.org/10.4161/rna.8.4.15396

    • PubMed
    • Google Scholar
31. 1. Ishikawa T
    2. Kashima M
    3. Nagano AJ
    4. Ishikawa-Fujiwara T
    5. Kamei Y
    6. Todo T
    7. Mori K

    (2017) Unfolded protein response transducer IRE1-mediated signaling independent of XBP1 mRNA splicing is not required for growth and development of medaka fish

    eLife 6:e26845.

    https://doi.org/10.7554/eLife.26845

    • PubMed
    • Google Scholar
32. 1. Jurkin J
    2. Henkel T
    3. Nielsen AF
    4. Minnich M
    5. Popow J
    6. Kaufmann T
    7. Heindl K
    8. Hoffmann T
    9. Busslinger M
    10. Martinez J

    (2014) The mammalian tRNA ligase complex mediates splicing of XBP1 mRNA and controls antibody secretion in plasma cells

    The EMBO Journal 33:2922–2936.

    https://doi.org/10.15252/embj.201490332

    • PubMed
    • Google Scholar
33. 1. Kanda S
    2. Yanagitani K
    3. Yokota Y
    4. Esaki Y
    5. Kohno K

    (2016) Autonomous translational pausing is required for XBP1u mRNA recruitment to the ER via the SRP pathway

    PNAS 113:E5886–E5895.

    https://doi.org/10.1073/pnas.1604435113

    • PubMed
    • Google Scholar
34. 1. Karagöz GE
    2. Acosta-Alvear D
    3. Nguyen HT
    4. Lee CP
    5. Chu F
    6. Walter P

    (2017) An unfolded protein-induced conformational switch activates mammalian IRE1

    eLife 6:e30700.

    https://doi.org/10.7554/eLife.30700

    • PubMed
    • Google Scholar
35. 1. Kimata Y
    2. Ishiwata-Kimata Y
    3. Ito T
    4. Hirata A
    5. Suzuki T
    6. Oikawa D
    7. Takeuchi M
    8. Kohno K

    (2007) Two regulatory steps of ER-stress sensor Ire1 involving its cluster formation and interaction with unfolded proteins

    The Journal of Cell Biology 179:75–86.

    https://doi.org/10.1083/jcb.200704166

    • PubMed
    • Google Scholar
36. 1. Korennykh AV
    2. Egea PF
    3. Korostelev AA
    4. Finer-Moore J
    5. Zhang C
    6. Shokat KM
    7. Stroud RM
    8. Walter P

    (2009) The unfolded protein response signals through high-order assembly of Ire1

    Nature 457:687–693.

    https://doi.org/10.1038/nature07661

    • PubMed
    • Google Scholar
37. 1. Kosmaczewski SG
    2. Edwards TJ
    3. Han SM
    4. Eckwahl MJ
    5. Meyer BI
    6. Peach S
    7. Hesselberth JR
    8. Wolin SL
    9. Hammarlund M

    (2014) The RtcB RNA ligase is an essential component of the metazoan unfolded protein response

    EMBO Reports 15:1278–1285.

    https://doi.org/10.15252/embr.201439531

    • PubMed
    • Google Scholar
38. Preprint

    1. Le Thomas A
    2. Ferri E
    3. Marsters S
    4. Harnoss JM
    5. Modrusan Z
    6. Li W
    7. Rudolph J
    8. Wang W
    9. Wu TD
    10. Walter P
    11. Ashkenazi A

    (2021) Noncanonical MRNA Decay by the Endoplasmic-Reticulum Stress Sensor IRE1α Promotes Cancer-Cell Survival

    bioRxiv.

    https://doi.org/10.1101/2021.03.16.435520

    • Google Scholar
39. 1. Li H
    2. Korennykh AV
    3. Behrman SL
    4. Walter P

    (2010) Mammalian endoplasmic reticulum stress sensor IRE1 signals by dynamic clustering

    PNAS 107:16113–16118.

    https://doi.org/10.1073/pnas.1010580107

    • PubMed
    • Google Scholar
40. 1. Lu Y
    2. Liang FX
    3. Wang X

    (2014) A synthetic biology approach identifies the mammalian UPR RNA ligase RtcB

    Molecular Cell 55:758–770.

    https://doi.org/10.1016/j.molcel.2014.06.032

    • PubMed
    • Google Scholar
41. 1. Mateju D
    2. Eichenberger B
    3. Voigt F
    4. Eglinger J
    5. Roth G
    6. Chao JA

    (2020) Single-Molecule Imaging Reveals Translation of mRNAs Localized to Stress Granules

    Cell 183:1801–1812.

    https://doi.org/10.1016/j.cell.2020.11.010

    • PubMed
    • Google Scholar
42. 1. Mechler B
    2. Rabbitts TH

    (1981) Membrane-bound ribosomes of myeloma cells. IV. mRNA complexity of free and membrane-bound polysomes

    The Journal of Cell Biology 88:29–36.

    https://doi.org/10.1083/jcb.88.1.29

    • PubMed
    • Google Scholar
43. 1. Mori K
    2. Ma W
    3. Gething MJ
    4. Sambrook J

    (1993) A transmembrane protein with A cdc2+/CDC28-related kinase activity is required for signaling from the ER to the nucleus

    Cell 74:743–756.

    https://doi.org/10.1016/0092-8674(93)90521-q

    • PubMed
    • Google Scholar
44. 1. Morisaki T
    2. Lyon K
    3. DeLuca KF
    4. DeLuca JG
    5. English BP
    6. Zhang Z
    7. Lavis LD
    8. Grimm JB
    9. Viswanathan S
    10. Looger LL
    11. Lionnet T
    12. Stasevich TJ

    (2016) Real-time quantification of single RNA translation dynamics in living cells

    Science (New York, N.Y.) 352:1425–1429.

    https://doi.org/10.1126/science.aaf0899

    • PubMed
    • Google Scholar
45. 1. Pakos-Zebrucka K
    2. Koryga I
    3. Mnich K
    4. Ljujic M
    5. Samali A
    6. Gorman AM

    (2016) The integrated stress response

    EMBO Reports 17:1374–1395.

    https://doi.org/10.15252/embr.201642195

    • PubMed
    • Google Scholar
46. 1. Pincus D
    2. Chevalier MW
    3. Aragón T
    4. van Anken E
    5. Vidal SE
    6. El-Samad H
    7. Walter P

    (2010) BiP binding to the ER-stress sensor Ire1 tunes the homeostatic behavior of the unfolded protein response

    PLOS Biology 8:e1000415.

    https://doi.org/10.1371/journal.pbio.1000415

    • PubMed
    • Google Scholar
47. 1. Plumb R
    2. Zhang ZR
    3. Appathurai S
    4. Mariappan M

    (2015) A functional link between the co-translational protein translocation pathway and the UPR

    eLife 4:53.

    https://doi.org/10.7554/eLife.07426

    • PubMed
    • Google Scholar
48. 1. Preibisch S
    2. Saalfeld S
    3. Schindelin J
    4. Tomancak P

    (2010) Software for bead-based registration of selective plane illumination microscopy data

    Nature Methods 7:418–419.

    https://doi.org/10.1038/nmeth0610-418

    • PubMed
    • Google Scholar
49. 1. Ricci D
    2. Marrocco I
    3. Blumenthal D
    4. Dibos M
    5. Eletto D
    6. Vargas J
    7. Boyle S
    8. Iwamoto Y
    9. Chomistek S
    10. Paton JC
    11. Paton AW
    12. Argon Y

    (2019) Clustering of IRE1α depends on sensing ER stress but not on its RNase activity

    FASEB Journal 33:9811–9827.

    https://doi.org/10.1096/fj.201801240RR

    • PubMed
    • Google Scholar
50. 1. Rüegsegger U
    2. Leber JH
    3. Walter P

    (2001) Block of HAC1 mRNA translation by long-range base pairing is released by cytoplasmic splicing upon induction of the unfolded protein response

    Cell 107:103–114.

    https://doi.org/10.1016/s0092-8674(01)00505-0

    • PubMed
    • Google Scholar
51. 1. Sanches M
    2. Duffy NM
    3. Talukdar M
    4. Thevakumaran N
    5. Chiovitti D
    6. Canny MD
    7. Lee K
    8. Kurinov I
    9. Uehling D
    10. Al-awar R
    11. Poda G
    12. Prakesch M
    13. Wilson B
    14. Tam V
    15. Schweitzer C
    16. Toro A
    17. Lucas JL
    18. Vuga D
    19. Lehmann L
    20. Durocher D
    21. Zeng Q
    22. Patterson JB
    23. Sicheri F

    (2014) Structure and mechanism of action of the hydroxy-aryl-aldehyde class of IRE1 endoribonuclease inhibitors

    Nature Communications 5:4202.

    https://doi.org/10.1038/ncomms5202

    • PubMed
    • Google Scholar
52. 1. Schindelin J
    2. Arganda-Carreras I
    3. Frise E
    4. Kaynig V
    5. Longair M
    6. Pietzsch T
    7. Preibisch S
    8. Rueden C
    9. Saalfeld S
    10. Schmid B
    11. Tinevez JY
    12. White DJ
    13. Hartenstein V
    14. Eliceiri K
    15. Tomancak P
    16. Cardona A

