MELK decreases replication stress Maternal embryonic leucine-zipper kinase (MELK) reduces replication stress in glioblastoma cells*

Results: A deficiency of MELK causes replication stress and is associated with a cell-cycle arrest and senescence.


INTRODUCTION
MELK is the only member of the subfamily of AMP-activated Ser/Thr protein kinases that is not activated through T-loop phosphorylation by protein kinase LKB1 (1). Instead, MELK is activated by autophosphorylation (2), but it is not known what triggers this auto-activation. In addition, MELK is regulated through phosphorylation by cyclin-dependent kinases (Cdks) and MAP kinases (3). It has also been established that the level of MELK is strictly controlled. The MELK-encoding gene is an E2F target that is induced in S-phase (4). The concentration of MELK reaches a maximum in the G2/M phases of the cell cycle, but drops again at the mitotic exit (5).
Since it is not known how the activity of MELK can be acutely modulated in intact cells, most functional studies have relied on knockdown or overexpression approaches. These studies have implicated MELK in various cellular processes, including (a)symmetric cell division (6-10), transcription (11), pre-mRNA splicing (12), DNA repair (13) and apoptosis (14-15). However, with some exceptions (15-17), the relevant MELK substrates are unknown, and a unifying hypothesis on the biological function of MELK is missing. It can be argued that at least some of the cellular effects of the loss or overexpression of MELK are indirect and stem from the activation of checkpoints, such as the replication, intra S-phase or the spindle-assembly checkpoint, that result in a cell-cycle arrest or apoptosis (18)(19). A further complication is that at least some of the reported outcomes of a knockdown of MELK might result from siRNA off-target effects (our unpublished data).
MELK is overexpressed in various cancers, including glioblastomas, which are notoriously resistant to chemo and radiotherapy (8, [20][21][22]. The high expression level of MELK clearly gives a proliferative advantage to tumor cells. This has been explained by the ability of MELK to inhibit apoptosis (14) or regulate p53 (17). In addition, MELK has been implicated in DNA-damage response pathways (13, 23). Consistent with this notion, an elevated MELK protein level increases the resistance to DNA-damaging treatments (24). MELK has also been suggested to contribute to cell-cycle progression but, remarkably, both the knockdown and overexpression of MELK were reported to cause a G2/M arrest (20). Other data are indicative for a role of MELK in the G1/S transition (8).
We have re-examined the role of MELK in cell-cycle progression using U87 MG glioblastoma cells and strictly controlled MELK knockdown and rescue experiments. Our results disclose a key role for MELK in progression through unperturbed S-phase and show that MELK is needed to cope with replication stress, in particular DNA double-strand breaks. Thus, MELK has potential as a therapeutic target for the treatment of glioblastomas.
For FACS analysis, cells were synchronized by overnight serum deprivation or a double thymidine block (2 mM thymidine for 16h, 7h release, 2 mM thymidine for 16h). Next, cells were washed twice with PBS, released in complete medium and harvested at the indicated time points. After fixation in 70% ethanol, cells were stained with propidium iodide (50 μg/ml, Sigma), as explained by Shapiro (25). DNA content was analyzed by flow cytometric analysis (FACS-Canto) and quantification was performed by using the FACS Diva software.
BrdU incorporation assays were performed as described by (26). Briefly, 3h after release from a thymidine block, the cells were pulsed for 1h with 10 µM BrdU. After trypsinization and fixation in 70% ethanol, the lysates were denatured in 2M HCl and neutralized with 0.1M sodium borate. Following staining with the primary (M0744, Dako) and secondary (A21202, Life Technologies) antibodies, DNA staining with propidium iodide (pI) was performed and fluorescence was measured for green (BrdU) or red (pI) signal by FACS-Canto and quantified by FACS Diva software.
SA-β-galactosidase staining assays were performed in triplicate in a 24-well plate format, as described by Dimri et al. (27). The number of the SA-β-galactosidase positive cells were expressed as a percentage of the total cell number. MTT assays were performed in triplicates in a 24well plate format, as described by Scudiero et al. (28).
Trypan Blue exclusion viability tests were performed as described by (29). Briefly, the trypsinized cell suspensions were washed and resuspended in PBS. Trypan Blue (0.4%) was added to the cell suspensions (1:1) and incubated at room temperature for 3 minutes. A drop of the cell suspensions was applied to a hemacytometer and counted. The calculations were done as described (29).
