Mechanisms of MCL-1 Protein Stability Induced by MCL-1 Antagonists in B-Cell Malignancies

The BCL-2 family proteins are divided into three major classes and all share a BCL-2 homology (BH) domain. Mcl-1 is a member of the anti-apoptotic cohort (1–3). Interactions among these proteins and their stoichiometry are central to survival and apoptosis of several types of cancer cells. The impetus to develop a small molecule pharmacologic agent to neutralize the actions of Mcl-1 survival protein was two-fold. First, the successful targeting of BCL-2 by venetoclax was encouraging. Second, investigations of somatic copy-number alterations identified Mcl-1 among the top genes amplified in tumors (4). Recently, several MCL-1 inhibitors (MCL-1i) have entered clinical trials including AstraZeneca's AZD5991 (NCT03218683; refs. 5, 6), Amgen's AMG-176 (NCT02675452; ref. 7), and Novartis's S64315 (NCT04629443; refs. 8, 9) with proven effectiveness in preclinical investigations of hematologic malignancies.

A common feature of treatment with these MCL-1i is an increased stability in Mcl-1 protein level (5, 7, 8, 10) as observed in breast cancer (11), acute myelogenous leukemia (7, 8), acute lymphoblastic leukemia, and multiple myeloma (7, 9) cell lines. However, detailed mechanistic insight into the ubiquitous observation of increased Mcl-1 protein level after treatment with MCL-1i is lacking.

Mcl-1 is a short half-life protein (12), which can be regulated at multiples levels including transcription (13), translation (14, 15), and degradation (16). Mcl-1 is degraded by the ubiquitin–proteasomal system and is under the control of ubiquitination by E3 ligases like Mule (17), Trim-17 (18), or β-TRCP (19). Conversely, deubiquitinases (DUB) like USP9x (20), KU-70 (21), and USP13 (22) reverse this process and contribute to Mcl-1 protein stability. Mcl-1 posttranslation modifications at Thr163 (23) and Ser159 (24) control ubiquitination.

We hypothesized that elevation of Mcl-1 protein level after treatment with MCL-1i may involve mechanisms of Mcl-1 production, stability, and/or degradation. To test our hypothesis, we utilized hematological malignancy cell lines and primary chronic lymphocytic leukemia (CLL) cells to determine mechanisms leading to increased Mcl-1 protein levels.

Two different heme-malignancy cell lines including MEC1 (CLL), and Mino (mantle cell lymphoma), HEK293, USP9x WT, and knockout HCT116 cell lines (RRID:CVCL_0291) and primary CLL cells were used in this study. CLL cells were obtained after patients (Supplementary Table S1) signed written informed consent for inclusion in the study, and the study was conducted in accordance with the Declaration of Helsinki. The protocol was approved by the Institutional Review Board at MD Anderson Cancer Center. Cell number and volume were determined with Coulter channelizer (Coulter Electronics). Detailed culture conditions are described in Supplementary Materials and Methods.

AMG-176 and AZD5991 were purchased from Chemietek; MG-132, Degrasyn (WP1130), maritoclax, and trametinib were purchased from Selleckchem. Agents were dissolved in DMSO (Sigma-Aldrich).

Cells were treated with MCL-1i and lysates were prepared and printed in five serial dilutions onto slides along with the normalization and expression controls. The array included 920 samples and slides were probed with one of 386 antibodies as detailed in the Supplementary Materials and Methods.

pCMV-Tag2A vector and HA-UB plasmids were kind gifts. pCMV-Flag-hMcl-1 (T163A; #25391) and pCMV-Flag-WT-hMcl-1 (#25392; RRID:Addgene_25392) were obtained from Addgene. Five different human Mission sh-USP9x clones, non-Target shRNA control (SHC002V), and Mission Lentiviral packaging mix (SHP001) were obtained from Sigma. Control and sh-RNA-USP9x along with Lentiviral packaging mix were individually transfected to HEK293T cells (RRID: CVCL_HA71) for virus preparation and infection as described previously (25).

Subconfluent HEK293 cells were transiently transfected with 3 μg of plasmids using Fugene 6 transfection reagent (Promega; ref. 25). Seventy-two hours after transfection, cells were lysed in IP Lysis Buffer (Pierce) supplemented with protease inhibitors cocktail (cOmplete, Protease, and PhosSTOP inhibitors; Roche). Cell extracts (500 μg) were immunoprecipitated with EZview anti-Flag M2 Agarose (Sigma) or EZview Red Anti-HA (Sigma) Affinity Gels overnight at 4°C and processed for immunoblot analysis. For endogenous co-immunoprecipitation (co-IP) experiments, MEC1 or Mino cells were processed as above, and complexes were captured by using Protein A/G-Agarose beads (Santa Cruz Technology) at room temperature for 1 to 2 hours. The beads were washed three times with lysis buffer and immunoblotted.

