Infliction of proteotoxic stresses by impairment of the unfolded protein response or proteasomal inhibition as a therapeutic strategy for mast cell leukemia

The intensity and duration of endoplasmic reticulum (ER) stress converts the unfolded protein response (UPR) from an adaptive into a terminal response. The first regulates homeostasis, the latter triggers apoptosis. Cells that rapidly proliferate and possess developed secretory capabilities, such as leukemia cells, depend on an efficiently operating UPR to maintain proteostasis. Activation of terminal UPR by either blockade of adaptive UPR or exaggeration of ER stress has been explored as a novel approach in cancer therapy. For mast cell leukemia (MCL) the efficacy of both approaches, by utilizing the KITV560G,D816V-positive MCL cell line HMC-1.2, was investigated. We show that HMC-1.2 cells display a tonic activation of the IRE1α arm of the UPR, which constitutively generates spliced XBP1. Inhibition of IRE1α by different types of inhibitors (MKC-8866, STF-083010, and KIRA6) suppressed proliferation at concentrations needed for blockade of IRE1α-mediated XBP1 splicing. At higher concentrations, these inhibitors triggered an apoptotic response. Blocking the proteasome by bortezomib, which confers an exaggerated UPR, resulted in a marked cytotoxic response. Bortezomib treatment also caused activation of the kinase JNK, which played a pro-proliferative and anti-apoptotic role. Hence, the combination of bortezomib with a JNK inhibitor synergized to induce cell death. In summary, the UPR can be addressed as an effective therapeutic target against KITD816V-positive MCL.


INTRODUCTION
For physiological performance, cells depend on a functional proteostasis network enabling proper protein expression, folding, transport and topology, as well as the capability to stringently control and, when necessary, to correct these processes upon stressful perturbations [1]. The endoplasmic reticulum (ER) is the port of entry of proteins into the secretory pathway. Particularly in cells with a pronounced proliferative and/or secretory phenotype, such as cancer cells, extensive protein and membrane production as well as protein folding tasks are required. Hence, cancer cells exhibit vulnerability to the accumulation of misfolded proteins in the ER. An elaborate system comprising sensors, transcription factors and chaperones, called the unfolded protein response (UPR), has been evolutionary established to maintain proteostasis [2,3]. Three transmembrane sensor proteins located in the ER membrane form the backbone of the UPR -inositol-requiring enzyme 1α (IRE1α), doublestranded RNA-activated protein kinase-like ER kinase (PERK), and activating transcription factor 6 (ATF6) -each containing a luminal domain that allows their Research Paper activation by recognition of misfolded proteins [2,3].
IRE1α contains a serine/threonine kinase and an endoribonuclease domain in its cytoplasmic domain. After activation, the latter splices the mRNA of the transcription factor XBP1, causing a shift in the reading frame and allowing for translation of the "spliced" transcription factor XBP1s, which enters the nucleus and controls the production of a large number of ER chaperones as well as proteins involved in ER-associated protein degradation (ERAD), amongst others [2,4]. Moreover, in a process called "regulated IRE1-dependent decay of mRNA (RIDD)" IRE1α can cleave additional ER-localized mRNAs [5]. Activated PERK phosphorylates and thereby inactivates the translation initiation factor eIF2α, resulting in a global attenuation of protein synthesis. Paradoxically, phosphorylated eIF2α promotes translation of several proteins, a prominent one being the pro-apoptotic transcription factor ATF4 [2,6]. The third sensor, ATF6, upon ER stress, de-oligomerizes and translocates from the ER to the Golgi apparatus, where it is cleaved by membrane-bound proteases in its cytoplasmic domain. Cleavage results in the release of the N-terminal domain, which subsequently enters the nucleus and acts as a transcription factor [2,7].
The context-dependent cooperation of these sensors and the respective transcription factors allow for cellular responses ranging from adaptation to apoptosis [2]. This apoptotic switch has been utilized already for the treatment of the plasma cell malignancy called multiple myeloma (MM), the cells of which exert marked secretory capacity and constitutively depend on the UPR [8,9]. To induce UPR-directed apoptosis, the proteasome inhibitor bortezomib (BZ), which has been established as a standard of care for MM patients, was applied to impede degradation of misfolded proteins by the proteasome in the ERAD pathway, thus enhancing the protein load of the ER and tipping the balance of the stress response towards apoptosis. In addition to proteasome inhibitors, the glycosylation inhibitor tunicamycin (TM), the inhibitor of the sarcoplasmic/endoplasmic reticulum Ca 2+dependent ATPase (SERCA), thapsigargin, as well as the reducing agent dithiothreitol (DTT) are frequently used to artificially induce the UPR [10].
