Chronic lymphocytic leukemia requires BCL2 to sequester prodeath BIM, explaining sensitivity to BCL2 antagonist ABT-737 (original) (raw)
ABT-737 kills CLL cells in the low nanomolar range. ABT-737 is a cell-permeant small molecule that binds with high affinity to BCL2, BCL-XL, and BCL-w in the subnanomolar range. The negative control enantiomer is a stereoisomer of ABT-737 that binds to BCL2 with reduced affinity and has been used as a loss-of-function control (14, 15). CLL cells have shown sensitivity to ABT-737, so we initially selected them as a cancer model of BCL2 dependence (15). To test CLL cell sensitivity to ABT-737, freshly isolated primary CLL cells were incubated with ABT-737 or negative control enantiomer. After 48 hours, CLL cell death was assessed using annexin V staining. In all 24 CLL samples tested, apoptosis was induced within 48 hours by ABT-737 with an EC50 of 4.5 ± 2.2 nM (Figure 2A) (range 1.9–9.4 nM; see Supplemental Table 1; supplemental material available online with this article; doi:10.1172/JCI28281DS1). The negative control enantiomer was approximately 100-fold less potent (mean EC50: 574 nM) (Figure 2A). Directly targeting the apoptotic machinery might be expected to induce apoptosis rapidly. To test this prediction, we examined apoptosis in 5 CLL patient samples treated for 4 hours. These samples responded similarly to those receiving 48-hour treatment (Figure 2B). Importantly, nonmalignant PBMCs from normal donors were resistant to ABT-737, with an EC50 of more than 1000 nM (Figure 2C).
BCL2 antagonist ABT-737 efficiently kills primary CLL cells. (A) CLL cells from 24 patient samples (numbered at right) were cultured for 48 hours with different concentrations of ABT-737 or negative control enantiomer (enant). Death was quantitated by annexin V staining and normalized to controls treated with solvent (DMSO). The effective concentration at which 50% killing is observed (EC50) is shown. (B) CLL death induced by ABT-737 is rapid. CLL cells harvested from 5 patient samples (numbered at right) were cultured for 4 hours with different concentrations of compounds. Death was quantitated as in A. (C) Normal PBMCs are not sensitive to ABT-737. Four normal PBMC samples (numbered at right) were cultured for 24 hours in the presence of the indicated concentrations of compounds. Death was quantitated as in A.
Table 1 summarizes the clinical characteristics of the source patients in Figure 2A. We compared EC50 values between groups dichotomized by factors previously identified as prognostically useful in CLL. This analysis revealed that in no case did a difference in mean EC50 between groups exceed 2 nM. A nonparametric statistical comparison of the groups showed that none differed with statistical significance except the groups dichotomized by leukocyte count (Table 2). Thus, biological response to ABT-737 appears to be largely independent of traditional prognostic factors. These studies confirm that primary CLL cells are indeed sensitive to ABT-737, validating the use of primary CLL cells as a human cancer model for BCL2 dependence.
Characteristics of CLL patients contributing samples used in Figure 2A
Comparison of clinical characteristics with in vitro sensitivity to ABT-737
BCL2 and BIM levels are consistent among CLL samples. Given the consistency of response to ABT-737, we wondered whether BCL2 protein expression was also uniform in CLL cells. Because the antiapoptotic BCL2 and the proapoptotic BIM (16, 17) proteins are important determinants of commitment to apoptosis in lymphocytic cells, we tested the levels of these proteins in CLL cells. Western blots using antibodies specific for BIM and BCL2 revealed that BCL2 and BIM levels among 15 CLL cell samples were remarkably uniform (Figure 3A). Conversely, levels of BCL2 and BIM in PBMCs were consistently lower (Figure 3B).