Cell fractionation and immunoblot analysis was performed as described earlier (10, 25). Detailed procedure is described in Supplementary Materials and Methods. List of antibodies and their sources are provided in Supplementary Table S2.

MEC1, Mino cell lines, and peripheral blood mononuclear cells from CLL patient samples were treated by DMSO, AMG176, or AZD5991 for 24 hours. Cells were then collected and processed for apoptosis assay by Annexin V-PI staining as described in Supplementary Materials and Methods.

Total RNA from MEC1 and Mino cells was isolated by using RNeasy Mini Kit. RT reaction was performed by using RevertAid H Minus First Strand cDNA Synthesis Kit (Thermo Fisher Scientific). Real-time RT-PCR was performed by using primers (Supplementary Table S3) and the SYBR Green PCR Master Mix (Applied Biosystems; ref. 25) as detailed in the Supplementary Materials and Methods.

To purify ubiquitinated FLAG-Mcl-1, HEK293 was cotransfected with HA-ubiquitin and FLAG-Mcl-1 plasmids in 10 cm plate. Seventy-two hours posttransfection, cells were harvested, and total cell lysate was extracted using IP lysis buffer (Pierce), followed by immunoprecipitation of FLAG-Mcl-1 using EZview anti-Flag M2 Agarose (Sigma) overnight, and eluted by 3x FLAG peptide (Thermo Fisher Scientific) according to the manufacturer's protocol. This was followed by a second immunoprecipitation using HA beads (Invitrogen) to isolate ubiquitinated FLAG-Mcl-1 as per manufacturer's protocol. For Mule and USP9x purifications, either rabbit anti-Mule or rabbit anti-USP9x antibodies were coupled to Dynabeads M-280 Sheep Anti-Rabbit IgG (RRID: AB_2783009) for 1 hour. The beads were then washed and HEK293 protein lysates were added to the antibody bound beads and incubated by vertical rotation for 2 hours at 4°C. To maintain the activity and functionality of the purified proteins, gentle elution was done by 2M NaCl for 10 minutes.

Full-length His-tag recombinant Mcl-1 protein was allowed to bind to magnetic His-beads (Invitrogen) according to the manufacturer's protocol. Detailed assay is described in Supplementary Materials and Methods.

In vitro ubiquitination of Mcl-1 protein by Mule was done using the Ubiquitinylation Kit (Enzo life Sciences, Inc.). To avoid extensive backgrounds produced by the biotin-conjugated ubiquitin, we replaced it with recombinant FLAG-tagged ubiquitin (R&D Systems). Briefly, recombinant Mcl-1 was incubated in the presence or absence of recombinant Mule with either DMSO, AMG-176, AZD5991, or maritoclax in ubiquitination buffer containing Mg-ATP, E1 enzyme, and E2 enzyme (Recombinant Human UbcH7/UBE2L3 Protein; R&D systems) according to the manufacturer's protocol. The reactions were incubated for 4 hours at 37°C and quenched by the addition of 2× laemmli. Ubiquitination was assessed by immunoblot analysis and probed with anti-FLAG. For the in vitro ubiquitination of Mcl-1 by purified Mule from HEK293 cells, 5 μmol/L recombinant human ubiquitin aldehyde protein (R&D Systems) was added to inhibit de-ubiquitinating enzymes.

Ubiquitinated HA-FLAG-Mcl-1 was purified as above and then incubated with purified USP9x from HEK293 lysate in DUB buffer at 37°C in the presence of either DMSO, AMG-176, or AZD5991 in a total reaction volume of 75 μL. Reactions were allowed for 20 minutes, then quenched by the addition of 4x laemmli and subjected to immunoblot analysis and probed with anti-HA to detect ubiquitination. For detecting in vitro DUB activity, a DUB Activity Assay Kit (Cayman Chemicals) was used to assess the direct effect of AMG-176 or AZD5991 on the de-ubiquitinating enzyme USP9x. USP9x (either recombinant or purified from HEK293) or MEC1 total cell lysates pretreated by 2 μmol/L AMG-176 or AZD5991 for 6 hours) were incubated in the presence or absence of DMSO, AMG-176, or AZD5991 in DUB buffer for 15 minutes in a 96-well plate according to manufacturer's protocol. The fluorogenic AMC-Ub was then added to the reactions and immediately assessed for fluorescence in the microplate reader for 1 hour in kinetic mode. Readings were taken every minute.

