Centrosome impairment causes DNA replication stress through MLK3/MK2 signaling and R-loop formation

Centrosomes function as organizing centers of microtubules and support accurate mitosis in many animal cells. However, it remains to be explored whether and how centrosomes also facilitate the progression through different phases of the cell cycle. Here we show that impairing the composition of centrosomes, by depletion of centrosomal components or by inhibition of polo-like kinase 4 (PLK4), reduces the progression of DNA replication forks. This occurs even when the cell cycle is arrested before damaging the centrosomes, thus excluding mitotic failure as the source of replication stress. Mechanistically, the kinase MLK3 associates with centrosomes. When centrosomes are disintegrated, MLK3 activates the kinases p38 and MK2/MAPKAPK2. Transcription-dependent RNA:DNA hybrids (R-loops) are then causing DNA replication stress. Fibroblasts from patients with microcephalic primordial dwarfism (Seckel syndrome) harbouring defective centrosomes showed replication stress and diminished proliferation, which were each alleviated by inhibition of MK2. Thus, centrosomes not only facilitate mitosis, but their integrity is also supportive in DNA replication. Highlights Centrosome defects cause replication stress independent of mitosis. MLK3, p38 and MK2 (alias MAPKAPK2) are signalling between centrosome defects and DNA replication stress through R-loop formation. Patient-derived cells with defective centrosomes display replication stress, whereas inhibition of MK2 restores their DNA replication fork progression and proliferation. Graphical abstract


Materials and Methods
Materials are listed in detail in Supplementary Table 1.

Cell culture
H1299 (non-small cell lung carcinoma, p53 -/-), SW48 (colon carcinoma) and retinal pigment epithelial (RPE) cells, as well as patient-derived skin fibroblasts were maintained in Dulbecco's modified Eagle's medium (DMEM). One set of fibroblasts was derived from a microcephaly patient carrying a homozygous mutation of CEP152, i.e. the splice site mutation c.261+1G>C.
This mutation and its consequences were previously described (Kalay et al., 2011). The mutation gives rise to several splice variants, many of which result in a frame shift. However, one of them leads to an in frame deletion of only two amino acid residues. Moreover, when overexpressed, this mutant localizes to centrosomes. These observations suggest that the mutant is hypomorphic and still capable of expressing some functional CEP152, but not at the level of its wildtype counterpart. Cell culture media were supplemented with 10% fetal calf serum (FCS) and antibiotics (penicillin, streptomycin, and ciprofloxacin). Cells were maintained at 37°C in a humidified atmosphere with 5% CO2. All cell lines were routinely tested for mycoplasma contamination and scored negatively.

Treatments and transfections
Stock solutions of small compounds were prepared in DMSO and then diluted in pre-warmed medium. For each siRNA-knockdown, cells were reverse transfected with 10nM siRNA (Ambion/Thermo Fisher), using Lipofectamine 3000 (Invitrogen). Culture media were changed after 24 hrs, followed by further incubation for 24 hrs. Plasmid transfections were carried out using 2µg of plasmid DNA with Lipofectamine 2000. Media were changed after 6 hrs followed by incubation for 48 hrs.

