Distinct roles of MCM2-7 subunits in replication licensing in human cells

Abstract

Eukaryotic DNA replication requires the precise assembly of MCM2-7 single hexamers (SHs) into head-to-head double hexamers (DHs) at replication origins. While DH formation is well-characterized in budding yeast, the underlying mechanisms in human cells remain poorly understood. Here, we report cryo-electron microscopy structures of endogenous human MCM2-7 SH isolated from G1-phase cells. In these structures, human MCM2-7 adopts a latched spiral conformation in an autoinhibited state where the carboxyl-terminal extension (CTE) of MCM5 occupies the central channel, and MCM3-CTE is capable of locking the MCM2-5 gate to occlude DNA entry. Systematic functional analysis demonstrates that the six CTEs of MCM2-7 play distinct roles in SH stability, MCM loading, and DH formation on chromatin. Surprisingly, unlike in yeast, the human MCM3-CTE is dispensable for cell viability but ensures efficient genome-wide replication initiation. Our findings establish how human MCM2-7 enables flexible yet precise MCM loading via its CTEs, providing a framework for understanding the regulation of DNA replication initiation in higher eukaryotes.

DNA replication initiation requires precise MCM2-7 assembly. Here, the authors reveal cryo-EM structures of human MCM2-7 showing unique roles of its C-terminal extensions, defining how flexible yet accurate helicase loading ensures genome-wide replication initiation.

Introduction

In eukaryotes, DNA replication initiates from multiple replication origins distributed along each chromosome^1,2. This process is tightly regulated to ensure each origin fires only once per cell cycle, maintaining genome integrity. Central to this regulation is the loading of the MCM2-7 complex, the catalytic core of the replicative helicase^3–5, onto origin DNA into a head-to-head double hexamer (DH) that encircles duplex DNA, a process known as replication licensing or pre-replication complex (RC) assembly^2,6.

The budding yeast Saccharomyces cerevisiae has served as a key model for understanding pre-RC assembly. Yeast origins contain defined autonomously replicating sequences (ARS) featuring an 11-bp ARS consensus sequence (ACS) and multiple downstream B elements^1,2,7. During the early G1 phase, the yeast origin recognition complex (yORC) binds to ACS and B1 sites, inducing DNA bending to provide a platform for pre-RC assembly^8–10. Cdc6 is then recruited to bridge the AAA+ domains of Orc1 and Orc2, creating binding pockets for the C-terminal winged-helix domains (cWHDs) of Mcm3 and Mcm7^11,12. This configuration facilitates recruitment of the first Cdt1-Mcm2-7 heptamer onto ORC-Cdc6, followed by DNA insertion through the gap between Mcm2 and Mcm5 into the MCM ring^13,14, forming an intermediate ORC-Cdc6-Cdt1-Mcm2-7 (OCCM) complex^2,6,12,15. ATP hydrolysis triggers the release of Cdc6 and Cdt1 and ring closure¹⁶, transitioning the OCCM to an Mcm2-7-ORC (MO) complex^17–19. During this rearrangement, ORC relocates to a downstream B2 site^17,18,20,21, accompanied by a second DNA bending and creating space for loading a second Cdt1-Mcm2-7 heptamer¹⁷. Sliding of intermediates during this process further facilitates pre-RC assembly at origins²², ultimately forming a tightly coupled MCM DH at origin DNA^23–25. Intriguingly, an alternative, MO-independent licensing pathway is also utilized on some yeast origins, particularly those with extended ACS-B2 spacing^22,26–28. This pathway is believed to involve two independently loaded MCM SHs for DH formation^27,28.

While the core licensing mechanism is conserved across eukaryotes, human systems exhibit greater complexity and flexibility in origin selection and DH formation²⁹. Unlike yeast, human ORC (hORC) recruitment is primarily governed by chromatin context rather than specific DNA sequences^30–32. Whereas yORC remains DNA-bound throughout the cell cycle, hORC associates transiently with chromatin during G1 phase^33,34. Human ORC1 specifically recognizes nucleosomes containing di-methylated histone H4 at its K20 residue (H4K20me2) through its bromo adjacent homology (BAH) domain^31,32, subsequently recruiting hORC2-5 to origin DNA. In contrast to yORC, human ORC6 does not stably associate with hORC1-5^35–37. Additionally, soluble human MCM2-7 (hMCM2-7) exists as a stable single hexamer (SH)^38–41, differing from the yeast Cdt1-Mcm2-7 heptamer^14,42, and its interaction with CDT1 is restricted by geminin⁴³.

Recent biochemical reconstitution studies using purified human proteins have revealed key differences in pre-RC assembly on DNA between yeast and humans. First, both hORC6 and hCDC6 enhance, but are not strictly required for, MCM loading and DH formation in vitro^39–41. Second, CDT1-MCM2-7 interaction is detectable only at the OCCM stage, even in the absence of geminin, suggesting that this interaction occurs exclusively during initial MCM loading. Third, head-to-head double hOCCM assemblies can form at origin DNA, indicating a DH formation pathway involving two independently loaded SHs in human. Fourth, hORC6 promotes formation of a human MO (hMO) complex, distinct from the yeast MO, implying that hDH formation could also proceed through an MO-dependent pathway^39,41. Lastly, the hDH engages origin DNA through a constrained central channel that induces initial melting of duplex DNA at the hexamer interface^39,41,44. Despite these advances, the specific roles of the six distinct human MCM2-7 subunits in pre-RC assembly remain unclear.

In this study, we determined high-resolution cryo-EM structures of the soluble MCM2-7 SH isolated from G1-phase HeLa S3 cells. Structural analysis showed that the MCM ring adopts an autoinhibitory state, in which the cWHD of MCM5 occupies the central DNA-binding channel via interactions with the inner surfaces of MCM2 and MCM3. This configuration stabilizes the MCM ring in a left-handed coiled conformation. Unexpectedly, we found that MCM3-cWHD docks onto MCM2-CTD at an interface overlapping with the MCM5-CTD binding site, potentially functioning as a molecular safety latch. Systematic analyses demonstrate that the six C-terminal extensions (CTEs) of MCM2-7 play unique roles in regulating SH formation, helicase loading, and DH assembly.

Results

Endogenous hMCM2-7 single hexamer adopts an autoinhibitory conformation

In budding yeast, Cdt1 regulates the nuclear localization of Mcm2-7 SH⁴², maintains its structural integrity^14,45,46, and stabilizes an open-coiled conformation prior to DNA loading⁴⁷. In contrast, hMCM2-7 forms a stable complex in solution without detectable interaction with CDT1^39–41. To investigate hMCM2-7 regulation before helicase loading, we purified endogenous hexamer complexes from G1-phase HeLa S3 cells expressing C-terminally 3xFLAG-tagged MCM2 from its genomic loci for structure determination (Supplementary Fig. 1a–d). The isolated complexes were stabilized by GraFix crosslinking to preserve complex integrity (Supplementary Fig. 1e). Following cryo sample preparation, data collection, and image processing, our cryo-EM analysis yielded a 3.47-Å structure of hMCM2-7 SH (Supplementary Fig. 1f–n; Supplementary Table 1). The high quality of the map allowed us to unambiguously assign the various motifs and domains of hMCM2-7 SH and build its atomic model.