Systemic mastocytosis (SM) is a clonal mast cell (MC) disease exhibiting various characteristics from indolent mastocytosis to aggressive mastocytosis and MC leukemia (MCL) [11,12]. SM features abnormal proliferation and accumulation of MCs in various tissues and organs [11,12]. The receptor tyrosine kinase KIT plays a central role in MC biology by promoting development, survival, proliferation, chemotaxis as well as proinflammatory mediator production and release [12,13]. Mutations in KIT can be found in most cases of SM, with KIT D816V being the most prevalent and detectable in more than 80 % of patients [11,12]. Since KIT D816V may be found in all variants of SM (from indolent SM to MCL) with comparable prevalence [14], mutations in additional genes might be present causing the differences in disease severity. Indeed, mutations in genes encoding for various signaling proteins, epigenetic regulators, transcription factors, and splicing factors were identified [15][16][17]. Moreover, a thorough molecular analysis of myeloid progenitor cells in multi-mutated advanced SM revealed KIT D816V as a late event [18]. Nevertheless, it is understood that KIT D816V is important in SM regarding MC proliferation and survival. Inhibition of KIT D816V by established tyrosine kinase inhibitors (TKIs), however, faces several problems. KIT D816V is resistant to the first generation TKI imatinib and high concentrations of the second generation TKI nilotinib are required for its inhibition [19,20]. Two further second generation TKIs, dasatinib and midostaurin, and the third generation TKI ponatinib can suppress KIT D816V -driven cellular activities in vitro [21][22][23][24], however, the determination of drugprotein interaction profiles as well as phosphoproteome analyses revealed restricted selectivity, offering the possibility of unwanted side effects [25][26][27][28]. Nevertheless, recent studies revealed effectiveness of nilotinib and midostaurin in a number of patients with advanced systemic mastocytosis, including highly fatal MCL [29,30]. However, further kinases except KIT, such as the SRC family kinase LYN, the TEC family kinase BTK, and the mitosis-regulating serine/threonine kinase PLK1, have been demonstrated to be involved in the regulation of proliferation and survival of MCL cell lines as well as patient cells [31,32], which might account for patient-and situation-specific restricted efficacy of the above mentioned TKIs. Hence, further TKI-independent therapies or the use of synergistically acting drug combinations should be developed.
In this study, we have approached the importance of the UPR in MCL and analyzed the efficacy of various UPR inhibitors and pharmacological inducers of ER stress to suppress proliferation and survival of the KIT V560G,D816Vpositive human MCL cell line HMC-1.2. In addition, we unraveled the potency of a combination of BZ and the JNK inhibitor JNK-IN-8 to efficiently induce apoptosis in KIT D816V -positive MCL cells.

