Distinct molecular mechanisms responsible for bortezomib-induced death of therapy-resistant versus -sensitive B-NHL cells

NonHodgkin lymphoma (NHL) is a heterogeneous group of neoplasms with distinct natural histories, clinical features, responsiveness to therapy, and prognosis. Rituximab, a chimeric anti-CD20 monoclonal antibody, has changed the treatment paradigm for patients with B-cell nonHodgkin lymphoma (B-NHL). The integration of rituximab into several standard chemotherapy regimens has been shown to be superior to systemic chemotherapy alone in several randomized phase III clinical trials in various subtypes of lymphoma.¹  The addition of rituximab (R) to standard doses of cyclophosphamide, doxorubicin, vincristine, and prednisone or fludarabine based-regimens has resulted in improved treatment outcomes in diffuse large B-cell and indolent B-cell lymphomas^2-5  in recent years.

Despite the improvement in the outcome of NHL patients treated with R-chemotherapy, a significant number of patients with diffuse large B-cell lymphoma and the majority of patients with indolent B-cell lymphomas relapse after treatment. The mechanisms by which lymphoma cells acquire resistance to rituximab and/or chemotherapy agents are multifactorial and can be intrinsic to the cancer cell or host.⁶  Using a rituximab-resistance preclinical model characterized by our group, we previously demonstrated the existence of shared, cancer cell–intrinsic pathways of resistance to rituximab and conventional chemotherapy.^7,8  Recent data in patients with diffuse large B-cell lymphoma undergoing salvage chemotherapy suggest that lymphomas that are resistant or relapse after upfront R-cyclophosphamide, doxorubicin, vincristine, and prednisone are indeed more resistant to subsequent treatment and further support the findings of our preclinical model.^9,10 

In 2 separate analyses, the level of expression of pro- versus antiapoptotic members of the B-cell lymphoma 2 (Bcl-2) family proteins proved prognostic for B-NHL patients treated with rituximab or chemotherapy.^11,12  In our preclinical model of rituximab/chemotherapy resistance. we demonstrated that deregulation of the expression of pro- and antiapoptotic proteins is associated with acquired resistance to rituximab.⁸  Similar deregulation of Bcl-2 family proteins was reported in independently derived rituximab/chemotherapy-resistant cells,¹³  further validating the importance of the Bcl-2 family of proteins in therapy-resistant B-NHL.

Mounting evidence suggests that many Bcl-2 family proteins are targeted for proteasomal degradation in malignant cells.¹⁴  Inhibition of proteasomes may, therefore, lead to the selective induction and/or activation of Bcl-2 family proteins resulting in modulation of the apoptotic potential of malignant cells. Bortezomib (PS-341, Velcade) is a peptide boronic acid inhibitor of the 26S proteasome that binds to and inhibits the chymotrypsin-like catalytic domain of the 20S proteasome core.¹⁵  Hematological malignancies of the B-lineage appear especially sensitive to the antitumor activity of bortezomib, potentially due to their constitutive production of large amounts of immunoglobulin and enhanced sensitivity to a terminal unfolded protein response.^16,17  In accord with this observation, bortezomib demonstrated clinical activity against and was approved by the United States Food and Drug Administration to treat relapsed or refractory multiple myeloma.¹⁸  Subsequently, bortezomib was Food and Drug Administration–approved for the treatment of relapsed/refractory mantle cell lymphoma¹⁹  and has shown activity against several other types of B-NHL in phase II trials.^20-22 

The mechanisms by which bortezomib induces cell death has yet to be fully elucidated. Bortezomib was thought to work by inhibiting nuclear factor-κB (NF-κB) activity via stabilization of IκB.²³  In a B-NHL model system similar to the one used here, Jazirehi et al¹³  demonstrated that bortezomib could sensitize resistant B-NHL cells to chemotherapy by inhibiting NF-κB activity. In multiple cell types, including B-NHL, multiple myeloma, and solid tumor cells, proteasome inhibitors are capable of killing malignant cells via induction of the proapoptotic Bcl-2 family proteins Noxa or Bik.^24-27  While each proposed mechanism of bortezomib action is likely valid, the only global inference that can be taken from the study of bortezomib's mechanism-of-action is that it can work via several distinct pathways that are dictated by the intrinsic nature of the tumor cells exposed to this agent.