We constructed a 3D model using RaptorX (RRID:SCR_018118), which is an ab initio method (http://raptorx.uchicago.edu/). The modeled structure was combined with the known Mcl-1 crystal structure complexed with AMG-176 (PDB entry: 6OQC). To understand the effect upon inhibitor binding, we conducted molecular dynamics simulations on both AMG-176–bound and AMG-176–removed Mcl-1 structures for 50ns using GROMACS (RRID:SCR_014565) with GROMOS96 force field (http://www.gromacs.org/) as detailed in the Supplementary Materials and Methods.

Statistical analysis was performed by using GraphPad Prism Software (RRID:SCR_002798). All numerical results are presented as mean ± SEM. The statistical significance of differences was analyzed using grouped analysis, ANOVA, and multiple t test (Student t test).

The data generated in this study are available within the article and its Supplementary Data files. For raw data request, please contact the corresponding author at [email protected].

RPPA of MCL-1i–treated MEC1 and Mino cell lines (P = 0.01) revealed selective upregulation of Mcl-1 (P < 0.003) and cleaved PARP (P < 0.004) protein levels, whereas SHIP2 protein level was downregulated (P = 0.009; Fig. 1A; Supplementary Fig. S1). Interestingly, at less stringent P value (P = 0.05), AMG-176 and AZD5991 induced several nonidentical proteomic changes (Supplementary Figs. S1A–S1C). Induction in Mcl-1 protein level was confirmed by immunoblot analysis in a time and dose-dependent manner (Fig. 1B) and in whole cells including mitochondria (Supplementary Fig. S1D). Mcl-1 upregulation was not due to transcription (Fig. 1C) but due to increased stability of Mcl-1 protein. Cycloheximide chase experiment confirmed increased half-life of the Mcl-1 protein following incubation with MCL-1i (DMSO 0.68 hours, AMG-176 2.94 hours, and AZD5991 3.50 hours; Fig. 1D), whereas co-incubation of cycloheximide and AMG-176 in both MEC1 and Mino cells blunted Mcl-1 protein accumulation (Fig. 1E). Despite higher abundance and stability of Mcl-1, MCL-1i induced apoptosis in Mino cell line (Supplementary Fig. S1E) as well as peripheral blood lymphocytes from CLL-patient samples (Supplementary Fig. S1F), indicating that the accumulated Mcl-1 protein did not provide any resistance to MCL-1i and lost its anti-apoptotic function. Interestingly, at 500 nmol/L, AZD5991 was more potent than AMG-176 in inducing apoptosis in Mino cell line (P = 0.0185) and CLL patient samples (P = 0.0001).

The ubiquitin (UB)–proteasome system tightly regulates Mcl-1 stability and degradation (26). The increased Mcl-1 stability induced by the inhibitors prompted us to explore Mcl-1 ubiquitination. IP and reverse co-IP analysis of MEC1 cells treated with DMSO, AMG-176, or MG-132 (positive control) revealed strong inhibition of Mcl-1 ubiquitination in AMG-176–treated cells. As expected, MG-132 treatment resulted in increased recruitment and accumulation of polyubiquitinated Mcl-1 (Fig. 2A). Conversely, AMG-176 treatment followed by a washout restored Mcl-1 ubiquitination. These data demonstrate that Mcl-1 protein accumulation occurs via inhibition in the Mcl-1 ubiquitination upon binding with these inhibitors. Consistently, IP and Co-IP of HEK293 transfected with WT-FLAG-Mcl-1 and HA-UB confirmed disruption in HA-FLAG interactions in AMG-176–treated cells that was reversed following withdrawal of AMG-176 from the cells (Fig. 2B).

