1. Cell Biology
Download icon

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

  1. Anush Bakunts
  2. Andrea Orsi
  3. Milena Vitale
  4. Angela Cattaneo
  5. Federica Lari
  6. Laura Tadè
  7. Roberto Sitia
  8. Andrea Raimondi
  9. Angela Bachi
  10. Eelco van Anken Is a corresponding author
  1. San Raffaele Scientific Institute, Italy
  2. Università Vita-Salute San Raffaele, Italy
  3. FIRC Institute of Molecular Oncology, Italy
Research Article
  • Cited 0
  • Views 950
  • Annotations
Cite as: eLife 2017;6:e27518 doi: 10.7554/eLife.27518

Abstract

Insufficient folding capacity of the endoplasmic reticulum (ER) activates the unfolded protein response (UPR) to restore homeostasis. Yet, how the UPR achieves ER homeostatic readjustment is poorly investigated, as in most studies the ER stress that is elicited cannot be overcome. Here we show that a proteostatic insult, provoked by persistent expression of the secretory heavy chain of immunoglobulin M (µs), is well-tolerated in HeLa cells. Upon µs expression, its levels temporarily eclipse those of the ER chaperone BiP, leading to acute, full-geared UPR activation. Once BiP is in excess again, the UPR transitions to chronic, submaximal activation, indicating that the UPR senses ER stress in a ratiometric fashion. In this process, the ER expands about three-fold and becomes dominated by BiP. As the UPR is essential for successful ER homeostatic readjustment in the HeLa-µs model, it provides an ideal system for dissecting the intricacies of how the UPR evaluates and alleviates ER stress.

https://doi.org/10.7554/eLife.27518.001

Introduction

Key to survival is that organisms adapt their behavior—or on the cellular level, their molecular machineries—according to need. An example is the ability of the endoplasmic reticulum (ER) to accomodate fluctuations in the load of client proteins that fold and assemble in this organelle before being dispatched to travel further along the secretory pathway (Walter and Ron, 2011). A vast array of ER resident chaperones and folding factors assist the maturation of client proteins (van Anken and Braakman, 2005). BiP is an ER resident HSP70 family chaperone, whose function is regulated by at least seven different HSP40 co-chaperones, named ERdj1–7, and two nucleotide exchange factors (Melnyk et al., 2015), as well as through reversible AMPylation (Preissler et al., 2015). The ER also hosts an HSP90 family member, GRP94, for which CNPY family proteins may be co-chaperones (Anelli and van Anken, 2013), and several peptidylprolyl isomerases (PPIases), such as cyclophilin B and a few FKBP family members. Special to the ER are the protein disulfide isomerases (PDI) that assist disulfide bond formation, and the lectin chaperones, such as calnexin (CNX) and calreticulin (CRT), that associate with client proteins through binding their N-glycans (van Anken and Braakman, 2005).

The chaperones and folding factors retain folding and assembly intermediates in the ER in order to minimize the secretion or deployment throughout the endomembrane system of immature client proteins (Ellgaard and Helenius, 2003). A consequence of client retention is that when the protein folding machinery is not adequate to facilitate maturation of clients, they accumulate in the ER, which causes stress and activates the unfolded protein response (UPR) (Walter and Ron, 2011). The circuitry of the UPR has been mapped to impressive detail: the three well-studied branches of the UPR are initiated by the UPR transducers IRE1α, PERK and ATF6α, each of which has an ER stress-sensing domain in the ER lumen and a UPR signaling effector domain in the cytosol. Activated IRE1α and PERK are kinases that trans-autophosphorylate. This step triggers IRE1α to assume endonuclease activity, such that it removes an intron from XBP1 mRNA. Upon its religation, the spliced mRNA encodes the XBP1 transcription factor (Yoshida et al., 2001; Calfon et al., 2002). Activated PERK transiently attenuates protein synthesis through phosphorylation of the translation initiation factor eIF2α (Harding et al., 1999). At the same time, eIF2α phosphorylation favors the expression of a few transcripts, in particular ATF4, a transcription factor that activates further downstream effectors, such as CHOP (Walter and Ron, 2011). The third UPR branch is activated by ATF6α, which undergoes regulated intramembrane proteolysis in the Golgi and thus a transcriptionally active N-terminal portion of 50 kDa is liberated that acts as a transcription factor (Ye et al., 2000). The UPR transcription factors jointly initiate genetic programs that drive the expression of all of the components that are necessary to expand the ER, including the chaperones and enzymes for membrane synthesis (Walter and Ron, 2011). In fact, overexpression of for instance XBP1 alone leads to ER expansion even in the absence of any perturbation of the ER client protein folding and assembly process (Sriburi et al., 2004). Altogether, the UPR homeostatically readjusts the ER folding machinery by expanding the organelle according to need, and regulates cell fate decisions depending on the severity of ER stress (Walter and Ron, 2011).

To date, most studies on the UPR circuitry have focused on the signaling pathways themselves, and little is known about how the UPR evaluates the severity of ER stress and the success of the homeostatic readjustment of the ER. Here, we show that the widely used strategy of employing ER stress-eliciting drugs obscures how ER homeostatic readjustment may be achieved, and instead, we present a HeLa cell model that allows us to evaluate just that. By inducible overexpression of orphan immunoglobulin M (IgM) secretory heavy chain (µs), we provoke a full-blown UPR, which is essential for the cells to cope with the proteostatic insult. As µs accumulates in the ER, it transiently eclipses BiP levels, at which point the UPR output is strongest. UPR-driven upregulation then allows BiP to reach levels that exceed µs levels again, while the ER expands in the process. The activation of the UPR is maximal only when there is a relative shortage of BiP, whereas it subsides to chronic, submaximal output levels when ER homeostatic readjustment is achieved. Co-expression of Ig light chain (λ) instead leads to productive IgM secretion, such that BiP is not sequestered by µs, the UPR is not activated and the ER does not expand. Thus, the UPR senses the levels of client proteins that sequester BiP versus those of BiP itself in a ratiometric fashion, which determines the amplitude of the response.

Results

Cytotoxicity of UPR-eliciting drugs

Drugs that are typically used to study the UPR include tunicamycin (Tm), which prevents the addition of N-glycans to nascent ER client proteins, dithiothreitol (DTT), which impedes disulfide bond formation, and thapsigargin (Tg), which depletes Ca2+ from the ER lumen. Their immediate effect is a general collapse of productive protein folding in the ER, and these drugs therefore activate the UPR (Walter and Ron, 2011). In the longer run, however, these drugs likely have pleiotropic effects. Evidently, non-productive folding in the ER causes ER client proteins to be retained in the ER (Ellgaard and Helenius, 2003), and thus they no longer reach their destination, be it anywhere throughout the endomembrane system of the cell or extracellular. As a result—depending on their half-life—these proteins will be depleted at the site where they have to exert their functions, which ultimately may cause a plethora of perturbations to the homeostasis of the cell.

In order to unmask pleiotropic effects of UPR-eliciting drugs, taking Tm as an example, we serially diluted the drug starting from a concentration (5 µg/ml), which is widely used to study the effects of ER stress and UPR activation, and monitored cell survival after 7 days by a cell growth/colony formation assay. At this high concentration, a full-blown UPR was triggered, as is evident from the maximal extent of XBP1 mRNA splicing in HeLa cells, that is, the appearance of a band of higher mobility, corresponding to the RT-PCR product of the XBP1S transcript from which the intron has been removed (Figure 1A). Yet, there was no cell survival after 7 days (Figure 1B), not even when the cells were treated with Tm only on the first day for 4 hr. At a 100-fold dilution of the ‘standard’ concentration, Tm induced the splicing of XBP1 mRNA to submaximal levels, and only transiently so (Figure 1A), but still provoked a substantial cell growth defect (Figure 1B,C). Even at a further two-fold dilution of Tm to 25 ng/ml, when no XBP1 mRNA splicing was detected (Figure 1A), there was some loss of cell growth (Figure 1B,C). These findings thus reveal that an overall impairment of ER client protein folding has cytotoxic consequences aside from the activation of the UPR. Indeed, Tm readily causes morphological aberrations of the ER at these low concentrations (Rutkowski et al., 2006). Strikingly, upon deletion of IRE1α by CRISPR/Cas9 and concomitant ablation of PERK and ATF6α by siRNA-mediated silencing (with good efficiency; Figure 1—figure supplement 1), cell growth was hardly impaired. The cytotoxic effects of Tm, however, became apparent in these cells at even lower concentrations than in cells with a functional UPR (Figure 1D), which implies not only that the UPR is predominantly cyto-protective but also that UPR-mediated pro-apoptotic pathways cannot play a decisive role in cell fate upon Tm-driven ER stress. Thus, the use of ER stress-eliciting drugs inevitably obscures how ER homeostatic readjustment can be achieved.

Figure 1 with 1 supplement see all
Tunicamycin has pleiotropic cytotoxic effects.

(A) HeLa S3 cells were treated with Tm at various concentrations and for various durations, as indicated, before mRNA was isolated from cells for RT-PCR analysis with oligos specific for XBP1. PCR fragments corresponding to spliced (XBPS) and unspliced (XBPU) XBP1 were separated on gel. A hybrid product that is formed during the PCR reaction (Shang and Lehrman, 2004) is denoted by an asterisk. The ratio of (XBPS)/(XBPS +XBPU) is indicative of the extent of IRE1α activation. (B) Cells as in (A) were seeded upon 1:5 serial dilution into 24-well plates and treated continuously or only for the first 4 hr upon seeding with various concentrations of Tm, as indicated. After 7 days of growth, cells were fixed and stained with crystal violet. (C) Quantitation of the crystal violet staining shown in (B) as a measure of cell growth, except for conditions that fully abrogated growth; n.a. = not assessed. Staining of untreated cells was set at 100%, and values are shown in a bar graph. Mean and s.e.m. are shown; n = 3. Statistical significance in a one-sample t-test of differences in crystal violet staining compared to untreated samples in (B) and replicate experiments was determined as a proxy for growth (**p≤0.01). (D) Cell growth assay as in (B) of HeLa S3 cells in which UPR transducers were deleted (IRE1α KO) or silenced (PERK KD and ATF6α KD) or not (WT), treated with Tm at various concentrations, as indicated.

https://doi.org/10.7554/eLife.27518.002

Proteostatic induction of the UPR as a model to assess ER homeostatic readjustment

We reasoned that pleiotropic effects of UPR-eliciting drugs likely obscure various important aspects of how cells cope with ER stress. We therefore sought to circumvent the shortcomings of using ER stress-eliciting drugs, and we hypothesized that overexpression of an ER client protein would be a better tool to allow us to appreciate UPR-driven adaptations to ER stress. The impressive ER expansion during B cell differentiation (Wiest et al., 1990; van Anken et al., 2003) inspired us to choose the bulk product of plasma cells, secretory antibody, as an ideal candidate to serve as a proteostatic insult that would drive the UPR. More specifically, we decided to employ the secretory heavy chain of IgM, µs, because µs expression in bulk coincides with elaborate ER expansion (Sitia et al., 1987) and UPR activation (Reimold et al., 2001) during plasma cell differentiation.

