MK3 Modulation Affects BMI1-Dependent and Independent Cell Cycle Check-Points

Although the MK3 gene was originally found deleted in some cancers, it is highly expressed in others. The relevance of MK3 for oncogenesis is currently not clear. We recently reported that MK3 controls ERK activity via a negative feedback mechanism. This prompted us to investigate a potential role for MK3 in cell proliferation. We here show that overexpression of MK3 induces a proliferative arrest in normal diploid human fibroblasts, characterized by enhanced expression of replication stress- and senescence-associated markers. Surprisingly, MK3 depletion evokes similar senescence characteristics in the fibroblast model. We previously identified MK3 as a binding partner of Polycomb Repressive Complex 1 (PRC1) proteins. In the current study we show that MK3 overexpression results in reduced cellular EZH2 levels and concomitant loss of epigenetic H3K27me3-marking and PRC1/chromatin-occupation at the CDKN2A/INK4A locus. In agreement with this, the PRC1 oncoprotein BMI1, but not the PCR2 protein EZH2, bypasses MK3-induced senescence in fibroblasts and suppresses P16INK4A expression. In contrast, BMI1 does not rescue the MK3 loss-of-function phenotype, suggesting the involvement of multiple different checkpoints in gain and loss of MK3 function. Notably, MK3 ablation enhances proliferation in two different cancer cells. Finally, the fibroblast model was used to evaluate the effect of potential tumorigenic MK3 driver-mutations on cell proliferation and M/SAPK signaling imbalance. Taken together, our findings support a role for MK3 in control of proliferation and replicative life-span, in part through concerted action with BMI1, and suggest that the effect of MK3 modulation or mutation on M/SAPK signaling and, ultimately, proliferation, is cell context-dependent.


Introduction
Sequential activation of kinases within the canonical M/SAPK (mitogen/stress activated protein kinase) cascades is a common and evolutionary-conserved signal transduction mechanism. The canonical M/SAPK cascades cooperate in transmitting and integrating intra-and extracellular signals, thereby controlling a large number of, sometimes opposing, cellular processes such as proliferation, differentiation, survival, stress response, and apoptosis [1]. Downstream of M/SAPKs, MAPK-activated protein kinases (MAPKAPKs), including RSK1-4, MSK1/2, MNK1/2 and MK2/3/5, signal to diverse cellular targets. Among the MAPKAPKs are three structurally related MKs MK2, MK3 and MK5. Despite their high homology, the three MKs display distinct spatio-temporal expression profiles and act in different biological processes [2,3]. Identification of MK-substrates suggests that MKs function in numerous cellular processes, including gene transcription, mRNA-stability and translation, cytoskeleton remodeling, cell proliferation and apoptosis. Besides their joint involvement in inflammatory responses, the biological relevance of substrate interaction and phosphorylation by MKs remains largely unclear.
MK3 (MAPKAPK3, 3pK) was identified as the first MK activated down-stream of all three mitogen-and stress-activated protein kinase (M/SAPK) cascades; consequentially, MK3 was considered an integration point of converging mitogenic and stress signaling [4]. Whereas the RAS-M/SAPK signalling pathways have a long-standing link to cancer, the involvement of MKs in cancer is currently unclear. MK3, originally referred to as 3pK (chromosome 3p kinase), was found frequently homozygously deleted as part of the 3p21.3 region in small cell lung cancer and other cancers and cancer cell lines [5,6]. Conversely, potential oncogenic 'driver' mutations have been identified in MK3 [7]. These records point to an involvement of MK3 in tumorigenesis and support the idea that it may act tumorigenic or tumor-suppressive.
We previously reported that MK3 associates with PRC1-complexes through direct SAM (Self-Association Motif) domain-mediated interaction with the Polyhomeotic orthologs PHC1 and PHC2 [8]. Polycomb Group repressive complexes (PRC1 and PRC2) act as part of a cellular epigenetic memory system and play an important role in the determination of cell fate [9]. Both core complexes harbor intrinsic PRC protein-associated epigenetic catalytic activity, and are known to interact with additional epigenetic regulators. These interactions and catalytic activities are controlled by post-translational modification [10,11]. In addition, we established that phosphorylation of the PRC1 complex controls PRC1/chromatin-association [8,12]. PRC proteins have been linked to oncogenesis: high expression or mutation of several PRC members has been etiologically implicated in the onset and malignant progression of cancer [13].
Recent data from our group identified MK3 as a regulator of the PRC1 target gene ATF3-expression via a negative feedback mechanisms on MEK1 and ERK1/2 in the context of mitognic stimulation. We found that MK3 ablation or inhibition resulted in prolonged ERK phosphorylation and increased ATF3 and premature and elevated EGR1 expression [14]. Using a genetic Drosophila model for wing development, we confirmed exagerated mitogenic ERK-signaling in the absence of dMK (the only Drosophila MK ortholog), thus supporting a negative regulatory role for MK3 in canonical ERK signaling [14]. In addition, besides ERK, modulation of MK3 levels also affects cellular P38 and JNK protein levels [14,15]. Combined, these observations suggest that altered cellular MK3 levels and or activity may contribute to deregulation of cellular proliferation through deregulation of M/SAPK, potentially contributing to tumorigenesis.
