The transcription accessory factor TIF1γ/TRIM33/RFG7/PTC7/Ectodermin functions as a tumor suppressor that promotes development and cellular differentiation. However, its precise function in cancer has been elusive. In the present study, we report that TIF1γ inactivation causes cells to accumulate chromosomal defects, a hallmark of cancer, due to attenuations in the spindle assembly checkpoint and the post-mitotic checkpoint. TIF1γ deficiency also caused a loss of contact growth inhibition and increased anchorage-independent growth in vitro and in vivo. Clinically, reduced TIF1γ expression in human tumors correlated with an increased rate of genomic rearrangements. Overall, our work indicates that TIF1γ exerts its tumor-suppressive functions in part by promoting chromosomal stability. Cancer Res; 75(20); 4335–50. ©2015 AACR.
Transcriptional intermediary factor 1 gamma (TIF1γ/TRIM33/RFG7/PTC7/Ectodermin) belongs to the evolutionarily conserved TIF1 family of nuclear factors. TIF1 proteins are involved in many biologic processes, such as stemness, embryonic development, and tumor suppression (1). The four identified members that compose this family (TIF1α/TRIM24, TIF1β/TRIM28, TIF1γ/TRIM33, and TIF1δ/TRIM66) share the TSS (TIF1 signature sequence) domain (2). These proteins are also characterized by several chromatin interaction domains, including B-boxes, coiled-coil domains, plant homeodomains (PHD), and bromodomains. The TIF1 proteins are also defined by the presence of a RING domain with E3-ubiquitin ligase activity and belong to the TRIM protein superfamily (tripartite motif). TIF1γ, whose activity and stability can be regulated by ubiquitination (3) and sumoylation (4), has been shown to be involved in various biologic functions, such as embryonic development (5–7), hematopoietic differentiation (8, 9), DNA repair (10), cell cycle regulation (11), and immune response (12). At the molecular level, TIF1γ is involved in chromatin remodeling and subsequent transcription modulation (3, 7, 10, 13, 14). Depending on the cellular context, TIF1γ has been described as a potent modulator of the TGFβ, acting either as a negative regulator (5, 14, 15–17) or a positive regulator of the pathway (12, 13, 18).
Several reports describe TIF1γ as a tumor-suppressor gene. Indeed, chromosomal rearrangements have been observed in the 1p13 locus containing TIF1γ gene in carcinomas (19), in hematopoietic neoplasia (20), in embryonic tumors (21), in conjunctive tissue tumors (22), and in nervous central system tumors (23). Moreover, decreased TIF1γ expression has been reported in different types of human tumors, including pancreatic ductal adenocarcinomas (24), breast tumors (17), non–small cell lung cancers (25), and hepatocellular carcinomas (HCC; ref. 26). Point mutations have not been described so far, possibly because epigenetic silencing may be primarily responsible for TIF1γ downregulation, as described in about 35% of chronic myelomonocytic leukemia (CMML) patients (27). In these tumors, TIF1γ is epigenetically silenced though promoter DNA methylation, and treatment of patient-derived cells with a demethylating agent is sufficient to restore its expression. In addition, mouse models have demonstrated the causal role of loss of Tif1γ in tumor progression. Indeed, we developed a genetically engineered mouse model that allowed us to demonstrate that conditional inactivation of Tif1γ could facilitate Kras-induced pancreatic carcinogenesis (24). The tumor-suppressive role of TIF1γ was subsequently confirmed by other groups in mouse models of cancers, such as CMML (27) and HCC (28). Finally, the closest family members of TIF1γ, i.e., TIF1α and TIF1β, have also been reported to behave as tumor suppressors, their hepatocyte-specific inactivation in mice resulting in HCC formation (28, 29).
Contrasting with the well-established link between TIF1γ deficiency and cancer, and despite recent advances in understanding the function of TIF1γ, little is known about the mechanisms through which TIF1γ exerts its tumor-suppressive role. We and others have proposed that the tumor-suppressive function of TIF1γ could result from its capacity to inhibit TGFβ-induced epithelial-to-mesenchymal transition (EMT; refs. 17, 26, 30, 31). However, we have also demonstrated in a transgenic mouse model of pancreatic cancer that the regulation of EMT by TIF1γ could not fully explain its tumor-suppressive effect (30), thus indicating that TIF1γ was able to regulate other oncosuppressive biologic functions. Recently, a novel function for TIF1γ during mitosis progression was identified, highlighting its role as part of the anaphase-promoting complex/cyclosome (APC/C; ref. 11). Mitotic progression relies on activation of the APC/C associated with the CDC20 protein, which promotes the segregation of sister chromatids and mitotic exit once chromosomes are correctly attached. This process is dependent upon inactivation of the spindle assembly checkpoint (SAC). SAC is a conserved mechanism in eukaryotes, which surveys the attachment of chromosomes to the mitotic spindle during the early steps of mitosis (32) and remains activated until microtubules are correctly attached to kinetochores, thus delaying anaphase onset by inhibiting APC/C. Sedgwick and colleagues reported that transient inactivation of TIF1γ in HeLa cells led to a blockade of cells in metaphase as a result of a defective APC/C–CDC20 complex and further sustained spindle checkpoint activation. However, this mechanism alone cannot account for the tumor-suppressive role of TIF1γ that we and others have previously reported (24, 27, 28), the long-term consequence of TIF1γ loss of function on cell cycle progression remaining unknown.
In the present study, we investigated the consequences of prolonged TIF1γ inactivation, better mimicking the in vivo situation found in tumors. Using Tif1γ−/−-immortalized mouse embryonic fibroblasts (MEF) and shTif1γ HRASV12/E1A-transformed NMuMG (normal murine mammary gland) cells, we explored the consequence of prolonged TIF1γ inactivation on mitotic cell division and tumor progression. We demonstrated in vitro and in vivo that, in contrast with transient Tif1γ inactivation, prolonged inactivation of Tif1γ caused the accumulation of mitotic defects leading to chromosomal abnormalities characterized by increased aneuploidy and chromosome rearrangements, which are associated with the acquisition of more aggressive properties. In addition, we confirmed the positive correlation between low levels of TIF1γ and increased chromosomal instability (CIN) in human tumors. Altogether, this study demonstrates that loss of TIF1γ results in increased tumor aggressiveness by promoting mitotic defects.
Materials and Methods
Details about additional procedures are provided in Supplementary Materials.
