Distribution of Bim determines Mcl-1 dependence or codependence with Bcl-x_L/Bcl-2 in Mcl-1–expressing myeloma cells

Oncogenic transformation results in up-regulation of apoptotic pathways; therefore, induction of antiapoptotic signals is required for cancer cells to maintain their survival. However, by overcoming the proapoptotic activity of oncogenic transformation, the tumor cells become dependent on the antiapoptotic signals. Thus, the regulation of proapoptotic versus antiapoptotic signals in cancer cells has been the focus of significant research efforts.^1-3 

Bcl-2 proteins integrate cell death and survival signals to determine whether caspase activation should occur after cellular stress. Caspases are activated after the release of the contents of the mitochondrial intermembrane space, and Bcl-2 proteins regulate the mitochondrial outer membrane permeabilization (MOMP). The effectors of MOMP are Bax and Bak, whose functions are negatively regulated by the antiapoptotic members of the Bcl-2 family (Bcl-2, Bcl-x_L, and Mcl-1).^1,2,4  The nature of this regulation remains somewhat controversial with one model suggesting that the sequestration of Bax and Bak by the antiapoptotic proteins inhibits MOMP.⁵  In a second model, Bax and Bak must be activated by the BH3-only proteins, Bim, tBid, and possibly Puma. Sequestration of these BH3-only proteins is the primary function of the antiapoptotic family members in this model, and their release is induced by additional BH3-only proteins.^6-9  We and others have previously suggested that these models may not be mutually exclusive.^10,11  Regardless of the mechanism tumor cells are generally dependent on antiapoptotic Bcl-2 proteins and would be predicted to be more susceptible to their inhibition.

One of the strategies to antagonize the function of antiapoptotic Bcl-2 proteins is to develop compounds that can mimic BH3-only proteins. One such agent ABT-737 mimics the BH3 domain of BAD and therefore selectively binds to Bcl-2, Bcl-x_L, and Bcl-w.¹²  Preclinical studies demonstrated that ABT-737 and the related orally active compound, ABT-263,¹³  were active in chronic lymphocytic leukemia, acute lymphocytic leukemia, lymphomas, and small cell lung cancer.^12,14-16  Furthermore, ABT-737 also induces apoptosis in multiple myeloma (MM) cells,^12,17-19  despite the fact that myeloma cells express Mcl-1,^20,21  which does not bind the BH3 domain of BAD and is associated with resistance to ABT-737.^22-25  Because ABT-737 is a good predictor of Bcl-2 dependence, we were prompted to investigate how the expression and/or interaction patterns of Bcl-2 family proteins play a role in determining the sensitivity of Mcl-1–expressing MM cells to ABT-737.

The 6 myeloma cells lines were obtained as previously described.¹⁰ 

ABT-737 and its less active enantiomer [(−)ABT] were provided by Abbott Laboratories.

The following primary antibodies were used: rabbit anti-Bak polyclonal antibody (pAb; Upstate Cell Signaling Solution); rabbit anti-Bax pAb (N-20, Santa Cruz Biotechnology); mouse anti-Noxa monoclonal antibody (mAb; Abcam); rabbit anti–Puma pAb (Cell Signaling), rabbit anti-Bim pAb (Chemicon), rabbit anti–Mcl-1 pAb (Stressgen); rabbit anti–Bcl-x_L pAb (13.6)²⁶ ; mouse anti–Bcl-2 mAb (sc-509, Santa Cruz Biotechnology), Bcl-w pAb (Chemicon); and the rabbit anti-actin pAb (Sigma-Aldrich). The ECL rabbit IgG, horseradish peroxidase-linked whole Ab (from donkey; GE Healthcare), and the anti–mouse IgG1-horseradish peroxidase conjugate (Roche Applied Science) were used as secondary antibody for Western blot. For coimmunoprecipitation, the following antibodies were used: mouse anti–Mcl-1 mAb (BD Biosciences), mouse anti–Bcl-x_L mAb (7B2.5),²⁶  and the mouse anti–Bcl-2 mAb sc-509.

