RAG-induced DNA lesions activate proapoptotic BIM to suppress lymphomagenesis in p53-deficient mice

Neoplastic transformation, the process leading from a normal cell state into a malignant tumor, is estimated to require at least five to seven oncogenic mutations (Stratton et al., 2009). These oncogenic mutations facilitate unrestrained cell proliferation, evasion from cell death, and escape from immune destruction (Hanahan and Weinberg, 2011). Defects that compromise the ability of cells to detect and repair DNA lesions increase the risk of acquiring oncogenic lesions and thereby predispose to cancer development. Lymphocyte differentiation requires DNA rearrangement at the antigen receptor (TCR and Ig) gene loci and this is catalyzed by the endonucleases RAG1 and RAG2 (Fugmann et al., 2000). Notably, aberrant RAG1/2 activity can promote tumor development by giving rise to chromosomal translocations that activate oncogenes (e.g., Myc-Ig: t[8;14] or Bcl-2-Ig: t[14;18]; Marculescu et al., 2002).

To confront the threat posed by acquisition of tumor-promoting mutations, elaborate mechanisms have evolved to ensure that DNA lesions are repaired in a timely manner or, alternatively, that the damaged cell is directed to undergo senescence or cell death to prevent expansion of a nascent neoplastic clone (Lord and Ashworth, 2012). Although apoptosis is recognized as a powerful means to suppress tumor development (Delbridge et al., 2012), it is surprising that no mechanisms have yet been identified by which potentially oncogenic DNA lesions trigger apoptosis to cause extinction of nascent neoplastic cells.

The BCL-2 regulated apoptotic pathway is critical for the initiation of apoptosis in response to many cytotoxic insults, including DNA damage (Czabotar et al., 2014). DNA damage causes activation of p53, which initiates apoptosis through direct transcriptional induction of the proapoptotic BH3-only proteins PUMA and (to a lesser extent) NOXA (Oda et al., 2000; Nakano and Vousden, 2001; Jeffers et al., 2003; Villunger et al., 2003; Yu et al., 2003). However, it has recently been demonstrated that p53 does not suppress tumor development solely through activation of apoptosis (Brady et al., 2011; Li et al., 2012; Valente et al., 2013). In striking contrast to p53-deficient mice, which all succumb to lymphoma (or, more rarely, sarcoma) by ∼280 d (Donehower et al., 1992; Jacks et al., 1994), mice lacking all three p53 target genes that are critical for its ability to induce apoptosis (Puma and Noxa) and cell cycle arrest, as well as cell senescence (p21), did not develop tumors (Valente et al., 2013). It has therefore been proposed that loss of p53-initiated DNA repair processes underpins the high cancer susceptibility in p53-deficient mice (Valente et al., 2013).

Because DNA damage elicited by chemotherapeutic drugs can induce apoptosis in the absence of p53 (Strasser et al., 1994), it is conceivable that potentially oncogenic DNA lesions (e.g., chromosomal translocations) may also be able to induce apoptosis in a p53-independent manner. It appears likely that BH3-only proteins (activated by DNA lesions in a p53-independent manner) may play a role in such a process (Huang and Strasser, 2000). Accordingly, one would expect that loss of such BH3-only proteins would accelerate tumor development in p53-deficient mice. We report that loss of BIM, but not loss of BMF, collaborates potently with loss of p53 in lymphomagenesis. Interestingly, this acceleration in lymphoma development exerted by loss of BIM could be prevented by concomitant loss of RAG1. These results reveal that BIM-dependent apoptosis is a critical tumor suppressive process activated by RAG1/2-induced DNA lesions.

BIM and BMF are proapoptotic BH3-only members of the BCL-2 family and both have been implicated in DNA damage–induced apoptosis. Because they can both be activated by DNA damage in a p53-independent manner (Bouillet et al., 1999; Labi et al., 2008; Kelly et al., 2010), we examined the impact of their loss on tumor development in p53-deficient mice. We thereby sought to determine whether loss of apoptosis induced by either BIM or BMF would synergize with loss of p53-initiated DNA repair signaling (and potentially loss of other p53-activated tumor suppressive processes), resulting in accelerated tumor development. Accordingly, mice deficient for both p53 and BMF or p53 and BIM were generated and aged to determine their rates of tumor development.

