RNAi Screen Identifies a Synthetic Lethal Interaction between PIM1 Overexpression and PLK1 Inhibition

Members of the PIM family of serine/threonine kinases (PIM1, PIM2, and PIM3) are overexpressed in a variety of malignancies, including leukemias, lymphomas, prostate, and pancreatic cancers (1, 2). The PIM kinases increase cell survival, proliferation, and tumorigenicity and appear to play these roles by phosphorylating multiple substrates including Cdc25A, NuMA, p21, Bad, C-TAK1, Cdc25C, and p27 (3–9). A highly notable feature of PIM kinase—driven tumorigenesis is the dramatic cooperativity between PIM kinases and MYC. In prostate cancers, PIM1 and MYC are frequently co-expressed (10, 11), and recent work in animal models has shown that PIM1 synergizes with c-MYC to induce advanced prostate cancer in a kinase-dependent manner (11). PIM1 can stabilize MYC protein levels and enhance MYC transcriptional activity (12). Importantly, PIM1 is required to maintain the tumorigenicity of MYC/PIM1-expressing tumor cells (13), supporting the notion that PIM1 could be a valid therapeutic target. Accordingly, there are many ongoing efforts aimed at developing small-molecule inhibitors of the PIM kinases as anticancer therapeutic agents. PIM1 inhibition is potentially an attractive strategy for treating prostate cancer as Pim kinase deficiency in mice is generally well tolerated, suggesting that PIM kinases are not required for essential cellular functions. Furthermore, the presence of a unique hinge region in the ATP-binding site of PIM1 facilitates the development of specific small-molecule kinase inhibitors.

Conceptually, another approach to targeting PIM kinase–expressing tumor cells is to identify the specific vulnerabilities of these cells. In this study, we used RNAi screening to systematically identify genes whose expression is required for the viability of PIM1-expressing prostate epithelial cells. RNAi screening has been used to identify synthetic lethal interaction between genes with relevance to cancer treatment (14, 15).

Here, we used a collection of siRNAs that target genes encoding selected serine/threonine kinases, tyrosine kinases, cell-cycle protein, and apoptosis proteins to identify genes that may be potential targets for inhibiting PIM1-expressing cells, leading to the identification of PLK1. PLK1 is a mitotic regulator that plays a crucial role at various steps of mitosis and is overexpressed in many tumor types including prostate cancer, where PLK1 overexpression was found to correlate with Gleason grade (16). The inhibition of PLK1 has been shown to have potent antitumor effects in experimental in vitro or in vivo models (17–20). The fact that PLK1 is required for normal mitotic progression has raised some concerns about the potential toxicity of anti-PLK1 therapeutic agents. However, in principle, the identification of molecular changes that make tumor cells more sensitive to the effects of PLK1 inhibition will lead to an increase in therapeutic index and better tolerability. Our findings demonstrate that molecular changes induced by oncogenes such as PIM1 can make cancer cells particularly sensitive to the inhibition of PLK1, a feature that can be exploited for therapeutic purposes.

RWPE-diploid and polyploid cells (21) were screened for siRNA libraries targeting a total of 570 genes involved in key cancer relevant pathways (111 cell-cycle, 318 apoptosis, 87 serine/threonine kinase, and 54 tyrosine kinase genes; siRNA library, Dharmacon). Before the screening, transfection condition and reagents were optimized and validated using reverse transfection (Dharmacon). The data from both cell lines were combined and hits were determined by Z-scores ≥2 or ≤−2 (22). For validation, we selected 10 genes including some of the top hits that reduced cell viability. The siRNAs against these 10 genes were custom ordered and tested using RWPE1-Pim1 and control RWPE1-Neo cells to identify genes whose knockdown specifically affected cell viability in Pim1-overexpressing cells. Cellular viability was determined after 72 hours of reverse transfection by using the CellTiter-Glo Luminescent cell viability assay (Promega).

