Decision letter: Inadequate BiP availability defines endoplasmic reticulum stress

Article Figures and data Abstract Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract How endoplasmic reticulum (ER) stress leads to cytotoxicity is ill-defined. Previously we showed that HeLa cells readjust homeostasis upon proteostatically driven ER stress, triggered by inducible bulk expression of secretory immunoglobulin M heavy chain (μs) thanks to the unfolded protein response (UPR; Bakunts et al., 2017). Here we show that conditions that prevent that an excess of the ER resident chaperone (and UPR target gene) BiP over µs is restored lead to µs-driven proteotoxicity, i.e. abrogation of HRD1-mediated ER-associated degradation (ERAD), or of the UPR, in particular the ATF6α branch. Such conditions are tolerated instead upon removal of the BiP-sequestering first constant domain (CH1) from µs. Thus, our data define proteostatic ER stress to be a specific consequence of inadequate BiP availability, which both the UPR and ERAD redeem. https://doi.org/10.7554/eLife.41168.001 Introduction It is well-established that accumulation of unfolded proteins in the endoplasmic reticulum (ER)—a condition referred to as ER stress—activates the unfolded protein response (UPR), which, in turn, mitigates the stress, most notably through enhancing the ER chaperone content to boost the protein folding capacity (Walter and Ron, 2011). What defines ER stress, and how ER stress may engender cytotoxicity, however, are poorly understood issues. Moreover, it is still debated what feature of ER stress activates the UPR. An important reason why these are still open questions is the wide-spread use of ER stress-eliciting drugs, such as tunicamycin (Tm), which inhibits N-glycosylation, or thapsigargin (Tg), which causes Ca2+ efflux from the ER (Walter and Ron, 2011). These drugs have pleiotropic effects and are inherently cytotoxic, hence obscuring important aspects of how ER homeostasis can be restored by virtue of the UPR or not. To overcome the shortcomings of ER stress-eliciting drugs, we recently have developed a HeLa cell-based model for proteostatically driven ER stress (Bakunts et al., 2017). Inducible overexpression of the IgM subunits µs and the λ light chain, in stoichiometric amounts, leads to bulk secretion of IgM with little if any UPR activation. In the absence of λ, however, µs is retained in the ER, and maximally activates the three main UPR branches, governed by IRE1α, PERK, and respectively, ATF6α. Yet, the cells successfully adapt to the proteostatic insult by expanding the ER both in size and in chaperone content, such that cell viability and growth are unaffected in the process, and UPR signaling subsides to a submaximal amplitude once homeostasis is restored (Bakunts et al., 2017). The ER resident chaperone BiP stands out in the course of the adaptation to µs expression in two ways. First, ER stress sensing and UPR signaling occur in a µs/BiP ratiometric fashion, that is the amplitude of UPR signaling is maximal when µs levels eclipse those of BiP, which is sequestered through binding to µs, while UPR signaling subsides to submaximal output when an excess of BiP over µs is restored (Bakunts et al., 2017). ER homeostatic readjustment is due to the UPR, since BiP is a key UPR target gene (Walter and Ron, 2011). Second, ER homeostatic readjustment to µs expression causes a ∼10-fold increase of BiP levels overall, which entails that BiP shifts from about one tenth to about one third of the total protein mass in the ER, such that BiP is the only chaperone in the ER of which the levels outmatch those of µs (Bakunts et al., 2017). The two main models that have been proposed for UPR activation are that it entails i) dissociation of BiP from the lumenal domains of the main ER stress sensors, IRE1α, PERK (Bertolotti et al., 2000) and ATF6α (Shen et al., 2002), and ii) direct binding of unfolded proteins (Gardner and Walter, 2011; Karagöz et al., 2017), including the Ig heavy chain CH1 domain (Karagöz et al., 2017), to these sensors. Based on insights obtained from µs-driven ER stress, we argue that these two UPR activation