Loss of BMF had no discernible impact on the incidence or rate of tumor development, the type of cancer that developed, and the severity of malignant disease in both p53^+/− and p53^−/− mice (Fig. 1 and not depicted). In contrast, loss of BIM provoked a substantial acceleration in the already rapid rate of tumor development that is caused by complete loss of p53 (Fig. 2 A; P < 0.0001). Strikingly, loss of even a single Bim allele was sufficient to accelerate tumor development in p53^−/− mice (Fig. 2 A; P = 0.028). Loss of BIM also significantly augmented lymphoma burden in the thymus of sick mice (Fig. 2 B), the primary site of tumor growth in p53^−/− mice (Donehower et al., 1992; Jacks et al., 1994).

Analysis of aged Bim^−/− mice revealed evidence of neoplastic growth, consistent with a previous study (Erlacher et al., 2006), providing evidence for the importance of BIM in tumor suppression. Serum analysis revealed the presence of paraproteins in 8/21 Bim^−/− mice at 4–9 mo of age (median age: 7 mo); this is indicative of abnormal neoplastic expansion of a plasma cell clone (Davidson et al., 1998). Moreover, histological examination revealed abnormal infiltration of lymphoid cells into the lungs and kidneys of aged Bim^−/− mice (Fig. 2 C). Spleen cells from three aged Bim^−/− mice were injected into two RAG1-deficient recipient mice and two wild-type recipient mice each. Interestingly, tumor growth was observed in 4/6 RAG1-deficient recipients but in none of the wild-type recipients. Thus, we conclude that loss of BIM is sufficient to allow cells to become partially transformed with additional oncogenic lesions required for complete escape from host tumor suppression.

To investigate the importance of BIM in a clinically relevant context, mice heterozygous for p53 (a model of Li Fraumeni syndrome) and also deficient for Bim were generated. Loss of BIM markedly accelerated tumor development in p53^+/− mice (Fig. 3 A; P = 0.014). In particular, loss of BIM substantially increased the incidence of lymphoma in p53^+/− mice. Consistent with published results (Donehower et al., 1992; Jacks et al., 1994), the few p53^+/− mice that developed tumors presented mostly with sarcoma (three out of four sick mice by 450 d). In contrast, the majority of sick p53^+/−Bim^−/− (13/15) and p53^+/−Bim^+/− mice (16/21) presented with lymphoma, commonly of thymic origin (Fig. 3 B).

To determine whether loss of BIM altered the frequency of loss of the remaining wild-type p53 allele (loss of heterozygosity) in the tumors that arise in p53^+/− mice, we tested malignant cells harvested from p53^+/−Bim^−/− mice for markers of p53 pathway activity. p53 pathway activity was abrogated in four of the eight tumors analyzed, as indicated by overexpression of p19 (Fig. 3 C), which is maintained at low levels in the presence of functional p53 (Eischen et al., 1999). DNA sequencing of the p53 locus in the four of eight tumors that had retained functional p53 signaling revealed, as expected, no evidence for p53 mutations. These findings are consistent with a previous study, which showed that ∼50% of tumors from p53^+/− mice (<18 mo of age) had retained the wild-type p53 allele (Venkatachalam et al., 1998).

RAG and TdT activities, which during lymphocyte differentiation are essential for antigen receptor gene rearrangement and nucleotide sequence diversification at the junctions, are known sources of oncogenic mutations (e.g., chromosomal translocations; Alt et al., 2013). Accordingly, loss of the activity of the DNA ligases that are critical for the fusion of gene segments during TCR and BCR gene rearrangement substantially accelerates lymphoma development in p53^−/− mice (Frank et al., 2000; Gao et al., 2000). We wondered whether BIM might be critical for the induction of apoptosis that can be triggered in response to potentially oncogenic DNA lesions that can arise from the RAG1/2 and TdT activities. This hypothesis was tested by generating chimeric mice with a hematopoietic system deficient for p53, Bim and Rag1 (p53^−/−Bim^−/−Rag1^−/−), and appropriate controls (i.e., mice with a p53^−/−Bim^−/−, p53^−/−Rag1^−/− or a p53^−/− hematopoietic system). Consistent with our analysis of mice lacking both p53 and BIM in all tissues (Fig. 2), combined loss of p53, and BIM only in hematopoietic cells provoked lymphoma development significantly faster than loss of p53 alone (Fig. 4; P < 0.0001). Remarkably, this acceleration in tumor development was abrogated by additional loss of RAG1, increasing median tumor-free survival from 111 d (p53^−/−Bim^−/−) to 168 d (p53^−/−Bim^−/−Rag1^−/−; P < 0.0001), which is comparable to mice with a p53^−/−Rag1^−/− (165 d) or a p53^−/− hematopoietic system (196 d).