    (2012) Fiji: an open-source platform for biological-image analysis

    Nature Methods 9:676–682.

    https://doi.org/10.1038/nmeth.2019

    • PubMed
    • Google Scholar
53. 1. Sidrauski C
    2. Walter P

    (1997) The transmembrane kinase Ire1p is a site-specific endonuclease that initiates mRNA splicing in the unfolded protein response

    Cell 90:1031–1039.

    https://doi.org/10.1016/s0092-8674(00)80369-4

    • PubMed
    • Google Scholar
54. 1. Sugimoto Y
    2. Vigilante A
    3. Darbo E
    4. Zirra A
    5. Militti C
    6. D’Ambrogio A
    7. Luscombe NM
    8. Ule J

    (2015) hiCLIP reveals the in vivo atlas of mRNA secondary structures recognized by Staufen 1

    Nature 519:491–494.

    https://doi.org/10.1038/nature14280

    • PubMed
    • Google Scholar
55. 1. Tinevez JY
    2. Perry N
    3. Schindelin J
    4. Hoopes GM
    5. Reynolds GD
    6. Laplantine E
    7. Bednarek SY
    8. Shorte SL
    9. Eliceiri KW

    (2017) TrackMate: An open and extensible platform for single-particle tracking

    Methods (San Diego, Calif.) 115:80–90.

    https://doi.org/10.1016/j.ymeth.2016.09.016

    • PubMed
    • Google Scholar
56. 1. Tirasophon W
    2. Welihinda AA
    3. Kaufman RJ

    (1998) A stress response pathway from the endoplasmic reticulum to the nucleus requires A novel bifunctional protein kinase/endoribonuclease (Ire1p) in mammalian cells

    Genes & Development 12:1812–1824.

    https://doi.org/10.1101/gad.12.12.1812

    • PubMed
    • Google Scholar
57. 1. Tran N-H
    2. Carter SD
    3. De Mazière A
    4. Ashkenazi A
    5. Klumperman J
    6. Walter P
    7. Jensen GJ

    (2021) The stress-sensing domain of activated IRE1α forms helical filaments in narrow ER membrane tubes

    Science (New York, N.Y.) 374:52–57.

    https://doi.org/10.1126/science.abh2474

    • PubMed
    • Google Scholar
58. 1. van Anken E
    2. Pincus D
    3. Coyle S
    4. Aragón T
    5. Osman C
    6. Lari F
    7. Gómez Puerta S
    8. Korennykh AV
    9. Walter P

    (2014) Specificity in endoplasmic reticulum-stress signaling in yeast entails a step-wise engagement of HAC1 mRNA to clusters of the stress sensor Ire1

    eLife 3:e05031.

    https://doi.org/10.7554/eLife.05031

    • PubMed
    • Google Scholar
59. 1. Voigt F
    2. Zhang H
    3. Cui XA
    4. Triebold D
    5. Liu AX
    6. Eglinger J
    7. Lee ES
    8. Chao JA
    9. Palazzo AF

    (2017) Single-Molecule Quantification of Translation-Dependent Association of mRNAs with the Endoplasmic Reticulum

    Cell Reports 21:3740–3753.

    https://doi.org/10.1016/j.celrep.2017.12.008

    • PubMed
    • Google Scholar
60. 1. Voigt F
    2. Eglinger J
    3. Chao JA

    (2019a) Quantification of mRNA Turnover in Living Cells: A Pipeline for TREAT Data Analysis

    Methods in Molecular Biology (Clifton, N.J.) 2038:75–88.

    https://doi.org/10.1007/978-1-4939-9674-2_6

    • PubMed
    • Google Scholar
61. 1. Voigt F
    2. Gerbracht JV
    3. Boehm V
    4. Horvathova I
    5. Eglinger J
    6. Chao JA
    7. Gehring NH

    (2019b) Detection and quantification of RNA decay intermediates using XRN1-resistant reporter transcripts

    Nature Protocols 14:1603–1633.

    https://doi.org/10.1038/s41596-019-0152-8

    • PubMed
    • Google Scholar
62. 1. Walter P
    2. Ibrahimi I
    3. Blobel G

    (1981) Translocation of proteins across the endoplasmic reticulum. I. Signal recognition protein (SRP) binds to in-vitro-assembled polysomes synthesizing secretory protein

    The Journal of Cell Biology 91:545–550.

    https://doi.org/10.1083/jcb.91.2.545

    • PubMed
    • Google Scholar
63. 1. Walter P
    2. Ron D

    (2011) The unfolded protein response: from stress pathway to homeostatic regulation

    Science (New York, N.Y.) 334:1081–1086.

    https://doi.org/10.1126/science.1209038

    • PubMed
    • Google Scholar
64. 1. Weidenfeld I
    2. Gossen M
    3. Löw R
    4. Kentner D
    5. Berger S
    6. Görlich D
    7. Bartsch D
    8. Bujard H
    9. Schönig K

    (2009) Inducible expression of coding and inhibitory RNAs from retargetable genomic loci

    Nucleic Acids Research 37:e50.

    https://doi.org/10.1093/nar/gkp108

    • PubMed
    • Google Scholar
65. 1. Wu B
    2. Miskolci V
    3. Sato H
    4. Tutucci E
    5. Kenworthy CA
    6. Donnelly SK
    7. Yoon YJ
    8. Cox D
    9. Singer RH
    10. Hodgson L

    (2015) Synonymous modification results in high-fidelity gene expression of repetitive protein and nucleotide sequences

    Genes & Development 29:876–886.

    https://doi.org/10.1101/gad.259358.115

    • PubMed
    • Google Scholar
66. 1. Yan X
    2. Hoek TA
    3. Vale RD
    4. Tanenbaum ME

    (2016) Dynamics of Translation of Single mRNA Molecules In Vivo

    Cell 165:976–989.

    https://doi.org/10.1016/j.cell.2016.04.034

    • PubMed
    • Google Scholar
67. 1. Yanagitani K
    2. Imagawa Y
    3. Iwawaki T
    4. Hosoda A
    5. Saito M
    6. Kimata Y
    7. Kohno K

    (2009) Cotranslational targeting of XBP1 protein to the membrane promotes cytoplasmic splicing of its own mRNA

    Molecular Cell 34:191–200.

    https://doi.org/10.1016/j.molcel.2009.02.033

    • PubMed
    • Google Scholar
68. 1. Yanagitani K
    2. Kimata Y
    3. Kadokura H
    4. Kohno K

    (2011) Translational pausing ensures membrane targeting and cytoplasmic splicing of XBP1u mRNA

    Science (New York, N.Y.) 331:586–589.

    https://doi.org/10.1126/science.1197142

    • PubMed
    • Google Scholar
69. 1. Yoshida H
    2. Matsui T
    3. Yamamoto A
    4. Okada T
    5. Mori K

    (2001) XBP1 mRNA is induced by ATF6 and spliced by IRE1 in response to ER stress to produce a highly active transcription factor

    Cell 107:881–891.

    https://doi.org/10.1016/s0092-8674(01)00611-0

    • PubMed
    • Google Scholar

Decision letter after peer review:

[Editors’ note: the authors submitted for reconsideration following the decision after peer review. What follows is the decision letter after the first round of review.]

Thank you for submitting the paper "Live imaging of the co-translational recruitment of XBP1 mRNA to the ER and its processing by diffuse, non-polarized IRE1α" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and a Senior Editor. The reviewers have opted to remain anonymous.

Comments to the Authors:

While there was significant interest in the topic and preliminarily findings, we are sorry to say that, after consultation with the reviewers, we have decided that this work will not be considered further for publication by eLife.

As you will read below, whilst all three reviewers concur with the importance of the paper's subject matter they independently concluded that the study would need significant additional work to become suitable for publication. The reviewers have laid out a non-trivial number of additional controls and experiments that are considered essential. In brief, the absence of these experiments indicates that many of the claims currently lack experimental support to the level required by ELife. More specifically, the clustering analysis is of limited scope and significance, temporal aspects of the observed events are absent, the tools used for this study need to be better validated, and single particle analysis was--in contrast to what was stated--not performed rigorously. There is also a consensus amongst the reviewers that a substantial portion of the paper is simply confirmatory based on prior work performed by the Kohno and Argon labs.

Reviewer #1 (Recommendations for the authors):

In this study, the authors explore key steps in the mammalian unfolded protein response (UPR). They focus on the IRE1a branch, specifically, the recruitment of target mRNA XBP1 to the endoplasmic reticulum (ER) surface, the fate of the mRNA, and organization of IRE1a, itself, in response to unfolded protein stress. The novelty of this study is the use of live cell imaging techniques to follow XBP1 mRNA localization and dynamics. The translational recruitment model of XBP1 to the ER membrane surface has been previously described by others and is confirmed in live cells, here. Different reporters were used to visualize localization and translation of spliced and unspliced forms of XBP1. The authors make additional claims regarding the stress-induced clustering behavior of IRE1a. Overall, this study adds visual data, with some dynamics information to potentially help refine models of steps of XBP1 mRNA trafficking and splicing in response to ER stressors. The novelty of the study approach is undercut by the need for more thorough characterization of the reporters, better descriptions of the reagents, and more accurate descriptions of experiments and results. With the additional requested details and modifications to the text, this manuscript could be a useful addition to the ER stress literature and of broader interest to groups interested in trafficking and translation on the ER membrane.

