Cyclin E-dependent localization of MCM5 regulates centrosome duplication

During mitosis, equal segregation of chromosomal DNA into two daughter cells is achieved by the formation of a bipolar spindle in which each pole is organized by a centrosome. At the end of mitosis, each daughter cell inherits a single centrosome, duplication of which is initiated near the subsequent G1-S transition of the cell cycle (Hinchcliffe and Sluder, 2002; Kochanski and Borisy, 1990; Kuriyama and Borisy, 1981). Control of centrosome number is essential for accurate chromosome segregation, and centrosome amplification is commonly associated with aneuploidy in many human tumors, suggesting a role in the pathogenesis of cancer (Duensing, 2005; Fukasawa, 2005; Wang et al., 2004). These observations have focused recent attention on the mechanism of centrosome duplication. Several independent groups have provided evidence that, during G1, cyclin-dependent kinase 2 (Cdk2) is an important regulator of the initiation and progression of both centrosome duplication and DNA replication (Coverley et al., 2002; Hinchcliffe et al., 1999; Lacey et al., 1999; Matsumoto et al., 1999; Woo and Poon, 2003). In Xenopus egg extracts, up to three rounds of centrosome duplication occur if the cell cycle is arrested in S-phase by aphidicolin (Hinchcliffe et al., 1999), and this amplification is dependent on active cyclin E–Cdk2 (Hinchcliffe et al., 1999; Lacey et al., 1999). Similarly, certain mammalian cell lines undergo multiple rounds of centrosome duplication during prolonged S-phase arrest, and expression of the Cdk2 inhibitor p21 blocks duplication (Balczon et al., 1995; Matsumoto et al., 1999). A crucial role for cyclin E–Cdk2 during centrosome replication in mammalian cells has been supported by the identification of specific substrates that are directly involved in centrosome duplication, including nucleophosmin B (Okuda et al., 2000; Tokuyama et al., 2001) and CP110 (Chen et al., 2002).

Recent studies have begun to uncover Cdk2-independent functions of cyclin E, including roles in re-entry into the cell cycle from quiescence, DNA endoreduplication, oncogene-mediated transformation, apoptosis and acceleration of the G1-S transition (Geisen and Moroy, 2002; Geng et al., 2007; Matsumoto and Maller, 2004; Mazumder et al., 2007). Furthermore, localization of cyclin E on centrosomes is also independent of Cdk2 binding (Matsumoto and Maller, 2004). This laboratory previously identified a 20-amino acid domain in rat cyclin E (aa 231-250) as a modular centrosomal localization signal (CLS) necessary for both the centrosomal localization of cyclin E and cyclin E-mediated acceleration of S-phase entry. Mutation of four highly conserved residues within the CLS to alanine (aa S234, W235, N237 and Q241) produces a quadruple cyclin E mutant (SWNQ-A) that no longer localizes to centrosomes or accelerates the G1-S transition. Additionally, expression of the GFP-tagged CLS domain of cyclin E was sufficient to target GFP to centrosomes and prevented endogenous cyclin E and cyclin A from localizing to centrosomes (Matsumoto and Maller, 2004). The cyclin E CLS binding partners on the centrosome are currently unknown. To identify potential protein interactions mediated by the CLS of cyclin E, we conducted a bacterial two-hybrid screen of a HeLa cDNA library using the CLS domain as interaction bait. With this approach, we discovered that minichromosome maintenance 5 (MCM5), a protein well documented as an essential DNA prereplication complex factor, interacted specifically with the wild-type CLS but not the SWNQ-A CLS mutant.

We show here that a direct interaction between cyclin E and MCM5 is dependent on the CLS of cyclin E but independent of the binding or activity of Cdk2. The minimal interacting region within MCM5 comprises a 37-amino acid domain (533-569) that has no reported role or function in DNA replication and shares no significant homology with other MCM family members. Additionally, centrosomal localization of MCM5 is dependent on the CLS of cyclin E, and expression of MCM5 but not MCM2 inhibits centrosome over-duplication in CHO cells arrested in S-phase. These studies support an emerging trend in which proteins previously thought to function only in DNA replication also regulate the centrosome duplication cycle (Prasanth et al., 2004b; Stuermer et al., 2007; Tachibana et al., 2005).