Fiber assays
DNA fiber assays to analyze replication fork progression and processivity were essentially carried out as described (L A Tibbles, 1996). Following drug treatment or transfection with pooled siRNAs and/or plasmid DNA, cells were incubated with 5-chloro-2′-deoxyuridine (CldU), 25M for 20 min, followed by 5-iodo-2′-deoxyuridine (IdU; both from Sigma-Aldrich), 25M for 60 min. DNA fibers were spread and lysed on glass slides using a spreading buffer (200mM Tris pH 7.4, 50mM EDTA, 0.5% SDS). After acid treatment (2.5M HCl), CldU-and IdU-labeled tracks were detected by 1 hr incubation with rat anti-BrdU antibody (1:400, AbD Serotec; Roche) and transferred into 1.5 ml tubes. Triton X-100 was added to each tube to a final concentration of 0.1%, followed by incubation on a rotating wheel for 15 minutes at 4°C. The tubes were centrifuged at 1300g for 5 minutes at 4°C. The supernatant (cytoplasmic fraction) was transferred to new tubes. The pellets were washed once with Buffer A and then further lysed with 250 μl modified RIPA buffer (1mM EDTA, 150mM NaCl, 0.1% Na-DOC, 1% NP-40, 50Mm Tris pH7.5 and Complete Roche protein inhibitors). 50U of benzonase (nuclease; Novagen) was added to the samples and incubated for 5-10 minutes at room temperature.
Samples were mixed by pipetting during the incubation time until they lost viscosity. Samples were diluted 1:3 with modified RIPA buffer. The supernatant was cleared by centrifuging the samples for 7 min at 16000g, 4°C, and the clear chromatin fraction was transferred to a new tube. After boiling the samples in Laemmli buffer at 95°C for 5 min, equal amounts of protein samples were separated by SDS-PAGE, transferred onto nitrocellulose, and stained with the following antibodies: JUN (Santa Cruz) , ATF2 (Cell signaling), MCM7 (Cell signaling), GAPDH (Abcam).

Immunoblot analysis
Cells were harvested in protein lysis buffer (20 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM Na2EDTA, 1 mM EGTA, 1 mM beta-glycerophosphate, 2 M urea, proteinase inhibitors pepstatin, leupeptin hemisulfate, aprotinin). The samples were briefly sonicated to disrupt DNA-protein complexes. Total protein concentration was measured using a Pierce BCA protein assay kit (Thermo Scientific Fisher). After boiling the samples in Laemmli buffer at 95°C

Depletion of centrosomal components or interfering with PLK4 activity each reduces DNA replication fork progression
To clarify how centrosome integrity affects DNA replication, we inhibited Polo-like Kinase 4 (PLK4) by the small compound Centrinone (Wong et al., 2015), to deplete centrosomes from H1299 cells (lung adenocarcinoma). To assess DNA replication in this context, we performed fiber assays and measured the progression of single replication forks using several cell types.
PLK4 inhibition not only led to a strong reduction in the number of detectable centrosomes  Upon PLK4 inhibition, we also found reduced proliferation of H1299 as well as HCT116 cells, but only after prolonged treatment for more than 2 days (Supplementary Figure 1 P-R).
Similarly, the overall DNA synthesis rate per cell, as revealed by EdU incorporation and fluorescence staining, was only mildly impaired initially, with severe reduction only seen after Taken together, these observations suggest that the disintegration of centrosomes reduces the progression of DNA replication forks, independent of its long term impact on cell division.

PLK4 inhibition or the depletion of centrosomal components each induces a DNA damage response
We hypothesized that the slow progression of DNA replication forks in response to the disintegration of centrosomes might be due to replication stress. To further investigate this possibility, we detected and quantified additional parameters indicative of a corresponding stress response, i.e. proteins that are phosphorylated by kinases of the DNA damage response.
First, we examined the accumulation of γH2AX as a marker of DNA damage in cells with centrosome impairment. Indeed, upon PLK4 inhibition by Centrinone, γH2AX levels markedly