In the current study, we used a previously described cell line model of rituximab/chemotherapy resistance⁷  to explore the therapeutic potential of bortezomib in therapy-resistant B-NHL. Data indicate that in our cell line model, and in a subset of primary B-NHL patient-derived tumor cell samples, bortezomib can induce apoptosis via stabilization of Bak. Surprisingly, death of therapy-resistant cells did not depend on Bak or caspase-dependent apoptosis. Knock-down of Bak or pharmacological inhibition of caspases was unable to inhibit their bortezomib-induced death, suggesting the activation of alternative cell death pathways (ie, necrosis or cell cycle arrest). The data presented here demonstrate the complexity of response to bortezomib treatment by demonstrating that within a cell population (ie, our resistant cells), bortezomib is capable of inducing apoptosis via Bak stabilization and/or caspase-independent cell death. Several caspase-independent mechanisms of cell death have been described and are currently being investigated as potential mediators of antitumor effects observed after exposure of B-cell lymphomas to bortezomib. Our data suggest that bortezomib may be useful in treating B-NHL patients who have failed previous therapy regimens because it can promote death through a variety of mechanisms distinct from other antineoplastic agents.

Therapy-sensitive and -resistant cell lines were described previously.⁷  SUDHL-6 cells were obtained from Dr Shawn Murphy (University of Rochester, Rochester, NY) and multiple myeloma cells were obtained from Dr Kelvin Lee (Roswell Park Cancer Institute, Buffalo, NY).

Antibodies raised against Bak and Actin were obtained from Sigma Chemicals, Bik from Santa Cruz Biotechnology, Moesin (antibody-1) from Lab Vision, poly(adenosine diphosphate-ribose) polymerase (PARP)–1 from BD Pharmigen, and Noxa from Calbiochem. Propidium iodide and 3-methyladenine were obtained from Sigma Chemicals, zVAD-fmk and Q-VD-OPh from MBL International, and DiIC₁ (1,1′-dimethyl-3,3,3′,3′-tetramethylindodicarbocyanine)⁵  from Invitrogen.

Primary malignant B-cells were isolated by magnetic-activated cell sorting (negative selection; Miltenyi Biotec) from pretreatment biopsy tissue obtained from B-NHL patients treated at Roswell Park Cancer Institute under an institutional review board–approved protocol (N = 18). Cells were seeded into 384-well plates at a density of 0.5 × 10⁶/mL and treated in triplicate with a several doses of a panel of therapeutic agents under investigation in our laboratory including rituximab (10 μg/mL), cisplatinum (CDDP; 0.1μM, 1μM, and 10μM), adriamycin (0.1μM, 1μM, and 10μM) and bortezomib (1nM, 5nM, and 10nM) with or without the pan-caspase inhibitors Q-VD-OPh (2.5μM; MBL International). Twenty-four and 48 hours after treatment, viability was determined using CellTiter-Glo (Promega). Viability was determined using the following formula: viability = relative light unit from treated cells/relative light unit from control cell × 100%.