In this study we tested the hypothesis that MK3 controls cell proliferation using MK3-gainand loss-of-function constructs in normal and cancer cells. In addition, we aimed to investigate the biological relevance of the interaction between MK3 and PRC1 in the context of cell proliferation. We provide evidence for a genetic interaction between the Polycomb Group protein BMI1 and MK3 in proliferative life span: BMI1 overcame the MK3-overexpression phenotype in normal cells. In contrast, BMI1 did not compensate for the negative effect of MK3 depletion on cell proliferation. MK3 overexpression inhibited the proliferation of normal and cancer cells. In contrast, MK3 depletion enhanced the cell cycling of cancer cell lines. Our data supports the idea that MK3-mediated abnormal M/SAPK signaling intersects with known checkpoints and contributes proliferative control. The potential implications of these observations for tumorigenesis are discussed.

Gain or loss of MK3 function induces senescence in normal human fibroblasts
To study the role of MK3 in cellular processes related to cell proliferation, we modulated cellular MK3 levels in different cell models using retroviral expression systems and measured the effects thereof. We initially focused on normal diploid human TIG3 fibroblasts to determine the effect of forced overexpression of wild type (non-mutated) MK3 (MK3 WT OE) on cell proliferation ( Fig 1A). Transduced TIG3 cells were expanded under selection pressure; proliferative capacity of these cultures was assessed at ±1 week and ±4 weeks after transduction. TIG3/MK3 WT OE displayed clearly decreased proliferative capacity relative to empty vector control cells at the early time point and had altogether stopped dividing at 4 weeks post-transduction ( Fig 1B). By means of reference, TIG3/Bmi1OE maintained proliferative capacity throughout the duration of the experiment, whereas primary empty vector-transduced TIG3 (con) progressively lost proliferative capacity due to limited proliferative lifespan in vitro (Fig 1B). DNA-profiling at late time points posttransduction revealed an increased number of cells in G0/G1 at the expense of cells in S-phase upon MK3-overexpression ( Fig 1C); overexpression of the RASV12 oncogene reduced S-phase cell numbers as anticipated as RASV12 is known to evoke oncogene-induced senescence (OIS) [16], Fig 1C). Reduced de novo DNA synthesis in TIG3/MK3 WT OE cells was confirmed by a markedly lower number of BrdU-positive cells in TIG3/MK3 WT OE cultures and (Figs 1D and S1A); equal relative distribution of cells throughout S-phase suggested that TIG3/MK3 WT OE cultures had undergone an intraS-phase arrest, a feature known to occur in the context of oncogene-induced senescence (OIS) (S1B Fig).
In further support of activation of a senescence-associated response by MK3, TIG3/MK3 WT OE cells displayed enlarged, flat-cell morphology (Fig 2A). The occurrence of senescence was further corroborated by expression of the senescence-associated beta-Galactosidase (SA-bGal) marker protein in large flat cells (Fig 2A). MK3 WT -overexpression also negatively affected cell division in immortal TIG3 hTERT cells, indicating that MK3 acts downstream or independent of hTERT in proliferative control (S1C Fig). TP53 is a senescence marker, activated downstream of e.g. increased mitogenic signaling through RAS/ERK [17][18][19][20]. Consistent with the observed MK3mediated senescence response, global expression of TP53 was elevated in TIG3/MK3 WT OE cells (Fig 2B). Increased staining of TP53 was readily detectable in senescent TIG3/MK3 WT OE nuclei ( Fig 2C); likewise expression of P21 CIP1/WAF1 , a direct transcriptional target of TP53, was increased ( Fig 2B). The increased expression of P16 INK4a in relation to sustained MK3 WT OE was consistent with its crucial role in establishing irreversible senescence ( Fig 2B) [17,21,22]. We next tested whether the adverse effect of MK3 WT OE was dependent on its kinase activity. Surprisingly, the early effect of MK3 WT OE on cell proliferation occurred independent of MK3 kinase activity, as expression of a kinase-dead mutant (MK3 KM OE) and a constitutively active mutant (MK3 CA OE) did not affect cell proliferation at 1 week post-transduction (S2A Fig). In agreement with this, cellular TP53 levels were only moderately increased in the TIG3/MK3OE cells, not in TIG3/MK3 KM OE or TIG3/MK3 CA OE cells. P16 INK4A levels were substantially increased at 1 week post-transduction in both TIG3/MK3 WT OE and TIG3/MK3 CA OE cultures, despite the differences in proliferation rate (S2A and S2B Fig). P16 INK4A was also induced by MK3 KM OE, however, at a lower level than in both other conditions. Despite the initial absence of adverse effects on proliferation, all three MK3 variants reduced proliferation rate at 4 weeks post-transduction (S2A and S2B Fig). The absence of correlation between TP53 and P16 INK4A levels and proliferation rate at 4 weeks suggests a potential contribution of additional molecular targets and mechanisms to the observed effects on proliferation. Combined this data shows that overexpression of MK3 WT or two MK3 kinase mutants in human diploid fibroblasts reduces their proliferative capacity.
As the MK3 locus is frequently missing in cancers as part of often larger genomic deletions [5,6], it is conceivable that cellular MK3 depletion affects cell cycle regulation and contributes to tumorigenesis. To study the effect of loss of MK3 on cell proliferation, we next reduced cellular MK3 levels using RNA-interference. To this end a retroviral shRNA construct was designed that selectively targeted MK3 (shMK3) (S2C Fig; cf. Fig 1A). Similar to MK3 overexpression in TIG3 cells, MK3 depletion resulted in reduced proliferation of TIG3/shMK3 cells and increased flat cell formation (Fig 2D), which was confirmed by reduced BrdU-incorporation in TIG3/ shMK3 cultures (S2D Fig). The shMK3-associated arrest in normal cells was accompanied by increased TP53, P21 CIP1/WAF1 and P16 INK4A expression ( Fig 2E). As MK2 is also expressed in TIG3 cells (cf. S5B Fig) we aimed to exclude the possibility that MK2 confounded the observed MK3 depletion-associated proliferative phenotype. To this end TIG3 cells were either transduced with a shMK3 or a shMK2/3 construct; the latter effectively reduced expression of both MK2 and MK3 (S2C Fig). Both constructs increased G1-arrest with equal efficiency; MK2 neither compensated for loss of MK3 nor added to the effect of MK3 depletion, suggesting that the observed proliferative reduction could be attributed to MK3 in this experimental setting (S2E Fig). Thus, overexpression as well as depletion of MK3 in TIG3 cells induced a proliferative block, which correlated with expression of known senescence markers. Taken together, these results suggested that MK3 controls proliferative lifespan in normal diploid human fibroblasts.

