An initial phase of JNK activation inhibits cell death early in the endoplasmic reticulum stress response

Perturbation of protein-folding homeostasis in the endoplasmic reticulum (ER) activates several signal transduction pathways collectively called the unfolded protein response (UPR) (Ron and Walter, 2007; Walter and Ron, 2011). In mammalian cells, the UPR is initiated by several ER-membrane-resident proteins, including the protein kinase-endoribonuclease (RNase) IRE1α (Tirasophon et al., 1998; Wang et al., 1998), the protein kinase PERK (also known as EIF2AK3) (Harding et al., 1999; Shi et al., 1999, 1998) and several type II transmembrane transcription factors such as ATF6α (Yoshida et al., 2000) and CREB-H (also known as CREB3L3) (Zhang et al., 2006). All of these signalling molecules activate prosurvival, but also proapoptotic, responses to ER stress.

These opposing signalling outputs are exemplified by IRE1α. The RNase activity of IRE1α initiates non-spliceosomal splicing of the mRNA for the transcription factor XBP1 (Calfon et al., 2002; Lee et al., 2002; Shen et al., 2001; Yoshida et al., 2001), which in turn induces transcription of genes encoding ER-resident molecular chaperones (Lee et al., 2003), components of the ER-associated protein degradation machinery (Oda et al., 2006; Yoshida et al., 2003) and several phospholipid biosynthetic genes (Lee et al., 2003, 2008) to promote cell survival. The IRE1α RNase activity also initiates the decay of several mRNAs encoding proteins targeted to the ER (Gaddam et al., 2013; Han et al., 2009; Hollien et al., 2009; Hollien and Weissman, 2006), which decreases the protein-folding load of the stressed ER. Degradation of DR5 (also known as TNFRSF10B) mRNA by IRE1α contributes to establishment of a time window for adaptation to ER stress (Lu et al., 2014). By contrast, IRE1α promotes apoptosis through both its RNase and protein kinase domains. Cleavage of several microRNAs (miRNAs), including miRNA-17, -34a, -96 and -125b, by the RNase domain of IRE1α stabilises and promotes translation of TXNIP and caspase-2 mRNAs (Lerner et al., 2012; Oslowski et al., 2012; Upton et al., 2012). TXNIP promotes apoptosis through activation of caspase-1 and secretion of interleukin 1β (Lerner et al., 2012). The role of caspase-2 in ER-stress-induced apoptosis has recently been questioned (Lu et al., 2014; Sandow et al., 2014). The kinase domain of IRE1α activates the JNK mitogen-activated protein kinase (MAPK) family (hereafter referred to as JNK) through formation of a complex with the E3 ubiquitin ligase TRAF2 and the MAPK kinase kinase (MAPKKK) ASK1 (also known as MAP3K5) (Nishitoh et al., 2002; Urano et al., 2000). Sequestration of TRAF2 by IRE1α might also contribute to activation of caspase-12 in murine cells (Yoneda et al., 2001). Pharmacological (Chen et al., 2008; Huang et al., 2014; Jung et al., 2014, 2012; Smith and Deshmukh, 2007; Teodoro et al., 2012; Wang et al., 2009; Zhang et al., 2001) and genetic (Arshad et al., 2013; Kang et al., 2012) studies have provided evidence that activation of JNK at 12 h or later after induction of ER stress is proapoptotic.

Much less is known about the role of JNK at earlier time points in the ER stress response. In tumour necrosis factor α (TNFα, also known as TNF)-treated cells two phases of JNK activation can be distinguished (Lamb et al., 2003; Roulston et al., 1998), an early and transient antiapoptotic phase and a later phase that coincides with activation of caspases (Roulston et al., 1998). In the early phase, JNK induces expression of JunD and the antiapoptotic ubiquitin ligase cIAP2 (also known as BIRC3) (Lamb et al., 2003). Furthermore, phosphorylation of Bad at T201 and subsequent inhibition of interaction of Bad with Bcl-x_L underlies the antiapoptotic role of JNK in interleukin (IL)-3-dependent hematopoietic cells (Yu et al., 2004), whereas JNK mediates IL-2-dependent survival of T cells through phosphorylation of MCL1 (Hirata et al., 2013). This functional dichotomy of transient and persistent JNK signalling prompted us to investigate whether an initial phase of JNK activation exists in the ER stress response and to characterise the functional significance of such an initial phase of JNK activation in ER-stressed cells.