Because enhanced Mcl-1^Thr163 phosphorylation is associated with increased stability of Mcl-1 protein (27), we examined MCL-1i-induced posttranslational modification. Treatment with MCL-1i resulted in increased total Mcl-1 protein as well as phosphorylated Mcl-1 (Mcl-1^Thr163 phosphorylation) in both cell lines and primary CLL patient lymphocytes (Fig. 3A and B). Moreover, a direct linear relationship was established between total Mcl-1 protein and Mcl-1^Thr163 phosphorylation in either inhibitor-treated (r² = 0.92 and P < 0.0001 for each drug) primary CLL cells (Fig. 3B). This upregulation of Mcl-1^Thr163 phosphorylation may be mediated by ERK activation as we noticed an increase in ERK phosphorylation with MCL-1i (Fig. 3C–E) which was downregulated upon treatment with the MEK/ERK inhibitor, trametinib (Fig. 3D and E) in cell lines as well as primary CLL lymphocytes. Trametinib partially decreased the MCL-1i–induced upregulation of both total Mcl-1 protein and Mcl-1^Thr163 phosphorylation in Mino cells and CLL lymphocytes (Fig. 3D and E). The use of QVD, a caspase inhibitor, did not rescue Mcl-1 downregulation in trametinib-treated cells, suggesting that the downregulation of Mcl-1 with trametinib in Mino cell line and CLL patient samples is due to ERK-mediated inhibition and decrease in Mcl-1^Thr163 phosphorylation rather than caspase activation and cleavage of Mcl-1 protein (Supplementary Fig. S2A). However, in MEC1 cells treated with trametinib, despite loss of Mcl-1^Thr163 phosphorylation, total Mcl-1 expression remained unchanged (Fig. 3D; Supplementary Fig. S2B). These data led us to examine the role of Mcl-1^Thr163 phosphorylation on Mcl-1 protein–protein interactions rather than stability. Interestingly, overexpression of T163A-FLAG-Mcl-1 in HEK293 cells showed loss of interaction with endogenous BAK, BIM, and Noxa but not BAX as compared with the WT-FLAG-Mcl-1 supporting a critical role for Mcl-1^Thr163 phosphorylation in modulating Mcl-1 interaction with the pro-apoptotic members (Supplementary Fig. S2C). Although Mcl-1^Ser159 phosphorylation also appeared to be increased in AMG-176–treated cells, it was less pronounced than the total protein of Mcl-1 (Supplementary Fig. S2D). In contrast, we did not see any changes in total and GSK3β^Ser9 phosphorylation (Fig. 3D). The use of the phospho-mutant T163A-Mcl-1 did not significantly reverse the effect of MCL-1i–induced Mcl-1 upregulation in HEK293 cells (Fig. 3F). Overall, we showed that Mcl-1^Thr163 phosphorylation via ERK partially contributed to the observed Mcl-1 stability.

Because BH3 proteins, including the E3 ligase Mule and the pro-apoptotic Noxa, can regulate Mcl-1 ubiquitination, stability, and half-life (17, 28), we assessed the impact of MCL-1i on expression of these proteins and their interaction with Mcl-1. Interestingly, a transient decrease in Mule expression was observed 1 hour following treatment with MCL-1i (Fig. 4A; one-way ANOVA, P = 0.02), which increased at later time points (6 and 24 hours). Noxa expression also dramatically decreased following treatment with MCL-1i, without any changes in Trim-17 or β-TRCP expression. To further support our observation, we incubated MEC1 cell line with AZD5991 and followed up change in protein levels at shorter time intervals. Indeed, treatment by AZD5991 decreased Mule protein expression at 30 minutes followed by gradual increase to reach normal level at 6 hours, whereas Noxa expression decreased in a time- and dose-dependent manner (Supplementary Fig. S3A). This transient decrease in Mule expression is thought to be due to decreased stability of Mule protein (Supplementary Fig. S3B) mediated by a decreased Mule:USP9x interaction (Supplementary Fig. S3C). Interestingly, our IP data showed that Mule:Mcl-1 interaction was not disrupted (Fig. 4B), despite both binding to the same BH3 pocket of Mcl-1 protein. In contrast, pro-apoptotic proteins including Noxa, BAK, and BAX were disrupted in either MEC1 (endogenous; Fig. 4C, left) or FLAG-Mcl-1 and HA-UB co-transfected HEK293 cells (Fig. 4C, right). In vitro pull-down assay of recombinant GST-tagged Mule (catalytic domain) or Mule IP lysate from HEK293 cells by His-tagged recombinant full length Mcl-1 confirmed that Mule:Mcl-1 interaction was maintained in the presence of inhibitors (Fig. 4D). Interestingly, Mule was found to dimerize under non-denaturing condition (Supplementary Fig. S3D). We performed an in vitro ubiquitination assay to assess Mule's role to ubiquitinate Mcl-1 in the presence of MCL-1i. AMG-176 was found to enhance Mcl-1 ubiquitination in similar magnitude as maritoclax, whereas AZD5991 failed to ubiquitinate Mcl-1 significantly compared with DMSO (Fig. 4E). Similar observations were seen with the full-length Mule IP lysate in an in vitro Mcl-1 ubiquitination study (Supplementary Fig. S3E), so much so, that even in the absence of the recombinant Mule catalytic domain, E2 conjugating enzyme can ubiquitinate Mcl-1 protein (Fig. 4E, lane 5). This observation indicates that MCL-1i can directly induce enhanced (AMG-176) or decreased (AZD5991) Mcl-1 ubiquitination independent of Mule. However, Mule as a catalyst can accelerate the ubiquitination process. No significant change in Bim or PUMA expression was noticed (Fig. 4F). The transient decrease in Mule expression, loss of Noxa:Mcl-1 interaction, and decreased Noxa protein may contribute to more Mcl-1 protein accumulation induced by MCL-1i.