The pcDNA3.1 (Invitrogen), pcDNA3.1-Mcl-1, and pcDNA3.1-Bcl-x_L-cDNA vectors were introduced into U266, MM.1s, and 8226/S and KMS11 cells by nucleofection (Amaxa) following the manufacturer's instructions (programs U266:X-005, MM.1s:O-023, 8226/S: G-015, KMS11: G-015). Nucleofected cells were plated in growth medium, and 0.5 μg/mL G418 was used for selection.

Cell death was measured by annexin V–fluorescein isothiocyanate (BioVision) and propidium iodide (PI) staining, as previously described.¹⁰ 

Western blotting was performed using standard techniques as previously described.¹⁰ 

Immunoprecipitation experiments were performed using the ExactaCruzTM C Kit (Santa Cruz Biotechnology) following the manufacturer's instructions as previously described.¹⁰ 

Silencing studies using small interfering RNAs (siRNAs) were obtained from Dharmacon RNA Technologies, selecting the ON-TARGETplus SMARTpool duplexes as the RNAi-specific technology platform. siRNA against human Mcl-1, Bak, Bax, Puma, Noxa, and Bim, and the siCONTROL nontargeting siRNA [siRNA(−)] were used. siRNAs were electroporated into cells by nucleofection (Amaxa) following the manufacturer's instructions as previously described.¹⁰ 

Myeloma cell lines were cultured with increasing concentrations of ABT-737 beginning with 50nM. Medium was replaced every 2 to 3 days increasing the drug concentration only when cell viability was higher than 50%. Final concentration was 0.5μM ABT-737 for 8226/S-ABTR and KMS18-ABTR and 2μM ABT-737 for U266-ABTR and KMS11-ABTR. Cells grown alongside the resistant cell lines in the absence of drug were used as controls (U266-CR2, 8226/S-CR2, KMS11-CR2, and KMS18-CR2).

Bone marrow aspirates from consenting myeloma patients were diluted to 20 mL with 1× PBS, and underlaid with lymphocyte separation medium (Mediatech). All samples were collecting following a University Emory Institutional Review Board–approved protocol. The buffy coat was collected and cells were washed, resuspended in culture medium, and stained with anti–CD38-phycoerythrin, anti–CD45-allophycocyanin-Cy7, and anti-CD138-fluorescein isothiocyanate antibodies (BD Biosciences) for fluorescence-activated cell sorter analysis. The cells were then prepared for plasma cell purification using the MACS Cell Separation MS Columns and CD138 magnetic microbeads per the manufacturer's protocol (Miltenyi Biotec). Once isolated, CD138-positive cells were checked for purity via flow cytometry. A total of 1 million cells (0.25 × 10⁶ cells/mL) were treated with the indicated concentrations of ABT-737 for 24 hours and apoptosis determined as described¹⁰ ; 1.5 to 2.0 × 10⁶ cells were used for coimmuoprecipitations and Western blot analysis.

Previous studies describing dependence on antiapoptotic proteins using BH3 profiling or ABT-737 sensitivity have concluded that cells are typically dependent on a primary antiapoptotic protein. We sought to determine whether cells could display codependence on several proteins, specifically Mcl-1 and Bcl-x_L or Bcl-2 as this could provide a therapeutic opportunity to use a Bad-mimetic, such as ABT-737, in tumors that are thought to be Mcl-1–dependent.

MM is a clonal plasma cell neoplasia of the bone marrow. Because myeloma cells are derived from a cell that is destined to be long-lived, it is not surprising that they express several antiapoptotic Bcl-2 proteins. Indeed, if one compares the expression of Mcl-1 from gene expression profiles of normal plasma cells, cells from patients with a premalignant plasma cell dyscrasia known as monoclonal gammopathy of undetermined significance, smoldering myeloma, or newly diagnosed MM, there is no difference in the median expression levels of Mcl-1 (Figure 1A). The only difference is in the top quartile of myeloma patients. This is consistent with a high-risk group of patients that have amplification of chromosome 1q where the Mcl-1 locus resides and is common in many malignancies.²⁷  Thus, nearly all myelomas express Mcl-1 at a level sufficient to predict they would be Mcl-1–dependent, a conclusion that was previously reached when Bcl-2 proteins were silenced with antisense oligonucleotides in 7 MM cell lines.^20,21  We confirmed and further expanded the results of others to 4 MM cell lines that were used in the course of these experiments. When Mcl-1 was silenced, all 4 cell lines were found to exhibit high levels of apoptosis consistent with the model that MM cell lines are Mcl-1–dependent (Figure 1B).

