Lats1 suppresses centrosome overduplication by modulating the stability of Cdc25B

Numerical aberration of the centrosome results in chromosome missegregation, eventually leading to chromosomal instability, a hallmark of human tumor malignancy. Large tumor suppressors 1 and 2 (Lats1 and Lats2) are central kinases in the Hippo pathway and regulate development and tumorigenesis by coordinating the balance between cell proliferation and apoptosis. Importantly, Lats1 and Lats2 also play pivotal roles in cell cycle checkpoint and mitosis. The Lats proteins localize at centrosomes, but their centrosomal functions remain elusive. Here, we generated Lats1-null knockout (Lats1−/−) mice and established Lats1-null mouse embryonic fibroblasts (MEFs). In Lats1−/− MEFs, centrosomes were markedly overduplicated, leading to severe mitotic defects such as chromosome missegregation and cytokinesis failure. We also found that Lats1 physically interacts with Cdc25B phosphatase that localizes both at the centrosome and in the nucleus and regulates the linkage between the centrosome cycle and mitotic progression. Although Lats1 did not phosphorylate Cdc25B, loss of Lats1 in MEFs caused abnormal accumulation of Cdc25B protein and hyperactivation of Cdk2 toward nucleophosmin (NPM/B23), one of the licensing factors involved in centriole duplication. Taken together, these data suggest that Lats1 regulates Cdc25B protein level and subsequent Cdk2 activity, thereby suppressing centrosome overduplication during interphase.

Scientific RepoRts | 5:16173 | DOi: 10.1038/srep16173 Cdk2-cyclin E and Cdk2-cyclin A complexes 8,9 . The Cdc25 phosphatase family, including Cdc25A, B, and C, is the principal regulator of the activity of the Cdk-cyclin complex during the cell cycle 10 . In particular, Cdc25B localizes to the centrosome throughout the cell cycle [11][12][13][14] and regulates centrosome duplication during interphase and microtubule assembly during mitosis 11 . Consistent with these functions, overexpression or depletion of Cdc25B causes centriole overduplication or loss of centrosome integrity, respectively, in cultured human cancer cell lines 15,16 .
The Hippo signaling pathway, a conserved mediator of growth control and cell fate decision, plays a crucial role in restraining cancer development 17 . Mammalian Large tumor suppressor 1 (Lats1) and Lats2, the main kinase components of the Hippo pathway, phosphorylate and inactivate Yap/Taz, a transcriptional activator of cell proliferation and anti-apoptotic genes. Lats2 localizes to the centrosome/ spindle pole and regulates mitotic progression 18,19 . Loss of Lats2 in mouse cells causes a wide variety of mitotic errors, including centrosome fragmentation, chromosome misalignment, and cytokinesis defects with multinucleation 20,21 . Moreover, centrosome stress induced by treatment of cells with the spindle poison nocodazole causes Lats2 to translocate from the centrosome to the nucleus, thereby preventing polyploidization via the p53 pathway 22 . On the other hand, although human Lats1 also localizes at the centrosomes in human cancer cell lines such as U2-OS osteosarcoma cells and HeLa cervical cancer cells [23][24][25] , to date no study has described the impact of Lats1 dysregulation on the centrosome cycle. Therefore, the biological role of centrosomal Lats1 and the molecular mechanism by which Lats1 regulates the centrosomal integrity remain unclear.
In this study, we generated Lats1-null knockout (Lats1 −/− ) mice, and established Lats1 −/− mouse embryonic fibroblasts (MEFs) from these animals. Lats1 −/− MEFs exhibited centrosome overduplication. Moreover, loss of Lats1 also led to multipolar spindle formation, chromosomal misalignment, micronuclei, and cytokinesis failure, even when the cells entered M phase. Importantly, Lats1 and Lats2 physically interacted with Cdc25B; however, Cdc25B associated more strongly with Lats1 than Lats2. Notably, Lats1 did not phosphorylate Cdc25B. The interaction between Lats1 and Cdc25B contributed to the destabilization of Cdc25B protein and, subsequently, activated Cdk2, thereby preventing centrosome overduplication. These findings suggest that Lats1 stringently regulates the duplication of the centrosome by restricting irrelevant stabilization of Cdc25B, thereby ensuring that the centrosome duplicates once per cell cycle.

