Plk1 overexpression suppresses tumor development by inducing chromosomal instability

Polo-like kinase 1 (Plk1) is a protein kinase currently considered as an attractive cancer target due to its critical role in the cell division cycle. Plk1 is overexpressed in a wide spectrum of human tumors, being frequently considered as an oncogene. However, its contribution to tumor development is unclear. Using a new inducible knock-in mouse model we report here that Plk1 overexpression does not favor cell proliferation but rather results in abnormal chromosome segregation and cytokinesis, leading to the formation of polyploid cells with reduced proliferative potential. Mechanistically, these cytokinesis defects correlate with defective loading of Cep55 and ESCRT complexes to the abscission bridge during cytokinesis in a Plk1 kinase-dependent manner. In vivo, elevated levels of Plk1 markedly prevent the development of mammary gland tumors induced either by KrasG12D or Her2, in the presence of increased rates of chromosome instability. In patients, higher Plk1 expression levels are associated with significantly increased overall survival in breast cancer subtypes. These data suggest that, despite the therapeutic benefits of inhibiting Plk1 due to its essential role in tumor cell cycles, Plk1 overexpression has tumor suppressive properties by perturbing mitotic progression and cytokinesis.


Introduction
Chromosomal instability (CIN) is a frequent feature both in solid and hematopoietic human tumors 1,2 . Although its causal role during tumor development is still under careful experimental scrutiny, it is now clear that CIN provides specific clones with a variety of chromosomal combinations that may favor either tumor growth or resistance to antitumor therapies [3][4][5] . Multiple oncogenic alterations may induce CIN although the copy number aberrations that ultimately arise do so as a consequence of defects in the cellular machinery that regulates chromosome segregation and protects from unequal chromosome inheritance during mitosis 1,2 . Whether alteration in the levels of mitotic proteins is a cause or consequence of CIN is not clear, although experimental overexpression of several mitotic regulators such as Mad2 (Ref. 6 ), cyclin B1 and cyclin B2 (Ref. 7 ) as well as Aurora B (Ref. 8 ) induces CIN and spontaneous tumor formation in mouse models 9 .
Plk1 is the most studied member of a conserved family of protein kinases (Plk1-5) involved in cell division as well as specific functions in postmitotic cells such as neurons 10 or smooth muscle cells 11 . Plk1 was originally identified in Drosophila as a protein involved in spindle formation and further studies have suggested critical functions for this kinase in centrosome biology, spindle dynamics, chromosome segregation and cytokinesis 12,13 . Genetic ablation of Plk1 or its chemical inhibition results in defective chromosome segregation commonly accompanied by cell cycle arrest or cell death in a variety of model organisms 13,14 . Plk1 induction has been proposed to play a role at early stages during the progression of certain carcinomas and its overexpression inversely correlates with the survival rate of patients with non-small cell lung, head and neck, and esophageal cancer among others [15][16][17][18] . Plk1 inhibition with specific small molecule inhibitors is currently considered as an attractive therapeutic strategy against specific tumor types such as leukemia and non-small cell lung cancer [19][20][21] .
From the previous studies, Plk1 has been frequently considered as a classical oncogene. However, the cellular effects of Plk1 overexpression in malignant transformation and their implications in tumor development have not been analyzed. In this study, we found that Plk1 overexpression functions as a tumor suppressor both in vitro and in vivo. Elevated levels of Plk1 delay mammary gland tumor formation driven by classical oncogenes such as Kras G12D or Her2. At the cellular level, these effects are accompanied by multiple aberrations during mitosis, as well as impaired loading of ESCRT complexes during cytokinesis leading to polyploidy. Importantly, increased levels of Plk1 in breast cancer patients is associated with better prognosis.

A new mouse model for inducible Plk1 overexpression.
