|Home | About | Journals | Submit | Contact Us | Français|
The polycomb group family protein BMI-1 is over-expressed by and functions as an oncogene in many different human cancers. We have previously shown that BMI-1 promotes the tumorigenicity of Ewing sarcoma family tumors (ESFT) and that this is mediated independently of CDKN2A repression. In the current study we have discovered that high levels of BMI-1 confer resistance to contact inhibition in ESFT cells. Using stable retroviral transduction we evaluated the consequences of BMI-1 knockdown on the growth of CDKN2A wild-type and mutant ESFT cells in sub-confluent and confluent conditions. Although knockdown of BMI-1 had no effect on proliferation in low-density cultures, at high cell densities it resulted in cell cycle arrest and death. The normal cell contact inhibition response is mediated, in large part, by the recently described Hippo pathway which functions to inhibit cell proliferation and promote cell death by inactivating the Yes-Associated Protein (YAP). Significantly, we found that YAP levels activity and expression did not diminish in confluent ESFT cells that expressed high levels of BMI-1. In contrast, YAP expression and nuclear localization were reduced in confluent BMI-1 knockdown cells suggesting that silencing of BMI-1 restored contact inhibition by restoring normal activation of the Hippo-YAP growth-suppressor pathway. Importantly, knockdown of YAP in ESFT cells resulted in profound inhibition of cell proliferation and anchorage-independent colony formation suggesting that stabilization and continued expression of YAP is critical for ESFT growth and tumorigenicity. Together, these studies reveal a previously unrecognized link between BMI-1, contact inhibition and the Hippo-YAP pathway and suggest that resistance to contact inhibition in BMI-1 over-expressing cancer cells may be in part a result of Hippo inhibition and aberrant stabilization of YAP.
Ewing sarcoma family tumors (ESFT) are poorly differentiated bone and soft tissue tumors that primarily occur in pediatric and young adult patients. Despite highly toxic multimodal treatment regimens, the overall survival for patients with ESFT remains 70% for localized tumors and only 20% for metastatic disease (Balamuth and Womer, 2010; Ban et al., 2006). Unfortunately, survivors of ESFT often suffer significant long-term side effects related to their curative therapy and new therapeutic approaches for these tumors are desperately needed (Balamuth and Womer, 2010). Importantly, recent data support a neural crest and/or mesenchymal stem cell origin of ESFT and deregulation of stem cell-associated genes and pathways contributes to tumor pathogenesis (Douglas et al., 2008; Meltzer, 2007; Richter et al., 2009; Tirode et al., 2007). Understanding the mechanisms of action of stem cell genes in ESFT initiation and progression is likely to identify novel therapeutic molecular targets.
The self-renewal gene and polycomb group family member BMI-1 is highly expressed by many human cancers where it functions as an oncogene (Gil et al., 2005). BMI-1 promotes stem cell self-renewal and malignant transformation in part via epigenetic repression of the CDKN2A locus, thus inhibiting expression of the cell cycle inhibitors p16INK4A and p14ARF (Jacobs et al., 1999a; Jacobs et al., 1999b). Importantly however, we have found that although ESFT over-express BMI-1, its oncogenic function in established tumors is largely independent of CDKN2A repression and is mediated, in part, through altered cell adhesion (Douglas et al., 2008). Consistent with this observation, other groups have shown that BMI-1 may exert its oncogenic function by affecting developmental and growth signaling pathways irrespective of CDKN2A status (Bruggeman et al., 2007; Datta et al., 2007; Wiederschain et al., 2007). In addition, recent data support roles for BMI-1 in regulating oxidative stress, mitochondrial function, and DNA damage response (Chatoo et al., 2009; Liu et al., 2009). In the current study we have discovered that high levels of BMI-1 in ESFT confer resistance to contact inhibition.
