Noxa/Bcl-2 Protein Interactions Contribute to Bortezomib Resistance in Human Lymphoid Cells*

Previous studies have suggested that the BH3 domain of the proapoptotic Bcl-2 family member Noxa only interacts with the anti-apoptotic proteins Mcl-1 and A1 but not Bcl-2. In view of the similarity of the BH3 binding domains of these anti-apoptotic proteins as well as recent evidence that studies of isolated BH3 domains can potentially underestimate the binding between full-length Bcl-2 family members, we examined the interaction of full-length human Noxa with anti-apoptotic human Bcl-2 family members. Surface plasmon resonance using bacterially expressed proteins demonstrated that Noxa binds with mean dissociation constants (KD) of 3.4 nm for Mcl-1, 70 nm for Bcl-xL, and 250 nm for wild type human Bcl-2, demonstrating selectivity but not absolute specificity of Noxa for Mcl-1. Further analysis showed that the Noxa/Bcl-2 interaction reflected binding between the Noxa BH3 domain and the Bcl-2 BH3 binding groove. Analysis of proteins expressed in vivo demonstrated that Noxa and Bcl-2 can be pulled down together from a variety of cells. Moreover, when compared with wild type Bcl-2, certain lymphoma-derived Bcl-2 mutants bound Noxa up to 20-fold more tightly in vitro, pulled down more Noxa from cells, and protected cells against killing by transfected Noxa to a greater extent. When killing by bortezomib (an agent whose cytotoxicity in Jurkat T-cell leukemia cells is dependent on Noxa) was examined, apoptosis was enhanced by the Bcl-2/Bcl-xL antagonist ABT-737 or by Bcl-2 down-regulation and diminished by Bcl-2 overexpression. Collectively, these observations not only establish the ability of Noxa and Bcl-2 to interact but also identify Bcl-2 overexpression as a potential mechanism of bortezomib resistance.

Regression of sensitive tumors such as lymphomas after effective chemotherapy is thought to reflect the induction of apoptosis (1). Chemotherapy-induced apoptosis results largely from activation of the mitochondrial or intrinsic apoptotic pathway (2,3), which is regulated by Bcl-2 family members (4 -8). This group of proteins consists of three functionally distinct subfamilies. The multidomain proapoptotic proteins Bax and Bak oligomerize upon death stimulation to induce mitochondrial outer membrane permeabilization, thereby allowing release of cytochrome c and subsequent caspase activation. Anti-apoptotic family members, including Bcl-2, Bcl-x L , Bcl-w, Mcl-1, and A1, prevent mitochondrial outer membrane permeabilization. Conversely, BH3-only proteins 3 such as Bim, Puma, and Noxa, which share only limited sequence homology with other Bcl-2 family members in a single 15-amino acid region known as the BH3 domain (9), serve as sensors of various cellular stresses and facilitate apoptosis when activated (6, 9 -14).
Although it is clear that BH3-only proteins are activated through transcriptional up-regulation or post-translational modification (7,10), the manner in which these proteins subsequently initiate apoptosis has been controversial. Two models have emerged to explain this process (15,16). The direct activation model postulates that certain BH3-only proteins, termed activators, bind and activate Bak and Bax directly, whereas the remaining BH3-only proteins, termed sensitizers, sequester anti-apoptotic Bcl-2 proteins, preventing neutralization of activators. In contrast, the indirect activation model postulates that BH3-only proteins trigger mitochondrial outer membrane permeabilization by binding and sequestering the anti-apoptotic proteins away from Bak and Bax, allowing the latter to oligomerize.
Regardless of which explanation for the activation of Bax and Bak is correct, both models indicate that binding of BH3-only proteins by anti-apoptotic Bcl-2 family members will diminish apoptosis (13,17,18). On the other hand, this interaction between BH3-only proteins and anti-apoptotic Bcl-2 family members has also been reported to exhibit important selectivity. In particular, although BH3 peptides derived from Bid, Bim, and Puma bind to all of the anti-apoptotic Bcl-2 family members, the human Noxa BH3 domain reportedly binds only Mcl-1 and A1 (16,19,20).
The importance of understanding the interactions of Noxa with other Bcl-2 family members stems from involvement of this BH3-only protein in the induction of apoptosis after exposure to certain stimuli (21)(22)(23), including the proteasome inhibitor bortezomib (24 -26). This anti-neoplastic agent is approved for the treatment of multiple myeloma (27,28) and refractory mantle cell lymphoma (29,30) and is being examined in additional lymphoid malignancies (31)(32)(33)(34)(35)(36)(37). The events leading from bortezomib-induced proteasome inhibition (38) to the unfolded protein response and subsequent cell death in highly secretory myeloma cells are well established (39,40). In contrast, there is less information about the manner in which bortezomib kills other lymphoid cells. Moreover, mechanisms of bortezomib resistance remain to be more fully elucidated.