Impairment of centrosomes causes replication stress independent of mitosis
These results suggested to us that centrosomal integrity is required to maintain the processivity of DNA replication. However, it was not clear yet whether this effect is a direct one, or whether centrosome disruption first impairs chromosome segregation during mitosis, which might then impair DNA replication during the next S phase. The latter scenario was plausible for two reasons. Firstly, centrosome disruption often impairs the function of the mitotic spindle and thus chromosome segregation (Meraldi et al., 2016). Moreover, even one additional chromosome (numerical aneuploidy) is sufficient to trigger DNA replication stress (Passerini et al., 2016). We therefore developed a strategy of disrupting centrosomes and assessing DNA replication without allowing the cells to go through mitosis during the time of centrosome impairment. The technical difficulty in doing so consisted in the prolongation of the period of time required to deplete centrosomal componentsa minimum of 72 hrs for siRNA knockdown or 48 hrs for PLK4 inhibition. Therefore, we sought to arrest the cells in G1 for 48 hrs and to disrupt the centrosome during this time. Only thereafter, the cells were released to S phase but not allowed to reach mitosis. To do so, we first arrested the cells in G1, using the cyclin dependent kinase 4 (CDK4) inhibitor Palbociclib (O'Leary et al., 2016). As shown in (Figure 3 A-B), this was achieved in less than 24 hrs. Washing off Palbociclib made the cells re-enter the cell cycle, but with variable time frames required for entering S (data not shown). To synchronize this entry, we released the cells from the CDK4 inhibitor but at the same time added thymidine, which is known to block the cell cycle right after entry into S (Whitfield et al., 2002). We then released the cells from the thymidine block for three hrs and thereby synchronized cells in S phase ( Figure 3 A-B). In this way, we were now able to disrupt the composition of centrosomes and analyze DNA replication without intermediate mitosis. Control experiments revealed that the preceding CDK4 inhibition did not reduce the number of centrosomes, as had been found in different cell species (Adon et al., 2010) (Supplementary Figure 3 A, B). Using this system, we observed a substantial decrease in the progression of DNA replication forks again (Figure 3 C).
Also, the overall DNA synthesis in the PLK4-inhibitor-treated cells was diminished after they had been released from the thymidine block (

Centrosomal disintegration induces p38/MK2 signaling, and this is required for replication stress
Searching for the mechanisms that impair DNA replication fork progression upon centrosomal impairment, we tested the activity of p38/MK2 signaling, a pathway that we had previously found required for reducing DNA replication by nucleoside analogues or CHK1 inhibition (Kopper et al., 2013). Indeed, the phosphorylated and thus active forms of p38 and MK2 were strongly augmented by PLK4 inhibition (Figure 4 A), and the same was found for the bona fide MK2 substrate Hsp27 (Zheng et al., 2006). This is in agreement with previous reports on p38 activity affected by centrosomes (Mikule et al., 2007). Thus, the disruption of centrosomes activates p38/MK2 signaling, independent of mitotic dysfunction. Next, we tested whether the activation of p38/MK2 signaling is a cause of impaired DNA replication upon centrosome disintegration. We treated H1299 cells with the PLK4 inhibitor Centrinone. While assessing DNA replication using fiber assays, we incubated the cells with a pharmacological inhibitor of MK2 (Anderson et al., 2007). And indeed, DNA replication was restored to normal levels by interfering with MK2 activity (Figure 4  Taken together, the observed MK2 activation upon centrosome disintegration causes reduction in DNA replication fork progression, accumulation of γH2AX, and partially restored cell proliferation. Thus, MK2 may act as part of a surveillance pathway that ensures centrosome integrity. In response to centrosome impairment, MK2 interferes with DNA replication and cell proliferation, perhaps avoiding the accumulation of cells undergoing aberrant mitoses and chromosome missegregation.

Upon centrosome disruption, the kinase MLK3 activates p38 and MK2
Given the crucial function of MK2 in replication stress, we sought to determine the upstream signaling pathway that leads to its activation in response to centrosome disruption. An upstream kinase of p38 that was previously found associated with the centrosome is MLK3 (Vertii et al., 2016). Accordingly, we found MLK3 associated with centrosomal structures, but only to a far lesser extent when the cells had been treated with the PLK4 inhibitor Centrinone such rescue of centrosome numbers through MLK3 depletion was observed after knocking down the centriolar component SASS6. Thus, much like MK2, MLK3 is required for a signal triggered by centrosome disruption to interfere with DNA replication. In addition, however, MLK3 functions to further disintegrate the centrosome when PLK4 is inhibited or when a peripheral component of the centrosome, CEP152, is depleted. In any case, MLK3 activates p38 and MK2 in response to centrosome impairment, and this then causes replication stress.

