Background
Polycomb Group (PcG) proteins originally discovered in Drosophila are evolutionarily conserved epigenetic regulators of development [
1‐
3]. These proteins regulate proliferation and differentiation of cells via epigenetic silencing of important growth regulatory genes [
3,
4]. The first mammalian PcG gene
BMI1 (B lymphoma Mo-MLV insertion region 1) was identified as a c-myc cooperating oncogene using an Eμ-myc transgenic mouse model [
5,
6]. There is increasing evidence that the deregulated expression of BMI1 contributes to cancer development. It is overexpressed in a number of cancers, such as mantle cell lymphoma [
7], B-cell non-Hodgkin's lymphoma [
8], myeloid leukemia [
9], non-small cell lung cancer [
10], colorectal cancer [
11], breast and prostate cancers [
12,
13], and head and neck cancers [
14,
15]. In addition to its role in cancer, BMI1 is also known to be required for self-renewal of neural, hematopoietic, intestinal and mammary stem cells [
16‐
21]. Consistent with its role in stem cell self-renewal, BMI1 expression is thought to promote stem-ness in tumor cells [
12,
22], and BMI1 is considered an important marker of breast cancer stem cells [
23]. Recent mouse xenograft studies using BMI1 and Ras co-overexpressing human mammary epithelial cells (HMECs) also support oncogenic roles for BMI1 in breast cancer development and metastasis of breast cancer cells [
24,
25].
PcG proteins assemble into polycomb repressive complexes (PRCs), which possess histone posttranslational modification (PTM) activities and act in a sequential fashion to mediate gene silencing [
3]. Biochemically, BMI1 is a core component of PRC1, which ubiquitinates histone 2A at lysine 119 residue [
26], and acts downstream of PRC2, which trimethylates lysine 27 residue of histone 3 [
27,
28]. Although BMI1 is a prominent component of PRC1, its exact role in PRC1 is unclear. BMI1 by itself does not appear to have an E3 ubiquitin ligase activity [
29], instead, the E3 ubiquitin ligase activity of PRC1 strictly depends on Ring1B (RING2) protein. However, it has been shown that Ring1B-mediated E3 ubiquitin ligase activity of PRC1 complex is enhanced by BMI1 [
29‐
31].
Structurally, human BMI1 is comprised of 326 amino acids [
32]. The primary structure of BMI1 in mice revealed the presence of a RING finger (RF) domain at the N-terminus, a potential HTH (helix turn helix) domain in the middle and a PEST (proline (P), glutamic acid (E), serine (S) and threonine (T) rich) -like domain at the C-terminus [
5,
6]. These domains of BMI1 are highly conserved across mammalian species including human. The BMI1 also contains two putative nuclear localization signals (NLS), NLS1 (KRRR, amino acid residues 92-95) and NLS2 (KRMK, amino acid residues 232-235). Of these two, only NLS2 appears to be functional in targeting BMI1 to the nucleus in mouse and human cells [
33,
34]. We have previously carried out functional analysis of BMI1 and shown that the RING finger and HTH domains of BMI1 are required for downregulation of p16INK4a tumor suppressor and bypass of senescence in human diploid fibroblasts (HDFs) [
35]. We also showed that both of these domains are required for immortalization of normal HMECs [
34].
PEST-like domains rich in proline (P), glutamic acid (E), serine (S) and threonine (T) residues have been described in the literature [
36]. Although the actual contribution of such domains to protein turn-over is not clear and such sequences may not mediate proteolysis per se, the PEST-like domains are present in many proteins that undergo rapid turn-over such as cyclins, NF-κβ, c-Myc, c-Fos, ODC, ABCA1 etc [
36‐
44]. Since a PEST-like domain, which is rich in proline (P) and serine (S) residues is present in BMI1 [
5,
6], and at present very little is known about its functional role, we carried out a structural-functional analysis of BMI1 to define the role of this domain in BMI1. As the PEST-like domain in BMI1 is not rich in E- and T- residues, here in, we refer this domain as the PS domain. Our data indicate that the PS domain of BMI1 may indeed regulate its proteolysis and that it may function as a negative regulatory domain of BMI1.
Discussion
The mouse
Bmi1 gene was cloned by two different groups in 1991 [
5,
6], and shortly after that, the human homologue of mouse Bmi1 was cloned [
32]. Since then most of the studies on BMI1 have been focused on its role in oncogenesis and stem cell phenotype, and very few studies have analyzed the regulation of BMI1. Only recently, we and others analyzed the transcriptional regulation of
BMI1 [
47,
48]. At present, virtually nothing is known about additional transcriptional or posttranslational regulation of BMI1. Such regulation of BMI1 is likely to be highly relevant for the role of BMI1 in development and pathobiology of various diseases, including cancer.
The initial characterization of mouse and human BMI1 predicted the presence of three structural domains in BMI1; a RING finger domain, an HTH domain and a potential PEST domain. Except for the RING finger domain, the functional role of these domains is largely undefined. The RING finger domain is most recognizable domain in BMI1. Although such domains are found in E3 ubiquitin ligases, BMI1 by itself lacks E3 ubiquitin ligase activity. Nonetheless, this domain has been shown to be functionally important [
15,
34,
35,
49].
A RING finger domain mutant of Bmi1 (the mouse homologue of human BMI1) was shown to be ineffective in collaborating with c-myc to induce B- and T- cell lymphomas [
49]. Moreover, the deletion of the RING finger domain of BMI1 may lead to dominant negative activity of the mutant BMI1 [
35], as the overexpression of the ΔRF mutant of BMI1 induces p16INK4a expression and premature senescence, whereas the expression of the wild type BMI1 suppresses p16INK4a expression and bypasses senescence in HDFs [
35]. Alkema et al. also showed that in addition to the RING finger domain, a centrally located HTH domain of Bmi1 was also required for B- and T- cell lymphomagenesis by Bmi1 and c-myc [
49]. Interestingly, the PS region of Bmi1 was not required to induce B- and T- cell lymphomas in the mouse model [
49]. Thus, at present, it is not clear what role the PS domain of BMI1 plays in oncogenesis.