We recently observed that interactions between Bcl-2 and Bak, previously reported to be undetectable when the affinity between isolated Bak BH3 peptide and Bcl-2 was examined, exhibited a K D of ϳ70 nM when full-length proteins were studied (41). Subsequent analysis demonstrated that the Bcl-2/Bak interaction occurred in intact cells and, because Bcl-2 is up to 40-fold more abundant than Mcl-1 in lymphoid cells, protected them from apoptosis triggered by Mcl-1 down-regulation (41). These observations prompted us to examine interactions between Bcl-2 and other potential binding partners. In the present study we set out to (i) examine the binding between Noxa and Bcl-2 proteins using surface plasmon resonance, (ii) determine whether certain lymphoma-associated Bcl-2 mutants exhibit increased binding to Noxa in vitro, (iii) assess whether Bcl-2 and Noxa interact in intact cells, and (iv) evaluate the possibility that this interaction can affect sensitivity of cells to Noxa overexpression or up-regulation.
Protein Expression and Purification-cDNAs encoding Bcl-2, Bcl-x L , and Mcl-1 lacking the transmembrane domain were cloned in-frame with glutathione S-transferase (GST) in pGEX-4T-1. cDNAs encoding Mcl-1 lacking the transmembrane domain and Noxa were cloned into pET29b(ϩ) to yield constructs containing His 6 and S peptide epitope tags. Plasmids encoding mutant or variant forms of the proteins were generated from wild type templates using a Qiagen QuikChange sitedirected mutagenesis kit and sequenced before use.
After clarified lysates containing His 6 -tagged protein were incubated with nickel-nitrilotriacetic acid-agarose (Novagen, La Jolla, CA) at 4°C for 4 h, they were applied to columns, washed with 20 volumes of PBS followed by 10 volumes PBS containing 40 mM imidazole, and eluted with PBS containing 200 mM imidazole. GST-tagged proteins were purified by incubating with glutathione-agarose overnight at 4°C. After beads were applied to a column and washed twice with 20 volumes of PBS containing 400 mM NaCl, bound protein was eluted with PBS containing 20 mM reduced glutathione.
Affinity Measurements by Surface Plasmon Resonance-Measurements were performed at 25°C on a Biacore 3000 biosensor (Biacore, Uppsala, Sweden). S peptide-His 6 -Noxa protein or the Noxa BH3 domain peptide was immobilized onto a CM5 sensorchip as instructed by the supplier. After washing with Biacore running buffer, purified anti-apoptotic proteins were injected at 30 l/min for 1 min. Bound proteins were allowed to dissociate by injection of protein-free Biacore running buffer at 30 l/min for 15 min. Residual bound proteins were desorbed with 2 M MgCl 2 . Binding kinetics were derived from sensorgrams using BIA evaluation software (Biacore).
Cell Culture-In addition to lymphoid cell lines obtained and cultured as previously described (44), cell lines were obtained from the following sources: RPMI1666 and Hs445 (Hodgkin lymphoma) were from American Type Culture Collection (Manassas, VA); Nalm6 (B-cell ALL (acute lymphocytic leukemia)) was from E. Hendrickson (University of Minnesota, Minneapolis, MN); RL (diffuse large B cell lymphoma), DoHH2 (immunoblastic lymphoma) and HT (diffuse mixed B cell lymphoma) were from T. Witzig (Mayo Clinic, Rochester, MN); MyLa (cutaneous T cell lymphoma) was from S. Ansell (Mayo Clinic, Rochester, MN). All nonadherent lines were maintained at densities below 10 6 cells/ml in RPMI 1640 medium containing 100 units/ml penicillin G, 100 g/ml streptomycin, and 2 mM glutamine as well as 15% (RPMI 1666 and Hs445) or 10% heat-inactivated fetal bovine serum (other lymphoid lines listed above). HCT116 clones stably expressing S peptide-tagged Bcl-2 or S peptide-tagged Mcl-1 were generated by transfecting parental HCT116 cells (45) with 40 g of pSPN (46) encoding Bcl-2 or Mcl-1, growing cells for 48 h in their usual medium, selecting with 800 g/ml G418, and isolating individual colonies using cloning rings.