All the cell lines were authenticated. RWPE-1, LNCaP, and PC3 cells were purchased from ATCC and were maintained in keratinocyte serum–free medium (KSFM for RWPE1) or RPMI medium (for LNCaP and PC3) with 10% FBS in a humidified 37°C incubator with 5% CO₂. RWPE-Neo, Pim1, diploid, and polyploid cells have been described in previous articles (21, 23, 24). LNCaP-Neo/Pim1 and PC3-Neo/Pim1 cells have also been described previously (23). NHPrE cells were maintained in F12/DMEM medium as described (25). To establish PLK1-knockdown cells, lentiviral PLK1 shRNAmir and control shRNAmir (Open Biosystems) were transduced into LNCaP-Neo/Pim1 and PC3-Neo/Pim1 cells, and stable clones were selected by using 1 to 2 μg/mL puromycin. For BI 2536 (Selleckchem) treatment, different doses of BI 2536 (10–100 nmol/L) were added to the cells and cell lysates were prepared 24 hours later.

Western blotting was performed as described previously (23). The following antibodies were used: PIM1 (Santa Cruz, sc-13513), PLK1 (Santa Cruz, sc-17783), phospho-PLK1 (Thr 210; Cell Signaling, #5472), phospho-histone H3 (Upstate, #06–570), cleaved PARP (Cell Signaling, # 9541), MYC (Abcam, ab32072), and phospho-serine 62-MYC (Abcam, ab51156). For immunoprecipitation, lysates prepared with the lysis buffer [50 mmol/L HEPES, pH 7.4, 150 mmol/L NaCl, 1 mmol/L EDTA, 0.2% NP-40, 1× protease inhibitor cocktail (Roche)] were incubated with mouse α-PIM1 or α-PLK1 antibodies overnight at 4°C, followed by incubation with the protein A/G agarose beads for 2 hours at 4°C. The lysates were washed with the immunoprecipitation/washing buffer 3 times, then the proteins bound to the beads were eluted in 2× SDS sample buffer, separated by SDS-PAGE, and blotted with the antibodies.

Cells were processed as reported previously (23). Briefly, cells on coverslips were fixed with 4% paraformaldehyde for 15 minutes at room temperature. After washing with PBS 3 times, the cells were permeabilized with 0.2% Triton X-100 in PBS for 5 minutes. Following washing and blocking, cells were incubated with the appropriate primary antibodies overnight and incubated with fluorescent secondary antibodies (molecular probe). After washing and staining with 4′,6-diamidino-2-phenylindole (DAPI), slides were mounted, sealed, and examined.

The kinase assay was carried out in the kinase buffer (20 mmol/L PIPES, pH 7.0, 5 mmol/L MnCl₂, 0.25 mmol/L β-glycerophosphate, 0.4 mmol/L spermidine, and 7 mmol/L β-mercaptoethnol) containing 10 μmol/L nonradioactive γATP, 10 μCi [³²P]ATP, 2 μmol/L aprotinin, and 0.1–2 μg PIM1 or PLK1. The reactions were then incubated at room temperature for 30 minutes, stopped by SDS sample buffer, separated by SDS-PAGE, fixed, amplified, dried, and the film was developed.

For xenograft studies, cells (10⁷ for both LNCaP and PC3) modified to overexpress Pim1 with or without stable PLK1 knockdown were mixed with 200 μL of Matrigel (Beckton Dickinson Labware) and injected subcutaneously into the flanks of 8-week-old male nu/nu mice, and tumor size was measured once a week for 5 weeks. For BI 2536 treatment, LNCaP-Pim1/Neo cells (10⁷ cells per graft) were grafted subcutaneously in nu/nu mice, and 1 week later, BI 2536 were injected intravenously at a dose of 25 mg/kg twice a week on 2 consecutive days for 3 weeks. Tumor size was measured once a week. Nude mice bearing PC3-Pim1 and Neo were also treated with BI 2536 i.v. for 6 cycles at a dose of 25 mg/kg twice a week starting from 5.5 weeks after grafting. All samples were processed for hematoxylin and eosin (H&E) and immunostaining with the following antibodies: phospho-histone H3 (Upstate, #06–570), activated caspase-3 (Cell Signaling, #9661), cleaved PARP (Cell Signaling, #9541), and MYC (Abcam, #ab32072).