Genetic interaction between MK3 and Polycomb in proliferative lifespan
MK3 has been identified as a PRC1 binding partner; we have shown that signaling through ERK1/2, P38 and MK3 controls gene expression via modulating chromatin-association of the PRC1 complex [8,14]. We further explored the functional interaction between PRC-function and MK3 in the context of cell proliferation. Senescent cells are known to release CDKN2A/INK4A repression and show increased P16 INK4A and/or P14 ARF levels [22][23][24]. Increased expression at the CDKN2A locus correlates with reduced local H3K27me3 [25,26]. Expression of the responsible histone H3K27 methyltransferase (HMT), the PRC2-class protein EZH2, was reported to be progressively reduced by TP53 in senescing cells [27,28]. Transcriptional repression of the CDKN2A-locus in proliferating cells correlates with PRC2-mediated histone H3 Lysine 27 trimethylation (H3K27me3) [25,26]. In good agreement with these reports, we observed that global TP53 and EZH2 protein levels inversely correlated in senescent TIG3/MK3 WT OE cultures and  Cell Context-Dependent MK3 Involvement in Cancer that reduced cellular EZH2 correlated with increased P16 INK4A expression (Fig 3A; cf. Fig 2B). Reduced EZH2 levels also inversely correlated with P16 INK4A expression in senescent TIG/shMK3 cultures (Fig 3B), in further support of a functional link between EZH2 reduction and senescence upon modulation of MK3 levels. In addition to EZH2, CBX8 levels were reduced in both senescent TIG3/MK3 WT OE and TIG3/shMK3 cultures, whereas PHC1 levels appeared relatively unaffected by gain or loss of MK3 function (Fig 3A and 3B). Thus far, the combined data shows that MK3 modulation induces a state of senescence that correlates with reduced cellular EZH2 levels and elevated P16 INK4A expression.
To assess changes in PRC1 protein/chromatin-association and Polycomb-related status at the CDNK2A/INK4A locus in relation to MK3 overexpression, we applied chromatin immunoprecipitation (ChIP) technology to pre-senescent TIG3/MK3 WT OE cell cultures. The CDKN2A/ INK4A locus has been mapped before in detail in regards to H3K27me3-marking and PRC1-occupation in proliferating and senescent cells [25,26]. A number of previously identified PRC1negative (p15exon1; p14exon1) and positive (HOXA10; HOXA11) regions were used as controls in our analyses [25]. MK3-overexpression moderately, but consistently enhanced chromatin-occupation of MK3 at all loci probed including non-PRC1 target loci (Fig 3C), in line with our earlier observation that part of cellular MK3 associates with chromatin [8]. Absence or presence of H3K27me3-marking in our experimental model correlated perfectly with absence or presence of CBX8 occupation, respectively ( Fig 3C); CBX8 binds the K27 trimethyl-mark through its chromobox domain [25,29]. Interestingly, chromatin-occupation of PHC1, a direct binding partner of MK3 [8], increased in parallel with MK3 at all promoter sequences analyzed, including at non-PRC1 target loci ( Fig 3C). In contrast to CBX8, PHC1 was found associated also with non-PRC1 target loci, suggesting that (part of) PHC1 may associate with chromatin in an H3K27me3-indendent fashion ( Fig 3C). Relevantly, MK3-associated induction of senescence correlated with loss of H3K27me3-enrichment at p16-exon1 and CBX8-displacement from the P16 INK4A locus (promoter and exon1; Fig 3C). This data is in good agreement with both the reduced cellular EZH2 and CBX8 levels, and with the increased P16 INK4A expression (S3B Fig) [25,26]. Combined, this data shows that in the MK3 WT -overexpressed state, cellular senescence correlates with reduced cellular EZH2 and CBX8 levels and consequential de-repression of transcription at the CDKN2A locus.