Cancer cells fail to undergo cell contact inhibition, a phenomenon that restricts the in vitro growth of normal cells at confluence (Abercrombie and Ambrose, 1958; Abercrombie and Heaysman, 1954). Loss of cell contact inhibition can, therefore, lead to cancerous outgrowth and invasion (Abercrombie, 1979; Eagle and Levine, 1967; Silletti et al., 1995). To assess whether BMI-1 contributes to loss of cell contact inhibition, we monitored the growth of ESFT cells following BMI-1 knockdown (Fig 1A). Consistent with our earlier findings (Douglas et al., 2008), knockdown of BMI-1 had no effect on cell expansion in cells grown in sub-confluent monolayers (Fig 1B). In contrast, confluent cells displayed dramatic changes in growth and phenotype (Fig 1C &D). Specifically, upon reaching confluence growth of BMI-1 knockdown cells ceased while control cells continued to divide (Fig 1C). In addition, when maintained in culture beyond confluence, BMI-1 knockdown cells lifted off cell culture dishes in sheets and cells adopted the rounded-up appearance of dying cells (Fig 1D). Importantly, despite their different genetic backgrounds (TC71 cell line is CDKN2A-null, TP53 mutant whereas CHLA9 cells are CDKN2A, TP53 wild-type) the effects of BMI-1 knockdown on contact inhibition were similar indicating that BMI-1 contributes to avoidance of contact inhibition and that this is mediated through mechanisms that are, at least in part, independent of CDKN2A repression.
To further characterize the growth inhibitory effects of BMI-1 knockdown in confluent cultures, we assessed cell death and proliferation in low- and high-density cultures. Trypan blue exclusion assays confirmed that viability was diminished in BMI-1 knockdown cells but only after they reached confluence (Fig 2A). In addition, cell cycle analyses revealed that upon reaching confluence, BMI-1 knockdown TC71 cells underwent a G1 cell cycle arrest (Fig 2C), which was rapidly followed by loss of attachment to cell culture plates and massive cell death (see Fig 1D). Analysis of CHLA9 cells revealed similar, although not identical findings (see Supplementary Fig 1). At both low and high cell density nearly 40% of control CHLA9 cells were actively cycling (Supplementary Fig 1A). However, upon reaching confluence BMI-1 knockdown CHLA9 cells experienced overwhelming cell death as evidenced by a marked increase in floating cells (see Fig 1D) and an increased frequency of cells with sub-G1 DNA content (Supplementary Fig 1B). Thus, in both cell lines, cell death was induced at confluence in BMI-1 knockdown populations. In TC71 cells this death was preceded by G1-arrest that was not detected in CHLA9 cells. We speculate that the wild-type p53 status of CHLA9 cells may have contributed to a more rapid cell death and prevented detection of prior G1 arrest. Elucidation of the potential contribution of the p53 pathway to contact inhibition-induced cell death requires further study.
Although loss of contact inhibition has long been recognized as a fundamental property of transformed cells, the molecular mechanisms underlying this phenomenon are only now being elucidated. In particular, recent studies have shown that the growth-regulating Hippo pathway is a key mediator of contact inhibition that is highly evolutionarily conserved (reviewed in Zhao et al., 2010a). Activation of serine/threonine kinases within the Hippo pathway results in phosphorylation and inactivation of the transcriptional coactivator Yorkie (in Drosophila) or Yes-associated protein (YAP; in mammals) (Zhao et al., 2010a). Appropriate phosphorylation, cytoplasmic retention and degradation of Yorkie/YAP is developmentally critical for the control of normal organ size and deregulation of this process can contribute to tumorigenesis (reviewed in Zhao et al., 2010a) Indeed the YAP1 gene is highly expressed by a number of human cancers and recent reports implicate YAP as a growth promoting oncogene in human cancer (Fernandez et al., 2009; Snijders et al., 2005; Zender et al., 2006).
Given the central role for the Hippo-YAP pathway in contact inhibition we investigated whether knockdown of BMI-1 induced changes in expression of YAP in confluent ESFT cells. Although no significant change in transcript levels was observed (Supplementary Fig 2A), BMI-1 knockdown cells demonstrated progressive down-regulation of YAP protein as cell density increased (Fig 3A). In contrast, YAP levels remained high in control cells (Fig 3A). In addition, expression of putative YAP target genes MYC and BIRC2 remained high in confluent control cells but was reduced in BMI-1 knockdown cells (Supplementary Fig 2B). Thus, our data confirm that avoidance of contact inhibition in BMI-1 over-expressing ESFT cells is associated with stabilization of YAP. Interestingly, we also observed a marked reduction in BMI-1 expression in confluent BMI-1 knockdown cells (Fig 3A). Like YAP, reduced BMI-1 was not associated with consistently reduced levels of the BMI-1 transcript (Supplementary Fig 2A) suggesting that the contact inhibition response in BMI-1 knockdown cells may have been further amplified by degradation of the BMI-1 protein itself. Whether BMI-1 and YAP degradation are controlled by the same effector pathways and/or a positive feedback loop exists between BMI-1 and YAP to potentiate growth inhibition in conditions of confluence is an intriguing question that requires further study.
