The ZEB1 Transcription Factor Acts in a Negative Feedback Loop with miR200 Downstream of Ras and Rb1 to Regulate Bmi1 Expression*

Background: Oncogenic Ras muations in cancer cells cause tumor-suppressing senescence in primary cells through repression of Bmi1. Results: Rb1 pathway status regulates a ZEB1-miR-200 loop downstream of Ras to control expression of Bmi1. Conclusion: Rb1 and ZEB1-miR200 are key downstream effectors of Ras that act on Bmi1. Significance: Rb1 and ZEB1-miR-200 link Ras to Bmi1 to regulate a cellular choice between oncogene-induced senescence and tumor progression when Ras is mutated. Ras mutations are frequent in cancer cells where they drive proliferation and resistance to apoptosis. However in primary cells, mutant Ras instead can cause oncogene-induced senescence, a tumor suppressor function linked to repression of the polycomb factor Bmi1, which normally regulates cell cycle inhibitory cyclin-dependent kinase inhibitors (cdki). It is unclear how Ras causes repression of Bmi1 in primary cells to suppress tumor formation while inducing the gene in cancer cells to drive tumor progression. Ras also induces the EMT transcription factor ZEB1 to trigger tumor invasion and metastasis. Beyond its well-documented role in EMT, ZEB1 is important for maintaining repression of cdki. Indeed, heterozygous mutation of ZEB1 is sufficient for elevated cdki expression, leading to premature senescence of primary cells. A similar phenotype is evident with Bmi1 mutation. We show that activation of Rb1 in response to mutant Ras causes dominant repression of ZEB1 in primary cells, but loss of the Rb1 pathway is a hallmark of cancer cells and in the absence of such Rb1 repression Ras induces ZEB1 in cancer cells. ZEB1 represses miR-200 in the context of a mutual repression loop. Because miR-200 represses Bmi1, induction of ZEB1 leads to induction of Bmi1. Rb1 pathway status then dictates the opposing effects of mutant Ras on the ZEB1-miR-200 loop in primary versus cancer cells. This loop not only triggers EMT, surprisingly we show it acts downstream of Ras to regulate Bmi1 expression and thus the critical decision between oncogene-induced senescence and tumor initiation.

Activating mutations in the Ras pathway are frequent in cancer, leading to the short-circuiting of growth factor signaling required for cancer cell proliferation and survival as tumors initiate in growth factor-deficient environments (1,2). Ras mutation also induces expression of the epithelial-mesenchymal transcription factor (EMT) 3 ZEB1, which drives tumor invasion and metastasis (3)(4)(5)(6)(7). As opposed to cancer cells, mutant Ras can trigger cell cycle arrest leading to tumor-suppressing oncogene-induced senescence in primary cells (8). A key downstream target of Ras is Bmi1, a component of polycomb repressive complex 1 (PRC1) (9 -14). Bmi1 is induced by mutant Ras in cancer cells, and induction of Bmi1 is linked to immortalization of cancer stem cells and to unrestricted proliferation of cancer cells via its repression of cell cycle inhibitory cyclin-dependent kinase inhibitors (cdki) (11,(15)(16)(17)(18)(19). By contrast, Bmi1 is repressed by mutant Ras in primary cells (9 -12). Heterozygous mutation of Bmi1 is sufficient for elevation of cdki and premature senescence of primary cells (9 -10). Therefore, this repression of Bmi1 leads to induction of cdki and oncogene-induced senescence triggered by mutant Ras in primary cells. But, the molecular pathway through which Ras regulates Bmi1 is unclear, and thus it is also unclear how mutant Ras causes opposing effects on Bmi1 expression in primary versus cancer cells. Nevertheless, these opposing effects on Bmi1 expression in primary versus cancer cells appear central to the cellular decision between oncogene-induced senescence and tumor progression when Ras is mutated. Mutant Ras classically represses Bmi1 and causes oncogene-induced senescence in primary cultures of mouse embryo fibroblasts (MEFs) (9,10,12). The Rb1 family consists of transcriptional repressors that can block cell cycle progression (20 -23). Mutation of the Rb1 family in MEFs prevents oncogene-induced senescence by mutant Ras, and the cells become tumorigenic in