Immunofluorescence-microscopy -Cells were grown on glass cover slips in 24-well plates. The cells were washed with PBS (phosphate-buffered saline at pH 7.2-7.4) and fixed using the following steps: (a) 15 minutes in 1.5 M HCl/4% PFA/PBS at room temperature, (b) 3X rinsing in PBS, (c) 15 minutes at -20°C in 100% methanol, and (d) 10 minutes at room temperature in PBS/0.5% NP40.
Subsequently, the samples were blocked for 1 h at room temperature in PBS/1% BSA and incubated for 1 h at room temperature with the primary antibodies, namely BrdU (1:300, DakoCytomation M0744) or H2AX-S139ph (1:500, Upstate 05-636). Following incubation for 1 h at room temperature with the secondary antibodies, DAPI staining was performed and cover slips were mounted in MowOil onto the microscope slides.
Single cell DNA-fiber analysis -Following a double thymidine block and a release for 2 hours, the cells were sequentially labeled with 25 μM IdU and 250 μM CldU for 30 minutes each. DNA fiber spreads were prepared as described previously, with minor modifications (30). Briefly, after trypsinization the cells were resuspended in PBS at approximately 1 × 10 6 cells/ml, spotted onto a microscope slide (3 µl), and lysed with 7 μl of lysis buffer containing 200 mM Tris-HCl at pH 7.5, 50 mM EDTA and 0.5% SDS. After 2-5 minutes, the slides were tilted 15° to allow lysates to slowly move down the slide. The resulting DNA spreads were air-dried and fixed for 10 min in 3:1 methanol/acetic acid. The slides were then treated with 2.5 M HCl for 80 min, washed and stored in PBS (1X) at 4°C overnight. For the stainings, the slides were first incubated for 30 minutes at room temperature in a humid chamber in blocking buffer (5% BSA/PBS) and then for 60 min with 1:400 rat anti-BrdU antibody (Abcam 6326; to detect CldU) plus 1:25 mouse anti-BrdU (BD Biosciences 347580; to detect IdU). After washing in PBS, the samples were blocked for 20 minutes in blocking buffer and then incubated in the secondary antibody solution (1:400, Alexa488 goat anti-rat, and 1:500, Alexa555 donkey antimouse; Invitrogen) for 1 hour. Slides were mounted in MowOil and microscopy was carried out using a Leica SPE confocal unit on DMI 4000B. The fiber lengths were measured using Leica (Las AF Lite) software. The fork progression rate and the frequency of the stalled forks were calculated as explained previously (31,32).

Statistics -
The results are presented as means ± S.E. for at least 3 independent experiments. Statistical significance between the control and the experimental groups was tested using Student's Ttest.

A loss of MELK causes a cell-cycle delay
and senescence -To delineate the contribution of MELK to cell-cycle progression in nonsynchronized U87 glioblastoma cells we first performed a FACS analysis 48h after the RNAimediated knockdown of MELK with 2 different siRNAs (Fig. 1A). The loss of MELK protein was associated with a significantly increased number of cells in G1 and a corresponding decrease in the S plus G2/M phases (Fig. 1B). However, the overexpression of MELK had no significant effect. Consistent with a delay in cell-cycle progression, the knockdown of MELK for 7 days also showed a reduced proliferation rate, as detected by MTT assays (Figs. 1C and D). Moreover, a prolonged knockdown of MELK eventually resulted in an enlarged and flattened cell morphology, and the accumulation of the cell-senescence marker galactosidase (Figs. 1E and F). Also, the number of viable cells, as measured by trypan blue exclusion, was significantly reduced 96 hours after transfection of the siRNA (Fig. 1G ). Similar data were obtained with MCF-7 breast carcinoma cells (data not shown). Thus, a loss of MELK results in a G1/S delay that, during subsequent cell cycles, gradually culminates in a cell-cycle arrest that results in either cell death or a senescence-like phenotype.