models are not mutually exclusive. Rather, the two models are complementary and should be unified, since in a three-way competition between UPR sensors, BiP, and an ER client protein (µs) for binding one another, the ratio of UPR sensors bound to the client versus those bound to BiP most robustly report on the client/BiP ratio, to which indeed the UPR signaling amplitude correlates (Bakunts et al., 2017). HeLa cells tolerate genetic ablation of the main three UPR transducers, but expression of µs in the context of UPR-ablated cells causes synthetic lethality through apoptosis, underscoring the key role the UPR has in restoring ER homeostasis (Bakunts et al., 2017). In this study we exploited this synthetic lethality to define how ER stress becomes proteotoxic. Results IRE1α and PERK are expendable, but ATF6α is key for µs-provoked ER homeostatic readjustment To investigate in detail how the UPR sustains ER homeostatic readjustment to bulk µs expression, we exploited cells in which IRE1α was deleted and PERK and ATF6α were silenced with good efficiency (Bakunts et al., 2017), either individually or in combinations. Surprisingly, ablation of IRE1α and PERK (either individually or in combination) had negligible effects on viability and growth of µs-expressing cells, (Figure 1A,B), or on ATF6α activation (Figure 1C). Thus, IRE1α and PERK are dispensable for restoring ER homeostasis upon bulk µs expression, and ER stress levels are not enhanced in their absence (although there is some ATF6α activation already under basal conditions when IRE1α and PERK are ablated; Figure 1C). Conversely, silencing of ATF6α alone caused reduced growth and/or viability of µs-expressing cells (Figure 1A,B), implying that ER homeostasis was not (fully) restored. Figure 1 Download asset Open asset ATF6α is essential but IRE1α and PERK are dispensable for restoring ER homeostasis upon µs expression. (A–D) In HeLa-µs cells, IRE1α was deleted (KO), and ATF6α and PERK were silenced (KD) either alone or in combination, or not (-), as indicated. (A) Cells were seeded upon 1:5 serial dilution into 24-well plates, and treated with 0.5 nM mifepristone (Mif) to induce expression of µs where indicated (+). After 7 days of growth, cells were fixed and stained with crystal violet. (B) Staining in (A) was quantitated as a measure for cell growth. Mean and s.e.m. are shown in a bar graph; n = 2. (C) Expression of µs was induced for 0 or 3 days. Immunoblotting of lysates from cells that were sufficiently viable upon the insult for analysis revealed levels of µs, BiP, CHOP, α-tubulin, and ATF6α processing (i.e. release of the p50 cleavage product from the p90 precursor); cross-reaction of the secondary antibody against anti-ATF6α with µs is denoted (µs). RT-PCR fragments corresponding to spliced (XBPS) and unspliced (XBPU) were separated on gel. A hybrid product that is formed during the PCR reaction is denoted by an asterisk. (D) BiP levels in (C) were quantitated and expressed as fold change upon µs expression compared to untreated cells. Mean and s.e.m. are shown in a bar graph; n=2-5. Statistical significance of differences in growth (B), or in expression levels (D), was tested by ANOVA (*p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001). https://doi.org/10.7554/eLife.41168.002 Figure 1—source data 1 https://doi.org/10.7554/eLife.41168.003 Download elife-41168-fig1-data1-v1.xlsx When µs is expressed for 3 days in wild-type cells, ER homeostasis is restored, and, consequently, IRE1α and PERK signaling subsides to submaximal output (Bakunts et al., 2017). In ATF6α-silenced cells, conversely, ER homeostasis is not restored, and, accordingly, signaling through the PERK and IRE1α pathways remained persistently high (Figure 1C); that is levels of CHOP, a key downstream effector of PERK (Harding et al., 2000), were increased, and IRE1α-mediated XBP1 mRNA splicing (Calfon et al., 2002) was enhanced, as was evident from the increased prominence of the higher mobility band, corresponding to the RT-PCR product of the XBP1S transcript from which the intron has been removed (Calfon et al., 2002). Ablation of ATF6α in combination with