Because both DUBs and multiple E3 ligases, including Mule (20–22), modulate Mcl-1 stability, we sought to explore the role of DUBs in Mcl-1 stability induced by MCL-1i. MCL-1i treatment did not significantly change protein levels of DUBs (USP9x, USP13, or KU-70, one-way ANOVA, P = 0.35; Fig. 5A). Co-IP experiment in MEC1 cells treated with MCL-1i showed increased USP9x interaction with Mcl-1 compared with DMSO control. However, this might be due to increased amount of Mcl-1 protein in the MCL-1i–treated samples (relative ratio of USP9x:Mcl-1 is 0.9 with AMG-176 and AZD5991 and 1 with DMSO; Fig. 5B). To test the DUB functional activity, we used in vitro de-ubiquitination assay and demonstrated that MCL-1i enhanced the de-ubiquitination of Mcl-1 by USP9x as compared with DMSO (Fig. 5C). However, the enhanced DUB activity is less likely to be a result of direct effect of MCL-1i on USP9x as indicated by normal DUB assay activity in the presence of MCL-1i as compared with DMSO (Fig. 5D; Supplementary Figs. S4A–S4C). Interestingly, AMG-176 showed an intrinsic fluorescence property as compared with AZD5991. Even in the absence of recombinant USP9x, AMG-176 showed a very high relative fluorescence (Supplementary Fig. S4D), which decreased with lower doses of AMG-176 as compared with AZD5991 (Supplementary Fig. S4E).

To further confirm the role for DUBs in the observed Mcl-1 protein stability, we tested the effect of the DUB inhibitor WP1130, which abrogated Mcl-1 induction in response to MCL-1i in both MEC1 and Mino cell lines (Fig. 5E) as well as in CLL lymphocytes, and upregulated Noxa level (Fig. 5F). WP1130 has been tested before to inhibit DUB activity in a dose-dependent manner (1–5 μmol/L); at a dose of 5 μmol/L, WP1130 inhibited 80% of DUB activity (29). These data demonstrate that inhibition of DUBs can completely reverse the effect of MCL-1i on Noxa and Mcl-1 protein. However, USP9x knockout in HCT116 (Fig. 5G) or USP9x knockdown in MEC1 cells (Fig. 5H) did not abrogate Mcl-1 protein stability. This may be due to the overexpression of other DUBs (e.g., USP24), which may retain overall DUB activity (30). Indeed, a compensatory increase in KU-70 (HCT116-USP9x-KO) and USP24 (MEC1-shRNA-USP9x) were observed whereas strong upregulation of Noxa in HCT116-USP9x-KO was seen similar to WP1130-treated cells. In contrast, Mule expression was dramatically decreased by loss of USP9x, reflecting an important role for USP9x in regulating Mule (Fig. 5G and H).

Primary process by which MCL-1i elicit their effect is through binding to Mcl-1 protein, which results in conformation change. Our in silico modeling suggests that residues 1 to 172 in the Mcl-1 protein might have largely disordered structures with high structural flexibility. This may be why, to date, no crystal structures have been obtained for these residues, which is in agreement with a previous report (5). However, our modeling also revealed that residues 150 to 172 fold back and form close interactions with the inhibitor binding site in Mcl-1. Molecular dynamics simulation showed that binding of AMG-176 or AZD5991 to Mcl-1 protein significantly affects the conformation of these 23 residues. In particular, the inhibitor binding pushes the residues away from one of the α-helices and increases the solvent-accessible surface area of Thr¹⁶³ from 22.01 Ų to 64.79 Ų and Ser¹⁵⁹ from 35.63 Ų to 42.76 Ų (Fig. 6A). These findings indicate that the two residues, especially Thr¹⁶³ (>42 Ų difference), may become more accessible to other partnering proteins, such as protein kinases, and may lead to increased phosphorylation and stabilization of the Mcl-1 protein (Fig. 3). AMG-176 and AZD5991 bind to the same BH3 pocket such as the BH3 domains of BAX, BAK (Fig. 6B and C), and Mule (Fig. 6D). So, the binding of MCL-1i can disrupt the binding of other BH3 proteins. Indeed, BAX–BAK interaction was disrupted but not Mule (Fig. 4B and C). Moreover, the interaction of USP9x, which binds to the N-terminus of Mcl-1, can be enhanced by these N-terminus conformation changes where several residues in Mcl-1, including Ser155, Ser159, and Thr163 (cyan sticks), are spaced out and become more exposed (see SASA data), and thus prone to phosphorylation. Once this occurs, the increased negative charges (phosphate groups) will enhance their binding to USP9x (orange) by interacting with several positively charged residues including Arg1936, Arg1940, Arg1941, and Lys1943 on USP9x (Fig. 6E). On the basis of our data, we provide a schematic summary for the mechanisms of MCL-1i–induced stability of Mcl-1 protein (Fig. 6F).