Mammalian Expression Plasmids and Transfection-Plasmid encoding S peptide-tagged Bcl-2 was constructed by inserting cDNA containing nucleotides 4 -720 of the human Bcl-2 open reading frame (GenBank TM X06487) into the Kpn1 and EcoR1 sites of pSPN (46). Plasmid encoding S peptide/ streptavidin-binding peptide-tagged Noxa was constructed by inserting cDNA containing nucleotides 1-165 of the human Noxa open reading frame (GenBank TM NM_021127) into the BamH1 and EcoRV sites of pSPN (46) with streptavidin-binding peptide inserted into the AscI site. Plasmids encoding mutant or variant forms of the proteins were generated from wild type templates using site-directed mutagenesis.
siRNA experiments were performed using custom-designed siRNAs directed against Noxa (5Ј-GGAGAUUUGGAGA-CAAACU-3Ј), Bak (5Ј-GUACGAAGAUUCUUCAAAU-3Ј), Puma (5Ј-GCCUGUAAGAUACUGUAUA-3Ј), or Bim (5Ј-GACCGAGAAGGUAGACAAU-3Ј) from Ambion (Austin, TX). Cells were transfected with 40 M concentrations of the indicated siRNA plus 5 g of plasmid encoding EGFP-Histone H2B (to mark transfected cells) using a BTX 830 square wave electroporator delivering a single pulse at 240 mV for 10 ms. Beginning 24 h after transfection, cells were treated with diluent or bortezomib for an additional 24 h and then assayed for annexin V binding by two-color flow cytometry as described below. Control experiments using fluorescently tagged siRNA molecules indicated a Ͼ95% transfection efficiency of siRNA into Jurkat cells under these conditions.
To assess the efficacy of knockdown, cell lysates were subjected to immunoblotting or quantitative RT-PCR, which was performed in triplicate using 100 ng of RNA and TaqMan One-Step RT-PCR Master Mix (Applied Biosystems, Carlsbad, CA) according to the vendor's instructions. Using Bim (Hs00197982_m1) and Puma (Hs00248075_m1) probe sets, PCR was performed on a ABI Prism 7900HT Real_Time System using a program that consisted of 48°C for 30 min, 95°C for 10 min, then 40 cycles of 95°C for 15 s and 60°C for 1 min. Data analysis was performed using the equations ⌬C t ϭ C t (sample) Ϫ C t (endogenous control), ⌬⌬C t ϭ ⌬C t (sample) Ϫ ⌬C t (untreated), and fold change ϭ 2 Ϫ⌬⌬Ct .
Bcl-2 or Mcl-1 shRNA experiments were performed using the plasmid pCMS5A, which contains an H1P promoter for shRNA silencing, a CMV promoter for driving the expression of shRNA-resistant cDNAs, and an additional SV40 promoter for driving the expression of histone H2B fused to EGFP. This plasmid was modified to contain shRNA targeting nucleotides 526 -546 (Bcl-2 shRNA #1) or nucleotides 907-925 of the Mcl-1 open reading frame. The rescue construct contained fulllength Bcl-2 mutated to TGGATGACAGAATATTTAAC at nucleotides 526 -546 (underlined nucleotides represent silent mutations rendering cDNA resistant to Bcl-2 shRNA #1). All plasmids were sequenced to verify the integrity of inserted DNA.
For analysis of Bcl-2 knockdown by shRNA, cells were transiently transfected with the indicated plasmid, incubated for 24 h, and sorted for EGFP-histone H2B ϩ cells using a BD Biosciences FACSAria cell sorter. The 10% of cells with the highest EGFP fluorescence were collected, lysed, and assayed by immunoblotting.
Immunoprecipitation and Pulldown Assays-Log phase cells growing in antibiotic-free medium were transiently transfected with the indicated plasmid. 24 h after transfection cells were treated with 15 nM bortezomib, 500 nM MG-132 or diluent for an additional 24 h. Cells were then washed with PBS and lysed in CHAPS lysis buffer (41). After centrifugation, 500 g of lysate was incubated with S-protein-agarose beads overnight at 4°C. After sedimentation, beads were washed 4 times with lysis buffer. For immunoprecipitations, 200 g of precleared extract was incubated for 1 h with anti-Bcl-2 or IgG1 that was precoupled to protein G-agarose (47,48). After sedimentation, beads were washed four times with lysis buffer. Proteins bound to the beads were released by heating for 20 min at 65°C in SDS sample buffer. Immunoprecipitated proteins and 1 ⁄10 the amount of input cell lysate were subjected to SDS-PAGE, transferred to nitrocellulose, and probed with antibodies as indicated.
Annexin V Staining and Analysis-Cells were transiently transfected by electroporation with the indicated plasmids or siRNAs along with plasmid encoding EGFP-histone H2B to mark successfully transfected cells (41). Beginning 8-48 h later as indicated, cells were treated with the indicated bortezomib concentrations for an additional 24 h. Cells were then analyzed for apoptosis using allophycocyanin (APC)-coupled annexin V (Pharmingen) as previously described (49). After collection on a BD Biosciences FACSCanto flow cytometer, data were analyzed by gating on EGFP-histone H2B ϩ cells and assessing APC-annexin V binding using CellQuest software.