We next explored the functional relationship between MK3 and BMI1 in proliferative lifespan. Genetic ablation of PRC1-members, among which BMI1 and PHC paralogs, is known to induce premature cellular senescence in vitro and in vivo [30,31]. In the light of the herein described local (i.e. at the CDKN2A locus) chromatin-displacement of PRC1 in senescing cells and our previous observation that prolonged MK3 overexpression leads to reduction of chromatinassociated BMI1 [8], we speculated that Bmi1OE would overcome the senescent state induced by gain or loss of MK3 function. To determine whether Bmi1OE could bypass the negative effect of MK3OE on proliferative capacity, TIG3 fibroblasts were sequentially transduced with murine Bmi1 and MK3-encoding retroviral expression vectors. Bmi1OE consistently produced a proliferative advantage for TIG3 fibroblasts (TIG3/Bmi1OE/MK3 WT OE) (Fig 4A and 4B). Whereas MK3 WT overexpression alone reproducibly showed reduced TIG3 proliferation, co-expression of Bmi1 compensated for the adverse effects of MK3 WT overexpression on cell proliferation ( Fig 4A). Importantly, in contrast to MK3 WT expression, the adverse effect of MK3 ablation on fibroblast proliferation was not rescued by BMI1 (Fig 4B). Bmi1 co-expression also reversed the MK3-induced senescent cell morphology in observed in TIG3/MK3 WT OE cultures, (Fig 4C). Coexpression of Bmi1 in TIG3/MK3 WT OE cells prevented loss of EZH2 as well as of the PRC1 complex proteins CBX4 and RNF2 ( Fig 4D). Finally, the global increase of H3K27me3 observed in TIG3/MK3 WT OE cells was prevented by co-expression of Bmi1 ( Fig 4D). TP53 expression in TIG3/MK3 WT OE cells was countered by BMI1 co-expression ( Fig 4E). Relevantly, sequentially transduced fibroblast cultures co-expressing Bmi1/MK3 WT self-selected for increased BMI1 protein levels over time compared to cultures overexpressing Bmi1 alone (Fig 4E), further supporting of a genetic interaction between MK3 and BMI1 in the context of cell proliferation. As EZH2 targets the CDKN2A/INK4A locus for transcriptional repression, we also examined whether EZH2 alone could rescue the MK3-induced proliferative arrest. Whereas Bmi1 co-expression in TIG3 cells clearly reduced MK3 WT OE-induced P16 INK4 expression in the same experiment, EZH2OE did not prevent MK3 WT -induced upregulation of cellular P16 INK4 ; instead, elevated basal P16 INK4 expression both at the mRNA and protein levels was already detectable in control cells in TIG3/EZH2OE cells (S4A and S4B Fig). Bmi1 co-expression clearly neutralized the adverse effects of MK3 WT OE on cell proliferation, the effect of EZH2OE on the MK3 WT OE-proliferation phenotype were inconsistent, yet never compensatory (data not shown). Taken together, the above findings support a functional genetic interaction between MK3 and BMI1/PRC1 in proliferative control of normal human fibroblasts, and suggest that EZH2/PRC2 is likely neither sufficient to bypass MK3-arrested proliferation nor to prolong proliferative lifespan.

Loss of MK3 provides a growth advantage in cancer cell models
We used the large-scale cancer genome-sequencing and expression analyses initiative Cancer Cell Line Encyclopedia (CCLE; http://www.broadinstitute.org/ccle) to assess MK3 expression in nearly 1000 cancer cell lines [32]. MK3 was relatively high expressed in various cancer cell types, among which bone, pancreatic, colorectal and endometrial cancers (S5A Fig). Conversely, MK3 showed significant focal deletion across a set of more than 3000 tumors, or its expression is low/absent in cancer cell types, including small-cell lung cancers, neuro-and medulloblastomas [32]. Although cancers of (neur)ectodermal origin tend to show low/absent MK3 expression, there was no obvious correlation between germ-layer origin and MK3 expression among the various cancer types (S5A Fig), suggesting that MK3 status is an acquired feature in tumors. When we next examined MK3 protein levels in a panel of cancer cell lines, we were able to confirm large variation in MK3 expression between different cell lines (S5B Fig); no obvious correlation was observed between MK3 and MK2 levels, in line with their distinct cellular functions [33,34]. Combined, this data points to a possible involvement of MK3 in tumorigenesis and suggests that the role of MK3 may be determined by cellular and/or genetic context. Given its possible dual involvement in tumorigenesis, we probed cancer cell lines for effects of gain or loss of MK3 function on cell proliferation. We here focused on U-2OS osteosarcoma cells and the HeLa cervical carcinoma cells. Of note, U-2OS/MK3 WT OE cultures also showed an MK3 WT OE-induced reduction of proliferation and enhanced flat cell morphology (9.6% ±1.8 MK3 WT OE vs.  Increased replication stress is known to induce a proliferative arrest which is often associated with double strand DNA breaks (DSB). The occurrence of replication stress-associated DSB is typically associated with an initial intraS-phase arrest, as part of a DDR [19,35,36]. U-2OS/ MK3 WT OE cells showed a ±1.5 fold increase of cells in S-phase within days after transduction ( Fig 5D). To determine whether the flat cell phenotype in U-2OS/MK3OE cells was associated with DNA damage, we use immunofluorescence to study the occurrence of DSB by measuring phosphorylation of histone variant H2A.X (γH2A.X) and of KAP1 phosphorylation at serine 824 (pKAP1). Double-positive γH2A.X/pKAP1 cells were found in considerable numbers throughout U-2OS/MK3 WT OE cultures; the nuclei of enlarged flat cells were prominently stained ( Fig 5E).
As the parental U-2OS and HeLa cell lines both express moderate levels of MK3 (cf. S5B  Fig), we studied the effect of RNAi-mediated MK3 depletion on cell proliferation. Remarkably, both U-2OS and HeLa cells proliferated faster in the absence of MK3 (Fig 6A). Thus, in contrast to the negative effects of gain and loss of MK3 function on proliferative capacity in in normal human fibroblasts, MK3 WT overexpression and ablation produce opposing effects in cancer cell models.