Hippo pathway-mediated inactivation of YAP is in part a consequence of YAP phosphorylation at serine 127 by the serine/threonine kinase LATS1 which leads to YAP cytoplasmic retention (reviewed in Zhao et al., 2010a). Therefore, to begin to elucidate the mechanism of aberrant YAP stabilization in BMI-1 over-expressing cells we evaluated expression of phosphorylated YAP and LATS1 as well as YAP subcellular localization. Notably, both pYAP-S127 and pLATS1 levels increased at high cell density in confluent control and BMI-1 knockdown cells indicating that in both conditions the Hippo cascade was activated at confluence. However, comparison of the relative levels of pYAP-S127 and pLATS1 compared to total YAP and LATS1 demonstrated that the relative expression levels of phosphorylated proteins were markedly increased in cells with reduced expression of BMI-1 (Fig 3B). Thus, these findings show that high levels of BMI-1 inhibit, but do not completely block, activation of the Hippo kinase pathway in confluent ESFT cells. Although phosphorylation of S127 is a key target of LATS1 activity, recent data suggest that phosphorylation of YAP at alternate residues by LATS1 and other effector kinases can also contribute to its destabilization (Zhao et al., 2010b). In particular, CK1-mediated phosphorylation of S381 has been implicated as a key effector of YAP degradation (Zhao et al., 2010b). Although we observed no change in the CK1 levels in BMI-1 knockdown cells at low- or high-density (not shown) it remains possible that alternate mechanisms and post-translational modifications might also contribute to stabilization of YAP in BMI-1 over-expressing ESFT cells.
Next, we assessed whether stabilization of YAP in confluent cells was associated with altered subcellular localization. As expected, YAP was predominantly nuclear in sub-confluent ESFT cells (not shown). Unexpectedly, however, we found that pYAP-S127 was also readily detected in cell nuclei, including cells at confluence (Supplementary Fig 3A & B). Importantly, however, in contrast to control cells, in BMI-1 knockdown cells nuclear expression of YAP and pYAP-S127 was reduced in high-density cultures (Supplementary Fig 3B). These data suggest that high levels of BMI-1 might directly and/or indirectly, promote nuclear retention of YAP and pYAP-S127. In support of this possibility BMI-1 and pYAP-S127 appear to co-localize in the nuclei of confluent ESFT cells (Supplementary Fig 3A) and immunoprecipitation studies suggest that YAP physically associates with BMI-1 in ESFT cells in vivo (Supplementary Fig 3C). Thus, in addition to relative inhibition of Hippo pathway activation these studies suggest that high levels of BMI-1 contribute to aberrant stabilization of YAP by promoting YAP sequestration in the nucleus.
Knockdown of BMI-1 in ESFT cells inhibits their tumorigenic potential (Douglas et al., 2008). Given our observation that BMI-1 stabilizes YAP and that YAP functions as an oncogene in several human cancers (reviewed in Zhao et al., 2010a) we hypothesized that YAP might also function as an important oncogene in ESFT. To assess the potential oncogenic function of YAP in ESFT cells we induced stable knockdown of YAP in TC71 and CHLA9 cells using two different shRNA lentiviral vectors and assessed the effects of loss of YAP on ESFT cell proliferation and anchorage-independent growth (Fig 4). Consistent with its role as a growth promoter, we observed a profound, dose-dependent inhibition of cell proliferation in YAP knockdown cells (Fig 4B). In addition, cell viability was also reduced in a dose-dependent fashion following knockdown of YAP (Figure 4C). Moreover, TC71 and CHLA9 cells with reduced levels of YAP demonstrated a marked reduction in colony formation in soft agar (Fig 4D). Impaired anchorage-independent colony formation in vitro was also observed in two additional ESFT cell lines (A4573 and A673) that were similarly genetically modified to express reduced levels of YAP (data not shown). These studies confirm that YAP functions as a potent oncogene in ESFT and support the hypothesis that stabilization of YAP by BMI-1, either alone or in combination with other as yet unidentified factors, contributes to ESFT tumorigenicity.