Inhibition of the IRE1α arm of the UPR suppresses proliferation and survival of HMC1.2 cells
In a situation-dependent manner, the UPR can result in an adaptive, pro-homeostatic or in a terminal, proapoptotic cellular response. Cells that rapidly proliferate and possess developed secretory functions are particularly dependent on a functional adaptive UPR to cope with the synthetic demand of the ER. Thus, we interrogated the KIT V560G,D816V -positive human MCL cell line HMC-1.2 for a constitutively active UPR by determining activation of the UPR sensor IRE1α. Occurrence of spliced XBP1 mRNA (XBP1s) was taken as the read-out for IRE1α activation. First, we applied an XBP1 splicing detection of spliced XBP1 mRNA was evaluated by an XBP1 splicing detection assay. HMC-1.2 cells were pre-incubated with vehicle (DMSO) or 30µM MKC-8866 (MKC) for 1h followed by 6h treatment with10µg/ml TM. Generated cDNA was used to amplify XBP1 by PCR. XBP1u amplicon digestion by PstI resulted in two fragments (u1, u2). Size of XBP1s amplicon lacking PstI sites was 473bp. B. HMC-1.2 cells were treated for 1h with vehicle (DMSO) or 30µM MKC followed by 6h treatment with10µg/ml TM. (n=3) Data show mean ± SD from n ≥ 3 independent experiments. Student's t-Test and one-sample t-Test were performed to calculate the p-values. *p < 0.05, **p < 0.01, ***p < 0.001. assay involving mRNA amplification by RT-PCR followed by diagnostic restriction digest. As a positive control, cells were treated with TM for 6 h. As expected, TM induced a strong splicing of XBP1 mRNA, which was suppressed by the IRE1α inhibitor MKC-8866, which targets the endonuclease domain of IRE1α ( Figure  1A). In single experiments, a faint band of XBP1s was already detectable in proliferating HMC-1.2 cells (data not shown), suggesting a weak basal activity of IRE1α in HMC-1.2 cells. The data obtained with the XBP1 splicing detection assay were corroborated using XBP1s-specific RT-qPCR ( Figure 1B). Owing to the enhanced sensitivity of the RT-qPCR approach, significant reduction of basal XBP1s mRNA by MKC-8866 was measurable, indicating once more the basal activation of IRE1α in proliferating HMC-1.2 cells. Noteworthy, constitutive activation of an UPR is not a feature of every cell type. Assuming that IRE1α activity is needed to promote growth of HMC-1.2 MCL cells, we next investigated if blocking IRE1α activity might confer inhibition of proliferation. HMC-1.2 cells were treated with increasing concentrations of MKC-8866 (10 -60 µM) or vehicle control and cell numbers were determined every 24h using an analytical cell counter. Indeed, inhibition of IRE1α resulted in significant suppression of HMC-1.2 proliferation after 72h treatment ( Figure 1C & 1D). To verify these data and to combine them with information on metabolic activity, XTT assays were performed. Incubation (72h) with MKC-8866 caused a dose-dependent decline in metabolic activity ( Figure  1E). Compared to the sole determination of cell numbers ( Figure 1D), a marked diminution of XTT positivity was evident from 30 µM to 60 µM of MKC-8866, suggesting appearance of an additional quality in the presence of 60 µM MKC-8866 ( Figure 1E). Therefore, we analyzed induction of cell death in MKC-8866-treated HMC-1. To corroborate the effect of MKC-8866-mediated IRE1α inhibition, two additional pharmacological inhibitors of IRE1α were tested: the IRE1α endonucleasespecific inhibitor STF-083010 [33] and the ATPcompetitive IRE1α kinase-inhibiting RNase attenuator KIRA6 [34]. Both inhibitors significantly and dosedependently suppressed proliferation determined by cell counting ( . This was intriguing all the more since an effective blockade of XBP1 splicing was already obtained using the lower concentrations of both inhibitors (STF-083010: 30 µM; KIRA6: 1 µM) (Supplementary Figure  2). Particularly KIRA6 was a potent inducer of cell death, potentially owing to its dual activity inhibiting both the endonuclease activity as well as the kinase activity of IRE1α. In conclusion, proliferation and survival of the human MCL cell line HMC-1.2 rely on a functional UPR, in particular signaling from the UPR sensor IRE1α, and indicate the potential of IRE1α inhibitors as pharmacological treatment for patients suffering from MCL.