The precise role of the UPR in ER expansion during plasma cell development is, however, difficult to assess. B cells anticipate that they will secrete IgM in bulk once they differentiate into plasma cells (van Anken et al., 2003), such that ER expansion is also driven by developmental cues (Iwakoshi et al., 2003). Accordingly, ER expansion occurs in differentiating B cells even when µs is ablated (Hu et al., 2009). To assess the effect on cellular homeostasis solely of sudden µs overexpression, we decided to create a HeLa cell model that allowed inducible expression of murine µs, for which we exploited the GeneSwitch system (Thermo Fischer Scientific). By subsequent lentiviral delivery and cloning of integrant cells, we introduced the hybrid nuclear receptor that is activated by the synthetic steroid mifepristone (Mif), yielding HeLa-MifON cells. The µs transgene under the control of the promoter element responsive to that hybrid nuclear receptor was then integrated into the HeLa-MifON cells, yielding HeLa-µs cells.

Expression of µs in HeLa-µs cells in the absence of Mif was undetectable by immunoblotting (Figure 2A) or immunofluorescence (Figure 2B). The GeneSwitch system is amplified through an autoregulatory feedback loop, whereby the Mif-bound hybrid nuclear receptor upregulates expression of its own gene. This feature likely means that the titration of µs expression levels by varying Mif concentration was not easily reproducible, and that the Mif-driven inducibility was instead an OFF/ON effect (data not shown). A proxy for tunable expression of µs was offered instead by comparing different HeLa-µs clones, which—due to differences in genomic locus or copy number of integrations—each expressed µs at a different level upon induction with Mif, as shown for three individual clones (Figure 2A). As illustrated by immunofluorescence of µs (Figure 2B), expression was remarkably uniform among cells because they were clonal. Moreover, µs is retained in the ER lumen, as shown by its strict co-localization with the ER marker CRT (Figure 2C).

Figure 2 with 1 supplement see all
A model for proteostatically induced ER stress.

(A–F,H–J) Expression of µs in HeLa-µs cells was induced with 0.5 nM Mif. (A,G,H) The UPR was pharmacologically induced as a reference for 4 hr with 5 µg/ml Tm, or—for analysis of ATF6α in (A) only—for 1.5 hr with 300 nM Tg (since Tm leads to deglycosylation of ATF6α). (A,D) Immunoblotting revealed levels of µs, CHOP, BiP and α-tubulin. A shift to a lower mobility phosphorylated form, P-PERK, and the appearance of CHOP revealed activation of PERK. The release of the p50 cleavage product from the p90 precursor revealed activation of ATF6α. Cross-reaction of the secondary antibody against anti-ATF6α with µs is denoted. (A,G) Splicing of XBP1 was assessed as in Figure 1A. (A) Three different clones of HeLa-µs cells with decreasing µs expression levels: high (Hi), medium (Me), or low (Lo) were induced for 16 hr. Non-treated (nt) or Tm-treated Hi HeLa-µs served as references. (B,C,H) Immunofluorescence revealed µs (red) and CRT (blue), which marks the ER (C) at the indicated times before or after induction (B) or after 8 hr induction (C,H). Nuclei were stained with DAPI (blue) (B). (C,H) The area that is boxed is shown by 3.5-fold magnification; scale bars represent 10 µm. (D) Samples were derived from equal numbers (7 × 104) of HeLa-µs or I.29µ+ lymphomas induced with Mif or stimulated with 20 µg/ml lipopolysaccharide (LPS), respectively, for the indicated times. Ponceau staining of the blot serves as a loading control. (E) HeLa-µs cells were pulse labeled with 35S labeled methionine and cysteine for 10 min at the indicated times after induction. Immunoprecipitated µs was resolved by gel electrophoresis (F). Levels of radio-labeled µs in (E) were quantified by phosphor imaging. The maximal signal for µs (at 64 hr) was set at 100. (G) HeLa-MifON cells were treated with a high dose of Mif (10 nM) for the indicated times. (H) In HeLa-µs cells in which IRE1α was replaced with doxycycline (Dox)-inducible IRE1α-GFP (green), IRE1α-GFP expression was tuned with 10 nM Dox to levels that allowed satisfactory detection of IRE1α-GFP by fluorescence microscopy. (I) Cell growth assay as in Figure 1B of HeLa-MifON and HeLa-µs induced continuously with Mif or not. (J) Quantitation of (I), performed as in Figure 1C. Mean and s.e.m. are shown, n = 2. There is no statistical significance in a one-sample t-test of differences in growth between conditions.

https://doi.org/10.7554/eLife.27518.005

The autoregulatory feedback loop of the GeneSwitch system has the advantage that it allows highly abundant expression of the transgene. Indeed, in the most highly expressing HeLa-µs clone, µs reached intracellular levels that outmatched those in model plasma cells: I.29µ+ lymphomas that were stimulated with lipopolysaccharide (LPS) for 3 days (Figure 2D). Synthesis of µs reached a plateau after 1–2 days of consistently high levels, as assessed by radiolabeling (Figure 2E,F). Thus, the proteostatic UPR stimulus is permanent, which implies that the cells must undergo homeostatic readjustment to meet the increased demand on the ER folding machinery.

Using a series of standard assays, we then demonstrated that µs triggered a full-blown UPR. Upon 16 hr of induction of µs expression, splicing of XBP1 mRNA, phosphorylation of PERK (appearance of a band of slightly lower mobility), expression of PERK’s downstream effector CHOP, cleavage of the precursor ATF6α-p90 and the concomitant release of its cytosolic portion p50 were all at levels comparable to those triggered by a conventional Tm- or Tg-elicited UPR in these cells (Figure 2A). Treatment of the parental HeLa-MifON cells with a high dose of Mif (10 nM) did not provoke XBP1 mRNA splicing (Figure 2G). Thus, the UPR induction in Mif-treated HeLa-µs was the result of µs expression. Moreover, UPR activation upon µs induction was dose-dependent, as the extent of XBP1 mRNA splicing or CHOP expression was commensurate with the levels of µs that were expressed by the different HeLa-µs clones (Figure 2A).

Another hallmark of full-blown UPR activation is the clustering of IRE1α (Li et al., 2010). We deleted IRE1α by CRISPR/Cas9 and reconstituted it with GFP-tagged murine IRE1α (IRE1α-GFP) under the control of a tight Tet-responsive element (Clontech) by lentiviral transduction. The GFP-tag was introduced such that it did not interfere with function (Figure 2—figure supplement 1), as has been shown before for human IRE1α (Li et al., 2010) or yeast Ire1p (Aragón et al., 2009). With doxycycline (Dox), we tuned IRE1α-GFP expression to near-endogenous levels that allowed visualization by fluorescent microscopy and found that IRE1α-GFP clustered not only upon Tm treatment but also upon µs expression (Figure 2H). Altogether, we concluded that µs expression drove the activation of all branches of the UPR that we tested, in a similar manner and to a similar extent as when they were triggered by ER stress-eliciting drugs.

Strikingly, in spite of the expression of µs in bulk, the cell growth of the HeLa-µs cells almost paralleled that of the Mif-treated parental MifON cells (Figure 2I,J). Thus, the full-blown UPR that was elicited by µs expression was well tolerated by the cells, unlike prolonged treatment with Tm at concentrations that triggered a full-blown UPR (Figure 1). In all, we concluded that the HeLa-µs cells offer a suitable model to assess ER homeostatic readjustment and the role of the UPR in this process, as the cells apparently cope well with the proteostatic insult that is unabated once µs expression is induced.

UPR activation correlates with accumulation of ER load rather than with the secretory burden

By expressing µs, we burden the cells with a Sisyphean task because, in absence of immunoglobulin light chain λ, no IgM can be produced, a situation that mimics the storage of mutant ER client proteins in disease. Inspired by B cell to plasma cell differentiation, we argued that λ in contrast to µs would likely provide a poor if any proteostatic insult to the cells, as it is expressed in B lymphocytes when they still are quiescent, and—by definition—minimally stressed. Moreover, ER retention of λ is far less stringent than that of µs, such that λ can be secreted (Hendershot and Sitia, 2005).

To assess whether it is the folding load of client proteins passing through the ER or their accumulation that determines the amplitude of UPR signaling, we created clones that co-express the two IgM subunits and selected clones based on stoichiometric differences in λ and µs expression. As expected, when λ and µs were co-expressed, antibody assembly took place, and we monitored this by immunoblotting of lysates and of culture media separated by reducing (Figure 3A) or non-reducing (Figure 3B) gel electrophoresis. The clone in which λ was in excess allowed successful assembly into polymeric antibodies that were efficiently secreted. In clones where µs was in excess, assembly and secretion were less efficient, leading to intracellular accumulation of disulfide-linked species that are characteristic intermediates of the IgM assembly pathway (Hendershot and Sitia, 2005). As anticipated, HeLa cells inducibly expressing λ—at similarly high levels as reached for µs—did not activate the UPR (Figure 3C). Strikingly, the µs-provoked UPR was also almost fully mitigated when λ was co-expressed in excess, while the other clones showed intermediate UPR activation commensurate with an excess of µs (or lack of λ).

The accumulation of clients in the ER rather than the flux of secretory load drives the UPR.

(A–D) HeLa clones constructed to express λ inducibly, either alone (λ) or in conjunction with µss + λ) at different stoichiometries, ranging from µs being in excess to λ being in excess, all under control of Mif, were induced with 0.5 nM in parallel to the HeLa-µs cells for 16 hr to express IgM subunits. Non-treated (nt) HeLa-µs cells served as a reference. Supernatants (secreted) and cell lysates (intracellular) were separated by reducing (A) or non-reducing (B) gel electrophoresis. Levels of µs, λ (A), CHOP, BiP (C) and α-tubulin (A,C), as well as activation of the UPR pathways (C) were assessed as in Figure 2A. (B) Immunobloting of µs (red) and λ (green) reveals disulfide linked assembly intermediates, as indicated: µλ, µ2, µ2λ2, and high molecular weight (HMW) polymeric assemblies of µs and λ. Simultaneous immunodetection of both IgM subunits resulted in yellow signal. (D) Quantitation of BiP levels shown in (C) and replicate experiments. Mean and s.e.m. are shown, n = 3. Statistical significance in a one-sample t-test of differences in BiP levels is indicated (*p≤0.05). (E) Schematic of BiP associating with the CH1 domain of µs until it is displaced by λ (if available).

https://doi.org/10.7554/eLife.27518.008

As the levels of ER client proteins (the sum of intracellular λ and µs) were comparable in the different clones, we concluded that the flux of proteins entering the ER is no key determinant for UPR activation, but rather that the nature of the proteostatic insult determines the extent of UPR activation, with λ being a very poor UPR activator, and µs being a particularly powerful one, provided it does not team up with λ. Of note is that levels of BiP increased markedly when µs was expressed alone but far less so or not at all if λ was present (Figure 3D). Indeed, it has been established in plasma cells that BiP binds to µs until it is displaced by λ (Bole et al., 1986; Figure 3E).

UPR signaling is commensurate with the extent of BiP being eclipsed by the ER folding load

As the cells that inducibly overexpress µs showed maximal UPR activation with no major negative impact on cell growth—unlike cells treated with UPR-eliciting drugs—we surmised that the µs-expressing cells underwent successful homeostatic readjustment of the ER. To appreciate the timing of the response to the proteostatic insult and the process of ER homeostatic readjustment with a more detailed temporal resolution, we undertook further analysis by immunoblotting of the upregulation of µs, BiP and of another ER resident chaperone, PDI, with time (Figure 4A).