To investigate how early JNK is activated in the ER stress response, we characterised JNK activation over an 8-h timecourse by monitoring phosphorylation of JNK in its T-loop on T183 and Y185 by western blotting with antibodies against phosphorylated and total JNK. In mouse embryonic fibroblasts (MEFs), phosphorylation of JNK in its T-loop increased as early as 10 min after addition of 1 µM thapsigargin (Fig. 1A,C) or 10 µg/ml tunicamycin (Fig. 1D,F). JNK phosphorylation returned to near basal levels by 8 h after addition of thapsigargin or tunicamycin to cells. The ability of these two mechanistically different ER stressors to elicit rapid phosphorylation of JNK, which over several hours declined to near basal levels, suggests that this initial phase of JNK activation is caused by ER stress invoked by these two chemicals and is not a response to secondary effects of these compounds. To compare the kinetics of JNK activation to the kinetics of the Xbp1 splicing reaction and phosphorylation of the PERK substrate eIF2α, we monitored Xbp1 splicing by using reverse transcriptase (RT)-PCR and phosphorylation of eIF2α on S51 by western blotting. Spliced Xbp1 mRNA differs from unspliced Xbp1 mRNA by lacking a 26-nt intron. Hence, the presence of a shorter RT-PCR product on agarose gels is indicative of activation of the IRE1α RNase activity and processing of Xbp1 mRNA. In thapsigargin-treated MEFs ∼45% of Xbp1 mRNA was spliced 20 min after addition of thapsigargin (Fig. 1B,C). Xbp1 splicing reached maximal levels only after several hours of thapsigargin treatment, suggesting that activation of JNK precedes maximal activation of XBP1. Phosphorylation of eIF2α was observed within 10 min after induction of ER stress with 1 µM thapsigargin, which indicates that both eIF2α and JNK are phosphorylated before substantial levels of Xbp1 mRNA are spliced (Fig. 1B,C). When ER stress was induced with 10 µg/ml tunicamycin, phosphorylation of JNK and eIF2α also preceded splicing of Xbp1 (Fig. 1D–F). Furthermore, Xbp1 splicing reached maximal levels only after JNK phosphorylation returned to near basal levels in tunicamycin-treated MEFs. In both thapsigargin- and tunicamycin-treated MEFs phosphorylation of eIF2α declined towards the end of the time course, which is consistent with the transient nature of the translational arrest mediated by eIF2α S51 phosphorylation (Kojima et al., 2003; Novoa et al., 2003).

To investigate whether a similar kinetic relationship between phosphorylation of JNK and eIF2α and Xbp1 splicing exists in other cell types, we repeated these experiments with Hep G2 hepatoma cells, 3T3-F442A adipocytes and C₂C₁₂ myotubes. In Hep G2 cells, JNK phosphorylation increased 30 min after addition of 1 µM thapsigargin to the cells and then declined to near resting levels after ∼120 min of thapsigargin exposure (Fig. S1A,C). By contrast, at 30 min after addition of thapsigargin, only ∼7% of XBP1 mRNA was spliced, and after another 15 min, XBP1 splicing was approximately half maximal (Fig. S1B,C). XBP1 splicing reached maximal levels only after 6 h of thapsigargin treatment. In 3T3-F442A adipocytes, phosphorylation of JNK reached a maximum as early as 10 min after application of 1 µM thapsigargin, then returned to basal levels before increasing again towards the end of the timecourse (Fig. S1D,F). Xbp1 splicing, however, was not detectable until 45 min after the addition of thapsigargin, required 4 h to reach maximal levels and remained at this level for at least another 4 h (Fig. S1E,F). Thus, activation of JNK also precedes activation of XBP1 in Hep G2 cells and 3T3-F442A adipocytes and also returns to near basal levels of activity after several hours of ER stress. We made the same observations in C₂C₁₂ myotubes. In these cells, an increase in JNK phosphorylation was detected as early as 10 min after induction of ER stress with 1 µM thapsigargin (Fig. S1G,H,J), whereas the earliest time point at which an increase in Xbp1 splicing was detected was 20 min (Fig. S1I,J). At the same time, activation of JNK diminished over time in C₂C₁₂ myotubes, whereas the level of Xbp1 splicing remained at maximal levels (Fig. S1H–J). In all three cell lines, phosphorylation of both eIF2α and JNK preceded splicing of XBP1 (Fig. S1). We conclude that activation of JNK preceding induction of XBP1 splicing and leading to an initial phase of JNK activity are phenomena that can be observed in several ER-stressed murine and human cell types.

Several different stresses activate JNK (Kyriakis et al., 1994). To examine whether the rapid JNK activation seen upon thapsigargin or tunicamycin treatment is in response to ER stress and thus mediated via IRE1α and TRAF2, we characterised whether this rapid JNK activation is IRE1α- and TRAF2-dependent. Activation of JNK in the first ∼60 min after induction of ER stress with 1 µM thapsigargin was decreased in Ire1a^−/− and Traf2^−/− MEFs compared to wild-type (WT) MEFs and did no longer reach statistical significance (Figs 1 and 2). In both Ire1a^−/− and Traf2^−/− MEFs JNK activation was delayed and reached maximal levels only towards the end of the 8-h timecourse (Fig. 2). This delayed activation of JNK might be explained by stresses other than and possibly secondary to ER stress, for example oxidative stress (Mauro et al., 2006). Before the onset of the delayed phosphorylation of JNK in Ire1a^−/− and Traf2^−/− MEFs, phosphorylation of JNK was higher in WT MEFs than in the Ire1a^−/− or Traf2^−/− MEFs (Fig. 2G), suggesting that the early JNK activation in ER-stressed cells requires both IRE1α and TRAF2.