Although all MM cell lines appear to be Mcl-1–dependent, it was also previously demonstrated that ABT-737 induces apoptosis in MM cells, suggesting that some of these cell lines may also be Bcl-x_L/Bcl-2–dependent.^10,17-19  Indeed, when we tested 6 cell lines, apoptosis was detected after ABT-737 treatment, albeit with various levels of sensitivity. Consistent with Mcl-1 dependence, 3 myeloma cell lines (U266, KMS11, and OPM2) were relatively insensitive to ABT-737 with 50% effective concentration values of 2.58, 1.6, and 2.77μM, respectively (Figure 1C). In contrast, 3 additional cell lines (MM.1s, 8226/S, and KMS18) were sensitive to ABT-737 with 50% effective concentration values of 0.58, 0.30, and 0.39μM, respectively (Figure 1C). We also found that CD138⁺ cells isolated from the bone marrow aspirates of MM patients were sensitive to ABT-737 in a similar manner as the codependent line, MM.1s (Figure 1D). Thus, codependence exists in 50% of the cell lines tested and appears to be relevant in patient-derived samples.

To explain the differences in the sensitivity of MM cell lines to ABT-737 and thereby codependence, we investigated the expression levels of its direct targets, Bcl-2, Bcl-x_L, and Bcl-w. As demonstrated in Figure 2A, the sensitive MM.1s and the insensitive U266 cell lines had higher levels of Bcl-x_L, whereas the sensitive KMS18 and the insensitive U266 cell lines express greater Bcl-w. Finally, all cell lines tested have similar Bcl-2 levels. Taken together, expression levels of ABT-737 targets do not predict ABT-737 sensitivity. It is also important to note that the sensitivity of MM cell lines to ABT-737 does not correlate with their sensitivity to other agents, including bortezomib²⁸  or arsenic trioxide,¹⁰  suggesting that the variation in the sensitivity of MM cell lines to ABT-737 cannot be explained by global differences in their apoptotic threshold.

Because ABT-737 mimics the BH3 domain of BAD, it cannot antagonize the antiapoptotic function of Mcl-l. Thus, cells that are dependent on Mcl-1 for survival would be expected to be resistant to the effects of ABT-737. Indeed, several reports demonstrate that increased expression of Mcl-1 correlated with resistance to ABT-737 treatment.^22-25  Furthermore, it was also demonstrated that the BH3-only protein, Noxa, which selectively binds and inhibits Mcl-1, renders cells sensitive to ABT-737.^29,30  However, when these 2 proteins were analyzed in the 6 MM cell lines, no correlation was detected between the expression of either Mcl-1 or Noxa and ABT-737 sensitivity (Figure 2A). Similarly, the levels of another BH3-only protein, Bim, as well as the main effectors of MOMP, Bax, and Bak, did not predict the sensitivity of MM cells to ABT-737 treatment (Figure 2A). On the other hand, the 3 sensitive cell lines express higher levels of Puma. However, after a 24-hour incubation with 0.4μM ABT-737, there was a marked decrease in Puma expression, suggesting that it does not play a role in ABT-737-induced apoptosis (Figure 2B). Overall, these results demonstrate that the relative levels of Bcl-2-family proteins in myeloma cells do not predict Bcl-2 dependence or ABT-737 sensitivity.