MK3 modulation or mutation induce signaling imbalance
The finding that proliferation capacity of U-2OS and HeLa cells was decreased in the context of MK3OE was unexpected. U-2OS osteosarcoma and HeLa cervical carcinoma cells both retain a "wild type" genetic status for TP53 and pRB and both carry a genetically intact CDKN2A/INK4A locus [37,38]. Both TP53 and pRB-dependent checkpoints are, however, compromised in these cells (see Discussion section). In U-2OS cells, the P16/INK4A promoter is transcriptionally silenced due to CpG methylation [38,39]. HeLa cells have an intrinsically high P16 INK4A level; the pRB/P16 INK4A checkpoint is dysfunctional in these cells, however, as pRB is continuously targeted for proteolytic degradation by HPV-expressed oncoprotein. Combined with our earlier observation that changes in proliferation rate of TIG3/MK3 kinase mutant cultures did not correspond well with TP53 and/or P16 INK4A levels, this prompted us to search for involvement of additional molecular mechanisms that may contribute to reduced proliferation in the context of gain or loss of MK3 function. MK3 is a downstream target of the M/SAPKs P38, JNK as well as ERK [4]; hence, MK3 acts as a convergence point of mitogenic ires.GFP (Bmi1OE) or GFP (con) virus, and MK3/puromycin (MK3OE) or control puromycin virus (con) at 48 hrs intervals. Retroviral vectors expressing murine Bmi1/GFP reporter were transduced first (or empty vector control), followed by a MK3/puromycin resistance marker (or empty vector control). Transduction of Bmi1OE and control transduced cells was simultaneously carried out with the same MK3 WT OE viral preparation (or control virus) to minimize inter-experimental variation. Cells were grown on selection medium and proliferation capacity was tested ± 2-3 weeks post-transduction. (A) Proliferation curves of normal human TIG3 fibroblasts transduced with a retroviral MK3 WT overexpression vector (MK3 WT OE; black symbols) or shcon vector (white symbols), in conjunction with either an empty vector control (con; circles) or a murine Bmi1 expression vector (Bmi1OE; triangles). (B) Proliferation curves of normal human TIG3 fibroblasts transduced with a retroviral MK3 knock-down vector (shMK3; black symbols) or shcon vector (white symbols), in conjunction with either an empty vector control (con; circles) or a murine Bmi1 expression vector (Bmi1OE; squares). Cell counts at t = 2 through t = 8 (A, B) were normalized to cell counts at t = 0 for each transduced cell culture individually (see Methods section for details); statistical significance was determined by two-tailed Student's t-test and is presented relative to the empty vector control (* p < 0.05). (C) Comparative morphology of TIG3 cells expression Bmi1 and/or MK3 versus control cells as recorded by GFP fluorescent imaging ± 3 weeks after transduction (D) Immunoblot analysis of EZH2, CBX4, RNF2 and H3K27me3 in MK3 WT OE, Bmi1OE, Bmi1OE/MK3 WT OE and control TIG3 cell lysates. (E) Expression analysis of BMI1, MK3, and TP53 at the indicated time points in (corresponding to experiment in Fig 4C). Cells were grown on selection medium and analysed at 1 or 4 weeks after serial transduction; expression vectors and antibodies are indicated in the figure. Early and late samples were loaded on the same gel for BMI1 analysis; corresponding sections are shown separately. and stress signaling cascades. We previously showed that gain, loss or inhibition of MK3 function affected expression levels of P38 and JNK and altered phosphorylation dynamics of ERK, P38 and JNK in response to mitogenic stimulation in U-2OS cultures [14]. As these findings pointed to an MK3-dependent M/SAPK signaling imbalance, we aimed to determine the effects of MK3 modulation in the context of cell proliferation of normal human fibroblasts and in cancer cell models. Whereas ERK and pERK (phosphorylated ERK) levels were only marginally af- Conversely, yet in keeping with the increased P38 levels in MK3OE cultures, P38 levels were reduced in response to MK3-depletion (shMK3) in both U-2OS and HeLa cancer cell lines (S7D Fig). In light of our previous finding that MK3 serves to provide negative regulatory feedback on canonical MEK/ERK signaling [14], and the current observation that P38 and JNK levels are increased in MK3 WT OE cells, we speculated that a dysbalance in MAPK and SAPK activity could contribute to the observed effects of MK3 modulation on cell proliferation in the various cell models. In keeping with this notion, the proliferation rate of U-2OS/MK3 WT OE cells was negatively affected by reduction of serum concentration in a dose-dependent manner and serum deprivation (1% FCS) significantly enhanced flat cell formation in U-2OS/MK3 WT OE cultures (S7E Fig; cf. Fig 5B). This data supports the idea that modulation of MK3 has direct functional consequences for mitogenic responses and supports a role for signaling imbalance caused by modulation of cellular MK3 levels in normal cells as well as in cancer cells.