The authors make two claims that are well supported by imaging data, mutants, and biochemical fractionation.

Splicing activity by IRE1a or the presence or expression of XBP1 message in an already spliced form decreases the fraction of XBP1 associated with the ER. These observations are supported by membrane fractionation experiments.

The authors have confirmed that stable ER association of unspliced XBP1 mRNA is established through HR2-dependent targeting and relies on active translation. The experimental data support this claim.

Other claims are less well supported or less accurate.

First, the authors claim to "directly monitor recruitment of individual XBP1 transcripts to the ER surface." I was unable to find a single movie or data series demonstrating the movement of a single particle from the cytoplasm to the ER membrane. Experiments show mRNAs on the ER or report amount of cytoplasmic vs. membrane bound. The results, presumably, of recruitment are observed and reported, not the process.

The authors state "Next, we quantified the mobility of individual particles with respect to their ER localization and therefore assessed when an mRNA particle associates with the ER." What appears to be assessed is whether a particle is associated with the ER, not "when." I assume the authors did not mean "when" as part of temporal order, but it would be less confusing to claim to assess whether or if.

The authors claim to show that IRE1a-catalyzed splicing mobilizes XBP1 mRNA from the ER membrane in response to ER stress. Here the data are mostly supportive of the claim. However, what the authors actually show is that splicing is necessary for an increase in XBP1 mRNA localization in the cytoplasm. No data are presented showing interactions with IRE1a and/or movement of XBP1 mRNA from the ER membrane to the cytoplasm. Based on multiple studies, the interpretation is likely correct, but the microscopy data do not show the process.

The matter of IRE1a clustering is more problematic.

The authors state, "Surprisingly, we find that XBP1 transcripts are not recruited into large IRE1a clusters, which only assemble upon overexpression of fluorescently-tagged IRE1a during ER stress. Our findings support a model where ribosome-engaged, ER-poised XBP1 mRNA is processed by functional IRE1a assemblies that are homogenously distributed throughout the ER membrane."

The authors state, "ER-poised XBP1 transcripts are processed by functional IRE1a assemblies that are homogenously distributed throughout the ER membrane." No data are provided on what fraction of the IRE1a is in an active state. There is no method used to detect the active state or information on the spatial distribution of stress-activated IRE1a. This claim is not supported.

"Our data have allowed us to visualize and uncover unanticipated features of one of the key steps of UPR initiation, the encounter of XBP1 mRNA with IRE1a to undergo splicing." This is another claim that is not quite accurate. The authors observed tagged XBP1 mRNA dynamics in co-expressing cells treated with an inhibitor of IRE1a activity. An encounter in which splicing could actually occur is never presented. Whether XBP1 mRNA is stimulated to dissociate from the ER membrane following an encounter with IRE1a-GFP is not presented. The ability of drug inhibited IRE1a-GFP to bind XBP1 mRNA is not tested or reported.

The authors have made a significant point of weighing in on the matter of IRE1a clustering in both the abstract and text. While their findings are potentially interesting, the investigation of this matter has not been sufficiently rigorous. First, the authors simply describe the attached fluorescent protein as a GFP. Is this monomeric GFP? Emerald? mNeonGreen? or? With the wide variety of fluorescent proteins available, the minimal expectation is to state what was inserted and imaged. Given the potential of many fluorescent proteins to oligomerize, the choice of fluorescent protein matters.

Second, the systems used may be inducing some constitutive stress. For example, in 2C, there appears to be splicing of the reporter, even in the absence of thapsigargin or expression of IRE1-fluorescent protein. Induction with Dox is for 15 hours. That's long enough for stimulation of a stress response and perhaps even some degree of adaptation. I'd like to see some other assays of ER stress before and during induction up to the time normally used to apply stress. Similarly, in the Ire1a-KO cells in 4B, there appears to be spliced XBP1, with and without stress. Importantly, the ratio of spliced XBP1 for the emi1 variant is much lower than for wt, even though immunoblotting suggests there is significantly more EMI-1-IRE1a-GFP protein present than for wt IRE1a. Please comment. Is the GFP tagged variant less active? Is it possible that the GFP tagged variant is less activatable?

More fundamentally, the authors claim that large IRE1a clusters… only assemble upon overexpression of fluorescently-tagged IRE1a during ER stress. At the very least, the more accurate thing would be to say fluorescently-tagged IRE1a only forms inducible observable clusters under overexpression conditions. In this study, there has been no testing of oligomeric/cluster status of native IRE1a. There has been no use of other assays to determine if low expressed IRE1a-fluorescent protein forms oligomers of any size during stress. Nor has there been investigation of whether a version with a small epitope tag might form clusters. What about the ability of other stress conditions, such as acute thapsigargin or DTT, for the ability to induce observable clustering for the lower expressed IRE1a-GFP? Ultimately, the clustering analysis has been too limited to draw any major conclusions. The ability to cluster and its relevance are important to popular models of IRE1a activity. A nuanced study of the clustering regulation of IRE1a would be immensely useful to the field. The observation that clustering observed in mammalian cells is stress inducible, suggests that observable clustering is probably reporting on some aspect of IRE1 regulation. Clusters may represent an extreme manifestation of normal physiology. That is, even an artifact of expression does not rule out the potential importance of forming small clusters in the normal stress response. Note that the overexpressed IRE1a-GFP is constitutively maximally active. There is no stress induced increase in splicing activity, yet there is a stress-inducible change in distribution. This redistribution could reflect some sort of stress inducible regulatory step, for example sequestering of excess unengaged IRE1a, inactivated IRE1a or something else. Whatever causes visible clustering seems to be more than a simple artifact.

It would be helpful for the authors to compare the functionality of their constructs/system relative to untagged versions. Specifically, rates of processing. While I think the authors have demonstrated reasonably well that their XBP1 MS2 reporter does get targeted to the ER and spliced in response to ER stress, it would be useful to know if the 24 MS2 sites and attached MCS-GFP reporters (5nm+ diameter for each) affect rates of XBP1 mRNA processing. Is it possible that the construct is so large that it is relatively immobilized on the ER membrane by sterics and/or sheer size? Could the size of the reporter slow the approach of IRE1a to the cleavage site? Does high expression of the XBP1 substrate impact the efficiency of processing? That is, if client is saturating, then encountering and splicing by activated IRE1a on the ER membrane could compensate for potential issues due to size or mobility. In addition, it would be useful to determine how much XBP1 reporter mRNA is expressed and how this compares to endogenous levels of XBP1 mRNA in unstressed and stressed cells.

Reviewer #2 (Recommendations for the authors):

This manuscript develops a different reporters to monitor XBP1 targeting to the ER, which are used to confirm previous results showing that XBP1 is directed to the ER through a mechanism involving translation of the HR2 mRNA sequence. As indicated in the manuscript, this mechanism had been previously reported by Kohno, and, while the work presented here confirms this model, it does not extend it. The major advance from this manuscript, apart from the reporter development, relates to the fact that IRE1 clusters are not observed in cells expressing endogenous levels of IRE1-GFP and subjected to ER stress. This is in contrast to previous reports where IRE1 clusters were proposed to be the primary site of XBP1 splicing; however, IRE1 clustering from XBP1s splicing has been shown to been separable previously in Ricci et al. (2019) FASEB J (where they showed that the flavinoid luteolin induces robust XBP1 splicing independent of clustering). Herein, the authors demonstrated that the clustering of IRE1-GFP is an artifact of overexpression, which is not observed upon expression of IRE1-GFP to endogenous levels. This is consistent with another recent report submitted to eLife from Peter Walters group showing that endogenous IRE1 does not cluster, despite previous reports (Belyy et al. 2021).

Ultimately, while the experiments appear well performed, the advance of this current manuscript is limited, although it does provide some of the controls requested of the Walter manuscript to compare to previous reports (specifically some of the experiments described in Figure 4). The data included in Figure 1-3 validate previous mechanisms proposed for XBP1 targeting to the ER using new approaches. While important to validate mechanisms using different approaches, there is no new insight included in this aspect of the work. In combination with the Walter manuscript, this work does correct the misinterpretation of the IRE1 activation mechanism resulting from overexpression artifacts, by supporting the fact that endogenous IRE1 does not appear to cluster, but instead splices XBP1 mRNA distributed through ER. Individually, this paper would not be considered strong enough to be published in eLife, but combined with the Walter manuscript they do correct a mechanism of IRE1 activation that is important to highlight in the literature.

Reviewer #3 (Recommendations for the authors):

1. Showing that the MCP-Halo and scAB-GFP do not associate with each more than chance would predict would help to at least show that the dots visualized by both techniques are not likely to be clusters of mRNAs. It is less clear to me how to show that MCP-Halo and scAB-GFP are detecting all of the relevant transcripts, but I would think this point would be important to address one way or another.

2. Would it be possible to engineer an Xbp1 RNA with both the MS2 tag and the SM tag? Presumably, in that case MCP-Halo and scAB-GFP signals should overlap for individual molecules until after splicing. More specifically for point 2, do Xbp1 RNAs that encounter IRE1 then leave the ER, or do they stay associated? Or can Xbp1 RNA leave the ER without encountering an IRE1 cluster?