RESULTS
Noxa Binds to the Anti-apoptotic Protein Bcl-2-Studies that analyzed the ability of the Noxa BH3 domain peptide to compete for binding to different anti-apoptotic proteins (16,19,20,50) have led to current models in which Noxa is thought to bind selectively to Mcl-1 but not Bcl-2. To further evaluate these models, full-length human Noxa was purified from bacteria and assayed for the ability to bind various anti-apoptotic proteins using surface plasmon resonance. Human Noxa bound not only Mcl-1 (Fig. 1A) and Bcl-x L (supplemental Fig. S1A) but also wild type Bcl-2 ( Fig. 1B). Results of multiple binding experiments using multiple chips indicated equilibrium dissociation constants of 3.4 Ϯ 0.8 nM for Mcl-1, 70 Ϯ 40 nM for Bcl-x L , and 250 Ϯ 15 nM for wild type Bcl-2 (Variant 2, Fig. 1E).
Our earlier study examining Bcl-2/Bak interactions demonstrated that a lymphoma-derived Bcl-2 mutant displayed enhanced affinity for Bak (41). To determine whether Bcl-2 sequence variation also affected Noxa binding, naturally occurring Bcl-2 mutants characterized previously (variants 1 and 3 in Fig. 1C) as well as a fourth variant (D31H, A60V) identified in the RL lymphoma cell line (supplemental Fig. S2) 4 were assayed for their ability to bind immobilized Noxa protein. As indicated in Fig. 1, D and E, the dissociation constants varied over a 40-fold range, with the lymphoma-associated variants 1 and 4 showing a higher affinity than wild type Bcl-2. In particular, Bcl-2 variant 4 bound Noxa with a mean K D of 11 nM, which approaches that of Mcl-1.
Further experiments examined the domains of Noxa and Bcl-2 that interact. To confirm that the Noxa BH3 domain is responsible for interaction with Bcl-2, we examined the interaction of wild type Bcl-2 with a 26-mer Noxa BH3 domain peptide. This analysis demonstrated a binding affinity similar to that of full-length Noxa ( Fig. 2A). To determine whether the BH3 binding groove of Bcl-2 participates in this interaction, we assessed the impact of mutating Arg 146 of Bcl-2, a conserved residue that sits in the BH3 binding groove (Fig. 2, B and C) and is critical for binding the invariant Asp in BH3 domains (e.g. Asp 83 in Bak (51)). Mutation of Arg 146 to Ala markedly diminished the ability of Bcl-2 to bind Noxa (Bcl-2 R146A, Fig. 2D). Collectively, these results suggest that the Noxa/Bcl-2 binding shown in Fig. 1 results from an interaction between the Noxa BH3 domain and the Bcl-2 BH3 binding groove.
Noxa and Bcl-2 Bind in Vivo-Further experiments examined whether Noxa and Bcl-2 interact in intact cells. For these experiments we initially examined control and bortezomib-treated Jurkat cells. Consistent with earlier studies in other neoplastic cell lines (24 -26, 52-57), we observed increased levels of Noxa and, to a smaller extent, the lower molecular weight Bim isoforms when Jurkat cells were treated with bortezomib in the presence of the broad spectrum caspase inhibitor Q-VD-OPhe (Fig. 3A). Comparison of whole cell lysates and known amounts of purified S peptide/His 6 -tagged Noxa demonstrated that bortezomib induced a Ͼ8-fold increase in Noxa levels, from Ͻ4 ϫ 10 3 molecules/cell to 3 ϫ 10 4 molecules/cell (Fig. 3B). Assuming that mitochondria, the site of Noxa localization (58), comprise 5% of the volume of Jurkat cells (mean volume 800 m 3 when determined as in Ref. 59), this corresponds to a mitochondrial Noxa level of at least 2 M after bortezomib treatment.
To assess whether Noxa was bound to Bcl-2 during bortezomib treatment, S peptide-tagged Bcl-2 was affinity-purified after transient expression in Jurkat cells. In the absence of drug treatment, basal levels of Noxa were not detectable in Bcl-2 pulldown assays from these cells (Fig. 3C, lane 3). After bortezomib-induced Noxa up-regulation, however, binding was readily detected (Fig. 3C, lane 4). In the reciprocal experiment, endogenous Bcl-2 could also be pulled down with tagged Noxa after treatment with the proteasome inhibitor MG-132 (Fig. 3D).
Further experiments assessed the effect of Bcl-2 sequence variation on Noxa binding in vivo. When Jurkat cells were transfected with S peptide-tagged versions of all four variants, the amount of Noxa recovered in S protein pulldown assays varied despite expression of the S peptide-tagged Bcl-2 species at similar levels. In particular, variant 4 bound more Noxa than the other variants (Fig. 3E). To address the concern that these results were seen as a result of artificially tagging the proteins, endogenous Bcl-2 protein was immunoprecipitated from RL cells, which constitutively express Bcl-2 variant 4 (supplemental Fig. S2) at high levels (41) due to the presence of the t(14;18) chromosomal translocation that juxtaposes the BCL2 gene and immunoglobulin heavy chain promoter (60,61). As indicated in Fig. 3F, Noxa was detected in immunoprecipitates of endogenous Bcl-2 from this cell line even without bortezomib treatment.
