The Zn-finger domain of MdmX suppresses cancer progression by promoting genome stability in p53-mutant cells

The MDMX (MDM4) oncogene is amplified or overexpressed in a significant percentage of human tumors. MDMX is thought to function as an oncoprotein by binding p53 tumor suppressor protein to inhibit p53-mediated transcription, and by complexing with MDM2 oncoprotein to promote MDM2-mediated degradation of p53. However, down-regulation or loss of functional MDMX has also been observed in a variety of human tumors that are mutated for p53, often correlating with more aggressive cancers and a worse patient prognosis. We have previously reported that endogenous levels of MdmX can suppress proliferation and promote pseudo-bipolar mitosis in primary and tumor cells derived from p53-deficient mice, and that MdmX-p53 double deficient mice succumb to spontaneously formed tumors more rapidly than p53-deficient mice. These results suggest that the MdmX oncoprotein may act as a tumor-suppressor in cancers with compromised p53 function. By using orthotopic transplantation and lung colonization assays in mice we now establish a p53-independent anti-oncogenic role for MdmX in tumor progression. We also demonstrate that the roles of MdmX in genome stability and in proliferation are two distinct functions encoded by the separate MdmX protein domains. The central Zn-finger domain suppresses multipolar mitosis and chromosome loss, whereas the carboxy-terminal RING domain suppresses proliferation of p53-deficient cells. Furthermore, we determine that it is the maintenance of genome stability that underlies MdmX role in suppression of tumorigenesis in hyperploid p53 mutant tumors. Our results offer a rationale for the increased metastatic potential of p53 mutant human cancers with aberrant MdmX function and provide a caveat for the application of anti-MdmX treatment of tumors with compromised p53 activity.


INTRODUCTION
The p53 transcription factor regulates expression of wide variety of genes involved in cellular response to stress, and mutation of p53 gene is the most common genetic lesion in human cancers. The level of p53 activity in a cell is tightly regulated by the MDM2 and MDMX (MDM4) oncoproteins. These MDM homologs bind to the amino-terminal p53 transactivation domain and inhibit p53 regulation of heterologous gene expression. MDM2 and MDMX heterodimerize to facilitate MDM2-mediated ubiquitination and degradation of p53. 1-3 MDM proteins are required to negatively regulate p53 activity in mice during the early and mid-stages of development, and embryonic lethality caused by MDM-deficiency is rescued by concomitant p53 deletion. [4][5][6][7] Mdm2 and MdmX also play critical roles in peri-natal and adult mice by controlling p53 regulation of cell differentiation, the DNA damage response, tissue homeostasis and aging. 8 MDM2 and MDMX also regulate p53-mediated tumor suppression. MDM gene amplification and/or overexpression have been observed in a variety of human cancers, including leukemia and lymphoma, breast cancer, glioblastoma, soft tissue sarcoma, osteosarcoma and retinoblastoma. [9][10][11] As most tumors with amplified copy numbers of MDM genes retain wild-type p53, 10,12 the increased level of MDM2 or MDMX proteins is thought to promote oncogenesis by inhibiting p53 activity in these cells. Experiments involving Mdm2 or MdmX overexpression in cells or mice also suggest that MDM oncoproteins may have p53-independent roles in regulating cell growth and tumorigenesis [13][14][15][16][17] although the precise contributions of these p53-independent, proto-oncogenic effects in human cancers remains uncertain.