3. Some statements that are in my mind not warranted:

a. Page 5, "…different from the canonical SRP-mediated recruitment…"; the SRP pathway is not examined. The pathway might use canonical SRP-mediated targeting, just relatively inefficiently.

b. Page 5, "…does not recruit XBP1 mRNA to these higher order oligomeric assemblies."; that there is no stable association is justified, but that there is no recruitment is not, in my opinion.

c. Page 12, "…drives the release of translationally active, translocon-engaged mRNAs."; there is no direct evidence for this claim, as translocons are not examined.

d. Page 12, "ER stress caused a reduction of ER association when compared to untreated conditions." There is no statistical analysis of this for these reporters, so this conclusion is not warranted.

4. Technical points:

a. The authors should explain why there is an NLS on the MCP-Halo protein.

b. The authors should state in the text or legend where the antibody used in Figure 1C recognizes XBP1 protein. If the antibody recognizes spliced wild-type XBP1 but not unspliced, that suggests that the antibody is downstream of the intron, which would also allow it to detect the HR2 mutant when unspliced, but then it shouldn't detect the HR2 mutant when spliced.

c. If the SM tag is place in frame with the spliced Xbp1 mRNA, is that species excluded from the ER membrane, as the authors' interpretations would suggest?

d. Why is 4u8C added to Figure 4F? The logic there is not clear.

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

We have considered carefully your appeal of our decision not to invite a revised version of the manuscript "Live imaging of the co-translational recruitment of XBP1 mRNA to the ER and its processing by diffuse, non-polarized IRE1α" for consideration at eLife. As a consequence of these deliberations, we are prepared to consider (with no guarantees of acceptance) a revised submission addressing that specific concerns that follow:

1. The authors have claimed that they "coordinated our study with the one from the Walter lab", which they provide as a reason for not needing to do certain suggested experiments/controls (since they are outlined in the paper from Belyy et al.), most notably the single molecule time-resolved tracking experiments and measurements of the IRE1 oligomeric state. Our understanding of the facts is that the papers were not submitted together as there are non adjoining cover letters indicating that these papers were linked. Thus, the papers were not considered as co-submitted. Moreover, key controls and experiments were not cross-referenced between the two papers. Only after the review process was complete for the Belyy et al. paper where we made aware of your paper. So, the current manuscript was considered as a stand alone study, and therefore we agree with the concerns of 2/3 reviewers that all of the requested experiments are imperative for publication.

If, however, revised versions of the papers are co-submitted and fully cross-referenced in the future, then in the next round of the editorial process it may be possible to consider the exclusion of some requested experiments from this study.

2. We find that in several instances your paper gives the impression that you were measuring dynamics and quantifying single molecules under these conditions, when in fact this may not have been the case. (The rebuttal letter states that the completion of the text was rushed to try to submit as soon as possible.) The authors note that the text will be edited, which is fine if, again, resubmitted papers are cross-referenced and returned to eLife in tandem. If the papers are returned independently without substantial cross-referencing, then the revised manuscript must stand on its own merits and dynamics and quantifying single molecules must be part of it to reach the level of novelty required by eLife

3. A reviewer rightfully asks for other measurements of ER stress since some splicing is apparent in the absence thapsigargin. Shorter induction times and, indeed, other measurements of ER stress should be shown.

4. We also agree that a side-by-side comparison of the tagged constructs to untagged constructs is valuable. It is more than reasonable that one should always confirm the full function of a tagged protein in relation to the untagged protein, thereby validating the behavior of the former. This would better lay out any caveats to the use of the new system, which is vital if others are to take advantage of it.

5. While the authors are indeed using a method that was previously pioneered by their lab, there are critical controls/information that should be added (e.g. spot intensity distributions) to the supplemental information.

6. The need for an experiment to measure detection efficiencies would indeed show "that the MCP-Halo and scAB-GFP do not associate with each more than chance".

7. In the rebuttal letter, it is noted that several other experiments to address the reviewers' comments are ongoing or can be started. The completion of these experiments will significantly strengthen the manuscript, and we encourage the authors to be as thorough as possible in completing these experiments.

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

Thank you for extensively editing and resubmitting your work entitled "Live imaging of the co-translational recruitment of XBP1 mRNA to the ER and its processing by diffuse, non-polarized IRE1α" for further consideration by eLife. As you will note below, we have asked the original three reviewers of this study to again offer their opinions and insights on your work. While they appreciate the extensive work that has gone into the revision, there are still some issues that need to be addressed prior to final acceptance of this report. Based on the reviewers' thoughtful comments, subsequent discussions, and our evaluation of the work, they are:

1. The need for improved quantitation of select data, which is required to better support some of the claims in the study (Reviewer #1).

2. Based on this analysis, a clear statement of how you are defining "puncta" in this study (Reviewer #1 and JLB). The need for this is accentuated by the fact that the companion paper also uses the word freely, and since scientists in the field will be reading both papers in tandem, it is vital that this definition is in harmony between the two studies.

3. An attempt to better position the paper, in the Abstract, Significance Statement, and Introduction, as a technological advance, rather than only a ground-breaking study on the UPR. As two of the Reviewers continue to firmly maintain (Reviewers #2-3), most of the scientific advances were embedded in the literature and/or are outlined in the companion paper.

4. With regard to this last point, please better coordinate with Belyy et al. to ensure that additional cross-references are included.

The specific comments from the Reviewers are:

Reviewer #1 (Recommendations for the authors):

The authors have satisfactorily addressed many of the issues raised by this reviewer. A few issues still remain unresolved.

1. Regarding the relationship between IRE1a expression level and stress inducible cluster formation, the data in figure 4 are supportive of this claim but insufficiently quantitative. Given the previously claimed importance of the IRE1a clusters in the literature, it would be important for this manuscript's claim and extremely helpful for the field for the authors to (a) provide a quantitative definition of stress inducible clusters and (b) perform quantitative analyses of the two reporter cell lines unstressed and stressed (what fraction of cells have puncta and how many puncta?). It would be even better if the authors could determine whether there is a relationship between ER intensity/IRE1a-GFP expression and propensity to form puncta, assuming only some fraction of cells form puncta, as described in the 2019 Belyy et al. PNAS paper. Note that a puncta definition would be useful for distinguishing between the structures observed in Figure 4E IRE1a-GFP and the relatively bright puncta in the "No stress" Emi1-IRE1a-GFP cell.

2. The wording of the manuscript matters. The authors have not changed their claim: "ER-poised XBP1 transcripts are processed by functional IRE1a assemblies that are homogeneously distributed throughout the ER membrane." No data are provided on what fraction of the IRE1a is in an active state.

Up to this point in the manuscript, this claim is not actually supported and the authors agree. Later, the authors cite the Belyy et al. manuscript, which also does not support the claim. There is no assay for visualizing the active form of IRE1a. The fully phosphorylated IRE1a and inactive dimer do not have physical characteristics that should significantly alter the diffusion coefficient or trajectory correlation. At this time, the Belyy et al. group can detect what appears to be dimers and tetramers/possibly larger oligomers which are presumably in the process of activating. The immunoblots do not establish what fraction of total IRE1a is phosphorylated. Therefore, the matter of where active IRE1a is distributed remains unexamined. It's an interesting question worthy of addressing. That said, the authors simply need to modify their text.

3. The authors have not mentioned sources for 4u8C, dox, tunicamycin or puromycin.

Reviewer #2 (Recommendations for the authors):

In the revised submission, the authors attempted to address many of the comments brought up in the previous review. However, some of the underlying problems still remain. Notably, the assay developed simply validates previous mechanistic insights into IRE1-dependent XBP1 splicing. The authors seem to agree with this in their rebuttal, making the point that this represents the first time that this process has been 'visualized' and that this assay now can be used to further probe XBP1 biology. However, this manuscript is written in such a way to suggest the focus was on improving our understanding of XBP1 splicing, not developing an assay for future work. Further, I would argue previous work clearly did a nice job of working out the mechanism, largely independent of microscopy, so I just don't see the advance here. This manuscript does provide support for a manuscript co-submitted with this revision by Peter Walter's group showing that IRE1 clusters are not required for activity, which is fine, but this is something that was previously described in published reports (see Ricci et al). I still don't feel like this work rises to the level suitable for publication in eLife on its own and it is relying on co-submission with the Walter manuscript to get over that bar.

Reviewer #3 (Recommendations for the authors):

I believe that the authors have addressed most of the technical concerns outlined, and I also find their rebuttal to be persuasive. I now appreciate that the sort of experiment I was envisioning, that underlay several of my comments (but was perhaps not obvious from them individually) would be to test the prediction that an Xbp1 mRNA would lose association with the ER membrane and its splicing could be detected by loss of scAB-GFP. But, as the authors point out, that would require simultaneous 3 channel visualization and also possibly a time scale that would be unfeasible (i.e., that the mRNA not only leaves the ER membrane), but that the ribosomes translating the GCN4 region completed their synthesis. It would also require presumably using milder conditions of ER stress, where it would be reasonable to expect some significant fraction of the Xbp1 mRNAs were unspliced.

I also think the revision benefits from more careful wording, and think the untempered claims in the original manuscript contributed to its perception.