Interestingly, combined loss of RAG1 and BIM resulted in an altered tumor spectrum on a p53-deficient background, with an increased proportion of pro–B/pre–B cell lymphomas, a tumor type that is only rarely observed in mice with either a p53^−/−Bim^−/− or a p53^−/− hematopoietic system, at least on a C57BL/6 background (Fig. 5 A). The reason for this is not clear, but the finding may indicate that in the absence of RAG1/2 activity, pro–B cells may be more genomically unstable than pro–T cells. The former would therefore be more prone to acquiring oncogenic lesions that drive lymphomagenesis.

Tumor burden at autopsy was comparable between mice of the different genotypes with the exception of increased tumor burden in the thymus of the p53^−/−Bim^−/− hematopoietic chimeric mice (Fig. 5 B), similar to what we observed in the germline p53 plus Bim-deleted animals (Fig. 2 B).

These results demonstrate that RAG1/2 DNA recombination activity is required for the ability of loss of BIM to accelerate lymphoma development in p53-deficient mice.

The ability of RAG1 deficiency to abrogate the synergy between loss of BIM and loss of p53 in lymphomagenesis suggests that aberrant RAG1/2 activity may be present in p53^−/−Bim^−/− lymphoma-initiating cells and, by extension, also in the resulting lymphomas. To search for evidence of aberrant RAG1/2 activity in the p53^−/−Bim^−/− tumors, we performed PCR analysis to discriminate between productive and nonproductive D-J TCRβ chain gene segment rearrangements, using published methods (Zhang et al., 2011). Productive TCRβ chain gene rearrangements were found in 9/9 p53^+/−Bim^−/− tumors, but only in 9/11 p53^+/−Bim^+/− lymphomas and 6/9 p53^+/− lymphomas (Fig. 6 A). This indicates that the absence of BIM may facilitate prolonged RAG1/2 activity in T cell progenitors, and thereby allow a higher rate of RAG1/2-induced productive recombination events.

We next wondered whether DNA lesions emanating from TCR or Ig gene rearrangement might cause an increase in BIM levels in lymphoid progenitors. To investigate this, we performed intracellular FACS analysis to compare the levels of BIM protein in Lin⁻CD25⁺CD44⁻ (DN3) T cell progenitors between wild-type and Rag1^−/− (no antigen receptor gene rearrangement possible) mice (Fig. 6, B and C). Consistent with our hypothesis, we observed higher BIM levels in wild-type DN3 cells compared with those lacking RAG1. To confirm these findings using an orthogonal approach, we sorted bone marrow progenitor cells from wild-type, Rag1^−/− and Bim^−/− (negative control for BIM protein detection) mice and performed in vitro differentiation on OP9-DL1 stromal cells (Carotta et al., 2011). Populations of DN1 cells (Lin⁻cKIT⁺CD25⁻) and DN2/3 cells (Lin⁻cKIT^+/−CD25⁺) were isolated by cell sorting. Western blot analysis revealed elevated BIM protein levels in the wild-type DN2/3 cells compared with the Rag1^−/− DN2/3 cells (Fig. 6, D and E). No BIM was observed in the DN1 cells of any genotype.

Collectively, these results indicate that loss of BIM may allow the survival of p53-deficient lymphoma-initiating cells with RAG1/2 induced (potentially oncogenic) DNA lesions, thereby fostering neoplastic transformation to malignancy. They also suggest that after select DNA damaging events, BIM is induced and activates apoptosis to enforce tumor suppression, even in the absence of p53.