Constitutive activation of PERK in HMC-1.2 cells does not support proliferation and survival
Motivated by the finding of basal UPR activity in HMC-1.2 cells, we investigated activation of the second UPR sensor, PERK, in proliferating HMC-1.2 cells. Phosphorylation of the translation initiation factor eIF2α was taken as a read-out for PERK activity. Phosphospecific immunoblot analysis of eIF2α in lysates of proliferating, vehicle (DMSO)-treated HMC-1.2 cells showed basal phosphorylation of eIF2α suggesting constitutive activation of the UPR ( Figure 3A). Enforced activation of the UPR by treatment with the glycosylation inhibitor TM resulted in enhanced eIF2α phosphorylation, indicating that the output of PERK was not maxed in proliferating HMC-1.2 cells ( Figure 3A). Using RT-qPCR we found that CHOP mRNA was markedly enhanced after 24h in TM-treated HMC-1.2 cells, which was significantly reduced by the PERK inhibitor GSK2606414 ( Figure  3B). Treatment with 0.1 µM GSK2606414 resulted in a near to maximum suppression of TM induced CHOP mRNA production ( Figure 3B) as well as corresponding eIF2α phosphorylation (Supplementary Figure 3). Having identified the effective concentrations of GSK2606414, we next investigated if blocking PERK activity might inhibit proliferation of HMC-1.2 cells. Hence, cells were treated with increasing concentrations of GSK2606414 (0.01 -1 µM) or vehicle and cell numbers were determined every 24h using an analytical cell counter. Inhibition of PERK had no significant effect on proliferation of HMC-1.2 cells after 72h treatment ( Figure 3C & 3D). This was corroborated by XTT assays measuring metabolic activity ( Figure 3E) as well as by analysis of apoptosis/cell death in GSK2606414-treated cells by staining with annexin V and propidium iodide (AV/PI) ( Figure 3F). In conclusion, the PERK pathway of the UPR appears slightly active in proliferating HMC-1.2 cells, however, an involvement in proliferation and survival, in comparison to the IRE1α/ XBP1s pathway, was not observed. www.impactjournals.com/oncotarget

Proteasome inhibition suppresses proliferation and survival of human MCL cells
Our data so far demonstrated that survival of HMC-1.2 MCL cells is supported by the constitutive activation of the UPR, especially the IRE1α arm. An important mechanism of cells to cope with rising levels of unfolded proteins is ERAD that depends on a functional ubiquitin proteasome system. Proteasome inhibitors, particularly the FDA-approved BZ (Velcade, originally codenamed PS-341), are used in MM cells and patients to block the ERAD resulting in enhanced accumulation of (misfolded) proteins in the ER, and thus to induce terminal UPR [35,36]. On the other side, various cancer cell lines have been shown to be resistant to BZ [37]. Thus, the possibility to reduce HMC-1.2 proliferation and survival by BZ treatment was thoroughly investigated. Measurement of total protein ubiquitination by immunoblotting of whole-cell lysates with ubiquitin-specific antibodies as readout for proteasome inhibition, proved 10 nM BZ to be sufficient for an optimal response ( Figure 4A). Moreover, 10 nM BZ triggered significant activation of the IRE1α path of the UPR as measured by RT-qPCR of XBP1s ( Figure 4B). Accordingly, BZ-induced XBP1s generation was suppressed by STF-083010 and KIRA6 ( Figure 4B). Next, proliferation of HMC-1.2 cells in the absence or presence of BZ was analyzed. Treatment with 5-30 nM BZ for 24-72h was found to completely block proliferation ( Figure 4C & 4D). This was in agreement with the suppression of metabolic activity measured by XTT assay, which suggested induction of cell death by BZ treatment ( Figure 4E). Hence, by determining AV/PI doublepositivity, BZ-triggered cell death was analyzed. Indeed, BZ was a potent inducer of HMC-1.2 cell death with 10 and 30 nM causing approximately 40% and 80% doublepositive cells, respectively ( Figure 4F; Supplementary Figure 1D). Thus, our data suggest that comparable to MM [35], BZ-mediated proteasome inhibition might be usable as a therapeutic option in KIT D816V -positive MCL.

Inhibition of BZ-induced JNK activity synergistically promotes cell death of HMC-1.2 MCL cells
The stress kinase JNK has been shown to be activated in response to proteasome inhibition and to initiate an apoptotic program [38]. On the other hand, JNK has been found in MCs to promote proliferation [39,40]. Moreover, JNK has been demonstrated in cancer cells to be able to mediate both pro-survival and pro-apoptotic signaling [41]. Providing that BZ treatment of HMC-1.2 cells results in activation of JNK, which might exert a proliferation-and/or survival-promoting role, combined use of proteasome and JNK inhibitors might be used as effective treatment of MCL. To address this question, HMC-1.2 cells were treated with 10 nM BZ for 24h and phosphorylation of JNK was measured by phosphospecific immunoblotting. Indeed, BZ induced pronounced phosphorylation of JNK, which was manifested in JNK activation since the characteristic JNK target, JUN, was phosphorylated as well ( Figure 5A). Inhibition of JNK by the selective pharmacological JNK inhibitor, JNK-IN-8 [42], did not block BZ-induced JNK phosphorylation, but completely suppressed phosphorylation of JUN ( Figure  5A). The markedly retarded electromobility of p-JNK from BZ-vs. BZ/JNK-IN-8-treated cells can be explained by covalent binding of JNK-IN-8 to a conserved Cys residue within JNK [42]. Moreover, BZ treatment caused marked induction of JUN mRNA, which was inhibited by JNK-IN-8 ( Figure 5B). The observed equivalence between JNK phosphorylation and activation was not self-evident, because TM treatment of HMC-1.2 cells also induced JNK phosphorylation, however, did not cause phosphorylation of JUN ( Figure 5A). This difference coincided with an altered running behavior of p-JNK in TM-vs. BZ-treated HMC-1.