Protein expression in CLL cells. (A) BCL2 and BIM levels were uniform across 15 CLL samples tested by immunoblot of 5 μg whole-cell lysates. Three isoforms of BIM (BIMEL, BIML, and BIMS) are shown. (B) BCL2 and BIM levels in PBMCs were lower than in CLL. Whole-cell lysates (5 μg) evaluated by immunoblot. Three normal PBMC lysates are depicted at left, CLL lysates at right. (C) Short-term culture does not affect BCL2 or BIM protein levels. Two independent CLL lysates generated before and 48 hours after culture were probed for BCL2 and BIM. (D) BCL2 protein levels in 6 primary CLL cells (a–f) and primary FL cells are similar as seen by indexing immunoblots to lysates from the t(14;18)-containing H2 human lymphoma cell line. (E) Purified glutathione-S-transferase–tagged BCL2 (GST-BCL2) and GST-MCL1 (10–300 ng) were run next to 5 independent CLL lysates (5 μg) to quantitate BCL2 and MCL1. MCL1 is detectable only upon long exposure (bottom panel). Relative amounts of BCL2 and MCL1 were calculated using densitometry and are depicted in Table 3. (F) Antibody detecting full-length and cleaved MCL1 does not demonstrate MCL1 cleavage products in CLL lysates. (G) CLL lysates (10 μg) reveal low levels of BID and no detectable cleaved BID. DHL4 (DHL) cell lysate (10 μg), known to express BID, and recombinant caspase-8–cleaved his-tagged BID (tB) (35 ng) were run as positive controls for antibody recognition. Note that the his-tag causes slower migration of uncleaved BID. (Numbers at top of gels correspond to patient sample numbers.)
To ensure that short-term culture did not affect BCL2 or BIM levels and perhaps alter response to ABT-737, protein lysates were made from CLL cell samples at the time of isolation and after 48 hours in culture. Neither BCL2 nor BIM levels changed during culture (Figure 3C). Because follicular lymphoma (FL) also overexpresses BCL2, albeit via the t(14;18) translocation, and exhibits sensitivity to ABT-737 (15), we compared BCL2 levels in CLL cells to levels in FL cells (Figure 3D). BCL2 levels were notably similar in the 2 diseases.
MCL1 and BID levels are low in CLL cells and likely have limited influence on the response to ABT-737. Next, we decided to determine the levels of MCL1 in CLL cells for several reasons: (a) MCL1 is an antiapoptotic relative of BCL2 expressed in CLL cells; (b) MCL1 can bind to BIM but is not antagonized by ABT-737 and might be expected to compensate for the loss of BCL2 function and maintain survival following ABT-737 treatment; (c) cleavage of MCL1 by caspases can allow release of BIM that can activate BAX (18, 19); and (d) higher levels of MCL1 expression have been suggested as portending inferior clinical outcome (20, 21). To investigate the relative importance of MCL1 in maintaining survival in CLL cells, we quantitated amounts of MCL1 and BCL2 by Western blot (Figure 3E and Table 3). Using defined amounts of recombinant MCL1 and BCL2 proteins as quantitative protein standards, we found that MCL1 was expressed at levels 4- to 14-fold lower than those of BCL2 in CLL cells. To measure cleaved forms of MCL1, we prepared immunoblots of CLL whole-cell lysates either before or after treatment with ABT-737 and probed using an antibody that can detect cleaved forms of MCL1 (22). We found that cleaved forms of MCL1 were undetectable (Figure 3F and Supplemental Figure 1A). These results suggest that MCL1 in intact or in cleaved form is present at relatively low amounts compared with BCL2 in CLL cells and does not play an important role in determining cellular response to ABT-737.
Quantitation of BCL2 and MCL1 levels in primary CLL samples in Figure 3E
Like BIM, BID has been identified as an activator BH3-only protein (8, 13, 23). Therefore, we examined BID levels in CLL cells. Using an antibody that can detect either full-length or caspase-8–cleaved p15 BID, we found that full-length p22 BID was barely detectable in whole-cell lysates. In addition, caspase-8–cleaved p15 BID was not detectable (Figure 3G). Treatment with ABT-737 did not induce increased levels of either p15 or p22 BID (Supplemental Figure 1B). Furthermore, neither form of BID was detectable as an immunoprecipitated complex with BCL2 (Supplemental Figure 1C). These data argue against an important role for BID in determining death following antagonism of BCL2 by ABT-737 in CLL cells.
