The BCL-2 inhibitor ABT-199/venetoclax synergizes with proteasome inhibition via transactivation of the MCL-1 antagonist NOXA

Enhanced expression of anti-apoptotic B-cell lymphoma 2 (BCL-2) protein is frequent in cancer. Targeting of BCL-2 with the specific inhibitor ABT-199 (Venetoclax) has significant clinical activity in malignant diseases such as chronic lymphocytic leukemia and multiple myeloma. The small molecule drug ABT-199 mimics the pro-apoptotic BCL-2 homology domain 3 of BH3-only proteins and blocks the hydrophobic BC-groove in BCL-2. We have previously shown that ABT-199 synergizes with the proteasome inhibitor (PI) bortezomib in soft tissue sarcoma derived cells and cell lines to induce apoptosis. Synergistic apoptosis induction relies on the pore-forming effector BAX and expression of the pro-apoptotic BH3-only protein NOXA. Bortezomib augments expression of NOXA by blocking its proteasomal degradation. Interestingly, shown here for the first time, expression of NOXA is strongly enhanced by ABT-199 induced integrated stress response (ISR). ISR transcription factors ATF3 & ATF4 mediate transactivation of the BH3-only protein NOXA which specifically inhibits the anti-apoptotic MCL-1. Thus, NOXA potentiates the efficacy of the BCL-2 inhibitor ABT-199 by simultaneous inhibition of MCL-1. Hence, ABT-199 has a double impact by directly blocking anti-apoptotic BCL-2 and inhibiting MCL-1 via transactivated NOXA. By preventing degradation of NOXA PIs synergize with ABT-199. Synergism of ABT-199 and PIs therefore occurs on several, previously unexpected levels. This finding should prompt clinical evaluation of combinatorial regimens in further malignancies.


INTRODUCTION
Mitochondria are the lead actors of the intrinsic apoptosis pathway and they are staged by pro-and anti-apoptotic members of the BCL-2 protein family. BCL-2 kinship is rooted to the presence of at least one of the four distinct BCL-2 homology (BH1-4) domains. The exclusive presence of the BH3-domain distinguishes the pro-apoptotic "BH3-only" proteins (BAD, BID, BIM, NOXA, PUMA) from their pro-apoptotic multidomain protein (MDP) siblings BAX, BAK, and BOK, that share BH1-4 [1]. BAX, BAK, and BOK are effector proteins that upon activation undergo conformational changes and oligomerize in the mitochondrial outer membrane (MOM). Upon oligomerization the effectors mediate MOM permeabilization (MOMP) thus releasing cytochrome c resulting in activation of caspases and cellular demise. Yet, a third clan of the BCL-2 family, the anti-apoptotic "BCL-2-like" proteins (BCL-2, BCL-x L , BCL-w, MCL-1, A1), antagonizes both proapoptotic clans, the BH3-only proteins and MDPs BAX and BAK, by accommodating their BH3-domain in a hydrophobic groove. Hence, BH3-only proteins neutralize the capacity of anti-apoptotic proteins to antagonize the effectors BAX and BAK and promote cell death [2][3][4]. The amino acid composition of the hydrophobic groove and the BH3-domain sequence determine the specific interaction, i.e. the antagonizing effect of BCL-2-like proteins on certain pro-apoptotic clan members, generating two signaling axes: the BCL-2-BAX and the MCL-1-BAK axis.
Another class of anti-cancer drugs are PIs. Improvement of the prototypical PI bortezomib (BTZ) [22] has yielded effective drugs for the therapy of MM and mantle cell lymphoma in particular [23]. PIs in clinical use, such as ixazomib (IXZ), carfilzomib (CFZ) and marizomib (MRZ), differ in pharmacokinetic and specificity towards the active subunits (β 1 , β 2 , β 5 ) in the proteasome. Each PI targets the chymotrypsin-like activity of β 5 . However, BTZ and IXZ also inhibit the caspase-like activity of β 1 while at increased concentration CFZ also blocks the trypsin-like activity of the β 2 subunit [24]. Proteasome inhibition induces accumulation of nonfunctional and misfolded proteins and activates cellular stress response pathways: i) the ER-lumen prompts the unfolded protein response (UPR), and ii) the cytosol the heat shock response (HSR), while iii) integrated stress response (ISR) reacts to both ER-lumen and cytosol. The ISR funnels diverse stress signals and reduces global protein synthesis while allowing translation of specific mRNAs or open reading frames (ORFs) from alternative start codons. The former mechanism is verifiable by enhanced eIF2 phosphorylation and the latter by induction of ATF4 expression [25].