ablation of IRE1α and/or PERK caused apoptosis (Bakunts et al., 2017) and, consequently, abrogated viability of µs-expressing cells (Figure 1A,B). We concluded that accumulation of µs in the ER per se confers proteotoxicity when the UPR is dysfunctional, and that the UPR counteracts this proteotoxicity, in particular through the ATF6α branch. IRE1α and PERK are expendable, but ATF6α is key for ER expansion in response to µs expression Despite the persistently maximal signaling through the PERK and IRE1α pathways upon µs expression in ATF6α-silenced cells (Figure 1C,D), upregulation of BiP was compromised (Figure 1C,D; Figure 2C,E), while upregulation of two other ER chaperones, PDI, and GRP94 was abolished (Figure 2—figure supplement 1), which confirms that also these ER chaperones are prominent ATF6α targets (Bommiasamy et al., 2009). ATF6α silencing did not affect accumulation of µs (Figure 2C, D), however, and the ER did not expand (Figure 2A, B), in accordance with the compromised upregulation of ER chaperones. Conversely, ER expansion (Figure 2A, B), and BiP upregulation (Figure 1C, D) upon µs expression was not compromised in PERK– and/or IRE1α–ablated cells. Thus, the ATF6α branch of the UPR is the main if not sole driver of ER expansion in response to µs expression. Figure 2 with 1 supplement see all Download asset Open asset ATF6α is essential but IRE1α and PERK are dispensable for upregulation of ER chaperones and ER expansion in response to µs expression. (A,B) HeLa-µs cells in which UPR transducers were ablated by silencing alone or in combination, or not (WT), as indicated, were induced with 0.5 nM Mif to express µs for 3 days or not. The cells harbor APEX-KDEL, a modified version of pea peroxidase that is targeted to the ER, and that catalyzes polymerization of 3,3'-diaminobenzidine tetrahydrochloride (DAB) upon treatment with H2O2 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 (A). The extent of ER expansion was assessed as described (Bakunts et al., 2017), and the percentage of the area within the cytoplasm corresponding to ER was determined and depicted in bar graphs (B). Mean and s.e.m. are shown, n = 10–20. (C–E) Cells were induced to express µs for the indicated times. Levels of µs (D) and BiP (E) were quantitated from (C), and replicate experiments. (D) Levels in WT of µs at 64 hr were set at 100 that was scaled to levels of BiP in WT at 64 hr such as to reflect a ratio of µs to BiP of 2:3, that is an estimate for this ratio at day three based on earlier quantitations that we have described (Bakunts et al., 2017). Mean and s.e.m. are shown in bar graphs; n = 2–5. Statistical significance in the extent of ER areas in the electron micrographs between µs-expressing or non-expressing cells (black), or between µs-expressing WT or ATF6α ablated cells (red) (B), or in expression levels (D,E) was tested by ANOVA (n.s., not significant; *p≤0.05; **p≤0.01; ***p≤0.001). https://doi.org/10.7554/eLife.41168.004 Figure 2—source data 1 https://doi.org/10.7554/eLife.41168.005 Download elife-41168-fig2-data1-v1.xlsx ER stress and ensuing cytotoxicity levels correlate with the extent of µs being chaperoned Since the UPR induces expression of ER resident chaperones, we surmised that µs-driven ER stress becomes cytotoxic when the UPR is compromised, in particular upon ATF6α ablation, due to 'under-chaperoning' of µs. Proteins that undergo folding tend to aggregate in absence of sufficient folding assistance. Upon ablation of IRE1α and ATF6α, µs indeed formed extensively disulfide-linked high molecular weight species that partitioned into a NP40-insoluble fraction, indicative of aggregation (Mattioli et al., 2006; Valetti et al., 1991)—with the single ablations showing intermediate phenotypes—(Figure 3A). Figure 3 Download asset Open asset ER stress correlates with the extent of ER chaperones being engaged and becomes cytotoxic when their capacity is exceeded. (A) HeLa-µs cells, in which IRE1α (KO) and/or ATF6α (KD) was ablated, or not (-), as indicated, were induced with 0.5 nM Mif to express µs for 24 hr. Samples were lysed in NP40 and equivalent amounts of soluble (S) and insoluble (I) fractions resolved under reducing (red) or non-reducing conditions, blotted and decorated with anti-µs. (B) HeLa-µs cells were induced with 0.5 nM Mif to express µs for the indicated times and treated with or without 100 μg/ml CHX for 3 hr before harvesting. Samples were analyzed by iso-electric focusing (IEF) to separate AMPylated (BiPAMP) from non-AMPylated BiP, which were detected by immunoblotting, as described (Preissler et al., 2015). To allow a better comparison between samples, considering the upregulation of BiP upon µs expression, approximately 15 µg of lysates were loaded for the 0 day samples, while only 2.5 µg were loaded for the other days. (C) HeLa-µs-derived cells, harboring Dox-inducible hamster BiP (HeLa-µs/BiPH), were treated for 2 days with 50 nM Dox to induce hamster BiP expression, while WT HeLa-µs cells were mock-treated with 50 nM Dox, before both cell lines were induced with 0.5 nM Mif to express µs for the indicated times. Immunoblotting of lysates revealed levels of µs, total BiP, hamster BiP, and GRP94. XBP1 mRNA splicing was assessed as in Figure 1C. https://doi.org/10.7554/eLife.41168.007 Under basal conditions, a significant proportion of BiP readily converts into an inactive, AMPylated state upon a three-hour block of protein synthesis with cycloheximide (CHX) (Figure 3B), which indicates that BiP gets to be dismissed from its chaperoning duties once its regular clients have had sufficient time to complete their folding, as has been reported before (Preissler et al., 2015). Conversely, in µs-expressing cells no AMPylation occurred upon CHX treatment at any time upon the onset of µs expression (Figure 3B), suggesting that the vast majority of BiP is permanently engaged in chaperoning µs even though the BiP pool is expanding massively in response to µs expression (Bakunts et al., 2017). As BiP stands out as a key chaperone for orphan µs, we reasoned that the level of BiP at basal conditions is a key determinant for µs-driven ER stress susceptibility. To test this idea, we created a derivative of the HeLa-µs cell line with an integrated copy of the hamster HSPA5 gene that encodes BiP under control of doxycycline (Dox). The induction of µs with Mif leads to it being the most abundantly transcribed gene (Bakunts et al., 2017) in the cells and concomitant induction of other transgenes would lead to competition for the transcription and/or translation machineries (not shown), thereby mitigating µs expression and, hence, µs-driven ER stress by default. We therefore decided to pre-emptively enhance BiP levels with Dox at least ~10 fold prior to induction of µs expression (Figure 3C). Even though exogenously driven BiP transcription ceased after that, exogenous (hamster) BiP levels remained high for a prolonged time (Figure 3C). In line with the notion that ER stress sensing in the HeLa-µs model occurs in a µs/BiP ratiometric fashion (Bakunts et al., 2017), and in line with earlier reports that BiP overexpression dampens UPR activation (Bertolotti et al., 2000), XBP1 mRNA splicing and upregulation of the UPR target GRP94 occurred with a delay when BiP levels were exogenously boosted as compared to when BiP was at endogenous levels, in spite of the similar extent and kinetics of µs accumulation (Figure 3C). Altogether the HeLa-µs model thus provides further support that sensing of ER stress correlates with the extent of the folding machinery being engaged in chaperoning its clients, and that BiP sequestration by client proteins appears to serve as the main proxy for that. Turnover of µs as afforded by ERAD is remarkably robust While µs levels increase, and the ER expands (~3–4 fold compared to basal levels), as wild-type cells are still adapting to the proteostatic insult, there is no further build-up of µs levels and ER expansion after ~2–3 days once homeostasis is restored (Bakunts et al., 2017), which implies that at that stage the influx of µs molecules into the ER must be matched by countermeasures. Translational attenuation through PERK activation can alleviate the burden on the ER folding machinery by diminishing the input of nascent clients entering the ER lumen (Harding et al., 1999). Yet, we ruled out that PERK-driven translational attenuation was a key determinant for ER homeostatic readjustment in the HeLa-µs model, considering that PERK ablation hardly impeded cell growth upon µs expression (Figure 1A,B). Accordingly, there was only a marginal reduction in overall protein synthesis (being at the lowest ~80% of that before induction) that was moreover transient (i.e. only manifest during the first 16 hr of µs-expression) (Figure 4—figure supplement 1). Following the same reasoning, we also ruled out that regulated IRE1α-dependent decay (RIDD) (Hollien and Weissman, 2006; Hollien et al., 2009) of mRNAs that encode ER client proteins (and thereby limiting their influx into the ER) is important for homeostatic readjustment upon µs expression, since ablation of IRE1α had negligible impact on cell growth (Figure 1A,B). However, µs is a target of ERAD, as has been shown in plasma cells (Fagioli and Sitia, 2001), and which is shown here for the HeLa-µs cell model, since the proteasomal inhibitor MG132 to a large extent stabilizes µs levels in pulse-chase assays (Figure 4A,B). Since µs is glycosylated, it is subject to mannose trimming (Aebi et al., 2010), which is a key step in delivering µs to the retro-translocation machinery that shuttles it to the cytosol for proteasomal degradation (Fagioli and Sitia, 2001). Accordingly, the ER mannosidase I inhibitor kifunensine (Kif) stabilized µs in a similar manner as MG132 (Figure 4A,B). Figure 4 with 2 supplements see all Download asset Open asset ERAD accounts for disposal of µs in a robust manner. HeLa-µs cells, in which ATF6α was ablated (D,E), as indicated, or not (A-C) were pulse labeled for 10 min and chased with excess unlabeled cysteine and methionine for the indicated times after 24 hr (A,B,D,E) or at various times (C), as indicated, after induction of µs expression with 0.5 nM Mif, in the absence (A,C,D) or presence (A) of 10 MG132 or as indicated (+). were quantitated and the after 0 hr was set at and s.e.m. are shown in bar graphs of the quantitations of (C) were to the of µs at various time after induction of its see on the of Figure data 1 Download while µs levels in the ER with ERAD kinetics hardly (i.e. the of µs was remarkably which implies that ERAD with the of µs (Figure ERAD are UPR target (Walter and Ron, and indeed various ERAD and which we to by (Bakunts et al., 2017), were upon µs expression (Figure 4—figure supplement Yet, their upregulation at most to to ERAD kinetics of the µs In ERAD kinetics of µs were not by ablation of ATF6α (Figure in line with the that µs accumulation was not upon ATF6α ablation (Figure Thus, ER homeostatic upon ATF6α ablation is not due to compromised the of ERAD is robust in HeLa-µs cells, since we can also out that IRE1α and/or PERK are essential for sufficient ERAD as their ablation hardly caused any growth of µs-expressing cells (Figure 1A,B), when ERAD is of µs through ERAD is key for homeostatic readjustment While prolonged proteasomal in is cytotoxic, ERAD of with per se did not affect cell viability (Figure and did not the UPR either (Figure supplement 1). We reasoned that ERAD would be however, to bulk accumulation of µs in viability was compromised in µs-expressing cells (Figure ERAD are the and its et al., which have been shown to ERAD of µs et al., ablation of and, to a of was in HeLa-µs cells upon µs expression (Figure Figure with 2 supplements see all Download asset Open asset ERAD of µs is through the (A) as in Figure of HeLa-µs cells, in which was deleted (KO), was silenced or not Cells were treated with 0.5 nM mifepristone (Mif) to induce expression of µs or not (-), and WT cells were treated with or not as indicated. of µs from or (C) HeLa-µs cells that were induced with 0.5 nM Mif to express µs for 4 hr and treated for the indicated times with 100 μg/ml CHX either alone in combination with 100 nM or not (B), or 10 MG132 (C), as indicated. The indicates the of µs. (D) HeLa-µs WT or cells were induced with Mif for 24 hr to express µs or not as indicated. Samples were lysed in and over a Levels of µs, and were detected by that in cells expression