CLL mitochondria reveal a tonic dependence on BCL2 function to maintain outer membrane integrity. Since BCL2 opposes the mitochondrial pathway of apoptosis, we hypothesized that the toxicity of ABT-737 was based on a mitochondrial requirement for BCL2 function in CLL. We have characterized a panel of peptides derived from the BH3 domains of BH3-only proteins that behave as selective antagonists of BCL2 function in several genetically defined model systems (8, 14). For instance, BH3 peptides from BAD, PUMA, BMF, and, with lower potency, BIK bind and inhibit BCL2 function whereas BH3 domains from NOXA, HRK, and BNIP-3A do not interact with BCL2. We have validated this panel with other antiapoptotic family members. The pattern of interaction is distinct for each antiapoptotic protein so that the function of each may be specifically detected by BH3 profiling.
Mitochondria that depend on BCL2 function for maintenance of their outer membrane integrity show induction of outer membrane permeability when treated with BAD, PUMA, and BMF but not NOXA, HRK, and BNIP-3A peptides. Therefore, CLL mitochondria were incubated with BH3 peptides as well as ABT-737 and negative control enantiomer (Figure 4). BAD, PUMA, and BMF induced cytochrome c release whereas the NOXA, HRK, and BNIP-3A peptides and a point-null mutant of BAD BH3 did not. This pattern is diagnostic of mitochondrial BCL2 dependence (14). ABT-737 also induced cytochrome c release, validating that its target is located at the mitochondria of CLL cells and is required to prevent MOMP. Therefore, these BH3 profiling experiments demonstrate that CLL mitochondria depend on BCL2 function to maintain mitochondrial outer membrane integrity. Furthermore, these data elucidate a mechanism for the exquisite sensitivity of CLL cells to ABT-737 treatment. Finally, since the BH3 peptides in the sensitizer panel lack the ability to directly activate BAX and BAK, these experiments also implicated the presence of an activator molecule constitutively sequestered by BCL2 in CLL cells. CLL cells are therefore “primed for death” (14).
BH3 profiling of CLL mitochondria reveals BCL2 dependence. Mitochondria were isolated from independent primary CLL patient samples and treated with BH3-only domain peptides as indicated (100 μM) or ABT-737 or negative control enantiomer (100 nM). DMSO (1%) is a solvent control. Cytochrome c release was analyzed via ELISA. Note that activators BID and BIM BH3 peptides interact with all antiapoptotic proteins tested and furthermore can directly activate BAX and BAK (8) so that they act as positive controls for cytochrome c release assays. BADmu, a point mutant of the BAD BH3-only domain, was used as a negative control. n = 7, except for BADmu, where n = 5, and ABT-737 and negative control enantiomer, where n = 3. Error bars represent SD.
BIM bound to BCL2 primes CLL cells for killing by ABT-737. ABT-737 and the sensitizer BH3 peptides act as antagonists of antiapoptotic BCL2 but lack the ability to directly activate BAX and BAK. In order to induce MOMP, sensitizers require the presence of an activator, such as BIM or BID (8, 14, 15). The results above therefore suggest that an activator is bound to BCL2, then displaced by ABT-737 or the BCL2-binding BH3 peptides. Following displacement, we hypothesized, the freed activator could induce MOMP via interaction with BAX and BAK.
Since BIM and BCL2 were uniformly present in all CLL cell samples examined (Figure 3), we tested to determine whether BIM was indeed bound by BCL2. Using immunoprecipitation with a BCL2 antibody, we showed that BIM was consistently sequestered by BCL2 in primary CLL cells (Figure 5A). Furthermore, nearly all the cellular BIM pool in CLL cells existed in complex with BCL2 (Figure 5B). Reciprocal coimmunoprecipitation of BCL2 with BIM confirmed that much of BCL2 likewise existed in complex with BIM (Figure 5C). Similarly, in other cell types, including nonmalignant lymphocytes, other investigators have found nearly all cellular BIM sequestered by antiapoptotic proteins or at the mitochondrion (24, 25). To further investigate the sequestration of BIM by mitochondrial BCL2, we examined the subcellular localization of BIM and BCL2 in the absence of detergents. Most BIM was present in the heavy membrane fraction that contained the mitochondria, with a smaller portion in the light membrane fraction (Figure 5D). Consistent with our immunoprecipitation results (Figure 5, B and C), BCL2 was distributed in a pattern similar to that of BIM. This finding offers further support for the tonic sequestration of BIM by BCL2 in CLL cells.