In this structure, hMCM2-7 SH adopts a left-handed coiled conformation with a narrow gap between MCM2 and MCM5 (Fig. 1a–d), resembling its yeast counterpart¹⁴ (Fig. 1e–h) and previously reported structures^38,41,48 (Supplementary Fig. 2a–c). The CTD tier binds ATP at MCM2-6, MCM6-4, and MCM4-7 interfaces and ADP at MCM3-5 and MCM7-3 sites (Fig. 1i–m). No nucleotide binds the MCM2-5 interface due to the ring breach. A comparison with hMCM-DH suggests that nucleotide exchange might occur at some of the ATPase interfaces, including MCM2-5, MCM5-3, MCM3-7, and MCM4-6 sites, during the SH-to-DH transition⁴⁴ (Fig. 1n, o).

Fig. 1. Overall architecture of the human MCM2-7 SH.

a–d Cryo-EM structure of the hMCM2-7 SH in top (a) and side (b–d) views, with the segmented density map superimposed with the atomic model. Relevant subunits are color-coded and labeled as indicated. e–h The segmented cryo-EM map of the yeast Mcm2-7 SH (EMD-6673). i–m Nucleotide occupancy at the MCM subunit interfaces. Walker A (WA) motifs are labeled, and segmented nucleotide densities (transparent gray) are shown at the same contour level. n–o Comparison of MCM nucleotide occupancy in the MCM2-7 SH (n) and in the MCM2-7 DH (PDB: 7W1Y) (o). p Domain organization of yMCM2 and hMCM2-7 subunits. q–r The cryo-EM density maps of hMCM2-7 SH (q) and yMcm2-7 SH (EMD-6673) (r), highlighting the conformations of the five CTEs (color-coded). s Schematic illustration of the CTE-mediated autoinhibition in hMCM2-7 SH.

Our structure resolved five stable CTEs with varying degrees of order (Fig. 1p, q). The CTEs of MCM2 and MCM6 dimerize atop the MCM ring (Fig. 1a–d, q; Supplementary Fig. 2d, e). MCM4-CTE extends over the central pore (Fig. 1a, q), though its overall density is weak. Further 3D subclassification focusing on the MCM4-CTE region revealed a bimodal distribution: in State I (49.1% of particles), MCM4-CTE is well-resolved; however, in State II (50.9%), it is substantially weaker or disordered (Supplementary Fig. 1j), indicating a dynamic equilibrium between ordered and flexible conformations. MCM5-CTE positions its cWHD within the DNA binding channel (Fig. 1q), where it contacts the inner surfaces of MCM2-CTD and MCM3-CTD (Supplementary Fig. 2f–h). Strikingly, MCM3-cWHD associates with MCM2-CTD (Fig. 1a, b, q), consistent with a recent study⁴⁸. Further analysis showed that MCM3-CTE is detectable in 77.6% of SH particles; within this subset, 29.2% exhibit a relatively weak interaction with MCM2-CTD (Supplementary Fig. 1h, i), reflecting conformational flexibility at this interface. This result was validated by cryo-EM analysis of an MCM3ΔCTE-containing hMCM2-7 (hMCM2-7Δ3CTE), which lacked the density for MCM3-cWHD (Supplementary Fig. 3). Removal of MCM3-CTE does not cause an obvious change in the overall MCM ring architecture (Supplementary Fig. 4a–g). Interestingly, only MCM7-CTE is fully flexible (Fig. 1a-d, q), highlighting its distinct role in MCM loading.

Structural comparison of hMCM2-7 with hOCCM revealed steric clashes between MCM4/MCM5-WHDs (hMCM2-7) and origin DNA (hOCCM), as well as between MCM6-CTE (hMCM2-7) and ORC (hOCCM) (Supplementary Fig. 2i, j). These observations suggest that CTEs help block premature MCM recruitment onto DNA, analogous to the regulatory mechanisms described in yeast^11,12,14,49. During loading, the flexible MCM7-CTE directly docks onto ORC1, while MCM3-WHD must undergo a major conformational change to disengage from MCM2-CTD and reposition at the CDC6-ORC2 interface (Supplementary Fig. 2k), a rearrangement that likely facilitates opening the MCM2-5 gate for DNA entry. However, our in vitro DNA pull-down assays showed that removal of MCM3-CTE alone is insufficient to confer DNA-binding ability (Supplementary Fig. 4h, i).

Together, these CTEs establish a multi-layer network that maintains hMCM in an autoinhibited state, physically occluding DNA entry from both the top and the side of the complex. This structural arrangement likely serves as an additional safeguard to ensure precise MCM loading orchestrated by ORC at replication origins throughout the genome.

A Meier-Gorlin syndrome-associated mutation (MCM3-Q761L) stabilizes the MCM2-MCM3 interface

Remarkably, a Meier-Gorlin syndrome-associated mutation, MCM3-Q761L, maps precisely to the interface between MCM3-CTE and MCM2-CTD (Fig. 2a–c). To probe its functional consequences, we purified the mutant complex and determined its cryo-EM structure (Supplementary Fig. 5). In this structure, the Q761L substitution introduces additional hydrophobic interactions with Arginine 751 of MCM2-CTD (Fig. 2b, c; Supplementary Fig. 5f–j), strengthening the MCM3-MCM2 interface. Unlike the wild-type MCM2-7 complex, where MCM3-CTE is absent in 22.4 % of particles (Supplementary Fig. 1h), all Q761L mutant particles show visible density for MCM3-CTE (Supplementary Fig. 5c). This structural observation is corroborated by yeast two-hybrid assays and in vivo affinity pull-down experiments, which confirm that the Q761L mutation enhances docking of MCM3-CTE onto MCM2 (Fig. 2d–f).

Fig. 2. Meier-Gorlin syndrome-associated mutation MCM3-Q761L impairs MCM loading.

a The cryo-EM map of the hMCM SH containing the MCM3-Q761L mutant. b, c Detailed interactions at the MCM3-MCM2 interfaces in the MCM3-Q761L mutant SH (b) and wild-type SH (c). Structures are shown in cartoon presentation with side chains of interacting residues displayed. d, e Yeast two-hybrid analysis of the MCM3-CTE-MCM2 interaction. Growth of AH109 yeast cells co-transformed with DNA-binding domain (BD)-MCM2 and either empty activation domain (AD) vector or AD-MCM3-CTE plasmids was assessed by spotting serial dilutions of cells onto synthetic complete medium (SCM) lacking the indicated amino acids. W, tryptophan; L, leucine; H, histidine. f Co-immunoprecipitation (IP) assay. Soluble extracts from HeLa S3 cells ectopically expressing MCM2-Twinstrep and 3×FLAG-MCM3-CTE were subjected to anti-FLAG IP. The bound proteins were detected by immunoblotting as indicated. Representative results from two independent experiments are shown. g Domain organization of MCM3 and MCM3-Δlinker (665-729) truncation mutant (left panels). Schematic illustration of the potential impact of MCM3-Δlinker mutation on MCM3-cWHD binding to MCM2-CTD (right panels). h Crude chromatin fractions were isolated and probed for the chromatin-bound 3×FLAG-tagged full-length MCM3, MCM3-Q761L, and their corresponding linker truncation mutants. Representative results from two independent experiments are shown.