To investigate the consequences of Plk1 overexpression we first generated KH2 mouse embryonic stem (ES) cells 22 in which a FLAG-tagged human Plk1 cDNA was introduced downstream of the ColA1 gene (Fig. 1a). In this construct, the FLAG-Plk1 cDNA is expressed under the tetracycline-inducible operator (tetO) sequences and it is therefore induced after the activation of the reverse tetracycline transactivator (rtTA; expressed in the Rosa26 locus) with the tetracycline derivative doxycycline (Dox) (Fig. 1a). Treatment of these ES cells with Dox resulted in rapid induction of FLAG-Plk1 (Fig. 1b), which was detected in the spindle poles and the spindle during mitosis suggesting a proper localization of the encoded FLAG-Plk1 protein (Fig. 1c). We then generated heterozygous [referred to as ColA1(+/Plk1) or (+/Plk1) in brief] or homozygous (Plk1/Plk1) knockin mice after microinjection of these ES cells into developing morulas. These knockin mice also expressed the Rosa26-rtTA (referred to as rtTA) transactivator either in heterozygocity (+/rtTA) or homozygocity (rtTA/rtTA) (Fig. 1a), resulting in an efficient induction of FLAG-Plk1 expression in different tissues after administration of Dox ( Fig. 1d and Supplementary Fig. 1a).
We first induced FLAG-Plk1 expression in vivo using heterozygous or homozygous knockin mice in the presence of two copies of the transactivator [(+/Plk1); (rtTA/rtTA) or (Plk1/Plk1); (rtTA/rtTA), respectively]. Unexpectedly, most of these animals died ( Supplementary Fig. 1b) as a consequence of a rapid loss of weight ( Supplementary Fig. 1c) during the first weeks on the Dox diet and with reduced counts of red blood cells (RBC), white blood cells (WBC) and lymphocytes (LYM) ( Supplementary Fig. 1d). Double heterozygous mutants [(+/Plk1);(+/rtTA)] treated with Dox since birth displayed a slightly reduced tumor-free survival (Fig. 1e), accompanied by a slight but non-significant increase in some tumors such as lymphomas and sarcomas ( Supplementary Fig. 1e,f). Overexpression of Plk1 in Dox-treated (+/Plk1);(+/rtTA) mice was accompanied by alteration in the nuclear size in some tissues such as bronchial epithelia, pancreas or liver (Fig. 1f).

Plk1 overexpression impairs cell proliferation and malignant transformation in
vitro. To analyze the cellular consequences of Plk1 overexpression during cell division we next used (+/Plk1);(+/rtTA) or (Plk1/Plk1);(+/rtTA) mouse embryonic fibroblasts (MEFs) generated from these mutant mice. The transgenic Plk1 was properly located in We next asked whether Plk1 overexpression could cooperate with oncogenic transformation by HrasV12 in immortal MEFs. As depicted in Fig. 2g, Plk1 overexpression dramatically reduced the number of Ras-transformed foci, and these transformed cells were unable to grow in soft agar (Fig. 2h). Additionally, HrasV12 and E1A transformed primary MEFs also grow slower after Dox administration compared to uninduced clones ( Supplementary Fig. 2f,g) and Plk1 expression continued to generate binucleated cells ( Supplementary Fig. 2h), despite being transformed. Finally, to provide further evidence that Plk1 overexpression facilitates polyploidization in nontransformed human cells, MCF10A-rtTA cells were infected with a Dox inducible Plk1 vector. As shown in Supplementary Fig. 2i, Plk1 overexpression resulted in a significant increase in binucleated cells after 48h on Dox. All together, these results suggest an antiproliferative effect of Plk1 overexpression in these assays.
We also checked whether the negative effects of Plk1 overexpression were uniquely present in p53-proficient cells. We generated MEFs derived from (+/Plk1);(+/rtTA);p53(-/-) and (Plk1/Plk1);(rtTA/rtTA);p53(-/-) mice and tested the effect of Dox-mediated induction of Plk1. Plk1 overexpression in p53-null cells also resulted in deficient cell proliferation ( Supplementary Fig. 3a) in the presence of highly polyploid cells (Supplementary Fig. 3b). In addition, lack of p53 did not rescue the defects in cell transformation induced by Plk1 overexpression in the presence of the HrasV12 oncogene ( Supplementary Fig. 3c), suggesting that the anti-proliferative defects induced by Plk1 overexpression do not require an active p53-mediated response.

Plk1 overexpression disrupts proper chromosome segregation and cytokinesis.
Since Plk1 has been involved both in DNA replication as well as in mitosis 23, 24 , we first tested whether Plk1 induction had any obvious effect in DNA replication. Treatment with Dox did not alter the number of cells entering S-phase at early time points in the first cell cycle after stimulation of cell cycle entry with serum ( Supplementary Fig. 4a).
