The Polycomb Group Protein Bmi-1 Is Essential for the Growth of Multiple Myeloma Cells

Bmi-1 was initially discovered through its ability to cooperate with c-Myc in the induction of lymphoma (1, 2) and was subsequently recognized as a member of the Polycomb group (PcG) family of proteins (3). The PcG proteins function within distinct multisubunit complexes and epigenetically regulate gene expression by altering chromatin states at specific promoters (reviewed in Schwartz and colleagues in ref. 4). In several contexts, including lymphoma, mouse embryonic fibroblasts, and human fibroblasts, Bmi-1 exerts its oncogenic function, at least in part, by silencing the INK4a/ARF (Cdkn2a) tumor suppressor locus encoding the p16 (INK4a) and p19 (ADP ribosylation factor, ARF) proteins (5–7). The proposed role for Bmi-1 as a key regulator of cell growth control/senescence mechanisms led to additional investigations about the role of Bmi-1 in tumorigenesis. Several studies have shown that Bmi-1 is overexpressed in medulloblastomas (8), lung (9), breast (10), prostate cancers (11), and multiple myeloma (MM; ref. 12–14), among others. Extensive microarray analyses of multiple cancer types also indicate that a distinct Bmi-1–driven gene expression signature is a predictor of metastasis and poor survival (15, 16). A functional contribution of Bmi-1 in tumorigenesis has been reported in several cancers (17–20), yet the precise mechanism(s) by which it promotes cell growth, especially in cases that are independent of INK4a/ARF suppression, have yet to be elucidated (19, 21). Mouse models of Bmi-1 deficiency have also established the importance of Bmi-1 in the self-renewal of neural and hematopoietic stem cells (8, 22–24). In concordance with its role in stem cells, Bmi-1 has been proposed to maintain cancer stem cell populations in leukemia (22), as well as in breast (25) and lung cancers (17).

Although analysis of published gene expression profiling data reveals an increase in Bmi-1 expression in MM patient samples and cell lines compared with normal plasma cells in three independent studies (12–14), the precise role of Bmi-1 in the growth of MM cells remains unknown and is therefore the focus of this study. In addition, the contribution of epigenetic mechanisms toward the pathogenesis of MM is not well understood. MM is a genetically heterogeneous and complex malignancy characterized by the abnormal proliferation of antibody-producing plasma cells (reviewed in Seidl and colleagues in ref. 26). Although current therapies including proteasome inhibitors, thalidomide, and lenalidomide have met with some success, the disease is still incurable (27).

In this study, we investigate whether Bmi-1 is necessary for the growth of MM cells both in vitro and in vivo. We not only delineate a contribution of this PcG protein in MM cell growth but also determine a novel mechanism by which Bmi-1 promotes survival. Transcriptional profiling studies in the RPMI-8226 myeloma cell line followed by functional rescue experiments showed that the proapoptotic gene Bim is negatively regulated by Bmi-1, and that this regulation is required for the Bmi-1–mediated inhibition of apoptosis in these cells. Overall, these studies highlight the importance of epigenetic mechanisms, as contributed by Bmi-1 function, toward the survival of MM cells.

RPMI-8226, NCI-H929, and U266 myeloma cells were obtained from the American Type Culture Collection. KMS12BM, LP-1, OPM-2, and EJM lines were obtained from the German Resource Centre for Biological Material. All cell lines were maintained in advanced RPMI-8226 medium (Invitrogen) with 1% fetal bovine serum, and 1% l-Glutamine.

Control short hairpin RNA (shRNA): GTGGACTCTTGAAAGTACTAT, Bmi-1 shRNA-1: CCAGACCACTACTGAATATAA (TRCN0000020155), and Bmi-1 shRNA-2: CCTAATACTTTCCAGATTGAT (TRCN000 0020156) were cloned into the inducible pLKO-Tet-On puromycin vector as previously described (28). The Bim shRNA sequence (29) was cloned into the pLKO-Tet-On neomycin vector.