Centrosome disintegration induces RNA:DNA hybrids, and this is required for replication stress
Replication stress is often driven by unscheduled transcription and the formation of R-loops, i. e. RNA hybridizing to DNA (often in association with transcription) and displacing the opposite DNA strand (Aguilera and Garcia-Muse, 2012). Accordingly, upon PLK4 inhibition, we detected the formation of RNA:DNA hybrids by immunostaining with the monoclonal antibody S9.6 directed against these structures (Figure 6 A, B). Upon staining fixed cells in situ, the immunofluorescence signal derived from antibody binding was prominent in discrete nuclear structures, compatible with the concept that R-loops mainly occur at specific sites of highly active transcription (Tsekrekou et al., 2017). In contrast, the overexpression of RNaseH1, an RNase that specifically cleaves the RNA component of RNA:DNA hybrids, strongly reduced the nuclear immunostaining signal, confirming the specificity of the antibody (Figure 6 A, B).
Similarly, using dot blot analyses, the accumulation of R loops was observed upon PLK4 inhibition but also when centrosomal components were depleted ( Figure 6 C-E, Supplementary To clarify the causal link between R-loop formation and replication stress, we performed fiber assays. Upon PLK4 inhibition, the progression of DNA replication forks was largely rescued by RNaseH1 overexpression (Figure 6 H, Supplementary Figure 6 E). We conclude that interfering with centrosomal integrity not only induces R-loop formation, but that this is a major cause of the observed DNA replication stress.

PLK4 inhibition leads to the activation of the transcription factors, thereby enhancing replication stress
We next sought to determine mechanisms that may drive unscheduled transcription and lead to the formation of the RNA:DNA hybrids to induce replication stress, exploring potential transcription factors downstream of p38/MK2 signaling. Upon activation of p38 and MK2, the transcription factors ATF2 and JUN (also known as c-Jun) are phosphorylated and form a dimer to stimulate transcription (Breitwieser et al., 2007). Accordingly, we detected increased phosphorylation of ATF2 and JUN upon Centrinone treatment (Figure 7  Given the multitude of JUN-activated genes, this argues that JUN-mediated transcription might be partially responsible for the occurrence of replication stress. To investigate whether global transcription is indeed responsible for compromised DNA replication, we used an inhibitor of Cdk9 to shut down transcription, as described previously (Klusmann et al., 2016). And indeed, Thus, R-loop formation requires transcriptional activity, which is partially but not fully conferred by JUN activation upon centrosome disintegration.
Regarding replication stress, the rescue by JUN removal was more complete (Figure 7 D). We speculate that this might be due to the robustness of DNA replication. If R-loops need to accumulate above a certain threshold before DNA replication is impaired, this would explain why even a moderate reduction in R-loops, as observed upon JUN depletion, would still restore the progression of DNA replication forks.

Seckel syndrome
The initial suspicion that centrosomes might govern DNA replication had come from the fact that genetic defects of centrosomes on the one hand, and the replication stress kinase ATR on the other hand, each lead to overlapping phenotypes in microcephaly (Kalay et al., 2011).
We therefore asked whether the cells from patients suffering from a centrosomal defect might also display the features of replication stress. And indeed, human fibroblasts from a patient with a defect of the centrosomal component CEP152 (Kalay et al., 2011) showed substantially slower replication fork progression than the fibroblasts from healthy donors (Figure 8 A, Supplementary Figure 8 A). Moreover, the patient-derived cells had increased MK2 activity, as determined by the phosphorylation of HSP27, and they displayed enhanced levels of γH2AX (Figure 8 B). Strikingly, these cells also had increased amounts of p53 and its target gene product p21/CDKN1A, an inhibitor of cell cycle progression. MK2 inhibition markedly reduced the amount of detectable p21 in these cells, perhaps reflecting reduced p53 activity, albeit without affecting the levels of p53 itself (Figure 8 C). This is in agreement with the previously described ability of the p38/MK2 system to activate p53 through phosphorylation at Ser20 (She QB, 2002). Moreover, although patient-derived fibroblasts grew substantially less efficiently than normal fibroblasts, incubation with MK2 inhibitor led to equally efficient growth of fibroblasts from the patient or from healthy donors alike (Figure 8 D). Thus, centrosome disintegration in patients with a Seckel-related syndrome mediates MK2 activation and replication stress.