Importantly, chromatin loading of the MCM complex harboring the Q761L mutation is significantly impaired (Fig. 2g, h). This finding suggests that a persistent MCM3-CTE-MCM2-CTD interaction may hinder subsequent steps required for productive loading onto chromatin. To test this hypothesis, we examined the role of the MCM3 linker region, which connects its CTD to cWHD. Deletion of the linker region (residues 665-729; MCM3Δlinker) does not affect chromatin loading of wild-type MCM3, indicating that the linker is dispensable for MCM loading under normal conditions. However, this same linker truncation largely rescues the chromatin loading defect of the MCM3-Q761L mutant, consistent with a recent study⁵⁰ (Fig. 2g, h). It is likely that the linker deletion restricts the ability of MCM3-CTE to engage MCM2-CTD, thereby restoring the loading of the Q761L mutant.

Together, these results suggest that efficient MCM loading requires precise temporal regulation of the MCM3-CTE/MCM2-CTD interaction. Failure to disrupt this interface in a timely manner may impair replication licensing.

Dynamic NTD tier of hMCM2-7

The NTD tier of hMCM2-7 exhibits pronounced flexibility, limiting its resolution in cryo-EM reconstructions (Supplementary Fig. 1 l). Through 3D classification, we identified three conformational states (I-III) of MCM2-7, which differ primarily in NTD stability (Fig. 3a-l; Supplementary Fig. 1i). In State III, the NTDs of MCM3 and MCM5 are highly flexible, while MCM2/6/4-NTDs remain relatively stable, and MCM7-NTD shows moderate destabilization. This variability correlates with the organization of zinc-finger (ZF) motifs. MCM2/6/4-ZFs cluster together, forming a stable ZF bundle through mutual interactions (Fig. 3m–o). MCM7-ZF connects to this bundle via the oligonucleotide/oligosaccharide-binding (OB) fold of MCM4, which binds to MCM2-ZF, whereas MCM3/5-ZFs are displaced from the ZF bundle and remain highly flexible (Fig. 3m–o). Notably, similar MCM3/5-NTD dynamics also occur in hOCCM (Fig. 3p). Upon DNA loading, however, the entire NTD tier is stabilized as shown in the structures of a fully loaded MCM-SH and MCM-DH^39,41 (Fig. 3q, r). These findings suggest that NTD flexibility may facilitate MCM loading and DH formation.

Fig. 3. Structural dynamics of the NTD tier in hMCM2-7 SH.

a–l Top and side views of the segmented cryo-EM density maps at the same contour level showing three conformational states of the hMCM2-7: State I (a–d), State II (e–h), and State III (i–l). m Bottom view of hMCM2-7 density map, highlighting the six ZF motifs (color-coded) with atomic models overlaid. n Zoomed-in view of the ZFs as in (m). Boxed region shows the ZF bundle formed by MCM2/6/4, with MCM4-OB (oligonucleotide/oligosaccharide-binding) fold indicated. o Schematic illustration of the ZF bundle organization and its relationship with MCM4-OB and MCM7-ZF. The interactions among these motifs are indicated by the double-headed arrows. p ZF conformations in hOCCM (EMD-19623; PDB-8S0E). The two stable ZFs from MCM4 and MCM6 are color-coded. q ZF arrangement of in hMCM2-7 DH (EMD-32258; PDB-7W1Y). r Structural comparison of ZFs between hMCM2-7 SH (gray) and DH (colored) states. Arrows highlight conformational changes during the SH-to-DH transition. MCM6-ZF was used as a reference for alignment.

Distinct roles of the six MCM-CTEs in hexamer integrity

Unlike budding yeast, where Mcm2 lacks a CTE (Fig. 1r), each hMCM subunit contains a distinct WHD-bearing CTE (Fig. 1p). To dissect their roles, we generated six CTE truncation mutants to assess their impacts on hexamer formation (Fig. 4a). Each MCM-CTE mutant bearing a 3xFLAG tag was overexpressed in HeLa S3 cells to perform anti-FLAG immunoprecipitation (IP) assays. We found that MCM2-CTE removal significantly reduced the ability of MCM2 in co-precipitating other subunits by 80-90% when compared with its WT version (Supplementary Fig. 6a, b), indicating a crucial role of MCM2-CTE in facilitating MCM2 incorporation into the MCM ring and/or stabilizing MCM2 associating with other MCM subunits. Deletion of MCM3-CTE or MCM4ΔCTE caused moderate reductions (50-60%) (Supplementary Fig. 6c–f). Similarly, MCM5ΔCTE showed a substantial defect in ring assembly, though it retained relatively stronger binding to MCM3 (Supplementary Fig. 6g, h). In contrast, truncation of the MCM6-CTEs or MCM7-CTE had minimal impact (Supplementary Fig. 6i-l). These results demonstrate that the CTEs of MCM subunits play unique roles in regulating hexamer stability in human cells.

Fig. 4. Functional analysis of MCM-CTEs in DH formation.

a Schematic illustration of MCM2-7 CTE truncation mutants used in this study. b–g Chromatin binding of FLAG-tagged MCM-WT or -ΔCTE subunits. Crude chromatin fractions were isolated from cells expressing the indicated constructs and probed by immunoblotting. h–m Quantification of chromatin-bound MCM levels from (b–g). Band intensities were normalized to the loading control. Data represent mean ± SEM from three independent biological replicates. Statistical analysis was performed using a two-tailed Welch’s unpaired t-test. *P < 0.05; **P < 0.01; ns (not significant). P-values of each analysis are provided in the Source Data file. n–y Representative negative-staining EM images of chromatin-bound MCM2-7 complexes isolated via FLAG affinity purification from HeLa S3 cells expressing 3xFLAG-tagged MCM WT or ΔCTE. Corresponding 2D class averages of the MCM particles are shown in the upper-right panels. Scale bar: 200 nm. Representative images from 2 independent experiments are shown.

MCM loading and DH formation on chromatin are regulated by MCM-CTEs

To understand the roles of individual CTEs in MCM loading in human cells, we analyzed G1 chromatin binding of each CTE-truncated mutant overexpressed in HeLa S3 cells. Compared to their WT MCM subunit, the mutants exhibited varying reductions in chromatin recruitment: MCM2ΔCTE (70% reduction), MCM3ΔCTE (5%), MCM4ΔCTE (50%), MCM5ΔCTE (10%), MCM6ΔCTE (75%), and MCM7ΔCTE (90%) (Fig. 4b–m). These results highlight the importance of MCM2/6/7-CTEs in MCM loading.