Western blots were performed as previously described.⁸ 

After overnight exposure to bortezomib, proteasome activity was determined as previously described.²⁸ 

Cells were incubated at a concentration of 0.5 × 10⁶ cells/mL in complete media containing bortezomib or CDDP with or without chemical inhibitors (see figure legends for specific doses). At 24-, 48-, and 72-hour time points, aliquots were removed and stained with Alexa Fluor488-conjugated annexin V (Invitrogen), 50nM DiIC₁,⁵  and 5 μg/mL propidium iodide in annexin-binding buffer (10mM HEPES [4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid]), 150mM potassium chloride, 1mM magnesium chloride, 1.3mM calcium chloride, 1 mg/mL glucose, 0.5% bovine serum albumin, pH 7.4). In addition, caspase-3/7 activity in equal numbers of treated versus untreated cells was detected in triplicate or quadruplicate wells using the luminometric Caspse-Glo 3/7 assay (Promega) following the manufacturer's protocols. Statistical differences were compared using unpaired Student t tests.

Expression constructs for Bak, Bik, and Noxa were created on the pIRES2-EGFP (Invitrogen) backbone using a previously described strategy.⁸  The coding regions of BAK1 or BIK were polymerase chain reaction (PCR) amplified from Raji cell cDNA, while PMAIP1 was PCR amplified from a Noxa expression construct kindly provided by Dr David Huang (Walter and Eliza Hall Institute, Melbourne, Australia). FLAG-tagged Bak constructs were generated by overlap-extension PCR. Resulting plasmids (2 μg) were transfected into resistant cells using an Amaxa Nucleofector following the manufacturer's protocol for Raji cells (Amaxa). Both mock transfected and empty pIRES2-EGFP vector transfected cells were used as negative controls.

Lysates were prepared from 2 × 10⁷ cells treated with 25nM bortezomib (PS-341) or vehicle control (dimethyl sulfoxide [DMSO]) for 24 hours. Protein concentration was determined by bicinchoninic acid and normalized between samples. Lysates were then precleared using protein G sepharose beads. Anti-FLAG M2 monoclonal or anti-Bak polyclonal antibody was added and allowed to incubate overnight at 4°C with gentle agitation. Protein G sepharose beads were used to remove antibody and bound protein from solution, washed 4 times, and boiled in sodium dodecyl sulfate loading buffer. Beads were then removed by centrifugation, and eluted protein was separated on 10% sodium dodecyl sulfate polyacrylamide gel electrophoresis gels followed by Western blot analysis.

Knock-down of Bak, Bik, and Noxa was achieved using ON-TARGETplus SMARTpool siRNA (Dharmacon) designed to specifically target BAK1, BIK, or PMAIP1, respectively. Nontargeting SMARTpool siRNA (Dharmacon) was used as a negative control. Two hours following transfection of 2 μg of siRNA using an Amaxa Nucleofector, cells were incubated with bortezomib (20nM) for an additional 72 hours.

We previously described a cell line model of therapy-resistant B-NHL in which apoptosis was not induced by standard chemotherapeutic agents⁸  or rituximab.⁷  This model has become the basis for investigation of novel agents that may be used to treat therapy-resistant B-NHL. Bortezomib was found to induce the death of resistant cell lines Raji 2R, Raji 4RH, and RL 4RH, along with sensitive, parental cell lines Raji and RL (Figure 1A). compared with sensitive Raji or RL cells treated with the same dose (25nM) for the same duration (72 hours), bortezomib induced the death of a similar (P < .05) percentage of therapy-resistant cells. Parallel treatment of these cells with cisplatinum (100μM) was used to confirm the inability of standard chemotherapeutic agents to kill resistant cells (Figure 1A). To further understand the mechanism underlying bortezomib-induced cell death of sensitive and resistant cells, we examined the ability of either bortezomib or CDDP to induce the activation of caspase-3/7, a hallmark of apoptosis. We found that while CDDP was only capable of inducing caspase-3/7 activity in sensitive cells, bortezomib induced caspase-3/7 activity in sensitive and resistant cells, albeit to a much greater extent in sensitive cells (Figure 1B). These data suggested that bortezomib may induce apoptosis in cells that are resistant to apoptotic cell death. During the initial characterization of bortezomib-induced cell death in sensitive versus resistant cells, we found that sensitive cells died more rapidly over the first 48 hours of incubation (data not shown). This delay in cell death correlated with a delay in bortezomib-induced PARP cleavage (Figure 1C). The delay in apoptosis induction of resistant cells was further confirmed by flow cytometric analysis, which indicated that there was an approximately 24-hour delay in phosphatidylserine exposure, DNA fragmentation, and cell death of bortezomib-treated resistant cells compared with bortezomib-treated sensitive cells (Figure 1D).