Results
Lats1 localizes at the centrosomes during G2/M phase but not during interphase. To elucidate the molecular function of Lats1 at the centrosome, we generated Lats1-null knockout mice by disrupting a part of exon 5 (E5, amino acids 684-853), which encodes the kinase domain ( Supplementary  Fig. S1A-C online). Both male and female Lats1 −/− mice exhibited growth retardation ( Supplementary  Fig. S1D online) and reduced body weight relative to wild-type mice (Supplementary Fig. S1E and F online), suggesting that depletion of Lats1 causes dwarfism in mice; this observation is consistent with the phenotypes of two other types of Lats1-knockout mice generated by truncation at the C-terminus 26 ( Supplementary Fig. S1A online) or N-terminus 27 . However, there was no difference in the incidence rate of spontaneous tumors between Lats1 +/+ and Lats1 −/− in 2-year-old mice. We established Lats1 +/+ and Lats1 −/− MEFs from Lats1 +/+ and Lats1 −/− embryos, respectively. In addition, we performed western blot analysis using two kinds of Lats1 antibodies, CST-C66B5 and Bethyl, which recognize the N-terminal and C-terminal portions of Lats1, respectively; these analyses confirmed that Lats1 −/− MEFs do not express full-length Lats1 protein or any truncated Lats1 fragments ( Supplementary Fig. S1G online). These results demonstrate that the Lats1 −/− MEFs we generated were Lats1-null cells.
Because human LATS1 localizes at the centrosomes during interphase and at the mitotic spindles and spindle poles from metaphase to anaphase 23 , we investigated whether mouse Lats1 actually localizes to the centrosome during the cell cycle. Immunofluorescence analysis revealed that in wild-type (Lats1 +/+ ) MEFs, Lats1 colocalized with γ -tubulin, a component of the pericentriolar material (PCM) that accumulated substantially at the centrosomes (corresponding to mature centrosomes) during late G2 phase, metaphase, and ana/telophase ( Fig. 1A-C, upper panels). During interphase, the Lats1 signal was weak or undetectable at the centrosomes (Fig. 1D, upper panels). To rule out the possibility that these Lats1 signals were experimental artifacts caused by antibody cross-reactions, we examined the centrosomal localization of Lats1 in Lats1 −/− MEFs. As expected, in Lats1 −/− MEFs, Lats1 was not detected at the centrosomes or spindle poles at any stage of the cell cycle ( Fig. 1A-D, lower panels), indicating that these Lats1 signals were not artificial. These results suggest that mouse Lats1 localizes substantially at the centrosomes during G2/M phase, but faintly or not at all during interphase. This distribution is distinct from that of the human LATS1 protein, which localizes at interphase centrosomes in human cancer cells.
Lats1 −/− MEFs exhibit centrosome overduplication. The delocalization of Lats1 from centrosomes in normal MEFs during interphase led us to hypothesize that Lats1 might play an interphase-specific role in centrosome regulation. To test this idea, we investigated the effect of Lats1 deficiency on centrosome duplication by immunofluorescence analysis, using an antibody against γ -tubulin. As shown in Figs 1A,D and 2A, we observed abnormal Lats1 −/− MEFs with more than two γ -tubulin foci per mononucleated cell. The frequency of these cells was markedly elevated in Lats1 −/− MEFs (Fig. 2B). To determine whether the generation of excessive γ -tubulin foci was a consequence of centrosome overduplication or  fragmentation, we co-stained cells for γ -tubulin and centrin, a component of the centriole (Fig. 2C). The excess γ -tubulin foci were colocalized with centrin foci in the majority of Lats1 −/− MEFs (78.7%) (Fig. 2D). These results suggest that loss of Lats1 leads to centrosome overduplication. The mechanism of centrosome overduplication is classified into two different categories: one is dysregulation of canonical templated centrosome duplication cycle 1 , and the other is de novo formation of centrosome with premature centriole 28 . In general, the centrosomes amplified by canonical duplication cycle tend to form a cluster around the centrosome with a mother centriole, whereas the centrosomes amplified by the de novo pathway tend to be dispersed in cytoplasm. Indeed, we found that centrosome overduplication of Lats1 −/− MEFs occurs by two different processes (Fig. 2C, second panels from top and bottom panels). To examine whether centrosome arises de novo in Lats1 −/− MEFs, the cells were co-immunostained with anti-ninein, a marker of mature centriole, and anti γ -tubulin antibodies. The number of ninein foci colocalized with γ -tubulin foci was counted in cells with clustering or scattered centrosomes. In Lats1 −/− MEFs with scattered centrosomes, the frequency of cells with 0 or 1 ninein foci were higher than that with > 2 ninein foci (Fig. 2E, second panels from top and 2F, left bar graph). On the other hand, in Lats1 −/− MEFs with clustering centrosomes, the frequency of cells with > 2 ninein foci were higher than that with 0 or 1 ninein foci (Fig. 2E, bottom panels and 2F, right bar graph). These results suggest that in Lats1 −/− MEFs, the clustering centrosomes are generated from a mature centrosome as a template, whereas the scattered centrosomes arise de novo before the centriole body becomes mature.