Lentiviral supernatants were generated according to our previously established protocol (28) and concentrated by ultracentrifugation followed by infection of MM cells as described in the Supplementary Methods.

Snap-frozen cell pellets of primary CD138⁺ plasma cells isolated from the bone marrow of healthy donors were purchased from Stem Cell Technologies. Cells were harvested in lysis buffer (Cell Signaling). Anti–Bmi-1 Clone F6 was from Millipore; anti–glyceraldehyde-3-phosphate dehydrogenase was from Imgenex; anti-TetR was from Boca Scientific; anti–β-actin was from Sigma; anti-cleaved poly ADP ribose polymerase (PARP) and anti–caspase-3 were from Cell Signaling Technology; and anti-Bim was from Thermoscientific.

Total RNA was isolated using the RNeasy Mini kit (Qiagen) or the TurboCapture 96 mRNA isolation kit (Qiagen). RNA samples were processed for quantitative reverse transcription-PCR (qRT-PCR) as previously described (30).

Two thousand to 8,000 cells per well were seeded in a 48-well plate in methycellulose medium (HSC-001; Stem Cell Technologies). A stock of 1.44% methylcellulose was made in medium and diluted to a final concentration of 1.17% by the addition of cell suspension. Colonies were stained with Hoechst 33342 after 14 days and visualized using the ChemGenius imaging system. Colony number and area were analyzed using the AlphaEase FC program [AlphaInnotech (Cell Biosciences)]. Cell cycle analysis was performed using the CycleTEST PLUS DNA Reagent kit (Becton Dickinson), whereas apoptotic cells were detected using the Annexin V-FITC/7-AAD kit (BD Biosciences).

RNA was isolated using the Qiagen RNeasy mini kit (Qiagen). Generation of labeled cDNA and hybridization to HG-U133 Plus2 arrays (Affymetrix) were performed as previously described (20). Expression values were normalized on a per-sample basis using the Affymetrix MAS5 algorithm (31). Microarray expression data have been deposited in Gene Expression Omnibus (accession number GSE 21912).

Female nonobese diabetic severe combined immunodeficient mice (Charles River Laboratories) were allowed to acclimate in the Novartis Institutes for BioMedical Research animal facility with ad libitum access to food and water for at least 3 days before manipulation. Animals were handled in accordance with the Novartis Animal Care and Use Committee protocols and regulations.

Stably infected RPMI-8226 cells were free of Mycoplasma and viral contamination (comprehensive mitogen-activated protein panel, RADIL, University of Missouri). Mice (ages 6–8 wk) were inoculated s.c. with 8 × 10⁶ cells in 50% Matrigel (BD Biosciences) in the right dorsal axillary region. Tumor dimensions were obtained using calipers, and tumor volume was calculated as (length × width²)/2. Two to three weeks after implantation, animals were randomized and sorted into treatment groups (n = 10) based on tumor volume (mean = 140 mm³). Animals received vehicle (5% sucrose) or doxycycline (Sigma Aldrich; 1 mg/mL in 5% sucrose) via drinking water for the duration of the study (27 d).

Tumor samples were fixed in 10% neutral-buffered formalin for 6 hours and processed using the Ventana Discovery System (CC1S_ R60_NH _OR) and as described in Supplementary Methods.

RPMI-8226 cells were treated with or without doxycycline for 48 hours, and cross-linking/chromatin immunoprecipitation (ChIP) was performed as described in Supplementary Methods. Antibodies used include the Rabbit IgG control (SC-2027 Santa Cruz Biotech) and a Bmi-1 polyclonal antibody generously provided by Drs. Caroline Woo and Robert Kingston (Massachusetts General Hospital, Boston, MA). DNA obtained after ChIP was analyzed through qPCR using SYBR green with oligos specific for the Bim promoter, and enrichments were calculated by normalizing to input DNA (see Supplementary Methods for primer sequences and calculation methods).