Several members of the Bcl-2 family of apoptosis regulatory proteins are known to be altered by treatment with proteasome inhibitors.¹⁴  We previously showed that the lack of chemotherapy-induced apoptosis in our resistant cells was primarily due to mutation or posttranscriptional down-regulation of the proapoptotic Bcl-2 family proteins Bax and Bak, respectively.⁸  We therefore hypothesized that bortezomib may overcome the block in apoptosis in resistant cells by altering the ratio of Bcl-2 family proteins. In vitro exposure of sensitive and resistant B-NHL cells to bortezomib led to no visible change in several members of the Bcl-2 family of proteins, including Bcl-2, Bcl-xL, Bax, Bim, and Puma (data not shown). Induced myeloid leukemia cell differentiation protein (Mcl-1) in bortezomib-treated sensitive and resistant cells was found to be stabilized and cleaved in a caspase-dependent manner (data not shown) as was previously described.^29,30  Most interestingly, however, 3 proapoptotic Bcl-2 family proteins (Bak, Bik, and Noxa) were induced in resistant cells following bortezomib exposure, while their expression remained constant in bortezomib-treated sensitive cells (Figure 2A). Moreover, quantitative analysis of the kinetics of Bak, Bik, and Noxa induction compared with PARP cleavage demonstrated clear induction of all 3 proapoptotic proteins before PARP cleavage in resistant cells (Figure 2B). These data suggest that increased expression of Bak and/or Bik and/or Noxa precede and may be necessary for induction of apoptosis in bortezomib-treated resistant cells.

To further study the role of Bak, Bik, and Noxa in the induction of apoptosis of resistant cells, we transfected resistant cells with pIRES2-EGFP containing either a BAK1, BIK, or PMAIP1 (Noxa) transgene. We found that expression of Bak, but not Bik or Noxa, in resistant cells led to a significant increase in caspase-3/7 activity and subsequent cell death (Figure 2C and data not shown). Western blots and flow cytometry were performed to verify the specific expression of enhanced green fluorescent protein and each transgene in transfected resistant cells (Figure 2D and data not shown). To illustrate the specificity of Bak induction of caspase-3/7 activity and cell death, an inactive, mutant form of Bak³¹  was transfected into resistant cells and was found to induce less caspase-3/7 activity than wild-type Bak (Figure 2E). These data demonstrate that induction of Bak is sufficient to induce apoptosis in resistant cells and support a role for increased Bak expression in bortezomib-induced death of resistant cells.

Data obtained from our B-NHL cell line model established 2 response patterns to in vitro bortezomib exposure: (1) sensitive cells die rapidly and do not show evidence of increased Bak expression following bortezomib exposure, and (2) resistant cells undergo slower cell death that is preceded by increased Bak, Bik, and Noxa expression following bortezomib exposure. When we examined several freshly isolated primary cultures of B-NHL cells from patients for their response to bortezomib, we found a similar pattern. Of the 10 primary isolates that yielded sufficient cells to perform cytotoxicity measurements and Western blots, 3 displayed a visible change in Bak expression following bortezomib exposure (Figure 3A), while the other 7 showed no evidence of increased Bak expression upon ex vivo bortezomib treatment (Figure 3B). Increased Bik expression was observed in the same 3 samples where increased Bak expression was seen while all 10 samples demonstrated increased Noxa expression upon ex vivo bortezomib exposure (data not shown). When the difference between Bak expression in vehicle-treated versus bortezomib-treated primary lymphoma cells was quantified, the 3 with visible Bak induction had greater than 1.5-fold changes in Bak (ΔBak ≥ 1.5), while the 7 with no visible changes in Bak expression had ΔBak values of < 1.5. We then separated the 10 samples into 2 groups based on their ΔBak values with 1.5 as our cutoff. When we compared the bortezomib-induced cytotoxicity between the 2 groups we found that, similar to our cell line model, primary cells in which Bak expression increased upon bortezomib exposure died slower over the 48-hour observation time than those with no change in Bak expression upon bortezomib exposure (Figure 3C). Sensitivity to rituximab-induced CMC (Figure 3D) or adriamycin treatment (Figure 3E) did not correlate with bortezomib sensitivity or changes in Bak expression following bortezomib. These data support findings from our cell line model and further suggest that Bak induction may precede and be necessary for bortezomib-induced cell death in those cells where Bak expression is controlled by the proteasome.