We next examined whether these mutants support DH formation by isolating their chromatin-bound complexes for imaging by negative-staining EM. Our results showed that MCM-DHs are associated with MCM2ΔCTE (Fig. 4n, o), MCM3ΔCTE (Fig. 4p, q), MCM4ΔCTE (Fig. 4r, s), or MCM5ΔCTE (Fig. 4t, u; Supplementary Fig. 7a-d), indicating that their CTEs are dispensable for DH formation in the presence of their endogenous subunits. However, MCM6ΔCTE yielded only low level of MCM-SHs on chromatin (Fig. 4v, w; Supplementary Fig. 7e, f), likely due to destabilized chromatin association which ultimately leads to a failure in DH formation. Strikingly, MCM7ΔCTE abolished chromatin association of both MCM-SH and MCM-DH (Fig. 4x, y; Supplementary Fig. 7g), underscoring its essential role in pre-RC assembly.

The CTEs of MCM2, MCM3, and MCM4 are dispensable for proliferation in human cells

In yeast, Mcm3-CTE is essential for MCM loading through binding to the interface between Orc2 and Cdc6 to promote OCCM formation^11,45. However, human MCM3-CTE showed a negligible effect on pre-RC formation as the overexpressed MCM3ΔCTE mutant can be incorporated into MCM-DH on chromatin in HeLa S3 cells (Fig. 4p, q). Given the conservation of MCM3-CTE in protein sequence across species (Supplementary Fig. 8a), we rigorously tested whether MCM3ΔCTE alone could support DH formation on chromatin. To deplete endogenous MCM3, we utilized mAID-MCM3 cells, where mini-auxin-inducible degron (mAID) was fused to both alleles of MCM3 at its N-terminus for an immediate degradation of the endogenous MCM3 upon auxin treatment^51–53 (Fig. 5a, b, lanes 1-11). As expected, MCM3 depletion eliminated MCM loading on G1 chromatin in cells after release from a nocodazole block (Fig. 5b, lanes 7–11). We found that this defect can be rescued by expressing either wild-type MCM3 (Fig. 5b, lanes 12–17) or MCM3ΔCTE mutant (Fig. 5b, lanes 18–23). Moreover, we also isolated the chromatin-bound MCM2-7 complexes containing MCM3ΔCTE for EM imaging (Fig. 5c–e). Our analysis showed that the mutant MCM indeed forms DH in the absence of the endogenous MCM3 (Fig. 5f, g).

Fig. 5. MCM3-CTE is not essential for MCM loading and cell proliferation.

a, b Nocodazole block-and-release assays for mAID-MCM3 cells expressing exogenous 3xFLAG-MCM3 or -MCM3-ΔCTE. Schematic of the experimental workflow shown in (a). Cell samples were collected at the indicated time points for chromatin binding assay and analyzed by SDS-PAGE and immunoblotting of the indicated proteins (b). The cell lines bearing an empty vector (NC) was used as a negative control. Representative results from two independent experiments are shown. c–g Purification and analysis of MCM2-7 complexes isolated from mAID-MCM3 cell lines. c The isolation scheme. d, e Glycerol gradient sedimentation of the isolated complexes. f, g Representative negative-staining EM images of MCM-DHs from fractions 7-8 highlighted in (d, e). Representative 2D class averages are shown in the upper-right panels. Scale bar is 200 nm. Representative results from two independent experiments are shown. h Proliferation of HCT116 WT and HCT116 MCM3-CTE-/- cells assessed by colony formation assay. i Growth complementation assay in GALS-MCM3 yeast cells. Serial dilutions of cells expressing the indicated constructs were spotted on both YPG (galactose) and YPD (dextrose) plates. Empty vector and wild-type MCM3 serve as controls. j Growth curve analysis of HCT116 WT and MCM3-CTE-/- cells. The curve was plotted based on samples collected from 3 biological replicates. Error bars represent mean ± SD. k–m Functional analysis of MCM3-CTE-/- cells. k Schematic illustration of the nocodazole block-and-release assay. l Immunoblot analysis of chromatin-bound MCM complexes at the indicated times after release. Representative results from two independent experiments are shown. m Flow cytometry (FACS) analysis of DNA content at the corresponding time points, assessing replication progression.

To determine whether MCM3-CTE is essential for cell viability, we attempted to generate a homozygous MCM3ΔCTE mutant (MCM3CTE-/-) in HCT116 cells using CRISPR/Cas9-mediated genome editing approach (Supplementary Fig. 8b–d). We found that the MCM3CTE (-/-) cells remained viable (Fig. 5h), contrary to its yeast counterpart (Fig. 5i). The mutant cells exhibited slower growth (Fig. 5j), and MCM loading occurred on chromatin with reduced efficiency (Fig. 5k, l, lanes 10-19) compared to the WT cells (Fig. 5l, lanes 1-9). Accordingly, FACS analysis showed that the mutant cells exhibited an obvious delay in DNA replication (Fig. 5m; Supplementary Fig. 10). Although DNA replication can be completed, an elevated level of γH2A.X signals indicated a challenge of replication stress (Fig. 5l), likely due to compromised pre-RC assembly genome-wide in the mutant cells.

We extended this analysis to the CTEs of MCM2, MCM4, and MCM5. Homozygous CTE-knockout cell lines were successfully established for MCM2 and MCM4 (Fig. 6a, b; Supplementary Fig. 8e–j). This contrasts with yeast, which lacks an MCM2-CTE homolog (Fig. 1p) and where Mcm4-CTE deletion severely impairs cell growth (Fig. 6c). Our subsequent characterization showed that in human cells, MCM2-CTE deletion obviously reduced, but did not abolish, MCM loading on chromatin (Fig. 6d). Moreover, MCM4-CTE deletion also caused a reduction in pre-RC assembly (Fig. 6e).

Fig. 6. The roles of the CTEs of MCM2, MCM4, and MCM5 in proliferation and MCM loading.

a, b Proliferation of wild-type and CTE-knockout cells assessed by colony formation assay. a MCM2-CTE + /+ and MCM2-CTE-/-; b MCM4-CTE + /+ and MCM4-CTE-/-.c Yeast complementation assay for yMCM4ΔCTE. Serial dilutions of GALS-MCM4 yeast cells expressing empty vector, wild-type yMCM4, or yMCM4-ΔCTE were spotted on the indicated medium. d, e Nocodazole block-and-release assays. Chromatin binding assays were performed to evaluate MCM loading in MCM2-CTE-/- (d) and MCM4-CTE-/- (e) mutant cells. Immunoblot of chromatin-bound proteins from relevant cells at the indicated times after nocodazole release. Representative images from two independent experiments were shown. f, g Proliferation of MCM5-mAID cells complemented with wild-type or mutant. Colony formation assay of cells expressing MCM5 or MCM5ΔCTE in the absence (f) or presence (g) of auxin (5’-Ad-IAA). h The yeast MCM5ΔCTE mutant was tested for complementation of Mcm5 depletion in GALS-MCM5 cells. Empty vector and wild-type MCM5 were used as controls. i MCM chromatin loading in MCM5‑mAID cells. Immunoblot analysis of chromatin‑bound proteins from MCM5‑mAID cells expressing empty vector (NC), MCM5-3FLAG, or MCM5‑ΔCTE-3FLAG after release from a nocodazole block.