Because the expression profile of the Bcl-2 proteins did not correlate with ABT-737 sensitivity, we next probed the activity of antiapoptotic proteins to determine whether the function of these proteins correlates with ABT-737 activity. The function of these proteins is to bind and sequester proapoptotic proteins; therefore, the interaction between the proapoptotic and antiapoptotic Bcl-2 members was analyzed. Although the initial ex vivo biochemical assays reported that ABT-737 has high affinity toward Bcl-w,¹²  in addition to Bcl-x_L and Bcl-2, these data were recently challenged by a report that demonstrated that Bcl-w is not a target of ABT-737 in vivo.³¹  Therefore, we focused our studies on investigating the binding of Bcl-2, Bcl-x_L, and Mcl-1 to determine which protein interactions may predict ABT-737 sensitivity. Consistent with our previous report as well as the findings of others, we found Bak, but not Bax, to be associated with either Bcl-x_L or Mcl-1.^10,32  Furthermore, silencing of Bak, but not Bax, inhibited ABT-737–induced apoptosis in 8226/S cells (supplemental Figure 1, available on the Blood Web site; see the Supplemental Materials link at the top of the online article). Taken together, these data suggest that, in myeloma cells, the interaction of Bak, but not Bax, is a potential target for ABT-737. A closer look at the association of these proteins revealed that there was a significant amount of Bak bound to Bcl-x_L in all 3 sensitive cell lines (MM1.s, 8226/S, and KMS18) and 2 insensitive cell lines (U266 and KMS11, Figure 3A). Only in one insensitive cell line, OPM2, Bak was bound primarily to Mcl-1 (Figure 3A). In addition, if we compare ABT-737–induced release of Bak from Mcl-1 and Bcl-x_L in an Mcl-1–dependent line (KMS11) and a codependent line (MM.1s), we observed similar patterns of concentration-dependent release of Bak from Bcl-x_L (supplemental Figure 2). Together, these data suggest that, although Bak release is necessary for ABT-737–induced apoptosis, the interaction between Bak and the prosurvival Bcl-2 proteins is not a determinant of ABT-737 sensitivity.

Among the BH3 proteins, we further analyzed Puma and Bim because they can bind to all 3 prosurvival Bcl-2 family proteins. In agreement with the data presented in Figure 2A, Puma was only detected in the 3 sensitive cell lines, and it was primarily associated with Bcl-2 and Bcl-x_L (Figure 3A). Furthermore, binding of Bim by either Bcl-x_L or Bcl-2 also appeared to correlate with ABT-737 sensitivity. In all 3 insensitive cell lines, Bim was found to primarily interact with Mcl-1; whereas in the 2 sensitive cell lines MM.1s and KMS18, Bim was primarily associated with either Bcl-x_L or Bcl-2 (Figure 3A). However, in one sensitive line, 8226/S, Bim was bound primarily to Mcl-1 compared with Bcl-x_L and Bcl-2. To determine whether this binding pattern was relevant in patient-derived samples, we performed coimmunoprecipation assays for Mcl-1 and Bcl-x_L from CD138⁺ plasma cells purified from bone marrow aspirates from 6 MM patients. As a reference for each sample, we performed parallel analyses on MM.1s cells. Two representative samples are presented in Figure 3B; and consistent with the sensitivity of these samples to ABT-737 (Figure 1D), Bim was found to be associated with both Bcl-x_L and Mcl-1.

To determine whether Puma and/or Bim plays a dominant role in predicting Bcl-x_L/Bcl-2 dependence, activity of ABT-737 was measured after silencing of these proteins. Interestingly, silencing of Puma had no protective effect on ABT-737 activity in the 3 lines tested (Figure 4). However, silencing of Bim in 3 MM lines inhibited ABT-737 activity (Figure 4). In addition, Bim is released from Bcl-x_L in MM.1s and KMS11 in a concentration-dependent fashion by ABT-737 (supplemental Figure 2). Together, these results suggest that the interaction of Bim with Bcl-x_L and Bcl-2 underlies Bcl-2 dependence and the sensitivity to ABT-737. The 8226/S cells appear to be the exception to this conclusion; therefore, we next investigated the possible reasons for codependence in this cell line.

A possible explanation for why 8226/S cells are sensitive to ABT-737 despite the majority of the Bim being associated with Mcl-1 is that endogenous expression of another BH3-only protein that has high affinity for this prosurvival protein may limit its function, thus rendering these cells to be functionally codependent. One such protein, Noxa, is constitutively expressed in this cell line (Figure 2A). Indeed, previous reports demonstrated that induction of Noxa in various tumor types sensitizes cells to ABT-737–induced apoptosis.^29,30  Consistent with this possibility, knock-down of Noxa abrogated the activity of ABT-737 in 8226/S cells, whereas it had no effect in KMS11 or MM.1s cells (Figure 4). Therefore, Noxa and Bim may be cooperating in ABT-737–induced apoptosis in these cells. Immunoprecipitation of Mcl-1 revealed that higher levels of Noxa were bound to Mcl-1 in 8226/S compared with KMS11 and MM.1s cells (Figure 5A).