When cells with excessive centrosomes enter M phase, supernumerary centrosomes are clustered and eventually form a bipolar spindle to prevent multipolarity, thereby ensuring successful cell division 29 . We measured the frequencies of bipolar cells with centrosome clustering and multipolar cells in M phase. The frequency of bipolar cells with centrosome clustering during M phase was not increased in Lats1 −/− MEFs compared with Lats1 +/+ MEFs (Fig. 2H), although Lats1 −/− MEFs have the excessive centrosomes during interphase (Fig. 2B). Alternatively, the number of Lats1 −/− cells with multipolar spindle during M phase was 2.4 times more abundant than that of Lats1 −/− cells with > 2 PCM foci during interphase ( Fig. 2G,H). These results suggest that spindle multipolarity in Lats1 −/− MEFs is due to centrosome declustering-dependent or -independent mechanisms, perhaps via spindle pole amplification.
Next, we examined whether enforced expression of Lats1 could rescue the centrosome overduplication in Lats1 −/− MEFs. The number of γ -tubulin foci in Lats1 −/− MEFs was markedly decreased by re-expression of 6× Myc-tagged Lats1, relative to expression of the 6Myc-vector alone (Fig. 2I,J), suggesting that Lats1 prevents centrosomal amplification. To rule out the possibility that centrosome overduplication was caused by DNA endoreplication in Lats1 −/− MEFs during S phase, we examined de novo DNA synthesis in Lats1 −/− MEFs using a 5-ethynyl-2'-deoxyuridine (EdU) incorporation assay. Counting of EdU-positive cells revealed that the proportion of cells that had undergone DNA synthesis was lower in Lats1 −/− MEFs than in Lats1 +/+ MEFs (Fig. 2K), suggesting that loss of Lats1 causes centrosome overduplication independently of DNA replication. Consistent with the delay of cell cycle, the cell growth rate was also slower in Lats1 −/− MEFs than in Lats1 +/+ MEFs ( Supplementary Fig. S2A online). To rule out the possibility that Lats1 −/− MEFs underwent apoptosis, apoptotic markers (cleaved caspase-3 and -9) were monitored by western blot analysis ( Supplementary Fig. S2B online). The band of cleaved caspase-3 was not detected in Lats1 +/+ and Lats1 −/− MEFs (top panel, lanes 3 and 4), whereas human cervical cancer cells, HeLa-S3, exhibited the increase of cleaved caspase-3 by UV irradiation (top panel, lane 2). Although in HeLa-S3 cells, pro-caspase-9 (shown by an arrow) was also cleaved and activated by UV irradiation (second panel from top, lanes 1 and 2, arrowheads), the active caspase-9 was not detected in Lats1 −/− MEFs (second panel from top, lanes 3 and 4). These results suggest that loss of Lats1 leads to cell cycle delay but not apoptosis. Taken together, these results suggest that Lats1 negatively regulates centrosome duplication.