To show that the Noxa/Bcl-2 interaction was not unique to bortezomib-treated cells, this interaction was also studied in HCT116 colon cancer cells after treatment with the topoi-  A, shown is surface plasmon resonance (relative units (RU)) observed when immobilized 26-mer Noxa BH3 peptide was exposed to increasing concentrations of purified wild type Bcl-2. B, a surface-filling model of human Bcl-2 (PDB 2021) using Pymol software (DeLano Scientific) shows Arg 146 at base of BH3 binding groove. C, amino acid alignment of human antiapoptotic Bcl-2 family members shows conservation of Arg in the BH3 binding groove. D, shown is surface plasmon resonance observed when immobilized Noxa was exposed to either 1 M wild type Bcl-2 (red), 1 M R146A Bcl-2 (orange), or 1 M GST-only control (blue). MAY 20, 2011 • VOLUME 286 • NUMBER 20 somerase I poison camptothecin, which up-regulates cellular Noxa levels (Fig. 3G, input lanes 2 versus 1 and 7 versus 6). Once again Noxa was detectable in pulldown assays of S peptidetagged Bcl-2 (Fig. 3G, lanes 7-10) as well as Mcl-1 (Fig. 3G, lanes 2-5) after drug treatment.

Noxa/Bcl-2 Interactions Contribute to Bortezomib Resistance
Bcl-2 Inhibits Noxa-induced Apoptosis-The observation that Bcl-2 binds Noxa raises the possibility that Bcl-2 might protect cells from Noxa-induced cytotoxicity. To assess this possibility, Jurkat cells were transfected with S peptide/streptavidin-binding peptide-tagged Noxa, which expresses at high levels relative to endogenous protein and induces apoptosis (Fig. 4, A and B). The apoptosis induced by forced Noxa overexpression was diminished by cotransfection with wild type Bcl-2 (Fig. 4B). Moreover, Bcl-2 variant 4, which binds Noxa more tightly, afforded more protection from Noxa-induced apoptosis. These results, which parallel the observed binding between Noxa and the Bcl-2 variants under cell-free conditions and in intact cells (Figs. 1E and 3E), are consistent with the possibility that the protection relied, at least in part, on the direct interaction between Noxa and Bcl-2.
Bcl-2 Inhibits Bortezomib-induced Noxa-mediated Apoptosis-Further studies examined the effects of Bcl-2 expression in bortezomib-treated Jurkat cells. Initial experiments showed that bortezomib induces apoptosis at nanomolar concentrations (Fig. 5A), providing guidance for drug dosing in subsequent experiments.
Previous studies have demonstrated that bortezomib induces up-regulation of death receptors in a number of cell types (62)(63)(64)(65)(66). In addition, FADD (FAS-associated death domain protein) and caspase 8 have been shown to play a role during bortezomib-induced apoptosis in melanoma and renal cell carcinoma cells (67)(68)(69). To assess the role of the death receptor pathway in bortezomib-induced apoptosis in Jurkat cells, we examined the bortezomib sensitivity of a series of Jurkat variants deficient in components of one apoptotic pathway or the other. In these studies the agonistic anti-Fas antibody CH.11 and camptothecin served as control stimuli that trigger apoptosis through the death receptor and mitochondrial pathways, respectively. Jurkat variants lacking FADD (I2.1 cells (70)) or caspase 8 (I9.2 cells (71)), two essential components of the death receptor pathway, were resistant to Fas-mediated death but were at least as sensitive as parental Jurkat cells to a 24-h treatment with nanomolar concentrations of bortezomib (Fig.  5B). In contrast, Jurkat cells lacking caspase 9 (JMR cells (72)) were resistant to bortezomib-induced apoptosis. Likewise, Bcl-2-overexpressing JB-6 cells (73) were resistant to bortezomib (Fig. 5B). The overall pattern of bortezomib sensitivity across these cell lines was distinct from that of CH.11 but similar to that of camptothecin. These results suggest that the mitochondrial pathway, not the death receptor pathway, plays a critical role in bortezomib-induced apoptosis in these cells.