In contrast, we were unable to isolate viable MCM5CTE-/- cells, suggesting MCM5-CTE is essential for proliferation. To investigate its function, we used an MCM5-mAID degron system complemented with either wild-type MCM5 or MCM5∆CTE. Under auxin-induced depletion of endogenous MCM5, only WT MCM5, but not MCM5∆CTE, rescued cell proliferation, confirming the essential role of MCM5-CTE for viability (Fig. 6f, g). This observation aligns with the fact that yeast Mcm5-CTE is indispensable for cell growth (Fig. 6h). We then examined the ability of MCM5∆CTE to support MCM loading after endogenous MCM5 depletion. Compared to the wild-type MCM5, MCM5∆CTE also showed reduced chromatin association of the MCM complex (Fig. 6i). These results suggest that MCM5-CTE plays a crucial and conserved role in regulating pre-RC assembly, and its complete loss impairs replication licensing.

In summary, our findings reveal a functional divergence between yeast and human MCM CTEs. In human cells, the CTEs of MCM2, MCM3, and MCM4 are individually dispensable for the core mechanics of replication licensing, although they contribute to maintaining licensing efficiency. Conversely, human MCM5-CTE, like its yeast ortholog, is strictly required for proliferation, suggesting its conserved role in replication initiation.

Discussion

DNA replication, whether in prokaryotes or eukaryotes, requires the unwinding of the duplex template by threading one of the strands through a hexameric ring helicase⁴. In bacteria, the helicase is a homohexameric ring formed by six identical DnaB subunits; whereas in archaea, six identical MCM subunits. In eukaryotes, the heterohexameric ring, MCM2-7, is formed by six highly conserved MCM core subunits with nonidentical N-terminal and C-terminal extensions^3,49. The essential nature of each MCM subunit implies they serve distinct, non-redundant functions in replication^2,54,55. In this study, we determined the cryo-EM structure of the endogenous MCM2-7 SH stabilized by GraFix crosslinking. Unlike prior studies, our structure reveals stable conformations for five of the six CTEs; only MCM7-CTE remains highly flexible (Fig. 1q; Supplementary Fig. 2a–c). This unique flexibility of MCM7-CTE appears critical for MCM loading, as its deletion completely abolished chromatin association of the MCM complex even in the presence of endogenous proteins.

Our study shows that the endogenous human MCM2-7 SH adopts a latched spiral conformation with two key autoinhibitory features: (1) MCM5-CTE blocks the central channel, physically occluding DNA entry; and (2) MCM3-CTE acts as a safety latch, locking the MCM2-5 gate. These observations contrast with the open coiled structure of yeast Mcm2-7, which is stabilized by Cdt1^14,47 and lacks the engagement between yMcm3-CTE and yMcm2-CTD. This species-specific MCM3-CTE conformation represents a unique regulatory layer ensuring precise replication licensing in human cells, where a large pool of free MCM2-7 SHs is maintained in the G1 nucleus^56,57. Supporting this hypothesis, our structural and functional analysis of the Meier-Gorlin syndrome-associated MCM3-Q761L, located at the MCM3- MCM2 interface, showed that it enhances the docking affinity of MCM3-CTE for MCM2-CTD while concurrently inhibiting MCM chromatin loading. Importantly, deletion of the linker region connecting MCM3-CTD to its cWHD largely suppressed this loading. This observation rules out the possibility that the impaired loading results from defective OCCM assembly rather than enhanced autoinhibition. Together, these results suggest that efficient MCM loading requires precise regulation of the MCM3-CTE/MCM2-CTD interaction. Failure to disrupt this interface in a timely manner could impair replication licensing, linking this mechanistic defect to the pathophysiology of Meier-Gorlin syndrome.

Through systematic mutagenesis, we identified the subunit-specific roles for the CTEs in maintaining MCM integrity and promoting MCM loading (Supplementary Table 2). Unlike yeast, human MCM2 possesses an additional WHD-containing CTE that forms intimate contacts with MCM6 (Supplementary Fig. 2d, e). CTE removal significantly reduces hMCM2’s binding affinity for other MCM subunits, suppressing chromatin association of the mutant MCM-DH. While yeast lacks an equivalent Mcm2-CTE, Cdt1 binding to the NTD interface between Mcm2 and Mcm6 strengthens subunit interactions. Deletion of the CTEs from MCM3, MCM4, or MCM5 also disrupts hexamer formation to varying degrees, likely through disrupting interactions with their binding partners. Although MCM4-CTE does not show stable contacts with other MCM subunits in our structure, it may transiently engage with other subunits during hexamer assembly. Notably, in our HeLa S3 overexpression system, where both endogenous MCM proteins and overexpressed mutants are present, the MCM3ΔCTE and MCM5ΔCTE mutants, despite being less efficient in ring assembly, were still efficiently loaded onto chromatin to form pre-RCs (Fig. 4c, e, i, k; Supplementary Fig. 7a–d). Given that MCM5-CTE occludes the central DNA-binding channel of the MCM ring, its removal may relieve an autoinhibitory block and enhance loading efficiency. Thus, it is plausible that the MCM5-ΔCTE mutant loads more efficiently than the wild-type complex, compensating for its reduced stability during SH assembly. In addition, chromatin-bound WT MCM SHs may facilitate the recruitment or stabilization of mutant complexes, enabling their subsequent conversion into DHs. Consistent with this model, chromatin association of the MCM complex was markedly reduced in homozygous CTE-knockout cells (MCM2-CTE-/- and MCM3-CTE-/-) and in MCM5-mAID cells expressing MCM5ΔCTE following endogenous MCM5 depletion.

Upon DNA loading, MCM2-7 transitions from a coiled spiral to a planar ring, involving precise opening and closure of the MCM2-5 gate. Our hMCM2-7 structure suggests that MCM3-cWHD binds MCM2-CTD, potentially blocking DNA entry, while the same MCM2 surface interacts with MCM5-CTD to close the MCM2-5 gate, as seen in the hOCCM structure^39,40 (Supplementary Fig. 9a–d). These findings suggest that timely displacement of MCM3-CTE from MCM2 is crucial for DNA entry and subsequent ring closure by MCM5. Although the exact mechanism of MCM3-CTE repositioning remains unclear, we propose that the structural reconfiguration during MCM loading onto origin DNA destabilizes its interaction with MCM2, liberating MCM3-CTE to engage the ORC2-CDC6 interface (Supplementary Fig. 9e–h). Moreover, MCM5-CTE occupies the central channel, sterically blocking DNA entry and likely preventing premature engagement with ORC. The release of this CTE is therefore a critical regulatory step in MCM loading. Recent in vitro pre-RC assembly studies have shown that hMCM-SH can be efficiently recruited onto dsDNA only in the presence of ORC1-5 and CDT1^39–41, indicating that ORC and CDT1 actively remodel the hMCM-SH into a conformation competent for stable DNA binding. Notably, previous work in yeast demonstrated that ORC-Cdc6 binding allosterically induces a conformational rearrangement in the MCM ring that is coupled to displacement of Mcm5-CTE from the central pore^12,58, thereby clearing the channel for DNA loading. Given the high structural conservation of MCM and ORC-Cdc6 complexes across eukaryotes, a similar mechanism is likely to reposition MCM5-CTE during human MCM loading.