In addition, we did not observe a significant increase of DNA damage foci (as detected with antibodies against phosphorylated (g)-H2AX or 53BP1) after Plk1 overexpression, suggesting no major defects in DNA replication ( Supplementary Fig. 4b,c).
We next followed progression throughout mitosis using time-lapse microscopy in cells co-expressing GFP-tagged histone H2B. Overexpression of Plk1 lead to a variety of mitotic defects such as monopolar and multipolar spindles in prometaphase, as well as lagging chromosomes and anaphase bridges (Fig. 3a-e) resulting in increased duration of mitosis (Fig. 3d). In line with these defects, a significant percentage of cells (26.7% versus 1.1% in control cells; Fig. 3a,b) exited mitosis in the absence of chromosome segregation. In addition, 38% of Plk1-overexpressing cells displayed abnormal cytokinesis resulting in binucleated cells or underwent mitotic regression (7.0% of Dox-treated vs. 3.4% untreated cells) thereby generating tetraploid cells with a single nucleus (Fig. 3a,b). Immunofluorescent analysis of these cultures revealed increased mitotic aberrations such as lagging chromosomes and cytokinesis bridges ( Fig. 3e). Finally, treatment of these cultures with Dox resulted in the formation of a significant number of binucleated cells (Fig. 3f), in agreement with a failure in cytokinesis.
To determine whether elevated Plk1 levels resulted in increased Plk1 activity at different cellular localizations, we stained Plk1-overexpressing cells with an antibody that recognizes phosphorylation on Thr210 at the Plk1 T-activation loop. Quantification of this signal revealed a significant increase in Plk1-pT210 levels in interphase cells with centrosomes separated into the two poles (characteristic of late G2 cells; Fig. 4a).
Furthermore, this signal was also significantly increased in prometaphase and metaphase cells (Fig. 4b), as well as in single prometaphase kinetochores during prometaphase in Plk1-overexpressing cells (Fig. 4c). The centromeric levels of Sgo1, a protector of centromeric cohesion, were reduced in Plk1-overexpressing cells (Fig. 4d) in agreement with previous data suggesting that Plk1 activity leads to dissociation of Sgo1 from centromeres 25 . In line with these observations, a significant number of Plk1overexpressing cells displayed reduced cohesion as observed in metaphase spreads after treatment of cells with the microtubule poison colcemid (Fig. 4e).  [26][27][28] . During mitosis, Plk1 phosphorylates Cep55 thereby preventing its premature association with the midzone until cytokinesis entry 26,27 . Plk1 inactivation during anaphase allows Cep55 dephosphorylation and translocation to the midbody thereby resulting in the localization of ESCRT proteins to the bridge. We therefore asked whether Plk1 overexpression could lead to defective abscission as a consequence of deregulation of this pathway.
A detailed analysis of cytokinesis defects in Plk1-overexpressing cells showed a significant number of cytokinesis aberrations (Fig. 5a) as well as increased length of the cytokinesis bridge in agreement with delayed abscission (Fig. 5b). During cytokinesis, the midbody is formed by compacted bundles of microtubules and proteins required for abscission such as RacGAP1 (also known as Cyk4) and MKLP1 (also known as Kif23 or centraspindlin). RacGAP1, a Plk1 substrate involved in RhoA activation and local actomyosin contraction 29,30 , was properly activated at the midbody in Plk1 overexpressing cells (Supplementary Fig. 5a). Similarly, loading of the midbody core kinesin MKLP1, a plus-end directed motor protein involved in the formation of the cleavage furrow in late anaphase and in cytokinesis 31 , was not affected (Supplementary Abscission is ultimately mediated by binding of the adaptor protein Cep55 to MKLP1 in the preassembled midbody. During mitosis, phosphorylation of Cep55 by Plk1 prevents its premature loading to this structure, whereas Plk1 degradation during anaphase results in Cep55 dephosphorylation and loading to the midbody 27 . Cep55 then promotes abscission by recruiting ESCRT-I membrane-remodeling proteins 28,32,33 . In agreement with a role for Plk1 in preventing Cep55 loading, Plk1 overexpression resulted in a significant reduction in the localization of Cep55 to the midbody (Fig. 5c).