Figure 4 with 4 supplements see all
ER homeostatic readjustment entails a transition from acute to chronic UPR signaling.

(A–C) HeLa-µs cells were induced with 0.5 nM Mif for various times as indicated. Levels of µs, BiP, PDI (A) and α-tubulin (A,C), as well as activation of the UPR pathways (C), were assessed as in Figure 2A. (B) Levels of µs and BiP were quantified using known quantities of BiP and µs as standards. The X-axis displays time in hrs in a logarithmic scale to show better detail of the early phase of the time course. Experiments were normalized to the level of BiP at t = 64 hr, which was reproducibly ~10 fold that at t = 0 hr. Quantities of BiP and µs are expressed in equivalents of BiP levels at t = 0 hr; mean and s.e.m. are shown, n = 5. Fitting was performed with Prism software to obtain a sigmoidal dose-response curve. As quantities were not assessed at the same time points in each experiment, some values were inferred from fit curves to obtain the s.e.m. Statistical significance of differences in expression levels between BiP and µs were tested by two-tailed t-test, and are depicted in a color-based heat map ranging from dark blue for [BiP] > [µs] (***p≤0.001), via purple for when there is no statistical significance for differences between the levels of BiP and µs, that is [BiP] ≈ [µs], to red for [µs] > [BiP] (*p≤0.05). (D) Activation of the UPR branches in (C) and replicate experiments were quantified (n = 2–20, depending on the time point assessed): for IRE1α, the percentage of XBP1 mRNA splicing; for PERK, the levels of expression of its downstream effector CHOP; and for ATF6α, the levels of ATF6α-p50 (both the latter normalized on the levels of α-tubulin). Mean values of UPR signaling output at each time point were calculated as a percentage of the maximal level, and are depicted in a color-based heat map ranging from no color (min) to full color (max).

https://doi.org/10.7554/eLife.27518.010

We found that levels of BiP only noticeably increased from 12 hr onwards (and PDI even later), whereas µs built up steadily in the first 12 hr upon induction. Indeed, using recombinant BiP and commercially obtained pure IgM (of which ~70% of the protein mass is µs) as standards, we assessed the absolute quantities in cell lysates of BiP and µs upon induction of µs expression. BiP levels increased ~12.5 fold. On the basis of cell counting, determination of the protein weight of the samples and the known molecular weight of BiP of ~70 kDa, we estimated that BiP levels rise from ~2×107 to ~2×108 copies per cell. Importantly, from the absolute quantitation, we deduced that BiP levels were in excess of µs once the cells had adapted to the proteostatic insult (with µs reaching an estimated ~1.5–1.8×108 copies per cell). Early upon the onset of µs expression and before BiP induction was fully underway, however, µs levels were transiently at a 1:1 stoichiometry with those of BiP, or possibly, µs levels even slightly exceeded BiP levels (Figure 4B).

Following the activation of the UPR over time upon onset of µs expression, we found that all three UPR branches transiently reached an output that was maximal, that is at least on a par with the output that was induced by a conventional drug-elicited UPR (Figure 4C; Figure 4—figure supplements 13). Levels of cleaved ATF6α (p50) reached a maximum at ~6 hr after the onset of µs expression before it subsided to submaximal levels. The full-length precursor ATF6α (p90) was in large part depleted after this intial burst of ATF6α (p50) release (Figure 4C), such that the determining factor for the amplitude of signaling through the ATF6α branch was inevitably, the de novo synthesis of p90 rather than the severity of the proteostatic insult, at least for a while before levels of p90 were fully restored.

Levels of spliced XBP1 mRNA likewise reached a maximum at ~6 hr of µs expression, and remained maximal for about 24 hr (with perhaps some oscillations) before they subsided to submaximal levels. Levels of CHOP, which provides a readout for PERK activity, reached a maximum slightly later, at 12–16 hr after the onset of µs expression, before reaching submaximal levels, as further illustrated in a ‘heatmap’ of UPR activation (Figure 4D). Treatment with ER stress-eliciting drugs during the chronic phase when the UPR had submaximal output readily triggered close to maximal output of all three UPR branches (Figure 4—figure supplement 4), indicating that prolonged µs expression had not exhausted the signaling capacity of the UPR, and thus, that homeostatic readjustment of the ER accounted for lowered UPR output levels upon chronic µs expression. Strikingly, the highest output of the three UPR branches did not coincide with µhaving reached maximal expression levels. Instead, UPR signaling reached maxima when the ratio µs/BiP was highest (Figure 4B,D). These findings support a scenario in which the UPR signals are commensurate with the extent to which the folding machinery, in particular BiP, is sequested by µs rather than by µs accumulation per se.

The ER expands in response to a proteostatic insult

The increase of BiP levels upon µs expression indicated that the homeostatic readjustment to accumulating µs levels entailed expansion of the ER. To directly visualize ER expansion, we targeted a modified version of pea peroxidase (APEX) (Hung et al., 2014; Lam et al., 2015) to the ER by use of an N-terminal signal peptide and a C-terminal KDEL. APEX-KDEL was expressed under control of the TetT promoter at such low levels that it did not interfere with the µs-driven UPR (Figure 5—figure supplement 1). We then exploited APEX to catalyze polymerization of 3,3’-diaminobenzidine tetrahydrochloride (DAB) upon treatment with H2O2 to obtain contrast in electron microscopy (EM), and stained the ER of cells before or after 1, 3, or 7 days of µs expression.

We determined that, before induction of µs expression, the ER lumen occupies 10–12% of the area within the cytoplasm (i.e. excluding the nucleus) in the electron micrographs (Figure 5), which by a rough estimate would correspond to (0.10–0.12)3/2≈3–4% of the cytoplasmic volume. Considering that the volume of the cytosol in HeLa cells is estimated to be ~2×103 µm3, that is ~2×103 fl (Milo, 2013), we estimated the ER to have a volume of ~60–80 fl. The ER expanded in the course of 1–3 days upon the onset of µs expression to cover ~18% of the cytoplasmic area in the electron micrographs, corresponding to roughly (0.18)3/2≈7–8% of the cytoplasmic volume, but we observed no further ER expansion after 3 days. On the basis of these estimates, the volume of the ER increased 2–3-fold as a result of µs expression, to ~120–240 fl per cell.

Figure 5 with 1 supplement see all
ER expansion in the course of homeostatic readjustment to µs overexpression.

HeLa-µs-derived cells, harboring Dox-inducible APEX-KDEL, were induced with 0.5 nM Mif to express µs for various days as indicated, and APEX-KDEL expression was induced with 100 nM Dox for 2 days. APEX-KDEL was exploited to obtain DAB precipitates (dark), revealing the extent of the ER in electron micrographs. Boxed areas are shown by 3-fold magnification; scale bars represent 1 µm. The percentage of the area within the cytoplasm corresponding to ER was determined and depicted in bar graphs; mean and s.e.m. are shown, n = 10. Statistical significance of differences in the extent of ER occupying cytosolic area in the electron micrographs was tested by ANOVA (*p≤0.05; ***p≤0.001).

https://doi.org/10.7554/eLife.27518.016

The ER becomes dominated by the chaperone BiP in response to a proteostatic insult

We further corroborated the µs-driven ER expansion by proteomic analysis and identified ~3000 proteins from lysates of cells before or after 1, 3, or 7 days of µs expression. We then used a label-free quantitation approach to obtain an approximation of the mass levels of proteins (Cox et al., 2014) (Supplementary file 1), such that we could estimate what share of the total protein mass content of the cell the ER would account for. Proteins that we identified as ranking among the 500 most abundant at any of the time points, altogether representing ~600 proteins, were categorized according to subcellular localization and/or function. We limited detailed analysis to this subset, because the label-free quantitation method is more reliable for abundant proteins (Cox et al., 2014), and because this subset already covered >90% of the total protein content in the cell. Changes in protein expression levels were assessed based on the label-free quantitations and in parallel by SILAC (Ong et al., 2002), which revealed mostly similar trends in differential expression (Supplementary file 1) and confirmed the usefulness of the label-free quantitation approach.

We then calculated the percentage of the total protein for each category and deduced that the ER accounts for ~3% of total protein content before µs expression in line with the findings by EM. Upon µs expression, the ER content, including µs, expanded to ~10% within 3 days, but did not expand much further after that (~12% at day 7) (Figure 6, upper panels; Supplementary file 1). The estimate of the percentage of the total protein that the ER accounts for that was obtained by proteomics was slightly higher than the EM-based estimate of the percentage of the total volume that the ER occupies within the cell, possibly indicating that continuous µs expression led to increased molecular crowding within the ER. In contrast to the ER, other organelles or cellular machineries were not markedly affected by µs expression (Supplementary file 1).

BiP becomes dominant in the ER when it expands upon µs overexpression.

HeLa-µs cells were induced with Mif to express µs for various numbers of days as indicated. Proteins were identified at each time point by mass spectrometry and expression levels were approximated by label-free proteomics. The ~600 most abundant proteins, accounting for over 90% of the total protein content, were categorized according to localization in the cell—cytosolic, nuclear, mitochondrial, ER resident, or residing in other organelles—by the use of Uniprot entries. From the approximated quantities of these proteins, the quantities of proteins per organelle were calculated as a percentage of the total proteome and depicted in pie charts (upper panels). For the ER, the resident proteins (gold) and µs (red) are shown separately. Quantities of ER resident proteins including µs were calculated as a percentage of the ER proteome and are depicted in pie charts (lower panels) that are size-proportioned commensurate with the difference in combined quantity of ER and µs at different days. For details on calculations, see Supplementary file 1. Color-coding of pie charts is as annotated in the legend embedded in the figure.

https://doi.org/10.7554/eLife.27518.019

Levels of ER resident chaperones underwent drastic changes upon µs expression (Figure 6, lower panels; Supplementary file 1). Most notably, between day 0 and 7, levels of BiP increased ~8-fold (and ~14-fold according to SILAC measurements; Supplementary file 1), in good agreement with the absolute quantitations we obtained by immunoblotting (Figure 4B). In fact, at day 7, BiP became the second most abundant protein in the cells (after β-actin), amounting to ~4% of the total proteome mass. Put together with the absolute quantitation of BiP levels, we estimate that the cells harbor an average of 5 × 109 proteins, which fits with earlier assumptions (Milo, 2013). Moreover, from these quantitations, we estimated BiP levels to be at ~0.5 mM (or 30–40 mg/ml) before induction and to reach ~1.5 mM (or 90–120 mg/ml) in the ~3-fold expanded ER at day 7, accounting for about a third of the estimated ~200–300 mg/ml overall protein concentration in the lumen of the organelle. Other ER resident chaperones, such as GRP94, CRT and some members of the PDI family (most notably ERp72 and P5), also increased albeit to a lesser extent. As a consequence, levels of BiP within the ER increased from ~15% before the onset of µs expression to 30–40% of the expanded ER proteome. Levels of GRP170, which acts as a nucleotide exchange factor of BiP (Behnke et al., 2015), also increased markedly. The ERdj co-chaperones of BiP were upregulated upon µs expression as well, but, surprisingly perhaps, to levels that were still about two orders of magnitude lower than the levels of BiP itself (Supplementary file 1). Thus, their co-chaperone function must be executed in a sub-stoichiometric fashion in the HeLa-µs model.