All data are shown as mean values ± SD and tested statistically using the two-tailed Student's t test. P ≤ 0.05 was considered to indicate statistical significance. In vivo tumor measurement data are shown as mean ± SEM and was tested statistically using the Kruskal Wallis One way ANOVA and the post hoc Dunn's method for pairwise comparisons. P < 0.05 was considered to indicate statistical significance.

We first examined the status of Bmi-1 protein expression in a panel of MM cell lines. Whole-cell extracts were subjected to immunoblot detection of Bmi-1, and the presence of Bmi-1 protein was verified in all MM cell lines that were tested. (Fig. 1A). Bmi-1 protein was also detected at higher levels in MM cell lines than in the normal donor derived CD138⁺ plasma cells that were analyzed (Fig. 1A). Interestingly, Bmi-1 migrated with distinct mobilities in certain cell lines (compare KMS12BM to RPMI-8226), raising the possibility that there is differential posttranslational modification of Bmi-1. To ensure that such species represent Bmi-1, immunoblots were also prepared in cell lines with and without Bmi-1 shRNA–mediated knockdown, and indeed, these protein species were each attenuated upon Bmi-1 knockdown (Supplementary Fig. S1C and E; Fig. 1B). To determine whether Bmi-1 protein was required for MM cell growth, we infected RPMI-8226, KMS12BM, and LP1 cells with a robust inducible single-lentiviral vector knockdown reagent pLKO-Tet-On (28) containing either control nontargeting shRNA or two distinct Bmi-1–targeting shRNA sequences (Bmi-1 shRNA-1 or Bmi-1 shRNA-2), and polyclonal stable lines were obtained after puromycin selection. Doxycycline induction of Bmi-1 shRNAs but not the control nontargeting shRNA led to the significant reduction of Bmi-1 mRNA (Supplementary Fig. S1A) and protein (Fig. 1B), which was more pronounced with Bmi-1 shRNA-1 than with Bmi-1 shRNA-2. To determine the effect of Bmi-1 knockdown on MM cell growth and viability, stable lines were grown in the presence or absence of doxycycline and the number of viable cells was determined using an MTS-assay. Doxycycline induction of both Bmi-1 shRNAs in RPMI-8226 cells correlated with >60% inhibition of cell growth over a time course of 7 days (Fig. 1C). Bmi-1 knockdown in KMS12BM and LP1 cells also led to the inhibition of cell proliferation but was less extensive and slightly delayed compared with RPMI-8226 cells (Supplementary Fig. S2A and B). To determine whether Bmi-1 was required for the clonogenic growth of MM cells, cells were plated in methylcellulose in the presence or absence of doxycycline. In this case, we observed that Bmi-1 knockdown led to the inhibition of colony formation by causing a reduction in colony number rather than size in all the three myeloma lines, RPMI-8226, KMS12BM (Fig. 2; Supplementary Fig. S3), and LP1 (Supplementary Fig. S4). Therefore, taken together, these results strongly suggest that Bmi-1 plays a necessary role in the in vitro growth of MM cells.