We recently found that BTZ synergizes with the BH3-mimetic ABT-199 to induce apoptosis in various soft tissue sarcoma cells and cell lines. The synergistic activity of ABT-199&BTZ depends on BAX and the MCL-1 antagonist NOXA [26]. In continuation of this work, we show here that ABT-199 transactivates PMAIP1/NOXA proposedly via stress-induced ATF4 with concomitant induction of ATF3 and thereby synergizes with clinically relevant PIs to overcome resistance and induce cell death.
Cell death induction by ABT199&PIs in SW982/BAK KO is similar to SW982/WT (Fig. 1A, B) implicating that BAK is not rate limiting for synergistic apoptosis induction by ABT-199&PI (in the presence of BAX). ABT-199 and BTZ, CFZ, or IXZ induced apoptosis is strongly reduced in BAX deficient cells, indicating that it largely depends on BAX. To investigate the molecular basis for synergistic apoptosis induction, we investigated expression of the antiapoptotic ABT-199 target protein BCL-2 and the short lived antiapoptotic MCL-1, that is degraded by the proteasome [27]. We also analyzed expression of BAX and BAK and the BH3-only protein NOXA, because NOXA is also degraded by the proteasome. We did not detect drug induced variations in the expression of BCL-2, BAK or BAX (Fig. 1D). Expectedly, PIs tended to induce enhanced expression of NOXA and enhanced expression of MCL-1 (Fig. 1D). Intriguingly, NOXA expression is increased by ABT-199, and increased NOXA expression is even more pronounced in combination with PIs. Results were confirmed in TP53 deficient H1299/WT and H1299/TP53 lung cancer cells (Suppl. Figure 2H). In conclusion, expression of NOXA correlates with efficacy of cell death induction in the order ABT-199 < ABT-199+IXZ < ABT-199 +BTZ < ABT-199+CFZ. Therefore, compound-specific differences between PIs exist.
Blocking of β 5 and β 2 activity correlates with synergistic cell death induction by  The PIs BTZ and IXZ block the β 1 and β 5 subunits of the 26 S proteasome, while CFZ blocks β 2 and β 5 subunits [24]. To analyze whether protease-specificity of PIs causes the differences in synergistic cell death induction we detected activity of proteasome subunits after incubation with BTZ, CFZ or IXZ. Chymotrypsin-like β 5 -activity is efficiently blocked by BTZ, CFZ, and IXZ ( Fig. 2A). CFZ is the most effective β 2 -inhibitor ( Fig. 2B) and weakest β 1 -inhibitor (Fig. 2C). Apparently, the combined inhibition of β 5 and β 2 plays a role in synergistic cell death induction while β 1 has minor impact.

