Mad2 overexpression is common in human tumors (Alizadeh et al., 2000
; Chen et al., 2002
; Garber et al., 2001
; Hernando et al., 2004
; Li et al., 2003
; Tanaka et al., 2001
; van ‘t Veer et al., 2002
) and has been shown to promote genomic instability in cell culture models (Hernando et al., 2004
). Here we extend this analysis and show that Mad2 overexpression can initiate tumorigenesis and cooperate with other oncogenic stimuli. Consistent with a role for Mad2 in promoting genomic instability, Mad2-induced tumors have frequent genomic rearrangements, whole chromosome gains or losses and sustained Mad2 overexpression is not required for continued tumor growth.
Although we show that Mad2 overexpression can initiate tumorigenesis, activating mutations have not been reported in human cancers. However, studies suggest that Mad2 is under the control of E2F which is deregulated in many human cancers (Hernando et al., 2004
). Thus, cells suffering mutations in the Rb pathway not only gain a proliferative advantage, but, as suggested in this study, can gain an instability that (again, as shown here) may contribute to tumorigenesis even if present only transiently. It should be noted that while the effect of Mad2 overexpression on tumor initiation and acceleration is likely to result from the observed chromosome instability, other unknown effects of Mad2 overexpression cannot be ruled out at this time.
Mad2 overexpression leads to a highly penetrant induction of a wide range of tumors in mice including lung adenomas, hepatomas and hepatocellular carcinomas, lymphomas and fibrosarcomas. Other cell types might also be susceptible to the effects of Mad2 overexpression but could have been masked by a variety of factors such as low expression levels from the CMV promoter or early lethality. This issue can now be addressed by the use of other tissue specific Mad2 alleles. As described, Mad2 overexpression was also observed to accelerate tumorigenesis in a well-established model of lymphoma driven by the expression of the myc oncogene in the B cell lineage. In addition, higher levels of Mad2 mRNA have been reported in DLBCL (Alizadeh et al., 2000
) as compared to most other B cell lymphoma subtypes, confirmed by our expression analyses. Interestingly, DLBCL display a highly aggressive biological behavior and represent the most aberrant B cell lymphomas in terms of ploidy alterations. These data suggest that in addition to tumor initiation, Mad2 overexpression may play an important role in tumor progression and mortality. Indeed, as reported previously, Mad2 is a poor prognostic marker for neuroblastoma (Hernando et al., 2004
) consistent with this hypothesis.
Clearly, in tumors the “penalty” for loss of a whole chromosome induced by Mad2 overexpression is balanced by growth advantages which likely result from LOH at tumor suppressor loci. Once a cell has acquired the CIN phenotype, theoretical considerations suggest that there is an optimal chromosome loss rate (between 10exp-2 and 10exp-3 per chromosome per generation) (Komarova and Wodarz, 2004
) which will maximize the loss of tumor suppressor genes and expansion of transformed clones. Indeed, we and others have observed a threshold for cell viability in cell culture and animal models as high level of expression of Mad2 (this study), complete loss of separase in mice (Kumada et al., 2006
; Wirth et al., 2006
) and complete loss of mitotic checkpoint function all lead to a profound cell death and early embryonic lethality (Babu et al., 2003
; Dobles et al., 2000
; Kops et al., 2004
; Michel et al., 2004
; Wang et al., 2004
). It is likely for this reason that no human tumors to date have been identified which have sustained a complete loss of Mad2 function, although partial loss of function has been observed (Percy et al., 2000
; Wang et al., 2002
; Wang et al., 2000
). It will be of interest to see if separase heterozygous mice will develop tumors with prolonged latency. While high levels of Mad2 overexpression in the rtTA system caused cell death of MEFs in culture, it is likely that in tumors that develop in the rtTA mice there is selection for levels of Mad2 which allow cell viability but promote cellular transformation.
The simplest explanation for the chromosome instability observed in the Mad2 overexpressing mice is that the stabilization of securin and cyclin B, observed previously in primary IMR90 cells (Hernando et al., 2004
) and now in Mad2 overexpressing lymphocytes (), inhibits the activity of separase leading to non-disjunction events and to cytokinesis inhibition. This is consistent with the oncogenic role of securin (PTTG) overexpression (Pei and Melmed, 1997
). Formal genetic demonstration of this hypothesis awaits the analysis of the Mad2 transgenics crossed with the securin knockout animals which is currently underway.
The cause of the observed interstitial deletions amplifications in the Mad2 overexpressing cells is unclear at this time. It is possible that when cohesiveness of sister chromatids is maintained during the exit from mitosis, chromosome breakage and rejoining events facilitate this type of chromosome instability. Indeed, our real time microscopy of Mad2 overexpressing cells shows evidence of chromatin trapped in extended cytoplasmic bridges during cytokinesis followed by a breakage event (see Supplemental movie 1
). In addition, karyotype analysis of MEFs overexpressing Mad2 show clear evidence of chromosome breakage in addition to whole chromosome gains and losses.