In several cellular contexts, Bmi-1 has been shown to promote cell cycle progression, and counteract senescence and apoptosis by silencing the expression of p16^INK4a and p19^ARF tumor suppressors (5, 6). To determine the mechanism by which Bmi-1 regulates the growth of MM cells, we stained cellular DNA with propidium iodide, allowing examination of the various phases of the cell cycle by flow cytometry during the loss of Bmi-1 function. Doxycycline induction of RPMI-8226 cells containing Bmi-1 shRNA led to a prominent appearance of a sub-G ₁ DNA content peak indicative of a population that has undergone cell death (Fig. 3A). The proportion of dead cells continued to increase through day 6 (Fig. 3B) and day 9 after doxycycline induction by which time >60% of cells showed sub-G₁ DNA content (Supplementary Fig. S5). To further confirm whether RPMI-8226 cells with Bmi-1 depletion undergo apoptotic cell death, cells were stained with FITC-labeled Annexin V. Induction of Bmi-1 shRNAs led to a statistically significant (P < 0.005) 15% to 30% increase in cells exclusively staining with Annexin V after 3 days of doxycycline treatment (Fig. 3C; Supplementary Fig. S12), indicating that this population is undergoing the early stages of apoptosis. Bmi-1 depletion in KMS12BM and LP1 cells did not significantly alter their cell cycle profile but yielded the detection of a 20% and 7% sub-G₁ DNA content, respectively, suggesting that only a proportion of these cells undergo cell death (Supplementary Fig. S6). Therefore, the induction of cell death partially explains the pronounced inhibition of colony formation observed in KMS12BM and LP1 cells upon Bmi-1 knockdown and leaves the possibility of additional mechanisms such as cellular senescence. Overall, these data show that although the apoptotic response upon Bmi-1 depletion varies among MM cell lines suggestive of multiple Bmi-1–driven growth mechanisms, they clearly define a function for Bmi-1 in protecting RPMI-8226 myeloma cells from apoptosis.

To determine whether Bmi-1 is also necessary for MM cell growth in vivo, we generated subcutaneous MM xenografts in nonobese diabetic severe combined immunodeficient mice using the inducible control shRNA and Bmi-1 shRNA-RPMI-8226 cells. To confirm that the doxycycline treatment led to inhibition of Bmi-1 expression, we analyzed a separate group of animals 4 days after doxycycline treatment. Doxycycline induction of Bmi-1 shRNAs led to 96% inhibition of Bmi-1 mRNA, as measured by qRT-PCR (Fig. 4A), and depletion of Bmi-1 protein (Fig. 4B). Knockdown of Bmi-1 was stable and persisted throughout the 27-day treatment period (Supplementary Fig. S7). In contrast to vehicle treatments and doxycycline-treated control shRNA tumors, doxycycline-treated animals bearing Bmi-1 shRNA RPMI-8226 xenografts showed a marked inhibition of overall tumor growth (Fig. 4C). Furthermore, in keeping with the notion that Bmi-1 might function as a prosurvival factor, we observed that Bmi-1–depleted tumors began to regress at day 5 of doxycycline treatment, displaying an average 70% regression by day 23. In agreement with the apoptotic phenotype observed upon Bmi-1 knockdown in vitro, immunohistochemical staining revealed a marked increase in cleaved caspase-3 in doxycycline-treated Bmi-1 shRNA cells compared with vehicle-treated animals (Fig. 4D). Quantitative analysis of the immunohistochemical data using a pixel-counting algorithm¹

Emerging evidence indicates that Bmi-1 also functions in an INK4a/ARF-independent manner in the normal as well as in the tumorigenic context, yet the precise targets involved have not been elucidated (19, 21). Bmi-1 knockdown neither resulted in the restoration of p16 nor an increase in p14 (human ARF) mRNAs in all three myeloma lines (RPMI-8226, LP1, and KMS12BM) tested (data not shown). Given the profound antiapoptotic effect of Bmi-1 in RPMI-8226 cells, we took advantage of our inducible Bmi-1 shRNA system to dissect the signaling that occurs downstream of Bmi-1 in these cells. HG-U133 Plus2 Affymetrix gene chips were used to analyze global changes in gene expression resulting from Bmi-1 knockdown. Because a significant reduction of Bmi-1 protein was detected beginning 24 hours after doxycycline treatment (Fig. 5A), transcriptional profiling was performed during doxycycline-mediated Bmi-1 shRNA induction at 24, 32, and 48 hours collecting data from biological triplicates at each time point.