To assess the contribution of Noxa to activation of the mitochondrial pathway, parental Jurkat cells were transfected with control siRNA or Noxa siRNA (along with plasmid encoding EGFP-histone H2B to mark successfully transfected cells), treated with bortezomib, and examined for annexin V binding. As previously reported for melanoma and myeloma cells (24, 26, 52, 53, 56), Noxa siRNA markedly diminished the ability of . At the end of the incubation, cell lysates were prepared in isotonic buffer containing 1% CHAPS and incubated with S protein-agarose beads to recover protein complexes. Pulldown assays and inputs were subjected to SDS-PAGE and immunoblotting with S peptide, Noxa, or Bcl-2 antibodies. E, lysates prepared from Jurkat cells transiently transfected with S peptide-tagged Bcl-2 variants and treated with 15 nM bortezomib for 24 h were incubated with S-protein agarose. Pulldown assays and inputs were subjected to SDS-PAGE and immunoblotting with the indicated antibodies. F, lysates prepared from untreated RL cells were immunoprecipitated (IP) with Bcl-2 or isotype control antibody (Ab). Immunoprecipitations and inputs were subjected to SDS-PAGE and immunoblotting with the indicated antibodies. G, HCT116 cells stably expressing S-peptide tagged Mcl-1 or wt Bcl-2 were treated with DMSO or 1 M camptothecin (CPT) for 48 h. Serial dilutions of lysates and pulldown assays were subjected to SDS-PAGE and immunoblotting with the indicated antibodies. Control represents untransfected, untreated HCT116 cells.

Noxa/Bcl-2 Interactions Contribute to Bortezomib Resistance
bortezomib to induce apoptosis in Jurkat cells (Fig. 5, C and D), confirming an important role for Noxa in this drug-induced cell death. The killing was, however, suppressed to an even greater extent by Bak siRNA, suggesting a possible role for other BH3only proteins as well. Because the smaller but more potent Bim isoforms (74) were also up-regulated by bortezomib (Fig. 3A), cells were also transfected with Bim siRNA or, as a control, Puma siRNA before bortezomib treatment. Bim siRNA was also able to diminish bortezomib-induced apoptosis (Fig. 5E), albeit to a much smaller extent than Noxa siRNA, suggesting that Bim plays a minor role in bortezomib-induced killing of Jurkat cells.
Further studies examined the ability of ABT-737, a BH3 mimetic that binds to Bcl-2 and Bcl-x L but not Mcl-1 (75), to enhance Noxa-induced apoptosis. In Jurkat cells and multiple other lymphoid cell lines, ABT-737 synergistically increased the cytotoxicity of bortezomib (Fig. 5, F and G, and supplemental Fig. S3). Synergy of these agents has previously been observed (40,76) and has been attributed to the ability of Noxa to down-regulate Mcl-1, thereby removing a mechanism of ABT-737 resistance (77,78). Because bortezomib did not cause Mcl-1 down-regulation in Jurkat cells (Fig. 3A), we examined the alternative possibility that ABT-737 facilitates the proapoptotic action of Noxa at least in part by neutralizing Bcl-2.  MAY 20, 2011 • VOLUME 286 • NUMBER 20

JOURNAL OF BIOLOGICAL CHEMISTRY 17687
As illustrated in Fig. 6A and summarized in Fig. 6B, shRNAmediated Bcl-2 down-regulation sensitized Jurkat cells to bortezomib, and this sensitization was reversed by expressing shRNA-resistant Bcl-2. These results suggest that endogenous Bcl-2 plays a role in protection from bortezomib. Conversely, overexpression of the Bcl-2 variants or Mcl-1 diminished bortezomib sensitivity, with the tightest binding species (variant 4 and Mcl-1, Fig. 1E) providing somewhat greater protection (Fig. 6, C and D, and supplemental Fig. S4). To further examine the potential importance of the Bcl-2/Noxa interaction in protection against bortezomib, the lymphoid lines CEM and Molt3 were also transfected with wild type Bcl-2, variant 3 (the weakest Noxa binding variant) or variant 4 (the tightest Noxa binding variant) before treatment with bortezomib. The same correlation between strength of the Noxa/Bcl-2 interaction and protection from bortezomib was also seen in these other cell lines (supplemental Fig. S5).
In further experiments we examined the effects of bortezomib in nine human lymphoid lines with varied expression of Bcl-2 and Mcl-1 (Fig. 6E). Despite the higher affinity of Mcl-1 for Noxa (Fig. 1E), there was not a clear-cut relationship between Mcl-1 expression and bortezomib sensitivity across this cell line panel. For example, RL, HT, and Hs445 cells expressed similar amounts of Mcl-1 (Fig. 6E), yet the percentage of cells undergoing apoptosis at 30 nM bortezomib varied from 55% (HT) to 10% (RL). On the other hand, bortezomib generally induced more apoptosis in lines with low Bcl-2 (e.g. CEM, MyLa, HT) than lines with high Bcl-2. Moreover, despite inducing Noxa up-regulation (Fig. 6G), bortezomib caused the least apoptosis in RL cells, which constitutively express high levels of Bcl-2 variant 4 (supplemental Fig. S2 and Fig. 6E). Combined with the results observed during forced overexpression of the Bcl-2 variants (Figs. 3E, 4B, and 6D, supplemental Figs. S4 and S5), these observations support the view that Bcl-2 is able to protect cells from bortezomib-induced apoptosis at least in part by interacting with Noxa.