of µs becomes due to the of expression levels, however, µs not high molecular weight indicative of the of the chaperoning and to target ERAD the ER to the where are by and, by the et al., Accordingly, µs was stabilized in or cells, as upon treatment of WT cells, while in WT cells µs was upon CHX treatment (Figure with MG132 stabilized µs in WT cells, and the of a of µs that, at least of of µs was to the and to in or cells no of µs that disposal of µs was at prior the step (Figure to be other that at the ER to et al., 2011; et al., of these can for the of treatment with the inhibitor did not lead to any of µs (Figure Thus, HRD1-mediated ERAD is the main if not disposal that is essential for ER homeostatic readjustment in the HeLa-µs model, even though has been reported to IgM and ER expansion in plasma cells et al., The synthetic lethality that once ERAD is compromised in the HeLa-µs model a to define which are to in with and in the disposal of µs. To that we ablated by without that is without In this we that cell viability upon was compromised, and that µs was stabilized in a CHX by ablation of and to significant (Figure supplement However, ablation of or of only µs-expressing cells. These two indeed have been shown to be as soluble ERAD upon mannose trimming of their before over these to et al., et al., In and fractions upon µs expression, while the ERAD did not (Figure These that disposal of µs is through of at the and at their These as its ablation abrogated their (Figure upon ERAD with µs levels BiP upregulation When ERAD is an excess of BiP over µs is restored upon 3 days of µs expression. The is as from a combination of and (Bakunts et al., 2017). As as BiP levels are in excess UPR signaling subsides to submaximal and ER homeostatic readjustment to µs expression is (Bakunts et al., 2017). The of viability in µs-expressing cells (Figure indicated that ER homeostatic readjustment and these cells indeed apoptosis (Figure in µs-expressing cells that ER stress was and accordingly, IRE1α and PERK at maximal levels (Figure Figure Download asset Open asset of µs disposal through ERAD leads to BiP being permanently ER homeostatic and HeLa-µs cells, harboring (D,E) or not (A) were induced with (µs) or without 0.5 nM Mif for 3 days in the presence or absence of (A) of cells were assessed by Mean and s.e.m. are shown in a bar n = HeLa-µs cells were induced to express µs for various times as indicated (B) or for 3 days in the absence or presence of (C) Levels of µs, BiP, and as as activation of the IRE1α and PERK of the UPR were assessed as in (Bakunts et al., 2017). (B) Levels of BiP and µs were assessed by as described (Bakunts et al., 2017), and depicted in bar graphs as in Figure such that the µs levels in the absence of were scaled to BiP levels at a ratio of Levels in the presence of are expressed as a fold change compared to levels in the absence of and s.e.m. are n = 2. (D) In cells harboring the extent of ER expansion was assessed as in Figure Boxed areas are shown by 3-fold magnification; scale bars represent 1 The percentage of the area within the cytoplasm corresponding to ER was determined and depicted in bar graphs and s.e.m. are shown, n = Statistical significance of differences in or the extent of ER area in the electron micrographs (E) were tested by ANOVA ***p≤0.001). Figure data 1 Download maximal UPR activation upon ERAD in µs-expressing cells that induction of BiP expression was persistently at maximal the build-up of BiP levels fold further increase after 3 not with the accumulation of µs fold further increase after 3 upon ERAD such that µs levels in the ER that were at about a with BiP (Figure aggregation of µs increased when ERAD was as by µs fractions in in WT cells (Figure Thus, under those conditions the chaperoning machinery becomes as upon ablation of IRE1α and/or ATF6α in cells (Figure 3A). We the of the ER under basal conditions to be of the and upon 3 days of µs expression to be of the corresponding to a fold increase of ER (Bakunts et al., 2017). Upon ERAD with the ER did not expand in cells. In µs-expressing cells, ERAD caused the area of ER within the cytoplasm