Loss of Lats1 generates micronuclei and enlarged nuclei. Lats1 deficiency causes cytokinesis failure 25,27,30 . Indeed, the Lats1-null MEFs exhibited a multinucleated phenotype caused by cytokinesis failure (Supplementary Fig. S3A and B online). Mitotic defects such as centrosome overduplication and cytokinesis failure induce micronuclei and enlarged nuclei through chromosome missegregation. colocalized with γ -tubulin foci in clustering and scattered centrosomes. Data represent the mean and standard deviation of three independent experiments (scattered; n > 30 cells, clustering; n > 15 cells). Predictably, Lats1 −/− MEFs exhibited micronuclei and enlarged nuclei (Fig. 3A,C). Compared with Lats1 +/+ cells, the frequency of mononucleated Lats1 −/− cells with micronuclei or enlarged nuclei was increased (Fig. 3B,D). These results suggest that loss of Lats1 causes not only centrosome overduplication and cytokinesis defects, but also chromosome missegregation and mitotic slippage.
Lats1 interacts with Cdc25B. Initiation of centriole duplication is triggered by activation of Cdk2, which is regulated by Cdc25 phosphatase family during late G1 phase, followed by phosphorylation of nucleophosmin (NPM) 8,31,32 . A recent study showed that overexpression of Cdc25B activates Cdk2, resulting in centrosome overduplication during interphase 16 . To elucidate the function of Lats1 in centrosome duplication, we focused on exploring the molecular relationship of Lats1 to Cdc25B and Cdk2 in some regulators of centrosome duplication. Because Lats1 does not interact with Cdk2 under non-stimulated conditions 33 , we first examined the interactions between Cdc25B and Lats1 or Lats2. To this end, we cotransfected human embryonic kidney 293T cells with 6× Myc-tagged Cdc25B and 3× FLAG-tagged Lats1 or Lats2, followed by immunoprecipitation and western blotting. Cdc25B associated more strongly with Lats1 than Lats2 (Fig. 4A). Previously, we showed that disruption of the N-terminal region of Lats1 resulted in centrosomal overduplication 27 , indicating that the N-terminus of Lats1 may be important for regulation of centrosome duplication. To determine whether the N-terminal region of Lats1 is required for the interaction with Cdc25B, we constructed a Lats1 deletion mutant lacking the N-terminus (amino acids 1-141), including the UBA (ubiquitin-associated) domain (Fig. 4B, Lats1-Δ LCD1). An immunoprecipitation assay revealed that Cdc25B interacted much more weakly with Lats1-Δ LCD1 than with full-length Lats1 (Fig. 4C), suggesting that the N-terminal region (including the LCD1 and UBA domains) of Lats1 is essential for binding to Cdc25B. Next, we investigated whether Lats1 phosphorylates Cdc25B. An in vitro kinase assay revealed that Cdc25B was not phosphorylated by Lats1; Lats1 did phosphorylate Yap (a positive control), but not MDM2 (a negative control) (Fig. 4D). These results suggest that Lats1, rather than Lats2, serves as a regulator of Cdc25B in a manner that is independent of kinase activity.  A defect in Lats1 contributes to stabilization of Cdc25B. Five major splicing variants of human Cdc25B have been identified to date 34 . An anti-Cdc25B antibody detected at least four bands, including two major bands (50-59 kDa) and two minor bands (60-80 kDa) in asynchronously growing wild-type MEFs (Fig. 4E, lane 1). In Lats1 +/+ MEFs, two minor bands (arrow and arrowhead) were attenuated by siRNA-mediated knockdown of Cdc25B, whereas two major bands (asterisks) did not change (Fig. 4E,  lane 2). Moreover, western blot analysis using serial dilution of the extract from Cdc25B knockdown cells revealed that two minor bands were increased in a dose-dependent manner ( Supplementary Fig.  S4A online, arrow and arrowhead). However, because only the slower-migrating of the two minor bands was successfully attenuated using other siRNAs against Cdc25 (siCdc25B-A, B and C), we concluded that endogenous Cdc25B protein in MEFs is primarily reflected by this slower-migrating band ( Supplementary Fig. S4B online, arrow).

