A kinase-independent function for AURORA-A in replisome assembly during DNA replication initiation

Abstract The catalytic activity of human AURORA-A kinase (AURKA) regulates mitotic progression, and its frequent overexpression in major forms of epithelial cancer is associated with aneuploidy and carcinogenesis. Here, we report an unexpected, kinase-independent function for AURKA in DNA replication initiation whose inhibition through a class of allosteric inhibitors opens avenues for cancer therapy. We show that genetic depletion of AURKA, or its inhibition by allosteric but not catalytic inhibitors, blocks the G1-S cell cycle transition. A catalytically inactive AURKA mutant suffices to overcome this block. We identify a multiprotein complex between AURKA and the replisome components MCM7, WDHD1 and POLD1 formed during G1, and demonstrate that allosteric but not catalytic inhibitors prevent the chromatin assembly of functional replisomes. Indeed, allosteric but not catalytic AURKA inhibitors sensitize cancer cells to inhibition of the CDC7 kinase subunit of the replication-initiating factor DDK. Thus, our findings define a mechanism essential for replisome assembly during DNA replication initiation that is vulnerable to inhibition as combination therapy in cancer.


INTRODUCTION
Human cancers originating in many different tissues frequently amplify or overexpress the AURKA gene (1-3), but how the 403 amino acid protein kinase encoded by the gene promotes carcinogenesis remains unclear. Consistent with normally elevated expression during the G2 and M phases of the cell cycle, AURKA has been implicated as a key regulator of mitotic chromosome segregation through its functions in chromosome condensation, mitotic spindle assembly and the bipolar attachment of kinetochores to spindle microtubules, as well as entry into and exit from mitosis (4)(5)(6)(7)(8)(9). Furthermore, AURKA-mediated phosphorylation of the replication licensing factor geminin on Thr25 during M phase has been reported. This event induces gem-inin stabilization by preventing its APC/C ubiquitin ligase complex-mediated degradation, ensuring its persistence during the M-G1 transition (10). However, recent evidence suggests that AURKA is also expressed during the G1 and S phases of the cell cycle and in non-cycling cells, where it has been implicated in different non-mitotic processes (11), including protection of DNA forks during replication stress (12), regulation of the expression of the DNA damageresponse genes BRCA1, CHK2 or BRCA2 (13,14), the disassembly of primary cilia during cell cycle entry (15,16), or control of mitochondrial dynamics and energy production (17). Moreover, accumulating evidence suggests that AU-RKA exerts functions that are kinase-independent, as well as through its catalytic activity. For instance, defects in mitotic spindle assembly induced by AURKA depletion are rescued by a kinase-dead catalytic mutant (18), and a similar catalytically-inactive mutant is capable of transactivating transcription driven by the MYC oncogene (19). In addition to kinase activity, which is targeted by small-molecule inhibitors now in clinical use (20,21), discrete AURKA protein conformations and protein-protein interactions that also underlie its biological functions have been identified. AURKA interacts with the mitotic protein TPX2, inducing an allosteric change that activates kinase catalytic activity, which is blocked by small-molecule inhibitors of the AURKA/TPX2 interaction (22)(23)(24). Likewise, a conformation of AURKA that interacts with the N-MYC protein can be blocked by allosteric (but not catalytic) small-molecule inhibitors, suppressing N-MYC activation in neuronal cancer cells (25).
Although AURKA has been associated with geminin phosphorylation during mitosis to prevent its degradation (10), no relationship between AURKA and the replication machinery in interphase has been described to date. Here, we have deployed both catalytic and allosteric inhibitors of AURKA to reveal a previously unrecognized non-catalytic function of the protein in DNA replication. We demonstrate that AURKA is necessary for the assembly of functional replisomes during the G1-S phase transition, and that allosteric but not catalytic inhibitors prevent the chromatin loading of replication factors required for the efficient initiation of DNA replication. We also show that allosteric but not catalytic AURKA inhibitors sensitize cancer cells to inhibition of the CDC7 kinase subunit of the replicationinitiating factor DDK. Thus, our findings provide fresh insights into the mechanisms that control human DNA replication, and suggest an approach for combination therapy in cancer.