To establish whether the initial phase of JNK activation is IRE1α- and TRAF2-dependent in cells other than MEFs, we characterised whether small interfering (si)-RNA-mediated knockdown of IRE1α or TRAF2 reduces the JNK activation caused by ER stress. Two IRE1A siRNAs (#2 and #3, Table S1) reduced IRE1A mRNA levels to ∼40% of control eGFP-siRNA-transfected cells by 72 h post-transfection (Fig. S2A) and decreased activation of JNK to 60±17% and 30±9% of eGFP siRNA-transfected cells, respectively (mean±s.e.m.; n=2) (Fig. S2B,C). Likewise, two siRNAs against human or murine TRAF2 reduced the ER-stress-dependent JNK activation in Hep G2 cells, 3T3-F442A fibroblasts and C₂C₁₂ myoblasts (Figs S2D–F and S3). Furthermore, a dominant-negative mutant of TRAF2, TRAF2Δ1-86 (Hsu et al., 1996; Reinhard et al., 1997), which lacks the RING domain (Fig. S4A) inhibited TNFα-induced JNK activation (Fig. S4B) and reduced the initial phase of JNK activation upon induction of ER stress with 1 µM thapsigargin in 3T3-F442A preadipocytes (Fig. S4C,D) and C₂C₁₂ myoblasts (Fig. S4E,F). Taken together, these data demonstrate that the initial phase of JNK activation upon induction of ER stress is mediated by both IRE1α and TRAF2.

An initial phase of JNK activation by stresses other than ER stress is viewed as being antiapoptotic (Chen et al., 1996a; Lee et al., 1997; Nishina et al., 1997; Raingeaud et al., 1995; Sluss et al., 1994; Traverse et al., 1994). To characterise whether JNK activation early in the ER stress response is also antiapoptotic, we studied whether the mitochondrial permeability transition (MPT) is more pronounced in JNK-deficient MEFs than WT MEFs, because the MPT is often observed in apoptotic cells (Bradham et al., 1998; Fulda et al., 1998; Narita et al., 1998; Scorrano et al., 1999). After exposure of cells to 1 µM thapsigargin or 10 µg/ml tunicamycin for up to 4 h, MPT was revealed by staining cells with the fluorescent dye 5,5′,6,6′-tetrachloro-1,1′,3,3′-tetraethylbenzimidazolylcarbocyanine iodide (JC-1) (Reers et al., 1991; Smiley et al., 1991) (Fig. 3A). The MPT inhibits accumulation of JC-1 in mitochondria and blue-shifts its fluorescence emission from a punctuate orange to a green fluorescence (Reers et al., 1991). After induction of ER stress with 1 µM thapsigargin for 45 min or 4 h MPT was observed in a greater percentage of Jnk1^−/− Jnk2^−/− MEFs than WT MEFs (Fig. 3A,B). Similar results were obtained when ER stress was induced with 10 µg/ml tunicamycin for 4 h (Fig. 3A,C). To provide further evidence for increased apoptotic cell death in JNK-deficient cells, we measured caspase-3/7-like protease activities early in the ER stress response (Fig. 3D,E). Two ER stressors, thapsigargin and tunicamycin, elicited a more pronounced increase of caspase-3/7-like protease activities in Jnk1^−/− Jnk2^−/− MEFs than in WT MEFs 4 h after induction of ER stress (Fig. 3D,E). These data suggest that JNK signalling early in the ER stress response inhibits apoptosis.

In the early antiapoptotic response to TNFα, JNK is required for expression of the mRNA for the antiapoptotic ubiquitin ligase cIAP2, also known as BIRC3 (Lamb et al., 2003). This motivated us to compare the expression of mRNAs for antiapoptotic genes including cIap1 (also known as Birc2), cIap2, Xiap and Birc6 at the onset of activation of JNK with 1 µM thapsigargin in WT and Jnk1^−/− Jnk2^−/− MEFs. Expression of the mRNAs for cIAP1, cIAP2, XIAP and BIRC6 increased in WT cells in the first 45 min of ER stress. By contrast, cIap1, cIap2 and Birc6 mRNA levels decreased in Jnk1^−/− Jnk2^−/− cells (Fig. 4). The increase in Xiap mRNA was more pronounced in WT than in Jnk1^−/− Jnk2^−/− MEFs, suggesting that JNK positively regulates expression of Xiap mRNA (Fig. 4C). To establish whether mammalian inhibitors of apoptosis (IAPs) delay the onset of apoptosis in the early ER stress response we compared caspase-3/7-like protease activity in WT, cIap1^−/− cIap2^−/− MEFs, and Xiap^−/− MEFs. Both cIap1^−/− cIap2^−/− MEFs, and Xiap^−/− MEFs displayed 4.4±1.2-fold higher (mean±s.e.m.; n=4) caspase-3/7-like protease activities than WT MEFs under unstressed conditions (Fig. 5A), which is consistent with increased susceptibility of these cells and cells treated with IAP antagonists to undergo apoptosis (Conte et al., 2006; Geserick et al., 2009; Schimmer et al., 2004; Vince et al., 2007; Yang and Du, 2004). ER stress induced for 4 h with thapsigargin or tunicamycin resulted in a greater increase in caspase-3/7-like protease activities in cIap1^−/− cIap2^−/− and Xiap^−/− MEFs than in WT MEFs (Fig. 5B,C). Taken together, the decreased transcriptional induction of several IAPs in Jnk1^−/− Jnk2^−/− MEFs, and increased MPT and increased caspase 3/7-like protease activities in JNK-deficient MEFs, cIap1^−/− cIap2^−/− and Xiap^−/− MEFs suggest that JNK-dependent transcriptional induction of several IAPs inhibits apoptosis early in the ER stress response.