MCL-1i–induced upregulation of Mcl-1 protein was shown to be due to increased stability, without any change in transcription, and it correlated with target engagement of MCL-1i to Mcl-1 protein, which can serve as a biomarker of target engagement (5). Although disruption/inhibition of Mule binding to Mcl-1 by the MCL-1i for targeting Mcl-1 proteasomal degradation could be an attractive explanation for the upregulation of Mcl-1 protein, data supporting this hypothesis are inconclusive and lack mechanistic experiments by MCL-1i (17). Consistent with the data of Mcl-1 upregulation by MCL-1i, we showed that AMG-176 and AZD5991 upregulated Mcl-1 protein in B-cell lines and primary CLL lymphocytes due to increased Mcl-1 stability rather than increased transcription. However, this increased Mcl-1 stability failed to offer any resistance to the action of MCL-1i as evidenced by loss of BAK:Mcl-1 interaction and concomitant induction of apoptosis.

These data indicate that disruption of Mcl-1 protein interaction with pro-apoptotic molecules is enough to induce apoptosis regardless of Mcl-1 protein accumulation. However, this Mcl-1 protein accumulation serves as a biomarker for target engagement, as shown previously using FRET studies by Tron and colleagues (2018), where Mcl-1 protein level increases with MCL-1i targeting Mcl-1 and is reversed with washout of Mcl-1i as shown in Fig. 2. Sensitivity to MCL-1i varies among cell lines and CLL patient samples. Current mechanistic investigations further confirm our prior report that there is no association of apoptosis with Mcl-1 protein induction (10). Because of the presence of caspase cleavage sites, Mcl-1 protein undergoes cleavage and disintegration once apoptosis initiates. We also showed decreased Mcl-1 ubiquitination by AMG-176 that was reversible upon AMG-176 washout, suggesting MCL-1i–induced defective Mcl-1 ubiquitination as a possible mechanism for this stability.

Treatment with MCL-1i induced an ERK-mediated increase in Mcl-1^Thr163 phosphorylation in B-cell lines and CLL lymphocytes that may be enhanced by increasing SASA to T163 and S159 due to Mcl-1 conformation change (Fig. 6A). Although the use of trametinib, a MEK inhibitor, abolished Mcl-1^Thr163 phosphorylation, the change in total Mcl-1 level was variable among cell lines and patient samples, suggesting a partial and cell-type–specific effect. Further, the use of the phospho-mutant T163A-Mcl-1 did not significantly reverse the effect of MCL-1i–induced Mcl-1 upregulation. Conversely, no significant changes were observed in Mcl-1^ser159 or GSK3β expression. Collectively, Mcl-1^Thr163 phosphorylation only partially associates with MCL-1i–induced Mcl-1 upregulation.

BH3-only proteins were shown to regulate Mule-dependent ubiquitination of Mcl-1, where Bim opposes Mule:Mcl-1 interaction leading to its stability (31). In contrast, Noxa promotes Mule-dependent Mcl-1 ubiquitination by inhibiting USP9x:Mcl-1 interaction leading to Mcl-1 degradation (32). MCL-1i are also BH3 mimetic molecules (5, 7) and hence may regulate Mcl-1 protein level. With this respect, Mule expression transiently decreased following treatment with MCL-1i and compensated by gradual increase at later time points. Further, this transient loss in Mule expression was associated with decreased USP9x:Mule interaction affecting Mule stability. In addition, MCL-1i induced loss of Noxa level following disruption of Mcl-1:Noxa interaction, triggering Noxa degradation. The Noxa downregulation may be due to decreased stability following its disruption. The disruption of Noxa interaction and its decreased expression may, in part, explain the decreased Mule:USP9x interaction, where Noxa overexpression can increase Mule:USP9x and decrease Mcl-1:USP9x interaction (32).

Despite the loss of interaction of pro-apoptotic BH3 proteins (e.g., BAX, BAK, and Noxa) to Mcl-1 by MCL-1i, Mule:Mcl-1 interaction was not disrupted by MCL-1i in both endogenous co-IP as well as in an in vitro pull-down assay. One possible mechanism for the observed lack of disruption of Mcl-1:Mule interaction in the presence of MCL-1i could be the access and availability of additional site(s) of interaction outside the BH3 binding groove of Mcl-1, where Mule could also bind to the N-terminal 30 amino acids of Mcl-1 (33). This in turn can allow unrestricted Mule:Mcl-1 binding even when a BH3-only protein interacts with Mcl-1. This is supported by the evidence that other BH3 mimetics like Noxa and maritoclax can bind to the BH3 groove of Mcl-1 and, at the same time, potentiate Mule interaction (32, 34, 35). However, this fails to explain why Bim inhibited Mule ubiquitination to Mcl-1 in in vitro studies (31). Hence, we characterize here two categories of BH3-only proteins that regulate Mule interaction to Mcl-1. BH3-only proteins, such as Bim and PUMA, which bind nonspecifically to Mcl-1 tend to block Mule interaction to Mcl-1 (31). On the other hand, BH3-only proteins that bind specifically to Mcl-1, such as Noxa, do not interfere with Mule interaction to Mcl-1. Specific MCL-1i such as AMG-176 and AZD5991 are predicted to bind to Mcl-1 in a similar manner as Noxa, although Noxa does not impair Mule interaction to Mcl-1.