We have previously shown that mice co-deleted for both MdmX and p53 succumb to spontaneous tumorigenesis faster than mice deleted solely for p53, 18 suggesting p53-independent antioncogenic function for MdmX in mice. In contrast, deletion of Mdm2 fails to alter the tumorigenic potential of mice lacking functional p53. 19 Down-regulation or loss of functional MDMX protein has also been associated with more aggressive or advanced osteosarcomas, soft tissue sarcomas, thyroid and prostate carcinomas, and chronic myelogenous leukemia. [20][21][22][23][24] In addition, an alternatively spliced MDMX variant is often found in high-grade glioblastomas, papillary thyroid carcinomas, soft tissue sarcomas and osteosarcomas. 20,21,24 In both human tumors 20,25 and in mouse model with targeted MdmX internal deletion 26 this altered splicing reduces the level of full-length (FL) MdmX transcript and generates a novel transcript encoding a severely truncated, unstable MdmX protein. The increase in short to FL transcript ratio in osteosarcomas correlates with reduced MDMX protein levels, faster metastatic progression and greatly reduced patient survival. 20 Lower MDMX protein levels in many osteosarcoma or breast cancer cell lines and in soft tissue sarcomas correlate with compromised p53 function. 20 Although it is likely that p53-mutant tumor cells have lost the selective pressure to maintain high levels of functional MDMX, it is unclear why loss of functional MDMX in these cells correlates with a more aggressive cancer.
We previously observed that p53-deficient mouse embryo fibroblasts (MEFs) and p53-deficient mouse tumor cells proliferate faster when MdmX is also deleted, and that MdmX/p53-doublenull cells have increased incidence of multipolar mitosis and reduced cell ploidy compared with p53-null cells. 18 These findings suggest a p53-independent role for MdmX in suppression of proliferation and in maintenance of genome stability in hyperploid mouse cells. In the present study, we use human tumor cells in mouse orthotopic transplantation and lung colonization assays to explore the relevance of these p53-independent effects of MdmX in tumorigenesis. We provide the evidence that MdmX suppresses tumor progression and metastases in these mouse models of human cancer. Furthermore, we find the inhibition of cell proliferation and maintenance of genome stability to be separable MdmX functions encoded by different MdmX protein domains. We demonstrate that the ability of MdmX Zn-finger domain to suppress multipolar mitosis and large-scale ploidy reduction in p53-mutant cells underlies the role of MdmX in tumor suppression. We discuss the implications of our findings on cancer treatment strategies and on current models of genome instability and cancer progression.

RESULTS
MdmX slows cycling of p53-deficient cells MdmX/p53 double-null MEFs and primary epithelial tumor cells from MdmX/p53 double-null mice proliferate faster than MEFs and tumor cells solely deficient for p53 (ref. 18 and Figure 1a). Multipolar mitosis (Figure 1b) are more common in populations of MdmX/p53-double-null than in p53-null cells (20% vs 10%, respectively, of all mitotic cells). Therefore, it is possible that the divisions that generate more than two daughter cells per division might contribute to the increased proliferation rate of MdmX/p53null cells. We have previously demonstrated 27 that polyploid cells undergoing multipolar mitosis can indeed generate more than two daughter cells but many of the resulting progeny dies during one or two subsequent divisions. Time-lapse video microscopy analyses now revealed that only 21% of all multipolar mitosis results in multipolar division and 71% of such progeny died or arrested during the 69 h of filming. A majority (79%) of multipolar mitosis produced only two viable daughter cells (Figure 1c) that underwent normal bipolar mitosis and continued to divide in bipolar fashion until the end of filming. Gamma-tubulin/ 4ʹ-6-diamidino-2-phenylindole-staining of cells in late multipolar anaphase typically revealed an unequal distribution of genetic material illustrated in Figure 1d. Therefore, it is unlikely that multipolar mitosis and the generation of more than two daughter cells per division accounts for faster proliferation rate of MdmX/p53-null cells. We applied live imaging to determine the duration of cell cycle at the single-cell level by measuring the length of time from the onset of anaphase in mother cell to the onset of anaphase in daughter cells (Figure 1e). The results showed that the absence of MdmX in p53-deficient cells significantly shortens cell cycle length ( Figure 1f). MEFs (left panel) or tumor cells (right panel) lacking both MdmX and p53 cycled 5 h and 2.5 h faster, respectively, than their p53-deficient counterparts retaining MdmX. Thus, it is a more rapid cell cycle progression and not a higher number of daughter cells per division that accounts for the faster proliferation rates of p53-deficient cells co-deleted for MdmX.