IRE1α mediates BZ-induced JUN and CHOP transcription
Via a complex with the adaptor protein TRAF2, IRE1α was demonstrated to activate JNK [43]. Moreover, in acute myeloid leukemia JUN was recently found to be involved in the transcription of the UPR transcription factors XBP1 and ATF4 [44]. As shown before, in comparison to TM, BZ induced strong production of JUN mRNA and pretreatment with the selective JNK inhibitor JNK-IN-8 suppressed BZ-induced JUN transcription ( Figure 6A). Interestingly, the IRE1α kinase/endonuclease inhibitor KIRA6 significantly reduced JUN production, whereas STF-083010, being an endonuclease inhibitor www.impactjournals.com/oncotarget    Figure 2A. Data shown are mean ± SD from n ≥ 3 independent experiments. Student's t-Test and one-sample t-Test were performed to calculate the p-values. *p < 0.05, **p < 0.01, ***p < 0.001. www.impactjournals.com/oncotarget only, was unable to attenuate JUN transcription ( Figure  6A). This suggested that IRE1α, via its kinase activity, is directly involved in the activation of the JNK/JUN pathway. Next, we analyzed if JNK is involved in the transcription as well as splicing of XBP1. BZ treatment of HMC-1.2 cells caused the production of XBP1s mRNA, which was significantly reduced by pre-incubation with JNK-IN-8 ( Figure 6B). These RT-qPCR data were corroborated employing the XBP1 mRNA splicing assay ( Figure 6C). This assay, in addition to XBP1s, allows the evaluation of changes in XBP1u. However, no obvious differences in the amounts of XBP1u were observed between BZ-treated cells pre-incubated with vehicle (DMSO) or JNK-IN-8 ( Figure 6C), indicating involvement of JNK in splicing rather than production of XBP1 in BZtreated HMC-1.2 cells. This was verified measuring the amount of XBP1u mRNA using RT-qPCR ( Figure 6D). Finally, the role of IRE1α and JNK with respect to BZinduced ATF4 and CHOP transcription was investigated.
Interestingly, compared to TM treatment, BZ did not cause a significant induction of ATF4 mRNA; this was not influenced by pharmacological blockade of JNK or IRE1α ( Figure 6E). This is in agreement with data by Nawrocki et al., who demonstrated PERK inhibition by BZ treatment in pancreatic cancer cells [37]. Unexpectedly and in contrast to ATF4, BZ did induce transcription of CHOP, though less pronounced than TM. This induction was weakly, but significantly attenuated by KIRA6, however not by STF-083010 and JNK-IN-8, suggesting a JNK-independent function of the activated IRE1α kinase domain ( Figure  6F). So far, we cannot exclude participation of ATF6 in BZ-triggered CHOP transcription [45,46]. In conclusion, JNK is mainly involved in BZ-induced XBP1s production and hence promotion of proliferation and survival, causing enhanced vulnerability for combined BZ and JNK-IN-8 treatment.