I continue to hold the view that the majority of the paper represents a technical advance of somewhat limited general interest because this group has already established the technique, and here uses it largely to confirm a pathway that is well-accepted. I think where this manuscript complements its companion from the Walter lab is that the Walter lab shows that higher clusters of IRE1 need not form under "normal" conditions, and this paper shows that, when they do form, Xbp1 is likely not spliced there.

[Editors’ note: The authors appealed the original decision. What follows is the authors’ response to the first round of review.]

Comments to the Authors:

While there was significant interest in the topic and preliminarily findings, we are sorry to say that, after consultation with the reviewers, we have decided that this work will not be considered further for publication by eLife.

As you will read below, whilst all three reviewers concur with the importance of the paper's subject matter they independently concluded that the study would need significant additional work to become suitable for publication.

We thank the reviewers for their feedback and genuinely appreciate the time and effort spent on improving our study.

The reviewers have laid out a non-trivial number of additional controls and experiments that are considered essential. In brief, the absence of these experiments indicates that many of the claims currently lack experimental support to the level required by ELife.

We respectfully disagree. Instead, we propose that there is not a single experiment proposed by the reviewers that we could not provide in a reasonable time frame.

More specifically, the clustering analysis is of limited scope and significance,

We argue that developing a mammalian model system in which XBP1 processing clearly takes place despite the absence of visible stress-induced IRE foci is in itself a major advance that, together with the manuscript of Belyy et al. (2021), clarifies a long-standing misunderstanding about the role of IRE1 clusters.

temporal aspects of the observed events are absent, the tools used for this study need to be better validated,

Please propose experiments for further validation of the RNA imaging tools. The IRE1a-GFP construct was already validated/established by the Walter lab in 2020 (Belyy et al., 2020, PNAS). We regret not having explained this more clearly.

and single particle analysis was--in contrast to what was stated--not performed rigorously.

This statement is incorrect. Single-particle analysis was performed with state-of-the-art methodology that has also been published before (e.g. Voigt et al., 2017). We argue that there is a conceptual misunderstanding amongst the reviewers. Please see the detailed response (i.e. to reviewer 3) below.

There is also a consensus amongst the reviewers that a substantial portion of the paper is simply confirmatory based on prior work performed by the Kohno and Argon labs.

This is correct, a substantial portion of the findings are confirmatory. Which is precisely, why it serves to validate the RNA imaging tools that we have established.

At the same time our work not only confirms a model but solves a debate about how/where UPR splicing takes place, favoring a model that supports the non-polarized splicing of XBP1 and discarding the model whereby ER stress signaling would initiate from specific sites within the ER. That is, we believe, the biological importance of our work.

Reviewer #1 (Recommendations for the authors):

In this study, the authors explore key steps in the mammalian unfolded protein response (UPR). They focus on the IRE1a branch, specifically, the recruitment of target mRNA XBP1 to the endoplasmic reticulum (ER) surface, the fate of the mRNA, and organization of IRE1a, itself, in response to unfolded protein stress. The novelty of this study is the use of live cell imaging techniques to follow XBP1 mRNA localization and dynamics. The translational recruitment model of XBP1 to the ER membrane surface has been previously described by others and is confirmed in live cells, here. Different reporters were used to visualize localization and translation of spliced and unspliced forms of XBP1. The authors make additional claims regarding the stress-induced clustering behavior of IRE1a. Overall, this study adds visual data, with some dynamics information to potentially help refine models of steps of XBP1 mRNA trafficking and splicing in response to ER stressors. The novelty of the study approach is undercut by the need for more thorough characterization of the reporters, better descriptions of the reagents, and more accurate descriptions of experiments and results. With the additional requested details and modifications to the text, this manuscript could be a useful addition to the ER stress literature and of broader interest to groups interested in trafficking and translation on the ER membrane.

Thank you for helping us improve our manuscript. We will be happy to provide more accurate descriptions as well as any characterization of the reporter transcripts asked for and agree that the manuscript is a useful addition to the ER stress literature.

The authors make two claims that are well supported by imaging data, mutants, and biochemical fractionation.

Splicing activity by IRE1a or the presence or expression of XBP1 message in an already spliced form decreases the fraction of XBP1 associated with the ER. These observations are supported by membrane fractionation experiments.

The authors have confirmed that stable ER association of unspliced XBP1 mRNA is established through HR2-dependent targeting and relies on active translation. The experimental data support this claim.

Other claims are less well supported or less accurate.

First, the authors claim to "directly monitor recruitment of individual XBP1 transcripts to the ER surface." I was unable to find a single movie or data series demonstrating the movement of a single particle from the cytoplasm to the ER membrane. Experiments show mRNAs on the ER or report amount of cytoplasmic vs. membrane bound. The results, presumably, of recruitment are observed and reported, not the process.

We apologize for this suboptimal phrasing and thank the reviewer for pointing it out. Of course, we are happy to provide individual example movies from amongst our large dataset to show single XBP1 transcripts as they are being recruited to the ER surface.

As the reviewer rightly points out, our analysis quantifies how wt and mutant XBP1 mRNAs are associated with the ER to different degrees. Since some of these transcripts only differ in individual nucleotide substitutions in elements involved in recruitment to the ER, we used their degree of association as a means to characterize the recruitment mechanism. At no point did we mean to insinuate that we had actually quantified recruitment dynamics.

The authors state "Next, we quantified the mobility of individual particles with respect to their ER localization and therefore assessed when an mRNA particle associates with the ER." What appears to be assessed is whether a particle is associated with the ER, not "when." I assume the authors did not mean "when" as part of temporal order, but it would be less confusing to claim to assess whether or if.

We thank the reviewer for pointing this out and did indeed not use “when” in a temporal context. We will use a different conjunction in the revised version of the manuscript.

The authors claim to show that IRE1a-catalyzed splicing mobilizes XBP1 mRNA from the ER membrane in response to ER stress. Here the data are mostly supportive of the claim. However, what the authors actually show is that splicing is necessary for an increase in XBP1 mRNA localization in the cytoplasm. No data are presented showing interactions with IRE1a and/or movement of XBP1 mRNA from the ER membrane to the cytoplasm. Based on multiple studies, the interpretation is likely correct, but the microscopy data do not show the process.

Again, we thank the reviewer for reading the manuscript attentively. What we actually show is an increase in ER association upon inhibition of IRE1a cleavage activity. Please point out how this result could be interpreted in any other way but the one proposed in the manuscript.

– We are happy to provide movies that show individual mRNAs leaving the ER, however, did not provide them in the first place because we did not consider them very informative (compared to the combined analysis of thousands of individual mRNAs and the heterogeneity of their mobility that we derived from that).

– With respect to showing interactions of individual transcripts with IRE1a clusters, we would like to point out that we did provide a specific example for this extremely rare event (Supplementary Movie 9, white arrow indicates a single XBP1 mRNA particle that colocalizes with an IRE1a cluster) to demonstrate that we were indeed able to detect those and that the absence of colocalization was no artifact of the experimental set-up.

The matter of IRE1a clustering is more problematic.

The authors state, "Surprisingly, we find that XBP1 transcripts are not recruited into large IRE1a clusters, which only assemble upon overexpression of fluorescently-tagged IRE1a during ER stress. Our findings support a model where ribosome-engaged, ER-poised XBP1 mRNA is processed by functional IRE1a assemblies that are homogenously distributed throughout the ER membrane."

The authors state, "ER-poised XBP1 transcripts are processed by functional IRE1a assemblies that are homogenously distributed throughout the ER membrane." No data are provided on what fraction of the IRE1a is in an active state. There is no method used to detect the active state or information on the spatial distribution of stress-activated IRE1a. This claim is not supported.

The reviewer is correct in that we do not provide any data to quantify what fraction of stress-activated IRE1a is functional or how it might be spatially distributed with respect to non-functional IRE1a assemblies. This is due to the nature of our experimental set-up and is exactly what Belyy et al. (2021) demonstrate in their complimentary manuscript. We urge the reviewer to consider both manuscripts together since they were submitted back-to-back. In respect of our agreement with Belyy et al., we refrained from performing the same single-molecule IRE1a imaging experiment that they have set-up.

However, we do demonstrate that IRE1a expressed at physiological levels is homogeneously distributed throughout the ER membrane, does not accumulate in large microscopically visible clusters (contrary to many previous reports) and that even when formed, these clusters are not sites of XBP1 processing. With this experiment, we address a major open question in the field where large IRE1a clusters have been proposed to function as the sites of XBP1 splicing for a long time.

"Our data have allowed us to visualize and uncover unanticipated features of one of the key steps of UPR initiation, the encounter of XBP1 mRNA with IRE1a to undergo splicing." This is another claim that is not quite accurate. The authors observed tagged XBP1 mRNA dynamics in co-expressing cells treated with an inhibitor of IRE1a activity. An encounter in which splicing could actually occur is never presented. Whether XBP1 mRNA is stimulated to dissociate from the ER membrane following an encounter with IRE1a-GFP is not presented. The ability of drug inhibited IRE1a-GFP to bind XBP1 mRNA is not tested or reported.

We thank the reviewer for raising this argument. As before, we should re-phrase and differentiate clearly between observing single-mRNA dynamics and the quantification of ER association of many individual mRNA transcripts, which is what we are showing in this manuscript.