nude mice (21). Surprisingly, we found that this Rb1 family mutation was sufficient to switch the effect of mutant Ras on Bmi1 expression from repression to induction in MEFs. These results then placed Rb1 pathway status in an unanticipated role as the determinant of whether mutant Ras represses or induces Bmi1. But, how might the Rb1 pathway function to modulate opposing effects of mutant Ras on Bmi1 expression? In cancer cells, Ras has been shown to cause repression of miR-200, which is known to repress Bmi1 (3,24,25). Thus, miR-200 might be a key link between Ras and Bmi1. ZEB1 classically represses miR-200 in the context of a mutual repression loop (3)(4)(5), and as noted above, ZEB1 is induced by Ras in cancer cells (6). Thus, we reasoned that ZEB1 might be a key downstream target of mutant Ras that causes induction of Bmi1 in cancer cells through repression of miR-200. Beyond its classic role in EMT, ZEB1 expression is closely linked to cell proliferation in vivo (26), and we found previously that heterozygous mutation of ZEB1 is sufficient for induction of cdki and premature senescence of MEFs in culture (26). As noted above, this is the same phenotype observed with Bmi1 mutation (9, 10), further suggesting that Bmi1 might be an important downstream target of ZEB1. We have shown that Rb1 binds to the ZEB1 promoter to represses its transcription (27), implying that when Rb1 is activated by mutant Ras it might lead to repression of ZEB1, induction of miR-200 and in turn loss of Bmi1. But, such repression of ZEB1 and induction of miR-200 in response to mutant Ras would be contrary to previous reports showing Ras induction of ZEB1 and repression of miR-200 in cancer cells (3,6,24,25). Here, we utilized primary MEFs with Rb1 family and ZEB1 mutations to examine regulation of Bmi1 by mutant Ras, and we analyzed the pathway in a mouse model of K-Ras initiated lung adenocarcinoma and in human lung adenocarcinomas with a Ras pathway mutation. We provide evidence that Ras controls Bmi1 expression through Rb1-dependent regulation of ZEB1. In primary cells, activation of Rb1 by mutant Ras leads to dominant repression of ZEB1, induction of miR-200 and loss of Bmi1. But, in cancer cells where dominant repression by the Rb1 pathway is lost, Ras induces ZEB1 to cause repression of miR-200 and in turn induction of Bmi1. These results place Rb1 pathway status and ZEB1 in unanticipated roles as key downstream effectors that determine whether Bmi1 is repressed by mutant Ras, leading to oncogene-induced senescence, or induced, leading to tumor progression.

EXPERIMENTAL PROCEDURES
Cell Culture-Rb1 family mutant MEFs and control wildtype MEFs were obtained from T. Jacks (21). Cells derived from four separate embryos were used with similar results. Cells were cultured in DMEM with 10% heat-inactivated fetal bovine serum (28). Ras-TKO MEFs were created by infection with a mutant Ras retrovirus and characterized with regard to Ras expression and activation as described in detail previously (29).
Tumor Formation in Nude Mice-Cells were injected subcutaneously into the hind limb of Balb/c nude mice as described (28). Tumors were fixed in 10% buffered formalin, embedded in paraffin, and sectioned for H&E.
RNA Extraction and Real-time PCR-RNA was extracted using TRIzol, and cDNA was synthesized using the Invitrogen RT kit (Invitrogen), and SYBR Green real-time PCR was performed using a Stratagene Mx3000P Real-time PCR system (28). PCR primer sequences have been described previously (26 -29). Three independent samples, each in triplicate, were analyzed for each real time PCR condition. The detection of miRNAs is as described (30). Briefly, polyadenylation of at least 5 g of the total RNA was completed by poly(A) polymerase kit (PAP, Ambion) in 20 l of reaction volume according to manufacturer's instruction. The polyadenylated RNA was thereafter directly utilized for cDNA preparation using a reverse transcription kit (M-MLV reverse transcriptase, Invitrogen) and an adaptor primer (5Ј-GCGAGCACAGAATTAATACGACTC-ACTATAGG(T)12VN*-3Ј) in 40 l of reaction volume. Realtime quantitative PCR was performed using a universal primer (5Ј-GCGAGCACAGAATTAATACGAC-3Ј) and a miRNAspecific primer (38).