Cell lines and reagents
Parental FRT/TO HeLa (kind gift from Stephen Taylor, University of Manchester), A569 (adenocarcinomatous human alveolar basal epithelial cells) and EUFA423 (fibroblast derived from Fanconi anemia subtype D1 patients) were grown in DMEM (Gibco) medium, SW48 (colorectal adenocarcinoma cell line) was cultured in RPMI (Gibco) medium and RPE (retinal pigment epithelial cells) were cultured in DMEM/F-12 (Gibco), all the media containing 10% (v/v) fetal bovine serum (Gibco). All cells were maintained at 37 • C with 5% CO 2 .
Parental FRT/TO HeLa cells were used to generate doxycycline-inducible cell lines as described previously (26). Briefly, HeLa FRT/TO cells were transfected with a pcDNA5/FRT/TO vector encoding human AURKA wildtype or K162R (a gift from Stephen Taylor (27), Addgene plasmids 59804 and 59805) together with a plasmid encoding Flp recombinase (pOG44). Stable integrants were then selected with 200 g/ml hygromycin (InvivoGen) and 4 g/ml blasticidin (InvivoGen). Transgene expression was induced by treatment with 100 ng/ml doxycycline (Sigma).

Viability assays
Cell lines were seeded overnight at 5000 cells per well in a 96-well plate in recommended media. Compounds were added the following day and viability was measured at 72 h using the sulforhodamine B (SRB) colorimetric assay as follows: cells were fixed by replacing media with 1% (v/v) trichloroacetic acid (TCA) (Sigma) at 4 • C and incubated for 1 h also at 4 • C. Wells were washed thoroughly with deionized water before staining with SRB (Sigma) solution (0.057% w/v in 1% v/v acetic acid (VWR)) for 30 min. Stain was aspirated and wells washed with several changes of 1% (v/v) acetic acid until unbound stain was removed, before air-drying. Bound stain was solubilized with 100 l/well 10 mM Tris (pH 8.0) (Trizma Base, Sigma) with agitation for 10 min. Fluorescence intensity was recorded at Ex540 nm/Em590 nm using a BMG Pherastar Plus plate reader (28).

Cell cycle analysis
DNA content per cell was measured by flow cytometry using an LSR II (BD Biosciences) flow cytometer. Cells were fixed in 70% ethanol, centrifuged, resuspended in staining buffer (40 g/ml propidium iodide (PI) (Sigma), 250 g/ml RNAase A (Sigma), in PBS) and incubated at 37 • C in the dark for 30 min before analysis.
In order to calculate the percentage of phospho-MPM-2 positive cells, cells were fixed in 70% ethanol and then washed and resuspended in blocking solution (0.1% Triton X-100 (ThermoFisher Scientific), 1% bovine serum albumin (Sigma) in PBS) for 30 min at room temperature. Then, cells were resuspended in 500 l antibody solution (anti-phospho-Ser/Thr-Pro MPM-2 Antibody (Millipore, 05-368) in blocking solution), incubated overnight at 4 • C, washed twice with blocking solution and resuspend in 500 l of secondary antibody solution (Alexa Fluor-488 conjugate anti-Mouse IgG secondary antibody (Invitrogen, A-11029) in blocking solution) for 1 h at room temperature. Finally, cells were washed with PBS, resuspended in staining buffer (40 g/ml PI, 250 g/ml RNAase A in PBS) and incubated at room temperature in the dark for 30 min before analysis using an LSR II (BD Biosciences) flow cytometer.
For analysis of EdU incorporation, 10 M EdU (Invitrogen) was added 7 h after cells were released from a mitotic arrest, before S phase entry, and cells were harvested 1 and 2 h after EdU addition. Samples were probed using the Click-iT Plus EdU Flow Cytometry Assay Kit (Invitrogen, C10645) following manufacturers' instructions and stained with 40 g/ml PI, 250 g/ml RNAase before analysis in an LSR II flow cytometer.

Chromatin fractionation
For chromatin fractionation cells were lysed in CSK buffer (10 mM PIPES, pH 7.0, 100 mM NaCl, 300 mM sucrose, 3 mM MgCl 2 ) supplemented with 0.7% Triton X-100, EDTA-free Protease Inhibitor Cocktail and Phosphatase Inhibitors (Cocktail 2 and Cocktail 3, Sigma), and centrifuged for 10 min at 13 000 rpm. After centrifugation, the supernatant contained the soluble fraction. The pellet was washed with CSK buffer, resuspended in CSK buffer for sonication and centrifuged for 10 min at 13 000 rpm. After centrifugation, the supernatant contained the chromatinenriched fraction. Laemmli buffer was added to both fractions and samples were heated at 90 • C 5 min. Finally, samples were spun at 13 000 rpm for 1 min before processing by Western blot.