Misslocalization of Cep55 was accompanied by reduced loading of the ESCRT-I component TSG101 (Fig. 5d). Importantly, direct inhibition of Plk1 kinase activity by 1-h treatment with the Plk1 inhibitor BI2536 significantly rescued the midbody levels of  (Fig. 6b). In addition, the number of tumors per animal was significantly reduced after Plk1 overexpression in these models (Fig. 6c). In line with the polyploid phenotype observed in cultured MEFs, overexpression of Plk1 resulted in increased nuclear volume in tumor cells ( Supplementary Fig. 6a,b). In addition, these Plk1-overexpressing tumor cells displayed a higher frequency of aneuploidy and polyploidy compared to the single oncogene tumors (Fig. 6d,e,f). To analyze if high Plk1 levels influence chromosome instability during tumor growth, we performed time lapse microscopy of tumor cells. Indeed, 50% of Her2/Plk1 tumor cells displayed mitotic errors compared to 24% of Her2 alone. In addition, 25% of Her2/Plk1 tumor cells became polyploid while this only occurred in 3% of Her2 tumor cells (Fig. 6g,h).
A common consequence of polyploidy is the activation of p21 through a p53 dependent mechanism resulting in cell cycle arrest. In agreement with our previous data in Plk1 overexpressing MEFs, we also observed a significant increase of p21 positive cells in Plk1 expressing tumors, compared to oncogene alone ( Supplementary Fig. 6c, d).
Moreover, the percentage of cells proliferating in these tumors was significantly lower than in Kras or Her2 tumors alone ( Supplementary Fig. 6c To monitor single cell fate of epithelial cells in vivo after Plk1 overexpression, we used a three-dimensional culture system of primary mammary epithelial cells 5,36 isolated from these mouse models. Single cells were embedded in matrigel and allowed to develop acinar structures for 6-8 days. Once spheres were formed, we induced transgene expression with Dox and followed cell division of these cells co-expressing GFP-tagged histone H2B using time-lapse microscopy. In line with previous results 5 , control and Kras G12D induced cultures had no obvious phenotype after 36 hours on Dox ( Fig. 7a) and the mitosis observed were normal. However, similar to the observations in cultured MEFs, overexpression of Plk1 in mammary epithelial cells resulted in a significant prolonged mitosis (Fig. 7a,b) in the presence of abnormal mitotic figures including lagging chromosomes, defective chromosome segregation and cytokinesis failure (Fig. 7c,d). All together, these data suggest that Plk1 overexpression results in defective mitosis and cytokinesis in mammary gland epithelial cells thus suppressing tumor development.
Plk1 expression correlate with genome-doubled breast cancers. Next, to explore the relationship between PLK1 expression and ploidy in human breast cancers, we obtained matched copy number and expression data for 953 breast tumors from the TCGA.
Consistent with an association between PLK1 expression and polyploidy, we observed a highly significant difference in levels of PLK1 expression in tumors which exhibited evidence of having undergone a genome doubling event during their evolution compared to their non-doubled counterparts (Fig. 8a, P=4.62e-09, t-test). Furthermore, the association between genome doubling and PLK1 expression levels remained significant in TP53 wildtype tumors, suggesting that PLK1 may facilitate polyploidization in a TP53 independent manner (Fig. 8b, P=3.37e-06, t-test).
Finally, to investigate the clinical significance of PLK1 expression in breast cancer, we grouped tumors into those with low PLK1 expression (bottom quartile PLK1 expression) and those with higher PLK1 expression (remaining quartiles). In keeping with the observed tumor suppressor properties of PLK1, patients with low PLK1 expression were associated with significantly shorter overall survival, compared to those with higher PLK1 expression levels (

Discussion
Plk1 belongs to a family of kinases with multiple roles in proliferative as well as postmitotic cells [10][11][12][13]23 . Among Polo-like kinases, Plk1 is considered an interesting target for cancer therapy due to the requirements for its kinase activity during cell division in tumor cells 13,17,21 . However, to what extent Plk1 expression is a cause or a consequence of carcinogenesis is under debate 37 . Plk1 expression is cell cycledependent and Plk1 is frequently overexpressed together with other mitotic genes in highly proliferating and chromosomally unstable tumors 38 . Plk1 is repressed by the tumor suppressor p53 (Ref. 39 ) and increased Plk1 levels can simply reflect p53 inactivation in cancer cells 40 .