Tif1γL/L mice bearing the homozygous Tif1γ conditional knockout allele were described previously (24). Five-week-old female Nude mice were purchased from Charles River. Mice were housed and bred in a specific pathogen-free animal facility. All experiments were validated by the local animal ethic evaluation committee (CECCAPP) in accordance with the animal care guidelines of the European Union and French laws.
Tif1γL/L-immortalized MEFs were isolated from E13.5 Tif1γL/L mutant embryos and immortalized according to a standard 3T3 protocol after 15 to 20 passages and were cultured in DMEM, supplemented with 0.03% l-glutamine and containing 10% FBS, 1% nonessential amino acids, and P/S (100 IU/mL penicillin and 100 μg/mL streptomycin sulfate). NMuMG epithelial cells were purchased from the ATCC and were cultured in DMEM, supplemented with 0.03% l-glutamine and containing 10% FBS, P/S, and 10 μg/mL insulin. All cells were propagated at 37°C under 5% (v/v) CO2 atmosphere. Nocodazole was purchased from Sigma-Aldrich and was used at a 600 ng/mL final concentration. Colcemid was purchased from Life Technologies and used at a final concentration of 500 ng/mL, whereas for the dose–response experiments, it was used at a final concentration ranging from 125 ng/mL to 3,000 ng/mL. MG132 was purchased from Sigma-Aldrich and was used at a 10 μmol/L final concentration.
Tif1γL/L-immortalized MEFs were transduced either with an empty (pBABE-empty) or a Cre-expressing (pBABE-Cre) retroviral vector and cultured in the presence of 3 μg/mL puromycin (Invivogen) to respectively obtain stable control- and Tif1γ−/−-immortalized MEFs. NMuMG cells were transduced either with an empty retroviral vector (pRS-empty) or a retroviral vector expressing an shRNA directed against Tif1γ (pRS-shTif1γ) and cultured in the presence of 10 μg/mL puromycin to respectively obtain stable control- and shTif1γ-NMuMG cells. Next, transformation of these NMuMG cell lines was performed by retroviral transduction with vectors encoding HRASV12 and E1A oncoproteins to generate respectively control- and shTif1γ-transformed NMuMG cells. Stably transduced cells were selected by culturing in the presence of 800 μg/mL G418 (Life Technologies) and 200 μg/mL hygromycin B (Life Technologies).
Analysis of human cancer cell lines and human carcinomas
Copy-number variation (CNV) and TIF1γ gene expression were analyzed for 990 human cancer cell lines available in the Cancer Cell Line Encyclopedia (CCLE; http://www.broadinstitute.org/ccle/home). Human breast, pancreatic, colorectal, and kidney tumors available in The Cancer Genome Atlas (TCGA) and in the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) were selected based on their tumor cells content (≥60% for TCGA and ≥70% for METABRIC), except for pancreatic adenocarcinomas, for which all samples were included due to the few number of available samples. TIF1γ expression levels were categorized into three classes for CCLE human cell lines (Fig. 7A) and for TCGA tumors in breast (Fig. 7B), in pancreas (Fig. 7D), and in kidney (Fig. 7E) according the following method. Tumors were ranked from lowest to highest according to their TIF1γ expression. The bottom 33% relative to their expression of TIF1γ was arbitrarily set as the lowest class. The top 33% of cells relative to their expression of TIF1γ were arbitrarily set as the highest class, whereas the remaining 33% between lowest and highest class were arbitrarily set as the intermediate class. For the METABRIC dataset (Fig. 7C), breast tumors were ranked from lowest to highest according to their TIF1γ expression. The bottom 20% relative to their expression of TIF1γ was arbitrarily set as the lowest class. The top 20% of cells relative to their expression of TIF1γ were arbitrarily set as the highest class, whereas the remaining 60% between lowest and highest class were arbitrarily set as the intermediate class. Percentiles varied depending on the number of samples available in different datasets.
CNV, TIF1γ gene expression, and TP53 gene status were analyzed for 131 human breast carcinomas available in TCGA. CNV and TIF1γ gene expression were analyzed for 120 human pancreatic adenocarcinomas and for 406 human clear cell renal cell carcinomas (CCRCC) available in TCGA. TIF1γ gene status and CIN or microsatellite instability (MIN) status were analyzed for 202 human colorectal carcinomas available in TCGA. CNV and TIF1γ gene expression were analyzed for 965 human breast carcinomas available in the METABRIC. Human colorectal carcinomas were divided into CIN and MIN phenotype based on their whole-genome hyper- or hypo-mutation status (<500 for CIN; ≥500 for MIN). For CCLE cell lines and TCGA breast and renal tumors, the provided log2-transformed copy number (CN) value was used to determine the wild-type or altered status of CN (CN > 0.3 or < –0.3). For TCGA colorectal tumors, the provided log2-transformed CN value was used to determine the wild-type or deleted status of TIF1γ (CN < −0.3). For TCGA pancreatic tumors, the percentage of CNV was obtained from the cBioPortal for Cancer Genomics Portal (33). For METABRIC breast tumors, fraction genome altered (FGA) was calculated as follows: For each segment i, CNi is given by CN = log2 (sample intensity/reference intensity), L(i) is the length of segment i, and T = 0.3 is the threshold value of the CNi above which the segments are considered altered. In other words, FGA is the ratio of the sum of the lengths of all segments with signal above the threshold to the sum of all segment lengths.
For cytology and histology quantifications, cell proliferation assays, luciferase reporter assays, chorioallantoic membrane assays, allografts in Nude mice and qRT-PCR, statistical significance were tested using the Student t test. Statistical significance of soft-agar assay results was tested using the Pearson χ2 test. CNV and gene expression data were represented as Tukey box plots using GraphPad Prism 6 software (GraphPad; the central point represents the median; the bottom and top of the box represent the 25th and 75th percentiles; the whiskers represent the minimum and maximum values excluding outliers, which represent any data point that are more than 1.5 interquartile ranges below the 25th percentile or above the 75th percentile). Statistical significance was tested using the Student t test. TP53 and TIF1γ status data were represented as percentage of tumors presenting TP53 mutation or TIF1γ deletion. Statistical significance was tested using the Pearson χ2 test. All statistical tests were considered to be statistically significant when P < 0.05.