RNAi and plasmid transfections
DharmaFECT transfection reagent 1 (Dharmacon) was used for RNAi transfections according to manufacturer's instructions. Hs AURKA 4 FlexiTube RNAi (Qiagen) and AllStars Negative Control RNAi (Qiagen) were used at 25 nM final concentration. Hanks' Balanced Salt solution (Gibco) was used to dilute the RNAi and the transfection reagent when preparing the transfection mixtures. Transfection reaction mixtures were mixed thoroughly and incubated at room temperature for 10 min before adding dropwise to cells.
Lipofectamine 3000 transfection reagent (ThermoFisher Scientific) was used following manufacturer's instructions for co-transfection of a plasmid encoding Flp recombinase (pOG44) and the modified pcDNA5/FRT/TO vector containing a single amino-terminal Myc tag and the human wild type or K162R AURKA transgenes into HeLa FRT/TO cells. Plasmid concentrations and ratios were used as specified at Flp-In T-REx system (Life Technologies) user manual. Transfection reaction mixtures were mixed thoroughly and incubated at room temperature for 15 min before adding dropwise to cells and culture media was replaced with fresh media after 24 h incubation at 37 • C

Co-immunoprecipitation
HeLa FRT/TO cells were synchronised in G1 by releasing them from a mitotic arrest with nocodazole. Three hours after the release (when cells were in early G1) 2× GI 50 equivalent concentrations of compounds were added. Cells were then collected 2 h later (before cells enter into S phase), and resuspended in NETN (50 mM Tris·Cl, pH 8, 150 mM NaCl, 1 mM EDTA) lysis buffer supplemented with 0.5% NP-40, EDTA-free Protease Inhibitor Cocktail (Roche) and Phosphatase Inhibitors (Cocktail 2 and Cocktail 3, Sigma) for 30 min on ice. Samples were sonicated and centrifuged before collecting the supernatant containing the proteins. In order to immunoprecipitate AURKA, 2 mg protein were incubated with 10l magnetic beads (Dynabeads ® , Invitrogen) and 0.5 g rabbit IgG (Sigma, I5006) for 1 h at room temperature with rotation, in a final volume of 200 l lysis buffer. Then, the beads were removed and the cleared sample was incubated overnight at 4 • C with 20 l beads previously coated, following manufacturer's instructions, with 6 g AURKA antibody (Thermofisher, PA1-26351) or 6 g rabbit IgG as a control. Finally, the beads were washed four times with cold lysis buffer, resuspended in 30 l 2× Laemmli buffer and heated for 10 min at 70 • C. Samples were then incubated for 5 min at 95 • C after the addition of 1 l 1 M ␤-mercaptoethanol (Sigma). 30 g protein from the input and the supernatant containing the immunoprecipitated proteins were used for western blot analysis. Membranes were probed against WDHD1 (Abcam, ab72436), MCM7 (Santa Cruz Biotecnology, sc-9966), POLD1 (Santa Cruz Biotecnology, sc-17776) and AURKA (IAK1 anti-Aurora A, BD Biosciences, 610939).

Mass spectrometry
AURKA was immunoprecipitated from G1 synchronised cell lysates (3 and 5 h after the release from a nocodazole arrest) as described above and the samples containing the co-immunoprecipitated proteins were sent to the Biological Mass Spectrometry and Proteomics Service, at the MRC Laboratory of Molecular Biology, Cambridge, for analysis (Supplementary Table S1). Gene ontology analysis was performed using the Princeton University online service https: //go.princeton.edu/cgi-bin/GOTermFinder (29).