The cell-cycle delay associated with the loss of MELK can be due to a deficiency in either G1 or early S phase. Since MELK is an established E2F target and shows a dramatically increased expression at the G1/S transition (4), we reasoned that MELK is likely to play a role in the G1/S transition or in early S phase. To explore this hypothesis, U87 cells were first synchronized in early S phase with a thymidine block and then released for 0-4h. A FACS analysis did not show an effect of a MELK knockdown at the beginning of the release but revealed a significantly delayed progression through S-phase after 4h ( Figs Exploration of the upstream signaling pathways -Cell-cycle arrests are often mediated by the stress-induced expression of inhibitors of cyclin-dependent protein kinases. One of these Cdk-inhibitory proteins is p21 WAF1/CIP1 , which inhibits the G1/S transition, the progression through S-phase as well as the G2/M transition (33). The upregulation of p21 in response to stress conditions, such as DNA damage, is also known to induce senescence (33,34). We found that the knockdown of MELK in U87 cells resulted in an increased expression of p21 (Figs. 3A and B). This effect was obtained with three different siRNAs (Fig. 3C) and was rescued by the ectopic expression of siRNA-resistant EGFP-tagged MELK (Figs. 3A and B). The increased expression of p21 in U87 cells correlated with the hypophosphorylation of the retinoblastoma protein (Rb) at Ser807/811 (Ser807/811ph). Rb phosphorylation by G1 and S phase Cdks serves to disrupt its inhibitory interaction with E2F-type transcription factors, which promote the expression of genes that are important for S-phase (33). As expected from the presence of hypophosphorylated Rb in MELK-deficient cells, the level of E2F-regulated Cdk2 and its associated cyclins (cyclin A and cyclin E) was decreased in this condition (Figs. 3A and B). These effects could be rescued by the expression of siRNAresistant MELK.
We have subsequently investigated whether the increased expression of p21 after the knockdown of MELK is due to the phosphorylation and activation of the transcription factor p53, a key regulator of p21 transcription (34). For that purpose, we compared the effects of a knockdown of MELK in U87 and U251 glioblastoma cells, which express wild-type p53 or an inactive, but stable p53 mutant (35) The activation of p53 as a transcription factor is often correlated with a phosphorylation at Ser15, which prevents its MDM2-mediated ubiquitination and degradation (36). However, we did not see a significant effect of a MELK knockdown on the phosphorylation of p53 at Ser15 (Figs. 5A and B). Neither was the level of p53 affected. Thus, the increased expression of p21 in MELK-deficient cells is p53-dependent, but cannot be explained by an increased steady-state level of p53 or an increased phosphorylation at Ser15.
To get more insights into the mechanism of the p53-dependency of the increased p21 expression, we have examined the effects of a knockdown of MELK on the level of MDMX, an established p53 inhibitor that acts by binding to the transactivation domain of p53 (37). MDMX was downregulated in MELK-deficient cells and this effect could be rescued by expression of siRNA-resistant MELK (Figs. 5A and B). Moreover, the 14-3-3 proteinwhich promotes the degradation of MDMX (38), was significantly upregulated after the knockdown of MELK (Figs. 5A and B). These data indicate that the increased level of p21 in MELK-deficient cells can, at least partially, be explained by the activation of p53 due to a loss of the inhibitor MDMX. This is in agreement with a study in MCF7 cells showing that the depletion of MDMX increases the expression of p21 without a significant increase in p53 levels (39).
MDMX is downregulated through phosphorylation by protein kinase Chk2, resulting in its ubiquitination and subsequent degradation (34,40). Moreover, 14-3-3 collaborates with Chk2 to stimulate MDMX ubiquitination (40). These data prompted us to examine whether a loss of MELK is associated with an activation of Chk2. We indeed found that the knockdown of MELK was associated with a significant increase of the priming phosphorylation of Chk2 at Thr68 (Figs. 5C and D), which leads to oligomerization, autophosphorylation and activation of kinase activity (41). Moreover, the inhibition of the Chk2 kinases ATM/ATR with CGK733 (42) prevented both the phosphorylation of Chk2 and the accumulation of p21 in MELK-deficient cells. Also, an accumulation of p21 in MELK-deficient cells was not detected after the prior knockdown of Chk2 (Figs. 5E and F). Collectively, these data suggest that the accumulation of p21 in MELKdeficient cells can be explained by the consecutive activation of ATM/ATR, Chk2 and p53. In general, Chk2 is activated by ATM whereas ATR phosphorylates and activates Chk1. However, the knockdown of MELK did not cause the activatory phosphorylation of Chk1 at S317 (Figs. 5G and H), indicating that the loss of MELK is associated with the selective activation of the ATM/Chk2 pathway.
The depletion of MELK induces persistent DNA double-strand breaks in unperturbed S phase -Various types of DNA damage, including double-strand breaks (DSBs), are a common feature of unperturbed replication (43). DSBs trigger the activation of the ATM-Chk2 pathway.