Prolonged inactivation of Tif1γ in MEFs results in accumulation of nuclear abnormalities and inactivation of the spindle checkpoint
In an effort to understand the long-term consequences on cell proliferation resulting from Tif1γ inactivation, we generated immortalized MEFs harboring a homozygous deletion for Tif1γ (24). For that, we immortalized Tif1γL/L MEF lines according to a classical 3T3 protocol. These cells were further transduced by a retrovirus containing either a vector expressing the Cre recombinase (Tif1γ−/−-immortalized MEFs) or a retrovirus containing an empty vector (control-immortalized MEFs). We observed that the inactivation of Tif1γ in immortalized MEFs rapidly resulted in impaired cell proliferation (Supplementary Fig. S1A) characterized by a decreased capacity of Tif1γ−/−-immortalized MEFs to progress through mitosis as illustrated by their accumulation in G2–M phases of the cell cycle (Supplementary Fig. S1B). Classically, SAC delays anaphase onset by inhibiting the APC/C complex until all kinetochores and microtubules are correctly attached. Anaphase onset is triggered by the APC/C associated with the CDC20 protein, which induces the proteasome-mediated destruction of CYCLIN A, CYCLIN B1, and SECURIN (32). The proteolysis of SECURIN releases active SEPARASE, which cleaves COHESIN thereby eliminating sister-chromatid cohesion and triggering the segregation of sister chromatids to the opposite poles of the mitotic spindle. Consistent with a sustained activation of the SAC, Western blotting experiments showed that the first answer of Tif1γ−/−-immortalized MEFs was to accumulate SAC components (BUBR1, AURORA B), APC/C-CDC20 targets (CYCLIN A, CYCLIN B1, SECURIN), and the mitotic marker phosphorylated Serine-10 HISTONE H3 (p-H3S10, a readout for AURORA B kinase activity; ref. 34; Supplementary Fig. S1C). We also observed in Tif1γ−/−-immortalized MEFs an increased senescence, a common mechanism that is observed after prolonged mitotic blockade and further mitosis catastrophe (Supplementary Fig. S1D; ref. 35). Our results support a recent work published by Sedgwick and colleagues, which demonstrated that TIF1γ was important for progression of HeLa cells through mitosis by activating APC/C (11). Indeed, they have reported that transient inactivation of TIF1γ could lead to extended mitotic blockade and accumulation of cells in mitosis as a result of sustained SAC activation. However, after this first prolonged delay, we demonstrated that Tif1γ−/−-immortalized MEFs subsequently regained a proliferation rate similar to control cells (Fig. 1A). Cell cycle analysis of control- and Tif1γ−/−-immortalized MEFs by flow cytometry using propidium iodide DNA staining, after this long-term culture, revealed an identical distribution of cells in G1, S, and G2–M phases (Fig. 1B). Thus, the accumulation in G2–M phases was severely compromised. Despite the equivalent proliferation rate and an identical distribution of the cells in the different phases of the cell cycle, cytological analyses of Tif1γ−/−-immortalized MEFs revealed severe mitotic abnormalities. Immunofluorescence staining of γ-TUBULIN showed that Tif1γ−/−-immortalized MEFs presented an increased proportion of cells with more than two centrosomes compared with control cells (44.8% vs. 17.4%; Fig. 1C). Immunofluorescence staining of α-TUBULIN showed that this abnormal number of centrosomes was associated with multipolar spindles along with cytokinesis defects (Fig. 1D). We then performed Western blotting experiments, which revealed in Tif1γ−/−-immortalized MEFs a clear decrease of steady-state levels of CYCLIN A, CYCLIN B1, and SECURIN (Fig. 1E). The steady-state level of BUBR1, a major component of the SAC, was also significantly decreased in the absence of TIF1γ, evocating an attenuated SAC (Fig. 1E). Altogether, these results strongly suggest that sustained inactivation of Tif1γ in immortalized MEFs results in SAC dysfunction, leading to the accumulation of mitotic abnormalities. To further confirm that long-term ablation of Tif1γ resulted in SAC defects, cells were treated with colcemid and nocodazole, two spindle microtubule-disrupting agents that maintain the spindle checkpoint active. As expected, flow cytometry analysis revealed that control-immortalized MEFs accumulated in G2–M phases in the presence of these poisons (74.4% with colcemid; 82.6% with nocodazole; Fig. 1F). In contrast, Tif1γ−/−-immortalized MEFs were significantly less sensitive to colcemid and nocodazole as illustrated by the decreased proportion of cells blocked in G2–M phases (45.0% with colcemid; 67.0% with nocodazole), confirming that SAC is less effective (Fig. 1F). The compromised capacity of Tif1γ−/−-immortalized MEFs to arrest in mitosis in response to spindle poisons was confirmed by the reduced accumulation of the mitotic marker p-H3S10 in the presence of colcemid visualized by Western blotting experiments (Fig. 1G) and immunofluorescence (Fig. 1H). Unrestrained mitosis in the presence of microtubule-disrupting drugs is expected to lead to chromosomal defects. Toluidine blue staining and subsequent quantification of multinuclei revealed an increased percentage of multinucleated Tif1γ−/−-immortalized MEFs treated with colcemid or nocodazole (Fig. 1I). Altogether, these results indicate that long-term inactivation of Tif1γ in MEFs leads to an attenuation of the SAC activity, a condition that may facilitate the accumulation of nuclear defects.
Inactivation of Tif1γ in epithelial-transformed NMuMG cells results in severe ploidy defects
In order to better understand the functional consequences of long-term TIF1γ inactivation in cancer, we extended our work to a cellular model amenable to analyze epithelial tumor progression. We generated control-transformed NMuMG cells (stably expressing HRASV12 and E1A oncogenic proteins) and shTif1γ-transformed NMuMG cells (stably expressing HRASV12, E1A, and an shRNA targeting Tif1γ). The expression of HRASV12 and E1A oncoproteins, along with the extinction of endogenous expression of TIF1γ, was validated by Western blotting experiments (Fig. 2A). Cell proliferation assay over a 5-day period revealed a similar growth rate between control- and shTif1γ-transformed NMuMG cells (Fig. 2B), mimicking long-term Tif1γ−/−-immortalized MEFs (Fig. 1A). We next investigated whether the absence of Tif1γ resulted in mitotic defects in this transformed cellular model as observed in immortalized MEFs. DAPI staining revealed that the shape and the size of nuclei were very heterogeneous in shTif1γ-transformed NMuMG cells compared with control cells (Fig. 2C). We therefore sought to finely characterize the nuclear defects that accumulate in the absence of TIF1γ. Costaining of nuclei and cortical actin revealed a significant increase of chromosomal abnormalities in shTif1γ-transformed NMuMG cells, such as tripolar alignments (0.22% vs. 0.03%), lagging chromosomes (0.50% vs. 0.03%), chromatin bridges (4.9% vs. 0.5%), multinucleated cells (6.1% vs. 1.2%), and micronuclei (9.2% vs. 3.6%; Fig. 2D). Note that the proportion of cells in mitosis is similar in control- and shTif1γ-transformed NMuMG cells (5.5% vs. 4.7%; Fig. 2D). Along with immortalized MEFs, these results indicate that transformed NMuMG cells display many mitotic abnormalities, continue to proliferate, and accumulate nuclear defects in the sustained absence of TIF1γ.