Surprisingly, MCM3-CTE is dispensable for replication licensing and cell viability, unlike in yeast, where it is essential for each round of MCM loading⁴⁵. We found that MCM3-CTE deletion in human cells only reduces the chromatin-bound MCM levels (Fig. 5l), indicating compromised pre-RC assembly genome-wide. This defect correlates with delayed DNA replication initiation and elevated replication stress (Fig. 5l, m). Intriguingly, recent in vitro studies showed that pre-RC assembly can occur, albeit less efficiently, without CDC6³⁹. As CDC6 insertion into the ORC ring on DNA creates the binding pocket for MCM3-CTE, CDC6 omission basically eliminates MCM3-CTE binding to ORC. These results indicate that MCM3-CTE enhances MCM loading efficiency rather than being absolutely required. Similarly, the CTEs of MCM2 and MCM4 are also not essential for proliferation but do impact replication licensing. These results suggest that the CTEs of MCM2/3/4 help maintain highly efficient pre-RC assembly on chromatin.

In contrast, the CTEs of MCM6 and MCM7 are essential for pre-RC assembly but dispensable for hexamer formation. MCM7-CTE is required for initial MCM recruitment to G1 chromatin, likely by mediating MCM-ORC contact. In contrast, MCM6ΔCTE permits low-level MCM2-7 SH recruitment. In yeast, Mcm6-CTE deletion slows down DNA insertion into the MCM ring^12,59. Since MCM6-CTE interacts with CDT1 specifically in the OCCM, it remains unclear whether CDT1 is recruited onto OCCM to stabilize the MCM ring with the help of MCM6-CTE. It is likely that the MCM6ΔCTE-containing MCM complexes may form non-productive intermediates on chromatin that are readily disassembled.

Based on our findings, we propose a working model for MCM loading in human cells (Fig. 7a). First, the flexible MCM7-CTE initiates contact with ORC through ORC1 binding. Second, MCM3-CTE disengages from MCM2-CTD and binds to the ORC2-CDC6 interface, opening the MCM2-5 gate for DNA entry. As a result, MCM2-7 transitions from a spiral to a planar ring conformation, accompanied by reconfiguration of MCM5-CTE and MCM6-CTE. These structural changes enable full MCM engagement with ORC, facilitating precise DNA insertion into the MCM ring. CDT1 likely binds to MCM2-7 during this transition. The association of CDT1 with the NTDs of MCM2/6/4 could compensate for MCM2-CTE flexibility, as shown in the OCCM structure^39,40, to maintain ring integrity during MCM loading. Next, ATP hydrolysis triggers the release of CDC6 and CDT1 and ring closure. Finally, two independently loaded MCM2-7 SHs converge to form a head-to-head DH. When MCM3-CTE is absent, pre-RC assembly can still be achieved by MCM7-CTE alone but with a reduced efficiency (Fig. 7b). We speculate that the dynamic nature of the NTD tier in hMCM2-7 SH, particularly the flexibility of MCM3 and MCM5 NTDs, also contributes to MCM loading and DH formation by accommodating necessary structural transitions from SH to DH.

Fig. 7. Distinct roles of human MCM2-7-CTEs in pre-RC assembly.

a Model of MCM loading mechanism mediated by MCM2-7 CTEs. MCM7-CTE initially contacts ORC to recruit the autoinhibited MCM2-7 onto DNA. MCM3-CTE is then repositioned to bind to the ORC2-CDC6 interface, driving highly efficient MCM loading. CDT1 stabilizes the complex by binding the MCM2/4/6 NTDs. Two loaded MCM2-7 SHs subsequently converge to form a head-to-head DH. b In wild-type cells, both MCM3-CTE and MCM7-CTE coordinate to achieve optimal ORC-CDC6 engagement and efficient loading (a). In the absence of MCM3-CTE, the MCM7-CTE alone can still mediate recruitment and DH formation, albeit with reduced efficiency, illustrating the functional flexibility of the human licensing system.

The ability of hMCM2-7 to form DHs independently of ORC6, CDC6, or MCM2/3/4-CTEs highlights the remarkable flexibility of replication licensing in human cells. This plasticity may reflect an adaptation to the complex life cycles of multicellular organisms. Our work provides crucial insights into how human cells balance precision and adaptation in regulating DNA replication initiation, a critical process for maintaining genome stability and preventing cancer.

Method

Cell culture

HeLa S3, HCT116, and HEK 293T cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin (100 U/mL, Thermo Fisher) in a 37 °C incubator with 5% CO₂.

Cell line construction

MCM2-3xFLAG HeLa S3 Cell

To knock-in a 3xFLAG tag at the C-terminus of endogenous MCM2, a sgRNA targeting the C-terminal MCM2 locus (sequence: GAGGCCCTATGCCATCCATA) was cloned into pX330 (Addgene #42230). A DNA cassette containing S-tag-3xFLAG-NeomycinR or S-tag-3xFLAG-HygromycinR was inserted into pUC57, flanked by homology arms surrounding the MCM2 stop codon. The three constructs were co-transfected into HeLa S3 cells using Lipofectamine 2000 (Invitrogen, 11668019). Transfected cells were selected with 600 μg/mL Geneticin (Gibco, 10131035) and 250 μg/mL Hygromycin B (Gibco, 10687010) for ~14 days. Single clones were isolated by sorting into 96-well plates and validated by genomic PCR and Western blotting to confirm homozygous tagging.

MCM ΔCTE (CTE-/-) HCT116 cell line construction

To construct the MCM3ΔCTE HCT116 cell line, a premature stop codon was introduced after the genomic sequence encoding MCM3 D670. Two sgRNAs were designed: one targeting the MCM3 D670 region and another targeting the MCM3 C-terminus region. The sequences were cloned into pX330 respectively. A donor cassette containing a stop codon followed by NeomycinR or PuromycinR was inserted into pUC57, flanked by homology arms upstream of MCM3 D670 and downstream of the MCM3 C-terminus. The four constructs were co-transfected into HCT116 cells using Lipofectamine 2000. Transfected cells were selected with 600 μg/mL Geneticin and 1 μg/mL Puromycin for ~10 days. Single clones were isolated by sorting into 96-well plates and validated by genomic PCR and Western blotting. MCM2ΔCTE or MCM4ΔCTE HCT116 cell line was constructed similarly as described above, with a stop codon introduced after MCM2 R825 or MCM4 D775. Plasmids and guide sequences used for cell line construction are listed in Supplementary Table 4.