One of the earliest responses to DSBs is the Chk2mediated phosphorylation of the histone variant H2AX at Ser139 (γH2AX), which accumulates at DSBs and triggers the recruitment of DNA repair proteins (43). Since a loss of MELK is associated with an activation of Chk2, we examined whether this activation is correlated with an accumulation of γH2AX. The knockdown of MELK in U87 cells that were arrested in G0 by serum starvation did not affect the number of cells with γH2AX foci (Figs. 6A and B). However, a MELK knockdown significantly increased the formation of foci in asynchronized cells (Figs. 6C and D) and in cells that were synchronized in S phase by a double thymidine block (Figs. 6E and F), and this effect was rescued by the expression of siRNA-resistant EGFP-MELK. A simultaneous staining for BrdUincorporation confirmed that the increased accumulation of γH2AX was only detected in replicating cells. Fig. 6G shows the quantification and statistics of the γH2AX data.
A loss of MELK results in the accumulation of stalled replication forks -To define the nature of the damage that results in the accumulation of γH2AX, we monitored the progression of individual replication forks by the successive labeling of newly synthesized DNA with iododeoxyuridine (IdU) and chloro-deoxyuridine (CldU), as visualized by red and green immunofluorescence analysis, respectively. The sequential incorporation of the two nucleotide analogues is required for the correct identification of firing sites and direction of their progression (30,31,44). U87 cells were first synchronized with a double thymidine block and then released for two hours (Fig. 7A). Subsequently, DNA synthesis was followed for two consecutive periods of 30 min each in the presence of CldU and IdU, respectively. The knockdown of MELK resulted in a 30% decrease of the average length of the fibers that were labeled with CldU, reflecting an average decreased fork progression speed (Figs. 7B and C). This effect was rescued by the expression of siRNA-resistant MELK. Since fork velocities vary over a wide range of values (31,44), we have also presented the distribution of frequency of fork velocities in a histogram (Fig.  7D). These data revealed that the loss of MELK causes a shift in the distribution of the frequencies, with a dramatic reduction in the number of fast progressing forks and an increased number of slowly progressing forks.
A reduced fork velocity can be due to a slowdown of DNA polymerase or to an increased number of (transiently) blocked forks. When the progression of DNA polymerase is reduced, the two forks emanating from a single origin move more slowly but bi-directionally with the same speed, forming symmetric replication bubbles. In contrast, fork blocks result in the accumulation of asymmetrical forks (30,31,44).We found that the knockdown of MELK was associated with a significantly increased frequency (≈ 30%) of stalled forks, an effect that was rescued by the ectopic expression of siRNA-resistant MELK (Figs. 7E and F). This increased incidence of stalled forks cannot be explained by mechanically broken DNA fibers since we considered for the quantitation assays only forks with the two halogenated labels.
Melk levels are increased during replication stress -The above data indicate that MELK may be required to cope with replication stress. Accordingly, MELK expression levels were markedly increased after a preincubation with the topoisomerase-I inhibitor camptothecin (45). (Fig   8A ), the replication inhibitor hydroxyurea (46) (Fig 8B) or the DNA-damaging agent bleomycin (47) (Fig. 8C).