Because Noxa and Bim are occupying Mcl-1 in 8226/S and should not be released by addition of ABT-737, we next determined the fate of Bim associated with Bcl-x_L and Bcl-2 in these cells. Consistent with 8226/S being a codependent line, Bim is easily displaced from Bcl-x_L and Bcl-2 by ABT-737 in these cells (Figure 5B). Whereas the released Bim is a minor fraction of the total bound Bim, the Noxa associated with Mcl-1 appears to interfere with Mcl-1's ability to bind drug-released Bim (Figure 5A). This is the probable explanation for the Noxa dependence of ABT-737–induced apoptosis in these cells. Similarly, in the other sensitive cell line tested (MM.1s), concentration-dependent displacement of Bim from Bcl-x_L and Bcl-2 is observed. However, because these cells do not express significant levels of Noxa and it is not detected in the Mcl-1 coimmunoprecipitation, marked drug-induced Bim association with Mcl-1 is observed (Figure 5A). However, MM.1s express significantly higher levels of total Bim (Figure 5C) compared with 8226/S and KMS11; therefore, sufficient levels of free Bim may still be available to induce MOMP. In the insensitive KMS11 cells, Bim is released from Bcl-x_L and Bcl-2 in concentration-dependent manner; however, this represents a minor faction of the total bound Bim (Figure 5A). Therefore, the concentration of ABT-737 required to release enough Bim to activate Bak appears to be greater in the Mcl-1–dependent cell lines.

To directly test the role of Noxa in these interactions, we silenced Noxa and performed coimmunoprecipitation of the antiapoptotic proteins. Consistent with our model, Noxa silencing resulted in a 61% to 71% increase in the amount of Bim associating with Mcl-1. Interestingly, the Bim appeared to be released from Bcl-2 where a comparable 74% to 83% decrease was observed, whereas no change in the association with Bcl-x_L is seen (Figure 5D-E). Furthermore, in the absence of Noxa, ABT-737–released Bim is sequestered by Mcl-1 (supplemental Figure 3). Taken together, these findings further indicate that Noxa expression perturbs the ability of Mcl-1 to bind Bim rendering 8226/S cells sensitive to ABT-737 treatment and thereby codependent.

We next tested whether forced overexpression of antiapoptotic proteins could change the dependence pattern and thus inhibit ABT-737–induced apoptosis. When Bcl-x_L is exogenously overexpressed, the capacity of Bcl-x_L–BH3 binding should significantly increase, which would require higher concentrations of ABT-737 to neutralize the antiapoptotic function. Indeed, in the codependent 8226/S and MM.1s cells, stable overexpression of Bcl-x_L resulted in diminished sensitivity to ABT-737 compared with their vector control-transfected counterparts (Neo, Figure 6A). However, Bcl-x_L did not significantly reduce the activity of ABT-737 in the Mcl-1–dependent lines, U266 and KMS11 (Figure 6A). This is consistent with Bim already being associated with Mcl-1 and the cells maintaining their Mcl-1 dependence. Similar to Bcl-x_L, overexpression of Mcl-1 also protected 8226/S and MM.1s, but not U266 cells from ABT-737–induced apoptosis (Figure 6B). Mcl-1 does have a significant effect on ABT-737–induced apoptosis compared with the Neo control; however, they were not different from the parental controls, which were less sensitive than the vector-transfected cells.

To determine whether overexpression of either Mcl-1 or Bcl-x_L alters the distribution of Bim bound to the antiapoptotic proteins, the association between these proteins were analyzed in the KMS11, MM.1s, and U266 transfectants. In KMS11 and MM.1s cell lines, stable overexpression did not alter the distribution of Bim (Figure 6C). In U266 cells, a slight increase in Bim–Mcl-1 association was observed in the Mcl-1–transfectants; whereas in the Bcl-x_L–transfected cells, Bim occupation of Bcl-x_L was increased (Figure 6C). However, these changes had no phenotypic consequences. Thus, forced expression of an antiapoptotic protein does not necessarily alter the distribution of Bim bound to Mcl-1 versus Bcl-x_L, suggesting that expression levels of antiapoptotic Bcl-2 family proteins is not the primary determinant of Bim distribution and therefore not adequate to predict the dependence of a cell type on a particular prosurvival protein.