BIM is sequestered by BCL2 and displaced by antagonism of BCL2. (A) Anti-BCL2 immunoprecipitation of CLL whole-cell lysates (50 μg). (B and C) CLL whole-cell lysates (50 μg) immunoprecipitated with anti-BCL2 (B) or anti-BIM (C) antibodies. Immunoprecipitation pellet (p), supernatant (s), and 100% of total lysate used for immunoprecipitation were loaded (40, 38 [B]; 38 [C]). (D) Subcellular fractions of primary CLL cells (50 μg) were immunoblotted for BIM and BCL2. Fractionation verified with marker proteins glucose 6 phosphate dehydrogenase (G6P), ER-binding protein (BiP), and mitochondrial superoxide dismutase (MnSOD). HM, heavy membrane; LM, light membrane. (E) Anti-BCL2 immunoprecipitation from 50 μg CLL lysates (Triton X-100 used for 2 samples at left, CHAPS for 2 at right). Incubation with DMSO (1%), 1 μM negative control enantiomer, or 1 μM ABT-737 displaced BIM to the supernatant identified by anti-BIM immunoblot. (F) Lysates of CLL mitochondria (15 μg) in RIPA buffer were immunoblotted for BAK, BAX, and MnSOD (loading control). DHL4 lysate (15 μg) was used as a positive control for BAK. Patient sample numbers shown at top of blots in A–F. (G) CLL cells were incubated with DMSO (0.1%), ABT-737, or negative control enantiomer (10 nM, 100 nM, or 1 μM) for 4 hours. Percentage of cells that were dead was assessed by annexin V. Oligomerization of BAX and BAK evaluated by anti-BAX or anti-BAK immunoblot of chemically cross-linked whole-cell lysates. (H) Mitochondria isolated from CLL samples were incubated with DMSO (1%) or BAD BH3 peptide (100 μM). Samples were preincubated with antibodies directed against either the human BIM BH3 domain or irrelevant antigen (CD56). n = 5. Error bars show ± SD.
To determine whether ABT-737 could indeed displace BIM from BCL2, we treated BCL2 immunoprecipitates with ABT-737 and tested for the presence of BIM protein in the supernatant. Since ABT-737 is a BH3 domain mimetic and since we have found that interactions of short BH3 peptides with BCL2 are perturbed by the presence of detergent (data not shown), we performed these experiments in detergent-free conditions. ABT-737, but not the negative control enantiomer, displaced BIM from BCL2 into the supernatant (Figure 5E). BIM has been shown to activate BAX and BAK and induce their oligomerization (8, 26, 27). If BIM displaced by ABT-737 induces death by interacting with BAX and BAK, then oligomerization of BAX and BAK should be observed following ABT-737 treatment. Examination of protein lysates from mitochondrial preparations revealed the presence of mitochondrial BAX and BAK (Figure 5F). Consistent with our model, both BAX and BAK formed oligomers within 4 hours of treatment of CLL cells with ABT-737 (Figure 5G).
Next, we attempted to independently validate the model that ABT-737 releases BIM from BCL2 and promotes BAX and BAK multimers, which lead to induction of MOMP and cytochrome c release. If BIM is indeed required for inducing MOMP following BCL2 antagonism, selective sequestration of BIM should cause a reduction in cytochrome c release following antagonism of BCL2 on CLL mitochondria. As shown in Figure 5H, we antagonized BCL2 function with the sensitizer BAD BH3 peptide. As shown previously in Figure 4, BAD BH3 by itself induced cytochrome c release. However, addition of an antibody that binds the BH3 domain of BIM caused a dramatic reduction in cytochrome c release induced by BAD BH3. An irrelevant antibody had no effect. Prevention of cytochrome c release by masking BIM supports a model in which antagonism of ABT-737 is toxic to CLL cells due to the displacement of BIM (Figure 5E) from a BCL2/BIM complex. Displaced BIM then induced BAX and BAK oligomerization (Figure 5G), MOMP, and commitment to programmed cell death. An important implication of these results is that BCL2 expression is necessary but not sufficient to dictate response to antagonism of BCL2 by ABT-737 or sensitizer BH3 peptides. Activator BH3-only proteins, such as BIM, must be sequestered by BCL2 for the cell to be sensitive to BCL2 antagonism.