Cdk2 activity evokes centrosome overduplication in Lats1 −/− MEFs. Because overexpression
of Cdc25B causes formation of excess centrosomes through Cdk2 activation 16 , we hypothesized that protein stabilization and upregulation of Cdc25B followed by Cdk2 activation would induce centrosome overduplication in Lats1 −/− MEFs. To test this idea, we examined the effect of Cdc25B-mediated Cdk2 activation on centrosome number in Lats1 −/− MEFs. Immunostaining with a γ -tubulin antibody revealed that depletion of Cdc25B by siRNA decreased the number of excess γ -tubulin foci in Lats1 −/− MEFs (Fig. 5A). Moreover, overduplication of the centrosome in Lats1 −/− MEFs was rescued by treatment with a Cdk inhibitor, roscovitine (Rosc), although the centrosome number in Lats1 +/+ MEFs was not altered in the presence or absence of Rosc (Fig. 5B). These results suggest that Cdc25B accumulation and activation of Cdk2 cause centrosome overduplication in Lats1 −/− MEFs.
Thr-199 (T199) of NPM is a target of phosphorylation by Cdk2, which initiates centrosome duplication by promoting centriole disengagement 32,35 . Hence, we next examined the impact of Lats1 deficiency on the phosphorylation level of NPM-T199. Levels of NPM pT199 were higher in Lats1 −/− MEFs than in wild-type MEFs (Fig. 5C, arrow). The increased phosphorylation of NPM-T199 in Lats1 −/− MEFs was suppressed by ectopic expression of wild-type Lats1 (Fig. 5D, lane 3). These results suggest that Lats1 negatively regulates Cdk2 activity toward NPM by inhibiting Cdc25B.
Based on these findings, we propose that Lats1 functions as a negative regulator of the Cdc25B-Cdk2 axis by binding directly to Cdc25B, thereby preventing overduplication or untimely duplication of centrosomes through phosphorylation of NPM (Fig. 5E).

Discussion
In a previous study, we generated Lats1 ΔN/ΔN mice by disrupting the region encoding the N-terminus of Lats1 27 . MEFs established from Lats1 ΔN/ΔN mice endogenously express an N-terminally truncated Lats1 protein, whose kinase activity is retained at least in vitro. These cells also exhibit centrosome overduplication. Here, we generated Lats1-null mice and derivative cells (Lats1 −/− MEFs). Using Lats1 −/− MEFs, we demonstrated that Lats1 contributes to the suppression of centrosome amplification by limiting the protein level of Cdc25B (Fig. 5E). Importantly, the N-terminal region of Lats1, including the LCD1 and UBA domain, is required for the binding of Cdc25B, although Lats1 does not directly phosphorylate Cdc25B (Fig. 4C,D). Because the UBA domain associates with the polyubiquitin chain of some proteins that are destined to be degraded, and recruits them to the proteasome 36 , it is possible that the UBA domain of Lats1 interacts with ubiquitinated Cdc25B, promoting protein degradation of Cdc25B by efficiently recruiting it to the proteasome. In support of this idea, Lats1 ΔN/ΔN MEFs expressing UBA domain-truncated Lats1 protein also exhibit centrosome overduplication 27 , potentially due to loss of the interaction between Lats1 and Cdc25B (Fig. 4C). Moreover, we found that accumulation of Cdc25B protein due to Lats1 deficiency causes aberrant activation of Cdk2 and subsequently promotes the phosphorylation of NPM (Fig. 5B,C).
were detected by western blot analysis. α -tubulin was used as the loading control. (G) Lats1 +/+ and Lats1 −/− MEFs were treated with cycloheximide (CHX) for the indicated periods. Lysates were probed by western blotting with anti-Cdc25B (arrow) and anti-Lats1 antibodies. α -tubulin was used as the loading control. (H) Lats1 −/− MEFs were transfected with 6Myc-Vec and 6Myc-Lats1. Forty-eight hours after transfection, the cells were treated with cycloheximide (CHX) for the indicated periods. Lysates were detected by western blotting with anti-Cdc25B (arrow) and anti-Lats1 antibodies. α -tubulin was analyzed as the loading control. These results suggest that Lats1 plays an important role in the licensing of centrosome duplication by fine-tuning the phosphorylation state of NPM via the Cdc25B-Cdk2 axis.