We next investigated whether the nuclear abnormalities described above were associated with a compromised capacity of the cells to arrest in mitosis in the presence of spindle poisons. As observed after toluidine blue staining in MEFs (Fig. 1I), a higher proportion of abnormal nuclei in shTif1γ-transformed NMuMG cells was observed compared with control cells after nocodazole treatment (70.3% vs. 28.2%; Fig. 3A). Notably, shTif1γ-transformed NMuMG cells present a more elevated rate of abnormal nuclei even in the absence of nocodazole treatment (20.9% vs. 5.3%; Fig. 3A). Flow cytometry analysis showed that untreated shTif1γ-transformed NMuMG cells displayed a higher steady-state number of cells with a >4n DNA content compared with untreated control cells (10.7% vs. 5.7%; Fig. 3B). This result is consistent with the large nuclei observed after DAPI staining (Fig. 2C and D), a phenotype classically associated with polyploidy (36, 37). The absence of TIF1γ also severely compromised the capacity of these cells to arrest in G2–M phases in response to colcemid (29.6% vs. 67.9%) and nocodazole (29.2% vs. 68.5%; Fig. 3B). Notably, in the presence of spindle poisons, shTif1γ-transformed NMuMG cells accumulate as a significant polyploid cell population with a >4n DNA content. In order to more directly prove that the lack of TIF1γ was associated with resistance to spindle poisons, we performed a dose–response curve with colcemid (Fig. 3C). As attested by flow cytometry analysis, whereas control cells massively arrest in G2–M phases starting from the lowest dose of colcemid (125 ng/mL), shTif1γ-transformed NMuMG cells did not significantly arrest in G2–M phases even at the highest dose of colcemid (3,000 ng/mL). In support to this, the proliferation of shTif1γ-transformed NMuMG cells was modestly decreased (∼20%) when cultured in the presence of colcemid at any concentration, whereas the proliferation of control cells significantly dropped down more (∼50%; Fig. 3D). These results demonstrate that shTif1γ-transformed NMuMG cells are insensitive to spindle poison clearly evocating an impaired SAC-mediated arrest.
Inactivation of Tif1γ in epithelial-transformed NMuMG cells results in mitotic checkpoints alterations
We further compared the timing of mitotic exit between control- and shTif1γ-transformed NMuMG cells. Mitotic enrichment by shake-off, release in fresh medium, and collection of cells at different time points after release and flow cytometry analysis of p-H3S10 staining revealed that shTif1γ-transformed NMuMG cell re-entered the cell cycle, more rapidly than control cells (Fig. 4A). The premature mitosis exit in shTif1γ-transformed NMuMG cells is a hallmark of an attenuated SAC (38). During metaphase, all chromosomes are aligned on metaphasic plate through centromere capture by mitotic spindle. In order to detect chromosome alignments defects, another hallmark of defective SAC, we stained the cells with DAPI, anti–α-TUBULIN antibodies, and CREST sera recognizing a centromeric antigen. We observed a high frequency of TIF1γ-depleted cells with chromosome alignments defects (precocious dissociation or congression defects; 71%) as attested by the location of centromeres (and thus chromosomes) between the spindle poles and the metaphasic plate (Fig. 4B). The compromised SAC is also supported by the fact that BUBR1, a core component of the MCC, displays reduced kinetochore localization visualized by immunofluorescence in shTif1γ-transformed NMuMG cells compared with control cells (29.0% vs. 80.4%; Fig. 4C). In accordance with a compromised capacity to arrest in mitosis, Western blotting experiments revealed that depletion of TIF1γ (Fig. 5A), resulted in (i) a significant decrease of the steady-state level of mitotic marker p-H3S10 along with decreased levels of CYCLIN A, CYCLIN B1, and SECURIN, which are targets of the APC/C-CDC20 complex; (ii) a reduced expression of SAC components, such as BUBR1, BUB1, and AURORA B; (iii) a less pronounced accumulation of BUBR1, BUB1, and AURORA B (SAC components); p-H3S10 (AURORA B target); and CYCLIN A, CYCLIN B1, and SECURIN (APC/C-CDC20 targets) after colcemid treatment in shTif1γ-transformed NMuMG cells; (iv) an increased γ-TUBULIN expression in shTif1γ-transformed NMuMG cells consistently with the supernumerary centrosomes we observed in Tif1γ−/−-immortalized MEFs (Fig. 1C). All these observations are evocative of an SAC deficiency after constitutive prolonged Tif1γ inactivation. Western blotting experiments revealed that BUBR1, BUB1, AURORA B, CYCLIN A, CYCLIN B1, and SECURIN level was still reduced in shTif1γ-transformed NMuMG compared with control when they are cultured in the presence of the MG132 proteasome inhibitor (Fig. 5B). This indicates that lower levels of these proteins do not exclusively result from protein stability regulation. Nevertheless, in shTif1γ-transformed NMuMG cells, the protein level is slightly increased in the presence of MG132 compared with basal level, attesting an activity of APC/C-CDC20 complex. Conversely, qRT-PCR experiments revealed a significant reduced expression of related transcripts in shTif1γ-transformed NMuMG cells compared with control cells, evocating a downregulation at the mRNA level (Fig. 5C). Altogether, premature mitosis exit, downregulation of SAC key effectors, decreased kinetochore localization of BUBR1 and chromosome misalignments, evocating either precocious dissociation of sister chromatids or congression defects, unequivocally demonstrate that sustained Tif1γ inactivation results in a defective SAC. Release in fresh medium after nocodazole treatment and mitotic shake-off revealed that shTif1γ-transformed NMuMG cells were capable of cycling as a population with a >4n DNA content, as attested by flow cytometry using propidium iodide DNA staining (Fig. 5D). This observation is highly evocative of an attenuated postmitotic checkpoint (also known as tetraploid checkpoint). This failsafe program is dependent on an active P53/P21 pathway allowing to block cells in G1 phase and eliminates those with aberrant chromosome numbers or chromosome abnormalities resulting from segregation errors, thus guaranteeing the integrity and the equal repartition of the genetic material between daughter cells (39). Then, we tested the activity of P53, the main effector of this checkpoint. Increased P53 activity is classically associated with increased phosphorylation on its serine 15 residue (p-P53-S15; ref. 40). Western blotting experiments showed that p-P53-S15 level in shTif1γ-transformed NMuMG was significantly reduced compared with control cells, both in the absence and the presence of colcemid (Fig. 5E). We next explored the expression of the cyclin-kinase inhibitor P21/WAF1/CIP1, one of the main targets of transcriptionally active P53 (39, 41). Western blot analysis revealed that P21 protein levels were markedly increased in control-transformed NMuMG cells treated