Here, we show that JNK is activated early in the mammalian UPR and that this immediate JNK activation is antiapoptotic. Activation of JNK early in the UPR by two mechanistically distinct ER stressors, thapsigargin and tunicamycin (Fig. 1; Fig. S1), and its dependence on IRE1α and TRAF2 (Fig. 2; Figs S2–S4) provides evidence that the early JNK activation occurs in response to ER stress. Greater activation of caspase-3/7-like protease activities and a more rapid MPT were observed in ER-stressed JNK-deficient MEFs than in WT MEFs (Fig. 3). These data support the view that early JNK activation protects ER-stressed cells from executing apoptosis prematurely and are consistent with the observation that Traf2^−/− MEFs are more susceptible to ER stress than WT MEFs (Mauro et al., 2006). Early JNK activation coincides with induction of several antiapoptotic genes (Figs 1 and 4). Maximal expression of these mRNAs was dependent on JNK (Fig. 4). MEFs lacking several IAPs, such as cIap1^−/− cIap2^−/− MEFs, and Xiap^−/− MEFs, displayed greater caspase-3/7-like protease activities than WT MEFs during short periods of ER stress (Fig. 5). These observations support the view that IAPs, whose transcriptional induction is dependent on JNK in the early ER stress response, protect cells against apoptosis early in the ER stress response.

Mostly pharmacological data support that activation of JNK late in the ER stress response promotes cell death (Arshad et al., 2013; Chen et al., 2008; Huang et al., 2014; Jung et al., 2014, 2012; Kang et al., 2012; Smith and Deshmukh, 2007; Tan et al., 2006; Teodoro et al., 2012; Wang et al., 2009; Zhang et al., 2001). Our work suggests that two functionally distinct phases of JNK signalling exist in the ER stress response – an early prosurvival phase and a late phase that promotes cell death. Biphasic JNK signalling with opposing effects on cell viability exists also in other stress responses. Transient activation of JNK in response to several other stresses is antiapoptotic (Chen et al., 1996a; Lee et al., 1997; Nishina et al., 1997; Raingeaud et al., 1995; Sluss et al., 1994; Traverse et al., 1994), whereas persistent JNK activation causes cell death (Chen et al., 1996a,b; Guo et al., 1998; Sánchez-Perez et al., 1998). These opposing functional attributes of transient and persistent JNK activation have also been causally established by using JNK-deficient MEFs reconstituted with 1-tert-butyl-3-naphthalen-1-ylmethyl-1H-pyrazolo[3,4-d]pyrimidin-4-ylemine (1NM-PP1)-sensitised alleles of JNK1 and JNK2 (Ventura et al., 2006). Hence, the antiapoptotic function of the initial phase of JNK activation in the ER stress response is another example for the paradigm that the duration of JNK activation controls cell fate. Identification of cIap1, Xiap and Birc6 as genes whose expression requires JNK in the early response to ER stress (Fig. 4) has allowed us to extend the repertoire of antiapoptotic JNK targets. These, and possibly other genes, might also contribute to how JNK inhibits cell death in other stress responses.

The existence of an initial antiapoptotic phase of JNK signalling in the ER stress response raises at least two questions. First, what are the molecular mechanisms that define this initial phase as antiapoptotic? And second, which mechanisms can restrict antiapoptotic JNK signalling to the early response to ER stress? Although future experiments will be necessary to answer these questions, possible explanations might be that the duration of activation affects the subcellular localisation of JNKs, that JNK signalling outputs are controlled by molecular determinants, or that the JNK signalling pathway functionally interacts with other signalling pathways, for example the NF-κB pathway.

Opposing signalling outputs of extracellular signal-regulated kinases (ERKs) in PC12 cells have been explained by different subcellular localisations of ERKs (Marshall, 1995). JNK, however, does not appear to relocalise upon stimulation, either in response to transient or persistent activation (Chen et al., 1996a; Sánchez-Perez et al., 1998). This is also the case for JNK activated early in the ER stress response (Fig. 6). An alternative possibility is that JNK substrates function as molecular determinants of the biological functions of transient and persistent JNK activation, respectively. This is, for example, the case for the ERK substrate c-Fos (Murphy et al., 2002).

In the ER stress response, NF-κB activation is transient and displays kinetics, in several cell lines, that are reminiscent of the initial phase of antiapoptotic JNK signalling reported in this study (Deng et al., 2004; Jiang et al., 2003; Wu et al., 2002, 2004). In TNFα signalling, JNK functionally interacts with the NF-κB pathway. JNK activation in the absence of NF-κB is apoptotic (Deng et al., 2003; Guo et al., 1998; Liu et al., 2004; Tang et al., 2002) or necrotic (Ventura et al., 2004), whereas NF-κB transduces an antiapoptotic response to TNFα (Devin et al., 2000; Kelliher et al., 1998). At the transcriptional level, NF-κB cooperates with JunD (Rahmani et al., 2001), whose phosphorylation is decreased in Jnk1^−/− Jnk2^−/− MEFs (Ventura et al., 2003). NF-κB induces cIap1, cIap2 and Xiap (Stehlik et al., 1998). JunD contributes to the transcriptional induction of cIAP2 in TNFα-stimulated cells (Lamb et al., 2003). This collaboration between NF-κB and transcription factors controlled by JNK, such as JunD, might explain the JNK-dependent induction of cIap1, cIap2, Xiap and Birc6 (Fig. 4), and potentially other antiapoptotic genes, early in the ER stress response.