Allosteric but not catalytic AURKA inhibitors provoke cell cycle arrest in interphase
We first examined the cell cycle profile of cells treated with ATP-competitive (MLN8237 or VX-689) (3,(30)(31)(32)(33) or allosteric (CD532) inhibitors of AURKA (25,(34)(35)(36)(37)(38). Gustafson et al. have shown that treatment with the allosteric inhibitor CD532 resulted in N-MYC degradation and consequently, a potent loss of S phase entry in N-MYC-amplified neuroblastoma cells, which rely on N-MYC to proliferate. In contrast, AURKA kinase inhibitors (MLN8237 or VX-689) had little or no effect on N-MYC levels and, therefore, resulted in mitotic arrest as a consequence of AURKA catalytic inhibition (25). Surprisingly, we also observed different effects in HeLa cells treated with either class of AURKA inhibitor using a range of equivalent fold GI 50 concentrations (defined as the dose inducing 50% maximal inhibition of cell proliferation over time) ( Figure 1A, B, Supplementary Figure S1A Figure  S1A and D). However, allosteric AURKA inhibitors proposed to alter the structure of the kinase induced only a mild increase in the proportion of cells in G2-M, suggestive of a kinase-independent function during interphase (Figure 1B). A similar phenotype is also observed after AU-RKA depletion using RNAi (Supplementary Figure S1E), albeit with certain caveats. Cell transfection for AURKA RNAi unavoidably causes some cytotoxicity, increasing the sub-G1 population, and because it takes time to deplete AURKA protein, exerts its effects over a longer duration (24/48 h) than the treatment of cells with AURKAi compounds. Furthermore, we observe similar results in addi- Viability as a percentage normalized to control cells exposed to DMSO plotted against the log 10 concentration in nM of the indicated inhibitor. Values represent the mean ± SEM of three observations. All results are representative of at least three independent repeats. tional cell lines including A549, EUFA423, SW48 or RPE when comparing cell cycle profiles after treatment with catalytic (MLN8237) or allosteric (CD532) inhibitors, despite differences in magnitude consonant with differences in their proliferation rate, indicating that such a non-mitotic, kinase-independent function of AURKA is not cell-type specific ( Figure 1C and Supplementary Figure S1F). In these cell lines, N-MYC levels are also virtually undetectable (Supplementary Figure S1C), again suggesting that the G1 arrest observed after CD532 addition is not a consequence of N-MYC proteolysis. Moreover, exposure of cells to a low dose of etoposide sufficient to suppress mitotic entry (Supplementary Figure S1G) sensitized cells to allosteric AU-RKA inhibitors, but not to catalytic inhibitors or taxol (Figure 1D). Collectively, our findings suggest that allosteric AURKA inhibitors block a distinct non-catalytic function independent of N-MYC during interphase that is not targeted by kinase catalytic inhibitors.

AURKA is required for the G1-S phase transition
Further experiments indicate that this non-catalytic interphase function of AURKA is essential for the G1/S transition, but not for the completion of DNA replication during S phase. Neither catalytic nor allosteric AURKA inhibitors affect S phase progression in cells released synchronously into S phase via a double thymidine (Td) block ( Figure  2A). However, when cells synchronized in early G1 phase (3 h after release from a nocodazole-induced mitotic arrest, Figure 2B) are exposed to allosteric but not catalytic AU-RKA inhibitors, the G1/S transition was impeded ( Figure  2B). Consistent with the notion that allosteric inhibitors reveal an interphase function of AURKA, CD532 exerts distinct effects depending on the point of exposure to the compound during the cell cycle. We exposed cells synchronized in mitosis and released into the cell cycle, to a single dose (2× GI 50 ) of CD532 or MLN8237, either in early G1 at 3 h after release (Supplementary Figure S2A Figure S2A). Moreover, treatment with CD532 in early G1 causes increased CHK1 phosphorylation (Supplementary Figure S2D) suggestive of replication stress, although the absence of functional p53 in HeLa cells discounts checkpoint activation as a cause of G1/S arrest. Collectively, these findings highlight the distinct effects of allosteric versus catalytic AURKA inhibition, and further strengthen our conclusions.
A similar effect to that observed after CD532 addition, that is, a impediment in the G1/S phase transition or sensitization of HeLa cells to a low dose of etoposide sufficient to suppress mitotic entry, also occurs when cells are treated with AurkinA, a small-molecule inhibitor of the AURKA/TPX2 interaction that has been structurally and biochemically validated but whose cellular potency is under optimization ((23), Supplemental Figure S2E-I). That two allosteric inhibitors with significantly different chemical structures reveal such a function suggests that it does not arise from off-target effects.
Indeed, genetic experiments provide further evidence supporting that such a function does not arise from offtarget effects. Experiments to test the function of AURKA during the G1/S transition proved technically challenging. AURKA is essential for the completion of mitosis, and currently available genetic methods preclude its timed depletion after exit from mitosis. Therefore, we devised a procedure to drive cells depleted of AURKA during the M phase into G1. To this end, we first depleted AURKA using RNAi in cells arrested in mitosis by nocodazole treatment (Figure 2C and Supplementary Figure S2J), before forcing mitotic exit with AZ3146, an MPS1 inhibitor that overrides the spindle assembly checkpoint and permits AURKAdepleted cells to complete mitosis and enter G1 ( (39), 'Release in DMSO' panels in Figure 2C versus Supplementary Figure S2K). In order to analyze the implication of AURKA in the G1/S transition, nocodazole exposure was maintained during AZ3146 treatment to prevent cell division, and thereby circumvent asynchrony in mitotic timing ('Release in AZ3146 + nocodazole' panels in Figure 2C). Hence, in these experiments, cells move from mitosis into G1 and S phase duplicating the genome in the absence of cell division, thereby increasing their DNA content from 4N to 8N. These findings cannot simply be attributed to DNA re-replication after prolonged mitotic arrest, because CDT1 levels fall when cells synchronized with nocodazole are treated with AZ3146 to bypass the spindle assembly checkpoint and re-initiate DNA replication ( Figure 2D). Instead, these findings imply that our approach enables licensing of new origins in a subsequent cell cycle. Whereas AURKAexpressing control RNAi-treated cells initiate DNA replication to enter S phase ('Release in AZ3146 + nocodazole' upper panels in Figure 2C), AURKA-depleted cells show an impediment to DNA replication ('Release in AZ3146 + nocodazole' lower panels in Figure 2C and E). Taken together, our results suggest that AURKA performs a kinaseindependent function that is essential for the G1/S phase transition, but dispensable for replication progression once DNA synthesis has initiated.