Knockdown of ZEB1-We have described lentiviral shRNA knockdown of ZEB1 protein and mRNA previously (28). An additional five lentivirus shRNA knockdown vectors obtained from Open Biosystems were also used for knockdown of ZEB1 with equivalent results. Lentivirus with a scrambled shRNA sequences was used as a control in the experiments (28).
Immunostaining, Western Blot, and in Situ Hybridization-Immunostaining and Western blotting were performed as described previously (26,29). Antibodies and conditions have been described (26 -29). For in situ hybridization to miR-200a and miR-200c, mouse lung tissues were fixed in 10% formalin solution immediately after removal, and then paraffin-embedded and sectioned at 10 m. LNA-modified, double-DIG labeled DNA probes were purchased from Exiqon (Denmark) and in situ hybridization was performed according to the manufacturer's instructions. Briefly, the paraffin-embedded slides were deparaffinized and then treated with protease K. The probes were hybridized to the sections at 57°C after dehydration, and then detected by an antibody against DIG conjugated with Cy3 (Jackson ImmunoResearch).
K-Ras LA1 mice-Housing and handling of all mice was in accordance with procedures approved by the University of Louisville Institutional Animal Care and Use Committee (IACUC). K-Ras LA1 mice (31) in a C57BL6 background were obtained from Jackson Laboratory. PCR genotyping was as described previously (31). Tumor pathology was evaluated independently by two experienced pathologists (ABJ and MC).
Human Lung Adenocarcinoma Microarray Analysis-Microarray data for human lung adenocarcinomas sequenced for mutations in K-RAS, EGFR, and p53, and patient-matched control lung tissue was obtained from the NCBI database (GSE11969) (32). In this study, 90 lung adenocarcinomas were examined. p53 exons 4 -10, EGFR exons 15-24, and K-RAS exons 1 and 2 were amplified from RNA samples used for microarray analysis and PCR products were sequenced for mutations. Tumors with a p53 mutation were excluded from the analysis. Data were corrected for background and normalized to median fluorescence. Genes whose expression varied less than a factor of three between samples were excluded from the analysis. Box plots and Pearson scores were done as described previously (33).

Mutant Ras Causes Repression of Bmi1, Induction of cdki and
Arf, Accumulation of Hypophosphorylated Rb1 and Senescence of MEFs-As demonstrated previously, infection of primary cultures of MEFs with a retrovirus expressing G12V mutant Ras triggered an acute arrest of cell proliferation and rapid senescence, occurring before the cells could be passaged (8) (Fig. 1A). Chronic stimulation of Ras signaling by serum growth factors in cell culture likewise leads to senescence, but only after multiple passages (13,34,35) (Fig. 1A). Such Ras-initiated senescence led to repression of Bmi1, induction of cdki and Arf and accumulation of hypophosphorylated, active Rb1 ( Fig. 1B and C). Induction of these of cdki and Arf is then thought to mediate the cell cycle arrest and senescence of primary cells that occurs with mutant Ras (12,13,35).