The primary sequence of MdmX contains four functional domains: an amino-terminal p53 binding (p53-BD) domain, a central acidic domain, a zinc-finger (ZnF) domain of unknown function and a carboxy-terminal RING domain critical for the interaction of MdmX with Mdm2. To characterize the ability of MdmX to suppress cell proliferation and multipolar mitosis, we utilized mouse MdmX expression constructs encoding either the FL, or mutant MdmX lacking carboxy-terminal RING domain (dRING), central zinc-finger domain (dZnF) or both (dZnF-dRING; Figure 2a). These constructs were co-transfected along with a puromycin drug selection marker into the epithelial tumor cells derived from MdmX/p53-double knockout (DKO) mice, and pools of puromycin-resistant stably transfected cells were recovered for further analyses.
Quantitative PCR analyses showed that the levels of MdmX messenger RNA (mRNA) in transfected cells (FL, 14.2 ± 5.7, dRING, 11.0 ± 5.8; dZnF, 13.3 ± 3.6; dZnF-dRING, 35.2 ± 4.5) are similar to the endogenous MdmX levels in p53 null mouse tumor cells (13.7 ± 4.5) and in wild-type MEFs (11.07 ± 4.6). Proliferation assays revealed that exogenous FL MdmX and dZnF mutant suppress proliferation of MdmX/p53-deficient cells (Figure 2b). In contrast, MdmX lacking either the RING domain or lacking both the ZnF and RING domain failed to alter the proliferation of the doublenull cells. These results indicate that the RING domain of MdmX encodes an anti-proliferative function. The role of the MdmX RING domain in suppressing cell proliferation was confirmed at the single-cell level by video microscopy (Figure 2c). The presence of MdmX RING domain, but not the central ZnF domain, correlates with an~2 h increase in cell cycle duration of p53-null cells.
Since MdmX interacts with Mdm2 and forms heterodimer through its RING domain, we explored whether the p53independent effect of RING domain on cell proliferation requires Mdm2. Tumor cells derived from Mdm2/MdmX/p53-triple deficient (TKO) mice were transduced with FL or MdmX deletion constructs (relative MdmX mRNA levels: Mock, 1.0; FL, 6.6 ± 3.6; dRING, 12.1 ± 7.0; dZnF, 23.5 ± 2.8; dZnF-dRING, 7.4 ± 2.3). Cell proliferation assay (Figure 2d) reveal that MdmX with intact RING domain suppresses proliferation of TKO cells. Analysis by video microscopy confirmed that the presence of RING domain prolonged the cell cycling time from 11.5 h (Mock) to 12.3 h (FL) or 12.8 h (dZnF) in TKO cells (Figure 2e). These results demonstrate that the ability of MdmX to suppress proliferation of p53-deficient cells does not require functional Mdm2.

Zn-Finger domain of MdmX maintains genome stability in mouse cells
We have reported that MdmX inhibits multipolar mitosis and chromosome loss in hyperploid p53-deficient cells. 18 To identify the domain(s) in MdmX protein responsible for promoting genome stability, we analyzed the DNA content and chromosome number of mouse DKO tumor cells transduced with MdmX expression constructs. Propidium iodide staining and metaphase spread analysis indicate that the expression of FL MdmX in DKO cells increased DNA content and chromosome numbers to the levels seen in tumor cells from p53-null mice retaining endogenous MdmX (Figures 3a and b). This MdmX-mediated increase in ploidy of DKO cells does not require the RING domain (see Figure 3a and b, FL and dRING panels), but it does require ZnF domain of MdmX (see Figure 3a and b, dZnF-dRING and dZnF panels). Metaphase spread analyses summarized in Figure 3c show that nearly the entire population of DKO cells with intact ZnF domain contained more than 60 chromosomes per cell, while the deletion of ZnF domain abolished this increase in chromosome number.