with colcemid as expected in the presence of an activated postmitotic checkpoint (Fig. 5E). In comparison, P21 protein levels in shTif1γ-transformed NMuMG both at the steady-state or after colcemid treatment were clearly lower compared with control cells, in accordance with an attenuated postmitotic checkpoint in the absence of TIF1γ (Fig. 5E). Reporter assays with a plasmid containing the luciferase gene driven by p21 promoter transiently transfected in control-transformed NMuMG cells revealed that colcemid could increase luciferase expression (Fig. 5F). This observation is consistent with an efficient activation of the P53/P21 postmitotic checkpoint. Conversely, shTif1γ-transformed NMuMG cells transiently transfected with the p21 reporter presented a 2-fold decrease in steady-state luciferase expression and did not respond to colcemid (Fig. 5F). Finally, we demonstrated by qRT-PCR that accumulation of p21 mRNA as well as other known transcriptional targets of P53, such as Puma and Gadd45, after colcemid treatment was significantly compromised in shTif1γ-transformed NMuMG cells (Fig. 5G). This suggests that the P53/P21 pathway is impaired in shTif1γ-transformed NMuMG cells. Altogether, proliferation of cells with a >4n DNA content, reduced activity of P53, and downregulation of its main transcriptional targets clearly demonstrates that sustained Tif1γ inactivation results also in an attenuated postmitotic checkpoint.
As a conclusion, these results demonstrate that sustained Tif1γ inactivation in transformed-NMuMG cells results in SAC inactivation, attenuation of the postmitotic checkpoint, proliferation after premature mitotic exit in the presence of abnormal chromosome segregation, and the accumulation of nuclear and ploidy defects.
Increased aggressiveness of shTif1γ-transformed NMuMG cells
Genomic instability plays an important role both in cancer development and in the response to therapies. Hence, we sought to determine whether the absence of TIF1γ could modulate the transforming and tumorigenic properties of transformed NMuMG cells. RAS and E1A were reported to constitute an inefficient oncogenic combination, and loss of P21 was shown to significantly cooperate with these two oncoproteins in promoting cell transformation (42). In support of our previous statement, depletion of TIF1γ is associated with decreased P21 expression (Fig. 5E–G) and increased aggressive properties of the cell line, as exemplified by their loss of contact growth inhibition (Fig. 6A), and the increased number and size of colonies observed in a soft-agar colony assay (Fig. 6B). Furthermore, despite tumor take is 100% in control- and shTif1γ-transformed NMuMG cells, TIF1γ depletion was found to drastically accelerate tumor growth when transformed NMuMG cells were either injected into chicken chorioallantoic membranes (CAM; Fig. 6C) or subcutaneously engrafted into Nude mice (Fig. 6D). Finally, histologic analysis demonstrated a significant increase in the proportion of abnormal mitosis in tumors arising from shTif1γ-transformed NMuMG cells compared with control-transformed NMuMG cells (57.7% vs. 28.6%; Fig. 6E), indicating that shTif1γ-transformed NMuMG-derived tumors grow faster with a higher index of mitotic defects. Altogether, these data are consistent with the concept that TIF1γ loss of function increases tumor aggressiveness by allowing chromosomal abnormalities.
Downregulation of TIF1γ in human tumors is correlated with increased CIN
We next assessed the clinical relevance of TIF1γ downregulation in human tumors, in which the chromosomal abnormalities are classically associated with quantifiable chromosomal rearrangements—(deletions, insertions, duplications, and complex multi-site variants) collectively termed CNV. We classified the cancer cell lines from different organs present in the CCLE (n = 990) according to the expression level of TIF1γ (Fig. 7A, top) and examined the CNV in each category. Our analysis revealed that human cell lines with a “low” expression of TIF1γ presented an increased CNV compared with cells with an “intermediate” or a “high” expression of TIF1γ (Fig. 7A, bottom). Because the 1p13.1 locus, which contains TIF1γ, was previously reported to be deleted in some human breast tumors (43), and because the NMuMG cell line is of mammary origin, we decided to focus our genetic analysis on human breast tumors. We examined in TCGA the CNV in human primary breast tumors (n = 131) according to the expression level of TIF1γ (Fig. 7B, top). Again, we observed that tumors with the lowest level of TIF1γ were those with the highest CNV (Fig. 7B, middle). Interestingly, low expression level of TIF1γ was found to be associated with an increased rate of TP53 mutations (Fig. 7B, bottom). The positive correlation between low TIF1γ expression and increased CNV was confirmed in the METABRIC annotated public dataset of 965 human breast carcinomas (Fig. 7C). The correlation between low expression of TIF1γ and increased CNV was confirmed in other cancers, such as pancreatic adenocarcinoma (n = 120; Fig. 7D) and in CCRCC (n = 406; Fig. 7E). There are two forms of genomic instability that differ from their underlying molecular mechanisms: MIN and CIN. MIN is characterized by microdeletions and microduplications of dinucleotide or trinucleotide repeats (microsatellites) at specific polymorphic sites in the genome and is classically considered to present a hypermutated phenotype due to inherited mutations in mismatch DNA repair genes (44). The second form, CIN, is characterized by large genomic deletions/amplifications, chromosomal translocations, and aneuploidy, and is classically considered to present a hypomutated phenotype (compared with MINs) and to occur as a consequence of chromosome segregation errors (44). Based on our results, TIF1γ depletion is expected to be mainly associated with the CIN phenotype. To investigate further whether or not a relationship existed between TIF1γ depletion, genomic instability, and cancer progression, we analyzed human colorectal carcinomas, which are characterized by their high rate of CNV and can be divided into MIN tumors (15% of colorectal carcinomas) and CIN tumors (85% of colorectal carcinomas). Analysis of TCGA colorectal tumors (n = 202) revealed that MIN (n = 30, hypermutated tumors with >500 mutations) and CIN (n = 172, hypomutated tumors with <500 mutations) respectively represented 14.8% and 85.2% of all colorectal tumors, a proportion in accordance with the literature (Fig. 7F, top pie chart). While a very small proportion of MIN tumors (3%) displayed a deleted TIF1γ locus, this proportion significantly increased in CIN tumors (23%; Fig. 7F, top pie chart). In support of this observation, a large majority of tumors with a deleted TIF1γ locus (n = 41) had a CIN phenotype (98%), whereas this percentage significantly dropped (82%) when the TIF1γ locus is intact (n = 161; Fig. 7F, bottom pie chart).