DISCUSSION
Our results indicate that interfering with centrosomal integrity results in replication stress, even when cells are not allowed to undergo mitosis. Activation of MLK3-p38-MK2 signaling, through R-loop formation, mediates the reduction in DNA replication fork processivity. Thus, supporting unperturbed DNA replication represents a novel function of centrosomes.
A seemingly plausible mechanism of how interfering with the centrosomal function would result in replication stress is through malfunction of the mitotic spindle. When centrosomes disintegrate, mitotic fidelity can be decreased, thus enhancing the missegregation of sister It is thus tempting to speculate that the role of centrosomes in DNA replication might be at least as important as their contribution to the accuracy of mitotic cell division. In terms of evolution, centrosomes may first have acquired their role as a microtubule organizing centre, maintaining the integrity of the cytoskeleton and of cilia. Such a role already requires precise timing of centrosome division. This, on the other hand, may have led to additional functions in the support of additional duplications during the cell cycle, i.e. the replication of DNA and the accuracy of the mitotic spindle.
We have previously characterized the role of MK2 in replication stress. In response to ultraviolet irradiation, treatment with the nucleoside analogue gemcitabine, or CHK1 inhibition, the resulting replication stress depends on MK2 and is relieved when MK2 is depleted or inhibited. At least in part, this rescue by MK2 inhibition is due to the reactivation of translesion synthesis polymerases, and accordingly, DNA polymerase eta was shown to represent an MK2 substrate (Kopper et al., 2013). We suggest that this signalling pathway is triggered by centrosome disruption as well, resulting in diminished progression of DNA synthesis when centrosomes are impaired.
The mixed-lineage kinase 3 (MLK3) was known to associate with centrosomes (Swenson et al., 2003) and to act upstream of MKK3/6 (Zhou et al., 2014). It is also known to activate JNKs and ERK, which may further induce replication stress (Chadee et al., 2006). Here we observed a functional interaction of MLK3 and the p38/MK2 complex. At least one of the functions of this interaction consists in the transmission of a signal that connects the replication of centrosomes and that of the cellular DNA.
Pharmacological inhibition of MLK3 using URMC099 was suggested for treatment of nonalcoholic steatohepatitis (NASH) (Tomita et al., 2017). Our results suggest that this drug may also re-enable DNA replication even under circumstances where MLK3 becomes active. It remains to be determined whether this might be beneficial by supporting liver regeneration, or whether it might support the outgrowth of cancer cell precursors by attenuating replication stress.
Initially, it was surprising to note that microcephaly associated syndromes often referred to as "Seckel syndrome" were found in patients with mutations of CEP152 (Kalay et al., 2011) or PCNT (Griffith et al., 2008), i.e. genes that encode centrosome components. The murine phenotype resembling human Seckel syndrome, including primary microcephaly, was also found in mice with a targeted deletion of CEP63, encoding another member of the centrosome (Marjanovic et al., 2015). Classically, both the human Seckel syndrome (O'Driscoll et al., 2003) and its murine model (Murga et al., 2009) were described in response to hypomorphic recessive alleles of ATR, the central mediator of the replication stress response which is required to dampen replication stress in most cells (Flynn and Zou, 2011). Mutations in the gene encoding the ATR-associated ATRIP protein also result in Seckel syndrome (Ogi et al., 2012). One way of explaining the similarity of syndromes consists in the concept that centrosomes might contribute to ATR signalling. Along this line, the ATR downstream kinase CHK1 was suggested to associate with centrosomes (Antonczak et al., 2016;Kramer et al., 2004) . However, the association of proteins with centrosomes was suggested to be treated with caution due to numerous cross-reactions (Arquint et al., 2014). At least in our hands, CHK1 was not associated with centrosomes. Instead, our results strongly suggest that impaired centrosome composition triggers the translocation of MLK3, followed by activation of p38 and MK2. This signalling cascade induces replication stress, much like the deletion of ATR. This similarity in outcome makes it tempting to speculate that the disruption of ATR signalling, or the impairment of centrosome integrity by genetic alterations, can each lead to replication stress and thus to similar clinical conditions. However, we can currently not rule out that alternative mechanisms, e.g. the induction of p53 (Bazzi and Anderson, 2014;Fong et al., 2016;Lambrus et al., 2016;Meitinger et al., 2016), are more important determinants of microcephaly when centrosomes are defective. Moreover, it should be noted that syndromes that deviate from the classical Seckel phenotype can also occur in response to genetic alterations in centrosomal components (O'Neill et al., 2018). We consider the observed induction of replication stress upon centrosome disintegration as one but certainly not the only mechanism that might contribute to the clinical features when corresponding genes are impaired.
It is also somewhat surprising that the depletion of different centrosomal components results in similar patterns of replication stress. What these components do have in common is their requirement for centrosome duplication. Indeed, it has been reported previously that CCP110 (Chen et al., 2002), SASS6 (Clech et al., 2008), andCEP152 (Cizmecioglu et al., 2010) are each needed to duplicate the centrosome. Also, the inhibition of PLK4 led to aberrant centriole duplication during early stages of the cell cycle (Lei et al., 2018). Correspondingly, we observed reduced centrosome numbers upon depletion of each component. The failure to duplicate centrosomes might trigger the signaling chain reaching from MLK3 through p38 and MK2 to R-loops and replication stress. Another clue to the underlying mechanism is provided by the location of MLK3 on centrosomes, which is transient in nature and is temporarily interrupted shortly before mitotic entry (Swenson et al., 2003). This transient nature of the association at least suggests that MLK3 is not a constitutive part of the centrosome but represents a peripheral and quite lose association partner. As a consequence, we propose that multiple alterations in centrosome composition can dissociate MLK3 from the centrosome, leading to signalling activation in a uniform manner.
Taken together, the findings reported here suggest a novel function of centrosomes in the cell cycle. On top of providing microtubule organizing centres and a supportive function in chromosome segregation, centrosomes also enable the unperturbed progression of DNA replication. In addition to ATR/Chk1 signalling (Dobbelstein and Sorensen, 2015) and histone supply (Alabert et al., 2017), centrosomes support the progression of DNA replication forks.