AURKA performs a function in DNA replication independent of its kinase activity
To verify that the role of AURKA revealed by allosteric inhibitors was independent of its catalytic activity, we next tested if the failure of AURKA-depleted cells to transit from G1 to S phase can be prevented by the expression of a wildtype AURKA, as well as by a catalytically-inactive AU-RKA protein. To this end, FRT/TO HeLa cells expressing a doxycycline-inducible wild-type or kinase-dead Myctagged AURKA transgene were depleted of endogenous AURKA by RNAi in the presence of doxycycline. Cells treated in this way, which are expected to express either the wild type-or kinase-dead AURKA protein, were synchronously released into cell cycle from mitotic arrest, as before by exposure to AZ3146 plus nocodazole, alongside Nucleic Acids Research, 2020, Vol. 48, No. 14 7849 DMSO-treated controls ( Figure 3A, Supplementary Figure  S3A). As expected, reconstitution with a wild type form of AURKA permits cells to progress both through S phase and mitosis ( Figure 3B). However, reconstitution with a catalytically-inactive form of AURKA suffices to overcome this block in the G1/S phase transition ('AZ3146 + nocodazole' panels in Figure 3C and D) but, consistent with the widely established kinase-dependent role of AURKA in mitosis, fails to support mitotic cell division, inducing cell death at the late time-points ('DMSO' panels in Figure  3C versus Supplementary Figure S3B). Collectively, these findings provide genetic evidence that a kinase-independent function of AURKA--distinct from its putative role during mitosis in supporting geminin stability (10)--is essential for the G1/S transition.