Knock-down of NOXA impairs synergism of ABT-199&PIs
To examine the role of NOXA we knocked down PMAIP1/NOXA  Figure 1A-G). PMAIP1/NOXA downregulation reduced cell death by~24% in case of ABT-199&BTZ (Fig. 2D) and 30% in case of ABT-199&CFZ (Fig. 2E) whereas ABT-199&IXZ induced cell death was blocked (<20%; Fig.  2F). Thus, knock-down of NOXA unambiguously impairs apoptosis induction by ABT-199&PIs. However, residual cell death indicates existênce of alternative pathways induced by ABT-199&BTZ or ABT-199&CFZ. We thought to elucidate the mechanisms underlying induction of NOXA. Because induction of PMAIP1/NOXA was associated with enhanced expression of MCL-1 whereas in siPMAIP1/NOXA transfected cells basal expression of MCL-1 was unchanged ( Fig. 2G-I), we excluded the possibility that enhanced MCL-1 expression caused increased NOXA expression. To analyze whether NOXA induction is the consequence of BAX/BAK induced MOMP and apoptosis we incubated HCT116 cells in the presence or absence of ABT-199 and/or BTZ for 24 h. Expectedly, knock-out of BAX and/or BAK reduces synergistic cell death induction (Suppl. Figure 1H) while NOXA expression is invariably induced by ABT-199 (in the presence of Q-VD-OPh) (Suppl. Figure 1I). Thus, induction of NOXA by ABT-199 is independent/upstream of MOMP and apoptosis.
Negligible TP53-dependent induction of NOXA by PIs PMAIP1 is a TP53-inducible gene [18,28] and NOXA and TP53 are post-transcriptionally regulated by proteasomal degradation [20]. Thus, we investigated whether proteasome inhibition augments expression of NOXA. SW982/WT cells were incubated with PIs and extracts were analyzed for expression of NOXA. However, PIs only very slightly augmented expression of NOXA (Figs. 1D, 3A). Because proteasome inhibition also stabilizes TP53 ( [29] Fig. 3A), we investigated whether PI-stabilized TP53 induces NOXA. We transfected SW982/WT with siRNA targeting TP53 (siTP53) and analyzed transcription of PMAIP1 by qRT-PCR upon inhibition of the proteasome. BTZ, CFZ and IXZ had no significant impact on expression of PMAIP1 (Fig. 3B). Knock-down of TP53 reduced PMAIP1 expression although not statistically significant (Fig. 3B). Western blot analysis showed slightly enhanced NOXA expression (BTZ, CFZ) and knock-down of TP53 reduced NOXA expression (Fig. 3C). Thus PI mediated stabilization of TP53 has negligible impact on PMAIP1/NOXA induction. Notably, ABT-199 and ABT-199&BTZ caused comparably stronger expression of NOXA that was hardly attenuated by siTP53. To further delineate induced To test whether ABT-199 mediates enhanced transcription we analyzed PMAIP1 mRNA by real-time PCR. Astonishingly, ABT-199 induced 10-fold expression of PMAIP1/NOXA that was even further enhanced (>25-fold) in response to ABT-199&BTZ. This enhanced transcription of PMAIP1/NOXA was blocked when transcription was blocked by actinomycin D (Act-D) (Fig. 3E, bar graph). Transcriptional regulation was further investigated by analyzing the impact of ABT-199 on expression of transgenic NOXA in   1 ABT-199&PIs synergize to induce cell death with concomitant induction of NOXA expression. A-C SW982/WT and corresponding knock-out cell lines SW982/BAK KO and SW982/BAX KO were cultured for 24 h in presence or absence of 15 µM ABT-199 and/or 5 nM PI (BTZ, CFZ or IXZ). Apoptotic cell death was assessed by flow cytometric analysis of phosphatidyl serine exposure (Annexin V-APC) and relative mitochondrial membrane potential (TMRM). Reduced apoptosis in SW982/BAX KO suggests BAX as relevant mediator of synergism, whereas BAK does not appear to play a major role in early (24 h) apoptosis induction. Graphs show mean and individual data points. Statistical significance was analyzed by an unpaired student´s t-test. D SW982/WT, SW982/BAK KO and SW982/BAX KO cells were incubated for 8 h in the presence or absence of 15 µM ABT-199 and/or 5 nM PI (BTZ, CFZ, IXZ; + 10 µM Q-VD-OPh). Expression of MCL-1, BCL-2, BAK, BAX, NOXA, and GAPDH (loading control) was analyzed by Western blot. Asterisks indicate identical Western blot due to cutting/reprobing. BTZ bortezomib, CFZ carfilzomib, IXZ ixazomib, CTRL control.