A loss of MELK induces a p21-mediated cellcycle arrest -
The role of MELK in cell-cycle progression is poorly understood and the literature on this topic is rather confusing. Some of the previous studies may have led to erroneous conclusions because they were based on improperly controlled siRNA experiments. We noted that 2 out of 5 tested MELK siRNAs caused off-target effects (not shown), which prompted us to adopt a rather rigorous approach and to include a rescue condition with ectopically expressed siRNA-resistant MELK in all critical experiments. This has enabled us to demonstrate unequivocally that a deficiency of MELK in U87 glioblastoma or MCF-7 breast carcinoma cells is associated with a delayed transition through G1-S. In accordance with recent data from Gu et al. (48), this delay was correlated with an increased expression of p21 WAF1/CIP1 , which mainly acts by inhibiting Sphase Cdks, i.e. cyclin E -Cdk2 and cyclin A -Cdk2 (33). The increased level of p21 can also explain why MELK-deficient cells show a decreased proliferation rate and, eventually, develop a cell-cycle arrest and a senescence-like phenotype (34). The expression of p21 is regulated by p53, which binds to p53-response elements in the promoter of p21 and enhances transcription (34). The induction of p21 in MELK-deficient cells is indeed p53-dependent since it was absent in cells that lack functional p53. Under basal conditions the level of p53 is kept low because it is rapidly degraded following ubiquitination by the associated ubiquitin ligase MDM2. In specific stress conditions p53 is stabilized through phosphorylation at Ser15, which prevents the recruitment of MDM2 (49,50). However, the steady-state levels of p53 and its phosphorylation at Ser15 in U87 cells were not affected by the knockdown of MELK, suggesting that p53 is activated in this condition by a distinct mechanism. Indeed, we found that a loss of MELK correlated with a decreased level of the p53 inhibitor MDMX and the activation of Chk2, which phosphorylates MDMX and targets it for MDM2-mediated degradation. In addition, the knockdown of MELK resulted in the upregulation of the p53 target 14-3-3, which is well known to promote the ubiquitination and degradation of MDMX (37). Chk2 knockdown experiments confirmed that this kinase is required for the induction of p21 in MELK-deficient cells. Moreover, the inhibition of the upstream Chk2 kinase ATM prevented both the activation of Chk2 and the induction of p21. In contrast, a deficiency of MELK was not associated with a hyperphosphorylation of Chk1, indicating that the ATR/Chk1 pathway is not activated. Collectively, these data suggest that the induction of p21 in MELK-deficient cells stems from the consecutive activation of ATM, Chk2 and p53. The increased level of p21 inhibits Cdks, which explains the hypophosphorylation of Rb and the decreased expression of E2F targets (Fig. 9).
MELK is required for the repair of DSBs in unperturbed replication -The ATM-Chk2 patwhay is activated by DNA double-strand breaks which recruit proteins, including H2AX, that are needed for repair by homologous recombination or nonhomologous end joining. The resulting repair foci can be visualized with phospho-epitope specific H2AX antibodies. Double-strand breaks can be generated by extrinsic factors (e.g. ionizing radiation) but often also represent a key step in repairing replication forks that collapse during an unperturbed S phase. DNA damage, including DSBs, activate the intra S-phase checkpoint, which delays S-phase progression until the damage is repaired. Our data suggest that the repair of DSBs during unperturbed S phase is hampered in MELK-deficient cells, as illustrated by the activation of the ATM/Chk2 pathway and the accumulation of H2AX foci. This deficient repair is associated with the accumulation of slow or stalled replication forks, hinting at the activation of the intra S-phase checkpoint. However, MELK-deficient cells can still progress to G2/M and even the next G1, indicating that the induced replication stress is relatively mild. The progression to the next G1 is not unexpected given that the G2/M checkpoint is 'imperfect' and can allow mitosis without complete repair (51,52).
The exact role of MELK in DSB repair pathway remains elusive. We speculate that MELK is implicated in the recruitment of repair proteins or the (non)homologous repair mechanism itself. In addition, it is possible that MELK is not involved in the repair process per se, but is a key factor in setting the signaling threshold for the repair of DSBs. In any case, our findings are consistent with previous data that identified MELK as a critical player of the DNAdamage response pathway (13, 23). Our data are also in accordance with the well-established correlation between the increased expression of MELK in glioblastoma cells and their increased resistance to ionizing radiation (24,53).
In conclusion, MELK is emerging as a key regulator of DSB-repair, both in unperturbed Sphase (this work) and after the extrinsic induction of DSBs (24,53). This function may be very important for cancer cells, in particular glioblastoma cells, which are under constant replication stress and have adapted by increasing the expression of MELK. Our data suggest that MELK is an attractive target for a targeted therapy to treat glioblastoma multiforme. Notably, the targeting of MELK with small-molecule inhibitors could be used to sensitize glioblastoma tumors or metastasis to DNA-damaging agents or radiation therapy. The recent reports that the thiazole antibiotic Siomycin A, which reduces the expression level of MELK, inhibits glioblastoma tumor growth in vivo when combined with radiotherapy or chemotherapy (treatment with the alkylating agent Temozolomide), fully supports this notion (53,54).

Footnotes
*This work was supported by the National Science Foundation -Flanders (G.0.686.10.N.10). We thank Chris Marine (VIB-KU Leuven) for the generous gift of MDMX antibodies. Annemie Hoogmartens and An Boeckx provided expert technical assistance. 1 To whom correspondence may be addressed: Mathieu.Bollen@med.kuleuven.be 2 The abbreviations used are: Cdk, cyclin-dependent kinase; DSB, double strand break; MELK, maternal embryonic leucine zipper kinase