Transient activation of NF-κB in the ER stress response might also contribute to control of the duration of antiapoptotic JNK signalling. NF-κB inhibits JNK activation by TNFα (De Smaele et al., 2001; Papa et al., 2004; Reuther-Madrid et al., 2002; Tang et al., 2002; Tang et al., 2001) through induction of XIAP (Tang et al., 2002, 2001) and GADD45β (De Smaele et al., 2001; Papa et al., 2004). TNFα also induces the dual-specificity phosphatase MKP1 (also known as DUSP1) (Guo et al., 1998). In murine keratinocytes, cis-platin induces persistent JNK activation but induces MKP1 only weakly, whereas transient JNK activation by trans-platin correlates with strong induction of MKP1 (Sánchez-Perez et al., 1998). Short hairpin RNA (shRNA)-mediated knockdown of MKP1 elevates JNK phosphorylation mediated by tunicamycin in C17.2 neural stem cells, which correlates with increased caspase-3 cleavage and decreased cell viability (Li et al., 2011). These observations suggest that MKP1 is a negative regulator of JNK in ER-stressed cells. However, it remains unresolved whether the effects of the MKP1 knockdown on caspase-3 cleavage and cell viability are causally mediated by JNK or other MKP1 substrates, such as the p38 MAPKs (Boutros et al., 2008). In tunicamycin-treated, but not DTT-treated, cerebellar granule neurons, S359 phosphorylation and stabilisation of MKP1 have been observed, which correlated with short-term JNK activation in tunicamycin-treated cells and prolonged JNK activation in DTT-treated cells (Li et al., 2011). Although these results suggest that MKP1 might control the duration of JNK activation in ER-stressed cells, they might also be the result of different pharmacokinetics or secondary effects of the two ER stressors, especially as JNK is activated by diverse stresses (Kyriakis et al., 1994). For example, DTT chelates heavy metal ions, including Zn²⁺ ions, with pK values of ∼10-15 (Cornell and Crivaro, 1972; Gnonlonfoun et al., 1991; Kręz˙el et al., 2001) and thus might affect many metal-dependent proteins. DTT can also alter proton gradients over membranes (Petrov et al., 1992), because of its pK_a of ∼9.2 (Whitesides et al., 1977), and might reduce lipoamide and through this affect pyruvate dehydrogenase and ATP generation, because its standard redox potential is more negative than the standard redox potential of lipoamide (Cleland, 1964; Massey, 1960). Hence, additional experimentation is required to characterise the role of MKP1 in the ER stress response.

The duration of JNK activation might also be regulated at the level of the ER-stress-perceiving protein kinase IRE1α. Activation of JNK by IRE1α requires interaction of TRAF2 with IRE1α (Urano et al., 2000). This interaction has not been observed in cells expressing kinase and RNase-defective K599A-IRE1α (Urano et al., 2000). JNK activation precedes XBP1 splicing (Fig. 1, Fig. S1). XBP1 splicing by mammalian IRE1α is stimulated by phosphorylation of IRE1α (Prischi et al., 2014). Hence, overall phosphorylation of IRE1α seems to be an unlikely explanation for the transiency of JNK activation. It is, however, possible that the specific pattern of phosphorylation of the approximately ten phosphorylation sites in IRE1α (Itzhak et al., 2014) controls its affinity towards TRAF2 and the activation of JNK by IRE1α.

In conclusion, we show that an initial phase of JNK activation produces antiapoptotic signals early in the ER stress response. Our work also identifies JNK-dependent expression of several antiapoptotic genes, including cIap1, cIap2 and Xiap, as a mechanism through which JNK exerts its antiapoptotic functions early in the ER stress response.

Rabbit anti-phospho-S51-eIF2α (cat. no. 9721S, batches 10-12), rabbit anti-JNK (cat. no. 9252, batch 15), rabbit anti-JNK2 (cat. no. 9258, batch 9), rabbit anti-phospho-JNK (cat. no. 4668, batches 9 and 11) antibodies and human recombinant TNFα (cat. no. 8902) were purchased from Cell Signaling Technology Inc. (Danvers, MA). The mouse anti-GAPDH antibody (cat. no. G8795, batch 092M4820V) was purchased from Sigma-Aldrich (Gillingham, UK), the rabbit anti-eIF2α antibody (cat. no. sc-11386, batch G1309) and the rabbit anti-TRAF2 antibody (cat. no. sc-876, batches G1508 and J2009) from Santa Cruz Biotechnology (Santa Cruz, CA), and the mouse anti-emerin antibody (cat. no. ab49499) from Abcam (Cambridge, UK). siRNAs against TRAF2, IRE1α and eGFP were obtained from Sigma-Aldrich. siRNA sequences are listed in Table S1. Tunicamycin was purchased from Merck Chemicals (Beeston, UK) and thapsigargin from Sigma-Aldrich (Gillingham, UK).

Plasmids were maintained in Escherichia coli XL10-Gold cells (Agilent Technologies, Stockport, UK, cat. no. 200314). Standard protocols for plasmid constructions were used. Plasmid pMT2T-TRAF2Δ1-86 was generated by amplifying a 1327-bp fragment from pMT2T-HA-TRAF2 (Leonardi et al., 2000) with primers H8215 and H8216 (Table S2). The PCR product was cleaved with ClaI and NotI and cloned into ClaI- and NotI-digested pMT2T-HA-TRAF2 to yield pMT2T-TRAF2Δ1-86. The TRAF2 region in pMT2T-TRAF2Δ1-86 was confirmed by sequencing.