We have proposed that genome instability in hyperploid p53-deficient cells lacking MdmX arises from aberrant chromosome segregation and loss during multipolar mitosis, an event that occurs much more frequently in p53-null cells that also lack MdmX. 18 We therefore expect that the suppression of chromosome loss by the Zn-finger domain may be linked to the suppression of multipolar mitosis. The immunofluorescence analyses of spindle formation in DKO cells (Figure 3d) showed that the expression of MdmX with intact ZnF domain reduced the frequency of multipolar spindles by 50%. In contrast, expression of MdmX lacking ZnF domain failed to decrease the incidence of multipolar mitotic events. In Figure 3e The Zn-finger domain of MdmX maintains genome stability in p53-deficient human cells To determine if MdmX has a p53-independent role in regulating genome stability in human cells, we utilized MB157 breast tumor cells that bear a truncating deletion in TP53 gene 28 (Figure 5h) suggesting that the proliferation rate itself does not underlay the differences in tumorigenic potential observed in vivo.

DISCUSSION
MDMX is frequently amplified and overexpressed in human cancers, and experiments utilizing human cells and genetically altered mice have established that the oncogenic potential of MdmX is largely due to its ability to inhibit the p53 tumor suppressor. However, our lab has previously noted the existence of p53-independent roles for MdmX in suppressing cell proliferation and in maintaining genome stability 18,29 suggesting that MdmX may have both oncogenic activity (via p53 inhibition) and anti-oncogenic capabilities (in tumors mutated for p53). Here, we establish that the carboxy-terminal RING domain of MdmX prolongs the duration of cell cycle in p53-deficient cells, thereby suppressing cell proliferation. We also determine that MdmX-mediated suppression of multipolar mitosis correlates with increased genome stability in hyperploid mouse and human tumor cells. This p53-independent function of MdmX maps to the central Zn-finger domain of the MdmX protein that matches the consensus sequence for the Ran binding protein 2 (RanBP2) type zinc fingers (ScanProsite results, release 20.119 of 12 October 2015, score 9.077). The Zn-fingers domain in RanBP2 nucleoporin facilitates its binding to the small GTPase Ran protein 30 which is associated with nuclear envelop assembly and mitotic spindle morphogenesis. 31,32 It can be speculated that MdmX demonstrates its centrosome clustering function as one of the downstream targets for Run in spindle formation. As such, MdmX might counteract the centrosome clustering inhibitory activity of previously identified Ran target, nuclear mitotic apparatus protein (Nu-MA). 33,34 Both MdmX and Mdm2 proteins have conserved RanBP2 zinc finger-like consensus sequences. However, there is only 43% identity in amino acid sequence in Zn-finger domain between the two proteins that may account for the difference in their role in p53-independent tumorigenesis.
Increased cell proliferation and chromosomal instability are hallmarks of tumorigenesis. Since we determined that the suppression of proliferation and genome instability by MdmX are independently regulated, we were able to examine relative contributions of these activities to MdmX-mediated suppression of tumorigenesis. Utilizing standard orthotopic transplantation assays to measure tumor growth, and lung colonization by in vivo imaging to measure metastatic activity, we determined that MdmX-mediated genome stabilization underlies the ability of MdmX to reduce tumorigenic potential of cells mutated for p53. Although we cannot formally exclude a role for MdmX-mediated inhibition of cell proliferation in tumor progression, it is noteworthy that exogenous MdmX fails to alter proliferative capacity of MB231 cells in vitro, yet it clearly suppresses tumor formation and metastasis in vivo. It suggests that the tumorigenic potential of hyperploid, p53-mutant cells is not driven by a proliferative advantage but rather by the genomic instability imparted by loss of MdmX.