DISCUSSION
In this study we demonstrate that the anti-apoptotic protein Bcl-2 and the proapoptotic BH3-only protein Noxa interact under cell-free conditions, identify the domains responsible for this interaction, and show that lymphoma-associated Bcl-2 sequence variants have up to 20-fold higher affinity for Noxa. Although previous studies have demonstrated enhanced affinity of Bcl-2 variants for Bak (41,79), these observations provide the first evidence that the enhanced affinity of some of these variants extends to a BH3-only family member as well. Additional experiments demonstrated that Bcl-2 protects Jurkat cells from the cytotoxicity of forced Noxa overexpression. Moreover, Bcl-2 down-regulation sensitizes these cells to bort-  ezomib, an agent that kills in a Noxa-dependent manner, and Bcl-2 overexpression protects from bortezomib. Collectively, these observations not only provide a more nuanced view of Bcl-2 family protein interactions but also identify Bcl-2 as a potential participant in bortezomib resistance.

Noxa/Bcl-2 Interactions Contribute to Bortezomib Resistance
Earlier studies examining the interactions of Noxa with anti-apoptotic Bcl-2 family members yielded inconsistent results. In particular, several previous studies examining the interaction of anti-apoptotic proteins with synthetic BH3 domain peptides reported nanomolar affinities of the Noxa BH3 peptide for Mcl-1 and A1 but little or no affinity for Bcl-2 and Bcl-x L (16,19,20). In contrast, the original description of Noxa indicated that it bound to Bcl-2 and Bcl-x L in pulldown assays (58). In addition, more recent studies found that Bcl-x L protected against Noxa-mediated death (54,80), although the nature and strength of the Bclx L /Noxa interaction was not reported.
Our experiments demonstrated that wild type Bcl-2 binds full-length Noxa with a K D of 250 Ϯ 15 nM, which is ϳ3-fold higher than the K D of Noxa and Bcl-x L and 70-fold higher than the K D of Noxa and Mcl-1 (Fig. 1). These observations confirm the previously reported high affinity of Noxa for Mcl-1 but also raise the possibility that Bcl-2, which is far more abundant than Mcl-1 in some lymphoid cells (41), might also bind Noxa under certain conditions. Additional analysis indicated that Bcl-2 bound the Noxa BH3 peptide and full-length Noxa with similar affinities (cf. Figs. 1B and 2C). In contrast, Bcl-2 R146A, which lacks a conserved arginine that binds the invariant aspartate in BH3 domains (51), exhibited markedly diminished Noxa binding (Fig. 2D). Collectively, these results suggest that the BH3 domain of Noxa interacts with the BH3 binding groove of Bcl-2.
Given these results, it is interesting to speculate why the Bcl-2/Noxa interaction might have been missed in earlier experiments. One possible explanation might be that some of the experiments utilized Bcl-2 variant 3, which exhibits lower affinity than wild type Bcl-2 for Noxa (Fig. 1E). In addition, several of the earlier experiments were performed as competition assays in which Bcl-2 was prebound to a competitor BH3 peptide and then injected onto a sensor chip containing immobilized Bim BH3 peptide and analyzed for its ability to dissociate from the "competitor" onto Bim. The ability of Noxa BH3 peptide to diminish the much stronger Bcl-2/Bim interaction is critically dependent on the ability of Noxa to remain bound to Bcl-2. Our experiments, which were performed by directly assessing the interaction between Bcl-2 and Noxa protein or BH3 peptide, revealed that Noxa bound rapidly to Bcl-2 but also dissociated rapidly (Figs. 1 and 2A). Under nonequilibrium conditions encountered during injection into a Biacore apparatus, this rapid dissociation could have made the interaction between Bcl-2 and the Noxa BH3 peptide appear weaker than it actually is, providing a second potential explanation for earlier results.
Despite the much higher affinity of Mcl-1 for Noxa, binding of Noxa to Bcl-2 is potentially important because of the much greater abundance of Bcl-2 in certain lymphoid cell lines (41). Mcl-1 can be viewed as a high affinity, low capacity neutralizer of Noxa. As Noxa levels increase after proteasome inhibition or other treatments and the binding capacity of Mcl-1 is saturated, Bcl-2 and other anti-apoptotic Bcl-2 family members can provide a lower affinity but higher capacity for neutralizing Noxa. Consistent with this view, we observed binding of Noxa to wild type Bcl-2 after bortezomib treatment (Fig. 3, C and D). This binding appears to be even more important when certain naturally occurring Bcl-2 mutants are present.