To identify genes selectively required for the viability of Pim1-expressing cells, we used RWPE-Pim1 cells. These cells stably overexpress Pim1 at expression levels previously determined to be within the range seen in human prostate tumors (23). With additional passaging, RWPE-Pim1 cells become polyploid and tumorigenic (21, 24). We used early-passage, diploid, and nontumorigenic RWPE-Pim1 and late-passage, polyploid, and tumorigenic RWPE-Pim1 cells in our screen to capture potential differences in genetic vulnerability between the diploid and polyploid cells. We screened a panel of siRNAs targeting 111 cell-cycle, 318 apoptosis, 87 serine/threonine kinase, and 54 tyrosine kinase genes in both cell lines. Each gene was targeted by 4 different siRNAs arrayed in a 96-plate format that facilitates reverse transfection. Cellular viability was determined after 72 hours of reverse transfection by using the CellTiter-Glo Luminescent cell viability assay, and the results were normalized to RISC-free siRNA as a control. We did not observe any consistent differences in viability between the diploid and the polyploid RWPE1-Pim1 cells upon knockdown of any of the genes examined.

We therefore combined the data from both cell lines to determine the effect of siRNAs on cell viability of RWPE-Pim1 cells regardless of ploidy status. Figure 1A shows a ranking of individual siRNA effects on cell viability relative to control set by Z-scores. Notably, the siRNAs that resulted in increased cell viability (Z ≥ 2) in this initial screen are known pro-apoptotic proteins including BAX, APAF1, APLP1, API5, ARHGEF6, BAG1, and ASC. Conversely, among the genes whose siRNAs reduced viability (Z ≤ −2) are MAPK1 and BRAF, which are part of the growth factor receptor/MAPK signaling pathway. Because RWPE1 cells are grown in media supplemented with mitogens including EGFR and bovine pituitary extract [which contains basic FGF and platelet-derived growth factor (PDGF)], these results might simply reflect the fact that pathways mediated by these growth factors are important for the viability for RWPE1 cells.

The most significant reduction in cell viability was seen for siRNAs targeting SON, a DNA-binding protein (also known as Bax antagonist selected in Saccharomyces 1) and PLK1 (Fig. 1B). SON is a Ser/Arg (SR)-related protein that functions as a splicing cofactor. Depletion of SON has been shown to result in severe impairment of proper spindle pole and microtubule function leading to genomic instability (26). Interestingly, these molecular defects result from inefficient RNA splicing of a specific set of cell-cycle–related genes that possess weak splice sites in SON-depleted cells (26). Inhibition of PLK1 has been shown to inhibit proliferation of prostate cancer cells preferentially compared with nontransformed prostate epithelial cells (27).

To determine whether any of these siRNAs selectively impairs the viability of Pim1-overexpressing cells, we performed a secondary screen including 10 genes in RWPE-Neo and RWPE-Pim1 cells. As shown in Fig. 1B (bottom), in this assay, SON depletion affected RWPE-Neo and RWPE-Pim1 viability equally, whereas PLK1 siRNA showed a selective effect on RWPE-Pim1 cells.