ZEB1 Is Repressed and miR-200 Is Induced by Mutant Ras in MEFs-Mutant
Ras has been shown to induce expression of ZEB1 through activation of Raf and in turn Erk2 (6), and such induction of ZEB1 has been linked to EMT and tumor invasion in cancer cells (3)(4)(5). miR-200 is tightly linked to ZEB1 in a mutual repression loop (4), and miR-200 is repressed by mutant Ras in cancer cells (3,24,25). As with Ras-mediated induction of ZEB1, this repression of miR-200 is associated with EMT and invasion/metastasis (3)(4)(5). In contrast to these previously published findings in cancer cells, we found that mutant Ras caused repression of ZEB1 and induction of miR-200 in primary cultures of MEFs (Fig. 1B). Consistently, ZEB1 was repressed and miR200 was induced as MEFs underwent passage-dependent senescence (Fig. 1D). Thus, Ras appeared to have an opposite effect on expression of ZEB1 and miR-200 in primary cells versus cancer cells.

ZEB1 Mutation in MEFs Leads to Gene Dosage-dependent Induction of miR-200, Repression of Bmi1, and Premature
Senescence-As with mutation of Bmi1, we found previously that ZEB1(ϩ/Ϫ) and ZEB1(Ϫ/Ϫ) MEFs undergo ZEB1 gene dosage-dependent premature senescence in culture with induction of cdki and Arf (Ref. 26). This occurred at P2 with ZEB1(Ϫ/Ϫ) MEFs and at P4 with ZEB1(ϩ/Ϫ) MEFs. We then compare gene expression in these senescent ZEB1 mutant MEFs to proliferating wild-type MEFs. As would be expected with down-regulation of ZEB1, miR-200a and miR-200c were induced in these ZEB1 mutant MEFs (Fig. 2B). Bmi1 is repressed by miR-200, and consistent with induction of miR-200 in the ZEB1 mutant MEFs, Bmi1 was repressed in these cells (Fig. 2B). These results raised the possibility that Ras-mediated repression of ZEB1, and in turn induction of miR-200, might be responsible for the down-regulation of Bmi1 by mutant Ras that is linked to oncogene-induced senescence in primary cells. But, it was unclear why ZEB1 would be repressed by mutant Ras in primary cells as opposed to being induced by Ras in cancer cells.
Repression of ZEB1 by Rb1 Leads to Loss of Bmi1-Rb1 family members have overlapping activities, and inactivation or mutation of all three family members (Rb1, Rbl1/p107, and Rbl2/ p130) to create TKO-MEFs is required to prevent oncogene-induced senescence by mutant Ras, and to prevent the passagedependent senescence resulting from chronic stimulation of the Ras pathway by continuous serum growth factor exposure in culture (21,22). We found previously that Rb1-E2F binds to the ZEB1 promoter and represses its expression (27), and accordingly ZEB1 was induced in an Rb1 family-dependent fashion as Rb1 family members were mutated in MEFs (Fig. 3A). And, this induction of ZEB1 coincided with induction of Bmi1 (Fig. 3A), providing evidence of linkage between Rb1 and Bmi1 via ZEB1. However, the E2F family member, E2F1, has been shown previously to bind the Bmi1 promoter (36). Rb1 can bind E2F1 to repress transcription (20), thus it was possible that beyond regulation of ZEB1, the Rb1 family was also directly regulating Bmi1 expression via E2F1. We then compared expression of Bmi1 in E2F1(ϩ/ϩ) and E2F1(Ϫ/Ϫ) MEFs, and found that E2F1 mutation did not affect Bmi1 expression (Fig.  3B), suggesting that E2F1-Rb1 is not directly affecting Bmi1 expression in these cells.
Experiments above show that mutation or knockdown of ZEB1 leads to loss of Bmi1 expression. We attempted to overexpress ZEB1 in MEFs to demonstrate induction of Bmi1, but this led to loss of cell viability (results not shown). However, when ZEB1 was overexpressed in mouse lung cancer cells (37), Bmi1 expression was induced (Fig. 3B). Lentiviral shRNA knockdown of ZEB1 in these cells also led to loss of Bmi1.