The first clear evidence that Bak was being regulated by proteasomes in resistant cells came from the observation that dose-dependent proteasome inhibition correlated with dose-dependent Bak induction in bortezomib-treated cells (Figure 4A). Treatment of cells with bortezomib is known to induce cell stress, particularly to the endoplasmic reticulum (ER),¹⁶  which could lead to induction of Bak. To address this possibility, we treated resistant cells with the ER-specific stress inducers brefeldin A, tunicamycin, or thapsigargin and compared Bak induction to proteasome inhibitor–treated cells. We found that Bak expression was induced by the proteasome inhibitors bortezomib, MG132, epoxomycin, and Ac-Leu-Leu-Nle (Figure 4B and data not shown). ER stressors and stresses induced by chemotherapeutic agents such as adriamycin and cisplatin did not increase Bak expression (Figure 4B). To determine whether the observed induction of Bak in bortezomib-treated resistant cells was dependent on transcription, we blocked transcription using actinomycin D and found that Bak protein expression still increased upon bortezomib exposure (data not shown). In support of the transcriptional independence of Bak induction, no change in BAK1 mRNA expression was detected by quantitative real-time PCR over the first 24 hours following bortezomib exposure of resistant cells. Bak protein expression examined in the same pool of cells increased approximately 2.5-fold over the same 24-hour time course, suggesting that the increase in Bak protein was independent of increased transcription of the BAK1 gene (Figure 4C). Because we found that proteasome inhibition in resistant cells led to increased Bak expression without increased BAK1 transcription, it seemed likely that Bak protein was stabilized by proteasome inhibition. We tested this idea using ³⁵S-methionine pulse-chase experiments in untreated versus bortezomib-treated cells. We found that Bak stability was reduced in resistant cells compared with sensitive cells and that treatment of resistant cells with bortezomib increased Bak stability to a level comparable with untreated sensitive cells (Figure 4D). Overall, these data suggest that decreased steady-state Bak expression observed in resistant cells is due to increased proteasomal degradation of Bak and that inhibition of proteasomes leads to increased Bak stability and expression in resistant cells.

The relative increase in proteasomal degradation of Bak observed in resistant cells could result from an increase in proteasome activity or selective targeting of Bak to the proteasome via increased ubiquitination. We previously demonstrated that resistant cells have a modest increase in total proteasome activity.⁷  However, this relatively small change in overall activity of proteasomes in resistant cells could not adequately explain the dramatic reduction of Bak protein expression. When FLAG-tagged Bak was immunoprecipitated from sensitive versus resistant cells and Bak Western blots were performed, we observed a distinct banding pattern in immunoprecipitates from resistant cells that was reminiscent of ubiquitination. In immunoprecipitates from sensitive cells, only a single band at the approximate molecular weight of Bak was observed (Figure 4E). We therefore examined the effect of bortezomib treatment on the ubiquitination of Bak in sensitive versus resistant cells. Immunoprecipitation of endogenous Bak followed by Western blot for Ubiquitin and Bak was carried out on lysates from sensitive and resistant cells treated with bortezomib. We observed a substantial increase in ubiquitinated Bak in bortezomib-treated resistant cells compared with bortezomib-treated sensitive cells and a banding pattern of ubiquitinated Bak similar to that observed in FLAG-Bak IPs (Figure 4F). To insure the increase in Bak ubiquitination in resistant cells was specific, immunoprecipitation of ubiquitin followed by Western blot for ubiquitin was preformed from the same lysates. As expected, bortezomib treatment increased the expression of ubiquitinated proteins in both sensitive and resistant cells to a similar extent (Figure 4F). These data suggest that Bak is specifically ubiquitinated and targeted for proteasomal degradation in resistant cells and helps to explain the reduced basal level of Bak observed in these cells.