Before induction, µs was barely detectable (~1 ppm) but levels rapidly increased to become ~10×103 ppm, that is ~1% (after 1 day) and ~2% (from day 3 onward) of the total protein content in the cell (Figure 6, upper panels; Supplementary file 1). Strikingly, 1 day of expression was sufficient for µs levels to exceed that of any chaperone in the ER except those of BiP, which still outmatched µs 3:2 (Figure 6, lower panels; Supplementary file 1). Yet, BiP levels had increased 3–4-fold during that first day to keep pace with the burden imposed on the ER folding machinery by µs. At later times, the margin by which BiP levels were in excess of µlevels increased again, with BiP and µs together constituting about half of the ER protein mass when it had fully expanded. Thus, upon µs expression in bulk, the ER content not only changed quantitatively (i.e. undergoing a ~ 3–4-fold expansion), but also qualitatively (i.e. becoming dominated by µs that is held in check by excess levels of its most devoted chaperone BiP).

UPR signaling is key for ER homeostatic readjustment

As cell growth and survival were hardly perturbed upon µs expression, we reasoned that ER homeostasis was successfully readjusted. The HeLa-µs model therefore permitted to ask the key question of whether UPR activation is essential for ER homeostatic readjustment. As we have shown above for wild-type HeLa cells (Figure 1D) ablation of all three UPR branches together hardly impeded cell growth in HeLa-µs cells in the absence of µs expression (Figure 7A,B), probably because HeLa cells by default handle a low secretory load (Supplementary file 1). Concomitant expression of µs, however, fully abolished cell growth (Figure 7A,B). Annexin V staining confirmed that µs expression caused synthetic lethality under conditions when all UPR branches were ablated (Figure 7C). Thus, we concluded that the UPR is essential for ER homeostatic readjustment upon µs expression.

The UPR is essential to sustain ER homeostatic readjustment upon µs expression.

(A,B,C) All UPR transducers were ablated, either by deletion (KO, IRE1α) or by silencing (KD, ATF6α and PERK) in HeLa-µs cells, and µs expression was induced with 0.5 nM Mif (+) or not (–), as indicated. (A,B) Cell growth was assessed as in Figure 1B and quantitated as in Figure 1C (B). Mean and s.e.m. are shown in a bar graph; n = 2. (C) Percentages of Annexin-V-positive cells before or upon induction of µs expression for 3 days were assessed by cytometric analysis. Means and s.e.m. are shown in a bar graph, n = 2. (B,C) Statistical significance of differences in growth (B) or Annexin V staining (C) were tested by ANOVA (**p≤0.01; ***p≤0.001).

https://doi.org/10.7554/eLife.27518.021

Discussion

By avoiding the use of ER stress-eliciting drugs that have cytotoxic side effects, we have created a cell model that allowed us to reevaluate several aspects of how cells respond to an accumulating load of client protein in the ER. First, our results highlight that the UPR predominantly acts as a pro-survival pathway. By confronting cells with a proteostatic stimulus (i.e. bulk µs expression) that these cells could well cope with, we confirmed that the UPR is key for their capacity to do so. Yet, the finding that µs expression readily causes cells to undergo apoptosis in the absence of these UPR transducers implies that other, as yet unknown, ER stress-induced mechanisms can be invoked to mediate cell death. These mechanisms may well play such a role also when the UPR is functional. For instance, hyperoxidizing conditions in the lumen of the ER activate NADPH oxidases, leading to efflux of radical oxygen species to the cytosol that may cause activation of pro-apoptotic pathways (Tabas and Ron, 2011).

Second, our data confirm that the accumulation of client proteins in the ER drives ER expansion. Interestingly, ER expansion is not necessarily provoked by bulk synthesis of client proteins per se, as co-expression of sufficient quantities of λ led to bulk secretion of IgM but failed to trigger a strong UPR, and, thus, did not result in augmented BiP production (which served as a proxy for ER expansion). Our findings indicate moreover that the ER folding machinery seems to be underused in HeLa cells. Apparently, mobilization of the default ER folding machinery is sufficient for a task that can be productively executed, such as the folding and assembly of IgM from subunits that are expressed in bulk but in the correct stoichiometry. In most cells, the basal size of the ER and its folding capacity may indeed exceed the need for folding assistance from the default secretory burden. In line with such a notion, a substantial fraction of BiP is kept in reserve in an inactive, AMPylated form under non-ER stress conditions (Preissler et al., 2015).

Third, our results highlight the key role of BiP in ER homeostasis. In fact, our choice to employ µs as a proteostatic insult to challenge the ER was motivated in part by the reasoning that if there were ever an ER client protein that would be excellent in sequestering BiP, it would be µs (Haas and Wabl, 1983; Bole et al., 1986). Furthermore, it is well-conceivable that even the slightest leakiness of unpaired µs from the ER has been evolutionarily selected against. As antigen recognition is likely to be compromised if µs is released from the cell when unaccompanied by λ, such a release would potentially lead to ill-fated off-target effects. Thus, BiP binding to µs must be highly stringent out of immunological necessity (Anelli and van Anken, 2013), which correlates with the fact that mutations in µs that lead to lowered affinity for BiP are associated with disorders of the immune system, such as lymphoproliferative heavy chain disease or myeloma (Hendershot et al., 1987; Anelli and Sitia, 2010).

To ascertain stringent ER retention of the accumulating load of µs, the pool of BiP and the ER at large expanded to an impressive extent, echoing ER expansion during B cell to plasma cell development. In the course of B to plasma cell differentiation, however, all cellular machineries develop in a strict sequential order that allows them to anticipate bulk IgM production (van Anken et al., 2003). Conversely, the purely µs-driven stimulus that we investigated in this study resulted almost exclusively in ER expansion. We found only meager indications of the expansion of even the remainder of the secretory machinery (Supplementary file 1). One could argue that the stringent retention of µs in the ER by BiP would obviate the need to reinforce the secretory pathway downstream of the ER. Accordingly, homeostatic regulation of the Golgi and compartments beyond appears to be governed foremost by dedicated signaling mechanisms distinct from the UPR (Luini and Parashuraman, 2016).

Importantly, we found that maximal UPR activation correlated with a shortage of BiP in matching the accumulating load of µs, rather than with accumulation of client protein in the ER per se. Indeed, the transitioning of UPR signaling to chronic submaximal levels coincided with the ER becoming dominated by BiP. Our findings are in line with the idea—which was first coined by David Ron and colleagues (Bertolotti et al., 2000)—that sequestering of BiP by ER clients (such as µs) is key for activating the UPR, as activation of the main three UPR sensors (IRE1α, PERK [Bertolotti et al., 2000] and ATF6α [Shen et al., 2002]) goes hand in hand with BiP dissociation from their respective luminal ER stress-sensing domains, whereas overexpression of BiP dampens UPR activation (Bertolotti et al., 2000). Accordingly, despite its accumulation in the ER, the PiZ mutant of α1-antitrypsin does not sequester BiP and fails to trigger the UPR (Graham et al., 1990). An in vitro study moreover suggests that sequestering of BiP by µs would be sufficient to activate IRE1α and PERK (Carrara et al., 2015). However, based on the crystal structure of yeast Ire1’s lumenal domain (Credle et al., 2005), Peter Walter and colleagues proposed instead that Ire1 activates through direct binding of unfolded proteins, for which in vitro evidence has subsequently been obtained for both yeast Ire1 (Gardner and Walter, 2011) and mammalian IRE1α (Karagöz et al., 2017).

Assuming that the UPR signaling amplitude is commensurate with that of ER stress sensing by the UPR transducers, it is tempting to conclude that the UPR transducers are in the OFF state (as ER stress sensors) when bound to BiP, and in the ON state when bound to clients. So in a three-way competition between formation of sensor-BiP, sensor-client, and BiP-client complexes, the ratio between sensor-client (ON) and sensor-BiP (OFF) will most robustly report on the ratio between client and BiP. Thus, the two models for ER stress sensor activation can be reconciled easily in a model of ratiometric ER stress sensing, in line with our observations that the UPR signaling output is highest when the levels of the ER client µs eclipse those of BiP, and that UPR signaling subsides to submaximal levels when an excess of BiP over the client is restored.

The finding that homeostatic readjustment of the ER entails that the UPR output becomes submaximal is also relevant for disease. Various genetic disorders stem from mutations in ER client proteins that lead to their misfolding and accumulation in the ER. On the basis of our results, cells that suffer from such conditions will likely adapt and display submaximal UPR signaling levels similar to those that we found for chronic µs expression. In other words, submaximal UPR signaling is not a sign of ‘mild’ ER stress, but rather reflects that the stress is chronic.

Finally, since successful ER homeostatic readjustment upon µs-driven proteostatic ER stress critically depends on the UPR, the HeLa-µs cell model provides an excellent tool to dissect in greater detail how the UPR alleviates ER stress, and how it may serve to evaluate the success of ER homeostatic readjustment. The latter aspect is key, because a variety of disorders, ranging from cancer to diabetes, tie in with aberrant or maladaptive UPR-driven cell fate decisions (Wang and Kaufman, 2012).

Materials and methods

Cell culture and Reagents

All reagents were obtained from Sigma-Aldrich, Milan, Italy, unless otherwise stated. HeLa S3 cells, of which the genotype was confirmed by PCR single locus technology, and all derivate lines (Supplementary file 2A) were cultured in DMEM (Thermo Fischer Scientific, Monza, Italy) containing glutamax (1 mM), 5% Tet-System approved Fetal Bovine Serum (FBS, Takara, Jesi, Italy), 100 U/ml penicillin and 100 μg/ml streptomycin. Cells were routinely tested (on a monthly basis) and found to be mycoplasm-free by use of a standard diagnostic PCR. Expression of transgenes was induced with 0.5 nM Mif unless indicated otherwise, and/or Dox at various concentrations as indicated. I.29µ+ lymphomas were used as model B lymphocytes (Alberini et al., 1987), and cultured in suspension in RPMI (Thermo Fischer Scientific) supplemented with 10% LPS free FBS (GE Healthcare, Milan, Italy), glutamax (1 mM), penicillin (100 U/ml), streptomycin (100 µg/ml), sodium pyruvate (1 mM) and β-mercaptoethanol (50 µM). I.29µ+ cells were induced to differentiate with 20 µg/ml LPS.

Generation of HeLa-derived cell lines that inducibly express transgenes

The pSwitch cassette (GeneSwitch system; Thermo Fischer Scientific), placed into a lentiviral vector as described (Sirin and Park, 2003)—a kind gift from Dr Frank Parker—was used to render cells Mif-responsive, as the pSwitch cassette encodes a hybrid nuclear receptor that is activated by Mif to drive expression of genes under the control of the GAL4 promoter. Similarly, a reverse tetracycline-dependent transactivator (rtTA) cassette (Zhou et al., 2006), under control of a bidirectional promoter with ∆LNGFR in the reverse orientation for selection purposes (Amendola et al., 2005), in a lentiviral vector was employed to render cells Tet (and thus Dox)-responsive. A pGene5b (GeneSwitch system) encoding lentiviral construct (Sirin and Park, 2003)—another kind gift from Dr Frank Parker—was used as backbone to place either Igµs or Igλ under control of the GAL4 promoter. The coding sequences for NP-specific murine µs and λ were derived from plasmids described elsewhere (Sitia et al., 1987; Fagioli and Sitia, 2001). pENTR223.1 #100061599 bearing murine IRE1α cDNA was obtained from Thermo Fisher Scientific. GFP was placed in frame at the same position in the juxtamembrane cytosolic linker portion of IRE1α as described (Li et al., 2010). The IRE1α-GFP cassette was placed under control of a TetTight (TetT)-responsive element (Takara) in a lentiviral vector. The APEX2 coding sequence (Lam et al., 2015), derived from the #49385 plasmid (Addgene, Cambridge, MA, USA), and modified to contain in frame extensions encoding the vitronectin signal peptide at the N-terminus and the tetrapeptide KDEL at the C-terminus, was placed under control of a TetT promoter in the same lentiviral vector as employed for IRE1α-GFP. Standard techniques were used for construction, transformation and purification of plasmids. Transgene cassettes were genomically integrated in a subsequent manner into HeLa S3 cells by lentiviral delivery, essentially as described (Amendola et al., 2005). Cells with genomic integrations of transgenes were cloned by limiting dilution to yield the cell lines used in this study as summarized in (Supplementary file 2A).