Repression of ZEB1 by Rb1 Is Dominant Over Its Induction by Ras-As reported previously, when mutant Ras was introduced into TKO-MEFs to create Ras-TKO-MEFs, the cells became tumorigenic in nude mice (21,38) and ZEB1 was induced and   (Fig. 1C). We demonstrated previously that Ras induction of ZEB1 is important for viability of the Ras-TKO-MEFs cancer cells (38), and similar results were also shown for lung and pancreatic cancer cells with Ras mutations (6,39). Knockdown of ZEB1 in the Ras-TKO-MEF cells led to induction of miR-200, repression of Bmi1 and induction of polycomb target genes including cdki and Arf (Fig. 3C). Taken together, our results suggest that repression of ZEB1 by Rb1 is dominant over its induction by Ras. Thus, Rb1 pathway status dictates whether mutant Ras will repress or induce the ZEB1-miR-200 loop-repression occurs in primary cells where the Rb1 pathway is active, whereas induction occurs in cancer cells where the Rb1 pathway is mutated or inactivated. A target of this ZEB1-miR-200 loop appears to be Bmi1, whose expression is thought to be important in the cellular decision between oncogene-induced senescence and tumor initiation when Ras is mutated.
There is interplay between the Rb1 family and p53 in senescence and tumor initiation. p53 can induce p21 to cause accumulation of hypophosphorylated Rb1, and in p53 mutant tumors Rb1 is not activated by mutant Ras (40), and accordingly mutant Ras fails to trigger oncogene-induced senescence in p53 mutant MEFs, instead leading to oncogenic transformation. p53 mutation had little or no effect on expression of ZEB1 or Bmi1 in proliferating MEFs (Fig. 3B). Because chronic serum growth factor treatment in culture fails to activate Rb1 in p53-null MEFs, the cells do not undergo senescence with passage in culture (resembling TKO MEFs), and thus ZEB1 and Bmi1 expression is maintained in the p53-null MEFs with passage (results not shown). However, knockdown of ZEB1 in the p53-null MEFs led to loss of Bmi1 expression (Fig. 3B).

PI3K Activity Is Required for Ras Induction of ZEB1 in TKO
MEFs-Next, we examined the pathway through which mutant Ras activates ZEB1 in TKO-MEFs. The resulting Ras-TKO-MEFs provide a defined primary cell genetic model of Ras-initiated tumorigenesis. GTP-Ras can recruit and activate Raf at the plasma membrane, and Raf in turn activates the MAP kinase signaling leading to Erk activation to stimulate cell proliferation (41). But, GTP-Ras also activates PI3 kinase (PI3K), leading in turn to phosphorylation and activation of Akt (42). Akt can also stimulate cell proliferation, for example by regulating subcellular localization of cyclin D and expression of GSK3␤. It also protects the cells from apoptosis by inhibiting proapoptotic factors. Further, Akt regulates cytoskeletal assemble required for transition of cancer cells to a migratory, pro-invasive phenotype, a hallmark of EMT. As opposed to functioning in parallel during Ras-initiated cancer progression, accumulation of constitutively active Akt can catalyze inhibitory phosphorylation of Raf, thereby blocking downstream activation of Erk (43,44). And, in tumors where the PI3K pathway inhibitor Pten is mutated, Akt becomes constitutively phosphorylated and Erk activation is inhibited (45). Additionally, mutation of Pten and Ras are generally mutually exclusive implying that Akt activation is a major function of Ras mutation in tumors (45)(46)(47). Accordingly, in mouse models of Ras-initiated cancer, activation of Akt but not Erk is required for tumor progression and invasion (48). A previous study found that ZEB1 could be induced by Ras through activation of Erk2 in a cancer cell line (6). But, loss of the Rb1 pathway can override the requirement for Erk activity (49). Indeed, we demonstrated previously that high levels of phosphorylated Akt accumulate in Ras-TKO MEFs, leading to inhibitory phosphorylation of Raf and a complete block in Erk activation (29) (Fig. 4A). Treatment of Ras-TKO-MEFs with the PI3K inhibitor LY294002 dimin- ished phospho-Akt in Ras-TKO MEFs, and it restored Erk activation ( Fig. 4B and C). Importantly, however, this block in PI3K activity, but not a block in Erk activity, led to down-regulation of ZEB1 and decreased proliferation of Ras-TKO-MEFs despite Erk activation (Fig. 5A, B, and C). By contrast, LY294002 did not significantly inhibit ZEB1 expression in MEFs or TKO-MEFs (Fig. 5A) or affect proliferation of these cells (Fig. 5C), suggesting that PI3K activity is specifically required for Ras induction of ZEB1. Taken together, these results demonstrate that in this defined genetic model of Ras-induced tumor initiation in pri-   (31). An H&E stained section shows the border of an AC. B, higher power view of the invasive front of an AC. C, immunostaining of a section at the invasive front of an AC for ZEB1. Note lack of ZEB1 in normal lung epithelium and induction of expression in the AC. D, immunostaining for E-cadherin at the invasive front of an AC. As opposed to ZEB1, E-cadherin is evident in normal lung epithelium and it is diminished in the AC. E and EЈ, immunostaining showing Bmi1 induction at the invasive front of AC. F and G, in situ hybridization for miR-200a (200a) and miR-200c (200c). Like E-cadherin, miR-200 is evident in normal lung, but it is diminished in AC. In AC expression, miR-200 is only evident on blood vessels (arrows). Bars are 100 m.