A recent systematic cancer genome re-sequencing effort predicted potential oncogenic driver mutations in known and novel genes [7]. Among these potentially tumorigenic MK3 variants were a number of MK3 missense mutants, among which P28S and E105A. As the effect of these mutations on MK3 function is currently unknown, we used the TIG3 model to evaluate their effect on cell proliferation, P38 and ERK levels and P38 and ERK phosphorylation. Comparable to the MK3 KM and MK3 CA mutant constructs, neither MK3 P28S nor MK3 E105A had an attenuating effect on cell proliferation at 1 week post-transduction ( Fig 6B); the proliferation rate of TIG3/MK3 P28S OE was modestly but significantly increased over the (empty vector) control and TIG3/MK3 E105A . In contrast, at 4 weeks post-transduction none of the in vivo identified MK3 mutants mimicked the negative effect of the experimental kinase mutants (MK3 KM , MK3 CA ) and the non-mutated MK3 WT kinase on proliferation (Fig 6B). Although all MK3 variants showed modestly enhanced total ERK1/2 levels, interestingly higher relative proliferation rate correlated with enhanced ERK levels ( Fig 6C)   Cell Context-Dependent MK3 Involvement in Cancer effect was less clear in the TIG3/MK3 KM OE and TIG3/MK3 P28S OE; the total P38 level in the latter cells seemed slightly lower at 4 weeks post-transduction (Fig 6C). In line with the earlier noted lack of correlation between proliferation characteristics and TP53 and/or P16 INK4A levels, TP53 levels were highest at 4 weeks post-transduction in the cells that had the highest relative proliferation rate. Similarly, absolute P16 INK4A levels were highest in TIG3/MK3 E105A OE cultures (Fig 6C). We previously established that gain, loss or inhibition of MK3 function resulted in abnormal mitogenic signalling as a result of altered phosphorylation dynamics of MEK/ERK [14]. We here show that mitogenic stimulation (i.e. serum starvation/serum stimulation), P38 levels and phosphorylation dynamics appeared dependent on the MK3 variant expressed. P38 levels were elevated in all TIG3/MK3 WT OE cultures, irrespective of MK3 type, which is suggestive of a MK3-kinase activity independent effect (S8A and S8B Fig). TIG3/ MK3 WT OE, TIG3/MK3 CA OE and TIG3/MK3 E105A OE showed more intense P38 phosphorylation compared to TIG3/MK KM OE and TIG3/MK3 P28S OE; in addition, pP38 levels were still high in the former three cultures at 2 hours post-stimulation, compared to TIG3/MK KM OE and TIG3/MK3 P28S OE (S8B Fig). Taken together the above data provides the first evidence that MK3-mutation affects cell proliferation. TIG3 fibroblasts thus provide a useful model to evaluate potential tumorigenic effects of novel MK3 mutants on cell proliferation. Our findings suggest that MK3 P28S and MK3 E105A mutants exert specific effects on M/SAPK signalling. In addition, the data support the notion that M/SAPK signalling intersects with TP53 and pRB mediated cell cycle checkpoints.

Discussion
The relevance of signaling through MK3 for cell proliferation was unknown. We here report that sustained gain or loss of MK3 induces a senescent state in normal human fibroblasts. Importantly, the MK3 overexpression-induced senescence is bypassed by co-expression of the oncoprotein BMI1, whereas the replicative senescent-like arrest induced by MK3 ablation is not. Surprisingly, MK3 WT OE overexpression also induces a significant reduction of proliferation in cancer cells, whereas loss of MK3 enhances cancer cell proliferation. In conjunction with our previous finding that modulation of cellular MK3 levels or inhibition of MK3 activity results in altered negative regulatory feedback to MEK/ERK we here show that gain or loss of MK3 function also affects P38. P38 levels appear to be increased by all MK3 variants, including a kinase-defective MK3, suggesting that this effect is independent of its catalytic activity. Interestingly, a number of potential oncogenic MK3 mutations showed distinct effects on ERK and P38 phosphorylation, and in contrast to non-mutant MK3 and MK3 kinase mutants, these mutants did not reduce proliferation capacity of TIG3 cells. Combined our observations support the idea that the effect of MK3 expression level modulation or MK3 mutant is dependent on cellular and genetic context. Known senescence checkpoints in relation to MK3 modulation TP53 and P21 CIP1/WAF1 levels are increased both in MK3 WT OE and shMK3 normal cells, in accordance with their pivotal role early in the senescence response [16,40]. The increased expression of P16 INK4a upon sustained MK3 overexpression is in good agreement with its crucial role in establishing irreversible senescence [17,21,22]. As the cancer cell line models used herein are all known to be defective in their P14ARF/TP53 and/or p16/pRB checkpoints, the opposing responses of normal fibroblasts and cancer cells to MK3-depletion suggest an involvement of CDKN2A/INK4A. The observation that U-2OS proliferation is negatively affected by gain of MK3 was unexpected in light of the reported epigenetic inactivation of P16 INK4A in osteosarcoma cells [38,39]. Importantly, however, we show in the present and a previous study [8] that P14 ARF is induced in senescent U-2OS/MK3 WT OE, which, like P16 INK4A , is associated with replication checkpoints in human cells [23,41]. Of note, MK2 was recently suggested to control murine haematopoietic stem-cell renewal through P19 ARF [42]. Although the mechanism by which CDKN2A/INK4A expression in U-2OS/MK3 WT OE cells is activated remains elusive at this time, our findings suggests that P14 ARF may be part of the anti-proliferative response in MK3 WT OE cells. In addition, TP53 is induced by MK3 WT OE in both cancer cell lines. Even though P14 ARF is known to stabilize TP53 [43], we cannot formally rule out mutually independent roles for TP53, P14 ARF and P21 CIP1/WAF1 in MK3-dependent proliferative control; such independent roles have been reported [44,45].