To confirm and extend this observation to additional cell lines, we decided to knockdown PLK1 in other prostate tumor cell lines using shRNA targeting PLK1 (shPLK1). We first examined PLK1 levels and found that PLK1 is robustly expressed in all 3 prostate cancer cell lines tested (DU145, PC3, and LNCaP) but not in the nontransformed cell line NHPrE (Supplementary Fig. S1A). We established stable PLK1-knockdown cells by shRNA in LNCaP and PC3 cells with and without Pim1 overexpression (Fig. 1D). Cell lines with severe PLK1 depletion were not viable, thus cells with modest amounts of PLK1 knockdown were selected and used in this study. Analysis of cell viability of these cells indicates that Pim1-overexpressing LNCaP and PC3 cells are more sensitive to the inhibitory effect of PLK1 knockdown than control Neo cells (Fig. 1D). To examine whether Pim1 affected PLK1 expression levels, Pim1-overexpressing cells as well as control Neo cells were synchronized in mitotic phase using nocodazole and PLK1 expression levels were examined. Results showed that there was no difference in PLK1 expression levels between Pim1-overexpressing cells and control Neo cells, indicating that increased sensitivity of Pim1-overexpressing cells to PLK1 knockdown is not due to the ability of Pim1 to affect PLK1 expression levels (Supplementary Fig. S1C).

To explore possible mechanisms for the observed dependency of Pim1-overexpressing cells on PLK1, we examined for a possible interaction between Pim1 and PLK1. For this, Pim1 and PLK1 were co-transfected into 293T cells and samples were processed for co-immunoprecipitation using either Pim1 or PLK1 antibodies. Our result indicated that Pim1 and PLK1 do interact (Fig. 2A). Next, immunofluorescence experiments were performed to determine the co-localization of PIM1 and PLK1. PLK1 was detected in centrosome, kinetochore, and midbody during mitosis (Fig. 2B), consistent with its multiple mitotic functions as reported (28). Interestingly, we detected co-localization of PIM1 and PLK1 in the centrosome and midbody, but not in the kinetochore (Fig. 2B).

Because both Pim1 and PLK1 are serine/threonine kinases, we next examined whether one can phosphorylate the other. We tested the activity of PIM1 by incubating it with histone H1, a known Pim1 substrate. Wild-type recombinant PIM1 was active as shown by PIM1 autophosphorylation and by phosphorylation of histone H1, but histone H1 was not phosphorylated by kinase dead mutant form of PIM1, K67M, or GST-only recombinant protein (Fig. 2C, left, lane 1–3). When we mixed wild-type PIM1 with PLK1, phosphorylated forms of both PIM1 and PLK1 were detected (Fig. 2C, left, lane 4). To distinguish which molecule phosphorylates which molecule, we performed a kinase assay with fixed amounts of PLK1 and increasing amounts of PIM1. PLK1 phosphorylation was not observed in the absence of PIM1 but became apparent with increasing amounts of PIM1 (Fig. 2C, right, lanes 1–5). Incubation of a fixed amount of PIM1 with increasing amounts of PLK1 revealed a dose-dependent increase in phosphorylated PLK1 but not in phosphorylated PIM1 (Fig. 2C, right, lanes 6–9). PIM1 is known to autophosphorylate itself and because the phosphorylated PIM1 band intensity remained constant with increasing amounts of PLK1 (Fig. 2C, right, lanes 6–9), these data indicate that PIM1 is not phosphorylated by PLK1 under these condition. Overall, these results indicate that PLK1 is a PIM1 substrate. To further investigate phosphorylation of PLK1 by Pim1, we overexpressed PLK1 and Pim1 and then examined phosphorylation of PLK1 using phospho-specific antibody (p-Thr210). Specifically, HA-PLK1 and Pim1 were transfected into 293 T cells and cells were treated with nocodazole to induce mitotic phase because PLK1 is reported to be phosphorylated in vivo in mitosis (29). Phosphorylation of PLK1 was increased after nocodazole treatment without Pim1 overexpression as expected (Fig. 2D, lanes 2 and 3), and Pim1 overexpression further increased phosphorylation of PLK1 after nocodazole treatment (Fig. 2D, lanes 6 and 7). These data strongly indicate that Pim1 phosphorylates PLK1 at threonine 210, a site previously reported to be phosphorylated by aurora A kinase during mitosis (30).