CONFLICT OF INTEREST
The authors declare no conflict of interest.    Mann-Whitney test.

ACKNOWLEDGMENTS
(D) Partial rescue or cell proliferation by MK2 inhibition, in the context of PLK4 inhibition. 5*10 3 H1299 cells were seeded in each well of a 24-well plate. Cells were treated with the DMSO solvent alone, or with 300 nM Centrinone, with or without 10 M MK2iIII.
Cell proliferation capacity was measured using the Celigo Cytometer (Nexcelom, software version 2.0). Confluence was measured every 24 hrs for 7 days. The experiment was carried out in three biological replicates with two technical replicates each. The mean and standard error were calculated from these six measurements at each time point. Note that most error bars are too narrow to be displayed.     Immunoblot analysis to confirm PLK4 overexpression (note that the apparent molecular weight is increased due to the Flag tag). MK2 activity, as revealed by HSP27 phosphorylation, is increased by gemcitabine, but not when PLK4 is overexpressed.
(H) Increased centrosome formation upon PLK4 overexpression. Synchronized H1299 cells were subjected to plasmid transfection (empty pcDNA3 or pcDNA3-PLK4) for 48 hrs. Centrosomes were detected by immunostaining PCNT, and the DAPI signal was used to detect the nuclei. Scale bar represents 20 μm. Arrows, amplified centrosomes in response to overexpressed PLK4.