WT, Ire1a^−/− (Lee et al., 2002), Jnk1^−/− Jnk2^−/− (Tournier et al., 2000), Traf2^−/− (Yeh et al., 1997), cIap1^−/− cIap2^−/− (Geserick et al., 2009), and Xiap^−/− (Vince et al., 2008) MEFs were provided by Randal J. Kaufman (Sanford Burnham Medical Research Institute, La Jolla, CA), Roger Davis (University of Massachusetts, Worcester, MA), Tak Mak (University of Toronto, Ontario Cancer Institute, Toronto, Ontario, Canada), and John Silke (Walter and Eliza Hall Institute for Medical Research, Victoria, Australia), respectively. 3T3-F442A preadipocytes (Green and Kehinde, 1976), C₂C₁₂ myoblasts (Blau et al., 1985) and Hep G2 cells (Knowles et al., 1980) were obtained from Christopher Hutchison, Rumaisa Bashir and Adam Benham (all Durham University, Durham, UK), respectively. All cell lines were tested for mycoplasma contamination upon receipt in the laboratory with the EZ-PCR mycoplasma test kit from Geneflow (cat. no. K1-0210, Lichfield, UK). Mycoplasma testing was repeated every ∼3 months with all cells in culture at that time. Contaminated cultures were discarded.

All cell lines were grown at 37°C in an atmosphere of 95% (v/v) air, 5% (v/v) CO₂, and 95% humidity. Hep G2 cells were grown in minimal essential medium (MEM) (Eagle, 1959) supplemented with 10% (v/v) foetal bovine serum (FBS) and 2 mM L-glutamine. All other cell lines were grown in Dulbecco's modified Eagle's medium (DMEM) containing 4.5 g/l D-glucose (Morton, 1970; Rutzky and Pumper, 1974), 10% (v/v) FBS, and 2 mM L-glutamine. The medium for Ire1a^−/− and corresponding WT MEFs was supplemented with 110 mg/l pyruvate (Lee et al., 2002). To differentiate C₂C₁₂ cells, 60–70% confluent cultures were shifted into low-mitogen medium consisting of DMEM containing 4.5 g/l D-glucose, 2% (v/v) horse serum, and 2 mM L-glutamine and incubated for another 7–8 days while replacing the low-mitogen medium every 2–3 days (Bains et al., 1984). Differentiation of C₂C₁₂ cells was assessed by microscopic inspection of cultures, staining of myotubes with Rhodamine-labelled phalloidin (Amato et al., 1983) and reverse transcriptase PCR for transcription of the genes encoding S-adenosyl-homocysteine hydrolase (AHCY), myosin light chain 1 (MYL1) and troponin C (TNNC1, Fig. S1G). 3T3-F442A fibroblasts were differentiated into adipocytes as described previously (Mihai and Schröder, 2015). Adipocyte differentiation was assessed by analysing Nile-Red-stained cells by flow cytometry as described previously (Mihai and Schröder, 2015). ER stress was induced with 1 µM thapsigargin or 10 µg/ml tunicamycin, if not stated otherwise.

Hep G2 cells were transfected with plasmids using jetPRIME (Polyplus Transfection, Illkirch, France, cat. no. 114) and with siRNAs using INTERFERin (Polyplus Transfection, cat. no. 409) transfection reagents. Plasmids and siRNAs were transfected into all other cell lines by electroporation with a Neon electroporator (Life Technologies, Paisley, UK) using a 10 µl tip. Manufacturer-optimised electroporation conditions were used for 3T3-F442A preadipocytes and C₂C₁₂ myoblasts. MEFs were electroporated with one pulse of 1200 V and a pulse width of 30 ms. 10–20 nM of each siRNA were transfected. The control siRNA was designed against the enhanced green fluorescent protein (eGFP) from Aequora victoria. Transfection efficiencies were determined by transfection of 2 µg of pmaxGFP (Lonza Cologne GmbH, Cologne, Germany) and detection of GFP-expressing cells with a Zeiss ApoTome fluorescence microscope. Transfection efficiencies were >80%. At 24 h after transfection cells were analysed or time courses initiated, if not stated otherwise.