We next sought to determine whether PLK1 depletion will impair the in vivo tumorigenicity of Pim1-overexpressing cells. We generated xenograft tumors using LNCaP-Pim1 cells or LNCaP-Neo cells with and without stable shPLK1 expression (Fig. 3A). As noted earlier, we selected cells with modest reduction in PLK1 because cells with drastic knockdown of PLK1 were not viable. In this assay, we found that Pim1-overexpressing LNCaP cells with PLK1 knockdown (LNCaP-Pim1/shPLK1 cells) formed significantly smaller tumors than the LNCaP-Pim1 with control shRNA (LNCaP-Pim1/Control cells; Fig. 3A, left and right). In contrast, no differences were observed between LNCaP-Neo/Control and LNCaP-Neo/shPLK1 cells (Fig. 3A, middle and right).

To confirm this finding, we used the PLK1-specific small-molecule inhibitor, BI 2536. In the first set of experiments, BI 2536 was given early during tumor development (1 week postgrafting). We found the dramatic inhibition of LNCaP-Pim1 tumor growth by BI 2536 treatment (Fig. 3B, left and right). In contrast, control LNCaP-Neo tumors treated with BI 2536 demonstrated the modest inhibition of tumor growth (Fig. 3B, middle and right). PLK1 inhibition is reported to arrest cells in mitotic phase (M phase) followed by apoptosis (28). Accordingly, BI 2536–treated LNCaP tumors showed evidence of arrest in M phase as determined by histone H3 phosphorylation (Fig. 3C, left, Supplementary Fig. S2A) as well as apoptosis determined by active caspase-3 staining (Fig. 3C, right, Supplementary Fig. S2B). Interestingly, LNCaP-Pim1 tumors treated with BI 2536 showed much higher rates of M phase arrest and apoptosis than BI 2536–treated control LNCaP-Neo tumors.

Next, we examined the efficacy of BI 2536 treatment in established tumors. For this experiment, we used the more aggressive PC3 cell line. BI 2536 was administered after tumors have developed (5.5 weeks postgrafting). The results showed that the PC3-Pim1 tumors regressed at a faster rate than the PC3-Neo tumors (Fig. 3D, left). Calculation of the slopes of tumor growth and regression showed that PC3-Pim1 tumors regressed at twice the rate of PC3-Neo tumors upon BI 2536 treatment (slope: 354.5 vs. 171.8; Fig. 3D, right).

To further understand the underlying mechanism responsible for the increased sensitivity of Pim1-expressing cells to PLK1 inhibition, we first checked expression levels of PLK1 in Neo and Pim1 cells. Because our data showed that Pim1 does not consistently modulate PLK1 expression levels even in nocodazole-synchronized cells (Supplementary Fig. S1C), we then treated LNCaP-Neo and-Pim1 cells with different dose of BI 2536 to find the differential effects of Neo and Pim1 cells to PLK1 inhibition. Results showed that there are more mitotic-arrested cells in Pim1 cells than in Neo cells as shown by phospho-specific histone H3 (p-HH3) immunofluorescence (Fig. 4A and B) and by Western blotting of p-HH3 (Fig. 4C). The difference is more obvious in lower dose of BI 2536, most likely due to saturation effects of BI 2536 in higher dose. In addition to the increased p-HH3, the apoptotic marker PARP, another characteristic of PLK1 inhibition, also increased more in BI 2536–treated Pim1-expressing cells than control cells, especially at a lower dose (Fig. 4C). These data suggest that Pim1 cells are more sensitive to the inhibitory effects of BI 2536 as shown by increased mitotic arrest and apoptosis.