Earlier studies demonstrated that Bcl-2 is mutated in some lymphoma cell lines and clinical samples bearing the t(14;18) chromosomal translocation (81,82). We previously reported that one such Bcl-2 variant exhibited a higher affinity than the wild type protein for Bak (41). In the present study we have extended these findings by identifying another lymphoma-associated Bcl-2 mutant (supplemental Fig. S2) and showing that this variant exhibits increased Noxa binding under cell-free conditions (Fig. 1D) and in immunoprecipitates (Fig. 3E). Indeed, the affinity of this mutant for Noxa is ϳ20-fold greater than wild type Bcl-2 (Fig. 1E). This finding provides an explanation for the binding of Noxa to this Bcl-2 variant even in the absence of bortezomib treatment (Fig. 3E).
To assess the potential biological consequences of the Bcl-2/ Noxa interaction, we initially focused on apoptosis induced by ectopic Noxa expression. Bcl-2 variant 4, which has an affinity for Noxa approaching that of Mcl-1 (Fig. 1E), afforded the cells more protection than wild type Bcl-2 ( Fig. 4 and supplemental Fig. S4). The fact that this protection parallels the ability of the variants to bind Noxa in a cellular context (Fig. 3E) raises the possibility that the direct interaction of Noxa with Bcl-2 might contribute to the protection.
While the present study was being revised, Du et al. (83) reported that Noxa appears under certain conditions to function as a direct activator. In contrast, Ren et al. (84) concluded that Noxa is only a sensitizer. The present results do not distinguish between these models. If Noxa is acting as a sensitizer, then a tighter-binding Bcl-2 variant would bind more of the overexpressed Noxa, leaving more Bcl-x L and Mcl-1 (higher affinity neutralizers of Bak (41)) available to bind any Bak and other proapoptotic Bcl-2 family members that become activated. On the other hand, if Noxa were acting as a direct activator, then a tighter-binding Bcl-2 variant would sequester more Noxa, diminishing the amount of Noxa available to interact with and activate Bax and Bak. Accordingly, tight binding Bcl-2 variants would be expected to protect better whether Noxa acts as a sensitizer or as a direct activator.
Further functional studies focused on the cytotoxicity of the proteasome inhibitor bortezomib. Our results demonstrated that Noxa siRNA protected Jurkat cells from bortezomib-induced apoptosis more effectively than down-regulation of other BH3-only proteins, establishing this as a largely Noxa-dependent cell death in this model system. Additional experiments demonstrated that Bcl-2 shRNA enhanced bortezomib-induced cell death (Fig. 6, A and B), suggesting a role for endogenous Bcl-2 in regulating Noxamediated cytotoxicity. Conversely, Bcl-2 overexpression protected these cells against bortezomib (Fig. 6, C and D). These results coupled with the observation that the tightest binding Bcl-2 variant protects the best against Noxa (Fig. 4) or bortezomib (Fig. 6D, supplemental Figs. S4 and S5) are consistent with a model in which Noxa is neutralized in part by Bcl-2 as well as Mcl-1.
On the other hand, the Bcl-2 family is involved in a complicated network of protein-protein interactions. Because perturbation of any particular element in this network can affect drug sensitivity, we cannot rule out the possibility that part of the protection by Bcl-2 is independent of the direct Noxa/Bcl-2 interaction. For example, Certo et al. (19) have reported that sensitivity of a different group of human lymphoid cell lines to a number of cytotoxic agents correlates with the amount of Bim that is bound to Bcl-2 before drug exposure. The current explanation for this "primed for death" state is that BH3-only family members induced or activated by the drugs displace Bim from Bcl-2, allowing the Bim to act as a direct activator. If this were the explanation for the bortezomib-induced apoptosis observed in our experiments, however, then Bim down-regulation would be expected to protect as effectively as Noxa or Bak down-regulation. Contrary to this prediction, Bim siRNA afforded Jurkat cells only modest protection against bortezomib-induced apoptosis despite down-regulation of Bim message and protein by ϳ80% (Fig. 5E). 5 These observations suggest that Noxa does more than simply displacing Bim from anti-apoptotic Bcl-2 family members during bortezomib-induced apoptosis. Conversely, the observation that Bcl-2 protects Jurkat cells from bortezomib-induced apoptosis despite the relatively minor role of Bim in this process again raises the possibility that the direct Bcl-2/Noxa interaction might play a role in this protection.
Interest in bortezomib resistance stems in part from the use of bortezomib to treat various neoplasms in the clinic. This agent has been approved for the treatment of mantle cell lymphoma (29,30) and is undergoing extensive testing in other lymphoma subtypes (31)(32)(33)(34)(35)(36)(37). The ability of Bcl-2 to protect from bortezomib-induced apoptosis provides a potential explanation for the recent observation that bortezomib is unable to induce remissions in follicular lymphomas (36), which universally express Bcl-2 at high levels because of the t(14;18) translocation (85,86). Accordingly, our observations not only demonstrate that Bcl-2 and Noxa can interact under certain conditions but also provide a biological context for understanding the potential importance of these interactions.