Ablation of UPR pathways, and pharmacological activation of the UPR

CRISPR/Cas9 was used to delete endogenous IRE1α in HeLa S3 or derivate cells (Supplementary file 2A), and clones were obtained by limiting dilution and verified by their lack of XBP1 mRNA splicing upon Tm treatment. Gene silencing of ATF6α, IRE1α, PERK, and XBP1 was performed with pooled ON-TARGET plus siRNA (GE Healthcare) (Supplementary file 2C) according to the manufacturers’ instructions. Cells were seeded the day after transfection with the siRNA pools. The UPR was pharmacologically elicited with either 300 nM Tg or 5 µg/ml (or lower) Tm.

XBP1 mRNA splicing assay

Total RNA was extracted from cells using the UPzol RNA lysis reagent (Biotechrabbit, Hennigsdorf, Germany) followed by a standard protocol provided by the manufacturer for assessing XBP1 mRNA splicing levels: cDNA was obtained from samples and amplified by PCR. Oligos used to amplify cDNA corresponding to XBP1 mRNA have been described (Calfon et al., 2002). PCR products were resolved on agarose gels and images were acquired with a Typhoon FLA-9000 reader (GE Healthcare). The percentage of spliced XBP1 mRNA was calculated as described (Shang and Lehrman, 2004).

Cell growth assay

To assess growth, cells were counted with a Burker chamber, and seeded in 24-multiwell plates in 1:5 serial dilutions (5000, 1000, and 200 cells per well) to grow for 7 days. Culture media and pharmacological agents were added as soon as cells attached after plating and refreshed every 2–3 days. Cells were fixed in methanol-acetone (1:1) for 10 min, stained with 0.5% crystal violet in 20% methanol for 10 min and washed with distilled water. Dried plates were densitometrically scanned at a resolution of 50–100 μm with the Typhoon FLA-9000 reader, employing the 647 nm laser and the photomultiplier 1000. Intensity of crystal violet staining was analyzed with ImageJ for quantitation of growth. Typically, wells seeded with 1000 cells were used for comparison of signal intensity between conditions. An empty well on the same plate served for background subtraction.

Protein analysis

Protein extraction, sample preparation, electrophoresis on 10% or 4–12% Bis/Tris precast polyacrylamide gels (Thermo Fischer Scientific), and transfer of proteins onto nitrocellulose (GE Healthcare) for immunoblot analysis with antibodies listed in Supplementary file 2B were performed using standard techniques with the following exceptions: for analysis of CHOP and ATF6α cells, total lysates were used to ensure that the nuclear pools of these proteins were solubilized; for analysis of disulfide-linked IgM assembly intermediates, reducing agents were omitted from the lysis buffer and free sulfhydryl groups were alkylated with N-ethyl-maleimide (NEM); for analysis of ATF6α, 1.5 mm thick 8% polyacrylamide gels, prepared from a 30% polyacrylamide-bis 29:1 solution (Biorad, Milan, Italy), were used as described previously (Maiuolo et al., 2011). Protein transfer onto nitrocellulose was confirmed by reversible Ponceau staining. Protein quantitation of samples was performed using a bicinchoninic acid.

Detection of fluorescent antibody signals on blots was performed by scanning with the Typhoon FLA-9000 reader, except for IRE1α and ATF6α, which were detected on films using HRP-conjugated secondary antibodies and ECL. Signal intensities were analyzed with Image J. For absolute quantitation of BiP and µs protein levels, standard µs and BiP curves were obtained. To that end, we used purified mouse-myeloma-derived IgM (#02–6800, Thermo Fischer Scientific), of which we estimated 70% of the protein mass to be µs, and purified recombinant hamster BiP, which save for two conservative changes, Y313F and A649S, is identical in sequence to human BiP and therefore, in all likelihood, is recognized equally well by the goat anti-BiP antibody (C-20; Santa Cruz, Heidelberg, Germany). Recombinant BiP was expressed and purified as described (Marcinowski et al., 2011) from a pPROEX expression construct that was a kind gift from Dr Johannes Buchner. To avoid potential crossreaction of secondary antibodies against µs, we used an anti-IgM antibody that was itself Alexa-546-conjugated to allow fluorescent detection.

Radiolabeling

Cells were starved for 10 min in standard medium, containing 1% dialyzed FBS but lacking cysteine and methionine, prior to 10 min pulse labeling with Express [35S] Protein labeling mix (Perkin Elmer, Milan, Italy) containing 40 µCi 35S-methionine and 35S-cysteine per 106 cells. Cells were harvested and washed twice in ice-cold HBSS (Thermo Fischer Scientific). After a wash in ice-cold PBS, cells were lysed in RIPA buffer containing NEM and protease inhibitors for 10 min on ice, as described previously (Fagioli and Sitia, 2001). Lysates were cleared for 10 min at 13,000 rpm at 4°C. Before immunoprecipitation, cell lysates were pre-cleared for 1 hr with 30 μl FCS-Sepharose (GE Healthcare) and μs was immunoprecipitated for 16 hr using a rabbit anti-mouse IgM (H) antibody (#61–6800, Life Technologies). Immunoprecipitates were collected on protein G-Agarose beads (Thermo Fischer Scientific), washed twice in 10 mM Tris (pH 7.4), 150 mM NaCl and 0.5% NP-40, and once in 5 mM Tris-HCl (pH 7.5), before gel electrophoresis. Gels were transferred and dried onto a 3 MM filter paper, and exposed to a LE storage phosphor screen (GE Healthcare). Signals were acquired on the Typhoon FLA-9000 with a phosphorimaging filter. Densitometric analysis of signals was performed with ImageJ.

Fluorescence microscopy

Samples were prepared for immunofluorescence as described previously (Sannino et al., 2014), and sample-containing coverslips were mounted on glass slides with Mowiol. Light microscopic images were acquired with an UltraView spinning disc confocal microscope operated by Volocity software (PerkinElmer). Images were processed with Photoshop (Adobe, San Jose, CA, USA). Antibodies used were: Alexa-546 anti-mouse IgM (µ) 1:1000 (Life Technologies, Monza, Italy); rabbit anti-calreticulin 1:200, and secondary Alexa-488-anti-rabbit 1:300 (Life Technologies). Nuclei were stained with DAPI.

Apoptosis assay

As a measure for apoptosis, the percentage of Annexin-V-positive cells was recorded by use of a standard detection assay (APC-Annexin V, BD Biosciences, Milan, Italy) on a Canto cytometer (BD Biosciences) following the manufacturers’ instructions.

Electron microscopic determination of ER size

Cells harboring APEX2-KDEL were fixed in 1% glutaraldehyde, 0.1 M sodium cacodylate (pH 7.4) for 30 min, and incubated for 20 min with 0.3 mg/ml DAB, and for 20 min with 0.03% H2O2 (to activate APEX) in 0.1 M sodium cacodylate (pH 7.4), all at room temperature. Samples were rinsed in 0.1 M sodium cacodylate (pH 7.4), and post-fixed with 1.5% potassium ferrocyanide, 1% OsO4, sodium cacodylate (pH 7.4) for 1 hr on ice. After en bloc staining with 0.5% uranyl acetate overnight at 4°C in the dark, samples were dehydrated with increasing concentrations of ethanol, embedded in EPON and cured in an oven at 60°C for 48 hr. Ultrathin sections (70–90 nm) were obtained using an ultramicrotome (UC7, Leica microsystem, Vienna, Austria), collected, stained once more with uranyl acetate and Sato’s lead solutions, and visualized in a transmission electron microscope (Leo 912AB, Carl Zeiss, Oberkochen, Germany). Digital micrographs were taken with a 2K × 2K bottom-mounted slow-scan camera (ProScan, Lagerlechfeld, Germany) controlled by EsivisionPro 3.2 software (Soft Imaging System, Münster, Germany). Images of randomly selected cells (10 for each condition) were acquired at a nominal magnification of 1840x. Using ImageJ software, cytoplasmic regions were selected manually and ER profiles were segmented, thanks to the electron-dense DAB precipitate, by means of an automatic macro based on threshold intensity, yielding the percentage of the area within the cytoplasm that was ER. For simplicity, we used the formula: volume (%) = (area (%))3/2—which would be fully accurate only if the ER were cuboidal, but which we considered a good approximation nonetheless—to obtain a rough estimate of the ER volume.

Proteomics analysis

SILAC heavy and light collected serum-free cell media were mixed proportionally to a 1:1 cell number ratio and lysed with 8 M Urea 10 mM Tris-HCl buffer. Proteins were then quantified by Bradford. About 5 µg of mixed proteins for each sample were reduced by TCEP, alkylated by chloroacetamide, and digested by Lys-C and trypsin (Kulak et al., 2014), before peptides were desalted on StageTip C18 (Rappsilber et al., 2003).

Samples were analyzed in duplo on a LC–ESI–MS-MS quadrupole Orbitrap QExactive-HF mass spectrometer (Thermo Fisher Scientific). Peptides separation was achieved on a linear gradient from 93% solvent A (2% ACN, 0.1% formic acid) to 60% solvent B (80% acetonitrile, 0.1% formic acid) over 110 min,and from 60% to 100% solvent B in 8 min at a constant flow rate of 0.25 µl/min on UHPLC Easy-nLC 1000 (Thermo Fischer Scientific) connected to a 23 cm fused-silica emitter of 75 µm inner diameter (New Objective, Inc. Woburn, MA, USA), packed in-house with ReproSil-Pur C18-AQ 1.9 µm beads (Dr Maisch Gmbh, Ammerbuch, Germany) using a high-pressure bomb loader (Proxeon, Odense, Denmark). MS data were acquired using a data-dependent top 20 method for HCD fragmentation. Survey full scan MS spectra (300–1650 Th) were acquired in the Orbitrap with 60,000 resolution, AGC target 3e6, IT 20 ms. For HCD spectra, resolution was set to 15,000 at m/z 200, AGC target 1e5, IT 80 ms; NCE 28% and isolation width 2.0 m/z.

For identification and quantitation, raw data were processed with MaxQuant version 1.5.2.8 searching against the database uniprot_cp_human_2015_03 + sequences of µs and the Mif responsive hybrid nuclear receptor Switch setting labeling Arg10 and Lys8, trypsin specificity and up to two missed cleavages. Cysteine carbamidomethyl was used as fixed modification, methionine oxidation and protein N-terminal acetylation as variable modifications. Mass deviation for MS-MS peaks was set at 20 ppm. The peptides and protein false discovery rates (FDR) were set to 0.01; the minimal length required for a peptide was six amino acids; a minimum of two peptides and at least one unique peptide were required for high-confidence protein identification.