Taken together our studies show that ESFT cells that express high levels of BMI-1 avoid contact inhibition and that this is mediated, at least in part, by inhibition of Hippo pathway activation and stabilization of YAP. The precise molecular mechanisms by which BMI-1 inhibits Hippo remain to be elucidated. Indeed, identification of the molecular components responsible for sensing cell-cell contact and initiating the Hippo-YAP pathway in mammalian cell remains an area of intense investigation. Engagement of several cytoskeleton-related proteins in Drosophila, including Merlin (Mer), Expanded (Ex) and Kibra – along with Mer and Ex homologues Nf2 and Frmd6 in mammalian cells – have been implicated directly or indirectly in contact inhibition and Hippo activation (Badouel et al., 2009; Baumgartner et al. 2010; Genevet et al., 2010; Hamaratoglu et al., 2006; Morrison et al., 2001; Striedinger et al., 2008; Yu et al.; Zhang et al., 2010). Significantly, we have previously observed that knockdown of BMI-1 in ESFT cells has a profound effect on expression of genes involved in cell adhesion (Douglas et al., 2008). Thus, it is conceivable that by altering expression of cell adhesion genes, high levels of BMI-1 might alter cytoskeletal protein engagement and thereby inactivate contact inhibition at the level of the cell surface. Ongoing studies are underway to investigate this possibility. With respect to our observation that stabilization of YAP might involve direct protein:protein interactions between YAP and BMI-1, it is noteworthy that polycomb group proteins have been shown to physically interact with members of the Hippo signaling cascade in Drosophila (Parrish et al., 2007). It will be interesting to learn whether BMI-1 or other polycomb proteins also interact with upstream members of the Hippo pathway in mammalian cells. In addition, given that BMI-1, along with other polycomb family members, functions to alter chromatin structure in part through mono-ubiquitination of histone H2A at lysine 119 (Bracken and Helin, 2009), the possibility that BMI-1, either by itself or as a member of the PRC1 complex, might post-translationally modify YAP to stabilize its function is worthy of further investigation.
Loss of BMI-1 can trigger premature differentiation of multipotent stem/progenitor cells (Oguro et al. 2010; Pietersen et al., 2008) and contact inhibition is a characteristic feature of more differentiated cell types (Lim et al., 1981; Rodesch, 1973). Knockdown of BMI-1 in ESFT cells results in up-regulation of differentiation genes (Douglas et al., 2008). Therefore, it is possible that knockdown of BMI-1 in these cells induces a degree of differentiation that restores activation of contact inhibition pathways, such as Hippo-YAP. Importantly, a recent study of murine embryonic stem cells demonstrated a critical role for YAP in activation of stem cell genes, leading the authors to speculate that YAP might also play a role in mediating the self renewal of tissue-specific stem cells (Lian et al., 2010). Indeed, it has been shown that YAP is preferentially expressed in the progenitor compartment of the intestine (Camargo et al., 2007) and in preliminary studies we have found that YAP1 is highly expressed by both neural crest and mesenchymal stem cells (unpublished data). Thus, we speculate that YAP might play an important role in promoting the self-renewal of both these putative cells of origin of ESFT as well as established ESFT cells themselves. Further studies are now required to address this novel hypothesis.
In conclusion, in the current study we have identified a previously unreported link between the polycomb group protein BMI-1, contact inhibition and the Hippo-YAP pathway. Our observation that continued high-level expression of YAP is necessary for ESFT cell proliferation and anchorage-independent growth, implicate YAP as an important oncogene in ESFT and a potentially critical effector of BMI-1-mediated tumorigenicity. It will be interesting to learn if BMI-1-mediated stabilization of YAP and prevention of contact inhibition contribute to the pathogenesis of other human cancers and to the tumorigenicity and invasiveness of BMI-1-positive cancer stem cells.
We gratefully acknowledge Long Hung and Asim Beg for technical support, members of the Lawlor lab for helpful discussion, Dr. T. Triche and the Children’s Oncology Cell Bank (COGCELL.org) for cell lines. This work was initiated at Childrens Hospital Los Angeles (CHLA) with support from the T.J. Martell Foundation. JHH was supported by CIRM training grant T1-00004 and ERL by NIH/NCI 1R01CA134604-01.
Grant support: California Institute for Regenerative Medicine Training Grant T1-00004 (JHH), 1R01CA134604-01 (ERL)
Conflict of interest statement: The authors have no conflicts of interest to declare.