Stable cell lines expressing exogenous MCM

Exogenous MCM2/3/4/5/6/7 (wild-type or CTE-truncated variants) were stably expressed via lentiviral transduction. The detailed residue numbers are as follows: MCM2ΔCTE (1-825), MCM3ΔCTE (1-670), MCM4ΔCTE (1-775), MCM5ΔCTE (1-685), MCM6ΔCTE (1-665), and MCM7ΔCTE (1-648). For lentivirus production, HEK 293 T cells were co-transfected with psPAX2 (Addgene #12260), pMD2.G (Addgene #12259), and a pLEX-Puro plasmid harboring 3xFLAG-tagged MCM cDNA. 16 hours post-transfection, the transfection mix was replaced with fresh medium. Viral supernatant was harvested after 48 h, clarified by centrifugation (1000 × g, 5 min), and used to transduce target cells. HeLa S3 cells were transduced with lentivirus for 18 h, then selected with 2 μg/mL Puromycin (Gibco, A11138-03). Protein expression was verified by western blotting. HCT116 mAID-MCM3 cells and HCT116 MCM5-mAID-3HA cells were transduced as above.

Protein purification

hMCM2-7 SH

MCM2-3xFLAG HeLa S3 cells were arrested at G2/M using 50 μg/mL nocodazole for 16 h, then released into fresh medium for 8 h before collection by centrifugation (200 × g). Cell pellets were lysed in extraction buffer (EBX) (150 mM L-glutamic acid potassium salt, 50 mM HEPES-KOH [pH 7.5], 10 mM magnesium acetate, 1 mM EDTA, 0.5% Triton X-100, 3 mM ATP, 2 mM NaF, 1 mM Na₃VO₄, 1 mM PMSF, 1× protease inhibitor cocktail [Roche cOmplete]) on ice for 15 min. After centrifugation (3000 × g, 5 min), the clarified lysate was incubated with ANTI-FLAG M2 affinity gel (Sigma, A2220) for 2 h at 4 °C. The resin was washed extensively with EBX, and bound proteins were eluted using 0.5 mg/mL 3xFLAG peptide in EBX. Eluates were subjected to glycerol gradient sedimentation (10–30%) in buffer (50 mM HEPES-KOH [pH 7.5], 150 mM potassium acetate, 10 mM magnesium acetate, 1 mM EDTA, 3 mM ATP) using a TLS-55 rotor (Beckman Optima MAX-XP Ultracentrifuge) at 105,000 × g for 13 h at 4 °C. For cryo-EM sample preparation, GraFix was applied by supplementing the 30% glycerol buffer with 0.05% glutaraldehyde (Sigma, G5882). Fractions containing MCM-SH were pooled, quenched with 10 mM ice-cold Tris-HCl (pH 8.0), and dialyzed against buffer (50 mM HEPES-KOH [pH 7.5], 250 mM KCl, 10 mM magnesium acetate, 1 mM EDTA, 3 mM ATP) to remove glycerol.

hMCM2-7-Δ3CTE and hMCM2-7(MCM3Q761L) mutant SH

The mutant SHs were purified and crosslinked for cryo-EM sample preparation as described above using HeLa S3 cells expressing exogenous 3xFLAG-MCM3(ΔCTE) or 3xFLAG-MCM3(Q761L).

Cryo-EM grid preparation and data acquisition

Cryo-grids were prepared by applying 4 μL of hMCM2-7 samples after buffer exchange to remove glycerol and concentration in an ultrafiltration unit (50 kDa cutoff) to holey carbon grids (C-flat R1.2/1.3, 400 mesh Au) after glow discharge (15 mA, 45 seconds). After incubation for 30 seconds at 4 ˚C and 100% humidity, the grids were blotted for 4 seconds and 0 blot force, and then quickly plunged into liquid ethane cooled by liquid nitrogen with a Mark IV Vitrobot (Thermo Fisher Scientific). The grids were loaded into a FEI Titan Krios G3i electron microscope (Thermo Fisher Scientific) equipped with a high-brightness field emission gun operated at 300 kV. Images were automatically collected with a K3 Summit direct electron detector (Gatan) using EPU software (Thermo Fisher Scientific) in counting mode with magnification of 81,000× (1.06 Å physical pixel size). The slit width of the Gatan Imaging Filter (GIF) Bio Quantum was set to 20 eV. The defocus ranges, dose rates and other parameters during image acquisition of each dataset are listed in Supplementary Table 1.

Image processing

hMCM2-7 SH, hMCM2-7Δ3CTE SH, and hMCM2-7(MCM3-Q761L) datasets were processed with cryoSPARC⁶⁰. Drift correction of the collected movies was performed using patch motion correction. Motion-corrected sums without dose-weighting were used for contrast transfer function (CTF) estimation with patch CTF correction. Motion-corrected sums with dose-weighting were used for all subsequent image processing. Particles were first automatically picked by blob picking and extracted for 2D classification. 2D classes with clear features were selected for ab initio reconstruction to generate initial models, which were then used in heterogeneous refinement to obtain a set of aligned MCM SH particles. These particles were then used for template picking to obtain additional SH particles. Topaz was also applied for particle picking with the hMCMΔ3CTE SH dataset^61,62. The picked particles were cleaned again with heterogeneous refinements. 3D classifications focusing on different regions of these final particle sets of MCM SH were performed to obtain different states of MCM SH. Local refinement was used to improve the resolution of M2/6/4 ZF bundle and MCM2/6 CTEs for model building. Flowcharts for data processing of MCM2-7 SH, MCMΔ3CTE SH, and hMCM2-7(MCM3-Q761L) are presented in Supplementary Figs. 1, 3 and 5, respectively.

Model building

Models of individual hMCM subunits without CTEs from human MCM DH structure (PDB: 7W1Y) were first docked into the final B-factor sharpened map of hMCM2-7 SH with the map-fitting function in UCSF ChimeraX⁶³. Models of MCM CTEs were then docked into the corresponding densities. AlphaFold predicted MCM3-cWHD model was docked into the extra density observed on the surface of MCM2 CTD^64,65. After removal of other extra residues that do not fit the observed densities and subsequent manual adjustments in Coot⁶⁶, the model was refined against the cryo-EM density map with phenix.real_space_refine module in PHENIX package⁶⁷.

Final MCM SH model without MCM3-cWHD was then fitted into B-factor sharpened maps of states 1 and 2 of hMCM2-7Δ3CTE. After manual adjustment in Coot⁶⁶, the models were refined against the corresponding cryo-EM density maps with phenix.real_space_refine module in PHENIX package⁶⁷. The qualities of the three final refined models were estimated with MolProbity⁶⁸. The figures on the maps and refined models were generated with UCSF ChimeraX and PyMOL (http://pymol.org). Refinement statistics are presented in Supplementary Table 1.

FLAG-IP, chromatin binding assay, and isolation of chromatin-bound MCM complex

G1 HeLa S3 cells expressing exogenous MCM were harvested as described in the above section for hMCM2-7 SH purification. Cell pellets were resuspended in EBX buffer and incubated on ice for 15 min. The lysate was fractionated into soluble fraction and crude chromatin pellet by centrifugation (3000 × g, 5 min). The soluble fraction was incubated with ANTI-FLAG M2 affinity resin for 2 h. The resin was then washed with EBX buffer, and bound proteins were eluted with 0.5 mg/mL 3xFLAG peptide in EBX buffer and analyzed by Western Blotting. The crude chromatin pellet was washed extensively with EBX buffer and analyzed by Western blotting for chromatin-bound proteins. Chromatin binding assay for HCT116 cells was performed using the same fractionation protocol.