To verify if Mule binding is still functional, we performed in vitro ubiquitination studies of Mcl-1 protein in the presence of MCL-1i with or without Mule. Interestingly, AMG-176 was shown to enhance Mcl-1 ubiquitination in a similar way to martioclax, an MCL-1i that is known to ubiquitinate and degrade Mcl-1 protein. Ubiquitination occurred even in the absence of Mule, suggesting a direct effect of AMG-176 on Mcl-1 ubiquitination, independent of Mule. Indeed, the addition of Mule catalyzed this process. On the other hand, the magnitude of Mcl-1 ubiquitination in the presence of AZD5991 is far less than AMG-176. These data suggest that a possible conformation change induced by AMG-176 and maritoclax results in Mcl-1 ubiquitination. Conversely, AZD5991 does not induce similar conformation change in the Mcl-1 protein and hence had little effect on Mcl-1 ubiquitination compared with AMG-176. The reason for this discrepancy between AMG-176 and AZD5991 is not clear. However, it may be due to binding differences between AMG-176 and AZD5991 to the BH3 groove of Mcl-1. AZD5991 has an extra binding site to the BH3 groove of Mcl-1, where the 17-chloride atom of AZD5991 was shown to bind to the backbone carbonyl alanine-227 whereas with AMG-176 the chlorine atom is buried in the induced fit hydrophobic pocket (5, 36). Also different BH3 molecules can induce different Mcl-1 Q221R222N223 motif conformation either towards a nonhelical (e.g., Bim) conformation that does not favor ubiquitination or a helical (e.g., maritoclax, Noxa, Mule) one that favors ubiquitination (35, 37). However, further studies are needed to explain this discrepancy. Elucidating differences between MCL-1i might help guide the development of potent MCL-1i since AZD5991 is shown to be more potent than AMG-176 (Supplementary Figs. S1E and S1F). Another interesting point, if AMG-176 can enhance Mcl-1 ubiquitination in in vitro ubiquitination assays like maritoclax, then why does AMG-176 stabilize Mcl-1 whereas maritoclax degrades Mcl-1 in treated cells? One possible explanation may be the difference in Noxa expression where maritoclax does not affect Noxa expression significantly (34), whereas we showed that AMG-176 downregulated Noxa expression. Normally, Noxa regulates USP9x interaction to Mcl-1.

The observed AMG-176–induced ubiquitination of Mcl-1 in in vitro ubiquitination studies triggered us to explore the role of DUBS in the observed stability, because it can reverse the ubiquitination process. Following treatment with MCL-1i, our IP data showed similar increase for either USP9x or Mcl-1. However, in vitro DUB assay revealed enhanced USP9x activity following treatment with MCL-1i compared with DMSO that is more prominent with AZD5991. Although the observed in vitro high DUB activity is not related to the direct effect of MCL-1i on USP9x, we predict that MCL-1i–induced change in Mcl-1 conformation may expose N-terminus offering enhanced USP9x activity. Interestingly, we also report here for the first time the intrinsic fluorescence properties of AMG-176, which can be clinically useful in tracking AMG-176 pharmacokinetics.

Pharmacologic inhibition of DUBs by WP1130 completely abrogated Mcl-1 induction following treatment with MCL-1i; however, genetic knockout of USP9x was ineffective in our study. Failure of USP9x knockout to control Mcl-1 stability in HCT116 has been reported previously (38). Later, Peterson and colleagues (30) showed that the HCT116 USP9x-knockout retained its DUB activity. They attributed it to the possible compensatory increase in other DUBs (e.g., USP24), which may compensate for USP9x deficiency and the need for complete depletion of USP9x because small level of the protein can take over the function. Besides, it is generally more effective to inhibit multiple DUBs (30, 39). This may explain why WP1130, but not USP9x knockout, abrogated the MCL-1i–induced stability of Mcl-1 protein. Here we also showed KU-70 upregulation in HCT116-USP9x knockout. Similar observation was noted in MEC1-shRNA-USP9x with compensatory increase in USP24. Further, we reported that the loss of USP9x downregulated Mule expression significantly in either HCT116-USP9x-KO or MEC1-shRNA-USP9x. This may stabilize Mcl-1 and explain why sometimes USP9x knockdown does not affect Mcl-1 stability. It is interesting to see a dramatic decrease in Mule compared with no significant change in Mcl-1 expression with USP9x downregulation. This highlights a dual function for USP9x where, on one hand, it stabilizes Mule, possibly enhancing Mcl-1 degradation, and, on the other hand, it de-ubiquitinates Mcl-1. The switch between the two functions can possibly be regulated by Noxa. Pharmacologic or genetic loss of USP9x results in an increase in Noxa. Although this was not fully evaluated, a possibility is that WP1130-mediated ER stress may induce ATF4, a transcription factor for Noxa (40).