Elevated ZEB1 Is Required to Maintain Bmi1 Expression in Ras-TKO-MEFs-As noted above, previous studies in cancer cells containing a Ras mutation demonstrated that the cells had become addicted to their elevated level of ZEB1. Consistent with increased dependence of cancer cells on Ras-mediated induction of ZEB1, it is of note that knockdown of ZEB1 in Ras-TKO-MEFs to a level that was still above that seen in wildtype MEFs (Fig. 3B), led to loss of viability (38,39). As would be expected, this knockdown of ZEB1 in Ras-TKO-MEFs led to an increase in miR-200 to a level resembling that in wild-type MEFs (Fig. 3B). But, Bmi1 was repressed to a level below that seen in wild-type MEFs; it resembled the level of Bmi1 seen in senescent MEFs and in ZEB1(Ϫ/Ϫ) MEFs that had undergone senescence at P2 (Figs. 3B and 5D). Consistent with this low level of Bmi1, cdki and Arf were dramatically induced in ZEB1 knockdown Ras-TKO-MEFs to levels similar to those seen in senescent MEFs (Fig. 3B). We conclude that Bmi1 expression is more sensitive to the level of ZEB1 in Ras-TKO MEFs than in wild-type MEFs, which we suggest accounts at least in part for dependence of Ras-TKO MEFs on elevated ZEB1 for viability. We noted that the Bmi1 promoter contains E-box sequences and thus potential binding sites for ZEB1. This raised the possibility that the elevated ZEB1 in the Ras-TKO MEFs might be directly repressing Bmi1 in addition to repressing miR200. But, we did not detect ZEB1 binding to the Bmi1 promoter in chromatin immunoprecipitation assays (results not shown).
Activation of the ZEB1-miR-200 Loop and Bmi1 in Ras-initiated Mouse Lung Cancer-As in cell culture, the Rb1 pathway serves as a barrier to Ras-initiated tumors. Mutation of Rb1 facilitates Ras-initiated lung adenocarcinoma, and in the absence of Rb1 mutation ZEB1 induction coincides with hyperphosphorylation and inactivation of Rb1 in lung and breast cancers (50,51). Activation of an oncogenic K-Ras allele by recombination in somatic cells leads to spontaneous lung cancer in  K-Ras LA1 mice, and these mice develop multiple and bilateral adenocarcinomas by four months of age (31) (Fig. 6A and B). Immunostaining demonstrated that ZEB1 was low or absent in normal lung epithelium, but it was induced in all adenocarcinomas (Fig. 6C). This pattern of ZEB1 expression was opposite to that seen for the epithelial maker, E-cadherin (Fig. 6D), which is a classic target of ZEB1 repression during EMT (45). Like ZEB1, Bmi1 was also induced in these adenocarcinomas ( Fig. 6E and EЈ). Using in situ hybridization, we found that miR-200a and miR-200c were high in normal lung epithelium, but they were repressed in adenocarcinomas where their expression was only evident in blood vessels ( Fig. 6F and G). These results provide evidence of a Ras-induced ZEB1-miR-200 loop in Ras-initiated lung adenocarcinoma that is linked to induction of Bmi1.
