BMI1, a new target of CK2α

AKT has been identified as the predominant kinase phosphorylating BMI1 on Ser 251, 253, 255 and 316 respectively, resulting in its altered cellular activity [3, 12]. However, inhibition of AKT by specific inhibitors (AKTi) reduced but did not eliminate the characteristic phospho-bands of BMI1 in ovarian cancer cells, CP20 and OV90 (Fig. 1a). That the slower migrating species represented phosphorylated form of BMI1 (Fig. 1a) was confirmed by λ-phosphatase treatment [15]. This suggests that in ovarian cancer cells additional kinases might be involved in phosphorylating BMI1. Stringent bioinformatics analysis (NetPhos2.0 Server [28], cutoff <0.9) of the BMI1 protein sequence revealed nine potential sites of phosphorylation (Table 1) including Ser 253, 255 and 316 that were previously described [3, 12]. Interestingly, all of the previously reported phosphorylation sites of BMI1 lie within the PS/PEST domain (236–326 amino acids), deletion of which increases half-life of the protein [15]. Interrogation of these phosphorylation sites with KinasePhos2 [29] analysis (Table 1) predicted previously reported kinases such as AKT [3, 12], as well as novel candidate kinases such as Ataxia Telangiectasia Mutated (ATM) and Casein Kinase II (CK2) (Table 1). While ATM/ATR dependent recruitment of BMI1 to sites of DNA damage has been reported [4], any association with CK2 remains unknown.

CK2 is a serine-threonine kinase comprising of two catalytic (α, α') and two regulatory (β) subunits. CK2 is upregulated in major lethal malignancies where its expression correlates with poor overall survival [16, 18, 19]. Interestingly of these subunits, high levels of CK2α correlated with poor overall survival in ovarian cancer [19]. Using immunoblotting, we determined that CK2α levels were significantly elevated in ovarian cancer cell lines compared to the non-malignant Fallopian Tube Epithelial (FTE188) cells (Fig. 1b). Since BMI1 is reportedly phosphorylated by AKT, we determined the expression of active AKT (Phospho AKT-S⁴⁷³) in these cell lines. With the exception of OV90, AKT was significantly phosphorylated in all the cancer cell lines compared to FTE188 (Fig. 1b). Interestingly, the phosphorylated species of BMI1 was present in all the cancer cell lines including OV90 that expressed minimal phospho-AKT, thereby implicating other kinases in phosphorylation of BMI1. Since the expression of CK2α correlated with the phospho-species of BMI1, we next determined if altering cellular levels of CK2α affected phosphorylation of BMI1. Silencing CK2α by siRNA significantly reduced both phospho-species as well as basal BMI1 levels in CP20 and OV90 cells (Fig. 1c). However the PRC1 complex partner of BMI1, RING1A levels remained unchanged (Fig. 1c). Subsequently we confirmed that in CK2α silenced cells, reduction in basal BMI1 level was not due to decreased transcription thus implicating post-translational events (Fig. 1d). These observations were further complemented by expressing CK2α in FTE188 cells that resulted in enhanced basal expression as well as phosphorylation of BMI1 (Fig. 1e). Together these results suggest that CK2α might potentially phosphorylate BMI1 and directly or indirectly regulate its post translational expression.

To determine if BMI1 is a direct substrate of CK2α, we performed reciprocal co-immunoprecipitation of endogenous BMI1 and CK2α using covalently cross-linked antibodies to protein A/G agarose beads. BMI1 and CK2α co-precipitated from lysates of both CP20 and OV90 cells (Fig. 2a), indicating that BMI1 might be a novel phosphorylation substrate of CK2α.

To confirm these observations, we performed an in vitro kinase assay using purified CK2α, GST-tagged BMI1 and [γ − ³²P] ATP. The autoradiograph clearly demonstrated that BMI1 was phosphorylated by CK2α (Fig. 2b). Additionally, as previously reported [30], autophosphosphorylation of CK2α was also observed (Fig. 2b). To further confirm these observations we first immunoprecipitated CK2α from cellular lysates of CP20 cells using antibody-conjugated agarose beads (Fig. 2c-left panel). Equal amount of these agarose beads were incubated with recombinant BMI1 and [γ − ³²P] ATP either in presence or absence of different concentrations of the CK2 inhibitor, TBB (4,5,6,7-tetrabromobenzotriazole) [31]. The autoradiograph clearly demonstrated that BMI1 was phosphorylated by CK2α and the signal significantly diminished with TBB (Fig. 2c-right panel). Together these results demonstrate that CK2α phosphorylates BMI1.