Inhibition of proteasome activity in resistant cells leads to increased expression of Bak, Bik, and Noxa and ultimately apoptotic cell death. To determine the contribution of Bak, Bik, or Noxa to bortezomib-induced apoptosis of resistant cells, siRNA specific for each were transfected 2 hours before incubation of cells with bortezomib, and their effect on bortezomib-induced apoptosis was determined. Western blot analysis of whole cell lysates from siRNA-transfected, bortezomib-treated resistant cells was used to confirm specific knock-down of each gene (Figure 5A). We found that siRNA-mediated inhibition of bortezomib-induced expression of Bak, but not Bik or Noxa, significantly (P < .01) reduced the ability of bortezomib to activate caspase-3/7 and induce DNA fragmentation (Figure 5B and data not shown). These data suggest that Bak is a critical mediator of bortezomib-induced apoptosis and that its stabilization in resistant cells allows for induction of apoptosis in cells that otherwise lack apoptotic potential. However, siRNA-mediated knockdown of Bak had no significant effect on the ability of bortezomib to kill resistant cells (Figure 5C). This suggests that bortezomib might be capable of activating additional Bak and caspase-independent death pathways in resistant cells.

To test the ability of bortezomib to kill cells in the absence of caspase activity, we incubated sensitive or resistant cells in bortezomib with or without the pan-caspase inhibitors zVAD-fmk or Q-VD-OPh. We found that in sensitive cells, caspase inhibition repressed bortezomib-induced cell death while in resistant cells caspase inhibition had no effect on or increased bortezomib-induced cell death (Figure 6A). Western blot was used to confirm that the doses of zVAD-fmk and Q-VD-OPh used inhibited caspase activity as measured by PARP cleavage (Figure 6B). These data suggest that in addition to triggering an apoptotic phenotype in resistant B-NHL cells, bortezomib is capable of killing them in a caspase-independent manner.

The ability of cell lines to activate “back-up” caspase-independent modes of cell death when apoptosis is inhibited has been extensively studied.³²  In our cell line model, we have found that bortezomib is capable of inducing caspase-independent death (Figure 6A). To determine whether this phenomenon is specific to our B-NHL cells that were selected for therapy resistance, we tested 18 primary tumor specimens obtained from patients with follicular lymphoma (Figure 7A), small lymphocytic lymphoma (Figure 7B), and diffuse large B-cell lymphoma of germinal center (Figure 7C) and non–B-cell lymphoma of germinal center (Figure 7D) phenotypes to determine whether caspase-independent cell death occurred following bortezomib exposure. Bortezomib was capable of reducing the viability of ex vivo treated B-NHL cells by greater than 20% in 15 of 18 samples tested. Of these 15 patient samples that were sensitive to bortezomib, addition of the caspase inhibitor Q-VD-Oph significantly inhibited bortezomib-dependent loss of viability in 5 cases (Figure 7 open triangles). No specific pattern was observed regarding histology, subtype, or disease status (de novo vs relapsed/refractory). These data support our cell line findings and demonstrate that bortezomib is capable of killing B-NHL cells in a caspase-dependent or -independent manner.