Computational quantification of somatic DNA alterations in human cancers reveals that tetraploidy generated by whole-genome doubling events occurs frequently during tumorigenesis. 35 Furthermore, polyploidy associated with increased genetic instability promotes aneuploidy and accelerated tumorigenesis in human cancers. 36,37 Pan-cancer analyses of somatic copy number alteration revealed that the whole-genome doubling occurred in 37% of cancers with average estimated ploidy of 3.31, suggesting the large-scale genome loss being associated with whole-genome doubling. 38 However, experiments with long-term culturing of tetraploid cells deriving from diploid cancer progenitors 39 revealed that, on per chromosome basis, tetraploidization by itself does not trigger additional instability. Dewhurst et al. 39 suggest that a tetraploid genome increases cell tolerance to segregation errors and aneuploidy. In agreement with this, our data provide direct evidence that it is not the presence of a large genome per se but rather the ability of these hyperploids to tolerate a greater range of genetic imbalances leading to the evolution of aggressive growth characteristics. We propose that MdmX exercises its tumor suppressive activity in hyperploid cells with compromised p53 by preventing multipolar mitosis and stabilizing the large genome of these cells, thus preventing chromosome loss. Loss of MdmX may trigger large-scale ploidy reductions (only tolerated in polyploid cells), reducing the ploidy to near-diploid levels until a new cycle of polyploidization occurs within the cell. Such a scenario of non-synchronized 'genomic breathing' in populations of tumor cells might easily contribute to tumor heterogeneity, and is concordant with the model of parallel evolution of different subclones within the same tumor. 40,41 Interestingly, overexpression of MdmX was recently shown to impair DNA damage response and promote chromosome and chromatid breaks in mouse embryo fibroblasts in vitro. 42 In contrast to those aberrations that occur at the level of individual chromosome(s), the genomic instability imparted by the loss of MdmX that we refer to represents the large-scale genome loss restricted to the polyploid cells that can lead to aneuploidy and tumorigenesis. These two quantitatively and mechanistically distinct MdmXmediated processes are not mutually exclusive and are likely dependent on cellular context.
Our findings also suggest that genome stabilization by MdmX is dependent upon the clustering of supernumerary centrosomes and the promotion of pseudo-bipolar mitosis. Although centrosome amplification and clustering is not limited to cancer cells, 43 it is believed that the centrosome clustering represents major route for hyperploid cancer cells with amplified centrosomes to escape multipolar mitosis and survive. Not surprisingly, identification and the inhibition of proteins involved in centrosome clustering has become an attractive anticancer strategy. 44,45 However, our experiments reveal that the majority of multipolar mitosis generates viable daughter cells with increased levels of aneuploidy and increased tumorigenic potential. Thus, our findings offer a caveat to proposed cancer treatment strategies that aim to promote multipolar mitosis. Furthermore, our data indicate that the loss of MdmX in p53-mutant cells promotes genome instability and cancer progression. As suggested by recent studies demonstrating the failure of anti-MdmX strategy in tumors expressing hypomorphic p53 mutant, 46 our data indicate that cancer treatments centered on MdmX inhibition should be employed only in cases where tumor retains functional p53.
Human tumor cell lines MDA-MB157 (obtained from ATCC) and MDA-MB231 (gift from Dr. Jane Lian, UMASS Medical School) were established from metastatic pleural effusion of human breast carcinomas. 48 MB231 cells were grown at 37°C with 5% CO 2 in DMEM with penicillin, streptomycin and 10% serum. For MB157 cells media was supplemented with 15% serum. Cell line MB231 has missense mutation in exon 8 resulting in mutant p53 protein 49 and it is tumorigenic in nude mice. 50 MB157 cells are p53-deficient due to the deletion in TP53 exon 4 resulting in the absence of p53 protein. 28 MB157 cells have hardly detectable levels of MDMX protein. 51 All cells were tested and were free of mycoplasma.