The lists of identified proteins were filtered to eliminate known contaminants and reverse hits. Normalized H/L ratios were analyzed via Perseus (version 1.5.0.6). Statistical analysis was performed using the Significance B outlier test where statistical significance based on magnitude fold-change was established at p<0.05. To look for proteins that changed over time, we considered Intensity L and Intensity H normalized by the correspondent Summed Intensity and the statistical ANOVA test analysis was performed using Perseus (ver. 1.5.0.6) with p<0.01. All proteomic data as raw files, total proteins, and peptides identified with relative intensities and search parameters have been loaded into Peptide Atlas repository (ftp://PASS01009:PJ3566i@ftp.peptideatlas.org/). To obtain approximations for absolute protein quantities, we followed a MaxLFQ label-free quantification strategy, as described previously (Cox et al., 2014). We then assessed the relative abundance of total protein per subcellular compartment, that is cytosol, nucleus, mitochondrion, ER, and other organelles, as well as µs, as well as the relative abundance of protein species, including or excluding µs, within the ER as detailed in the legend of Supplementary file 1.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
    Accumulation of the insoluble PiZ variant of human alpha 1-antitrypsin within the hepatic endoplasmic reticulum does not elevate the steady-state level of grp78/BiP
    1. KS Graham
    2. A Le
    3. RN Sifers
    (1990)
    The Journal of Biological Chemistry 265:20463–20468.
  17. 17
  18. 18
  19. 19
  20. 20
    Immunoglobulin assembly and secretion
    1. LM Hendershot
    2. R Sitia
    (2005)
    In: T Honjo, F. W Alt, M. S Neuberger, editors. Molecular Biology of B Cells. Elsevier Academic Press. pp. 261–273.
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
    Regulation of membrane IgM expression in secretory B cells: translational and post-translational events
    1. R Sitia
    2. MS Neuberger
    3. C Milstein
    (1987)
    The EMBO Journal 6:3969–3977.
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52

Decision letter

  1. Davis Ng
    Reviewing Editor; Temasek Life Sciences Laboratory and National University of Singapore, Singapore

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "The unfolded protein response turns from pro-survival to pro-death to cap the extent of endoplasmic reticulum expansion" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom, Davis Ng (Reviewer #1), is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Randy Schekman as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Peter Walter (Reviewer #2).

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

Summary:

The authors set the case that commonly used agents to induce the UPR, while useful to study signaling, are less desirable for studying the physiological decisions made by the cell to cope with stress due to their pleiotropic effects. The system set up is the controlled expression of the immunoglobulin M heavy chain in HeLa cells. The inspiration for their choice is the established system of B cell development into antibody secreting plasma cells. The process involves the UPR to remodel the ER for differentiation. Here, µs high-level expression causes strong UPR induction that is moderated when the lamda chain is co-expressed consistent with other studies where it is not the volume of the flux but the nature of the protein that is the primary determinant for the extent of UPR induction. The authors then combine quantitative proteomic analysis with ultrastructural measurements of growing ER to their induction profile. Together, these data were used to estimate the changes in steady state protein levels throughout the cell with a focus on ER localized proteins involved in protein homeostasis. Perhaps not surprisingly, BiP and other glucose regulatable chaperones showed the greatest changes over an expanding ER proteome that correlates with ER membrane expansion. The system is well suited the answer the long-standing question of when the UPR switches from being cyto-protective towards being pro-apoptotic.

Although there are specific concerns (see below), the study has many strengths. Of particular interest is the analysis of µs accumulation and how that correlates with the extent of UPR induction. One interesting finding was the following. "Instead, UPR signaling reached maxima when the ratio μs/BiP was highest (Figure 4B,D). These findings support a scenario in which the UPR signals are commensurate with the extent of μs sequestering the folding machinery, in particular BiP, rather than its accumulation per se." The idea of BiP limitation has been proposed for how the UPR activates, particularly Ire1 but to my knowledge, this is the first time the ratio of a specific stressor has been measured against BiP coordinately with all 3 arms of the UPR. However, the authors should be cautioned that the data in the manuscript do not distinguish the alternate mechanisms so such speculation should be stated as such.

The study is generally well designed and executed. It does an excellent job in validating the protein-based approach to induce ER stress. However, it is in the application of the method where authors encounter problems. In a number of experimental lines, the authors draw conclusions that are not fully supported by the data. In such cases, it is recommended that the text be changed to better reflect the data and any speculation be moved from the Results to the Discussion. In some cases, it might be better to move the data for future publication when the conclusions can be better substantiated. Of broad concern to reviewers is the writing style. There is an overuse of colorful/ornate terms that muddle up the authors' intended points. Because this occurs throughout the manuscript (some examples are outlined below), another round of careful editing is necessary. In the revision, it is also recommended that the text be shortened significantly to improve readability.

The manuscript in the present form is not yet suitable for publication. However, with substantial reorganization of the manuscript and figures, along with clarifications, this study will be of significant interest to the broad audience of eLife.

Revisions:

1. Figure 1. The authors state that 50 ng/ml Tm induces "marginal" XBP1 mRNA splicing. Looking at the data, the level of XBP1 mRNA splicing seems to approach 50%, which is by no means modest even when not maximal. Please correct. In addition, it is not accurate to state that a dose of 25 ng/ml Tm "fails" to engender XBP1 mRNA splicing. This is an assumption based on their PCR data, but it should be tested whether or not such low dose induces XBP1 mRNA splicing using specific primers to pick up on spliced from only.

2. In Figure 2A, there seems to be a missing a gel crop in the PERK blot between lanes 4-5. In Figure 2D, the Ig levels are not "comparable" to plasma cells; they appear to be much higher. Please correct or rephrase.

3. Figure 3. The authors overstate how their results distinguish flux of proteins entering the ER from simple accumulation of unfolded proteins in the ER lumen. Flux is not measured directly in this experiment, and an alternative explanation for the data is that IgL simply competes with BiP for binding sites on μs, resulting in an apparent "rescue" that results from the freeing of BiP that would otherwise be bound to μs.

4. Figure 4. The rise in BiP levels over those of Ig heavy chain is marginal. The authors should include metrics for statistical significance for each data point.

5. Figure 6. The authors should include quantification next to the corresponding pie chart slices for other organelles.

6. The conclusion that BiP behaves essentially as the ER "fuel gauge" enabling the cell to continue to expand the ER and attempt homeostasis is well supported by the data. However, there are no data to dismiss that direct binding of unfolded Ig heavy chain peptides could not engage the ER stress sensors as well. It could be possible that by flooding the ER with a single type of client, UPR activation results from the concomitant actions of the single ER client serving as BiP sink and unfolded peptides engaging the ER stress sensors. Moreover, it could be possible that the direct binding model could also help in fine-tuning the response. The data in this paper in no way refute the model that ER sensors are activated directly by unfolded protein binding.

7. Why is there no increase in BiP levels in the positive control (far left lane)?

8. The authors claim that Ire1a was replaced with Ire1a-GFP in cells. I don't see any description of CRISPR tagging, use of IRE1 KO cells or otherwise. This needs to be fully described.

9. How long were cells induced with Mif? While growing in the plates, only prior to placing on plates? This is not clear from the figure legend.

10. Figure 4D is a misrepresentation of the data. The data are not collected continuously. The time line and shading suggest continuous sampling. A standard plot with time points is the proper way to display these data.

11. Figure 5 lower panels. The dark blue histograms are often difficult to see against the dark EM background. Please use a better color or make a distinct histogram figure. Separating the histograms is not a great way to compare the data. The histograms should be presented side by side. This is true for 8F and 9E, too.

12. Figure 6. The choice of two dark blues for the pie charts is confusing as it inadvertently suggests the total protein increase is BiP. Pick a different color for the ER in the total proteome.

Since lumenal J proteins are key BiP co-chaperones and known to increase robustly with ER stress, it would make sense to include these in the analyses.

13. Abstract: "the UPR transitions from acute to full-geared activation…". The terms "acute" and "full-geared" are confusing in this context. Aren't they the same? Acute is generally meant to mean severe.

14. The authors may wish to restate that the µs induced HeLa cells "triggered a full-blown UPR" (page 5 bottom) for a number of reasons. First, the time of induction with tunicamycin or thapsigargin is 1.5 h versus 16 h for µs expression. They really aren't directly comparable. Second, the extent of induction is different. The extent of XBP1 splicing and ATF6 processing is less in the µs lanes. There is more BiP protein at steady state in the µs expressed cells but that's misleading because that's 16 h of induced expression versus 1.5 h and BiP is a very stable protein. Third, if µs expression did lead to "full-blown" UPR activation (vague, but defined in the text as Tm/Tg treatment), I see that as problematic because the authors have argued persuasively that such treatment is non-physiological and toxic. Based on that argument, I would expect that ER stress that is manageable would cause a more modest level of induction, which is exactly what is observed.

15. The passage: "Thus, the full-blown UPR that was elicited by μs expression was well tolerated by the cells, unlike prolonged treatment with Tm at concentrations that triggered a full-blown UPR (Figure 1)." Is tolerance due to UPR induction? Has this experiment been performed in cell lines deficient in any of the 3 UPR branches?

16. The passage: "Since the cells that inducibly overexpress μs showed maximal UPR activation with no major negative impact on cell growth-unlike cells treated with UPR eliciting drugs-we surmised that the μs expressing cells underwent successful homeostatic readjustment of the ER." This statement seems to contradict the stated basis for setting up this system, that the drugs are toxic, not because they simply induce the UPR, but that they disrupt the function of many proteins.

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

Thank you for resubmitting your work entitled "The unfolded protein response turns from pro-survival to pro-death to cap the extent of endoplasmic reticulum expansion" for further consideration at eLife. Your revised article has been favorably evaluated by Randy Schekman (Senior editor), a Reviewing editor, and two reviewers.

Although the manuscript is improved from the original version, there are claims that reviewers view as not fully supported. While the consensus view remains that the µs overexpression system of ER stress is well designed and characterised, the use of the system to understand specific questions of UPR physiology are not as well designed. Because the study is extensive, it is the advice of the reviewers to restate the claims as instructed and/or remove data that can be published elsewhere with additional work. It is our view that the manuscript can be published after some extensive reorganisation of the data and major text changes.

[Editors’ note: the following was also passed to the authors via correspondence with the editorial staff, further an author query.]

Regarding splitting the paper into two, it is the consensus view of the reviewers that the innovation of establishing a more physiological form of inducing ER stress and its validation to be the strength of the manuscript. That should be the focus of paper one. For paper two, we would caution the authors to consider very carefully how to proceed. We believe this is where the authors wish to use the system to answer specific questions regarding how the UPR responds to and alleviates ER stress. While the goal is important and an excellent use of the system, it is clear that the authors tried to do too much here, which made the supporting data incomplete at this point. Unless the authors have expanded this part of the study significantly and completely addressed reviewer concerns, it is not recommended that this part to be co-submitted. It should be submitted separately at another time.

[Editors' note: following the editors’ advice, the authors resubmitted and further revisions were requested prior to acceptance, as described below.]