The rate of acquisition of CIN in tumors must be comparable to spontaneous mutation rates in order to compete with mutational LOH at tumor suppressor loci and therefore play a role in tumor initiation. The acquisition of CIN may in fact be the second hit after a mutation at a TS locus since whole chromosome loss is not rate limiting in a CIN cell (see discussion in (Nowak et al., 2002
)). Interestingly, since Mad2 overexpression induces both interstitial deletions and amplifications and whole chromosome loss it might induce both the initial loss of function event at tumor suppressor loci as well as LOH. This would serve to minimize the deleterious effects of whole chromosome loss. Alternatively, spontaneous mutation of tumor suppressor genes may be the event that is selected for in the clonal expansion of the initiating tumor cell. Further analysis of the tumors that arise in the Mad2 transgenic mice is required to address this question.
It has been suggested that lagging and bridging chromosomes in human cells in culture are insufficient to induce cleavage furrow regression and tetraploidization (Shi and King, 2005). Rather, these mislocalized chromatids must end up in the wrong daughter cell in order to induce tetraploidization and aneuploidy is generally acquired after the tetraploidization event. However, this notion is at odds with recently published studies in S. cerevisiae
in which chromatin in the cleavage furrow induces an ipl dependent signaling cascade which results in furrow regression (Norden et al., 2006). Mad2 overexpression would be predicted to induce a high rate of non-disjunction events due to persistence of cohesion and tetraploidy prior to the appearance of an aneuploid cell. However, we observe aneuploidy early after Mad2 induction in murine cells with chromosome numbers in the 2N and 4N range. A similar result has been reported recently in CENP-E knockout MEFs (Weaver et al., 2006
) This discrepancy is unlikely to be a murine specific effect as has been suggested (Shi and King, 2006
) since a similar result has been observed in Mad2 overexpressing human cells (Hernando et al., 2004
) and in human cells which show high rates of non-disjunction due to the loss of one copy of Mad2 (Michel et al., 2001
). We conclude that in several different settings aneuploidy can be established independently of tetraploidization.
Turning off Mad2 transgene expression in established tumors has little effect on tumor progression, at least in the case of the hepatomas examined. This is in contrast to the oncogene-dependence observed in other systems (for review see (Jonkers and Berns, 2004
)). We presume that in the case of Mad2, the lack of dependence is a reflection of the early induction of chromosome instability by Mad2 which would persist after Mad2 levels are normalized. This hit-and-run effect of Mad2 overexpression may lead to an underestimation of the fraction of human tumors which have experienced Mad2 overexpression or overexpression of other mitotic checkpoint components during the early phases of the oncogenic process. In sum then, our results suggest that deregulation of mitotic checkpoint pathways by Rb inactivation or other mechanisms may be an early and transient event in the initiation and evolution of a wide variety of common cancers.
Generation of Mad2 inducible Mice
The pTRE vector from Clontech, containing the tetracycline operator and the SV40 polyadenylation sequence, was linearized with EcoRI and BamHI. The murine Mad2 cDNA was amplified with specific primers containing the HA epitope tag and the corresponding restriction enzymes and ligated into the pTRE vector. Restriction digests and sequencing were used to identify clones in which the Mad2 cDNA had inserted into the correct orientation.
Animal husbandry and genotyping
TetO-Mad2 transgenic mice, CMV-tTA and CMV-rtTA mice were kept in a pathogen-free housing under guidelines approved by the MSKCC Institutional Animal Care and Use Committee and Research Animal Resource Center. Eμ-myc mice (Adams et al., 1985
) and CMV-tTA mice (Furth et al., 1994
) have been previously described. CMV-rtTA mice contain an interstitial deletion on chromosome 5 and will be described elsewhere. Doxycycline was administered by feeding mice with doxycycline-impregnated food pellets (625ppm; Harlan-Teklad). Tail DNA was isolated using Qiaprep Tail DNeasy isolation kit (QIAGEN) according to the manufacturer’s protocol. TetO-Mad2 transgenic mice were genotyped using the following primers: Mad2F: 5′-CCATCCACGCTGTTTTGACCTC-3′; Mad2R: 5′-GGCTTTCTGGGA CTTTTCTCTACG-3′ and CMV-rtTA mice: rtTAF: 5′- GTGAAGTGGGTCCGCGTA CAG-3′ and rtTAR: 5′- GTACTCGTCAATTCCAAGGGCATCG-3′.
Preparation of MEFs and lymphocytes and tissue culture
Mouse embryonic fibroblasts (MEFs) were isolated from E13.5 embryos and cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 2 mM glutamine, 1% penicillin/streptomycin, 10% tetracycline free fetal bovine serum (FBS) and 1 μg/ml of doxycycline when indicated. For proliferation assays, 1x105
cells were plated on 6-well plates in duplicate as described previously (Sotillo et al., 2001
). Primary lymphocytes were isolated from the spleen of 6-month-old mice, cultured in RPMI + 10% FBS and stimulated with PMA and ionomycine (Sigma) in the presence or absence of doxycycline and cell cycle profiles were analyzed by cytometry.