Lats1 and Lats2 are classified as members of the Dbf2 kinase family, which includes nuclear Dbf2-related proteins 1 and 2 (NDR1 and NDR2) 37 . Previous work showed that overexpression of NDR1 and NDR2, but not Lats1 and Lats2, causes centrosome amplification in U2-OS cells 38 , suggesting that Lats1 and Lats2 are dispensable for the promotion of centrosome duplication. Consistent with this, our results suggest that Lats proteins, unlike NDR proteins, function as suppressors of centrosome duplication, especially overduplication.
Although centrosome duplication is induced during S phase, the majority of Lats1 localizes in the cytoplasm and nucleus during this phase, with little or no Lats1 detectable at the centrosome 27 (Fig. 1D). Fluctuations in the total Cdc25B protein level in cells affect the abundance of centrosomal Cdc25B and the subsequent accumulation of centrosomal proteins such as centrin, ultimately affecting centrosome number 16 . Therefore, one possibility is that cytoplasmic or nuclear Lats1 may influence the level of Cdc25B at the centrosome by regulating the total Cdc25B level. Another possibility is that the abundance of centrosomal Lats1 itself may be stringently regulated by cellular degradation machinery, such as the proteasome, in the vicinity of the centrosome during interphase, in order to prevent inappropriate inhibition of centrosome duplication. On the other hand, between late G2 and M phase, Lats1 also localize to the centrosomes and the spindle poles of mouse cells (Fig. 1A-C), consistent with previous reports regarding the subcellular localization of Lats1 in human cells 23,24 . Salvador (also known as WW45 and hSav) and Mst2, which are core components of the Hippo pathway, also localize at the centrosome; together with Nek2A kinase, these proteins cooperatively regulate the disjunction of centrosomes at mitotic entry 39 , Lats1 might colocalize with Salvador and/or Mst2 at the mitotic centrosome and coordinate the functions of these proteins in centrosome disjunction. Moreover, Lats1 appears to be phosphorylated by Cdk1/cyclin B at the spindle poles during mitosis 40 . However, the biological function of Lats1 at the mitotic centrosome remains unclear.
In our previous study, Lats2 −/− MEFs exhibited centrosome fragmentation (excess γ -tubulin foci without centriole amplification) but not centrosome overduplication (excess centrin foci with centriole amplification) 21 . Here, we showed that Cdc25B interacts more weakly with Lats2 than Lats1 (Fig. 4A). Although the N-terminal regions of Lats1 and Lats2 have a relatively high conserved region, LCD1, which is required for binding to Cdc25B, the overall sequence similarity between their N-terminal regions is very low. Therefore, it is thought that Cdc25B binds to Lats1 and Lats2 via different mechanisms, leading to the dramatically different effects of Lats1 and Lats2 deficiency on centrosome integrity. The Lats1-null knockout mice generated in this study, as well as other Lats1-deficient mice generated by disrupting different genome regions of Lats1 26,27 , were born normally but exhibited growth retardation and smaller body size (i.e., dwarfism), whereas Lats2 deficiency is lethal before embryonic day 12 21 . Primordial dwarfism, a human disorder that involves an extreme reduction in growth and body size, is associated with loss-of-function mutations in genes coding several centrosomal proteins, including core components involved in centrosome duplication (e.g., PCNT, CPAP, and CEP152), the pre-replicative complex (e.g., ORC1, ORC4, ORC6, CDT1, and CDC6), and the DNA damage checkpoint (e.g., ATR and MCPH1) 41 . Notably, in Lats1 −/− MEFs, the proportion of cells that had undergone DNA synthesis was reduced, despite the overduplication of centrosomes (Fig. 2I). Moreover, Lats1 interacts with CDK2 and inhibits its kinase activity toward BRCA2 on S3291 downstream of the ATR-RASSF1A-MST2 axis, thereby preventing replication fork instability in response to replication stress 42 . Therefore, centrosomal Lats1 seems to coordinately regulate precise progression of DNA replication and centrosome during S phase, which might explain its role in disorders associated with body growth failure, such as dwarfism, although cytoplasmic or nuclear Lats kinases are known to suppress overgrowth of certain organs (e.g., liver) by inhibiting Yap and Taz in the Hippo pathway.