To minimize the effect of doxycycline treatment alone or of the control shRNA vector alone in the data analysis, we grouped all the 33 “control” samples (biological triplicates of Bmi-1 shRNA without doxycycline; control shRNA with and without doxycycline at 24, 32, and 48 h, and Bmi-1 shRNA and control shRNA before doxycycline treatment). The 9 Bmi-1 shRNA + doxycycline samples (24, 32, and 48 h, in biological triplicates) were grouped together and were compared through the two-tailed Student's t test to the control sample group described above. Table 1 shows genes that were differentially regulated in response to Bmi-1 knockdown. False discovery rates were calculated using the Benjamini-Hochberg step-down method (32). Fold changes were modest, ranging between 1.5 and 2 for genes most consistently upregulated, and 1.5 to 6 for genes consistently repressed, but were highly significant. The two-tailed student t test and false discovery rate calculation was also performed for each time point of doxycycline induction (Supplementary Table S1).

Interestingly, several genes that were found to be upregulated upon Bmi-1 depletion, such as Bcl2l11 (Bim), P115, GAS2, and FAM26B, have been linked to apoptosis (33–36). Among these genes, Bim has well-defined roles in the execution of apoptosis (33, 37). Our analysis also revealed genes that were repressed upon Bmi-1 knockdown, suggesting that Bmi-1 may function directly or indirectly in maintaining the expression of such genes. Indeed, the most highly repressed gene in the doxycycline-treated Bmi-1 shRNA group was Bmi-1, therefore validating a crucial factor in this experiment. In agreement with the induction of proapoptotic genes upon Bmi-1 silencing, repression of the antiapoptotic gene, apoptosis inhibitor-5/AAC-11(API5; Supplementary Fig. S9D; Table 1; ref. 38) was observed. The changes in gene expression for several genes from the microarray analysis, including Bim, API-5, and MAPK6, were reconfirmed in the original RNA samples by parallel qRT-PCR analysis (Supplementary Fig. S9). In summary, global transcriptional profiling led to the identification of several potential downstream mediators of Bmi-1 function in myeloma cells, most notably a group of apoptosis regulators.

The finding that the expression of the proapoptotic gene Bim is upregulated upon Bmi-1 knockdown in RPMI-8226 cells is in accordance with the prominent apoptotic phenotype that is detected in these cells and reveals a novel potential target downstream of Bmi-1. We further investigated this regulation in vitro and in vivo. In a similar time course experiment to that performed for the transcriptional profiling studies, a gradual increase in Bim mRNA was detected from 16 to 72 hours postdoxycycline induction of Bmi-1 shRNA (Fig. 5A), thereby corroborating the results from the microarray experiment. A maximal 1.6-fold increase was detected 32 hours after doxycycline induction, a time point at which Bmi-1 protein is significantly reduced (Fig. 5A). Therefore, these data indicate a correlation between the loss of Bmi-1 protein and an increase in Bim mRNA transcript levels. Immunoblots of doxycycline-treated Bmi-1 shRNA-RPMI-8226 cells also indicated an increase in Bim protein, in particular the Bim L and Bim s isoforms, upon Bmi-1 knockdown (Fig. 5B). Although the changes in Bim mRNA and protein seem modest, previous studies have shown that minor changes in Bim expression are sufficient to confer functional effects (39–41). The regulation of Bim was also observed in vivo, wherein Bim protein increased 4 days after doxycycline treatment of Bmi-1 shRNA-RPMI-8226–transplanted animals (Fig. 5C). Taken together, these data suggest that Bmi-1 can regulate Bim both in vitro and in vivo. We next determined whether a potential mechanism by which Bmi-1 controls Bim expression is through the regulation at the Bim promoter. We performed ChIP studies in RPMI-8226 cells using a Bmi-1–specific antibody previously used for ChIP (42). Although detection of Bmi-1 binding at the Bim promoter varied through the 2,500-bp region upstream of the transcriptional start site (TSS), it was highest at ∼700 bp upstream from the TSS as determined by quantitative PCR (Fig. 5D). Importantly, cells that were treated with doxycycline to induce Bmi-1 shRNAs failed to show enrichment at such loci within the Bim promoter (Fig. 5D), confirming that the observed interaction was dependent on Bmi-1. Therefore, these studies suggest that Bmi-1 regulates Bim expression through interaction at the Bim promoter.