The reviewer is correct, we do not show an individual mRNA molecule that is cleaved by IRE1a and leaves the ER surface in response to that. However, we do show that IRE1a is homogenously distributed throughout the ER membrane (Figure 4E) and that XBP1 transcripts are present/absent there in response to IRE1a cleavage activity (Figure 3C). This observation is not only dependent on treatment with an IRE1a inhibitor but was also observed for non-cleavable mutant transcripts that accumulate on the ER surface. We argue that the increase in ER association that we observe for these non-spliceable mutant mRNAs clearly shows that ER dissociation depends on mRNA cleavage even though we do not show (and do not claim to show) the actual dynamics of the process.

The authors have made a significant point of weighing in on the matter of IRE1a clustering in both the abstract and text. While their findings are potentially interesting, the investigation of this matter has not been sufficiently rigorous. First, the authors simply describe the attached fluorescent protein as a GFP. Is this monomeric GFP? Emerald? mNeonGreen? or? With the wide variety of fluorescent proteins available, the minimal expectation is to state what was inserted and imaged. Given the potential of many fluorescent proteins to oligomerize, the choice of fluorescent protein matters.

We apologize for the lack of specificity. As mentioned in the methods, we have reproduced the construct design established by the Walter lab in 2019. Our fusion protein design includes a GFPuv tag.

Second, the systems used may be inducing some constitutive stress. For example, in 2C, there appears to be splicing of the reporter, even in the absence of thapsigargin or expression of IRE1-fluorescent protein. Induction with Dox is for 15 hours. That's long enough for stimulation of a stress response and perhaps even some degree of adaptation. I'd like to see some other assays of ER stress before and during induction up to the time normally used to apply stress.

We thank the reviewer for this comment. Doxycycline-inducible expression of imaging reporter constructs does not induce ER stress, as indicated by the splicing rate of endogenous XBP1 mRNA or the low expression of other UPR transcripts or trigger any kind of detectable adaptation. In the experiments where splicing/translation of reporter mRNAs were characterized, we typically induced their expression for 15 hours to reach steady-state levels of mRNA expression. Please note that IRE1a-GFP expression was not controlled by Dox-inducible promoters, but by constitutive promoters.

We would be happy to include this information in the manuscript, or to repeat the characterization experiments using shorter induction times.

Similarly, in the Ire1a-KO cells in 4B, there appears to be spliced XBP1, with and without stress. Importantly, the ratio of spliced XBP1 for the emi1 variant is much lower than for wt, even though immunoblotting suggests there is significantly more EMI-1-IRE1a-GFP protein present than for wt IRE1a. Please comment. Is the GFP tagged variant less active? Is it possible that the GFP tagged variant is less activatable?

We thank the reviewer for this accurate comment. The reviewer is right, in the gel provided there is a faint band that co-migrates with the band that corresponds to the spliced mRNA. Since quantitative RT-PCR of XBP1 mRNA failed to detect any spliced XBP1 mRNA, and Western blot analysis shows no XBP1s protein in IRE1a-KO cells, this band could represent a non-specific PCR product that migrates like the “spliced” PCR band. We are conducting PCRs with a different set of primers to rule out that possibility.

Also, Emi1-IRE1a-GFP promotes splicing with lower efficiency than endogenous IRE1a, as the PCR and qPCR analyses indicate, we thank the reviewer for giving us the opportunity to mention this aspect properly.

More fundamentally, the authors claim that large IRE1a clusters… only assemble upon overexpression of fluorescently-tagged IRE1a during ER stress. At the very least, the more accurate thing would be to say fluorescently-tagged IRE1a only forms inducible observable clusters under overexpression conditions. In this study, there has been no testing of oligomeric/cluster status of native IRE1a.

We thank the reviewer for pointing this out and will re-phrase to “no observation of microscopically visible IRE1a clusters”.

There has been no use of other assays to determine if low expressed IRE1a-fluorescent protein forms oligomers of any size during stress. Nor has there been investigation of whether a version with a small epitope tag might form clusters. What about the ability of other stress conditions, such as acute thapsigargin or DTT, for the ability to induce observable clustering for the lower expressed IRE1a-GFP? Ultimately, the clustering analysis has been too limited to draw any major conclusions. The ability to cluster and its relevance are important to popular models of IRE1a activity. A nuanced study of the clustering regulation of IRE1a would be immensely useful to the field. The observation that clustering observed in mammalian cells is stress inducible, suggests that observable clustering is probably reporting on some aspect of IRE1 regulation. Clusters may represent an extreme manifestation of normal physiology. That is, even an artifact of expression does not rule out the potential importance of forming small clusters in the normal stress response. Note that the overexpressed IRE1a-GFP is constitutively maximally active. There is no stress induced increase in splicing activity, yet there is a stress-inducible change in distribution. This redistribution could reflect some sort of stress inducible regulatory step, for example sequestering of excess unengaged IRE1a, inactivated IRE1a or something else. Whatever causes visible clustering seems to be more than a simple artifact.

Thank you for raising all of these points. As above, we would like to refer to the manuscript of Belyy et al. (2021) that is also under review at eLife and does exactly this. The focus of our work (which purposefully does not repeat the excellent study performed in the Walter lab) is on the recruitment of XBP1 transcripts to IRE1a assemblies. We fully agree that visible clusters might have other physiologically relevant functions but simply state that they are not sites of XBP1 processing.

This is no obvious finding but has been controversially discussed in the literature for more than a decade (Li et al., 2010; Korennykh et al., 2009) and is still discussed today (Belyy et al., 2019; Tran et al., 2021) especially since HAC1 mRNAs have been shown so early on to be recruited to discrete Ire1p foci in yeast (Aragón et al., 2009; Kimata et al., 2007).

It would be helpful for the authors to compare the functionality of their constructs/system relative to untagged versions. Specifically, rates of processing. While I think the authors have demonstrated reasonably well that their XBP1 MS2 reporter does get targeted to the ER and spliced in response to ER stress, it would be useful to know if the 24 MS2 sites and attached MCS-GFP reporters (5nm+ diameter for each) affect rates of XBP1 mRNA processing. Is it possible that the construct is so large that it is relatively immobilized on the ER membrane by sterics and/or sheer size?

Thank you for proposing this experiment. While we agree with the reviewer that it is always useful to assay mRNA processing rates, our data clearly show that the ER association of our MS2 tagged reporter transcripts is abolished upon introduction of a single point mutation in the HR2 sequence (Figure 1F,G) or removal of the ER intron (Figure 3C). In addition, our RT-PCR assays clearly show that reporter transcripts are efficiently spliced only in response to ER stress (Figure 1B,C).

Therefore, we do not understand how differences in processing rates (compared to endogenous transcripts) would affect the findings presented here, where the same kind of bias (potentially introduced by the MS2 labeling) affects all reporter transcripts in the same way. The observed differences in ER association in between different reporters can therefore only be due to differences in experimental set-up (mutants, inhibitors) and are not caused by the labeling method.

Could the size of the reporter slow the approach of IRE1a to the cleavage site? Does high expression of the XBP1 substrate impact the efficiency of processing? That is, if client is saturating, then encountering and splicing by activated IRE1a on the ER membrane could compensate for potential issues due to size or mobility.

All of these are fair questions and could be assessed at a later time. However, they are not relevant to our main conclusions, which are qualitative (microscopically visible IRE1a clusters are not sites of XBP1 processing) and highly relevant to the research community.

In addition, it would be useful to determine how much XBP1 reporter mRNA is expressed and how this compares to endogenous levels of XBP1 mRNA in unstressed and stressed cells.

We can provide this information based on the qPCR assays that we have performed.

Reviewer #2 (Recommendations for the authors):

This manuscript develops a different reporters to monitor XBP1 targeting to the ER, which are used to confirm previous results showing that XBP1 is directed to the ER through a mechanism involving translation of the HR2 mRNA sequence. As indicated in the manuscript, this mechanism had been previously reported by Kohno, and, while the work presented here confirms this model, it does not extend it.

Please see comment above. The experiments provide important validation and were used to establish the single molecule imaging tools, which will be extremely useful for further characterization of mRNA processing during the UPR on the ER.

The major advance from this manuscript, apart from the reporter development, relates to the fact that IRE1 clusters are not observed in cells expressing endogenous levels of IRE1-GFP and subjected to ER stress. This is in contrast to previous reports where IRE1 clusters were proposed to be the primary site of XBP1 splicing; however, IRE1 clustering from XBP1s splicing has been shown to been separable previously in Ricci et al. (2019) FASEB J (where they showed that the flavinoid luteolin induces robust XBP1 splicing independent of clustering). Herein, the authors demonstrated that the clustering of IRE1-GFP is an artifact of overexpression, which is not observed upon expression of IRE1-GFP to endogenous levels. This is consistent with another recent report submitted to ELIFE from Peter Walters group showing that endogenous IRE1 does not cluster, despite previous reports (Belyy et al. 2021).

Please see comment above. Apart from the study of Ricci and colleagues that the reviewer mentions, earlier works from the Walter lab (Belyy et al. PNAS 2020) as well as their manuscript currently under review at eLife (Belyy et al., 2021), are consistent with our observation that large IRE1a foci only form whenever IRE1a is overexpressed.