BIM is a potent tumor suppressor in a MYC-driven mouse lymphoma model (Egle et al., 2004), and the BIM gene is lost or silenced in several human cancers, including mantle cell lymphoma or renal cancer (Tagawa et al., 2005; Sturm et al., 2006; Mestre-Escorihuela et al., 2007; Zantl et al., 2007; Greenhough et al., 2010; Richter-Larrea et al., 2010). A previous study using a low-dose γ-irradiation–induced lymphoma model in p53-deficient mice has highlighted the contribution of genetic factors in addition to p53 in the suppression of hematopoietic tumor development (Kemp et al., 1994). Our data shed new light on the mechanisms by which BIM suppresses tumor development and reveal how apoptosis can be activated by potentially oncogenic DNA lesions in a p53-independent manner to eliminate nascent cancer-initiating cells. Specifically, we show that loss of BIM accelerates lymphoma development in p53^−/− and even p53^+/− heterozygous mice. We also show that RAG1/2 gene rearrangement activity is required for the synergy between loss of BIM and loss of p53 in lymphomagenesis. Thus, we conclude that BIM is activated in response to DNA lesions caused by RAG1/2 DNA recombination activity to eliminate nascent cancer-initiating cells harboring potentially oncogenic DNA lesions.

Previous studies have demonstrated normal thymic cellularity in Bim^−/− mice, albeit with abnormally increased numbers of mature T cells and reduced immature CD4⁺CD8⁺ thymocytes (Bouillet et al., 1999; Pellegrini et al., 2003; Erlacher et al., 2006). These mice also have normal (or even enhanced) immune function (Pellegrini et al., 2003). Thus, it appears unlikely that impaired antitumor immunity contributes to the acceleration in tumor development caused by loss of BIM on a p53-deficient background.

Although the study presented here focuses predominantly on T lineage lymphomas, sustained survival of lymphoid progenitors with RAG1/2 activity permitted by loss of BIM is likely to also promote development of B lineage lymphomas. Loss of BIM is found in ∼15% of human mantel cell lymphoma cases (Tagawa et al., 2005), and loss of BIM enhances tumor development in a mouse model of this malignancy, which is driven by transgenic cyclin D1 overexpression (Katz et al., 2014). Given that BIM’s critical role as an inducer of apoptosis is most evident in lymphoid and myeloid cells (Delbridge and Strasser, 2015), it appears likely that our findings will be of most relevance to hematological cancers. This tissue-predominant function of BIM probably also underpins the selective effect of its loss in enhancing the development of lymphoma in p53^+/− mice, resulting in increased lymphoma incidence (with concomitantly reduced sarcoma incidence).

The molecular mechanisms that activate BIM in response to DNA lesions, including those caused by aberrant RAG1/2 activity, are unknown. Indeed, recent studies have further highlighted the absence of mechanistic insight into how BIM is activated by cell death stimuli more generally (Clybouw et al., 2012; Herold et al., 2013), questioning previous studies that found transcriptional regulation by FOXO3a (Dijkers et al., 2000) or direct phosphorylation by ERK (Ley et al., 2004) to be critical. The mechanisms that control BIM expression therefore remains a topic of high interest to the field of cell death research. In light of the data presented here, it is interesting to speculate that pathway components upstream of BIM may be disrupted in those cases of mantel cell lymphoma that retain wild-type BIM.

Experiments with mice were conducted according to the guidelines and with approval from the Walter and Eliza Hall (Parkville, Victoria, Australia) Institute Animal Ethics Committee. p53^−/− (Jacks et al., 1994), Bim^−/− (Bouillet et al., 1999), Bmf^−/− (Labi et al., 2008), and Rag1^−/− (Mombaerts et al., 1992) mice have been previously described. All strains, except the Bmf^−/− mice, were originally generated on a mixed C57BL/6 × 129SV background using 129SV-derived ES cells and were backcrossed to C57BL/6 mice for >20 generation before the current study. The Bmf^−/− mice were generated on an inbred C57BL/6 background using C57BL/6-derived ES cells, and the Bmf^−/− mice were maintained on this background. Compound mutant mice were generated by intercrossing the relevant single-knockout strains. Genotyping protocols will be made available on request. This study was aided by gifts of mice from the following investigators: T. Jacks (Koch Institute for Integrative Cancer Research at MIT, Boston, MA), A. Villunger (Medizinische Universität Innsbruck, Innsburck, Austria), J. Adams (Walter and Eliza Hall Institute of Medical Research [WEHI], Parksville, Victoria, Australia), S. Cory (WEHI), L. Corcoran (WEHI), C. Scott (WEHI), and P. Bouillet (WEHI).