Although Plk1 is frequently classified as an oncogenic protein, its contribution to malignant transformation is arguable. In pioneer assays in the 90's, ectopic expression of Plk was shown to increase DNA synthesis and mitosis in quiescent NIH 3T3 cells and to induce oncogenic foci 41 . Although the molecular basis for these observations is unknown, Plk1 can inhibit p53 (Ref. 42 ) and leads to the stabilization of Myc proteins 43 , suggesting possible mechanisms for proliferative functions. However, its sufficiency in triggering cell cycle entry and progression is unclear and more recent data suggest that Plk1 overexpression leads to cell proliferation defects at least partially due to aberrant mitosis and the activation of the spindle assembly checkpoint 44 .
Interestingly, there are early reports showing that other Polo like family members might also play as tumor suppressors such as Plk3 or Plk5 (Refs. [45][46][47]. Recent data also suggests that constitutive overexpression of Plk1 in mice does not lead to tumor formation 48 , although the mechanism is not investigated. In our hands, overexpression of Plk1 using an inducible knockin model in murine cells results in decreased cell proliferation accompanied by multiple mitotic aberrations, including defects in chromosome congression and segregation, and abnormal cytokinesis (Fig. 2-5). Plk1 overexpression also prevents malignant transformation of primary cells by Ras oncogenes (Fig. 2) and impairs breast cancer development induced by Kras or Her2 oncogenes (Fig. 6).
The pleiotropic defects caused by Plk1 overexpression are likely a consequence of the multiple roles of this protein in several cellular structures such as the centrosomes, kinetochores or the cytokinesis bridge 12,13,23 . Among these defects, failure in cytokinesis and abscission is the most abundant defect in Plk1-overexpressing cells, resulting in the formation of binucleated cells as well as tetraploid mononucleated cells generated after mitotic regression in the absence of abscission (Fig. 3). Lack of abscission in these cells correlates with defective loading of Cep55, a Plk1 substrate that recruits the ESCRT component TSG101 to the cytokinesis bridge 27 . In the presence of high Plk1 activity, loading of the ESCRT complex into the cytokinesis bridge is deficient (Fig. 5), likely as a consequence of impaired loading of Cep55 during late stages of anaphase. In fact, treatment of Plk1-overexpressing cells with the Plk1 inhibitor BI2536 partially rescued these defects, suggesting that cytokinesis defects in Plk1-overexpressing cells are kinase-dependent and at least partially mediated by the Cep55-ESCRT pathway. Since Plk1 is also involved in cell migration and metastasis 49 , authophagy 50 , pentose phosphate metabolism 51 or blood pressure regulation 11 Fig. S8). Our analysis of human breast cancer datasets shows that tumors with higher Plk1 expression generally have improved prognosis (Fig. 8). All together, these data indicate that, despite being generally considered as an oncogene, Plk1 may have tumor suppressive activities.
The fact that Plk1 may function as a tumor suppressor instead of an oncogene does not necessarily argue against the use of Plk1 inhibitors in cancer therapy. Many essential components of cell proliferation may be used as cancer targets despite having no oncogenic activity, owed to non-oncogene addiction of cancer cells for specific cellular processes such as cell division 62 . Plk1 is a frequent hit in chemical or genomewide genetic screens to uncover new targets under different oncogenic backgrounds [63][64][65][66][67][68][69][70][71][72] .
In fact, Plk1 inhibition may be particularly effective in Ras-(Refs. 63,73 ) or Her2-(Ref. 74 ) induced tumors indicating that, independently of a possible tumor suppressor role of overexpressed Plk1, efficient kinase inhibition of this protein may impair tumor cell proliferation and survival. Plk1 inhibitors are currently being tested in multiple solid and hematopoietic tumors and BI6727 (volasertib), a derivative of BI2536, has recently received the FDA Breakthrough Therapy designation for its effect in acute myeloid leukemia 21,75 . Understanding the specific requirements for Plk1 as compared to other essential cell cycle kinases will likely contribute to a better design of therapeutic strategies against its kinase activity.