To determine the functional significance of Bim upregulation upon Bmi-1 knockdown, we generated RPMI-8226 cells containing inducible expression of both Bmi-1 and Bim shRNAs. We hypothesized that if suppression of Bim plays an important role for Bmi-1–imposed cell survival, the depletion of Bim should rescue cells from undergoing apoptosis upon Bmi-1 knockdown. Doxycycline induction of Bim shRNA led to a decrease in Bim mRNA as measured by qRT-PCR (Fig. 6A). Importantly, doxycycline induction of Bim shRNA was sufficient to block the increase in Bim mRNA levels that is otherwise detected upon Bmi-1 knockdown alone (Fig. 6A). Bmi-1 was efficiently depleted in doxycycline-induced Bmi-1shRNA and Bmi-1shRNA/Bim shRNA cells (Fig. 6B). Furthermore, doxycycline induction of control shRNA/Bim shRNA and Bmi-1shRNA/Bim shRNA cells resulted in a significant decrease of Bim protein (Fig. 6B). Additionally, although cleaved PARP, a well-established marker of apoptosis, was upregulated upon Bmi-1 knockdown, this increase was not detected in the presence of simultaneous Bmi-1 and Bim knockdown (Fig. 6B). Therefore, these results show that Bim knockdown is sufficient to negatively affect known markers of apoptosis (Fig. 6B). In agreement with the attenuation of cleaved PARP observed upon dual Bmi-1/Bim knockdown and in sharp contrast to the inhibition of cell growth observed upon Bmi1 knockdown, simultaneous knockdown of Bim and Bmi-1 resulted in a phenotypic rescue with ∼80% of the cells surviving Bmi-1 depletion as long as 8 days after doxycycline induction (Fig. 6C). We confirmed the rescue observed upon dual Bmi-1/Bim knockdown in clonogenic assays (Supplementary Fig. S10) and Annexin V apoptosis assays (Supplementary Fig. S11) with both Bmi-1 shRNAs. Quantitation of colony formation (Supplementary Fig. S10B) and Annexin V staining (Supplementary Fig. S12) confirmed statistically significant differences (P < 0.05) between induction of Bmi-1 shRNA alone compared with induction of both Bmi-1 and Bim shRNAs. These results show that the regulation of Bim by Bmi-1 is an important mechanism by which Bmi-1 inhibits apoptosis in RPMI-8226 cells.

Although MM represents a malignancy in which several genetic and tumor microenvironment factors have been proposed to promote the tumorigenic phenotype (43, 44), the contribution of epigenetic mechanisms toward this disease is not clearly understood. The PcG proteins, of which Bmi-1 is a critical member, serve evolutionarily conserved functions in epigenetic gene regulation during development, with emerging evidence implicating their role in tumorigenesis (45). For example, in keeping with its ability to accelerate lymphoma in E-μ-Myc mice, Bmi-1 overexpression can also cause lymphoma in the E-μ-Bmi-1 transgenic model (1, 2, 46). In addition, Bmi-1 can immortalize mouse embryonic fibroblasts as well as cooperate with oncogenic RAS to transform fibroblasts and mammary epithelial cells (6, 47, 48). Here, we show, using an inducible shRNA system, that Bmi-1 is essential for the maintenance of the transformed phenotype of myeloma cells both in vitro and in vivo. As a core constituent of the Polycomb-repressive complex 1, Bmi-1 stimulates the Histone H2A ubiquitin ligase activity of Polycomb-repressive complex 1, thereby maintaining silencing that is initiated by the Polycomb-repressive 2 complex (reviewed in ref. 45). In future studies, it will be interesting to interrogate the relative contribution of Bmi-1 versus the other PcG proteins in MM as well as the mechanisms leading to their constitutive expression. Several possibilities exist since previous studies have shown that c-MYC can induce Bmi-1 expression (49), and c-MYC is highly expressed in MM lines including the ones analyzed in this study.