Chimeric mice reconstituted with the hematopoietic system of compound mutant mice and controls were generated through timed matings of mice lacking various combinations of p53, Rag1, and Bim. Fetal liver cells were isolated from embryonic day 13 embryos and used to reconstitute four lethally irradiated (2 × 5.5 Gy, 3 h apart) C57BL/6-Ly5.1 mice. Recipient mice were maintained on neomycin sulfate–supplemented (Sigma-Aldrich) drinking water for 21 d after γ-irradiation to prevent infection. Such reconstituted mice were monitored thrice weekly for signs of lymphoma development (lymphadenopathy, splenomegaly, and respiratory difficulties).

Immediately before sacrifice, peripheral blood was obtained by retroorbital bleed and collected into heparinized tubes. Total white blood cell, red blood cell, and platelet numbers were determined in an Advia 120 machine using the mouse analysis software module (Bayer). At autopsy, the weights of the spleen, thymus, and lymph nodes (inguinal, axillary, and brachial) of sick mice were determined and the major organs were collected for histological analysis. Mice were deemed to be lymphoma bearing based on abnormal enlargement of the lymphoid organs, elevated peripheral blood counts, infiltration of tumor cells into nonlymphoid organs (such as the liver, lungs, and kidneys), and by abnormal cell surface marker expression. In the case of chimeric (hematopoietic system reconstituted) mice, lymphoma cells were verified to be of donor origin based on expression of Ly5.2. Ambiguity on whether a mouse indeed had a lymphoma was resolved through transplantation and/or tissue culture to determine whether lymphoma cell lines could be established in vivo and/or in vitro. Sarcomas were identified on the basis of characteristic presentation at autopsy and by histological analysis.

Retention versus loss of the wild-type p53 allele was examined by PCR; primers and PCR conditions will be made available on request. D-J β-chain TCR gene segment ligation efficiency was investigated by PCR using previously published methods (Zhang et al., 2011).

Sequencing of genomic DNA was performed on coding regions of the Trp53 gene from exon 4–10 inclusive, encompassing the entire DNA-binding domain. Sequencing was performed using the Illumina MiSEQ platform as previously described (Aubrey et al., 2015). Primer sequences were designed to target intron genomic DNA sequence flanking each exon as follows: exon 4–1 for-5′-CAGAGCAGAAAGGGACTTGG-3′ and rev-5′-CGCCATAGTTGCCCTGGTAAG-3′; exon 4–2 for-5′-TTTTGAAGGCCCAAGTGAAG-3′ and rev-5′-AGGCATTGAAAGGTCACACG-3′; exon 5 for-5′-CGACCTCCGTTCTCTCTCC-3′ and rev-5′-GAGGCTGCCAGTCCTAACC-3′; exon 6 for-5′-CGGCTTCTGACTTATTCTTGC-3′ and rev-5′-CCCTTCTCCCAGAGACTGC-3′; exon 7 for-5′-GTAGGGAGCGACTTCACCTG-3′ and rev-5′-CCCTAAGCCCAAGAGGAAAC-3′; exon 8 for-5′-TCTTACTGCCTTGTGCTGGTC-3′ and rev-5′-TGTGGAAGGAGAGAGCAAGA-3′; exon 9 for-5′-CCCAAAGTCACCTCTTGCTC-3′ and rev-5′-GAGAACCACTGTCGGAGGAG-3′; and exon 10 for-5′-GGTTGTGTGACCTTGTCCAG-3′ and rev-5′-AGCAGGGTGGGGTTTTTATC-3′. Each primer set was made to include a nested sequence for secondary amplification and indexing as previously described (Aubrey et al., 2015).