DISCUSSION
For cancer therapy, both blockade of adaptive UPR as well as exaggeration of ER stress resulting in terminal UPR might be employed. Our data concentrating on MCL and utilizing the KIT V560G,D816V -positive MCL cell line HMC-1.2 as a model showed for the first time the efficacy of both approaches with respect to induction of cytostatic and/or cytotoxic responses. Inhibition of the IRE1α arm of the UPR by different types of inhibitors (MKC-8866, STF-083010, and KIRA6) suppressed proliferation at concentrations needed for blockade of IRE1α-mediated XBP1 splicing. At higher concentrations, these inhibitors triggered apoptosis by terminal UPR. Enhanced activation of the UPR by proteasome inhibition, which blocks the ERAD pathway and promotes the excessive accumulation of proteins in the ER, caused a marked cytotoxic response. Additional inhibition of the compensating activation of the kinase JNK allowed for synergistic induction of cell death at comparatively low inhibitor concentrations. Hence, our data suggest that the UPR can be addressed as an attractive and effective therapeutic target against KIT D816V -positive MCL. Comparable low (2-10 nM), pharmacologically achievable concentrations of BZ were already shown to suppress growth of MM cells as well as to sensitize them to conventional chemotherapeutic agents [35,47].
Cells of the plasma cell neoplasm, MM, are dependent on a functioning IRE1α/XBP1 arm of the UPR. Inhibition of the RNase activity of IRE1α by the small-molecule inhibitor STF-083010 was demonstrated to have cytotoxic activity against human MM cells in a xenograft model as well as ex vivo [33]. Moreover, the covalent IRE1α inhibitor 4µ8C attenuated growth of MM cell lines, however, was not associated with acute toxicity [48]. Enhancing ER stress by blocking proteasome activity using BZ (PS-341) was shown to be cytostatic and cytotoxic to human MM cell lines as well as freshly isolated patient MM cells [35]. BZ treatment lead to XBP1 splicing and the IRE1α inhibitor MKC-3946, which, by its sole use, modestly inhibited growth of MM cells, could significantly enhance BZ-induced cell death [49]. Motivated by these reports, we analyzed the importance of the UPR for growth and survival of human MCL cells using the KIT V560G,D816V -positive MCL cell line HMC-1.2. Employing three different IRE1α inhibitors, cytostatic as well as cytotoxic activities could be observed dependent on the inhibitor concentrations applied. These inhibitors were MKC-8866 and STF-083010 blocking the RNase activity of IRE1α [33] as well as KIRA6, an ATP-competitive IRE1α kinase inhibitor allosterically coinhibiting IRE1α´s RNase activity by breaking respective oligomers [34].
Inhibitor concentrations that maximally blocked XBP1 splicing caused strong suppression of HMC-1.2 proliferation, however, only weak cytotoxic effects were observable. On the other hand, concentrations higher than needed for the full blockade of XBP1 splicing were able to induce significant apoptosis. The reason for this qualitative difference, the switch from a cytostatic to a cytotoxic response, has not been clarified yet. One possibility could be the potential need for higher concentrations of IRE1α inhibitors for the suppression of RIDD. While this cannot be excluded for the two IRE1α RNase inhibitors, MKC-8866 and STF-083010, it appears rather unlikely for KIRA6 in the light of work from Papa and coworkers [34]. In their experiments, the IC50 of KIRA6 for RIDD inhibition was lower than the IC50 for inhibition of XBP1 splicing. Another possibility might be cytotoxic side effects of these inhibitors at higher concentrations. KIRA6 as an ATP-competitive inhibitor might additionally block other kinases possibly involved in pro-survival signaling. Indeed, our data indicate that KIRA6, either directly or indirectly, is able to suppress phosphorylation/activation of JNK. Moreover, all three inhibitors, at higher concentrations, might interact with completely different targets causing pro-apoptotic signals. The principle variability of inhibitors has previously been shown convincingly by Rix et al, who found by chemical proteomic profiling that the oxidoreductase NQO2 is inhibited at physiologically relevant concentrations by the otherwise rather selective tyrosine kinase inhibitors, imatinib and nilotinib [27]. A further example of IRE1α-independent KIRA6 activity might underlie the small, though significant pro-proliferative effect of low concentrations of this IRE1α inhibitor ( Figure 2B). At this point, the involved target of KIRA6 is not known. Ghosh et al. tested a panel of seven Ser-/Thr-kinases for potential cross-reactivity of KIRA6 [34]. Among these kinases only IRE1α was inhibited in in vitro kinase assays. However, the human kinome is considerably bigger and it cannot be excluded that one or more of the many non-tested kinases i) can be inhibited by KIRA6 and ii) has anti-proliferative effects. At higher concentrations of KIRA6, the measured pro-proliferative effect would then be outcompeted by the pro-apoptotic effect of complete IRE1α inhibition and suppression of XBP1s expression.