Culturing cells from tumors
Following tumor dissection, tumors were minced in high glucose DMEM supplemented with antibiotics and collagenase (2mg/ml), incubated at 37°C for 3 h with shaking, washed with PBS containing 5% FBS, resuspended in DMEM with antibiotics and 2% FBS and plated on collagen-coated plates. Cells were maintained in 2% FBS DMEM/F12 with antibiotics. Fibroblasts were removed by differential trypsinization. Once cells reached growth crises, cells were fed with low glucose 10% FBS DMEM.
Transfection of mouse and human cells Cells were grown in DMEM with no antibiotics until 70-80% confluent. Linearized MdmX-expression constructs were co-transfected with PGK promoter-puromycin drug selection marker using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA) for transfection of mouse and MDA-MB231 human cells or XtremeGene-9 (Roche, Mannheim, Germany) for MDA-MB157 human cells. Cells were split 24 h after transfection and selection for puromycin resistance started 24 h later. Pools of puromycin-resistant stably transfected clones were recovered for further analyses.
Quantitative real-time PCR Total mRNA was isolated using RNeasy kit (Qiagen, Hilden, Germany). For the first-strand cDNA synthesis SuperScript III First-Strand Synthesis System (Invitrogen) was used according to manufacturer's instructions. Quantitative PCR reaction was performed with SYBR Select Master Mix (Invitrogen) using the ABI-9300 PCR machine. The MdmX mRNA data were normalized to actin and the level in Mock control was assigned a value 1. Primers used for quantitative PCR are as follows: Mouse  . The duration of cell cycle was defined as the time from anaphase onset in mother cell manifested as the chromosome disjunction and/or sudden start of cell elongation till the anaphase onset in daughter cells.

Metaphase spreads and chromosome counting
Actively proliferating cells were incubated with Colcemid (Gibco; 0.02 μg/ml media) for 90 min, harvested and incubated in hypotonic solution (0.075 M KCl) for 15 min at 37°C and fixed with methanol-acetic acid. Drops of cell suspension were placed on microscope slides, dried, 4ʹ-6-diamidino-2-phenylindole stained and analyzed by fluorescent microscopy.
Orthotopic transplantation studies MDA-MB231 cells stably transduced with FL or MdmX deletion mutants were resuspended in PBS with 30% Matrigel (BD Biosciences, San Jose, CA, USA). Cells (4 × 10 6 ) in 200 μl were injected into the mammary fat pad of 4 weeks old NSG mice, five animals per group. Tumor growth was monitored externally and animals were killed 16 days after injection. Tumors were harvested and tumor volume was calculated: total volume = ½ (length x width 2 ). 54 All animal work in this manuscript was approved by the Institutional Animal Care and Use Committee of UMMS.

In vivo bioluminescence imaging
For tail vein injection, cells were resuspended in PBS at 1 × 10 6 cells/ml. Using a 26G needle, 200 μl of cells were injected into the tail vein of 8-week-old NSG mice, five animals per group. Immediately following injection, animals were imaged for the luciferase activity at time zero. D-Luciferin (Gold Biotechnology, St Louis, MO, USA; LUCK-500) was dissolved in DPBS (15 mg/ml), filter-sterilized and injected intraperitoneally (150 mg/kg body weight). Animals were anesthetized with isoflurane for 5 min and imaged from both dorsal and ventral position by Xenogen IVIS 100 Imaging System, Caliper Life Sciences in Optical Imaging Core Facility at UMMS. Mice were monitored for luciferase activity every other day for the first week and then weekly for the rest of the experiment. Bioluminescence images of upper dorsal region corresponding to the lung position were quantified using Living Image 2.60.1 software. The values for photon flux were normalized to those obtained at day 1.

Statistical analysis
The two-sample two-sided unpaired t-test was used to compare the mean between two experimental conditions with the significance level 0.05 and confidence interval 95%. Number of samples and P-values for specific group of experiments are provided in figure legends. The normality of data were tested by the stem and leaf plot or by Kolmogorov-Smirnov test (for time-lapse video microscopy data). For animal studies sample size was estimated using Mead's resource equation with the degrees of freedom between 10 and 20. No blinding or randomization methods were used.