A well-established and extensively studied mechanism by which Bmi-1 promotes tumorigenesis is through the transcriptional silencing of the p16 and p19 tumor suppressor genes (5, 6). Whereas p16 inhibits cell cycle progression by disrupting the cyclinD/CDK4/6 complex, p19 inhibits Mdm2 function resulting in p53 stabilization and subsequent apoptosis. In particular, by inhibiting p19, Bmi-1 is able to overcome c-Myc induction of p19 and apoptosis (5). Therefore, the tumor-promoting function of Bmi-1 has been proposed to be in part due to the inhibition of apoptosis. In our myeloma system, knockdown of Bmi-1 resulted in the potent induction of apoptosis in RPMI-8226 cells, as well as the increased production of cleaved caspase-3 in the corresponding xenograft model. However, we did not observe any changes in p16 or p14 (human ARF) mRNA, suggesting INK4a/ARF-independent Bmi-1 function(s). We therefore undertook a global transcriptional profiling approach to elucidate novel genes that are regulated by Bmi-1. Our analysis uncovered a major proapoptotic protein, Bim, to be upregulated upon Bmi-1 knockdown. The repression of Bim expression by Bmi-1 seems to occur through a mechanism that involves Bmi-1 localization at the Bim promoter as indicated by our ChIP studies. Strikingly, dual knockdown of Bmi-1 and Bim caused an almost complete rescue of cells from undergoing apoptosis. To date, this is the first study that functionally links Bmi-1 to the regulation of the proapoptotic gene Bim. Therefore, our results establish that the repression of Bim by Bmi-1 plays an important role toward the survival of MM cells, as provided by the RPMI-8226 cell line model. Given the considerable genetic heterogeneity within MM, our data suggest that Bmi-1 promotes MM growth through potentially distinct mechanisms because we did not observe a strong apoptotic phenotype upon Bmi-1 knockdown in KMS12BM or LP1 cells. These findings are partly in agreement with reports that LP1 cells lack expression of Bim (50). In future studies, the mechanisms of Bmi-1–driven growth in such cell lines will need to be elucidated.

Our findings about the suppression of Bim downstream of Bmi-1 are intriguing in light of several studies that implicate INK4a/ARF-independent mechanisms by which Bmi-1 affects tumorigenesis (19, 21). In addition, even in Bmi-1 knockout models, INK4a/ARF deficiency does not confer a complete rescue against the senescent and apoptotic phenotypes observed, again suggesting the presence of additional Bmi-1 targets (5, 6). Indeed, genome-wide studies in human embryonic fibroblasts have revealed that the PcG proteins regulate the expression of various genes involved in differentiation and stem cell maintenance, some of which belong to the Wnt and Notch signaling pathways (51). As such pathways are deregulated in several cancers including MM, whether these targets are similarly regulated by the PcG proteins in the tumorigenic context and specifically in MM will be of interest to investigate in future studies.

In conclusion, we present evidence for the involvement of Bmi-1 function in MM as well as characterize a novel pathway by which Bmi-1 functions in inhibiting apoptosis. An understanding of Bmi-1–regulated pathways in normal as well as in the tumorigenic context will be crucial toward evaluating it as a potential target in anticancer therapy.

Z. Jagani, D. Wiederschain, A. Loo, D. He, R. Mosher, P. Fordijour, J. Monahan, M. Morrissey, Y-M. Yao, M. Warmuth, W.R. Sellers, and M. Dorsch are employees of Novartis Institutes for BioMedical Research. C. Lengauer was a previous employee of Novartis Institutes for BioMedical Research, and is currently an employee of sanofi-aventis.

We thank Frank Stegmeier, Raymond Pagliarini, Ramesh Shivdasani, and William Matsui for critical advice; Kristy Chung for the assistance; and Shilpa Kadam, Felix Lohmann, David Ciccone, and Gilles Buchwalter for the advice on ChIP experiments.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.