Co-induction of ZEB1 and Bmi1 in Human Lung Adenocarcinomas with a Ras Pathway Mutation-Next, we examined human lung adenocarcinomas microarrays, which had been sequence for Ras pathway mutation (either K-Ras or epidermal growth factor receptor (EGFR), whose mutation leads to constitutive Ras pathway signaling) (32), for expression of ZEB1 and Bmi1. Bmi1 was significantly induced in these lung adenocarcinomas compared with adjacent uninvolved normal lung, and a Pearson plot demonstrates a linear relationship between ZEB1 and Bmi1 in the tumors (Fig. 7A and B).

Bmi1 is classically repressed by mutant Ras in primary cells.
This loss of Bmi1 leads to induction of cdki that inhibit the cell cycle, and Arf, which activates p53. The result is then a combination of oncogene-induced senescence and apoptosis that suppresses tumor progression. Beyond genes targeting cell proliferation and survival, Bmi1 expression is also important for maintenance of hematopoietic, skeletal, and neural stem cells, repression of imprinted genes, and its induction in tumors is linked to generation of cancer stem cells as well as tumor initiation and progression. By contrast to its repression by Ras in primary cells, Bmi1 is induced by Ras in cancer cells, including breast cancer and pancreatic cancer where K-Ras is mutated in more than 95% of tumors (14 -17, 52). The molecular basis for these opposing effects of Ras on Bmi1 expression is then central to understanding the balance between tumor suppression via oncogene-induced senescence versus tumor progression when Ras is mutated. Previous studies have linked Bmi1 expression to miR-200 and to ZEB1 in stem cells and cancer stem cells (3)(4)(5)19). And, we have demonstrated previously that ZEB1 is a target of repression by Rb1, and it has a role beyond EMT in regulation of cdki and Arf and thus in preventing cell senescence (26). Here, we tie these previous observations to our new studies and suggest that Rb1 pathway status dictates whether mutant Ras will repress Bmi1 to trigger oncogene-induced senescence or induce Bmi1 leading to tumor initiation (Fig. 8). The Rb1 pathway is mutated or inactivated in cancer cells, and in the absence of dominant repression by Rb1, we show that Ras induces ZEB1 to repress miR-200, leading in turn to induction of Bmi1. Surprisingly, these studies place ZEB1 in a key position downstream of Ras and Rb1 in regulation of Bmi1 and thus oncogene-induced senescence. Accordingly, we show a linear correlation between expression of Bmi1 and ZEB1 in lung adenocarcinomas with Ras pathway mutation.
It has been presumed that transition of cancer cells to an invasive, metastatic phenotype is a multi-step process that culminates in later stage tumors. But, emerging evidence suggests that cancer stem cells represent an early-metastasizing population in tumors (53). Indeed, in pancreatic adenocarcinoma, which is driven by K-Ras mutation, early metastasizing cells with properties of cancer stem cells are evident in precancerous PanIN lesions, before their transition to adenocarcinoma (54). These cancer stem-like cells express ZEB1 and have undergone EMT. The origin of cancer stem cells in tumors remains controversial, but our previous findings (28), together with more recently studies from Weinberg's group (54), demonstrate that induction of ZEB1 can drive reprogramming of an existing cancer cell population to cancer stem cells. Taken together, these findings suggest an unanticipated role for induction of ZEB1 early in cancer initiation and progression. Bmi1 has a central role in maintenance of cancer stem cells (11,19), and our studies raise the interesting possibility that Ras-mediated induction of ZEB1 is driving reprogramming of cancer cells to cancer stem-like cells at least in part through induction of Bmi1 and thus activation of the global reprogramming complex PRC1.