To further investigate the role of the interaction and expression patterns of Bcl-2-family proteins in codependence and ABT-737 sensitivity, ABT-737–resistant 8226/S, KMS18, KMS11, and U266 cell lines were generated. The resistant counterparts of the codependent cell lines, 8226/S and KMS18, displayed 7.5- and 5.9-fold resistance to ABT-737, respectively, compared with parental cells (CR2) that were maintained in culture in parallel to the selected cells to assure that the time in culture during selection was not a contributing factor to resistance (Figure 7A). Acquired resistance of Mcl-1–dependent cell lines, U266 and KMS11, resulted in 12.1- and 17-fold resistance respectively, compared with CR2-cells. (Figure 7A).

Analysis of the expression of Bcl-2 family proteins in the resistant cell lines demonstrated that both ABT-737–resistant 8226/S and KMS18 displayed increased expression of Mcl-1 compared with their control counterparts (Figure 7B). Interestingly, there was also a modest reduction in Bak expression in ABT-737–resistant KMS18 cells. Puma expression was lower in both 8226/s and KMS18 (Figure 7B), although the significance of this reduction is unclear. In contrast to 8226/S and KMS18, the expression of Mcl-1 did not change in the resistant counterparts of U266 and KMS11 (Figure 7B). Strikingly, in both of these lines, there was a marked reduction in Bim levels.

We next investigated the effects of acquired resistance on the distribution of Bim and Bak (Figure 7C). Interestingly, when ABT-737 resistance was acquired in the codependent cell lines, Bim is found to be exclusively associated with Mcl-1 (Figure 7C). In contrast, the distribution of Bak did not change in either of the resistant cell lines. As expected, in the ABT-737–resistant U266 and KMS11 cell lines, almost all the remaining Bim was found to be associated with Mcl-1 while there was no apparent change in the interaction of Bak (data not shown). Overall, these data indicate that the total amount of Bim bound to Bcl-x_L or Bcl-2 underlies the Bcl-2/Bcl-x_L dependence and sensitivity to ABT-737.

Recent advances in our understanding of the complex interactions behind the induction of apoptosis revealed that changes that result in an increased apoptotic threshold are not simply limited to overexpression of prosurvival Bcl-2 proteins.³³  To determine which Bcl-2 protein interactions are indispensable for a particular tumor cell type, a technique called BH3 profiling has been developed.³⁴  In addition, development of the small molecule BH3-mimetic, ABT-737, provided another effective tool to emulate BH3 profiling in whole cells.³⁴  Goldsmith et al recently demonstrated in neuroblastoma cells that BH3 profiling in isolated mitochondria closely correlates with the sensitivity of these tumors to ABT-737.³⁵  Thus, studies using ABT-737 can shed light on targeting Bcl-2 dependence in tumor cells.

Because ABT-737 mimics Bad, which selectively binds to Bcl-x_L, Bcl-2, and Bcl-w, a tumor cell that has a strong Mcl-1 dependence would be expected to be resistant to this agent. Indeed, several reports demonstrate that expression of Mcl-1 as well as its antagonist BH3 protein, Noxa, modulates the sensitivity of acute myeloid leukemia, small cell lung cancer, melanoma, head and neck cancer, and colon cancer cells to ABT-737.^{22,23,25,29,30,36-38}  Similarly, cancers, such as chronic lymphocytic leukemia, in which oncogenic transformation is associated with Bcl-2 up-regulation, are shown to be highly sensitive to ABT-737.^14,15  These findings help provide the rationale for phase 1/2 clinical trials with an orally available BH3-mimetic, ABT-263 (Navitoclax), in chronic lymphocytic leukemia, lymphoma, and small cell lung cancer.¹ 