MK3 modulation and M/SAPK signaling imbalance
Our previous and current data shows that modulation of MK3 levels causes an imbalance in mitogenic and stress signaling: we find that SAPK P38 and JNK levels are increased in cells overexpressing MK3, whereas loss of MK3 correlates with reduced P38. We recently showed that MK3 controls SAPKs at multiple levels (i.e. JNK and P38 expression/stabilization) and that MK3 inhibition increased pERK levels in human fibroblasts, which correlated with aberrant expression of immediate early genes (IEG; e.g. ATF3, EGR1) [14]. The finding that MK3 affects P38 and ERK also agrees well with the fact that we established genetic interaction between dMK2 and rolled (dERK) and dP38 in Drosophila [14]. MK3 was shown to be targeted by all three canonical M/SAP-kinases [4], which uniquely positions MK3 at the convergence point of potentially conflicting signalling input. We here provide evidence that MK3-overexpression or ablation changes signaling through MAPK and SAPK, and has profound effects on proliferative control both in normal and cancer cell lines, positioning MK3 function both upand down-stream of M/SAPKs, likely as part of regulatory feedback loops. The involvement of MAPKs and SAPKs in cell proliferation and proliferative life span is well documented [46,47]. Immediate-early response genes represent a standing response mechanism to a variety of triggers; their activation is closely linked to M/SAPK and MK action [48][49][50][51]. Our previous studies suggest that altered M/SAPK signalling to IEGs in the context of altered MK3 function plays a vital role in mitogenic and thereby potentially in oncogenic responses [14]. Analogously, the outcome of oncogenic signalling imbalance was reported to be dependent on critical threshold expression levels of the IEG N-MYC and, equally relevant, on the cells' intrinsic (i.e. genetic) state and its microenvironment [52]. Similarly, tumor suppressor function has been proposed to involve dosage and context, rather than all-or-none type effects [53]. As holds true for M/ SAPKs, relative concentrations of MK2, MK3, and MK5 and cellular context have been proposed to dictate protein-protein interactions and thus signalling events and outcome [2,3,49].

MK3 in tumorigenesis
We used the Cancer Cell Line Encyclopedia (CCLE) to assess MK3 expression in numerous cancer cell lines [32]. This survey revealed that MK3 is relatively highly expressed in various cancer cell types whereas its expression is low/absent in other cancer cell types. MK3 was originally proposed as a potential tumor-suppressor gene (TSG), as it located in chromosomal region 3p21.3, which frequently carries deletions in cancer. Of interest, MK3 was recently suggested to harbour potentially oncogenic (driver) mutations, in contrast to MK2 [7]; these included a series of missense mutations: P28S, E105A and D276Y. The P28S mutation occurs N-terminally to a predicted P-loop (AA 52-57) and the ATP-binding pocket (AA 71-77) at a position which may be conserved between MK2 and MK3 as part of a relatively proline-rich area. The E105A mutation occurs in a group of relatively polar adjacent amino acids, in between the ATP-binding pocket and two regulatory threonines (T201/T313), which constitutively activate MK3 when mutated (TT>EE in MK3 CA ). The D276Y occurs in between the aforementioned regulatory threonines. In the current study we used our normal diploid human fibroblast model to read out effects of such two of these mutations on M/SAPK signalling and cell proliferation. Although the exact role of these mutations in tumorigenesis awaits further analysis, our observations suggest that the MK3 P28S and MK3 E105A affect cell proliferation and M/SAPK cell signalling via distinct mechanisms compared to MK3 WT . Expression of both MK3 mutants correlates with a relatively higher pERK level compared to MK3 WT or MK3 kinase mutants (KM or CA). Although both novel mutants enhance P38 signalling (level and phosphorylation), and prolonged expression of the MK3 E105A form correlates with relatively high P16 INK4A expression, neither of the two missense mutants exerts a proliferative disadvantage relative to empty vector control transduced TIG3 cells. The MK3 E105A mutant was originally identified in an endometrial carcinoma (EC); of note endometrial cancers show a relatively high mean MK3 expression level among cancer types [32]. The MK3 P28S mutant was found in a glioblastoma, a rather heterogeneous cancer type in respect to MK3 expression. Although any statement on the effect of these mutations on MK3 activity would be speculative at this point, it is conceivable that such MK3 mutants result in altered subcellular MK3 interactions that tilt the balance toward mitogenic and/or survival signalling. With respect to the MK3 P28S variant an obvious caveat is its low expression level in the current study. In depth examination of their biochemical properties, their interactome and their general effect on cell biology in the absence of wild type MK3 would prove useful to fully understand the possible implications of these potential oncogenic MK3 mutations.
The highly homologous MK2 and MK3 proteins have acquired divergent cellular-context dependent functions [2,54]. MK5 (PRAK), although required for RASV12-mediated OIS, is by itself not capable of inducing senescence and is not required for damage-induced responses [55]. We find that MK3 overexpression induces a phenotype reminiscent of oncogene-induced senescence (OIS), as evidenced by accumulation of DNA damage, activation of a DNA damage response, an intraS-phase arrest accompanied by enhanced TP53/P21 CIP1/WAF1 expression, and ultimately P16 INK4 and SA-bGal expression and morphological alterations typical of senescent cells. Conversely, MK3 depletion provides a selective growth advantage for cancer cell lines, which fits with enhanced mitogenic signalling through MEK/ERK, and is consistent with loss of TSG function [14]. Although the consequences of MK3 modulation on proliferative life span are likely secondary to abnormal signalling and altered checkpoint activity, combined our findings suggest that both MK3 gain-or loss-of-function may contribute to tumorigenesis depending on cellular and/or genetic context.