AURKA has a role in the activation of replication origins independent of its kinase activity
Our findings prompted us to examine the role of AURKA in the mechanisms that initiate DNA replication. Several events crucial in ensuring that DNA replication occurs only once per cell cycle take place before entry into S phase. First, an inactive form of the hexameric MCM2-7 helicase assembles with the CDT1 factor on chromatin at sites that mark origins of replication to form the pre-replicative complex ('licensing'). Pre-replicative complex assembly occurs within a specific time window between late mitosis and early G1, when cyclin-dependent kinase (CDK) activity is low, and the activity of the anaphase-promoting complex (APC/C) is high. Subsequently, pre-replicative complexes are activated to initiate DNA replication ('firing') by the sequential chromatin recruitment of proteins such as CDC45 and TOPBP1, followed by factors like WDHD1, MCM10 and PCNA among others, collectively serving to induce the formation of the replicative polymerase holoenzyme. Origin firing occurs only after APC/C activity has been extinguished whereas CDK activity reaccumulates, triggering the G1/S transition (40,41).
We find that allosteric but not catalytic AURKA inhibitors suppress DNA replication initiation during the G1/S transition ( Figure 2B, Supplementary Figure S2A). Thus, the incorporation into nascent DNA of the nucleotide analogue EdU in cells transiting from G1 to S phase is inhibited by the allosteric inhibitor CD532, but not the catalytic inhibitor MLN8237 ( Figure 4A). Moreover, we have already shown that AURKA is dispensable for progression through S phase, since the effect of the allosteric inhibitors on replication is only observed when cells are treated in G1, before the initiation of DNA duplication has taken place ( Figure 2A versus B, compare Supplementary Figure S2A versus B), consistent with the fact that the majority of the origins required to complete DNA replication during an unperturbed S phase are activated during the G1/S transition.
The vast majority of origins licensed during the G1/S transition remain dormant during an unperturbed S phase, but may be activated later during S phase should cells encounter lesions that prevent or impede replication fork progression (42). To address whether AURKA was necessary for the activation of these late origins, we arrested cells using a double thymidine block and then released them synchronously into S phase. One hour later, during S phase pro-gression, we exposed the synchronized cells to a low dose of the replication inhibitor hydroxyurea ( Figure 4B), which is reported to trigger the activation of dormant origins to overcome replication inhibition and complete genome duplication (42,43). Whereas inhibition of AURKA kinase catalytic activity had no effect in S phase execution and cells only arrest when they reach mitosis (MLN8237 treatment in Figure 4B), we observe that an allosteric inhibitor of AU-RKA prevented the completion of DNA replication, but only in the presence of low doses of hydroxyurea (CD532 treatment in Figure 4B), when new origin activation is necessary (42,43). Collectively, our results therefore suggest that AURKA is essential not only for DNA replication firing during the G1/S transition, but also for the firing of dormant, licensed origins when cells encounter replication stress later during S phase.
Consistent with these observations, neither catalytic nor allosteric AURKA inhibitors affect the chromatin loading of CDT1, MCM2 or MCM7 ( Figure 4C), suggesting that AURKA is dispensable during G1 for pre-replicative complex assembly and replication licensing, in contrast to a prior report (10). Similarly, both types of AURKA inhibitors do not affect the phosphorylation of MCM2 (Supplementary Figure S4A) or the loading of CDC45 or TOPBP1 ( Figure 4D), suggesting that the early events during replication firing proceed normally. Remarkably however, the allosteric AURKA inhibitor CD532--but not the catalytic inhibitor MLN8237--markedly diminished the loading of WDHD1, MCM10, PCNA ( Figure 4D). Moreover, chromatin loading of SLD5, a component of the GINS complex necessary for replisome firing, was also impaired (Supplementary Figure S4B). Interestingly, the observed defect in MCM10 loading is evident 5 hrs after release from a nocodazole arrest even when CD532-treated cells remain in G1 (as marked by the absence of PCNA on chromatin, Figure 4D). No detectable effect on the level of chromatin-bound AURKA was observed for either CD532 or MLN8237 (Supplementary Figure S4C). Thus, our findings demonstrate that a kinase-independent function of AURKA is required for DNA replication firing by promoting the chromatin loading of proteins required for activation of the replicative polymerase.
Interestingly, the chromatin loading of TOPBP1 remains unaffected ( Figure 4D). Thus, our results suggest that the point at which a novel non-catalytic function of AURKA is required for replication initiation falls between TOPBP1 recruitment and GINS assembly. These results not only speak against the notion that the requirement for AURKA in G1 is somehow a consequence of prolonged mitosis or replication fork instability, but also better define a novel noncatalytic function of AURKA during the temporal progression of replisome assembly that is distinct from the role of AURKA kinase activity in replication fork protection during S phase shown previously (12).
Budding yeast Mcm10 is di-ubiquitinated during S phase, permitting recognition by PCNA (44). However, it remains unclear whether human MCM10 undergoes similar modifications, and whether such modifications regulate its function, even though two forms of human MCM10 whose gel migration corresponds to mono-and diubiquitinated species can be detected in late G1 and through S phase (45,46). In G1-synchronised cells (5 h after noco- dazole release), a slow-migrating form of MCM10 suggestive of a di-ubiquitinated species is not detected on chromatin, consistent with the absence of PCNA recruitment on chromatin before S phase initiation ( Figure 4D and E). When cells initiate replication (9 hrs after nocodazole release), this slow-migrating MCM10 form can be detected in the nuclear-soluble, but not chromatin-bound, fraction of CD532-treated cell extracts ( Figure 4E). Moreover, CD532 exposure decreases the chromatin-bound level of the faster-migrating MCM10 form and markedly reduces the level of the slow-migrating form of MCM10, compared to AURKA catalytic inhibitor or control DMSO exposure. These findings suggest that CD532 suppresses the loading of MCM10 onto chromatin, which in turn diminishes PCNA recruitment, to inhibit replication origin firing.
Interestingly, mass spectometry data showed that AU-RKA forms a macromolecular complex with several proteins involved in DNA replication initiation (Supplementary Table S1): it co-immunoprecipitates with MCM7, WDHD1 and POLD1 in an extract prepared from cells synchronized in G1. However, complex formation is not affected by either catalytic or allosteric AURKA inhibitors (Supplementary Figure S4D). Thus, these results suggest that chromatin loading of factors involved in replication firing requires an AURKA function that is blocked by al-losteric but not catalytic inhibitors; this function is, however, dispensable for complex formation.