After demonstrating the suitability of inhibition of an adaptive UPR with respect to suppression of MCL cell proliferation and induction of cell death, we next sought to address the possibility of switching an adaptive to a terminal UPR. To reach this, we applied the 26S proteasome inhibitor BZ [50], and found marked triggering of an apoptotic response at reasonable low inhibitor concentrations, which were also found active in the treatment of MM cells [35]. Interestingly, proteasome inhibition has been demonstrated to induce phosphorylation/activation of the MAPK JNK [51,52], which promoted the cytotoxic response ignited by blockade of the proteasome. In MCs, however, JNK has been found to promote proliferation [39,40]. In addition, pro-survival SOX4-dependent upregulation of JNK1 (MAPK8) was demonstrated as a critical factor in acute lymphoblastic leukemia [53]. We hypothesized that BZinduced JNK activation might be compensatory for the MCL cells and additional inhibition of JNK could thus result in enhanced, synergistic cell death of MCL cells. Indeed, combined use of BZ and the selective JNK inhibitor JNK-IN-8 was most effective in causing cell death of HMC-1.2 cells, indicating that in the case of MCL, proteasome inhibition causes pro-proliferative/ anti-apoptotic JNK activation. Since active JNK was able to induce JUN transcription as well as JUN production in HMC-1.2 cells, a pro-survival role of JUN might be envisaged. Interesting in this respect is a recent publication by Zhou et al., who showed a pro-survival role of JUN in acute myeloid leukemia, multiple subtypes of which frequently overexpress JUN [44]. JUN was demonstrated to bind to promoters of several UPR effectors, such as XBP1 and ATF4. However, in HMC-1.2 cells ATF4 upregulation upon BZ treatment was not observed; moreover, inhibition of JNK did not affect expression of XBP1u, suggesting cell type-specific regulatory mechanisms in the UPR.
Interestingly, our data in HMC-1.2 cells hint at a mutual functional dependence of IRE1α and JNK. While the kinase activity of IRE1α promotes activation of JNK, JNK appears to be able to facilitate the endonuclease activity of IRE1α (Figure 7). JNK has been reported to specifically phosphorylate serine and threonine residues, which are followed by a proline residue. In several JNK targets, such as in JUN and JUND, proline is followed by an acidic amino acid [54]. Intriguingly and in line with the proposed model (Figure 7), T973 in the endonuclease domain of IRE1α has been reported to be phosphorylated [55] and resides in the context of the sequence T-P-D and thus would fulfill both location in the endonuclease domain and presence of a JNK target-specific sequence. It will be important to explore the phosphorylation of T973 by JNK and further analyze the IRE1α (kinase) -JNK -IRE1α (endonuclease) circuit in further leukemia cells.
Similar to the IRE1α arm of the UPR, the PERK arm, depending on the strength of ER stress, is able to induce both an adaptive and a terminal response. The respective adaptive response involves phosphorylation of eIF2α preventing the formation of the ternary initiation complex consisting of unphosphorylated eIF2α, GTP, and tRNA Met , thus inhibiting general protein synthesis. The terminal response involves selective translation of the transcription factor ATF4 resulting, amongst others, in the induction of the proapoptotic transcription factor CHOP. ATF4 has also been shown recently to induce autophagy [56]. In case of increased production of mutated or misfolded proteins in the ER, as it may occur in neurodegenerative diseases, blockade of translation by eIF2α phosphorylation appears appropriate; however, such response would not be advantageous for proliferating leukemia cells. Although we detected basal phosphorylation of eIF2α in HMC-1.2 cells, pharmacological inhibition of PERK by GSK2606414 did neither have cytostatic nor cytotoxic effects. Speaking against basal PERK activation, GSK2606414 suppressed increased eIF2α phosphorylation in response to TM treatment, however, it did not reduce basal phosphorylation of eIF2α, suggesting involvement of other kinases, like PKR or mGCN2. Moreover, as with every measurable phosphorylation event, which is the net result of kinase and phosphatase activities, eIF2α-specific phosphatases might be active in MCL cells to allow for protein production as well as to prevent ATF4/CHOPmediated apoptosis. Further analysis of these phosphatases and their regulation in HMC-1.2 cells might enable the identification of novel drug targets in MCL.