A recent study suggested a link between PLK1 and the regulation of MYC protein levels in the G₂ phase of the cell cycle (31). The ubiquitin ligase β-TrCP was found to ubiquitylate MYC, leading to increased MYC stability. Phosphorylation of MYC by PLK1 increases its association with β-TrCP, thereby enhancing MYC stability. Thus PLK1 inhibition might be expected to lead to a reduction in MYC protein levels. We tested this idea by examining MYC protein levels in cells following PLK1 inhibition by BI 2536. Interestingly, MYC levels were much lower in Pim1 cells than in control Neo cells after BI 2536 treatment (Fig. 4C), and this is correlated with phosphorylation of MYC (Fig. 4C). Furthermore, immunostaining showed reduced levels of MYC expression in both LNCaP-shPLK1 cells and BI 2536–treated Pim1 xenografts compared with control Neo cells (Fig. 4D, Supplementary Fig. S2C and S2D). MYC expression levels were also lower in BI 2536–treated PC3-Pim1 cells than in control Neo cells (Supplementary Fig. S3B). Decreased MYC expression levels were also observed in parental LNCaP, DU145, and PC3 cells after BI 2536 treatment (Supplementary Fig. S3C). The reduced MYC expression seen in cells in which PLK1 is inhibited is particularly interesting in light of the known cooperativity between MYC and PIM1. Together our results suggest that PIM1-expressing cells are hypersensitive to PLK1 inhibition most likely due to increased mitotic arrest followed by apoptosis as well as reduced MYC protein levels upon PLK1 inhibition.

On the basis of the data presented so far, we decided to examine whether there is a relationship between PIM1 and PLK1 in human prostate tumor tissue microarrays. Expression levels of PIM1 or PLK1 were much higher in tumor samples than in normal prostate tissues, and they were frequently co-localized (Fig. 5A). Of the 162 specimens examined, PIM1 staining was observed in 77 (48%) cases and PLK1 staining was present in 55 (34%) cases (Fig. 5B). There was a considerable overlap between samples that overexpress both PLK1 and PIM1 (49 cases; 30. 2%; Fig. 5B). In addition, co-expression of PLK1 and PIM1 was significantly correlated to higher Gleason scores. The majority (66.8%) of PLK1/PIM1 high expressing samples are of high Gleason score 8–10, whereas the majority (69.6%) of the PLK1/PIM1 low expressing samples are of low Gleason score less than 7 (P = 0.0012; χ² test; Fig. 5C). These results support the idea that PLK1 and PIM1 are frequently co-expressed in human prostate tumors.

In these studies, we used a focused RNAi screen to identify PLK1 as a target whose inhibition impairs the viability of PIM1-expressing prostate tumor cells. PIM1 is overexpressed in numerous solid tumors including prostate cancers (1, 2, 32) and has been known to play significant roles in tumorigenesis and chemoresistance in various cancers cells (33–36). As such, a number of PIM1 inhibitors have been developed (37); however, most of these have not been tested yet either in vitro or in vivo. In this study, we used an alternative strategy to target PIM1; we reasoned that the molecular changes induced by overexpression of PIM1 might render cancer cells particularly sensitive to the knockdown of certain genes. Our studies led to the finding that PIM1-expressing cells are particularly sensitive to PLK1 inhibition.

PLK1 is overexpressed in a wide variety of malignancies including prostate cancer and its expression frequently correlates with poor patient prognosis (16, 28, 38, 39). It plays a key role in cell division and its activity is elevated in cells with a high mitotic index including cancer cells (27, 28). Notably, a genome-wide RNAi screen has identified PLK1 as the kinase selectively required for the viability of activated Ras (14). All these data coupled with the unique structure of PLK1 have made PLK1 an attractive anticancer drug target. Several inhibitors targeting PLK1 have been developed so far and they have been under investigation in multiple clinical trials (28, 40).