Thank you for resubmitting your work now entitled "Ratiometric sensing of BiP-client versus BiP levels by the unfolded protein response determines its signaling amplitude" for further consideration at eLife. Your revised article has been favorably evaluated by Randy Schekman (Senior editor), a Reviewing editor, and one reviewer.

The manuscript has been improved but there is one minor, but significant, issue that needs to be addressed before acceptance, as outlined below:

Reviewer #3:

The heavily revised manuscript is dramatically improved and I will be interested to see the companion manuscript. The wording now agrees well with the presented data and I consider the manuscript acceptable with one small correction.

I remain adamant that Figure 4D is not appropriate as presented. The data in 4C are not continuous. They are discrete. I have no problem with the authors presenting their data as a heat map. That is fine. However, it should be clear that each data point is a defined block representing a specific time point. The data, as currently plotted, in a smeared form indicate that there are multiple data points that blur together, as if data were sampled every few minutes. Just present the heat map as blocks instead of a smear and I accept the revised manuscript.

The request from the reviewer does not involve any additional experimentation but modification in presentation of the data.

[Editors’ note: the original Decision letter and Author response included references to data which have subsequently been omitted and are therefore not discussed in this edited version.]

https://doi.org/10.7554/eLife.27518.027

Author response

We were of course delighted to read that the reviewers found that “there are many strengths to the study”, that “the study is generally well designed and executed”, and that “this study will be of significant interest to the broad audience of eLife”. As you may recall, in the paper we present a HeLa cell model for endoplasmic reticulum (ER) stress driven by a proteostatic insult in the form of the heavy chain of IgM (µs). Since the cells can cope with the insult, we created a system that the reviewers agree “is well suited to answer the long-standing question of when the UPR switches from being cyto-protective towards being pro-apoptotic”.

Your letter moreover highlighted that “one interesting finding was the following: ‘Instead, UPR signaling reached maxima when the ratio μs/BiP was highest (Figure 4B,D). These findings support a scenario in which the UPR signals are commensurate with the extent of μs sequestering the folding machinery, in particular BiP, rather than its accumulation per se.’ (Results). The idea of BiP limitation has been proposed for how the UPR activates, particularly Ire1 but to my knowledge, this is the first time the ratio of a specific stressor has been measured against BiP coordinately with all 3 arms of the UPR”, and that our study”does an excellent job in validating the protein-based approach to induce ER stress”. We include point-by-point answers to the reviewers’ questions below.

[…] Revisions:

1. Figure 1. The authors state that 50 ng/ml Tm induces "marginal" XBP1 mRNA splicing. Looking at the data, the level of XBP1 mRNA splicing seems to approach 50%, which is by no means modest even when not maximal. Please correct. In addition, it is not accurate to state that a dose of 25 ng/ml Tm "fails" to engender XBP1 mRNA splicing. This is an assumption based on their PCR data, but it should be tested whether or not such low dose induces XBP1 mRNA splicing using specific primers to pick up on spliced from only.

In view of the reviewers’ considerations we adapted the text.

2. In Figure 2A, there seems to be a missing a gel crop in the PERK blot between lanes 4-5. In Fig. 2D, the Ig levels are not "comparable" to plasma cells; they appear to be much higher. Please correct or rephrase.

We thank the reviewers for their concern, but lanes 4 and 5 were adjacent already in the original PERK blot, so no separator line is necessary.

In Figure 2D, the Ig levels are not "comparable" to plasma cells; they appear to be much higher. Please correct or rephrase.

In view of the reviewers’ considerations we adapted the text.

3. Figure 3. The authors overstate how their results distinguish flux of proteins entering the ER from simple accumulation of unfolded proteins in the ER lumen. Flux is not measured directly in this experiment, and…

In view of the reviewers’ considerations we adapted the text.

[…] an alternative explanation for the data is that IgL simply competes with BiP for binding sites on μs, resulting in an apparent "rescue" that results from the freeing of BiP that would otherwise be bound to μs.

BiP indeed acts as a “place-holder” for 1 and tightly binds to µs (exclusively at the CH1 domain where 1 will bind), which we refer to in the text (and further highlight in a schematic figure). We adapted the text to make this point slightly clearer.

4. Figure 4. The rise in BiP levels over those of Ig heavy chain is marginal. The authors should include metrics for statistical significance for each data point.

We have performed the statistical analysis. Evidently, before the onset of µs expression, the excess of BiP over µs is statistically significant. Between 2-8 hours there is no statistically significant difference between BiP and µs levels. Then, between 8-12 hrs, µs exceeds BiP levels. The statistical significance is marginal, but in this time span also UPR output levels are highest. Between 16-32 hrs again there is no statistical difference between µs and BiP levels, whereupon BiP levels exceed µs levels in a highly significant manner in the chronic phase.

To report on the significance of the differences in expression levels in an intuitive manner (see also the response below to minor point 18), we thought to display a time bar with a color-coding scale, ranging from dark blue (BiP > µs; p < 0.001) to purple (BiP ≈ µs) to red (µs > BiP; p < 0.05).

Since we do not think that the method is accurate enough to establish exactly by what margin µs eclipses BiP levels, we want to emphasize the opposite of which we are very confident, namely, that BiP levels transiently do not eclipse those of µs. We therefore feel that the accompanying sentence in the text still best expresses how the data should be interpreted: “Early upon the onset of µs expression and before BiP induction was fully underway, however, µs levels transiently were at a 1:1 stoichiometry with those of BiP, or possibly, μs levels even slightly exceeded BiP levels (Figure 4B).”

5. Figure 6. The authors should include quantification next to the corresponding pie chart slices for other organelles.

The way the calculations of the data and the representation in this figure were designed was to emphasize that µs expression causes few if any changes in the proteome except that the ER increases considerably in protein mass in comparison with the rest of the cell. While we believe the approach we have chosen was ideal for demonstrating just that, we would be hesitant to give too much credence to the absolute values we obtained per organelle. To display them next to the graphs would distract from the message and suggest an accuracy that we cannot vouch for. Besides the real quantifications are already in Supplementary File 1 and easily traceable for who might be interested in those details.

6. The conclusion that BiP behaves essentially as the ER "fuel gauge" enabling the cell to continue to expand the ER and attempt homeostasis is well supported by the data. However, there are no data to dismiss that direct binding of unfolded Ig heavy chain peptides could not engage the ER stress sensors as well. It could be possible that by flooding the ER with a single type of client, UPR activation results from the concomitant actions of the single ER client serving as BiP sink and unfolded peptides engaging the ER stress sensors. Moreover, it could be possible that the direct binding model could also help in fine-tuning the response. The data in this paper in no way refute the model that ER sensors are activated directly by unfolded protein binding.

We reconsidered this issue and concluded that the models of UPR sensor activation are only in apparent contradiction, but physiologically interchangeable and therefore easy to reconcile. We adapted the Discussion to make this point clear.

7. Why is there no increase in BiP levels in the positive control (far left lane)?

We assume the reviewers allude to the BiP levels upon Tm treatment in Figure 2. Since cells were treated for only 4 hrs with 5 µg/ml Tm no increase of BiP is detectable at the protein level – it would require a longer Tm treatment (12 hrs) to perceive changes for BiP at the protein level, but at that stage cells are already quite sick from the Tm treatment.

8. The authors claim that Ire1a was replaced with Ire1a-GFP in cells. I don't see any description of CRISPR tagging, use of IRE1 KO cells or otherwise. This needs to be fully described.

The reconstitution is amply described in the Material and methods section of the manuscript and the validation of the reconstitution is shown in the figure supplement to Figure 2. Since the tool we built is almost identical to that built by Han Li (PNAS, 2010), we felt it would be superfluous to present it as a result, which is why we “relegated” the description of the knock-out and reconstitution to the Material and methods, but if the reviewers insist, we will promote them to the Results section.

9. How long were cells induced with Mif? While growing in the plates, only prior to placing on plates? This is not clear from the figure legend.

The cells were treated for 7 days with mifepristone, starting from when cells attached (usually 8 hrs) after plating. We now state that in the figure legend and describe the procedure better in the Material and methods section.

10. Figure 4D is a misrepresentation of the data. The data are not collected continuously. The time line and shading suggest continuous sampling. A standard plot with time points is the proper way to display these data.

We now provide standard plots for each UPR output that we have assessed in the Figure 4—figure supplements 13. We do understand the criticism that the color-based heat maps would suggest continuous sampling, but since we provide one example (Fig. 4B) and now also the plots summarizing the data for that and repeat experiments on which the heat maps are based as supplements, we believe that the reader will not be led astray in thinking that there was continuous sampling.

We believe that the color-based heat maps, in particular in combination with the color-based heat map we now produced to depict statistical significance of differences in expression levels between BiP and µs (see response above to minor point 4), give an easy to grasp and intuitive visual summary of the message we wish to convey, namely that UPR signaling output levels are commensurate with BiP levels running short relative to µs levels.

11. Figure 5 lower panels. The dark blue histograms are often difficult to see against the dark EM background. Please use a better color or make a distinct histogram figure. Separating the histograms is not a great way to compare the data. The histograms should be presented side by side. This is true for 8F and 9E, too.

We adapted the figures following the reviewers’ suggestions.

12. Figure 6. The choice of two dark blues for the pie charts is confusing as it inadvertently suggests the total protein increase is BiP. Pick a different color for the ER in the total proteome.

We adapted the figures following the reviewers’ suggestions.

Since lumenal J proteins are key BiP co-chaperones and known to increase robustly with ER stress, it would make sense to include these in the analyses.

Reports on the robust upregulation of ER lumenal J (ERdj) proteins are all based on fold changes of transcripts. The tendency in -omics studies to focus on fold changes leads to a particular emphasis in the interpretation of biological phenomena, which sometimes even may obscure other biologically meaningful aspects. In our study we have chosen to estimate protein levels in ppm of the total proteome, which gives another perspective on what are robust increases.

For instance, according to our proteomics data, PDI is upregulated “only” 1.6 fold, but since PDI is an abundant protein at basal conditions, being already 4.5x104 ppm, which equates to approximately 2x107 copies per non-stressed cell, it implies that more than 1x107 extra molecules of PDI are present after 7 days of µs expression. Ten million extra copies of PDI seems to us a rather robust change, despite the low fold change!

From the data presented in Supplementary File 1 it is evident that four ERdj proteins were so lowly expressed or not expressed at all in our model cell line that we did not detect them by proteomics: ERdj1 (DNAJC1), ERdj2 (SEC63), ERdj4 (DNAJB9), and ERdj7 (DNAJC25), while ERdj5 (DNAJC10) was barely detectable by proteomics. ERdj3 (DNAJB11) and ERdj6 (DNAJC3) were well-detectable by proteomics, but still in low amounts. Our proteomics survey estimated ERdj3 and ERdj6 to go up about 3-6 fold from about 2-6x102 ppm to 2x103 ppm, which would amount to 1x106 copies per cell for either ERdj protein after 7 days of µs expression; in other words there are about 1-2x106 total ERdj protein extra copies per cell, which arguably is a less robust change than that of PDI, even though at the mRNA level the fold change may be much more flagrant than for PDI. For comparison: we estimated BiP levels after 7 days of µs expression to be about 2x108 copies per cell. Thus, altogether it seems that BiP levels exceed those of its co-chaperones about a hundred fold. From these estimates it thus appears that either a single ERdj protein can cater 100 BiP proteins simultaneously or that a majority of BiP molecules operates without co-chaperoning.