In the present study, we report that deletion of the PS domain of BMI1 increases its half life and that the mutant protein becomes very stable. Although naturally occurring deletions in the PS region of BMI1 have not been reported, our studies form the basis of detailed analysis of posttranslational regulation of BMI1 and biochemical analysis of its degradation pathways. With respect to biochemical analysis of BMI1 proteolysis, we noticed the presence of a degron motif DSG(X)
2+nS, in the PS domain of BMI1. The degron motif is a recognition site for the βTRCP (β-transducin repeat containing protein), which is a member of the F-box protein family and constitutes one of the subunit of the ubiquitin ligase complex known as SCF (SKP-Cullin-F-Box) [
50]. The SCF complex is involved in proteosome-mediated degradation of several proteins known to be involved in cell cycle and oncogenesis such as CDC25A [
51‐
53], Gli [
54,
55] Mcl-1 [
56], and PDCD4 [
57]. It is tempting to speculate that a βTRCP-based SCF complex posttranslationally regulates the stability of BMI1. Identification and detailed characterization of such an SCF complex and its constituents remain to be undertaken.
We reasoned that if the deletion of the PS region makes BMI1 more stable, overexpression of the ∆PS mutant of BMI1 may augment known biological activities of the wild type protein and/or exhibit additional biological activities. The best known biological activity of BMI1 is regulation of proliferation, which it upregulates via p16INK4a-dependent and -independent pathways. The p16INK4a-independent pathway includes AKT and its targets, and possibly other growth regulatory genes. Indeed, our results indicate that deletion of the PS domain of BMI1 results in augmentation of pro-proliferative activity of BMI1 via both p16INK4a-dependent and -independent pathways in different cell types. Specifically, overexpression of the ∆PS mutant in HMECs results in upregulation of AKT and its targets such as cyclin D1, and increased proliferation. On the other hand in primary HDFs, which express p16INK4a, overexpression of the ∆PS mutant results in significant downregulation of p16INK4a accompanied by increase in proliferation and decrease in senescence.
Apart from the increase in proliferation in both epithelial cells and fibroblasts, overexpression of the ∆PS mutant also results in induction of an EMT phenotype, which is often accompanied by downregulation of E-cadherin and upregulation of fibroblastic markers such as fibronectin and vimentin. The EMT phenotype is thought to be important not only for development of an organism but also for certain pathological conditions such as cancer and organ fibrosis [
58]. PcG proteins, in particular EZH2 has been shown to downregulate E-cadherin in breast and prostate cancer cells [
59,
60]. In addition, a recent report suggests that BMI1 can downregulate E-cadherin expression and induce EMT via downregulation of PTEN in nasopharyngeal epithelial cells (NEPC) [
61]. However, our unpublished data suggests that in HMECs, overexpression of the wild type BMI1 is not sufficient to downregulate PTEN. It is possible that unidentified potential pro-oncogenic lesions in immortalized NEPC and nasopharengeal carcinoma cooperate with the wild type BMI1 to downregulate PTEN expression and induce EMT. A different recent report suggests that BMI1 interacts with nuclear PTEN [
62]. Interaction of BMI1 with PTEN may also inhibit its function, and this interaction in part could induce an EMT phenotype. Stable BMI1 in the form of the ∆PS mutant may also upregulate EZH2, which in turn could directly downregulate the
CDH1 promoter and induce EMT. These hypotheses remain to be examined. In any case, it appears that by increasing BMI1 stability via deletion of its negative regulatory sequences, one can potentially induce EMT; although the mechanism of this induction is not clear at this point.
Although the exact role of the PS domain of BMI1 in the pathobiology of cancer remains to be explored, our studies imply that under certain pathological conditions, BMI1 may be more stable due to deregulation of its degradation machinery, which is likely to be regulated by the sequences present in the PS region of BMI1. So far there is no report of somatic oncogenic mutations or deletions occurring in the PS region of BMI1; however such a possibility cannot be ruled out. In summary, our data implicates the PS domain of BMI1 in its negative regulation, possibly via posttranslational mechanisms. Based on our new and published data, we propose that the RING finger domain of BMI1 can be classified as a catalytic domain, while the PS domain of BMI1 can be termed as a regulatory domain. In view of the well established roles of BMI1 in oncogenesis and stem cell maintenance, our studies have important implications for the pathobiology of cancer and the development of novel treatment strategies for cancer patients.
BMI1 and its target p16INK4a are also important regulators of in vivo aging. Recently it was shown that
Bmi1-/- mice present several growth defects, progeroid phenotype, such as hair loss, lens cataracts and reduced locomotor activity, and a high lethality rate exceeding 75% [
63]. Furthermore, the levels of p16INK4a , which is a well established target of BMI1, but not other CDK inhibitors increase with aging in vivo, and inhibition of p16INK4a can ameliorate the physiological impact of aging on stem cells and improve injury repair in aged tissues [
64‐
68]. As the deletion of the PS domain of BMI1 results in a stable and constitutively active BMI1, and leads to a robust downregulation of p16INK4a, we speculate that the PS region of BMI1 could also be targeted for the development of treatment strategies for age-related pathologies.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
GPD conceived, coordinated the study, interpreted data and wrote the manuscript. AKY, AAS, MD, PVB and RS performed all the experiments, and helped in interpretation of the data and the preparation of the manuscript. All authors read and approved the final manuscript.