In our study, we chose to use PLK1-specific, small-molecule inhibitor, BI 2536. BI 2536 is an ATP-competitive PLK1 inhibitor identified through high-throughput screening. BI 2536 showed high efficacy in vivo at well-tolerated doses and caused tumor regression in several xenograft models (19). Several clinical trials that include hormone-refractory prostate cancer have also revealed that BI 2536 exhibits some antitumor activity in patients (41, 42). A clinical trial with BI 2536 as a single agent administered to patients with prostate cancer showed some sign of antitumor activity measured by prostate-specific antigen (PSA; ref. 43). However, this study was performed on a small scale in patients with undefined genetic backgrounds. It is well known that not all patients respond to the same drugs and the extent of PSA decline and measurable tumor regression are variable. One explanation for this could be different genetic backgrounds of each individual as well as their different prior treatment options. In this regard, our finding that PIM1-overexpressing prostate cells show better response to PLK1 inhibition is intriguing.

One of the major concerns in drug targeting for cancer therapy is the potential toxicity in normal tissues. Inhibition of PLK1 induces apoptosis and cancer cells seem to be more sensitive to PLK1 inhibition than normal cells (14, 44–46). In this study, we found that even in tumor cell lines such as LNCaP and PC3 cells, there exists differential sensitivity to PLK1 inhibition because Pim1-overexpressing LNCaP or PC3 cells, which are more tumorigenic than their control cells, are much more sensitive to PLK1-inhibitory effects than control cells. Thus, identifying genetic changes such as PIM1 overexpression in individual tumors might be of value in selecting patients to be put on PLK1 inhibitor therapy regimens.

Mitotic arrest and apoptosis after PLK1 inhibition are well documented in previously published literatures (19, 27, 47). In this study, we observed more mitotic arrested cells and apoptotic cells in Pim1-overexpressing cells after PLK1 inhibition than in control cells. Previously, we reported that Pim1 overexpression induces genomic instability characterized by polyploidy, abnormal tubulin, and defects in mitotic checkpoint and cytokinesis (23). As mentioned earlier, PLK1 plays key roles in regulating cell-cycle–related events such as bipolar spindle formation, centrosome maturation, chromosome segregation, activation of the anaphase-promoting complex/cyclosome (APC/C), and cytokinesis (28). Thus, the mitotic stress that Pim1-overexpressing cells are experiencing might be exacerbated when the function of key cell-cycle–related molecules such as PLK1 is disturbed. Previous reports that cells with activated Ras or p53 mutation–bearing cancer cells depend more on PLK1 for their viability than their isogenic cells support this idea (14, 45). It is possible that PIM1 and PLK1 have common downstream effectors that are required for cell-cycle progression. Our data showing the co-localization of PIM1 and PLK1 in centrosome and midbody are consistent with this notion.

A possible common effector for both PIM1 and PLK1 is MYC. PIM1 interacts with MYC and increases its transcriptional activity (48). On the other hand, PLK1 stabilizes MYC as BI 2536 treatment decreases MYC levels in cells in culture (31). We observed reduced MYC expression in Pim1 xenografts and cells after BI 2536 treatment (Fig. 4D, Supplementary Fig. S2C and S2D). Thus, one of the potential mechanisms by which inhibition of PLK1 impairs the tumorigenicity of PIM1-expressing cells is by destabilizing MYC.

In summary, this study identified PLK1 as a synthetic partner of PIM1 and provides a rationale for the potential clinical use of PLK1 inhibition in PIM1-overexpressing prostate cancer. Our study represents an attempt to assess the impact of PLK1 inhibition in genetically defined tumor model systems with Pim1 overexpression. It would be interesting to test PLK1 inhibition in additional model systems that faithfully reflect human prostate cancer.

No potential conflicts of interest were disclosed.

Conception and design: S.A. Abdulkadir, M. Roh

Development of methodology: H.Y. Song, S.-H. Park, S.A. Abdulkadir, M. Roh

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.A. Abdulkadir, M. Roh

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.A. Abdulkadir, M. Roh

Writing, review, and or revision of the manuscript: S.A. Abdulkadir, M. Roh

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. van der Meer, H.Y. Song, S.-H. Park, M. Roh

Study supervision: S.A. Abdulkadir, M. Roh

This study was supported by the Department Of Defense (grant no. W81XWH-10-1-0246) and the National Cancer Institute (grant no. CA123484).

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.