These data clearly indicate that genomic deletion of TIF1γ in colorectal carcinomas is preferentially associated with the CIN phenotype. Altogether, genetic analysis of human tumors revealed that decreased expression of TIF1γ or the deletion of TIF1γ-containing locus was associated with increased rate of chromosomal rearrangements, likely resulting from segregation errors due to an impaired spindle checkpoint and an attenuated P53/P21-dependent postmitotic checkpoint.
We and others have demonstrated that TIF1γ was a tumor suppressor (24, 27–29), whose mechanism of action has remained elusive. In this study, we demonstrated that stable Tif1γ inactivation resulted in SAC and postmitotic checkpoint attenuation, leading to the accumulation of severe chromosomal abnormalities. As a result, Tif1γ-inactivated cells present mitotic defects increasing their tumor aggressiveness in animal models. Finally, we observed that low TIF1γ expression was associated with increased CIN in different types of human tumors. Therefore, this work highlights an original mechanism by which TIF1γ behaves as a tumor suppressor through its role in the control of mitosis, whose impairment may represent a major tumor-suppressive process.
First of all, we revealed here that the immediate consequence of TIF1γ depletion in different cell types (primary and immortalized MEFs, transformed or immortalized epithelial cells) resulted in a proliferation arrest, mitotic blockade, accumulation of SAC components and APC/C-CDC20 targets, and increased mitotic catastrophes illustrated by senescence and/or apoptotic cell death (35, 45). These defects were demonstrated in immortalized MEFs (Supplementary Fig. S1A–S1D) and confirmed in primary MEFs (Supplementary Fig. S2A–S2C), HEK-293T (Supplementary Fig. S3A–S3C), and HepG2 cells (Supplementary Fig. S3D and S3E). These observations are in accordance with the work published by Sedgwick and colleagues in HeLa cells (11). Originally, we show here that the long-term consequence of TIF1γ inactivation, better mimicking the situation found in human tumors, is radically different. Using Tif1γ−/−-immortalized MEFs (Fig. 1A) and shTif1γ HRASV12/E1A-transformed NMuMG cells (Fig. 2B), we found that long-term Tif1γ-deficient cells eventually display normal proliferation rates. Precise characterization of these cells revealed that normal proliferation was accompanied with the accumulation of various mitotic and nuclear abnormalities, such as tripolar alignments, cytokinesis bridges, multi- and micro-nuclei, or lagging chromosomes (Figs. 1C and D and 2D), suggesting that proliferation was recovered at the expense of mitotic control. In line with this, the sustained absence of TIF1γ was found associated with hallmarks of SAC inactivation sequentially observed during progression in mitosis: (i) a decreased steady state of SAC components (BUBR1, BUB1, and AURORA B) by Western blotting experiments (Figs. 1E, 5A and B) and qRT-PCR (Fig. 5C); (ii) a reduced recruitment of the MCC core component BUBR1 at the kinetochores upon colcemid treatment (Fig. 4C); (iii) a defective alignment of chromosomes (Fig. 4B); (iv) a decreased steady state of APC/C-CDC20 targets (CYCLIN A, CYCLIN B1, SECURIN) by Western blotting experiments (Figs. 1E and 5A and B) and qRT-PCR (Fig. 5C); (v) a shortening of the mitosis (Fig. 4A); (vi) an increased resistance to spindle poisons (colcemid or nocodazole), which normally induced metaphase blockade by maintaining the SAC active (Figs. 1F, 3B and C, and 5D); (vii) the emergence of a polyploid cell population both in the absence and in the presence of spindle poisons treatments (Figs. 3B and C and 5D).
SAC attenuation in Tif1γ-deficient cells described above is however not sufficient to explain that these cells recover a proliferation rate similar to control cells. Indeed, cells experiencing defective chromosome segregation leading to chromosomal abnormalities upon mitotic exit should arrest proliferation in the G1 phase of the next cell cycle as a result of the P53/P21-dependent postmitotic checkpoint (46). Actually, postmitotic checkpoint inactivation, that is frequently observed in cancer, allows cells that accomplished abnormal mitosis to keep proliferating, thus leading to the onset and the spreading of a cell with growth advantages. In this context, we further demonstrated an attenuated P53-dependent postmitotic checkpoint in shTif1γ-transformed NMuMG cells, through (i) a proliferation with a >4n DNA content (Fig. 5D); (ii) a decreased level of the active form of P53 (p-P53-S15) in absence or in presence of spindle poisons (Fig. 5E); (iii) an altered P53 transcriptional activity visualized in a reporter assay (Fig. 5F); and (iv) a compromised expression of P53-target genes such as p21, Puma, and Gadd45 both in untreated and in colcemid-treated shTif1γ-transformed NMuMG cells (Fig. 5G).