To isolate hematopoietic progenitors, femurs, and tibias were isolated, and flushed into balanced salt solution (BSS; Thermo Fisher Scientific). Isolated cells from two mice were pooled into a single sample. The cell suspensions were later prepared in BSS supplemented with 2% (vol/vol) fetal calf serum. Antibodies to the following surface molecules were used as supernatants for immune-magnetic bead depletion of lineage (Lin)⁺ BM cells: CD3 (KT3-3.1), CD8 (53–6.7), CD2 (RM2-1), B220 (RA3-6B2), MAC-1 (M1-70), GR-1 (RB6-8C5), and erythrocyte (TER119). Cell suspensions were mixed with BioMag goat anti–rat IgG antibody-coupled beads (QIAGEN), and depleted of Lin⁺ cells with a Dynal MPC-L magnetic particle concentrator (Invitrogen). Lin⁻ BM cells were then stained using fluorescently conjugated antibodies as described: goat anti–rat IgG (H+L) antibodies conjugated to Alexa Fluor 680 (Invitrogen), c-Kit APC (2B8), SCA-1 PECy7 (E13161.7), and CD25 Pacific Blue (PC61). Cells were stained for 20–30 min on ice with fluorescently conjugated antibodies, and sorted on FACSAria II (BD) for LSK (Lin⁻ SCA-1⁺ c-KIT⁺) cells. Dead cells were excluded by staining with propidium iodide (PI; 2 µg/ml).

The sorted cells were seeded on OP9-DL1 stromal cells in the presence of 100 ng/ml stem cell factor (SCF), 5 ng/ml FLT3L, and 2% IL-7 supernatant for T cell development, as previously described (Carotta et al., 2011). The cells were passaged into a larger well, as necessary, on a new OP9-DL1 layer, when they had achieved maximum confluency. The concentration of SCF was lowered to 50 ng/ml on day 10, and subsequently to 25 ng/ml on day 14 to promote T cell differentiation. Cells were harvested for FACS sorting on day 18. Cells were stained with the fluorescently conjugated antibodies as mentioned above, and sorted on a FACSAria II for DN (double-negative, CD3/TCR⁻CD4⁻CD8⁻), DN1 (c-KIT⁺ CD25⁻), and DN2+3 (CD25⁺) fractions. OP9-DL1 cells were excluded in the sort by their GFP fluorescence.

In vitro–differentiated T cell samples were prepared from 150,000 sorted cells. Tumor samples were prepared from cryopreserved single-cell suspensions. In both cases, lysates were prepared using RIPA buffer with Complete Ultra Protease Inhibitors (Roche).

Protein lysates, loaded at equal quantity, were separated on a NuPAGE Novex 10% Bis-Tris Protein Gel (Thermo Fisher Scientific), followed by iBlot transfer to nitrocellulose (P3 program, 7 min; Thermo Fisher Scientific). Membranes were blocked in 5% skim milk powder in PBS with 0.1% Tween-20, and then probed with the following antibodies, as indicated in the figure legends: polyclonal rabbit anti-BIM (ADI-AAP-330-E; Enzo Life Sciences), rabbit polyclonal anti-p19ARF (Ab80; Abcam), polyclonal rabbit anti-p53 (NCL-p53-CM5p; Leica Biosystems), and monoclonal mouse anti-actin (#A2228, clone AC-74: Sigma-Aldrich), followed by goat anti–rabbit IgG antibodies conjugated to HRP or goat anti–mouse IgG antibodies conjugated to HRP, respectively (both from Southern Biotech). Blots were developed using the ChemiDoc Touch Imaging System and quantitated with Image Lab Software (both from Bio-Rad Laboratories).

Lymphoma phenotype was determined by flow cytometric analysis after staining for hematopoietic subset-specific cell surface markers (B220 [RA3-6B2; WEHI], CD4 [YTA3.2.1; WEHI], CD8 [YTS169; WEHI], IgM [5.1; WEHI], IgD [11-26C; WEHI], TCRβ [H57.59.1; WEHI], Ly5.1 [A20.1; WEHI], and Ly5.2 [5.450.15.2; WEHI]). Staining was performed on ∼10⁶ cells for 30 min on ice in the dark. Before analysis, samples were resuspended in medium with propidium iodide (PI, 2 µg/ml; Sigma-Aldrich) and cells were analyzed using a FACSCalibur flow cytometer (BD). Dead cells (PI⁺) were excluded from analysis.