In conclusion, we have demonstrated the importance of the UPR for the proliferation and survival of MCL cells, in particular the IRE1α-dependent arm of the UPR. We have shown that both pharmacological inhibition of IRE1α catalytic activities and exaggeration of the UPR by pharmacological stressors suppress proliferation and survival of MCL cells. Moreover, prosurvival JNK activity induced by BZ-triggered IRE1α activation is introduced as an additional drug target to boost the efficacy of BZinduced MCL cell death. In conclusion, our data suggest that the UPR can be addressed as an effective therapeutic target against so far unsatisfactorily treatable KIT D816Vpositive MCL.

Apoptosis assay
Cells were seeded at a density of 3.5 x 10 5 cells/ml and treated with the indicated substances for 72h. After treatment cells were incubated with Annexin V-Alexa Fluor 647 (Alexis Biochemicals) in culture medium for 20 min at RT in the dark. Immediately before analysis by flow cytometry, propidium iodide (1 μg/ml was added and analyzed on a FACScan (BD Biosciences).

Proliferation assays
Cells were seeded at a density of 3.5 x 10 5 cells/ ml or 5.5 x 10 5 cells/ml ( Figure 1C/1D) and treated with different concentrations of the test substances; solvent (DMSO)-treated cells served as controls. After a 24h-treatment, cells were resuspended completely and 50 μl from each well was diluted in 10 ml PBS for automated multi-parameter cell counting using a Casy cell counter (Innovatis). Metabolic activity was measured using the XTT Cell Proliferation Kit II (XTT)(Roche). Cells were seeded in microplates at a density of 3.5 x 10 5 cells/ ml (suspension culture grade, 96 wells, flat bottom) in a final volume of 100 µl culture medium per well in a humidified atmosphere (37°C, 5% CO 2 ) for 72h. After the incubation period, 50 µl of the XTT labeling mixture was added to each well (final XTT concentration 0.3 mg/ml). Incubation of the microplate was for 3 -4 h in a humidified atmosphere (e.g., 37°C, 5% CO 2 ).
Spectrophotometrical absorbance of the samples were measured using a microplate reader. The wavelength used to measure absorbance of the formazan product of the XTT assay was 475nm and the reference wavelength was 650nm. Sample values at 475nm were subtracted with medium controls (blanked) resulting in delta blanked values. Total absorbance was calculated by subtraction of delta blanked values (475nm) with their reference values at 650nm. These absorbance values (A 475nm -A 650nm ) are shown in the respective figures.

XBP1 mRNA splicing
RNA was isolated and reverse transcribed as above to obtain total cDNA. Then, XBP1 primers were used to amplify an XBP1 amplicon spanning the 26 nt intron from the cDNA samples in a regular 3-step PCR. Thermal cycles were: 5 min at 95 °C, 30 cycles of 30 s at 95 °C, 30 s at 60 °C, and 1 min at 72 °C, followed by 72 °C for 15 min, and a 4 °C hold. PCR was performed using h-XBP1.3 fwd: AAA CAG AGT AGC AGC TCA GAC TGC and h-XBP1.12 rev: TCC TTC TGG GTA GAC CTC TGG GAG primers. PCR fragments were then digested by PstI and separated on Ethidium Bromide containing 2% agarose gels. The restriction digest of unspliced XBP1 (XBP1u) resulted in two fragments of 290 and 183bp. The size of the XBP1s amplicon lacking PstI sites was 473bp.

Statistical analysis
Data were generated from independent experiments. P-values were calculated by the unpaired two-tailed Student's t-test or the one-sample t-Test using GraphPad Prism. Figures represent means ± SD of n ≥ 3 independent experiments (with n indicated in the respective figure legends). P-values of * < 0.05, ** < 0.01, and *** < 0.001 were considered statistically significant. Values higher than a p-value of 0.05 were regarded as not significant (ns).

ACKNOWLEDGMENTS
K. Maschke-Neuß is acknowledged for excellent technical assistance.

CONFLICTS OF INTEREST
The authors declare that they have no conflicts of interest.