Proteasome inhibition by bortezomib can alter the expression of several pro- and antiapoptotic Bcl-2 family proteins in a variety of cell types and via distinct mechanisms.¹⁴  Bortezomib treatment of rituximab/chemotherapy resistant B-NHL cell lines developed by Dr Benjamin Bonavida's group led to down-regulation of Bcl-xL and sensitization to chemotherapy in an NF-κB–dependent manner.¹³  Here we report that bortezomib is capable of killing therapy resistant B-NHL cell lines in the absence of additional chemotherapeutic agents. Unpublished observations demonstrated that inhibition of NF-κB activity did not kill our therapy-resistant cell lines, even at high doses. Therefore, inhibition of NF-κB by bortezomib is not sufficient to kill therapy-resistant B-NHL cells, further suggesting that bortezomib must be acting on additional factors to promote cell death in our model of resistance.

Activation of BH3-only members of the Bcl-2 family of proteins is critical for initiation of the apoptotic cascade. BH3-only proteins are tightly regulated in healthy cells and can be activated transcriptionally or by posttranslational modification and/or stabilization. After bortezomib treatment, we found that 2 BH3-only proteins, Bik and Noxa, were induced in resistant, but not sensitive, cells. Previous studies in mantle cell lymphoma and multiple myeloma cell lines have demonstrated Noxa is constitutively degraded by proteasomes and is therefore induced by treatment with proteasome inhibitors.^24,25  Similarly, Bik stabilization has been observed in cell lines derived from solid tumors treated with bortezomib.^27,33  Recent evidence suggests that because of their limited binding specificities, expression of Noxa or Bik alone should not be sufficient to induce cell death in cells that express Bcl-xL and Mcl-1.³⁴  Noxa should inhibit the binding of Mcl-1, but not Bcl-xL, to Bak while Bik should inhibit Bcl-xL, but not Mcl-1, binding to Bak and/or Bax. Simultaneous induction of Bik and Noxa in resistant cells (Figure 2) could potentially lead to apoptosis by blocking the inhibitory binding of both Mcl-1 and Bcl-xL to Bak, the only multidomain proapoptotic Bcl-2 family protein expressed in resistant cells. We were, however, unable to demonstrate a role for either Bik or Noxa in bortezomib-induced apoptosis (Figure 5). Additional experiments were conducted in which Bik and Noxa were simultaneously expressed or knocked down, and still no effect on apoptosis was observed (data not shown). These data suggest that while the expression of Bik and Noxa are clearly enhanced by proteasome inhibition, in the absence of adequate levels of Bak, this effect does not make a significant contribution to bortezomib-induced apoptosis in our resistant cells. Additional factors (perhaps other BH3-only proteins or stabilization of adequate levels of Bak) must be activated or induced in bortezomib-treated cells to promote Bak-dependent apoptosis.

Regardless of how bortezomib activates Bak- and/or Bax-dependent apoptosis, either Bax or Bak must be expressed in order for a cell to undergo apoptosis.³⁵  In resistant cells, however, Bax is not expressed, and the expression of Bak is reduced. Reintroduction of either Bax or Bak led to apoptosis of resistant cells.⁸  Similarly, bortezomib treatment of resistant cells led to increased Bak expression and Bak-dependent apoptosis. Our data suggest that Bak is ubiquitinated and targeted for proteasomal degradation in resistant, but not sensitive, cells (Figure 4). Furthermore, we were able to observe Bak stabilization upon bortezomib treatment of additional B-NHL cell lines that were not selected for resistance (data not shown) and in isolated, primary B-NHL cells treated ex vivo with bortezomib (Figure 3). These observations suggest that increased proteasomal degradation of Bak may be a novel pathway that could contribute to therapy resistance in B-NHL.