To biochemically verify a potential interaction of cyclin E with MCM5, endogenous cyclin E immunoprecipitates from lysates of asynchronous HeLa S3 cells were analyzed for co-precipitation of MCM5 by immunoblotting (Fig. 1A). MCM5 was clearly associated with cyclin E, and there was no significant background binding of MCM5 to an IgG control antibody, demonstrating a specific interaction. To exclude the possibility that co-precipitation was mediated by a bridging protein, the interaction was also evaluated in an in vitro GST pull-down assay (Fig. 1B). GST or GST–cyclin E purified from baculovirus-infected Sf9 insect cells was incubated with MCM5 protein that was produced and radiolabeled with [³⁵S]methionine in an in vitro transcription/translation reaction. The resultant pull-down was separated by SDS PAGE and analyzed by autoradiography. Whereas GST alone had little background binding to MCM5, GST–cyclin E exhibited a strong specific interaction with MCM5 (Fig. 1B). Taken together, these data indicate that cyclin E is capable of directly interacting with MCM5 in vivo.

Given the observation in the two-hybrid assay that interaction between cyclin E and MCM5 is potentially dependent on the CLS of cyclin E, we next tested the necessity of a wild-type CLS for MCM5 binding to cyclin E in mammalian cells. To this end, multiple tetracycline-inducible stable CHO cell lines were generated that expressed either full-length 6Myc-tagged wild-type rat cyclin E, full-length 6Myc-tagged rat cyclin E containing the four (SWNQ-A) CLS point mutations, or full-length 6Myc-tagged rat cyclin E containing a single point mutation (S180D) that blocks nearly all Cdk2 binding and activity (Matsumoto and Maller, 2004) (Fig. 1C, lower panel). Expression of the various Myc-tagged cyclin E constructs was induced for 18 hours by addition of doxycycline, and Myc immunoprecipitates were analyzed for the presence of MCM5 by immunoblot (Fig. 1C, top panel). Binding to MCM5 was essentially equivalent with wild-type cyclin E and the S180D mutant of cyclin E, whereas binding to MCM5 was greatly reduced with cyclin E containing the SWNQ-A mutant CLS. The finding that an in vivo interaction between cyclin E and MCM5 does not require binding to Cdk2 but does require a wild-type CLS verifies the binding observed in the two-hybrid screen.

A number of domains in MCM5 have been identified that function in DNA replication or regulation of transcription (Barry et al., 2007; Forsburg, 2004; Jenkinson and Chong, 2006; Kearsey and Labib, 1998). It was therefore important to determine what region within MCM5 mediates an interaction with cyclin E. Sequencing during the initial two-hybrid screen resulted in isolation of eight different partial cDNA constructs encoding various regions of MCM5 (data not shown). Alignment of all eight sequences revealed a minimal 72-amino acid region (aa 496-569) within MCM5 that was shared among all constructs. We tested both a C-terminal HA-tagged construct of MCM5 (aa 496-734) as well as an HA-tagged construct encoding only the 74 shared amino acids (aa 496-569) in an in vitro GST pull-down assay (Fig. 2A). Both of the truncated MCM5 proteins were able to specifically bind cyclin E, with little or no background binding to GST alone. To further elucidate the region necessary for an interaction with cyclin E, aa 496-569 of MCM5 were subdivided into two equal GFP-tagged segments and examined for their ability to interact with GST–cyclin E (Fig. 2B). Whereas neither GFP alone nor GFP-MCM5 aa 496-532 showed any significant ability to interact with cyclin E, aa 533-569 of MCM5 specifically bound to cyclin E, albeit at a lower level than that observed for longer constructs. However, it is evident that aa 533-569 of MCM5 are sufficient for an interaction with cyclin E, and this region in MCM5 is largely conserved from yeast to human (Fig. 3).

Given the specificity of this interaction for a wild-type CLS in cyclin E, we next investigated the potential for centrosomal localization of endogenous MCM5. Both CHO-K1 and HeLa S3 cells were fixed and immunostained with antibodies to MCM5 and γ-tubulin, a marker for centrosomes (Fig. 4A). In an asynchronous population, endogenous MCM5 colocalized with γ-tubulin in a majority of the cells. Additionally, ectopically expressed HA-tagged full-length MCM5 also localized to centrosomes (Fig. 4B). In order to determine whether an interaction with cyclin E is necessary for MCM5 to be centrosomally localized, the same MCM5 constructs used in GST pulldown assays (Fig. 2) were transiently transfected into CHO-K1 cells and examined by indirect immunofluorescence for localization. Both the full C-terminal fragment (aa 496-734) as well as the 72-amino-acid `shared region' (aa 496-569) of MCM5 colocalized with γ-tubulin (Fig. 4B), but GFP alone or GFP-MCM5 aa 496-532 – the fragment that did not bind cyclin E (Fig. 2B) – did not localize to centrosomes. Significantly, GFP-MCM5 aa 533-569 – the fragment that did retain the ability to bind cyclin E – was centrosomally localized (Fig. 4C).

Although these data suggest that an interaction with cyclin E is necessary for MCM5 to localize to centrosomes, we also examined the localization of endogenous MCM5 in cells expressing either the GFP-tagged wild-type CLS or SWNQ-A mutant CLS of cyclin E. It has previously been shown that endogenous cyclin E is displaced from centrosomes presumably by competition with the expressed intact CLS (Matsumoto and Maller, 2004), and therefore expression of the CLS might also displace MCM5 from centrosomes. CHO-K1 cells were fixed 20 hours after transient transfection of either the wild-type or SWNQ-A-mutant CLS construct and immunostained for endogenous MCM5. GFP-positive cells were identified by fluorescence of the expressed GFP tag rather than by antibody detection. Importantly, expression of GFP alone (control) or the SWNQ-A mutant CLS had no significant effect on MCM5 localization, whereas the wild-type CLS greatly disrupted centrosomally localized endogenous MCM5 (Fig. 5B).

Although these results support the hypothesis that the centrosomal localization of MCM5 is dependent on the CLS of cyclin E, it could not be excluded that expression of the CLS displaced MCM5 through removal of a different protein, e.g. cyclin A (Matsumoto and Maller, 2004). Therefore, we examined the localization of MCM5 during expression of the wild-type CLS while also forcing cyclin E to return to centrosomes via a different targeting motif. Subcloning of the pericentrin-AKAP450 centrosomal targeting (PACT) domain (Gillingham and Munro, 2000) upstream and in-frame of each 6Myc-tagged cyclin E construct created fusion constructs that should be centrosomally localized regardless of the CLS. This 91-amino-acid domain (aa 3699-3790) has been reported to be sufficient to target GFP to centrosomes (Gillingham and Munro, 2000). Fusion of this domain to the N-terminus of cyclin E was sufficient for strong centrosomal localization of all cyclin E constructs, including cyclin E containing an SWNQ-A mutant CLS (Fig. 5A). Presumably, if cyclin E is re-targeted to centrosomes despite expression of the wild-type CLS, it could re-establish the centrosomal localization of endogenous MCM5.

CHO cells were co-transfected with wild-type GFP-CLS along with the indicated 6Myc-tagged PACT-fused cyclin E constructs, fixed and analyzed for the co-expression of both plasmids within cells as well as for the localization of the PACT–cyclin E proteins. In general, approximately half of the Myc-positive cells (i.e. expressing Myc-tagged PACT-fused cyclin E) were also GFP positive (expressing GFP-CLS), whereas nearly 99% of the GFP-positive cells were also Myc-positive. This allowed analysis of the localization of MCM5 in GFP-positive cells with approximately 99% confidence that these cells were also co-expressing the desired PACT-fused cyclin E protein. Importantly, cells that were GFP positive also displayed a strong centrosomal localization of the PACT–cyclin E proteins, indicating that expression of the CLS does not interfere with localization that is mediated by the PACT domain (data not shown).

Analysis of endogenous MCM5 localization within GFP-positive cells showed that co-expression of the PACT-fused wild-type cyclin E or the S180D non-Cdk2-binding mutant of cyclin E restored the centrosomal localization of MCM5 to near normal levels. However, co-expression of PACT-fused cyclin E containing the SWNQ-A CLS point mutations that prevent binding to MCM5 (Fig. 1C) did not rescue MCM5 localization even though the CLS-mutant cyclin E protein had been re-localized to the centrosome (Fig. 5A,B). Taken together, these data strongly indicate that an interaction with cyclin E, mediated by the CLS, directly recruits MCM5 to centrosomes and that MCM5 is co-displaced from centrosomes with cyclin E when the CLS domain is expressed.

We next wanted to explore whether MCM5 has a functional role in centrosome duplication. In certain cell lines, S-phase inhibition uncouples the DNA replication cycle from the centrosome duplication cycle and allows multiple rounds of centrosome duplication to occur, even in the absence of DNA replication. Furthermore, apparent centrosome amplification in this system has been shown by electron microscopy to be a result of template-driven centrosome duplication events rather than from centrosome splitting or de novo generation (Balczon et al., 1995). Using this system, we examined centrosome duplication in hydroxyurea (HU)-arrested CHO-K1 cells expressing MCM5. After 48 hours of HU-mediated S-phase arrest, control CHO cells contained an average of seven γ-tubulin staining foci per cell, whereas cells transfected with untagged or HA-tagged full-length wild-type MCM5 contained an average of only four γ-tubulin staining foci (Fig. 6A). Strikingly, the same degree of inhibition of centrosome duplication was observed in HU-arrested cells expressing only the HA-tagged 74 amino acid (496-569) fragment of MCM5 (Fig. 6B), a fragment that binds cyclin E (Fig. 1A) and localizes to centrosomes (Fig. 2C). Importantly, as an additional control for an MCM5-specific effect, cells that were expressing MCM2, a related member of the MCM family, exhibited no significant inhibition of centrosome duplication (Fig. 6B). Taken together, these results suggest that MCM5 has a specific negative regulatory function in centrosome duplication. Finally, we examined the localization of endogenous cyclin E while expressing MCM5 (Fig. 7). If expression of MCM5 displaced cyclin E from centrosomes, this could be predicted to result in inhibition of duplication. However, the presence of endogenous cyclin E at centrosomes was evident even in cells highly expressing MCM5 (Fig. 7).

In many cell types, both centrosome duplication and DNA replication are initiated at the G1-S transition. This essential coordination of DNA and centrosome duplication is achieved at least in part by the activation in late G1 of Cdk2 coupled to cyclin E (cyclin E–Cdk2). Numerous studies have demonstrated that cyclin E–Cdk2 plays an essential role in S-phase entry. At the G1-S transition, active cyclin E–Cdk2 phosphorylates Orc-bound pre-replication complexes that are assembled on DNA replication origins, thereby promoting the recruitment of DNA polymerases and the initiation of DNA synthesis (Forsburg, 2004; Masuda et al., 2003; Strausfeld et al., 1996; Zou and Stillman, 2000). However, continued high levels of Cdk2 activity prevent the loading of pre-replication-complex proteins onto DNA and inhibit inappropriate re-licensing and re-replication (Dahmann et al., 1995; Forsburg, 2004; Hua et al., 1997; Ishimi and Komamura-Kohno, 2001). Although the precise functional roles and target substrates of cyclin E–Cdk2 with regards to DNA replication have been extensively studied, only within recent years has it become evident that Cdk2 also regulates centrosome duplication (Hinchcliffe et al., 1999; Hinchcliffe and Sluder, 2002; Lacey et al., 1999; Matsumoto et al., 1999; Winey, 1999). In both Xenopus embryos and mammalian cells, over-duplication of centrosomes is blocked by Cdk2 inhibitors, and it has been suggested that, under some conditions, cyclin A–Cdk2 might also play a crucial role (Meraldi et al., 1999). The identification of centrosomal cyclin E–Cdk2 substrates, such as nucleophosmin B and CP110, which are essential for initiation of centrosome duplication, has provided preliminary indications of how centrosome duplication is positively regulated by cyclin E (Chen et al., 2002; Okuda et al., 2000; Tokuyama et al., 2001). The results in this paper suggest that, by recruiting MCM5 to the centrosome, cyclin E–Cdk2 might also provide negative regulation of centrosome replication, thereby preventing over-duplication (Fig. 8).

Recent studies have presented compelling evidence for Cdk2-independent functions of cyclin E (Geisen and Moroy, 2002; Geng et al., 2007; Honda et al., 2005; Matsumoto and Maller, 2004; Mazumder et al., 2007). In 2004, Matsumoto and Maller demonstrated that localization of cyclin E to the centrosome requires an intact CLS but not Cdk2 binding (Matsumoto and Maller, 2004). Differences between the knockout phenotypes of mice lacking either cyclin E or Cdk2 have also challenged the traditional view that cyclin E functions only as an activator of Cdk2. Cdk2 knockout mice are viable and do not display anatomical or behavioral abnormalities, except for severe gonadal atrophy (Berthet et al., 2003). Additionally, Cdk2^–/– mouse embryo fibroblasts proliferate normally in culture, indicating that Cdk2 activity is not essential for mitotic cell division. In these cells, cyclin E expression levels and centrosomes are normal, presumably because cyclin E is bound to Cdk1 (Aleem et al., 2005; Berthet et al., 2003). However, a double knockout of both cyclin E1 and cyclin E2 was embryonic lethal owing to placental defects that were not observed in Cdk2 knockout embryos (Geng et al., 2003). In culture, fibroblasts isolated from double cyclin E knockout embryos were found to actively proliferate and increase in cell number similar to wild-type controls as long as the cells were continuously cycling. However, cyclin E-deficient cells in G0 after serum starvation failed to re-enter the cell cycle after serum stimulation, owing to a failure to load MCM proteins onto origins of replication (Geng et al., 2003). Further work with these cells showed that cyclin E knockout phenotypes could be at least partially rescued after retroviral-mediated expression of cyclin E, including a construct impaired for activation of Cdk2 as well as a construct with a mutant CLS (Geng et al., 2007). This rescue was associated with restoration of MCM loading and is consistent with multiple studies in egg extracts, yeast and mammalian cells indicating that Cdk2 activity is not needed to load MCM proteins onto origins but is required for initiation of DNA replication (Lei and Tye, 2001; Prasanth et al., 2004a). In this connection, our work here demonstrates that cyclin E-mediated recruitment of MCM5 to centrosomes also does not require Cdk2 binding or activity.

We biochemically verified an in vivo interaction between cyclin E and MCM5 and mapped the region within MCM5 that is sufficient for interaction to amino acids 533-569. This portion of MCM5 is located in the C-terminal region and has not previously been attributed to any known function of MCM5. Whereas alignment analysis shows no significant conservation within this region to other MCM family members, residues in this region are highly conserved in MCM5 sequences from different species (Fig. 3). Additionally, we found that both endogenous and exogenously expressed MCM5, as well as C-terminal cyclin E-interacting fragments, are centrosomally localized. This localization appears to require an interaction with cyclin E, because only those constructs of MCM5 that interact with cyclin E are also capable of centrosomal localization. Moreover, PACT-driven localization of cyclin E is sufficient to direct MCM5 to the centrosome only if the CLS is wild type (Fig. 4). These data are in agreement with binding assays indicating that the SWNQ-A CLS mutation greatly impairs MCM5 binding and should therefore be unable to recruit MCM5 to centrosomes. Although the CLS is necessary for the binding of MCM5 to cyclin E, it is probably not sufficient, because GFP-CLS expression displaced MCM5 from centrosomes along with cyclin E (Fig. 5B) (Matsumoto and Maller, 2004). Taken together, these data indicate that the most likely mechanism by which MCM5 is centrosomally localized is direct recruitment by interaction with cyclin E that has a wild-type CLS.

The likelihood that MCM proteins have diverse functions is suggested by the finding that they are highly abundant, with their numbers greatly exceeding the number of replication origins (Bell and Dutta, 2002; Donovan et al., 1997; Lei et al., 1996). Even though only a fraction of the total MCM proteins can be mapped to replication sites in S-phase, it has been estimated that anywhere from 40 to 100 MCM complexes are loaded onto chromatin for each origin of replication (Edwards et al., 2002; Forsburg, 2004; Young and Tye, 1997). Additionally, this large pool of MCM proteins is required, because even a partial reduction in MCM level that does not greatly affect DNA replication leads to increased recombination events, chromosome loss and checkpoint sensitivity (Bell and Dutta, 2002; Liang et al., 1999). The excess level of MCM proteins along with broad binding on chromatin has been termed the `MCM paradox' (Forsburg, 2004; Hyrien et al., 2003; Laskey and Madine, 2003) and further studies have shown that MCM proteins have multiple functions outside of DNA replication, including transcriptional activation and repression, chromatin remodeling, meiotic recombination, chromatin cohesion, and checkpoint responses (Agarwal et al., 2007; Forsburg, 2004; Gauthier et al., 2002; Snyder et al., 2005; Sterner et al., 1998). It is significant that, in addition to the localization of MCM5 on centrosomes that is reported here, localization of both Orc2 and geminin has also been reported on centrosomes (Prasanth et al., 2004b; Stuermer et al., 2007; Tachibana et al., 2005). These findings support an emerging trend in which essential DNA-replication proteins have other functions that are mediated by distinct domains in the protein.

Because a significant reduction in MCM5 protein level has been reported to cause S-phase arrest (Ryu et al., 2005), we performed MCM5 siRNA knockdown in HeLa S3 cells, a line previously reported to not reduplicate centrosomes during prolonged S-phase arrest (Kasbek et al., 2007; McDermott et al., 2006). At 96 hours after siRNA transfection, a near complete knockdown of MCM5 protein was observed, with no detectable difference in MCM2 protein level, and S-phase progression was inhibited (data not shown). Approximately 30% of the siRNA-treated cells displayed an increase in centrosome number, which was not observed in cells treated with negative-siRNA controls. However, HeLa S3 cells placed under HU arrest alone for 96 hours had nearly the same percentage of cells with an abnormal number of centrosomes (data not shown). Consequently, increased centrosomal numbers in cells after treatment with MCM5 siRNA was probably an indirect result of inhibiting S-phase and cannot be necessarily attributed to loss of MCM5 function in centrosome duplication.

At present, it is unclear how the recruitment of MCM5 by the CLS inhibits cyclin E-mediated centrosome over-duplication in HU-arrested cells. In vitro kinase assays revealed no phosphorylation of MCM5 by purified cyclin E–Cdk2 (data not shown). In addition, even in cells highly expressing MCM5, colocalization of endogenous cyclin E with γ-tubulin was not altered (Fig. 7). Therefore, inhibition of centrosome duplication by MCM5 expression is not due to displacement of endogenous cyclin E from the centrosome. Because inhibition of centrosome duplication can be achieved with a discrete small domain of MCM5 (aa 496-569), it probably reflects a mechanism that is independent of MCM5 function in DNA replication or transcription. It has long been known that coupling of DNA replication and centrosome duplication is essential for ensuring genomic stability, but the precise mechanisms by which one cycle exerts control over the other is not fully understood. Our data potentially provide evidence that, in addition to cyclin E–Cdk2, other proteins associated with the initiation and progression of DNA replication also regulate centrosome duplication. It is attractive to speculate that the release of MCM5 from chromatin during S-phase progression could potentially act as a negative feedback mechanism that prevents inappropriate centrosome re-duplication (Fig. 8).

CHO-K1 and HeLa S3 cells were grown at 37°C in a 5% CO₂ atmosphere in Ham's F12K medium (Invitrogen) supplemented with 10% fetal bovine serum (FBS) (HyClone) and penicillin-streptomycin (50 U/ml and 50 g/ml, respectively) (Invitrogen). Flp-In CHO cells were purchased from Invitrogen and maintained in Ham's F12 media supplemented with 10% FBS, antibiotics and 100 g/ml Zeocin (Invitrogen). Stable integration of the tetracycline repressor to create a tetracycline-responsive CHO Flp-In T-Rex cell line was carried out according to the manufacturer's directions. Briefly, Flp-In CHO cells were transfected with linearized pcDNA6/TR and maintained in medium containing 6.5 g/ml Blasticidin (Invitrogen) for approximately 2 weeks. Colonies were trypsinized, and individual cells isolated and placed into 96 wells. Expanded cells were screened by western blotting for expression of the tetracycline repressor (Abcam, 14075) and colonies displaying the highest expression level of the tetracycline repressor were tested for tetracycline-inducible gene expression by transient transfection of the pcDNA5/FRT/TO/CAT positive control vector supplied by Invitrogen in the Flp-In T-Rex core kit. Cells were analyzed by western blotting for expression of CAT (Sigma, C9336) with and without doxycycline treatment. The final host CHO Flp-In T-Rex cell line was expanded from the clone that exhibited the lowest level of basal expression without doxycycline and the highest expression of CAT 20 hours after addition of 1 g/ml doxycycline. Establishment of Flp-In T-Rex Myc-tagged cyclin E expression cell lines was performed following the manufacturer's directions and selection of stable transfectants was carried out by culturing cells in the presence of 600 g/ml Hygromycin B (Invitrogen).

All transfections were carried out with Lipofectamine 2000 transfection reagent (Invitrogen) according to the manufacturer's directions. At 6 hours after transfection, cells were washed with sterile PBS and given fresh media without antibiotics.

For endogenous cyclin E immunoprecipitation (IP), HeLa S3 cells were lysed in NP-40 lysis buffer (150 mM NaCl, 50 mM Tris pH 8.0, 1% NP-40) supplemented with 1× complete protease inhibitor (Roche), 1 mM AEBSF (Sigma-Aldrich), 1 mM NaF and 1 mM Na₃VO₄. Protein concentration was determined by Bradford assay and 1 mg of lysate was used for each IP. Magnabind protein-G beads (Pierce) were incubated with either anti-human cyclin E antibody (Upstate, HE12) or control mouse IgG (Santa Cruz) for 2 hours at 4°C. Beads were thoroughly washed with NP-40 lysis buffer and blocked with 5% BSA/PBS for 6 hours at 4°C. Beads were then incubated with protein lysate overnight at 4°C, washed twice with NP-40 lysis buffer and twice with high-salt (250 mM NaCl) NP-40 lysis buffer. After elution with 1× Laemmli sample buffer, samples were separated on 10% SDS-PAGE, transferred to PVDF membranes and immunoblotted for the protein of interest. For blotting, the following primary antibodies were used against cyclin E (Upstate, 06-459), MCM5 (Abcam, 17967) and Cdk2 (Sigma, C5223). As a secondary antibody, HRP-conjugated goat anti-rabbit (Pierce) was used.

For Myc IP, Flp-In T-Rex stable-expression cell lines were induced for 20 hours using 1 g/ml doxycycline. Cells were lysed using NP-40 lysis buffer supplemented with 1× complete protease inhibitor, 1 mM AEBSF, 1 mM NaF and 1 mM Na₃VO₄. Lysate protein concentration was determined by Bradford assay and equal amounts of protein were used for each IP (600 g). Preconjugated Myc-antibody agarose beads (Santa Cruz, 9E10) or control preconjugated mouse IgG agarose beads (Santa Cruz) were washed in NP-40 lysis buffer and blocked in 5% BSA/PBS for 6 hours at 4°C. Beads were then thoroughly washed with lysis buffer and incubated with lysate overnight at 4°C. Beads were washed three times with NP-40 lysis buffer and eluted with 1× Laemmli sample buffer. Samples were separated on 10% SDS-PAGE, transferred to PVDF membranes, and the protein of interest was detected by immunoblot analysis. The following primary antibodies were used against Myc (Santa Cruz, 9E10-HRP conjugated), MCM5 (Abcam, 17967) and Cdk2 (Sigma, C5223). As a secondary antibody, HRP-conjugated goat anti-rabbit (Pierce) was used.

GST and GST-tagged human cyclin E were purified from baculovirus-infected Sf9 cells obtained from the University of Colorado Cancer Center Tissue Culture Core Facility. Briefly, Sf9 cell pellets were resuspended in Sf9 lysis buffer (50 mM Tris pH 7.4, 300 mM NaCl, 1 mM MgCl₂, 0.1% TX-100) supplemented with 1× complete protease inhibitor mix and 1 mM DTT, and sonicated three times for 30 seconds each. Lysates were spun twice at 20,000 g for 10 minutes and the supernatant incubated with glutathione-Sepharose 4B beads (Amersham Biosciences) for 1 hour at 4°C. Beads were washed twice with Sf9 lysis buffer supplemented with 450 mM NaCl and twice with Sf9 lysis buffer containing 150 mM NaCl. The beads were resuspended in 50% ethylene glycol/50% SF9 lysis buffer with 1 mM DTT, aliquoted and stored at –20°C. For pulldown analysis, equal amounts of GST and GST–cyclin E bound to glutathione beads were washed three times with TIF buffer (20 mM Tris pH 8.0, 150 mM NaCl, 1 mM MgCl₂, 0.1% NP-40 and 10% glycerol) and blocked in 5% BSA/PBS for 1 hour at 4°C. Beads were washed and incubated overnight at 4°C with MCM5 protein produced in an in vitro transcription-translation (TNT)-coupled reaction according to the manufacturer's protocol (Promega). MCM5 protein was radiolabeled by incorporation of [³⁵S]methionine during the TNT reaction. Beads were washed three times, eluted with 1X Laemmli sample buffer and separated on a 10% SDS-PAGE. Gels were dried and exposed to Kodak Biomax MR film.

Cells were seeded onto collagen-coated glass coverslips (Iwaki, Japan) and fixed with –20°C methanol for 15 minutes after the indicated treatments. After rehydration in PBS for 30 minutes, cells were blocked in 5% BSA/PBS overnight at 4°C and immunostained with the following primary antibodies against MCM5 (Santa Cruz, 22780), -tubulin (Sigma, GTU-88), hemagglutinin (HA)-Alexa-Fluor-488 conjugate (Molecular Probes, 16B12), c-Myc-FITC conjugate (Santa Cruz, 9E10), -tubulin (Sigma, AK-15) and cyclin E (Abcam, 2094) as indicated. GFP constructs were visualized by GFP excitation of the expressed GFP-tagged protein rather than immunostaining. All primary antibody incubations were for 1 hour at room temperature followed by appropriate secondary antibody staining. Secondary antibodies, either goat anti-rabbit or anti-mouse conjugated to either Alexa Fluor 488 or Alexa Fluor 594 (Molecular Probes) were incubated with cells for 45 minutes at room temperature. Cells were thoroughly washed, DAPI stained (Invitrogen) and mounted with Prolong Gold anti-fade reagent (Invitrogen). Microscopic observation was made on a Nikon Eclipse TE 300, PCM 2000 inverted microscope with a 100× oil-immersion objective (NA 1.4). Images were obtained with an air-cooled charge-coupled device (CCD) camera (SenSys Photometrics) attached to a 0.76× coupler (Diagnostic Instruments). For microscopic analysis, Simple PCI (Compix) acquisition software was used.

The CLS of rat cyclin E was cloned into the bait vector (pBT) of the BacterioMatch II two-hybrid system (Invitrogen) using EcoRI and XhoI digestion sites. Competent BacterioMatch II validation reporter cells were then co-transformed with the pBT vector containing the CLS and a commercially prepared HeLa plasmid cDNA library (Stratagene). Screening for potential interactions was preformed according to the manufacturer's directions. Briefly, cells were plated onto M9+ His-dropout selective screening media containing 5 mM 3-amino-1,2,4-triazole (3-AT) and incubated at 37°C overnight. Verification of an interaction was carried out by patching colonies onto M9+ His-dropout dual selective screening media containing 3-AT as well as streptomycin (12.5 g/ml). Target vector was isolated by tetracycline selection, sequenced and subjected to BLAST analysis for identification. Clones of interest were then re-screened in the two-hybrid system using a pBT bait vector containing the SWNQ(A) CLS mutant to exclude any interactions not specific for a wild-type CLS.

CHO-K1 cells were seeded onto collagen-coated glass coverslips and allowed to attach overnight. Cells were then placed in media containing 4 mM hydroxyurea for 24 hours followed by transient transfection of expression vectors. At 6 hours after transfection, cells were washed and fresh media containing HU was added. At 24 hours after transfection (for a total of 48 hours of HU treatment), cells were methanol-fixed and processed for immunofluorescent staining. Cells were immunostained for centrosomes using γ-tubulin and MCM5 or HA, depending on the expressed protein.

A vector encoding the C-terminus of AKAP450 was kindly provided by Sean Munro (University of Cambridge, Cambridge, UK). The PACT domain of AKAP450 (aa 3699-3790) was amplified by standard PCR techniques and subcloned upstream and in-frame with cyclin E constructs. All constructs were verified by PCR analysis and DNA sequencing.

We thank Eran Silverman for early contributions to the bacterial two-hybrid screen for CLS-interacting proteins. We thank Sean Munro for the PACT cDNA construct and Jillian Zhang (Weill Medical College, Cornell University) for the HA-MCM5 cDNA construct. Additionally, we thank Robert Sclafani, Frank Eckerdt and Gaetan Pascreau for valuable discussions. This work was supported by the Howard Hughes Medical Institute. J.L.M. is an Investigator of the Howard Hughes Medical Institute.