SW982 cells. In line with previous data ABT-199 enhanced the expression of endogenous NOXA while expression of transgenic NOXA was unaffected (Suppl. Figure 3A). Noteworthy, these findings demonstrate that ABT-199 alone and in combination with BTZ induces transcription of PMAIP1/NOXA that translates into enhanced expression of NOXA (Fig. 3E, lower panel).
We investigated a role of TP53 in enhanced expression of NOXA in response to ABT-199 with or without PIs. As noted, expression of NOXA was enhanced by ABT-199 and unleashed in combination with BTZ or CFZ. In contrast, TP53 was slightly augmented by PIs but unaffected by ABT-199, implying TP53-independent NOXA induction (Fig. 3F). Nevertheless, we knocked down TP53 and analyzed ABT-199 and/or PI mediated PMAIP1 gene expression. ABT-199 alone and in combination with BTZ, CFZ and to a lesser extent IXZ induced expression of PMAIP1 that was not affected by TP53 knock-down. Furthermore, PMAIP1 induction by  alone and in combination with PIs was significantly stronger than induction by proteasome inhibition (Fig. 3G). Western blot showed that induction of NOXA expression mainly depends on ABT-199 and is augmented by PIs. Thus, qRT-PCR (Fig. 3G) and Western blot (Fig. 3C, H) indicate that ABT-199 strongly induces expression of NOXA by a transcription (Fig. 3E) and translation (Fig. 3D, lower panel) dependent mechanism, independent of TP53.
TP53 is dispensable for ABT-199&PI mediated induction of NOXA expression and cell death To broaden the analyzes, we investigated ABT-199&PI induced apoptosis in the rhabdomyosarcoma cell line RD with mutant TP53 [30], the non-small cell lung cancer cell line H1299 with deleted TP53 [31], and H1299 with reconstituted expression of TP53. In each cell line, ABT-199&PIs synergistically induced cell death (Suppl. Figure 2A  both, H1299/WT and H1299/TP53, ABT-199 strongly induced PMAIP1 mRNA expression (Suppl. Figure 2G). Western blot showed that in H1299 NOXA induction was induced by ABT-199 rather than PIs (Suppl. Figure 2H). These data confirm cell type and TP53independent induction of NOXA by ABT-199 and substantiate synergistic cell death induction by ABT-199&PIs.
ABT-199 activates ATF3&ATF4 that mediate NOXA expression Data show that the BCL-2 inhibitor ABT-199 induces expression of the MCL-1 antagonizing NOXA by a TP53-independent transcriptional mechanism. Roca-Portoles et al. [32] postulated that ABT-199 induces mitochondrial metabolic reprogramming, whereby ABT-199 impairs complex I/II activity of the electron transport chain (ETC), resulting in enhanced reductive carboxylation. To analyze metabolic reprogramming we detected Citrate and αketoglutarate by mass spectrometry. Analysis evidenced an increased ratio of α-ketoglutarate:citrate after incubation with ABT-199 and ABT-199&BTZ (Suppl. Figure 4A-C) confirming metabolic reprogramming. A disbalance in the ETC activates the integrated stress response pathway (ISR) and the activating transcription factor ATF4 that, together with ATF3, can induce expression of PMAIP1/NOXA [28]. Thus, we analyzed whether ATF3 and ATF4 mediate NOXA induction in response to ABT-199 by qRT-PCR and Western blot. Proteasome inhibition had negligible impact on ATF3 expression but ABT-199 enhanced transcription (6-fold) and expression (WB) of ATF3 (Fig. 4A,

ABT-199 induces activation of the integrated stress response pathway
We next analyzed the role of ISR for NOXA expression and cell death induction. ISR induces phosphorylation of the translationinitiation complex eIF2 subunit α (eIF2α) and (P)-eIF2α inhibits the translation complex eIF2B. This abolishes translation initiation from regular AUG start codons while enhancing translation from alternative AUG-codons resulting in increased expression of the activating transcription factor 4 (ATF4) and it´s downstream target ATF3. To analyze the role of ISR in ABT-199 induced NOXA expression we utilized the specific ISR-inhibitor ISRIB which antagonizes the inhibitory effect of (P)-eIF2α. ISRIB had no significant impact on ABT-199 and BTZ mediated transcriptional induction of ATF3, ATF4 and PMAIP1 (Fig. 4B-D). In cells incubated with ABT-199&BTZ, the increased expression of ATF3, PMAIP1, and ATF4 was even more pronounced but ISRIB had no significant impact on the expression of ATF3, ATF4 and PMAIP1. Thus, ISRIB showed limited effect on transcriptional induction of ATF3&ATF4 and PMAIP1.