The observation that bortezomib is capable of inducing apoptosis in cells that are otherwise resistant to apoptosis is important in our understanding of how these cells became therapy resistant. Identification of the molecular mechanisms underlying Bak degradation (eg, specific ubiquitin ligase[s]) could help guide the logical use of bortezomib to resensitize resistant B-NHL cells to apoptosis. However, identification of Bak as a major contributor to bortezomib-induced apoptosis of resistant cells did not complete our understanding of how bortezomib killed these cells. Bak induction, while necessary for caspase activation, was not required for bortezomib-induced death of resistant cells (Figure 5C), likely because bortezomib is capable of killing resistant cells in a caspase-independent manner (Figure 6A). Caspase-independent cell death is considered to be a secondary form of cell death that is only activated when apoptotic, caspase-dependent cell death is inhibited. Our data are consistent with this hypothesis yet further suggest that bortezomib may induce a form of caspase-independent death that is capable of killing some B-NHL cells (ie, our resistant cells) better than others (ie, our sensitive cells). To confirm and advance these data, we tested an additional 18 primary tumor cell specimens derived from patients with various subtypes of B-NHL and found that bortezomib was able to induce both caspase-dependent (n = 5) and -independent (n = 10) cell death. The caspase dependence of bortezomib-induced death did not correlate with histology or disease status (de novo vs relapsed/refractory). This could be explained by the heterogeneity within disease subtypes studied or could be a result of the limited number of samples currently available. Further study will be required to uncover correlations between the caspase dependence of cell death induced by bortezomib and clinical/molecular features of B-NHL.

Several caspase-independent, nonapoptotic modes of cell death have been described.³⁶  Autophagy had been linked to proteasome activity,³⁷  and its activation by proteasome inhibition made the most intuitive sense because the ubiquitin-proteasome system and autophagy are parallel protein degradation pathways that can compensate for one another. On the other hand, irreversible cell cycle arrest and/or necrosis had been observed following proteasome inhibition in preclinical cancer models^38,39  Our data suggest that under normal circumstances apoptosis is the predominant mechanism by which bortezomib kills cells.

In addition, further studies to enhance our understanding of caspase-independent death pathways induced by bortezomib are ongoing in our laboratory. Caspase-independent cell death observed in resistant B-NHL cell lines and primary B-NHL cells may have clinical implications for the treatment of resistant B-NHL that have severe defects in their ability to undergo apoptosis. If caspase-independent pathway(s) are identified, bortezomib-based combination therapies could be designed to bypass the requirement for B-NHL cells to die by apoptosis and instead engage caspase-independent death pathways. Additional preclinical studies will be required to further understand bortezomib-induced cell death and to determine optimal treatment combinations that could benefit patients with resistant B-NHL.

The authors thank Ping-Chaio Tsai, Elizabeth Repasky, PhD, Paul Wallace, PhD, and John Byrd, MD, for constructive criticism of this work.

This work was supported by National Institutes of Health R01 grant CA136907-01A1 awarded to Roswell Park Cancer Institute, Buffalo, NY; National Institutes of Health P01 grant CA103985-1 awarded to Garden State Cancer Center, Belleville, NJ; and the Eugene and Connie Corasanti Lymphoma Research Fund.

National Institutes of Health

Contribution: S.H.O., J.B., A.B.-V., W.R., and C.M. performed experiments; J.G. provided surgical specimens; S.H.O., F.J.H.-I., and M.S.C. analyzed data and made the figures; and S.H.O., N.B., J.L.C., F.J.H.-I., and M.S.C. designed the research and wrote the paper.

Conflict-of-interest disclosure: M.S.C. has served on an advisory board and has received clinical research support from Millennium Pharmaceuticals. The remaining authors declare no competing financial interests.

Correspondence: Myron S. Czuczman, MD, Chief, Lymphoma/Myeloma Section, Roswell Park Cancer Institute, Elm and Carlton Streets, Buffalo, NY 14263; e-mail: myron.czuczman@roswellpark.org.