Altogether, our findings clearly indicate that long-term inactivation of TIF1γ results in attenuation of both the spindle and the postmitotic checkpoints, commonly associated with elevated genomic instability rate in human tumors (47). Indeed, CIN and aneuploidy resulting from uncorrected mitotic errors due to an impaired spindle checkpoint and postmitotic checkpoint inactivation is a common mechanism, which promotes tumorigenesis and increase tumor aggressiveness. Therefore, we further addressed a putative causal role of TIF1γ depletion in tumorigenesis. In vitro and in vivo, we showed that shTif1γ-transformed NMuMG cells presented enhanced aggressive properties, illustrated by their abilities to form colonies of large diameters in soft agar (Fig. 6B) and to generate tumors very rapidly when injected into chicken CAM (Fig. 6C) or subcutaneously engrafted into Nude mice (Fig. 6D). We also showed that tumors arising from shTif1γ-transformed NMuMG cells presented an increased number of aberrant mitotic figures (Fig. 6E), confirming the importance of Tif1γ for mitosis control in vivo. Finally, we provide evidence that our observations are transposable to human tumors. We performed computational analyses of human tumors and cell lines available in public databases in order to explore the consequence of TIF1γ deficiency in human tumors. The CCLE, TCGA and the METABRIC datasets allowed us to demonstrate that TIF1γ low expression was correlated with accumulation of chromosomal rearrangements (increased CNV, a hallmark of CIN) and increased rates of TP53 mutations, providing a clinical relevance for decreased expression of TIF1γ in human tumors (Fig. 7A–F). Furthermore, this high CNV amount is observed in cells of different tumor origins from the CCLE database, together with breast, pancreatic, renal, and colorectal primary carcinomas from TCGA and METABRIC datasets. Thus, we put in evidence that, in human neoplasia, low TIF1γ expression is associated with high genomic instability, validating our observations in in vitro and in vivo systems.
A previous study reported that TIF1γ could interact with APC/C-CDC20 and positively regulate E3-ubiquitin ligase activity of the complex (11). In their study, the authors proposed that mitotic blockade observed after transient inactivation of TIF1γ resulted from compromised APC/C-CDC20 activity due to the absence of TIF1γ in the complex and SAC activation. In the present study, we reported that after prolonged mitotic arrest, some cells are able to adapt by bypassing SAC activation and exit mitosis through downregulation of the expression of key players of the SAC (BUB1, BUBR1, and AURORA B) and decreased recruitment of BUBR1 at kinetochores. This probably does not permit to reach the required threshold for a fully active and efficient SAC. We also reported that prolonged inactivation of Tif1γ resulted in the decrease of CYCLIN A, CYCLIN B1, and SECURIN, which are APC/C-CDC20 targets proteins. It is noteworthy that MG132 treatment increases the level of CYCLIN A and CYCLIN B1 proteins both in the presence and the absence of TIF1γ, indicating that APC/C is still active in long-term Tif1γ-deficient cells. This observation argues that TIF1γ is dispensable, at least in part, for APC/C-CDC20 E3-ubiquitin ligase activity. This supports our original model that TIF1γ negatively regulates SAC activity rather than positively regulated APC/C-CDC20 activity. Indeed, while a positive effect of TIF1γ on APC/C-CDC20 cannot be excluded, we propose that TIF1γ mainly behaves as an inhibitor of the SAC. This model would explain why stable Tif1γ-deficient cells are still able to progress through mitosis as a consequence of downregulated activity of SAC resulting from diminished level of SAC protein expression (BUB1, BUBR1, AURORA B), allowing then the recovery of APC/C-CDC20 activity. This alternate mechanism, also proposed by Sedgwick and colleagues (11), might explain several unexpected results reported by these authors. First, they demonstrated that TIF1γ preferentially interacted with APC/C-CDC20 when the SAC is activated, strongly suggesting an active role of TIF1γ on APC/C-CDC20 before the SAC is satisfied. They also demonstrated that BUBR1 knockout could rescue the mitotic arrest initially caused by TIF1γ inactivation, constituting clear evidence that TIF1γ is not mandatory for APC/C-CDC20 full activity. Finally, Segdwick and colleagues showed that TIF1γ could interact with the APC/C-MCC in SAC-activated cells, which is compatible with a potential role of TIF1γ in inhibition of the SAC.
At the molecular level, it is possible that TIF1γ, through its E3-ubiquitin ligase activity, ubiquitinates SAC components and/or APC/C-CDC20 targets proteins. Further experiments need to be performed to address this point. Also, because TIF1γ associates with the chromatin and belongs to a family of transcriptional repressors, TIF1γ might repress the expression of mitotic factors. This hypothesis is supported by microarray experiments we performed on primary MEFs that allowed us to identify numerous mitotic genes that were upregulated in primary Tif1γ−/− MEFs readily after Tif1γ inactivation (Supplementary Fig. S2D and S2E and Supplementary Table S1). In long-term Tif1γ-deleted cells, cell cycle arrest escape is associated with a downregulation of these genes expression (Fig. 5C), representing, in that case, an escape mechanism and not a direct effect of TIF1γ loss. Moreover, MG132 treatment, even if it allows the accumulation of SAC proteins and APC/C-CDC20 targets in absence or in presence of TIF1γ, do not lead to similar levels observed in control cells. Thus, proliferation recovery and the associated mitotic abnormalities seem to be linked to reduced SAC genes expression rather than posttranslational regulation. However, the possible contribution of the TIF1γ E3-ubiquitin ligase activity in that context remains to be addressed.
It is important to note that the mechanism by which the loss of TIF1γ promotes SAC and postmitotic checkpoint attenuations probably results from a stepwise selection process, conferring to the cells lacking TIF1γ a selective growth advantage. An attenuated postmitotic checkpoint in Tif1γ-deficient cells most likely allows the accumulation of nuclear defects and would therefore facilitate the positive selection of cells with more aggressive properties. Indeed, it is commonly accepted that the detrimental effect resulting from chronic CIN is counterbalanced by the growth advantage resulting from the loss of tumor-suppressor genes to promote the amplification of more aggressive clones (48). Interestingly, hepatic tumors arising from Tif1α knockout in mice present a high rate of polyploid cells (20%), highlighting a possible conserved role of TIF1 family member in mitosis regulation (29). In contrast with NMuMG cells, it is interesting to note that we did not see obvious accumulation of cells with a >4n DNA content in Tif1γ−/−-immortalized MEFs (Fig. 1B and F), potentially because these cells are not transformed and retained other active safeguard mechanisms preventing the onset of a prominent aneuploidy cell population. Polyploidy has been recently described to enhance adaptation abilities through genetic plasticity (49). Furthermore, tetraploidy associated with deficient P53 pathway enhances chromosomal rearrangements and thus supports the emergence of abnormal clone prone to tumor development (50). Altogether, loss of TIF1γ should be seen as a triggering event for chromosomal abnormalities that will lead to the onset of genomic alterations giving rise to more aggressive cells. However, whereas all the defects are very different, they all are typical of attenuated SAC and further postmitotic checkpoint inactivation. We have not precisely addressed which ones are likely initiating events of mitosis defects in the absence of TIF1γ. However, it is tempting to speculate that the chromosomal abnormalities observed in Tif1γ-deficient cells arise from supernumerary centrosomes (51). Cells with more than two centrosomes can perform tripolar mitosis, but daughter cells will die by apoptosis, indicating that this defect is not a triggering event for genomic instability. Alternatively, cells with more than two centrosomes can undergo centrosomes clustering and thus form a pseudo-bipolar spindle, further generating lagging chromosome as a result of noncorrected merotelic attachment (51, 52). As a final consequence, the chromosome missegregation errors are increased, resulting then in CIN. Interestingly, some of these defects can be causally related. For instance, lagging chromosomes will generate chromatin bridges and micronuclei upon chromosome decondensation during the next G1 phase progression (53). The large range of mitotic abnormalities observed in Tif1γ-deficient cells (multiple centrosomes, multipolar mitosis, lagging chromosomes, chromatin bridges, micronuclei, multinucleated cells, and giant nuclei) clearly indicated SAC defects. However, these abnormalities remain to be further investigated in order to decipher if they are dependent of each other. Furthermore, despite an obvious effect of TIF1γ loss on SAC, we cannot rule out other mechanisms, such as endo-replication or cytokinesis alterations, to explain the polyploid phenotype (54).