As shown in Fig. 2C, individual γ -tubulin foci in Lats1 −/− MEFs do not appear to colocalize with two centrin foci but rather with only one centrin focus, suggesting that each γ -tubulin focus contains only one centriole (these have been called "singlet centrioles" in the literature). In Lats1 +/+ MEFs, centrin-staining area in each γ -tubulin focus was almost completely merged with γ -tubulin-staining area, although we could not clearly discriminate the number (one or two) of centrin foci in each γ -tubulin focus (Fig. 2C,  top panels). This result does not show that γ -tubulin focus contains only one centriole, suggesting that this centrosome is not a "singlet centriole". Likewise, the scattered centrosomes in Lats1 −/− MEFs do not seem to be singlet centrioles (Fig. 2C, second panels from top). In contrast, the clustering centrosomes of Lats1 −/− MEFs seem to be singlet centrioles (Fig. 2C, bottom panels). Therefore, singlet centrioles may be a unique phenotype of clustering centrosomes of Lats1 −/− MEFs. The proximity of these singlet centrioles to one another in the clustering centrosomes of Lats1 −/− MEFs suggests that premature centriole disengagement may be occurring, with consequent "templated" centrosome overduplication. These results suggest that in Lats1 −/− MEFs the scattered centrosomes arise from de novo overduplication, whereas the clustering centrosomes arise from canonical centrosome overduplication with premature disengagement; however, since it seems that γ -tubulin foci may colocalize with only single centrioles, cells with > 2 γ -tubulin foci (specifically, cells with 4 foci) may not actually have supernumerary centrosomes. Thus, our claim about the extent of overduplication in the clustered centrosomes may be overstated.
Centrosome overduplication potentially induces chromosome fragmentation and missegregation, followed by formation of micronuclei 43 . Because micronuclei can indicate chromosomal instability, they have been used as a tool to understand the pathogenesis and the malignancy in human tumor cells 43 . We also found that the proportion of abnormal cells with micronuclei is elevated in Lats1 −/− MEFs, suggesting that in MEFs, Lats1 plays an important role in chromosomal instability through the centrosome overduplication. Indeed, overexpression of Cdc25B and downregulation or mutation of Lats1 and Lats2 are frequently observed in various human cancers 11,37,44 , and loss of the N-terminus (LCD1) of Lats1 can cause tumor formation in nude mice 19 . Thus, in human malignant tumors, deficiency in Lats1 and/or Lats2 might promote overexpression of Cdc25B, leading to centrosome overduplication and, ultimately, to cancer progression. However, even though human Lats1 can interact with human Cdc25B (Fig. 4A), it remains unclear whether Lats1 is involved in centrosome duplication during interphase in human normal and/or cancer cells. Therefore, further studies are needed to determine whether human Lats1 contributes to centrosomal integrity, including duplication and its licensing, as in the case of mouse Lats1.

Methods
Generation of the Lats1 targeted allele. The mouse Lats1 gene was disrupted by replacing part of exon 5 (E5) (amino acids 684-853), which encodes a large part of the kinase domain ( Supplementary Fig.  S1A online). An ES cell (clone #40) identified as harboring a correctly targeted construct was injected into 8-cell stage embryos, which were transferred to pseudopregnant females to generate chimeric mice. PCR analysis using primers B and C ( Supplementary Fig. S1A online) was used to identify bona fide chimeric mice ( Supplementary Fig. S1B online, #40-2, -5, -12, and -13). Next, these chimeric mice were bred with C57BL/6 mice to produce F1 heterozygotes. Homozygous C57BL/6 mice [Lats1 −/− (Accession No. CDB0938K: http://www.clst.riken.jp/arg/mutant%20mice%20list.html)] were obtained by intercrossing the heterozygous offspring. Mouse genotypes were confirmed by PCR analysis of genomic DNA derived from the tails of the offspring using primer pairs A (primer A and C) and B (primer B and C) ( Supplementary Fig. S1C online).