Although MM survival has been known to be Mcl-1–dependent,^20,21  data presented here and in other reports^17-19  demonstrate the sensitivity of these cells to ABT-737. Three of the 7 MM cell lines that were found to be Mcl-1–dependent in these earlier studies were tested in the current study. Our findings confirm that U266 and OPM-2 are Mcl-1–dependent²¹  and explain why silencing of Mcl-1 in 8226/S cells would induce apoptosis,²⁰  despite the cells appearing to be Bcl-2/Bcl-x_L–dependent. Together, these data suggest that 8226/S is codependent on Mcl-1 and Bcl-2/Bcl-x_L. Although all the cell lines tested in this study are Mcl-1-dependent, 50% are codependent on Bcl-x_L/Bcl-2. This suggests that expression or even dependence on Mcl-1 does not preclude targeting of Bcl-x_L/Bcl-2 in cancer cells. Thus, although it is well documented that overexpression of Mcl-1 confers resistance to ABT-737,^25,29,37  our findings here emphasize that interactions, rather than expression levels of Bcl-2 proteins, play a more significant role in determining the sensitivity of a cell to this compound and by extension Bcl-2 dependence.

The analysis of interactions between various Bcl-2 proteins suggests that binding of Bim and Bak to Bcl-x_L and Bcl-2 correlates with sensitivity of MM cell lines to ABT-737. However, a closer look at the interactions of these 2 proapoptotic proteins indicate that the distribution of Bim is a better predictor of ABT-737 sensitivity (supplemental Figure 4, codependence models I and II). Bak was equally distributed between Bcl-x_L and Mcl-1 in most of the insensitive and sensitive MM cell lines, whereas Bim was highly bound to Mcl-1 in the former cells. Furthermore, cells selected for ABT-737 resistance, either the expression or the interaction of Bim, was found to be altered in all 4 cell lines, whereas Bak levels were slightly reduced in only 1 cell line and its interaction pattern did not change in any of them. Therefore, it appears that the amount of Bim that is displaced from Bcl-x_L and Bcl-2 underlies the sensitivity of MM cell lines to ABT-737. These findings suggest that cells that are codependent on Mcl-1 and Bcl-x_L/Bcl-2 have levels of Bim priming such that release from any antiapoptotic is sufficient to induce MOMP (supplemental Figure 4). This was not limited to cell lines as we found copriming of Bim in 6 freshly isolated patient samples that displayed sensitivity to ABT-737.

The question of how the distribution of Bim among various antiapoptotic Bcl-2 proteins is determined in each cell line remains elusive. Although phosphorylation of either Bcl-2³⁹  or Bcl-x_L⁴⁰  after treatment with various chemotherapeutics has been shown to inhibit their antiapoptotic function, there are no data, to our knowledge, that demonstrate how the interactions between proapoptotic and antiapoptotic proteins are regulated in cells in the absence of external stimuli. A common notion is that expression of an antiapoptotic protein may render a cell dependent on its function for survival. However, findings presented here argue that this is not the only possibility. For example, forced overexpression of either Mcl-1 or Bcl-x_L did not result in a shift of Bim binding between these antiapoptotic proteins in 2 of the 3 cell lines tested. Interestingly, in U266 transfectants, overexpression of both Mcl-1 and Bcl-x_L leads to greater levels of Bim associated with the transfected antiapoptotic protein, suggesting that in certain cell types Bim distribution may be determined by the expression levels of prosurvival proteins. However, there was no consequence to the change in Bim distribution in these cells; therefore, the relevance of this finding is not completely clear. Nevertheless, taken together, these findings suggest that how the affinity of Bim for each prosurvival Bcl-2 protein is regulated appears to be complex and not solely depend on the expression levels of these proteins. A plausible explanation is that post–translational modifications may alter the structure of the BH3-binding groove in antiapoptotic Bcl-2 proteins modulating the affinity of these proteins for BH3 domains. Alternatively, post–translational modifications of Bim could result in altered affinity for antiapoptotic proteins. Thus, post–translational modification of Bim may play a role in its association with individual antiapoptotic Bcl-2 proteins. Indeed, Bim is reported to contain various serine/threonine phosphorylation sites that regulate its stability and apoptotic activity.^41-45  Further research is required to determine whether the phosphorylation of these residues may also affect the affinity of Bim toward each antiapoptotic Bcl-2 protein.