Functional interactions between MK3 and PRC1 in checkpoint control
The MK3-induced proliferative arrest in normal fibroblasts can be overcome by BMI1 co-expression, and clearly correlates with down-regulation of CDKN2A/P16INK4A expression, suggesting that MK3 and PRC1 cooperatively control proliferative lifespan. The finding that BMI1 only bypasses the effects of MK3 overexpression, not of MK3 depletion in normal cells, suggests that PRC1-dysfunction (i.e. complex disruption, inactivation) is unlikely to provide a common PRC1-dependent mechanistic explanation for the proliferative arrest under MK3 gain or loss-of-function. Instead these findings support the involvement of additional, PRC1-independent mechanisms in MK3-mediated checkpoint and replicative lifespan control. We did not observe a bypass-effect of EZH2 overexpression in TIG3/MK3 WT OE cells in this study. EZH2 is implicated in CDNK2A/INK4A repression, in stem-cell regulation and senescence [26,31,[56][57][58][59][60]. Although published studies and our observations show reduced local H3K27me3 marking at the CDNK2A locus, globally enhanced H3K27me3-chromatin association appears to correlate to physiological stress (e.g. replication, hypoxia, oxidation stress; unpublished observations JWV). Given the concurrent drop in EZH2 levels and the fact that it fails to bypass MK3-induced senescence, this, by inference, implicates other regulators, among which likely H3K27me3 demethylases, in senescence responses [61]. Regulation of gene expression by histone methyl transferases like EZH2 and demethylases is known to involve regulatory non-coding RNAs, functional association with DNMTs and reciprocal functional interactions with TP53 and/or P16 INK4A [28,[62][63][64][65]. The exact role of EZH2 in bypassing senescence is as yet not clear and may depend on tumor cell type [64,66]. As transcriptional regulatory interdependency among PRC genes has been reported [9], whether or not the reduced PRC1 levels observed in this study represent a hallmark of senescence remains to be established. The combined data herein further supports a functional interaction between PRC1 and MK3. The PRC1 oncoprotein BMI1 may thus control human cell proliferation and differentiation in ways not solely dependent on INK4A, but for instance by repression of proto-oncogenes [67]. In keeping with this idea, BMI1/RAS oncogenic collaboration and RAF1-induced senescence also involve p16 INK4A -independent mechanisms [68,69].

Conclusion
In summary, our findings suggest that MK3 controls cell proliferation via multiple pathways, including M/SPAKs, TSGs including TP53, P21 cip1/waf1 , CDKN2A/INK4A and PRC1. The effect of gain or loss of MK3 function is context dependent and is likely to be influenced by the large variation of genetic events (heterozygous and homozygous deletions, loss of heterozygocity) involving chromosome 3p and other chromosomes in cancer. Our findings provide a starting point for systematic evaluation of these effects in different cancer types. The finding that MK3-overexpression apparently reactivates dormant checkpoints in cancer cells is an important observation, as it holds the promise of identification of novel therapeutic targets and stresses the relevance of personalized anti-cancer approaches.
cDNAs and expression systems MK3 KM and MK3 CA mutant cDNAs were produced by S. Ludwig (Münster, GE). The MK3 P28S and MK3 E105A mutants were generated using the QuickChange site-directed mutagenesis method (Agilent Technologies Netherlands B.V.) and sequence verified. Retroviral expression vectors were used to maximize the percentage of expressing cells and to minimize integration effects [74,75]. Retroviral vectors (pBABE-PURO, pBMN-LZRS.ires.GFP, pBMN-LZRS.ires. NEO) expressing murine Bmi1 and human MK3 have been described [8]. Expression vectors encoding the murine ecotropic receptor or hTERT and human HA-tagged EZH2 were kindly provided by R. Bernards (Amsterdam, The Netherlands) and T. Jenuwein (Freiburg, Germany), respectively. Criteria used for shRNA-sequence design and the retroviral expression system were as described before [76]. Targeting sequences are listed in S1 Table. Transduced cells were selected for 1 week on 4-16 μg/ml puromycin (Sigma).

Proliferation Assays
Growth curves were standardized: to compare cell genotypes, all viral transductions (i.e. various expression vectors within one experiment) were performed at the same time; all cells were seeded at equal density (± 20.000 cells/cm2) in 12-multiwell plates (Greiner Bio-one) the preceding evening and allowed to adhere overnight, before the first time point was fixed (t = 0) and followed over time. Cells were collected at the indicated time point cells, washed twice with phosphate-buffered saline and fixed for 10 minutes with 3.7% formaldehyde at room temperature. Next, cells were rinsed 5 times with demineralized water. Cells were stained with 0.1% Chrystal violet for 30 minutes or overnight, and washed 5 times with demineralized water. Chrystal violet was extracted with 10% acetic acid and absorbance was measured spectrophotometrically at 590 nm (Benchmark, Biorad). Data points are based on triplicate measurements within one experiment. Data (cell counts at t = 2 to t = 8) were normalized to cell counts at t = 0 for each transduced cell culture individually. Statistical significance (p<0.05) was determined by two-tailed Student's t-test and presented relative to the empty vector control. All experiments, including transductions and selections, were repeated at least three times using fresh or frozen viral stocks. Immuno-histochemistry, fluorescent in situ hybridization (FISH), cell staining Cells were seeded at 5-10% confluence on glass slides, infected at low MOI (± 1:1) and subjected to a selection (i.e. 8 (TIG3) -16 μg (U-2OS) puromycin/ml). Cells were fixed (2% formaldehyde/PBS, 10 min RT), permeabilized (0.2% Triton-X100/PBS) or directly fixed and permeabilized (100% methanol, 20 min, -20°C). Antisera were diluted in blocking buffer (5% normal goat serum, 5% FCS, 0.02% TritonX100/PBS). Polyclonal (Pab) rabbit antiserum glutathione S-transferase (GST) and monoclonal antiserum (Mab) against MK3 were kindly provided by S. Ludwig [77,78]; Urovysion probe mixture (Abbott Molecular) containing probes for the chromosome 3, 7 and 17 centromeric regions and the 9p21 locus, labelled with SpectrumRed, SpectrumGreen, SpectrumAqua and Spectrum-Gold, respectively, was used; cells were counter-stained with DAPI. SA-bGal assays were performed as previously described [79].