ATF3&ATF4 double knock-down reduces ABT-199&BTZ mediated NOXA expression and cell death Finally, we directly investigated the relevance of ATF3&ATF4 for ABT-199 mediated NOXA induction. Simultaneous knock-down of ATF3 and ATF4 (Suppl. Figure 3B) reduced both PMAIP1 induction in the presence of ABT-199 alone from 1.8x to 0.9x and significantly reduced induction by ABT-199&BTZ by 50% from 3.8x to 2x (Fig. 5A). Western blot showed that strongest induction of NOXA by ABT-199&BTZ was reduced by knock-down of ATF3&ATF4. Thus, it is expected that knock-down of ATF3&ATF4 also reduces NOXA mediated cell death. Therefore, we investigated cell death in a time kinetic over 15 h using CellTOXGreen + . ABT-199&BTZ incubation at time points beyond 12 h resulted in significant fewer CellTOXGreen + cells in ATF3&ATF4 double knockdown cells as compared to siCTRL (Fig. 5B). Thus, ABT-199 induces ATF3&ATF4 thereby transactivating PMAIP1/NOXA and NOXA potentiates ABT-199&BTZ induced apoptosis.

DISCUSSION
We previously demonstrated that ABT-199 synergizes with BTZ to induce apoptosis in a BAX and NOXA dependent manner. Here, we show that ABT-199 also synergizes with PIs carfilzomib and ixazomib to induce apoptosis. Similar to ABT-199&BTZ, also ABT-199&CFZ enhances expression of NOXA with concomitant cell death induction. ABT-199&IXZ less efficiently induces NOXA and cell death. Reduced efficacy of IXZ may result from the lowest inhibitory activity for the chymotrypsin-like β 5 activity, but might also result from biological availability or stability. Regardless of PI efficacy, knock-down of PMAIP1 confirms that NOXA is a key mediator of synergistically induced apoptosis by ABT-199&PIs [26].
Wang et al. demonstrated NOXA induction by ISR, already considering ATF3 and ATF4 [38]. This is conceivable since a) ATF3 and ATF4 bind to the CRE (cAMP-responsive) element within the PMAIP1 promotor and activate NOXA expression [28,36,39] and b) ATF3 and ATF4 are crucially involved in ISR. Here, we show that ABT-199 and ABT-199&BTZ activate ATF3, ATF4 and induce ATF3&ATF4 dependent induction of PMAIP1/NOXA. Consequently, simultaneous knock-down of ATF3 and ATF4 reduces induction of PMAIP1 expression by ABT-199 and ABT-199&BTZ. In turn, reduced induction of PMAIP1/NOXA was functionally associated with a lower number of apoptotic cells, reflecting the importance of NOXA for synergistic cell death induction by ABT-199&BTZ.