Microtubule-disrupting agents (taxol, vincristine, colcemid, nocodazole) maintain the spindle checkpoint active. Cells treated with microtubule poisons arrest in mitosis because SAC cannot be satisfied and undergo apoptosis (after postmitotic checkpoint activation), explaining the extensive use of these drugs in chemotherapeutic treatments. Yet their efficiency is highly variable between cancer cells and the molecular basis of the adaptation process leading to drug resistance remains unclear (55). One explanation is that some cells are predisposed to SAC inactivation (as a result of genetic or epigenetic alterations), preventing the drugs from arresting the cells in mitosis and their elimination by apoptosis. In such a context, it is also possible that spindle poison drugs may even have the reverse effect of accelerating the onset of chromosomal defects. Indeed, cells with such abnormalities may not be eradicated, with the possible consequence of providing a deleterious positive selection pressure for cells that have acquired a growth advantage and more aggressive properties due to genomic instability. We demonstrated here that long-term inactivation of Tif1γ was associated with attenuated mitotic checkpoints allowing cells harboring chromosomal abnormalities to enter a new cell cycle thus maintaining these genomic abnormalities to daughter cells (Figs. 3B and C and 5D). In this context, it is possible that the status of TIF1γ in human tumors may represent a predictive marker of microtubule drug resistance, as illustrated in vitro in the present study with TIF1γ-deficient cells, which are insensitive to colcemid and nocodazole. Besides, the accumulation of chromosomal alteration may even drive acquisition of aggressive features in TIF1γ-deficient cells when exposed to spindle poisons. Targeting the activity of the APC/C may represent a pertinent therapeutic strategy for these tumors.
In conclusion, we propose an original model to explain the tumor-suppressor role of TIF1γ (Fig. 7G). When TIF1γ is lost, the inhibitory signal emanating from the SAC toward the APC/C-CDC20 is exacerbated, leading to an immediate mitotic arrest. In the longer term, the attenuation of the SAC in escapers lacking TIF1γ releases the APC/C-CDC20 from this inhibition and allows unrestrained mitotic progression. This enables the completion of an abnormal mitosis and the production of daughter cells with chromosomal defects. The postmitotic checkpoint that should detect and eliminate those abnormal cells is attenuated in TIF1γ-deficient cells. CIN, by allowing the loss of tumor-suppressor genes, would then confer to those cells a growth advantage, facilitating tumor progression. This cascade of events provides an original molecular scenario to understand the tumor-suppressive function of TIF1γ.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: R.M. Pommier, J. Gout, L.B. Alcaraz, P. Bernard, A. Puisieux, U. Valcourt, S. Sentis, L. Bartholin
Development of methodology: R.M. Pommier, J. Gout, L.B. Alcaraz, V. Arfi, B. Kaniewski, S. Sentis, L. Bartholin
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R.M. Pommier, J. Gout, L.B. Alcaraz, V. Arfi, B. Kaniewski, G. Fourel, S. Sentis, L. Bartholin
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R.M. Pommier, J. Gout, L.B. Alcaraz, V. Arfi, G. Devailly, C. Moyret-Lalle, S. Sentis, L. Bartholin
Writing, review, and/or revision of the manuscript: R.M. Pommier, J. Gout, D.F. Vincent, L.B. Alcaraz, G. Fourel, C. Moyret-Lalle, A. Puisieux, S. Sentis, L. Bartholin
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L.B. Alcaraz, N. Chuvin, S. Martel, B. Kaniewski, S. Sentis, L. Bartholin
Study supervision: L.B. Alcaraz, S. Sentis, L. Bartholin
Other (help in interpreting some data and consequently in proposing some experiments): S. Ansieau
This work was supported by the Institut National de la Santé Et de la Recherche Médicale (INSERM Avenir program), the Association pour la Recherche sur le Cancer (J. Gout, D.F. Vincent, and B. Kaniewski salaries), fellowships from the Ministère de l'Enseignement Supérieur et de la Recherche of France (R.M. Pommier), from the Ligue Nationale Contre le Cancer (V. Arfi and L.B. Alcaraz), and from the Ecole Normale Supérieure of Lyon (N. Chuvin and G. Devailly).
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.
The authors thank the Centre Léon Bérard–specific pathogen-free animal facility platform AniCan, ALECs-SPF facility, the Laboratoire des Modèles Tumoraux platform for animal care, Daniel Pissaloux and Isabelle Treilleux (Pathology Department, Centre Léon Bérard) and Nicolas Gadot (AniPath platform, Université Lyon I) for histologic processing and their expert analyses, and Christophe Vanbelle for his expert analyses in confocal microscopy (Centre d'Imagerie Quantitative Lyon-Est, SFR Santé Lyon-Est). They also thank Servane Tauszig-Delamasure and Gabriel Ichim for their help with CAM assays, Emmanuelle Ruiz for her help in the genetic analyses of human samples, Anne-Pierre Morel and Jonathan Lebeau for the critical reading of the article, Stéphanie Monod for her technical assistance, and Sarah Kabani for English language editing of the article.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
- Received November 26, 2014.
- Revision received July 8, 2015.
- Accepted July 24, 2015.
- ©2015 American Association for Cancer Research.