AURKA allosteric inhibitors and DDK inhibitors mutually cross-sensitize HeLa cells
Our observation that a kinase-independent function of AU-RKA is essential for the initiation of DNA replication prompted us to test whether dividing cells would be sensitized to the effects of an allosteric AURKA inhibitor in combination with an inhibitor of replication initiation. To this end, we tested the viability of cells exposed to PHA-767491, a CDC7 inhibitor that blocks the Dbf4-dependent kinase essential for DNA replication (47), in combination with different types of AURKA inhibitors. While the combination of catalytic AURKA inhibitors with PHA-767491 had no additive effect ( Figure 5A and Supplementary Figure S5A), exposure to the allosteric AURKA inhibitors CD532 or AurkinA cross-sensitized HeLa cells to the cytotoxic effect of PHA-767491 ( Figure 5B, Supplementary Figure S5B, C). Similar results obtained when cells were treated with AURKA inhibitors in combination with Simurosertib, an alternative CDC7 inhibitor (48), further confirm this cross-sensitization (Supplementary Figure S5D, E). These findings raise the possibility that such a drug combination might exhibit enhanced efficacy in cancer therapy.   25 ) (lower panels). Viability was measured by the SRB assay and represented as a percentage normalized to control cells exposed to DMSO, plotted against the log 10 concentration of the indicated inhibitor. Values represent the mean ± SEM of three observations. (B) Cell viability of HeLa cells treated with combinations of CD532 together with PHA-767491 as indicated in A. All results are representative of at least three independent repeats. (C) Diagram depicting kinase dependent and independent functions of AURKA in mitosis and DNA replication respectively (see text for details).

Implications
The findings we report here identify a novel function of human AURKA in the initiation of DNA replication. AU-RKA operates after pre-replicative complex assembly and early steps in replication firing to promote the chromatin loading of essential factors for replicative polymerase activation. This AURKA function is independent of its kinase catalytic activity and, accordingly, can be blocked by a new class of allosteric inhibitors but not by classical catalytic-site inhibitors, causing cell cycle arrest at the G1/S phase transition ( Figure 5C). By contrast, the canonical, kinase-dependent role of AURKA in mitosis can be blocked by both catalytic and allosteric inhibitors ( Figure  5C). Moreover, this function is also required for the reactivation of dormant replication origins following DNA damage during S phase. Thus collectively, our findings define an AURKA-dependent step required for replicative polymerase activation via the chromatin loading of SLD5, WDHD1, MCM10 and PCNA, providing fresh insights into the mechanisms that control human DNA replication. We speculate that the frequent overexpression of AURKA in many different forms of human cancer may serve to dysregulate this replicative function, promoting its role as an oncogenic driver. Interestingly, allosteric AURKA inhibitors that disrupt its replicative function effectively crosssensitize cancer cells to inhibitors of the CDC7 replicative kinase under clinical development, suggesting a new approach with potential value in the combination therapy of human cancers. Thus, our work demonstrating a novel nonmitotic function of AURKA offers fresh biologic insights as well as potential therapeutic impact.