Intracellular staining for BIM was performed on isolated thymocytes that had been fixed using the eBioscience Foxp3/Transcription Factor Staining Buffer Set (eBioscience). Surface staining for lineage markers (B220 [RA3-6B2; WEHI] or CD19 [1D3; WEHI], CD4 [YTA3.2.1; WEHI], CD8 [YTS169; WEHI], Gr-1 [RB6-8C5; WEHI], MAC-1 [MI/70; WEHI], Ter119 [TER-119; WEHI], NK1.1 [PK136; WEHI], CD3 [KT3-1; WEHI], and MHC Class II [M5/114; WEHI], as well as CD25 [PC61; WEHI] and CD44 [IM781; WEHI]) was performed. Cells were then fixed in Fix/Perm buffer for 30 min on ice and, afterward, washed with Perm/Wash buffer. The cells were then stained using the 3C5 anti-BIM antibody conjugated to Alexa Fluor 647 (Life Technologies) for 30 min. Cells were washed and analyzed using a LSR-II flow cytometer (BD). DN3 cells were defined as Lin⁻CD25⁺CD44⁻. Specific BIM signal was determined relative to a Bim^−/− sample (negative control).

In all cases, sample and mouse numbers were chosen based on previous experience with mice deficient for various BCL-2 family members. All data points were included in final analysis, without exclusions. All animal survival experiments were conducted in a blinded manner with sick mice identified by an experienced mouse technician blinded to their genotype and the hypothesis. Kaplan-Meier animal survival curves were prepared using Prism 6 (GraphPad Software) and statistical comparisons were performed in a pairwise manner using the log-rank (Mantel-Cox) test; P <0.05 was deemed to be statistically significant. Organ weights and blood counts of tumor-bearing mice were plotted using Prism 6 and comparisons were made using the Student’s t test (unpaired, two-tailed).

We thank all members of the Strasser, Gray, and Herold laboratories for their support and advice, and Dr. P. Fedele for advice and expertise with plasmacytoma analysis. We thank Drs. T. Jacks, A. Villunger, J. Adams, S. Cory, L. Corcoran, C. Scott, and P. Bouillet for gifts of mice; K. Humphries, K. Walker, E. Lanera, J. Mansheim, G. Siciliano, S. O’Connor, K. Trueman, G. Dabrowski, K. Vella, and K. McKenzie for expert animal care; B. Helbert, H. Ierino, K. Mackwell, and C. Young for genotyping; J. Corbin and J. McManus for automated blood analysis; and D. Quilici and T. Nikolaou for γ-radiation services.

This work was supported by grants and fellowships from the Cancer Council of Victoria (S. Grabow and A.R.D. Delbridge [Sydney Parker Smith Postdoctoral Research Fellowship]), Leukemia Foundation National Research Program Clinical PhD Scholarship (B.J. Aubrey), Lady Tata Postdoctoral Fellowship (S. Grabow), the National Health and Medical Research Council (Program grant #1016701; SPRF Fellowship 1020363, A. Strasser; Project Grant #APP1049720, M.J. Herold), the Leukemia and Lymphoma Society (SCOR grant #7001-13), the estate of Anthony (Toni) Redstone OAM, Melbourne International Research Scholarship (University of Melbourne, S. Grabow), Melbourne International Fee Remission Scholarship (University of Melbourne, S. Grabow), Australian Postgraduate Award (A.R.D. Delbridge), Cancer Therapeutics CRC Top-up Scholarship (S. Grabow and ARDD), and the Ian Potter Foundation. This work was made possible by operational infrastructure grants through the Australian Government Independent Medical Research Institutes Infrastructure Support Scheme and the Victorian State Government Operational Infrastructure Support Program.

The authors declare no competing financial interests.

Author contributions: Study and experiments were designed and conceived by A.R.D. Delbridge and A. Strasser. Experiments were conducted and analyzed by all authors. Manuscript was prepared by A.R.D. Delbridge and A. Strasser, with assistance from the other authors.