To isolate chromatin-bound MCM, the crude chromatin pellet was digested with 1 U/μL benzonase (7sea Biotech, RPE002) in benzonase buffer (50 mM HEPES-KOH [pH 7.5], 150 mM L-glutamic acid potassium salt, 8 mM MgCl₂, 1 mM EDTA, 0.02% NP-40, 3 mM ATP, 2 mM NaF, 1 mM Na₃VO₄, 1 mM PMSF, 1× protease inhibitor cocktail [Roche, cOmplete]) at 37 °C for 10 min, followed by incubation on ice for 1 h. The clarified chromatin extract was collected by centrifugation (3000 × g, 5 min) and incubated with ANTI-FLAG M2 affinity resin for 2 h at 4 °C. Bound proteins were eluted with 0.5 mg/mL 3xFLAG peptide in benzonase buffer. For NS-EM imaging, the eluate was directly applied. For complex size analysis, the eluate was resolved by glycerol gradient sedimentation (20–40%) in buffer (50 mM HEPES-KOH [pH 7.5], 150 mM L-glutamic acid potassium salt, 8 mM magnesium chloride, 1 mM EDTA, 3 mM ATP) using a TLS-55 rotor (Beckman Optima MAX-XP Ultracentrifuge) at 105,000 × g for 13 h at 4 °C. 14 fractions were collected and analyzed by SDS-PAGE. The same protocol was used to isolate chromatin-bound MCM2-7 complexes containing 3xFLAG-tagged WT or CTE-truncated MCM3 from HCT116 mAID-MCM3 cells.

Yeast two-hybrid assay

The full-length hMCM2 was cloned into pGBKT7, while the wild-type or Q761L mutant of the hMCM3-CTE (residues 671-808) was cloned into pGADT7. Binary combinations of these pGBKT7 and pGADT7 constructs were co-transformed into the Saccharomyces cerevisiae strain AH109. Transformants were selected on synthetic complete medium lacking leucine and tryptophan (SCM-L-W). Protein-protein interactions were assessed by plating the transformants on selective medium lacking leucine, tryptophan, and histidine (SCM-L-W-H) and incubating at 30 °C for three days.

In vitro DNA pull-down assay

4 pmol of biotinylated DNA was immobilized on 10 µL of Invitrogen Dynabeads M-280 streptavidin magnetic beads in 200 µL of binding buffer (50 mM HEPES-KOH pH 7.6, 100 mM potassium glutamate, 10 mM magnesium acetate, 5 mM ATP, 1 mM DTT, 0.02% NP-40, 5% glycerol) at room temperature for 30 minutes. After immobilization, the beads were washed with binding buffer, and 4 pmol of purified TetR was added to block free DNA ends for 5 minutes. The beads were then washed three times with binding buffer. Next, purified yORC, hMCM-SH(WT), hMCM-SH(Δ3CTE), or hMCM-SH(Q761L) was added to 200 µL of binding reaction mixture to a final concentration of 20 nM and incubated with the DNA-coupled beads at 37 °C for 30 min. Following incubation, the beads were washed four times with binding buffer. DNA-bound proteins were eluted using 500 U of Micrococcal Nuclease (NEB) in 30 µL of binding buffer supplemented with 5 mM calcium chloride at 37 °C for 30 minutes. yORC was purified as previously described⁹.

Co-immunoprecipitation

The pLEX-MCM2-Twin-Strep and pLEX-MCM3-CTE-3×FLAG (or the Q761L mutant) plasmids were co-transfected into HEK 293 T cells. Cells were harvested 48 hours post-transfection, and the pellet was lysed in EBX buffer containing 75 mM potassium glutamate. After centrifugation at 3000 × g for 5 minutes, the supernatant was subjected to anti-FLAG immunoprecipitation. Bound proteins were eluted with 0.5 mg/mL 3×FLAG peptide and analyzed by Western blot.

Negative-staining EM

A total of 3 μL of protein sample was applied to glow-discharged grids and stained with 2% (w/v) uranyl acetate. Images were acquired at 57,000× magnification (2.49 Å pixel size) on a Talos L120C microscope (Thermo Fisher Scientific) operating at 120 kV with a 4k×4k Ceta CMOS camera. Micrographs were processed with RELION 3.0⁶⁹ for 2D class averaging of MCM-containing particles.

Colony formation assay

800 HCT116 cells were seeded into 35-mm plates and cultured for 9 days with medium replacement on day 4. Cells were fixed in 100% ice-cold methanol for 10 min and stained with crystal violet solution (0.5% crystal violet, 25% methanol). For HCT116 MCM5-mAID cell lines, 1 μg/mL doxycycline and 2.5 μM 5-Ad-IAA were added to the culture medium.

Growth curves

A total of 5 × 10⁵ cells were seeded into 90-mm dishes in DMEM + 10% FBS. At indicated time points, triplicate samples were trypsinized and counted using an automated cell counter (Countess 3, Thermo Fisher) to generate growth curves.

Flow Cytometry

MCM3CTE-/- HCT116 cells were collected at indicated time points, trypsinized, and fixed in 75% ice-cold ethanol for ≥30 min. After washing twice with PBS, cells were resuspended in FACS buffer (50 μg/mL propidium iodide, 100 μg/mL RNase A, 0.1% Triton X-100 in PBS) and incubated in the dark for 30 min. Samples were analyzed on a BD FACSAria™ III Cell Sorter. The results were analyzed using FlowJo™ v10.4 Software (BD Life Sciences). Gating strategy is explained in Supplementary Fig. 10.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Y. Zhai conceived and supervised the project; X.F., J.L. and M.K. constructed cell lines; X.F., Y.C.H., Z.L., H.J., W. Li, and Q.Z. purified proteins and performed functional experiments; X.F., W. Lam, D.Y., Y. Zhang, and S.D. prepared cryo samples and determined cryo-EM structures; Y. Zhai, X.F., W. Lam, Z.Y., W.W., and N.L. analyzed the data, prepared the figures, and wrote the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

Cryo-EM density maps of the human MCM SH (State I-III), MCM2-7Δ3CTE SH (State I-II), MCM2-7 (MCM3-Q761L) and related supported maps have been deposited in the Electron Microscopy Data Bank (EMDB) with accession codes EMD-64408, EMD-68350, EMD-68363, EMD-65161, EMD-65170, and EMD-68728 respectively. Atomic coordinates of the human MCM SH (State I), MCM2-7Δ3CTE SH (State I-II) and MCM2-7 (MCM3-Q761L) have been deposited in the Protein Data Bank (PDB) with accession code 9UQ0 10.2210/pdb9UQ0/pdb, 9VLN 10.2210/pdb9VLN/pdb, 9VLW 10.2210/pdb9VLW/pdb, and 22VT 10.2210/pdb22VT/pdb respectively. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-73118-9.