An exception to the model in which Bim association with Bcl-x_L/Bcl-2 is shown to correlate with sensitivity to ABT-737 was the 8226/S cell line where Bim is found to be primarily associated with Mcl-1. Interestingly, these cells also highly express another BH3 protein, Noxa, which selectively binds to Mcl-1. Furthermore, knock-down of Noxa significantly protected 8226/S from ABT-737–induced apoptosis. Thus, in the presence of Noxa, Mcl-1 may not be able to adequately sequester the Bim that is displaced from Bcl-x_L and Bcl-2; therefore, ABT-737 can effectively kill 8226/S cells (supplemental Figure 4, codependence model II). We have also observed a similar effect in KMS18, suggesting why Mcl-1 does not serve as a sink for Bim in these codependent cells (not shown). Consistent with this idea, when 8226/S and KMS11 cell lines were compared, it was determined that after incubation with ABT-737, the Bim/Mcl-1 association does not change in the former cell line, whereas the association was enhanced in the latter (supplemental Figure 4, Mcl-1–dependent model vs codependent model II). Moreover, in MM.1s cells, in which there is a higher proportion of Bim bound to Bcl-x_L and Bcl-2, Bim/Mcl-1 association increases significantly after ABT-737 treatment, indicating that in these cells Mcl-1 cannot neutralize enough of the Bim that is displaced by ABT-737 to prevent MOMP (supplemental Figure 4, codependence model I). Taken together, these data suggest that the ratio of “unprimed” Mcl-1 to the Bim that is bound to Bcl-x_L and Bcl-2 predicts sensitivity to ABT-737. This would also explain why increasing Mcl-1 in KMS11 and U266 cells had no effect on ABT-737–induced cell death.

Based on this prediction, there are 2 possible ways to sensitize cells to ABT-737 treatment, either through the up-regulation of Bim bound to Bcl-x_L and Bcl-2 or the down-regulation and/or inactivation of Mcl-1. The first possibility is supported by a recent report by Chen et al, which demonstrates that the histone deacetylase inhibitor, suberoyl bis-hydroxamic acid, synergizes with ABT-737 in both leukemia and MM cell lines through induction of Bim, which is sequestered by Bcl-x_L and Bcl-2.⁴⁶  Thus, certain agents not only increase Bim expression but also result in Bcl-2/Bcl-x_L dependence, making them excellent choices to combine with ABT-737. In this regard, previous reports have demonstrated either a synergistic or additive activity of dexamethasone,^18,19  which is also reported to induce Bim.^47,48  The second possibility is supported by the demonstration that agents that induce Noxa, such as bortezomib and arsenic trioxide, synergize with ABT-737.^10,19 

Overall, our findings indicate that the interactions in conjunction with expression of Bcl-2 proteins determine whether a cell is Bcl-2/Bcl-x_L–dependent and can be targeted with ABT-737. These data have several clinical implications, such as selecting the correct subset of patients to treat with ABT-737 as well as searching for agents that may synergize with this BH3-mimetic. Furthermore, although Mcl-1 expression is observed in all MM cells, it appears that some of these cells may still rely on Bcl-x_L and Bcl-2 activity; therefore, ABT-737 could have activity in multiple myeloma beyond sensitizing cells to other therapies.

The authors thank John Shaughnessy for helpful advice in the analysis of the gene expression profile datasets he deposited in GEO.

This work was supported in part by the National Institutes of Health (grants R01 CA127910 and CA129968) and the T.J. Martell Foundation (L.H.B.) and the Deets Foundation (S.L.). The ABT-737 and its enantiomer were a gift from Abbott Laboratories (Abbott Park, IL). L.H.B. is a Georgia Cancer Coalition Distinguished Scholar.

National Institutes of Health

Contribution: A.A.M., M.K., and S.M.M. designed the experiment, executed the experiments, and wrote the manuscript; J.L., D.S., and D.M.G. performed experiments; J.L.K., K.P.L., and S.L. designed the experiment, critically reviewed the data, and prepared the manuscript; and L.H.B. oversaw the project, designed the experiment, interpreted data, and prepared the manuscript.

Conflict-of-interest disclosure: L.H.B. is an author on a patent related to the development of ABT-737 and ABT-263. In the past 3 years, he has received a milestone payment related to the development of ABT-263. The remaining authors declare no competing financial interests.

Correspondence: Lawrence H. Boise, Department of Hematology and Medical Oncology, Emory University School of Medicine, 1365C Clifton Rd, Atlanta, GA 30322; e-mail: lboise@emory.edu.