Since our aim was to obtain an idea of how the cell, and, in particular the ER, changes in terms of protein (mass) content, we chose the “cut off” for in-depth analysis such that it pertained only to the proteins that were among the 500-600 most abundant proteins, as explained in the manuscript. The ERdjs did not make that cut and if we would display them in the way we display our data in the proteomics figure, the ERdjs would be still invisible, as the most lowly abundant protein within this top 500-600, TRAPd, is already barely visible in the pie diagram.

We therefore do not feel we should add data on ERdjs to the figure. For who would be interested in the ERdjs, their expression data are readily available in Supplementary File 1.We did however decide to briefly touch in the results on the finding that BiP vastly outnumbers its co-chaperones, the ERdjs. We moreover acknowledge better the current knowledge on ERdjs in the introduction by citing a recent review by Melnyk et al.

13. Abstract: "the UPR transitions from acute to full-geared activation…". The terms "acute" and "full-geared" are confusing in this context. Aren't they the same? Acute is generally meant to mean severe.

To avoid confusion we corrected the abstract such that there is now a comma between acute and full-geared (which lacked before): “acute, full-geared activation”.

14. The authors may wish to restate that the µs induced HeLa cells "triggered a full-blown UPR" (page 5 bottom) for a number of reasons. First, the time of induction with tunicamycin or thapsigargin is 1.5 h versus 16 h for µs expression. They really aren't directly comparable. Second, the extent of induction is different. The extent of XBP1 splicing and ATF6 processing is less in the µs lanes. There is more BiP protein at steady state in the µs expressed cells but that's misleading because that's 16 h of induced expression versus 1.5 h and BiP is a very stable protein. Third, if µs expression did lead to "full-blown" UPR activation (vague, but defined in the text as Tm/Tg treatment), I see that as problematic because the authors have argued persuasively that such treatment is non-physiological and toxic. Based on that argument, I would expect that ER stress that is manageable would cause a more modest level of induction, which is exactly what is observed.

15. The passage: "Thus, the full-blown UPR that was elicited by μs expression was well tolerated by the cells, unlike prolonged treatment with Tm at concentrations that triggered a full-blown UPR (Figure 1)." Is tolerance due to UPR induction? Has this experiment been performed in cell lines deficient in any of the 3 UPR branches?

16. The passage: "Since the cells that inducibly overexpress μs showed maximal UPR activation with no major negative impact on cell growth-unlike cells treated with UPR eliciting drugs-we surmised that the μs expressing cells underwent successful homeostatic readjustment of the ER." This statement seems to contradict the stated basis for setting up this system, that the drugs are toxic, not because they simply induce the UPR, but that they disrupt the function of many proteins.

The above three points that the reviewers raise are a bit bewildering to us as they seem to suggest (contrary to the earlier points that were raised) that the reviewers do not see what the merits of our model using a proteostatic insult are with respect to “traditional” ER stress eliciting drugs.

In short: Tm etc. may trigger full-blown UPR output levels, i.e. maximally obtainable levels of splicing, ATF6a cleavage and CHOP expression, but ER homeostatic readjustment cannot be achieved, since the cells succumb to pleiotropic effects before. From the time course experiment that we show (Figure 4) it is evident that inducible µs expression leads to similar full-blown UPR output levels (albeit that maximal splicing levels are reached at a slower pace), but the cells do not suffer from pleiotropic effects and, thus, ER homeostatic readjustment can be achieved. As such, we were in a position to demonstrate that the UPR is essential for this ER homeostatic readjustment (Figure 7).

Thus, full-blown UPR output levels are not detrimental to the cells per se. It is the pleiotropic effects of the ER stress eliciting drugs that kills them. Cells die from Tm treatment likely due to glycoproteins being depleted at destinations where they have to exert their function, and similar reasoning would explain the toxic effects of DTT or Tg. The UPR to some extent mitigates these problems and thereby even prolongs the life of cells treated with Tm etc. Accordingly, we found that cells in which UPR pathways are ablated succumb at even lower doses of Tm than WT cells. This finding demonstrates that strategies to investigate potential pro-apoptotic roles of the UPR in the context of cells treated with ER-stress-eliciting drugs by default are in vain.

Since we apparently did not make this notion clear enough to the reviewers, as it seemed from minor points 24-26, we decided to include this finding in Figure 1 to highlight better why strategies to elucidate how the UPR can transition from a cyto-protective to pro-apoptotic output are in vain when ER-stress eliciting drugs are used. We adapted the text accordingly and moved the original Figure 7—figure supplement 1 to become Figure 1—figure supplement 1.

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

Although the manuscript is improved from the original version, there are claims that reviewers view as not fully supported. While the consensus view remains that the µs overexpression system of ER stress is well designed and characterised, the use of the system to understand specific questions of UPR physiology are not as well designed. Because the study is extensive, it is the advice of the reviewers to restate the claims as instructed and/or remove data that can be published elsewhere with additional work. It is our view that the manuscript can be published after some extensive reorganisation of the data and major text changes.

We were delighted to hear that our revised work entitled "The unfolded protein response turns from pro-survival to pro-death to cap the extent of endoplasmic reticulum expansion" has been favorably evaluated.

In spite of the positive evaluation, your letter stated that "there are claims that reviewers view as not fully supported", and the reviewers advised us "to restate the claims as instructed and/or remove data that can be published elsewhere with additional work", while the view at eLife was that "the manuscript can be published after some extensive reorganization of the data and major text changes."

If we follow the reviewers' advice, we simply remove the parts to which they have raised objections, and they still would like to see it published. If the editors would agree with this option, it would be ideal for us, since we are anxious too to get this work out. Yet, the second part contains almost all issues that have been central to the reviewing history of our manuscript so far at eLife, such that if we remove the second part from this submission, it would make sense to resubmit that second part as a separate manuscript to eLife again, since it would have been extensively reviewed at eLife already. We moreover think that the study becomes really too extensive for a single paper, in particular once extra data would be included if according to the reviewers they would be sufficient to satisfyingly support our claims.

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

Reviewer #3:

The heavily revised manuscript is dramatically improved and I will be interested to see the companion manuscript. The wording now agrees well with the presented data and I consider the manuscript acceptable with one small correction.

I remain adamant that Figure 4D is not appropriate as presented. The data in 4C are not continuous. They are discrete. I have no problem with the authors presenting their data as a heat map. That is fine. However, it should be clear that each data point is a defined block representing a specific time point. The data, as currently plotted, in a smeared form indicate that there are multiple data points that blur together, as if data were sampled every few minutes. Just present the heat map as blocks instead of a smear and I accept the revised manuscript.

The request from the reviewer does not involve any additional experimentation but modification in presentation of the data.

We adapted the figure as requested.

https://doi.org/10.7554/eLife.27518.028

Article and author information

Author details

  1. Anush Bakunts

    Division of Genetics and Cell Biology, San Raffaele Scientific Institute, Milan, Italy
    Contribution
    Conceptualization, Data curation, Formal analysis, Supervision, Validation, Investigation, Visualization, Methodology, Writing—review and editing
    Contributed equally with
    Andrea Orsi and Milena Vitale
    Competing interests
    No competing interests declared
  2. Andrea Orsi

    1. Division of Genetics and Cell Biology, San Raffaele Scientific Institute, Milan, Italy
    2. Università Vita-Salute San Raffaele, Milan, Italy
    Contribution
    Conceptualization, Data curation, Formal analysis, Supervision, Validation, Investigation, Visualization, Methodology, Writing—review and editing
    Contributed equally with
    Anush Bakunts and Milena Vitale
    Competing interests
    No competing interests declared
    ORCID icon 0000-0003-2839-1640
  3. Milena Vitale

    1. Division of Genetics and Cell Biology, San Raffaele Scientific Institute, Milan, Italy
    2. Università Vita-Salute San Raffaele, Milan, Italy
    Contribution
    Conceptualization, Data curation, Formal analysis, Supervision, Validation, Investigation, Visualization, Methodology, Writing—review and editing
    Contributed equally with
    Anush Bakunts and Andrea Orsi
    Competing interests
    No competing interests declared
    ORCID icon 0000-0001-7007-402X
  4. Angela Cattaneo

    IFOM, FIRC Institute of Molecular Oncology, Milan, Italy
    Contribution
    Data curation, Investigation
    Competing interests
    No competing interests declared
  5. Federica Lari

    Division of Genetics and Cell Biology, San Raffaele Scientific Institute, Milan, Italy
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  6. Laura Tadè

    Division of Genetics and Cell Biology, San Raffaele Scientific Institute, Milan, Italy
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  7. Roberto Sitia

    1. Division of Genetics and Cell Biology, San Raffaele Scientific Institute, Milan, Italy
    2. Università Vita-Salute San Raffaele, Milan, Italy
    Contribution
    Methodology, Funding acquisition
    Competing interests
    No competing interests declared
    ORCID icon 0000-0001-7086-4152
  8. Andrea Raimondi

    Experimental Imaging Center, San Raffaele Scientific Institute, Milan, Italy
    Contribution
    Investigation, Visualization, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
  9. Angela Bachi

    IFOM, FIRC Institute of Molecular Oncology, Milan, Italy
    Contribution
    Conceptualization, Data curation, Supervision, Investigation, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
  10. Eelco van Anken

    1. Division of Genetics and Cell Biology, San Raffaele Scientific Institute, Milan, Italy
    2. Università Vita-Salute San Raffaele, Milan, Italy
    Contribution
    Conceptualization, Data curation, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing
    For correspondence
    evananken@mac.com
    Competing interests
    No competing interests declared
    ORCID icon 0000-0001-9529-2701

Funding

Giovanni Armenise-Harvard Foundation

  • Eelco van Anken

Ministero della Salute (RF-2011-02352852)

  • Eelco van Anken

Associazione Italiana per la Ricerca sul Cancro (MFAG 13584)

  • Eelco van Anken

Ministero della Salute (PE-2011-02352286)

  • Roberto Sitia
  • Eelco van Anken

Associazione Italiana per la Ricerca sul Cancro (IG 2016-18824)

  • Roberto Sitia

Fondazione Telethon (GGP15059)

  • Roberto Sitia

Fondazione Cariplo (2015-0591)

  • Roberto Sitia

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Drs Bernhard Gendtner, Frank Park, and Johannes Buchner for sharing plasmids. We thank Dr Caterina Valetti for her advice on immunoblotting of ATF6α, Dr John Christiansson for help with RNA silencing, Dr Andrew Bassett for gRNA design, and Dr Stefano Bestetti for assistance with FACS analysis. All members of the ALEMBIC imaging facility and the Anna Rubartelli, Paola Panina, Sitia, Bachi and Van Anken labs are acknowledged for stimulating discussions and advice. We thank Drs Maurizio Molinari, Alison Forrester, John Christiansson, Tomás Aragón, and Luca Rampoldi for proofreading of the manuscript.

Reviewing Editor

  1. Davis Ng, Reviewing Editor, Temasek Life Sciences Laboratory and National University of Singapore, Singapore

Publication history

  1. Received: April 6, 2017
  2. Accepted: December 15, 2017
  3. Accepted Manuscript published: December 18, 2017 (version 1)
  4. Version of Record published: January 31, 2018 (version 2)

Copyright

© 2017, Bakunts et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 950
    Page views
  • 246
    Downloads
  • 0
    Citations

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

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading