Cyclin‐A–CDK2‐mediated phosphorylation of CIZ1 blocks replisome formation and initiation of mammalian DNA replication

Eukaryotic DNA replication is initiated at multiple replication origins that fire in a temporally and spatially regulated process. Multiple layers of regulation ensure that the entire genome is duplicated completely and only once in each S phase. The initiation proteins and the principles of their regulation by the combined activities of cyclin dependent kinases (CDKs) and Dbf4–Cdc7, also referred to as Dbf4‐dependent kinase (DDK), are conserved among eukaryotes (Araki, 2010; Labib, 2010). Sites of initiation are specified by the pre‐replication complex (pre‐RC), which includes ORC, CDC6 and CDT1, leading to recruitment of the replicative helicase complex containing MCM2–7, CDC45 and GINS (CMG complex) (Ilves et al., 2010; Pacek et al., 2006). DDK activates the CMG complex leading to localised DNA unwinding (Heller et al., 2011), and in yeast this is followed by CDK‐mediated phosphorylation of GINS subunits SLD2 and SLD3, recruitment of DPB11 and polymerase ε to form the pre‐initiation complex (pre‐IC), and activation of MCM helicase activity (Heller et al., 2011; Muramatsu et al., 2010; Tanaka et al., 2007; Zegerman and Diffley, 2007). Subsequent recruitment of PCNA and polymerases α and δ complete the replisome and enable initiation of DNA replication.

In mammals, replication origin licensing and initiation of DNA replication require the sequential activities of cyclin‐E–CDK2 (cyclin E is encoded by CCNE1 and CCNE2), which supports MCM complex assembly, and cyclin‐A–CDK2, which activates DNA synthesis (Coverley et al., 2002; Geng et al., 2007; Geng et al., 2003). Importantly, cyclin‐A–CDK2 plays additional inhibitory roles at lower and higher concentrations than that which activates DNA synthesis (Copeland et al., 2010; Coverley et al., 2002), illustrating tight control over initiation and implying a complicated array of substrates. In recent years, studies in mammalian cells have identified replication factors that have evolved with little homology to the ‘core’ replication proteins identified in yeasts, yet are required as drivers of the initiation process. These include CIP1‐interacting zinc finger protein 1 (CIZ1) and geminin coiled‐coil containing protein 1 (GEMC1, also known as GMNC), which acts under the regulation of cyclin‐E–CDK2 to promote TopBP1‐mediated recruitment of CDC45 to chromatin (Balestrini et al., 2010). We identified the DNA replication role of CIZ1 using a cell‐free system that reconstitutes spatially organized replication of chromatin, using isolated intact nuclei as a template (Coverley et al., 2005). Consistent with a role in DNA replication at the interface with its nuclear organisation, CIZ1 is normally immobilized in the nucleus after removal of chromatin, soluble proteins and membranes (Ainscough et al., 2007), indicating association with an underlying nuclear matrix. A dysfunctional nuclear matrix is thought to be a defining feature of cancer cells (Munkley et al., 2011; Zink et al., 2004), and aberrant nuclear architecture is one of the most overt visible indicators of pathogenesis. Related to this, aberrant expression of alternatively spliced CIZ1 transcripts is now implicated in a range of adult and paediatric cancers (den Hollander et al., 2006; Higgins et al., 2012; Rahman et al., 2007; Rahman et al., 2010; Warder and Keherly, 2003), in some cases specifying altered sub‐nuclear localisation. In addition, CIZ1 has been associated with the apparently unrelated human disorders Alzheimer's disease and cervical dystonia (Dahmcke et al., 2008; Xiao et al., 2012).

CIZ1 interacts with cyclin E or cyclin A through distinct sites, in a two‐step process that reflects the order of cyclin expression in late G1 phase (Copeland et al., 2010). This work suggests that CIZ1 acts to temporally and spatially coordinate cyclin function during the assembly and activation of the DNA replication machinery. Interaction with cyclin‐A–CDK2 is through a conserved cyclin‐binding motif at Cy‐ii (K321) in CIZ1, and mutation of this site prevents both cyclin‐A–CDK2 interaction and CIZ1‐mediated activation of DNA replication (Copeland et al., 2010). Previous results also hint at a complex regulatory relationship between CIZ1 and cyclin‐A–CDK2. Titration of recombinant cyclin‐A–CDK2 into cytosolic extracts normally elicits a biphasic response in DNA replication assays, promoting initiation within a narrow range of kinase concentrations (Copeland et al., 2010; Coverley et al., 2002); however, when supplemented with recombinant CIZ1, this range is extended both below and above the peak concentration (Coverley et al., 2005). Moreover, when assayed at the peak active concentration of cyclin‐A–CDK2, recombinant CIZ1 also elicits a biphasic response but this is extended by mutation of a putative CDK phosphorylation site (Coverley et al., 2005). Here, we show that CIZ1 is a direct substrate for cyclin‐A–CDK2, and we identify sites for which phosphorylation downregulates the functional interaction, impacting on both cyclin A delivery to the nuclear matrix and the formation of a functional replisome.

There are 14 putative CDK phosphorylation sites (S/TP) in the full‐length (845‐amino‐acid) form of murine CIZ1, specified by the reference sequence NP082688 (accessed 24 April 2014). Six are within the minimal functional replication domain referred to as N471 (Fig. 1A) and are highly conserved in mammals (supplementary material Fig. S1). None of these are specified by sequences that have been reported to be alternatively spliced in disease contexts (Dahmcke et al., 2008; den Hollander et al., 2006; Higgins et al., 2012; Rahman et al., 2007; Rahman et al., 2010; Warder and Keherly, 2003; Xiao et al., 2012). To investigate how phosphorylation might regulate the function of CIZ1 in initiation of DNA replication, phosphomimetic aspartate mutations in each of the six sites were analysed in cell‐free DNA replication experiments. When incubated in mid G1 phase extracts, 18% of late G1 phase template nuclei incorporated labelled nucleotides, indicating the fraction that was in S phase prior to isolation and establishing the background of this assay. Parallel incubation in S phase extract increased the fraction to 30% (by inducing licensed nuclei to enter S phase), and this was increased further to 44% by recombinant CIZ1‐N471 (Fig. 1B). This illustrates the positive influence of CIZ1 on initiation, and defines the fraction of nuclei in this population that is capable of initiation in vitro. Analysis of the phosphomimetic mutant proteins revealed that three of the six putative CDK phosphorylation sites in N471 suppress initiation of DNA replication in vitro (Fig. 1C). At the optimal concentration (0.05 nM), mutation of threonine residues at positions 144, 192 and 293 reduced activity relative to wild‐type N471, and this was confirmed over a broad concentration range (0.02–0.5 nM, Fig. 1D). Importantly, phosphomimetic mutation at each of these sites not only blocked the ability of N471 to boost initiation, it also suppressed initiation in nuclei that would normally respond to S phase extract. This contrasts with individual alanine mutations at these sites, all of which promoted initiation of DNA replication to a similar degree as wild‐type CIZ1 (supplementary material Fig. S2).

To establish whether phosphorylation at multiple sites has a cooperative effect, N471 with three phosphomimetic mutations (T144D, T192D and T293D; N471‐DDD) was tested under the same conditions as the single mutants (Fig. 1D). As expected, N471‐DDD failed to promote initiation and, in fact, suppressed initiation to a greater extent than any of the single mutations. The data clearly demonstrate that phosphomimetic CIZ1 mutants reduce initiation of DNA replication in licensed nuclei and identify a dominant inhibitory effect.

In vitro kinase assays confirmed that CIZ1 is a substrate for recombinant cyclin‐A–CDK2 and cyclin‐E–CDK2 in vitro (supplementary material Fig. S3A). However, as the interaction between CIZ1 and these two cyclins appears to be sequential in vivo (Copeland et al., 2010), we reasoned that it is unlikely that cyclin‐E–CDK2 would mediate inhibitory phosphorylation events that suppress interaction with cyclin A, and therefore the ability of CIZ1 to promote initiation. For this reason, and the likelihood that site specificity would not be faithfully recapitulated in a two‐component purified system, in vitro kinase assays were focussed on cyclin‐A–CDK2 only. This generated kinetic analysis of phosphorylation of CIZ1 species with alanine mutations at each of the six putative CDK sites (Fig. 2A,B), providing insight into the relationship between them. This revealed two distinct effects – a low‐level reduction in phosphorylation rate relative to that of the wild type for alanine mutations at T138, T144 and T187, and a more severe reduction for alanine mutations in the C‐terminal residues T192, T293 and S331. The data imply a hierarchy of modifications in which phosphorylation at all of the C‐terminal sites facilities phosphorylation at one or more of the N‐terminal sites. Efficient phosphorylation does not distinguish between functional and non‐functional sites, where ‘function’ is defined by reduced initiation activity of phosphomimetic equivalents, as only two of the three high impact sites downregulate activity in replication assays. Lack of impact of phosphomimetic S331 in replication assays (Fig. 1C) does not contradict this interpretation, but argues that phosphorylation at this site alone is not sufficient.

Consistent with a fundamental regulatory role for the threonine residue at T293, proteomic analysis of wild‐type CIZ1 showed that T293 is phosphorylated by cyclin‐A–CDK2 in vitro (supplementary material Fig. S3B), but phosphorylation(s) at other sites were not detected in this system.

To begin to dissect the role of phosphorylation in regulation of CIZ1 function, protein interaction studies were performed. Direct interaction with both endogenous cyclin A (Fig. 2C) and recombinant cyclin‐A–CDK2 (Fig. 2D) was confirmed using CIZ1‐N471 immobilized by a GST tag. Consistent with the idea that CDK phosphorylation reduces the stability of the cyclin‐A–CDK2–CIZ1 ternary complex, the amount of cyclin A recovered through interaction with CIZ1‐N471 was modulated by titration of ATP into the reaction, and stabilized by the presence of the CDK inhibitor roscovitine (Fig. 2D). However, single phosphomimetic mutation of N471 did not significantly weaken the CIZ1–cyclin‐A complex (supplementary material Fig. S3C), requiring instead phosphomimetic mutation of all three of the functionally implicated regulatory sites in triple mutant CIZ1‐N471‐DDD (Fig. 2E). This shows that phosphorylation of CIZ1 at multiple sites is required to regulate interaction with cyclin‐A–CDK2, potentially modulating initiation of DNA replication by influencing the targeting of cyclin‐A–CDK2 to chromatin (Copeland et al., 2010).

Direct interaction between CIZ1 and cyclin A requires motifs within the 80 amino acids encoded by the active N471 fragment of CIZ1, but absent from the inactive fragment N391 (Fig. 2A) (Copeland et al., 2010). In contrast, both N471 and N391 interact with endogenous CDC6 in S phase HeLa cell cytosolic extract, whereas neither recovered MCM2 or PCNA under the same conditions (Fig. 2F). Moreover, direct interaction between N471 and recombinant CDC6 was not significantly impaired by individual phosphomimetic mutations (Fig. 2G) or by triple mutant N471‐DDD (Fig. 2H), a conclusion supported by reciprocal binding reactions in which GST–CDC6 was used as bait for CIZ1 (Fig. 2I). This shows for the first time direct interaction between CIZ1 and CDC6 but, more importantly, it shows contrasting interaction profiles, which suggest that CDC6 and cyclin A associate with CIZ1 through distinct sites and are not regulated in the same way. Thus, CIZ1 could associate with licensed replication origins through CDC6, independently of CIZ1 phosphorylation status.

The majority of the CIZ1 protein in NIH3T3 cells exists in nuclear foci that are immobilized on a DNase‐I‐ and high‐salt‐resistant nuclear matrix (Ainscough et al., 2007). Therefore, direct interaction with CDC6 raises the possibility that CIZ1 might play a role in the recruitment of DNA replication complexes to the nuclear matrix, where replication proteins and their regulators are believed to function in mammalian cells, and initiation and elongation take place (Anachkova et al., 2005; Djeliova et al., 2001; Pospelov et al., 1982). Neither of the CIZ1 fragments N471 or N391 encode nuclear‐matrix‐binding domains; therefore, in order to probe the effects of CIZ1 phosphorylation on immobilization and function in cell‐based assays, we used full‐length protein incorporating nuclear‐matrix‐binding sequences, with and without mutation at the three key sites. When transfected into cycling NIH3T3 cells (Fig. 3A), unphosphorylatable GFP–CIZ1‐AAA boosts the number of cells engaged in DNA synthesis (P<0.001), whereas GFP alone (control) has no significant effect on DNA replication, compared with that of untransfected cells. Importantly, phosphomimetic GFP–CIZ1‐DDD reduced the number of replicating cells to below the level in untransfected populations (P = 0.013), consistent with a dominant‐negative phenotype that inhibits DNA replication at licensed origins.

Wild‐type GFP–CIZ1, and also AAA and DDD derivatives, were detected in the nuclear matrix fraction after DNase I digestion and elution of chromatin and associated proteins (Fig. 3B), localised in subnuclear foci (Fig. 3C). We conclude therefore that CIZ1 phosphorylation does not regulate its function in DNA replication by affecting its immobilization on the nuclear matrix.

To evaluate the effect on cyclin A recruitment, asynchronous 3T3 cells were transfected with GFP–CIZ1 or derived DDD and AAA mutants, and nuclear matrix fractions were isolated by DNase I extraction of chromatin. Endogenous cyclin A was dramatically reduced in the nuclear matrix fraction from cells transfected with GFP–CIZ1‐DDD, suggesting that phosphorylated CIZ1 recruits cyclin A much less efficiently than wild‐type or AAA does (Fig. 3D). Moreover, the low‐level residual endogenous cyclin A protein that was detected by immunofluorescence in chromatin‐depleted GFP–CIZ1‐DDD transfected nuclei failed to colocalise with recombinant GFP‐tagged protein (Fig. 3E). In contrast, colocalization was apparent in the chromatin‐depleted nuclei population transfected with GFP–CIZ1 or GFP–CIZ1‐AAA. Analysis of the extent of colocalisation with cyclin A in populations of nuclei is shown in supplementary material Fig. S4, revealing very low levels of overlap with CIZ1‐DDD that are significantly different (P<0.01) to the overlap with CIZ1 or CIZ1‐AAA. The data support the premise that CIZ1 facilitates cyclin A recruitment to the nuclear matrix, but only when in a hypophosphorylated state.

An affinity‐purified phosphospecific antibody, raised against a phosphorylated peptide encompassing position phospho‐T293, was validated using in vitro kinase and phosphatase treatment of recombinant CIZ1 and derived mutants (supplementary material Fig. S3D). Western blot analysis of detergent‐insoluble nuclear protein (chromatin and nuclear matrix) harvested from chemically synchronised 3T3 cells (Fig. 4A) shows that CIZ1 phospho‐T293 is evident in S phase, peaks during G2 and is lost in nocodazole‐arrested cells, although CIZ1 protein itself remains present (Fig. 4B). Similar results were seen when 3T3 cells were synchronised by contact inhibition and serum depletion, then induced to re‐enter the cell cycle (Fig. 4C). Progression through G1 and S phase was monitored by EdU incorporation (Fig. 4D), indicating entry to S phase between 16 and 22 hours, whereas CIZ1 phospho‐T293 is evident from 20 hours and accumulates through S phase. Cell‐cycle‐dependent phosphorylation in intact cells synchronized by this physiologically relevant method is strong evidence that CIZ1 activity is normally regulated by phosphorylation at this site.

The data suggest that CIZ1 promotes DNA replication in a hypo‐phosphorylated state and inhibits DNA replication in a hyper‐phosphorylated state. Consistent with this, using a cell‐free assay under conditions that induce a biphasic response to cyclin‐A–CDK2 in isolated G1 nuclei (Copeland et al., 2010; Coverley et al., 2002), we saw no increase in the phosphorylation of endogenous CIZ1 at T293 at the activating concentration of cyclin‐A–CDK2 (0.1 ng/µl, Fig. 4E,F), indicating that initiation of DNA replication is not accompanied by detectable phosphorylation at this site. In contrast, a tenfold higher concentration of cyclin‐A–CDK2 (1 ng/µl), which fails to activate DNA synthesis (Fig. 4E), did phosphorylate CIZ1 at T293 (Fig. 4F), supporting the idea that phosphorylation of CIZ1 occurs after initiation and plays a role in suppressing re‐initiation.

To look at how phosphorylation of CIZ1 influences the DNA replication machinery, we monitored the recruitment of selected replication factors in 3T3 cells synchronised by release from quiescence, after transfection with phosphomimetic GFP–CIZ1N471‐DDD or unphosphorylatable GFP–CIZ1N471‐AAA (Fig. 5A). Cells entered S phase with tight synchrony after 16 hours for control and transfected populations, except those expressing DDD (Fig. 5B). This reinforces results with unsynchronised cells and with isolated nuclei, and shows that DDD has a potent dominant‐negative effect on initiation. Western blot analysis at 24 hours after release from quiescence shows that recombinant CIZ1 is expressed at similar levels for all three constructs, and that they are present in the detergent‐resistant (chromatin/nuclear matrix‐associated) fraction at similar levels (Fig. 5C). CDC6, MCM2 and CDT1 levels are similar, suggesting that pre‐RC formation is unperturbed and therefore not responsible for the observed reduction in DNA replication. CDC45 levels are also similar, indicating that its recruitment, and by implication that of MCM2–7 and GINS (the CMG complex), is also not blocked. However, consistent with weakened cyclin A interaction in vitro (Fig. 2) and loss of cyclin A from the nuclear matrix fraction (Fig. 3), there is a marked decrease in cyclin A recruitment in the presence of GFP–CIZ1N471‐DDD. This is accompanied by failure to complete replisome assembly as evidenced by reduced PCNA recruitment. This suggests that cyclin‐A–CDK2 positively regulates late steps in the transition from G1 to S phase, and implicates CIZ1 in the execution of this process.

Cell cycle progression is governed by cyclin‐dependent protein kinases as they reach levels that exceed thresholds required to phosphorylate their substrates. Temporal separation of replication licensing and activation of DNA synthesis prevents replicated DNA from re‐establishing licensed origins (Blow and Dutta, 2005) and is controlled by both positive and negative phosphorylation events. Within this process, an array of substrates have been documented; MCM complex assembly in G1 phase is dependent upon cyclin‐E–CDK2 (Coverley et al., 2002; Geng et al., 2007; Mailand and Diffley, 2005), and recruitment of GEMC1 and the candidate functional orthologues of lower eukaryote proteins SLD2 (an orthologue of mammalian RecQ4) and SLD3 (an orthologue of mammalian treslin) is promoted by CDK activity (Abe et al., 2011; Balestrini et al., 2010; Kumagai et al., 2010; Kumagai et al., 2011; Piergiovanni and Costanzo, 2010; Sangrithi et al., 2005). The proteins for which phosphorylation promotes and inhibits DNA replication in mammalian cells are still to be fully elucidated, but some key events have been found to correlate with defined kinase activity thresholds (Coudreuse and Nurse, 2010; Spencer et al., 2013). In the case of cyclin‐A–CDK2, distinct positive and negative events can be ordered along a concentration gradient that reflects increasing endogenous kinase levels in late G1 phase – lower concentrations switch off pre‐RC assembly by phosphorylating and inactivating CDC6, higher concentrations activate DNA synthesis, and still higher concentrations phosphorylate MCM2 and are associated with loss of ability to promote initiation (Coverley et al., 2002). This is consistent with studies of DNA polymerase α, where CDK‐mediated phosphorylation promotes initiation at low levels, but inhibits activity at high concentrations (Schub et al., 2001; Voitenleitner et al., 1997; Voitenleitner et al., 1999). Thus, cyclin‐A–CDK2 both activates and inhibits initiation by regulating components of the pre‐RC and the replisome. However, this level of analysis is not capable of reporting on localized environments within the highly organized nucleus of higher eukaryotes, because global levels measured in cell lysates mask localised peaks and troughs established by immobilization of kinase at sites of action.

In fact, individual replication factories might experience changes in local kinase concentration in an asynchronous manner, guided by delivery or receptor proteins that spatially limit kinase‐regulated events. We find that CIZ1 is both a delivery factor that mediates the nuclear matrix recruitment of cyclin‐A–CDK2, and is itself phosphorylated by cyclin‐A–CDK2 at multiple sites that inhibit its function in DNA replication. Notably, direct interaction with the pre‐RC protein CDC6 is unaffected by CIZ1 phosphorylation. This raises the possibility that CDC6 is a receptor for CIZ1, and their interaction brings licensed origins and the nuclear matrix into close proximity. CDC6 was previously suggested to be a chromatin‐associated receptor for cyclin E (Furstenthal et al., 2001), which is also a direct interaction partner of CIZ1 (Copeland et al., 2010). However, this earlier study did not apply nuclease digestion techniques and so was not able to distinguish between recruitment to chromatin verses the nuclear matrix. Nevertheless, when considered alongside our previous work, which showed that CIZ1 interacts sequentially with cyclins E and A (Copeland et al., 2010), the data suggest that CIZ1 (and cyclin E) might cooperate with CDC6 to bring activating kinases to replication origins.

Based on the data presented here, which shows that CIZ1 mediates the regulated recruitment of cyclin A to the nuclear matrix, and independent data sets that identify nuclear‐matrix‐binding sites in the C‐terminal part of CIZ1 (Ainscough et al., 2007), we suggest that CIZ1 contributes nuclear matrix localization to the events leading to initiation. On the assumption that both cyclin A interaction and nuclear matrix interaction domains are present and active within the same molecule, our integrated model proposes the following. In late G1 phase when kinase levels are low, CIZ1 exists in a hypophosphorylated state that is permissive for interaction with cyclin‐A–CDK2 and localization of cyclin‐A–CDK2 activity to nuclear‐matrix‐associated sites of initiation (Fig. 6). As kinase levels rise during early S phase, cyclin‐A–CDK2 phosphorylates CIZ1 at three sites (T144, T192 and T293), inactivating its DNA‐replication‐promoting activity by preventing further interaction with cyclin‐A–CDK2. This describes a negative‐feedback loop that could limit cyclin‐A–CDK2 recruitment to origins to within a narrow local range of CDK activity. Thus, CIZ1 appears to function as a nuclear‐matrix‐associated kinase sensor, which could influence replisome formation independently at individual sites.

Mouse 3T3 and HeLa cells were cultured in Dulbecco's modified Eagles Media (D‐MEM) containing 10% (v/v) fetal calf serum (FCS) and penicillin‐streptomycin‐glutamine (all Gibco). Synchronised populations of 3T3 cells were generated by contact inhibition and serum starvation in G₀ followed by release into fresh medium prior to the isolation of nuclei or cytosolic extracts for cell‐free analysis, or chromatin for analysis of CIZ1 phosphorylation (Coverley et al., 2002). S phase and G2 cells were produced using a double thymidine block (2.5 mM) followed by release into fresh medium for 1 hour or 10 hours, respectively, and G2/M phase cells were produced by addition of 0.02 µg/ml nocodazole to S phase cells for 12 hours prior to isolation. For transfection of asynchronous populations of 3T3 cells, 15‐cm dishes at ∼80% confluence were transfected with 5 µl of vector as indicated (2.5 µg) using nucleofector kit R (Lonza) and cultured for the indicated times. Where indicated, synchronous cells were transfected after contact inhibition and serum starvation in quiescence with 5 µl of vector as indicated (2.5 µg) using nucleofector kit R (Lonza), plated into fresh medium to re‐enter the cell cycle and cultured for up to 24 hours after release.

Growing cells were harvested, fixed overnight in 20% ethanol at −20°C and stained with 100 µg/ml propidium iodide in PBS containing 0.5% (v/v) Triton X‐100. Cell cycle profiles were determined using a BD FACScanto flow cytometer and FACSDiva software. Data were collected for 10,000 cells with consistent gating applied for all samples.

Cytosolic extracts were prepared from early S phase HeLa cells by Dounce homogenisation in hypotonic buffer (10 mM HEPES pH 7.8, 5 mM potassium acetate, 0.5 mM MgCl₂, 1 mM DTT) followed by centrifugation to remove insoluble material (Krude et al., 1997). G1 nuclei were isolated at 17 hours after release from quiescence and used as substrates in cell‐free DNA replication assays. Evaluation of the composition of nuclei populations was described previously, and consists of (1) nuclei that incorporate labelled nucleotides in the absence of CDK (in a pre‐restriction point early G1 phase cytosolic extract, isolated at 15 hours after release) and which were in S phase at the time of isolation, (2) nuclei that respond to CDK (supplied by S phase extract or recombinant kinase) and which are ‘licensed’ late G1 phase nuclei, and (3) those that do not respond and are therefore not licensed (early G1 or G0). For each independent experiment, at least 100 nuclei were scored for the presence of fluorescent DNA replication foci and expressed as a percentage of the total population. All experiments were performed with n≥3, as indicated. Calculation of the percentage initiation in the licensed fraction was as described previously (Coverley et al., 2005).

Cells were washed twice in cold PBS, then cytoskeleton (CSK) buffer [10 mM PIPES pH 6.8, 100 mM NaCl, 300 mM sucrose, 1 mM MgCl₂, 1 mM DTT, with EDTA‐free complete protease inhibitors (Roche)]. Detergent‐insoluble fractions were prepared by inclusion of Triton X‐100 (0.1% v/v) in CSK buffer and incubation on ice for 5 minutes, followed by centrifugation at 6000 g for 5 minutes. Further inclusion of 0.5 M NaCl followed by centrifugation generates a high‐salt‐resistant fraction. The nuclear matrix fraction was obtained by washing the high‐salt‐resistant pellet twice in DNase I buffer (10 mM Tris‐HCl pH 7.4, 10 mM MgCl₂, 5 mM CaCl₂) before digestion with RNase‐free DNase I (Roche) at 37°C for 1 hour, and a further wash with 0.5 M NaCl in CSK buffer (Munkley et al., 2011). For immunofluorescence analysis, cells were treated on coverslips and DNase I efficiency was monitored by staining residual nuclei with Hoechst 33258. For interaction studies with endogenous CDC6, S3 HeLa cells were resuspended in buffer A [50 mM HEPES‐KOH pH 7.8, 0.5 mM ATP, 5 mM MgCl₂, 10 mM CaCl₂, 1 mM EDTA, 1 mM EGTA, 5 mM magnesium acetate, 10% (v/v) glycerol, 1 mM DTT and 1×Halt protease inhibitors (Pierce)]. Cells were Dounce homogenised, and nuclei were enriched by centrifugation at 6000 g, then resuspended in two volumes of the same buffer supplemented with 450 mM KCl and 0.02% (v/v) NP40. Soluble extract was removed after centrifugation at 10,000 g, diluted in three volumes of buffer A and incubated with GST–CIZ1‐N471, GST–CIZ1‐N391 or GST alone, immobilised on glutathione–Sepharose, as indicated, for 1 hour at 4°C. Beads were washed five times, in ten times bead volume buffer A containing 100 mM KCl and 0.02% (v/v) NP40, and analysed by western blotting.

Cells on coverslips were extracted as described previously and immunostained to reveal chromatin and nuclear matrix or just matrix‐associated proteins (Munkley et al., 2011). BrdU (GE Bioscience) or EdU Click IT reactions (Invitrogen) were used as recommended to identify replicating cells. Colocalisation experiments using cells expressing GFP–CIZ1 constructs were performed using a Zeiss 510 confocal microscope, with 1 Airy unit pin‐hole for all fluorophores. Images were captured using multitrack mode, setting imaging parameters for GFP–CIZ1 using palette range finder to set parameters that were applied for all images. Post acquisition, images were prepared using Adobe Photoshop CS5 and colocalisation was determined using BioImage XD (http://www.openbioimage.org). Colocalisation was determined using constant parameters for all images. Percentage colocalisation refers to the total percentage of CIZ1 and cyclin A colocalised pixels in each image. Population data for each construct was used to generate a box and whisker plot using SPSS statistics 21, showing the distribution of the extent of colocalisation for each construct.

Monoclonal antibodies against cyclin A (CY‐1A, Sigma Aldrich), actin (AC15, Sigma Aldrich), PCNA (sc56, Santa Cruz Biotechnology), CDC6 (sc‐9964), anti‐BM28 (Becton Dickinson) and lamin A (sc‐20680), and polyclonal antibodies against CDC45 (Santa Cruz Biotechnology, sc‐20685), histone H3 (Abcam, ab1791) and CDT1 (Santa Cruz Biotechnology, sc‐28262) were used as directed for western blotting and/or immunofluorescence. CIZ1 was detected with anti‐CIZ1 polyclonal antibody 1793 (Coverley et al., 2005) or a rabbit polyclonal antibody raised against purified bacterially expressed recombinant ECIZ1‐N471 for this study (Covalab) (Fig. 4 only). For western blots, goat anti‐mouse‐IgG conjugated to peroxidase (POD) and goat anti‐rabbit‐IgG–POD (both Sigma Aldrich) were used with an ECL detection system (GE Lifescience). For immunofluorescence, Alexa‐Fluor®‐488‐conjugated goat anti‐mouse‐IgG (H⁺L) (A10667) and Alexa‐Fluor®‐568‐conjugated goat anti‐rabbit‐IgG (H⁺L) (A11036) (both Invitrogen) were used. Phosphospecific anti‐CIZ1 phospho‐T293 (anti‐pT293) was produced by immunisation of rabbits with TAPKQTQTpTPDRL peptide, where pT indicates phosphothreonine (Covalabs). Antibodies were purified by sequential subtraction on unphosphorylated peptide columns and enrichment on a phosphopeptide column, to purify phosphospecific anti‐CIZ1 antibodies.

Single amino acid substitutions were created with the Stratagene QuikChange method, using pGEX‐ECIZ1 N471 or pEGFP‐ECIZ1 as templates (Copeland et al., 2010). For multiple mutations, each site was mutated in separate reactions and all vectors were sequence verified (Eurofins).

All CIZ1 fragments and derived mutants were expressed as GST fusions in pGEX6P3 in Escherichia coli Rosetta (DE3) and purified as described previously (Coverley et al., 2002; Coverley et al., 2005). GST–CDC6 was expressed in BL21 (DE3) E. coli and the tag was removed by digestion with PreScission protease (GE). For binding studies, proteins were immobilised on glutathione beads (as indicated) incubated with cell extracts or purified recombinant proteins for 1 hour at 4°C in binding buffer (10 mM HEPES pH 7.8, 135 mM KCl, 20 mM MgCl₂, 10 mM CaCl₂, 0.04% NP40, 1 mM DTT, with EDTA‐free complete protease inhibitor cocktail). ATP (Sigma Aldrich) and/or 100 µM roscovitine (Cell Signaling Technology/New England Biolabs) were included as indicated. Beads were washed five times in cold binding buffer and analysed by western blotting.

Assays were performed in 50 mM HEPES KOH pH 7.8, 20 mM MgCl₂ 10 mM ATP, 1.5 µl [γ‐³³P]ATP (Perkin Elmer), 1 mM DTT. CIZ1 was used at 1 µM, and cyclin‐A–CDK2 or cyclin‐E–CDK2 were used at 50 nM under pseudo‐first conditions. For samples treated with λ protein phosphatase (P0753S, New England Biolabs), reactions were performed according to the manufacturer's instructions. Samples were separated by SDS‐PAGE, and gels were dried and developed on CL‐XPosure Film (Thermofisher). Densitometry was performed using NIH ImageJ.

Significant differences in DNA replication activity and colocalisation analysis were performed by one‐way analysis of variance (ANOVA) followed by Tukey's test post‐hoc using IBM SPSS statistics 21. All significant results are shown in figures as appropriate. ***P<0.001, **P<0.01 and *P<0.05.

In vitro kinase assays were performed using cyclin‐A–CDK2 and CIZ1‐N471 fragments and analysed by SDS‐PAGE, and in‐gel proteolytic digestion was performed after reduction and S‐carbamidomethylation with iodoacetamide. Gel pieces were rehydrated in 25 mM ammonium bicarbonate and digested with sequencing‐grade modified porcine trypsin at 0.004 µg/µl (Promega) overnight at 37°C. Eluted peptides were vacuum desiccated and resuspended in 1 M glycolic acid in 80% aqueous acetonitrile (v/v), 5% trifluoroacetic acid (v/v). Phosphopeptides were selectively enriched using manually packed titanium microcolumns (Larsen et al., 2005). Enriched phosphopeptides were purified by Poros Oligo R3 reversed‐phase a material (Larsen et al., 2005) and eluted directly onto a ground steel MALDI target plate with a freshly prepared 10 mg/ml solution of 2,5‐dihydroxybenzoic acid (Sigma‐Aldrich) in 50% aqueous acetonitrile (v/v), 1% phosphoric acid (v/v).

Positive‐ion MALDI mass spectra were obtained using an ultraflex III mass spectrometer (Bruker, Bramen, Germany) in reflectron mode, equipped with a Nd∶YAG smart beam laser. Mass spectrometry (MS) spectra were acquired over a mass range of 800–4000 m/z. Final mass spectra were externally calibrated against an adjacent spot containing six peptides of known m/z [des‐Arg¹‐bradykinin, 904.681; angiotensin I, 1296.685; Glu¹‐fibrinopeptide B, 1750.677; ACTH (1–17 clip), 2093.086; ACTH (18–39 clip), 2465.198; ACTH (7–38 clip), 3657.929]. Peptides were manually selected for MS/MS fragmentation, which was performed in LIFT mode without the introduction of a collision gas. Fragmentation spectra were acquired manually with laser intensity and number of summed spectra optimised for each acquisition. The default calibration was used for MS/MS spectra, which were baseline‐subtracted and smoothed (Savitsky‐Golay, width 0.15 m/z, cycles 4); monoisotopic peak detection used a SNAP averaging algorithm (C 4.9384, N 1.3577, O 1.4773, S 0.0417, H 7.7583) with a minimum S/N of 6. Bruker flexAnalysis software (version 3.3) was used to perform the spectral processing and peak list generation for both the MS and MS/MS spectra. Tandem mass spectral data were submitted to database searching using a locally running copy of the Mascot program (Matrix Science Ltd, version 2.3), through the Bruker ProteinScape interface (version 2.1). All spectral data were searched against an in‐house database containing the expected protein sequence. Search criteria specified: Enzyme, trypsin; Fixed modifications, carbamidomethyl (C); Variable modifications, oxidation (M) and phosphorylation (S,T); Peptide tolerance, 250 ppm; MS/MS tolerance, 0.5 Da; Instrument, MALDI‐TOF‐TOF.

We thank Adam Dowle and Jerry Thomas for technical support with MS and analysis. We are grateful to Clive Price, Emma Hesketh and Justin Ainscough for critical comments on the manuscript.