BH3 profiling discriminates BCL2 versus MCL1 requirement in myeloma cells. BH3 profiling revealed a requirement for BCL2 in CLL, which is reflected at the cellular level by sensitivity to ABT-737. To determine whether BH3 profiling could predict antiapoptotic requirements in other human malignancies, we examined human multiple myeloma cell lines LP1 and L363. When mitochondria isolated from LP1 were profiled, cytochrome c release was induced by BH3 domains of BIK, NOXA, PUMA, and BMF but not BAD, HRK, and BNIP-3A (Figure 6A). This interaction pattern is inconsistent with BCL2 dependence; rather, it matches that of MCL1 (9, 14, 28). Mitochondria from L363 cells, in contrast, revealed a pattern of sensitivity to BH3 domains of BAD, PUMA, BMF, and BIK but not NOXA, HRK, or BNIP-3A (Figure 6A). As described above and shown in Figure 2, this profile suggests a BCL2 requirement. It is notable in this context that LP1 cells showed significantly greater sensitivity to antisense oligonucleotide downregulation of MCL1 than did L363 cells (29). We examined expression levels of MCL1 and BCL2 in these 2 cell lines, and found, consistent with our mitochondrial functional data, more BCL2 present in L363 than in LP1, but more MCL1 in LP1 than L363 (Figure 6B). In both cases, ample BIM was present to provide the activator function needed to signal BAX and BAK to induce MOMP.
BH3 profiling correctly predicts ABT-737 sensitivity in myeloma cell lines. (A) Mitochondria isolated from LP1 cells treated with 100 μM BH3 peptides show an MCL1-dependent pattern of cytochrome c release. Mitochondria isolated from L363 cells display a BCL2-dependent pattern. (B) Lysates (10 μg) of LP1 and L363 cell lines subjected to Western blot, comparing levels of MCL1, BCL2, and BIM. (C) L363 cells are sensitive and LP1 cells resistant to 48-hour treatment with ABT-737 and negative control enantiomer. n = 3. Error bars show ± SD. (D and E) Reduction of BIM confers resistance to ABT-737. BIM levels in L363 cells were reduced using lentiviral-delivered shRNA against exon 5 of BIM. (D) Western blot demonstrating knockdown of BIM in shRNA BIM–treated cells compared with negative control shRNA luciferase–treated (nc) cells. BCL2 levels are also shown at bottom. Quantification of protein bands by densitometry reveals 40% total BIM knockdown (white bars) with more than half of BIML and BIMS isoforms reduced, as compared with shRNA luciferase (black bars). Samples were normalized to actin. (E) L363 cells infected with lentiviruses expressing shRNA BIM are less sensitive to 48-hour treatment with ABT-737 (1 μM) than L363 cells infected with lentiviruses expressing shRNA luciferase. These data indicate that the presence of BIM complexed with BCL2 is responsible for ABT-737 sensitivity. n = 3. Error bars show ± SD.
If mitochondrial BH3 profiling predicts cellular behavior and sensitivity to ABT-737, then L363 cells, but not LP1 cells, should be sensitive to antagonism of BCL2 by ABT-737. To test this prediction, we treated cells with ABT-737 and negative control enantiomer and found that L363 cells were vastly more sensitive to ABT-737 than were LP1 cells (Figure 6C).
Both L363 and CLL cells demonstrated sensitivity to ABT-737 and expression of BIM. The sensitivity of both cell types to this compound and to peptides that antagonize BCL2 in BH3 profiling suggests the presence of at least 1 activator BH3-only protein sequestered by BCL2. To formally determine whether BIM primes these cells for death and provokes sensitivity to ABT-737, we silenced BIM expression using a lentiviral short hairpin RNA (shRNA) construct. The BIM lentiviral shRNA construct reduced BIM mRNA levels (data not shown) as well as protein levels of all 3 BIM isoforms (Figure 6D). Cells with reduced BIM levels showed a significant reduction in sensitivity to ABT-737 (Figure 6E). These data further support the critical role of BIM in conducting death signals following displacement from BCL2 by ABT-737.
These data offer promising confirmation that cellular dependence on BCL2 or any other individual antiapoptotic protein can be predicted based on mitochondrial sensitivity to our BH3 peptide panel, BH3 profiling. They furthermore suggest that myeloma cases may be discriminated and perhaps differentially targeted based on dependence on either BCL2 or MCL1.