Retroviral-mediated gene transfer and lymphoma generation
Eμ-myc HSCs derived from fetal livers at embryonic day 13–15 were transduced with retroviruses expressing Mad2 or the MSCV vector alone and used to reconstitute the hematopoietic compartment of lethally irradiated C57BL/6 mice (Schmitt et al., 2002
; Schmitt et al., 2000
). Mice were observed for lymphoma onset with periodic palpation of peripheral lymph nodes, overall morbidity and by whole body fluorescence imaging. After the appearance of well-palpable lymphomas, tumors were harvested and either fixed for histological evaluation or rendered as single-cell suspensions, analyzed or stored frozen in 10% DMSO.
Magnetic resonance imaging
Individual mice were subjected to MRI assessment for detection of tumors. In brief, mice were anesthetized with 2% isofluorane and images were obtained on a Bruker 4.7T 40cm bore magnet with a commercial 7-cm inner diameter birdcage coil in the Animal Imaging MRCore Facility at MSKCC. Low-resolution axial scout images were obtained initially, followed by a high-spatial-resolution T2-weighted axial images (repetition interval (TR)=3,800ms, effective echo time (TE)=35ms, eight echoes per phase encoding step, spatial resolution=1.0mm slice thickness x 112μmX112μm in plane resolution, and four repetitions of data acquisition for 8–9 min of imaging time).
FACS, Karyotyping, FISH, and Live Cell Imaging
For FACS analysis, trypsinized cells were washed in PBS, fixed in 70% ethanol and stained with propidium iodide (50μg/ml). 104
cells were analyzed by using a FACScalibur (Becton Dickinson). Apoptotic cells were labeled by fluorescent TUNEL assay (In Sity Cell Death Detection Kit, Roche) and quantified by FACS. For karyotyping, cells were incubated in medium containing Colcemid (0.05 μg/ml) for 40 minutes and harvested by standard cytogenetic procedures. Metaphase spreads were stained with DAPI (0.08%) in 2xSSC. For FISH analysis we made probes using pairs of BAC clones near the centromeres for each chromosome. Additional details are listed in the Supplemental Material. Mitotic index was quantified by measuring MPM2 expression (anti-MPM2, Upstate Biotechnology) versus DNA content (PI) by FACS. For live cell imaging, primary MEFs were infected twice with a retrovirus expressing H2B-GFP (Yamamoto et al., 2004
) and were cultivated in a glass-bottom culture (Delta TPG) dish. Imaging was performed as previously described (Michel et al., 2004
Array CGH analysis
Genomic DNA extracted from normal and tumor livers from TetO-Mad2/CMV-tTA and CMV-rtTA mice was subjected to comparative genomic hybridization array analysis at the MSKCC Genomics Core Lab. For each mouse, genomic DNA extracted from the liver of a wild type littermate was used as a reference and hybridized into mouse CGH Agilent arrays (44A version). Results were analyzed using a special normalization method correcting for the GC content of the probes (adapted from (Tonon et al., 2005
RNA and protein analysis
RNA was isolated using the RNeasy kit (Qiagen, Valencia, CA). RNA was treated with DNaseI (Ambion) to eliminate any contaminating DNA. RT-PCR reactions were performed with SuperScript III (Invitrogen) according to the manufacturer’s instructions. For quantitative RT-PCR, reactions were performed using the ABI7900 Sequence Detection System (Applied Biosystems). Primer sequences and amplification conditions and protein expression are described in the Supplemental Material.
Tissue microarrays of human lymphomas
We analyzed the Oncomine database for expression of Mad2 on different set of microarray data comparing normal vs. cancer samples and established a p<0.05 as cut-off limit. Several tissue microarrays (TMA) comprising 168 cases of FL (Hedvat et al., 2002
), 81 cases of DLBCL, 35 small lymphocytic lymphoma/chronic lymphocytic leukemia (SLL/CLL), 35 mantle cell lymphomas (MCL), 15 T cell lymphoblastic lymphomas, 10 angioimmunoblastic lymphomas, 40 peripheral T cell lymphoma (PTCL), 6 anaplastic large cell lymphoma (ALCL), 7 Burkitt’s lymphomas, 9 plasmatocytomas and 4 plasmablastic lymphomas were analyzed for Mad2 expression by immunohistochemistry analysis. Patient samples were obtained through institutionally approved protocols.
For immunohistochemistry analysis, representative sections were deparaffinized, rehydrated in graded alcohols, and processed using the avidin-biotin immunoperoxidase method. Sections were subjected to antigen retrieval by microwave oven treatment using standard procedures. Diaminobenzidine was used as the chromogen and hematoxylin to counterstain nuclei. The antibodies used for inmunohistochemistry are listed in the Supplemental Material. TMAs were scored (by JTF and EH) evaluating % of positivity of tumor cells and intensity of nuclear staining.