The AZD5991 clinical trial (NCT03218683) has been on hold recently after reporting a case with asymptomatic elevation in cardiac parameters (https://www.ashclinicalnews.org/online-exclusives/fda-places-clinical-hold-trials-azd5991). In concert, another MCL-1i, S63845, impaired contractility and myofibril assembly in human-induced pluripotent stem cell cardiomyocytes (41). Our data showed that MCL-1i decreased Mule and Noxa stability with a potential role in MCL-1i–induced cardiotoxicity, although further investigations are required. Mule was suggested to protect the heart against oxidative stress; its cardiac knockdown in mice led to cardiac hypertrophy and left ventricular dysfunction (42). Also, Noxa was shown to regulate autophagy, and its downregulation mediated phenylephrine-induced cardiac hypertrophy (43–46). Collectively, these indirect mechanisms and direct targeting of Mcl-1 may be responsible for drug-induced cardiotoxicity.

In conclusion, MCL-1i are not only important in clinic, they serve as powerful tools to explore Mcl-1 protein regulation by BH3 mimetics. MCL-1i enhanced Mcl-1 de-ubiquitination through Noxa disruption, enhanced de-ubiquitination and Mule destabilization. They bind to Mcl-1 specifically, like Noxa, and hence they do not impair the BH3-only protein E3 ligase Mule interaction to Mcl-1, unlike other BH3 proteins (e.g., Bim, PUMA). Although MCL-1i are similar in their mechanism of action, striking differences were elaborated that can help guide development of more potent MCL-1i. Molecular changes of Noxa downregulation and Mule de-stabilization following MCL-1i treatment may be potential inducers of the reported cardiotoxicity induced by MCL-1i, although further investigations are required.

S.I. Tantawy reports other support from Administration of Cultural Affairs and Missions, Egypt, during the conduct of the study. V. Gandhi reports grants from CLL Global Research Foundation Alliance, as well as other support from MD Anderson's Chronic Lymphocytic Leukemia Moon Shot® during the conduct of the study. No disclosures were reported by the other authors.

S.I. Tantawy: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing–original draft, writing–review and editing. A. Sarkar: Conceptualization, data curation, formal analysis, supervision, validation, investigation, methodology, writing–original draft, writing–review and editing. S. Hubner: Resources, data curation, formal analysis, writing–review and editing. Z. Tan: Software, formal analysis, investigation, writing–review and editing. W.G. Wierda: Resources, writing–review and editing. A. Eldeib: Supervision, writing–review and editing. S. Zhang: Software, formal analysis, investigation, writing–review and editing. S. Kornblau: Resources, data curation, formal analysis, writing–review and editing. V. Gandhi: Conceptualization, resources, formal analysis, supervision, funding acquisition, validation, investigation, writing–original draft, project administration, writing–review and editing.

This work was supported in part by an Alliance grant from the CLL Global Research Foundation (to V. Gandhi); by MD Anderson's Chronic Lymphocytic Leukemia Moon Shot program (to V. Gandhi and W.G. Wierda); and by the NIH/NCI through MD Anderson's Cancer Center Support Grant, CA016672. Z. Tan receives research funding (RR220039) from Cancer Prevention and Research Institute of Texas (CPRIT). S. Tantawy received a scholarship from the Administration of Cultural Affairs and Missions, Egypt, to work as a visiting student at MD Anderson. Authors are thankful to Dr. Rong Chen, Dr. Betty Lamothe, and Dr. H. Amin, The University of Texas MD Anderson Cancer Center, for providing pCMV-Tag2A vector, HA-UB plasmids, and Mino cell lines, respectively. Authors are thankful to Dr. Fred Bunz, John Hopkins University, for providing WT and USP9x knockout HCT116 cell lines. We thank Stephanie Deming of Scientific Publications, Research Medical Library, MD Anderson Cancer Center, for editing the manuscript and Danielle Walsh, Scientific Project Director, CLL Moon Shot, for reviewing final revised manuscript.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.