More importantly, however, we show for the first time that the ability of such foci to recruit IRE1a is indeed very limited. The direct observation of this lack of recruitment of XBP1 mRNA solves a long-standing debate in the field. Of course, our findings do not rule out the possibility that under specific conditions or in specific cell types (for instance, in multiple myeloma cells, where IRE1a is strongly overexpressed), these foci may be able to recruit other RNA substrates, or to promote other processes in the stressed cell.

Ultimately, while the experiments appear well performed, the advance of this current manuscript is limited, although it does provide some of the controls requested of the Walter manuscript to compare to previous reports (specifically some of the experiments described in Figure 4). The data included in Figure 1-3 validate previous mechanisms proposed for XBP1 targeting to the ER using new approaches. While important to validate mechanisms using different approaches, there is no new insight included in this aspect of the work.

We agree with the reviewer but would like to stress that even if XBP1 targeting mechanisms have been proposed and discussed in the literature for a while, our study is the first to directly test them.

In combination with the Walter manuscript, this work does correct the misinterpretation of the IRE1 activation mechanism resulting from overexpression artifacts, by supporting the fact that endogenous IRE1 does not appear to cluster, but instead splices XBP1 mRNA distributed through ER. Individually, this paper would not be considered strong enough to be published in eLife, but combined with the Walter manuscript they do correct a mechanism of IRE1 activation that is important to highlight in the literature.

We thank the reviewer for this important comment.

Reviewer #3 (Recommendations for the authors):

1. Showing that the MCP-Halo and scAB-GFP do not associate with each more than chance would predict would help to at least show that the dots visualized by both techniques are not likely to be clusters of mRNAs. It is less clear to me how to show that MCP-Halo and scAB-GFP are detecting all of the relevant transcripts, but I would think this point would be important to address one way or another.

To answer your question concerning detection efficiencies: This could be done in e.g. a FISH experiment where two different probe sets are used to detect the same transcript species. The degree of colocalization between both labels can then be used to calculate detection efficiencies for each fluorescent label. For further information please see Voigt*, Gerbracht* et al., 2019, Nat Prot. We are happy to provide such an experiment in a revised manuscript if needed.

2. Would it be possible to engineer an Xbp1 RNA with both the MS2 tag and the SM tag? Presumably, in that case MCP-Halo and scAB-GFP signals should overlap for individual molecules until after splicing. More specifically for point 2, do Xbp1 RNAs that encounter IRE1 then leave the ER, or do they stay associated? Or can Xbp1 RNA leave the ER without encountering an IRE1 cluster?

This is precisely what we have done. Please see the cartoon in Figure 2A. XBP1u translation reporter transcripts contain both, the SM and MS2 tags. This is why MCP-Halo and scAB-GFP spots need to colocalize. XBP1 mRNAs that are processed by IRE1a eventually leave the ER as illustrated by differences in ER association upon induction of ER stress and treatment with 4µ8C in Figure 3C. The dynamics of the cleavage reaction cannot be inferred from the kind of live imaging experiment that we have performed here.

Last, yes, it is possible that XBP1 mRNAs could leave the ER without encountering IRE1a, yet, this would only mean that we are underestimating the efficiency of XBP1 recruitment to the ER. Since our conclusions from the ER association analysis are qualitative, not quantitative, we believe that this possibility is of minor relevance.

3. Some statements that are in my mind not warranted:

a. Page 5, "…different from the canonical SRP-mediated recruitment…"; the SRP pathway is not examined. The pathway might use canonical SRP-mediated targeting, just relatively inefficiently.

Please excuse the vague phrasing. This is exactly what we mean. As well as the lack of a canonical signal sequence in the XBP1 ORF.

b. Page 5, "…does not recruit XBP1 mRNA to these higher order oligomeric assemblies."; that there is no stable association is justified, but that there is no recruitment is not, in my opinion.

We will re-phrase this sentence.

c. Page 12, "…drives the release of translationally active, translocon-engaged mRNAs."; there is no direct evidence for this claim, as translocons are not examined.

This is correct. We are working on this at the moment.

d. Page 12, "ER stress caused a reduction of ER association when compared to untreated conditions." There is no statistical analysis of this for these reporters, so this conclusion is not warranted.

This is correct. We will add the analysis.

4. Technical points:

a. The authors should explain why there is an NLS on the MCP-Halo protein.

To recruit excess/unbound MCP-Halo protein away from the cytoplasm and thereby increase signal/noise. We will add this detail to the revised manuscript.

b. The authors should state in the text or legend where the antibody used in Figure 1C recognizes XBP1 protein. If the antibody recognizes spliced wild-type XBP1 but not unspliced, that suggests that the antibody is downstream of the intron, which would also allow it to detect the HR2 mutant when unspliced, but then it shouldn't detect the HR2 mutant when spliced.

As indicated by the provider (Santa Cruz Biotechnology), the antibody recognizes a region within the N-terminal half of XBP1 protein, encoded upstream the UPR intron. Thus, this antibody should recognize both XBP1u and XBP1s proteins. Yet, we can only detect XBP1s protein. We believe this lack of recognition is due to the translational regulation of XBP1u mRNA and the low stability of XBP1u protein.

c. If the SM tag is place in frame with the spliced Xbp1 mRNA, is that species excluded from the ER membrane, as the authors' interpretations would suggest?

This experiment is in progress.

d. Why is 4u8C added to Figure 4F? The logic there is not clear.

In order to inhibit IRE1a cleavage activity and allow accumulation/detection of XBP1 mRNA in IRE1a clusters. We have performed the experiment with the same result in absence and presence of the compound. Here, we show the lack of accumulation in 4µ8C treated cells to illustrate that even a “kiss-and-run” mechanism (very fast turnover) could be detected in this experimental set-up.

[Editors’ note: what follows is the authors’ response to the second round of review.]

1. The authors have claimed that they "coordinated our study with the one from the Walter lab", which they provide as an reason for not needing to do certain suggested experiments/controls (since they are outlined in the paper from Belyy et al.), most notably the single molecule time-resolved tracking experiments and measurements of the IRE1 oligomeric state. Our understanding of the facts is that the papers were not submitted together as there are non adjoining cover letters indicating that these papers were linked. Thus, the papers were not considered as co-submitted. Moreover, key controls and experiments were not cross-referenced between the two papers. Only after the review process was complete for the Belyy et al. paper where we made aware of your paper. So, the current manuscript was considered as a stand alone study, and therefore we agree with the concerns of 2/3 reviewers that all of the requested experiments are imperative for publication.

Thank you for explaining the details of the editorial process, through which our manuscript proceeded. While we did indeed submit our manuscript a few weeks after the Belyy et al. paper, both cover letters actually mentioned the other’s manuscript and suggested a parallel revision process. We regret that this could not happen in a more coordinated manner and aimed to address both, missing controls as well as cross-references, in the revised version of the manuscript.

If, however, revised versions of the papers are co-submitted and fully cross-referenced in the future, then in the next round of the editorial process it may be possible to consider the exclusion of some requested experiments from this study.

We co-submit our revised and cross-referenced manuscript together with the revised manuscript of Belyy et al..

2. We find that in several instances your paper gives the impression that you were measuring dynamics and quantifying single molecules under these conditions, when in fact this may not have been the case. (The rebuttal letter states that the completion of the text was rushed to try to submit as soon as possible.) The authors note that the text will be edited, which is fine if, again, resubmitted papers are cross-referenced and returned to eLife in tandem. If the papers are returned independently without substantial cross-referencing, then the revised manuscript must stand on its own merits and dynamics and quantifying single molecules must be part of it to reach the level of novelty required by eLife.

We apologize for any confusion caused by the single-molecule terminology used.

Our study indeed quantifies single particles and their degree of association with the ER using both live and fixed single-molecule imaging approaches. While it is correct that we quantify instantaneous diffusion coefficients as a measure of the mobility (or “dynamics”) of individual mRNA particles, this does not mean that we are also able to characterize the “recruitment dynamics” (as in the complete trajectory travelled) of individual mRNAs on their way to the ER. This is due to experimental limitations such as the high particle number and mobility as well as low signal/noise and rapid photo bleaching of diffraction-limited spots (as which we detect single particles) that are inherent to any single-molecule imaging experiment. Instead, our analysis relies on the quantification of the mobility and subcellular localization of single particles that were imaged over short periods of time at high temporal resolution.

In summary, our work relies on the quantification of single molecules and their diffusive properties (“dynamics”) over short time frames (ms scale). However, this is not the same as the analysis of “recruitment dynamics” that would involve tracking individual mRNAs over the extend of time (min scale) that a particle requires to travel after translation initiation in the cytoplasm to the ER surface, which is not possible in currently available live single particle imaging set-ups.

We thank the reviewers for pointing out how confusingly we have applied the term “dynamics” and have made sure to re-phrase accordingly in the revised version of our manuscript.

3. A reviewer rightfully asks for other measurements of ER stress since some splicing is apparent in the absence thapsigargin. Shorter induction times and, indeed, other measurements of ER stress should be shown.

We thank the reviewers for requesting a more thorough analysis of the behavior of our reporter mRNAs and their effect on ER stress signaling. We now include a more detailed characterization of UPR signaling in the cell lines used in this study both under non-stress and stress conditions (Figure 1—figure supplement 1 and Figure 4—figure supplement 1).