Roca-Portoles et al. [32] described a BCL-2 independent effect of ABT-199 [32] that involves the transcription factor ATF4. ABT-199 affects the ETC and the TCA cycle mediating "metabolic reprogramming", leading to accumulation of α-ketoglutarate and succinate. Concomitantly, the ABT-199 induced imbalance in ETC activates the ISR and ATF4 [32]. In line, activity of the ETC affects sensitivity of MM to ABT-199 induced cell death [40]. Thus, it is conceivable that ABT-199 induced effects (metabolic reprogramming, ISR, and apoptosis) are exacerbated in the combined presence of ABT-199 and BTZ. A likely mechanism is enhanced ATF3&ATF4 mediated expression of PMAIP1/NOXA and simultaneously reduced degradation of NOXA by proteasome inhibition. Thus, knock-down of PMAIP1 reduces ABT-199&BTZ induced cell death. death induction. The ABT-199 induced transactivation of NOXA described here might re-sensitize resistant cancer cells or MCL-1 overexpressing tumors. Importantly, the mechanism described here is independent of TP53thus, ABT-199&BTZ is a therapeutic option for TP53-mutant or deleted tumors. Clinical studies propose that ABT-199 is effective not only in the treatment of AML [41] but also in patients with relapsed chronic lymphocytic leukemia (CLL) [6] or del(17p) CLL [7,42]. Furthermore, the combination of ABT-199 with BTZ and dexamethasone demonstrated promising efficacy in patients with relapsed/ refractory MM in phase 1b and 3 trials [43,44]. New PIs with different subunit-specificity and pharmacokinetic profiles have been developed and a phase 2 trial investigates ABT-199&CFZ + dexamethasone (NCT02899052), whereas the novel combination ABT-199&IXZ has not achieved clinical trial success to date (NCT03856112). These results might reflect the mechanisms proposed here. In view of the molecular aspects, i.e. BCL-2 inhibition and transactivation of NOXA, we propose that both, expression of BCL-2 and expression of NOXA are relevant for the efficacy of ABT-199&BTZ.

CONCLUSION
Molecular aspects elucidated in this study are summarized in Fig. 6. We show here for the first time that the BCL-2 inhibitor ABT-199 transactivates NOXA via ATF3&ATF4. Thus, in addition to BCL-2 inhibition, transactivated NOXA blocks MCL-1. Therefore, ABT-199 has a double impact on both, the BCL-2-BAX axis and the MCL-1-BAK axis of apoptosis signaling. Additional inhibition of the proteasome reduces NOXA degradation and exacerbates apoptosis induction by ABT-199&PI.
Solid cancers have shown to be less dependent on BCL-2 as compared to leukemias explaining why BH3-mimetics showed little clinical activity in solid tumors [45]. However, intrinsic BH3mimetic resistance may be overcome by combinatorial treatment with PIs utilizing previously unexpected synergistic mechanisms of action described here. Our data strongly suggest that ABT-199&PI combinations could be active in various solid malignancies.

Flow cytometry
Annexin V-APC, TMRM staining: Apoptotic cell death was assessed as previously described [46]. Cells were harvested using trypsin/ EDTA, resuspended in supernatant and washed in ice-cold PBS. Then, cell pellets were resuspended in 300 μL Annexin V-APCbinding buffer (PBS, 2.5 mM CaCl 2 ) supplemented with recombinant chicken Annexin V-APC (ImmunoTools, Friesoythe, Germany) and incubated for 10 min on ice. Subsequently, samples were analyzed using a FACS Lyric flow cytometer (Becton Dickinson, Heidelberg, Germany). To detect loss of mitochondrial membrane potential, cell pellets were resuspended in PBS supplemented with 2% FCS and 50 nM tetramethyl rhodamine (TMRM) (Merck, Darmstadt, Germany) of the potentiometric dye. Cells were incubated at 37°C for 20 min and fluorescence was analyzed using a FACS Lyric flow cytometer. The proportion of TMRM low and Annexin V-APC + cells was calculated using FACS Suite software.   Fig. 6 The proposed mechanism underlying synergistic cell death induction by ABT-199&PIs by transactivation of NOXA. Besides stabilization of TP53, proteasome inhibition by BTZ, CFZ, or IXZ prevents proteasomal degradation of the BH3-only protein NOXA. (I) The BH3mimetic ABT-199 antagonizes the anti-apoptotic activity of BCL-2 thereby releaving BAX. (II) ABT-199 reduces activity of complexes I and II of the mitochondrial ETC mediating metabolic reprogramming that induces ISR manifesting in (P)-eIF2α and increased expression of ATF4 and ATF3. ATF3&ATF4 mediate transactivation of PMAIP1/NOXA that inhibits the anti-apoptotic activity of MCL-1. Thus, unrestrained effectors BAX and BAK are free to oligomerize and mediate mitochondrial outer membrane permeabilization (MOMP) that sets the intrinsic apoptosis machinery in motion.