RNA was extracted with the EZ-RNA total RNA isolation kit (Geneflow, cat. no. K1-0120) and reverse transcribed with oligo-dT primers (Promega, Southampton, cat. no. C1101) and Superscript III reverse transcriptase (Life Technologies, cat. no. 18080044) as described previously (Cox et al., 2011). Protocols for detection of splicing of murine and human XBP1 have been described previously (Cox et al., 2011). In brief, 2.5 µl of the cDNA synthesis reaction were amplified with 1 µM of primers H8289 and H8290 for human XBP1 and primers H7961 and H7962 for murine Xbp1 in a 50-µl reaction containing 1×GoTaq reaction buffer (Promega, cat. no. M7911), 1.5 mM MgCl₂, 200 µM dNTPs, and 0.05 U/ml GoTaq hot start polymerase (Promega, cat. no. M5001). The reaction was incubated for 2 min at 94°C, and then cycled for 35 cycles consisting of subsequent incubations at 94°C for 1 min, 59°C for 1 min and 72°C for 30 s, followed by a final extension step at 72°C for 5 min. Actb was amplified under the same conditions as described for Xbp1 except that GoTaq G2 Flexi DNA polymerase (Promega, cat. no. M7801) was used. Human ACTA1 was amplified with primers H8287 and H8288 and murine Actb with primers H7994 and H7995. Primer sequences are listed in Table S2. Band intensities were quantified using ImageJ (Collins, 2007) and the percentage of XBP1 splicing calculated by dividing the signal for spliced XBP1 mRNA by the sums of the signals for spliced and unspliced XBP1 mRNAs. Quantitative PCRs (qPCRs) were run on a Rotorgene 3000 (Qiagen, Crawley, UK). Amplicons were amplified with 0.5 µl 5 U/µl GoTaq® Flexi DNA polymerase (Promega, cat. no. M8305), 2 mM MgCl₂, 200 µM dNTPs, and 1 µM of each primer and detected with a 1:167,000-fold dilution of a SybrGreen stock solution (Life Technologies, cat. no. S7563) or the GoTaq qPCR Master Mix from Promega (cat. no. A6002). Primers for qPCR are listed in Table S2. qPCR using GoTaq DNA polymerase was performed as follows. After denaturation for 2 min at 95°C, samples underwent 40 cycles of denaturation at 95°C for 30 s, primer annealing at 58°C for 30 s and primer extension at 72°C for 30 s. After denaturation at 95°C for 2 min, qPCRs with the GoTaq qPCR Master mix were cycled 40 times at 95°C for 15 s, 60°C for 15 s, and 72°C for 15 s for cIap1, cIap2, Xiap and Birc6 and 40 times at 95°C for 15 s, 60°C for 60 s for Actb. Fluorescence data were acquired during the annealing step or in case of qPCR amplification of Actb with the GoTaq qPCR Master Mix during the first 30 s at 60°C. Amplification of a single PCR product was confirmed by recording the melting curves after each PCR run. Average amplification efficiencies in the exponential phase were calculated using the comparative quantification analysis in the Rotor Gene Q software and were between 0.6 and 0.7 for all qPCRs. C_T values were calculated and normalised to GAPDH, ACTA1 or Actb mRNA levels as described by Pfaffl (2001) taking the average amplification efficiencies into account. Results represent the mean±s.e.m. of three technical repeats. qPCR results were confirmed by at least one other biological replicate. Murine Ahcy, Myl1 and Tnnc qPCRs were standardised to Gapdh, murine Birc6, cIap1, cIap2, Traf2 and Xiap qPCRs to Actb, the human IRE1A qPCR to GAPDH and the human TRAF2 qPCR to ACTA1.

Cells were washed three times with ice-cold phosphate-buffered saline (PBS, 4.3 mM Na₂HPO₄, 1.47 mM KH₂PO₄, 27 mM KCl, 137 mM NaCl, pH 7.4) and lysed in RIPA buffer [50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.5% (w/v) sodium deoxycholate, 0.1% (v/v) Triton X-100, 0.1% (w/v) SDS] containing Roche complete protease inhibitors (Roche Applied Science, Burgess Hill, UK, cat. no. 11836153001) as described previously (Cox et al., 2011).

For isolation of cytosolic and nuclear fractions, cells were washed two times with ice-cold PBS and gently lysed in 0.32 M sucrose, 10 mM Tris-HCl pH 8.0, 3 mM CaCl₂, 2 mM Mg(OAc)₂, 0.1 mM EDTA, 0.5% (v/v) NP-40, 1 mM DTT, 0.5 mM PMSF. Nuclei were collected by centrifugation for 5 min at 2400 g at 4°C. The supernatant was used as the cytosolic fraction. The nuclear pellet was resuspended in 0.32 M sucrose, 10 mM Tris-HCl pH 8.0, 3 mM CaCl₂, 2 mM Mg(OAc)₂, 0.1 mM EDTA, 1 mM DTT, 0.5 mM PMSF by flipping the microcentrifuge tube. The nuclei were collected by centrifugation for 5 min at 2400 g at 4°C. After aspiration of all of the wash buffer, the nuclei were resuspended in 30 µl low-salt buffer [20 mM HEPES pH 7.9, 1.5 mM MgCl₂, 20 mM KCl, 0.2 mM EDTA, 25% (v/v) glycerol, 0.5 mM DTT, 0.5 mM PMSF] by flipping the microcentrifuge tube. One volume of high-salt buffer [20 mM HEPES pH 7.9, 1.5 mM MgCl₂, 800 mM KCl, 0.2 mM EDTA, 25% glycerol (v/v), 1% NP-40, 0.5 mM DTT, 0.5 mM PMSF] was added dropwise while continuously mixing the contents of the microcentrifuge tube by flipping. The tubes were then incubated for 45 min at 4°C on an end-over-end rotator. The tubes were centrifuged at 14,000 g for 15 min at 4°C and the supernatant transferred into a fresh microcentrifuge tube to obtain the nuclear extract.