4. We also agree that a side-by-side comparison of the tagged constructs to untagged constructs is valuable. It is more than reasonable that one should always confirm the full function of a tagged protein in relation to the untagged protein, thereby validating the behavior of the former. This would better lay out any caveats to the use of the new system, which is vital if others are to take advantage of it.

We agree with the reviewer’s sensible comment and include these control experiments (Figure 4—figure supplement 1) in the updated version of our manuscript.

5. While the authors are indeed using a method that was previously pioneered by their lab, there are critical controls/information that should be added (e.g. spot intensity distributions) to the supplemental information.

Thank you for pointing this out. We now provide intensity distributions for single-particle signal from live (Figure 3figure supplement 1E) as well as fixed cell (Figure 2—figure supplement 1D) imaging experiments.

6. The need for an experiment to measure detection efficiencies would indeed show "that the MCP-Halo and scAB-GFP do not associate with each more than chance".

To address this concern and test if MCP-Halo and scAB-GFP could associate unspecifically, we have performed a targeted imaging experiment (Figure 2—figure supplement 2A), for which we provide data (Figure 2—figure supplement 2B) and quantification (Figure 2—figure supplement 2C) in the revised manuscript.

In brief, we imaged mRNA reporters that encode MCP-Halo and scAB-GFP binding sites either combined on a single or separately on two individual transcripts. Image data analysis shows that the fraction of co-localizing MCP-Halo and scAB-GFP spots is reduced to background levels (Mean ± SD = 0.01 ±0.03) when binding sites are co-expressed from separate reporter transcripts in the same cell (as opposed to 0.50 ±0.17 when expressed from a single transcript), and thus demonstrates that scAB-GFP and MCP-Halo do not associate or co-localize by more than chance (Figure 2—figure supplement 2C).

7. In the rebuttal letter, it is noted that several other experiments to address the reviewers' comments are ongoing or can be started. The completion of these experiments will significantly strengthen the manuscript, and we encourage the authors to be as thorough as possible in completing these experiments.

Several experiments to characterize the translation of XBP1 reporters on the ER are on their way. We have performed ribosome run-off experiments (after Harringtonine treatment) to characterize translation elongation rates on ER associated transcripts and found that ribosome occupancy was too low to fit our data and quantify specific elongation rates. Thus, we are now in the process of engineering a new, brighter XBP1u translation site reporter. In addition, we have also started to generate XBP1s reporter cell lines. Unfortunately, their generation was beyond the time frame that we agreed on for the coordinated re-submission of the two manuscripts.

[Editors’ note: what follows is the authors’ response to the second round of review.]

1. The need for improved quantitation of select data, which is required to better support some of the claims in the study (Reviewer #1).

We have added the quantification in Figure 4F.

2. Based on this analysis, a clear statement of how you are defining "puncta" in this study (Reviewer #1 and JLB). The need for this is accentuated by the fact that the companion paper also uses the word freely, and since scientists in the field will be reading both papers in tandem, it is vital that this definition is in harmony between the two studies.

Thank you for pointing this out. We now define IRE1a-GFP foci as intensity aggregates that can be detected as an enrichment of GFP intensity, which is ≥ 5-fold over background. Similarly, we define cells as containing IRE1a-GFP foci if at least 1% of all cellular GFP signal is contained in foci. This quantification is in accordance with the quantification employed in the manuscript of Belyy et al.

3. An attempt to better position the paper, in the Abstract, Significance Statement, and Introduction, as a technological advance, rather than only a ground-breaking study on the UPR. As two of the Reviewers continue to firmly maintain (Reviewers #2-3), most of the scientific advances were embedded in the literature and/or are outlined in the companion paper.

We have modified Abstract, Significance Statement and Introduction accordingly.

4. With regard to this last point, please better coordinate with Belyy et al. to ensure that additional cross-references are included.

We have included additional cross references and discussed this with Belyy et al.

The specific comments from the Reviewers are:

Reviewer #1 (Recommendations for the authors):

The authors have satisfactorily addressed many of the issues raised by this reviewer. A few issues still remain unresolved.

1. Regarding the relationship between IRE1a expression level and stress inducible cluster formation, the data in figure 4 are supportive of this claim but insufficiently quantitative. Given the previously claimed importance of the IRE1a clusters in the literature, it would be important for this manuscript's claim and extremely helpful for the field for the authors to (a) provide a quantitative definition of stress inducible clusters and (b) perform quantitative analyses of the two reporter cell lines unstressed and stressed (what fraction of cells have puncta and how many puncta?). It would be even better if the authors could determine whether there is a relationship between ER intensity/IRE1a-GFP expression and propensity to form puncta, assuming only some fraction of cells form puncta, as described in the 2019 Belyy et al. PNAS paper. Note that a puncta definition would be useful for distinguishing between the structures observed in Figure 4E IRE1a-GFP and the relatively bright puncta in the "No stress" Emi1-IRE1a-GFP cell.

Thank you for this input. We have added a quantification of our data in Figure 4F. In agreement with Belyy et al., we define cells as IRE1-foci containing if ≥ 1% of the total cellular IRE1a-GFP fluorescence can be attributed to puncta that are detected via thresholding (of normalized images) using a 5-fold fluorescence intensity enrichment over background as cut-off.

2. The wording of the manuscript matters. The authors have not changed their claim: "ER-poised XBP1 transcripts are processed by functional IRE1a assemblies that are homogeneously distributed throughout the ER membrane." No data are provided on what fraction of the IRE1a is in an active state.

Up to this point in the manuscript, this claim is not actually supported and the authors agree. Later, the authors cite the Belyy et al. manuscript, which also does not support the claim. There is no assay for visualizing the active form of IRE1a. The fully phosphorylated IRE1a and inactive dimer do not have physical characteristics that should significantly alter the diffusion coefficient or trajectory correlation. At this time, the Belyy et al. group can detect what appears to be dimers and tetramers/possibly larger oligomers which are presumably in the process of activating. The immunoblots do not establish what fraction of total IRE1a is phosphorylated. Therefore, the matter of where active IRE1a is distributed remains unexamined. It's an interesting question worthy of addressing. That said, the authors simply need to modify their text.

Thank you for pointing this out. We had originally misunderstood the reviewer’s point and have now modified the text to reflect that we have no information on what fraction of IRE1a assemblies is functional or how it is distributed throughout the membrane.

3. The authors have not mentioned sources for 4u8C, dox, tunicamycin or puromycin.

We have added their sources in the methods section.

Reviewer #2 (Recommendations for the authors):

In the revised submission, the authors attempted to address many of the comments brought up in the previous review. However, some of the underlying problems still remain. Notably, the assay developed simply validates previous mechanistic insights into IRE1-dependent XBP1 splicing. The authors seem to agree with this in their rebuttal, making the point that this represents the first time that this process has been 'visualized' and that this assay now can be used to further probe XBP1 biology. However, this manuscript is written in such a way to suggest the focus was on improving our understanding of XBP1 splicing, not developing an assay for future work. Further, I would argue previous work clearly did a nice job of working out the mechanism, largely independent of microscopy, so I just don't see the advance here. This manuscript does provide support for a manuscript co-submitted with this revision by Peter Walter's group showing that IRE1 clusters are not required for activity, which is fine, but this is something that was previously described in published reports (see Ricci et al). I still don't feel like this work rises to the level suitable for publication in eLife on its own and it is relying on co-submission with the Walter manuscript to get over that bar.

Seeing is believing, which is why we strongly advocate the development of single-molecule tools that can test biological hypotheses. In addition, we have modified Abstract, Significance Statement and Introduction to also highlight the methodological advance of our work.

Author details

1. Silvia Gómez-Puerta

   Department of Gene Therapy and Regulation of Gene Expression, Center for Applied Medical Research (CIMA), University of Navarra, Pamplona, Spain

   Contribution

   Investigation

   Competing interests

   No competing interests declared

2. Roberto Ferrero

   Department of Gene Therapy and Regulation of Gene Expression, Center for Applied Medical Research (CIMA), University of Navarra, Pamplona, Spain

   Contribution

   Investigation

   Competing interests

   No competing interests declared

3. Tobias Hochstoeger

   1. Friedrich Miescher Institute for Biomedical Research, Basel, Switzerland
   2. University of Basel, Basel, Switzerland

   Contribution

   Investigation

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-8061-7857

4. Ivan Zubiri

   Department of Gene Therapy and Regulation of Gene Expression, Center for Applied Medical Research (CIMA), University of Navarra, Pamplona, Spain

   Contribution

   Investigation

   Competing interests

   No competing interests declared

5. Jeffrey Chao

   Friedrich Miescher Institute for Biomedical Research, Basel, Switzerland

   Contribution

   Funding acquisition

   Competing interests

   No competing interests declared

6. Tomás Aragón

   Department of Gene Therapy and Regulation of Gene Expression, Center for Applied Medical Research (CIMA), University of Navarra, Pamplona, Spain

   Contribution

   Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Supervision, Validation, Visualization, Writing – original draft, Writing – review and editing

   For correspondence

   taragon@unav.es

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-1700-2729

7. Franka Voigt

   Friedrich Miescher Institute for Biomedical Research, Basel, Switzerland

   Contribution

   Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing – original draft, Writing – review and editing

   For correspondence

   franka.voigt@fmi.ch

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-9515-0367

• 1,483

  Page views

• 271

  Downloads

• 1

  Citations

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