To determine the potential site/s on BMI1 that are phosphorylated by CK2α, full-length purified BMI1 was subjected to the in vitro kinase reaction in presence or absence of purified CK2α and non-radioactive ATP and then analyzed by mass spectrometry (MS). 19% of the detected BMI1 peptides were phosphorylated at Ser110 compared to none without CK2α as shown in the phosphopeptide spectrum (Fig. 3a and Additional file 1: Figure S1). The extracted ion chromatogram for the peptide is shown in the inset of Fig. 3a. By site-directed mutagenesis of the FLAG tagged wild-type BMI1 (WT) construct, we generated a S110A mutant and expressed these constructs in CP20 cells. Immunoblotting and densitometric analysis demonstrated that comparable expression could be achieved by transfection 1.5 μg of WT and 2 μg of the S110A-BMI1 mutant (Mut-BMI1) constructs respectively (Fig. 3b). Next, WT or Mut-BMI1 was immunoprecipitated using anti-FLAG antibody conjugated to protein A/G agarose beads (Fig. 3c, left panel). Equal amount of beads were incubated with full-length GST-tagged CK2α and [γ − ³²P] ATP for the in vitro kinase assay. Autoradiograph signals were obtained with WT-BMI1 but not with Empty vector (EV) control or with the Mut-BMI1 (Fig. 3c, right panel). Together these results indicate that BMI1 is phosphorylated by CK2α at the S110 residue.

Modulation of cellular CK2α levels impacted not only BMI1 phosphorylation but also total protein levels (Fig. 1c-d). In addition, though the plasmid backbones were identical we consistently observed lower expression of S110A compared to wild-type BMI1 (Fig. 3b) suggesting issues with stability of the mutant (non-phosphorylatable) protein. To investigate this possibility, the half-life of the FLAG tagged WT and S110A-BMI1 was determined in CP20 cells using 100 μg/ml cyclohexamide (CHX). As evidenced by immunoblotting and densitometry, WT-BMI1 had a half-life of ~23 min while S110A with a half-life of ~11 min degraded faster (Fig. 4a). Prior studies have reported a half-life of ~25–30 min for WT-BMI1 that is specifically degraded by the 26S proteasome-mediated pathway [15]. To determine if decreased stability of mutant BMI1 was due to proteasome-mediated degradation, we treated plasmid transfected CP20 cells with the proteasome inhibitor MG132 (10 μM) for the indicated times. Compared to the wild-type protein, consistently more S110A-BMI1 accumulated after treatment with MG132 suggesting that the mutation and thereby lack of phosphorylation rendered the protein less stable and more amenable to degradation by the proteasome (Fig. 4b). Similarly, compared to the control more accumulation of endogenous BMI1 could be observed in the MG132 treated CK2α silenced cells (Fig. 4c) further supporting that phosphorylation at S110 affects stability of BMI1.

BMI1 confers clonal self-renewal property to cells. Therefore using clonal growth assays, we determined the effect of WT or S110A-BMI1 in OV90 and CP20 cells. Simultaneously, endogenous BMI1 was silenced by a 3'UTR targeted siRNA and comparable levels of WT or S110A-BMI1 was expressed in these cells. Significant knockdown of endogenous BMI1 by siRNA and re-expression of the FLAG tagged WT and S110A-BMI1 was confirmed by immunoblotting (Fig. 5a). Compared to the control, significant decrease in clonal growth was observed in the BMI1 silenced cells (~55% in OV90 and ~45% in CP20; Fig. 5b-c). Re-expression of WT-BMI1 almost completely rescued clonal growth in endogenous BMI1 silenced cells while the S110A-BMI1 had no effect (Fig. 5b-c).

Based on these results we posited that if CK2α mediated BMI1 phosphorylation led to protein stability, then a correlation between CK2α and BMI1 protein levels would be expected in ovarian cancer clinical specimens. Therefore, we analyzed BMI1 and CK2α levels in a panel of twenty high-grade serous ovarian tumors and two normal fallopian tube epithelial (FTE) tissues [32] by immunoblotting. BMI1 expression was significantly higher in primary tissues (p = 0.0398) and CK2α was detected in a significantly higher proportion of primary tissues (p = 0.013). Remarkably within the tumor samples, significant correlation between CK2α and BM11 expression was observed (Fig. 5d, Spearman correlation coefficient is 0.62, p = 0.0021). While patients with higher expression of both BMI1 and CK2α showed a trend towards shorter progression free survival, the data did not reach statistical significance likely due to the limited sample size (Additional file 1: Figure S2). Together, these results demonstrate that phosphorylation of BMI1 at Ser110 by CK2α is important for both stability and functionality of the protein.