RNA interference
Knock-down experiments were performed according to the manufacturer´s protocol. Briefly, 0.6 × 10 5 cells/12-well or 2.5 × 10 5 cells/6-well were seeded 24 h prior to transfection. Then, cells were transfected with either 100 µl or 200 µl of Opti-MEM (Gibco) containing 50 nM ON-TARGET Plus Smartpool siRNAs targeting PMAIP1, or ATF3/ATF4 or non-targeted (NT) (Horizon Discovery, Waterbeach, UK) using Dharmafect Ι reagent (Thermo Fisher Scientific) according to the manufacturer's protocol. After 24 h cells were incubated with ABT-199 and/or BTZ for additional 24 h. Then, cells were harvested and analyzed by flow cytometry. For verification of knock-down efficacy, cells were lysed and assessed by Western blot analysis.
Cytotoxicity assay 0.6 × 10 5 cells were seeded in 12 well plates 24 h prior to transfection. Then, cells were transfected with siRNA specific for ATF3 and ATF4 or siCTRL. 8 h post transfection, non cell-permeable CellTOX Green Dye (Promega GmbH, Walldorf, Germany) was added and fluorescence was monitored in a Cytation 1 Cell Imaging Multi-Mode Reader (BioTek, Bad Friedrichshall, Germany). Cell death was assessed at indicated time points by counting the number of CellTOX Green positive cells.
Quantitative RT-PCR (qRT-PCR) Total RNA from cells was extracted using RNeasy Kit (Qiagen, Valencia, USA). 500 ng of RNA was reverse-transcribed by Moloney Murine Leukemia Virus Reverse Transcriptase (M-MLV RT, Pro mega GmbH, Walldorf, Germany). For gene expression analysis following Taqman Gene Expression Assay primer/probes were used: Hs00231069 for ATF3, Hs00909569 for ATF4, Hs00560402 for PMAIP1, Hs01034249 for TP53, Hs03023943 for β-ACTIN, Hs02758991 for GAPDH, Hs00362387 for TUBA1A (Life Technologies, Darmstadt, Germany). Gene expression analysis was performed in real-time PCR system (7900HT Real-time PCR system, Applied Biosystems, Singapore) and relative quantification was calculated with SDS2.4 software based on the expression level of either β-ACTIN alone or the three references genes β-ACTIN, GADPH and TUBA1A.
Proteasome-Glo™ Chymotrypsin-Like, Trypsin-Like and Caspase-Like Cell-Based Assays 5 × 10 3 cells were seeded in 96 well plates 24 h prior to incubation with 5 nM or 100 nM PI (BTZ, CFZ or IXZ) for 4 h. Then, Proteasome-Glo TM chymotrypsin-like, trypsin-like and caspaselike cell based reagent was added accordingly (Promega GmbH, Walldorf, Germany). Each reagent contains a luminogenic protease substrate specific for the three different subunits of the 26 S proteasome (chymotrypsin-like assay for the β 5 subunit, trypsinlike assay for the β 2 subunit and caspase-like assay for the β 1 subunit). Luminescence was detected in an Enspire Multimode Plate Reader (Perkin Elmer, Waltham, Massachusetts, USA). Reduction of luminescence indicates efficacy of the respective PI to inhibit the specific subunits of the 26 S proteasome.

Statistical analysis
Continuous variables are presented as mean as indicated and categorical variables are given by number and percentages. The statistical significance of differences was analyzed using students t-test. All statistical tests were considered significant when p < 0.05. Statistical analyses were calculated using GraphPad Prism (v5.04).

DATA AVAILABILITY
The data generated or analyzed during this study are included in this published article and its supplementary information files. Original data is available from the corresponding author on reasonable request.