Proteins were separated by SDS-PAGE and transferred to polyvinylidene difluoride (PVDF) membranes (Amersham HyBond™-P, pore size 0.45 μm, GE Healthcare, Little Chalfont, UK, cat. no. RPN303F) by semi-dry electrotransfer in 0.1 M Tris, 0.192 M glycine and 5% (v/v) methanol at 2 mA/cm² for 60–75 min. Membranes were blocked for 1 h in 5% (w/v) skimmed milk powder in TBST [20 mM Tris-HCl, pH 7.6, 137 mM NaCl, and 0.1% (v/v) Tween-20] or 5% bovine serum albumin (BSA) in TBST and then incubated overnight with the primary antibody at 4°C and gentle agitation. Blots were washed three times with TBST and then probed with secondary antibody for 1 h at room temperature. The anti-eIF2α, anti-phospho-S51-eIF2α, anti-JNK, anti-JNK2, anti-phospho-JNK, and anti-TRAF2 antibodies were used at a 1:1000 dilution in TBST plus 5% (w/v) BSA. Membranes were then developed with horseradish peroxidase (HRP)-conjugated goat anti-rabbit-IgG (H+L) secondary antibody (Cell Signaling, cat. no. 7074S, batch 24) at a 1:1000 dilution in TBST plus 5% (w/v) skimmed milk powder. The mouse anti-GAPDH antibody was used at a 1:30,000 dilution in TBST plus 5% (w/v) skimmed milk powder and developed with HRP-conjugated goat anti-mouse IgG (H+L) secondary antibody (Thermo Scientific, cat. no. 31432, batch OE17149612) at a 1:20,000 dilution in TBST plus 5% (w/v) skimmed milk powder. For signal detection, Pierce ECL Western Blotting Substrate (cat. no. 32209) or Pierce ECL 2 Western Blotting Substrate (cat. no. 32132) from Thermo Fisher Scientific (Loughborough, UK) was used. Blots were exposed to CL-X Posure™ film (Thermo Fisher Scientific, Loughborough, UK, cat. no. 34091). Exposure times were adjusted on the basis of previous exposures to obtain exposures in the linear range of the film. Films were scanned on a CanoScan LiDE 600F scanner (Canon) and saved as tif files. Bands were quantified using ImageJ exactly as described under the heading ‘Gels Submenu’ on the ImageJ web site (http://rsb.info.nih.gov/ij/docs/menus/analyze.html#plot). In case of unphosphorylated proteins, intensities for the experimental antibody were divided by the intensities obtained with the antibody for the loading control in the same lane to correct for differences in loading between lanes. Intensities for phosphorylated eIF2α were divided by the intensities obtained for total eIF2α in the same lane. For phosphorylated and total JNK, the sums of the intensities at 54 kDa and 46 kDa, which both represent several JNK1 and JNK2 isoforms (Gupta et al., 1996), were used to calculate the fraction of phosphorylated JNK in a similar way as described for phospho-eIF2α. Normalisation of phospho-JNK signals to JNK2 or GAPDH gave qualitatively the same results. All loading control- or unphosphorylated-protein-corrected intensities obtained for one western blot were then expressed relative to the loading-control-corrected intensity of the 0 h sample in the same western blot. To reprobe blots for detection of nonphosphorylated proteins, membranes were stripped using Restore Western Blot Stripping Buffer (Thermo Fisher Scientific, cat. no. 21059) and blocked with 5% (w/v) skimmed milk powder in TBST.

Caspase-3- and -7-like activities were determined with the Caspase-Glo 3/7 kit from Promega (cat. no. G8091). Luminescence was read with a Synergy H4 Multi-Mode Microplate Reader (BioTek, Swindon, UK) and standardised to total protein concentrations determined with the DC protein assay from Bio-Rad Laboratories (Hemel Hempstead, UK, cat. no. 500-0116).

For confocal microscopy, cells were grown in lumox dishes (Sarstedt, Leicester, UK, cat. no. 94.6077.331). After incubation with 1 µM thapsigargin cells were incubated with 2 µg/ml JC-1 (Life Technologies, cat. no. T3168) at 37°C for 20 min (Ankarcrona et al., 1995; Cossarizza et al., 1993; Reers et al., 1991; Smiley et al., 1991). The cells were washed twice with PBS before addition of fresh medium for live-cell imaging on a Leica TCS SP5 II confocal microscope (Leica Microsystems, Mannheim, Germany). JC-1 fluorescence was excited at 488 nm with an argon laser set at 22% of its maximum power. Green fluorescence between 515–545 nm was collected with a photomultiplier tube and orange fluorescence between 590–620 nm with a HyD 5 detector. Cells showing fluorescence emission between 515–545 nm only were counted as having undergone MPT, whereas cells that displayed punctuate fluorescence emission between 590–620 nm were counted as not having undergone MPT.

Samples sizes (n) were derived from experiments with independent cell cultures. Experimental data are presented as the mean±s.e.m. For composite parameters, errors were propagated using the law of error propagation for random, independent errors (Ku, 1966). Statistical calculations were performed in GraphPad Prism 6.07 (GraphPad Software, La Jolla, CA, USA).

We thank A. Benham (Durham University), R. Bashir (Durham University), R. Davis (University of Massachusetts), C. Hutchison (Durham University), R. J. Kaufman (Sanford Burnham Medical Research Institute), T. Mak (University of Toronto), and J. Silke (Walter+Eliza Hall Institute for Medical Research) for providing cell lines. We thank U. Siebenlist (NIAID, NIH) for providing plasmid pMT2T-HA-TRAF2.