Mcm2 promotes stem cell differentiation via its ability to bind H3-H4

Mcm2, a subunit of the minichromosome maintenance proteins 2–7 (Mcm2-7) helicase best known for its role in DNA replication, contains a histone binding motif that facilitates the transfer of parental histones following DNA replication. Here, we show that Mcm2 is important for the differentiation of mouse embryonic stem (ES) cells. The Mcm2-2A mutation defective in histone binding shows defects in silencing of pluripotent genes and the induction of lineage-specific genes. The defects in the induction of lineage-specific genes in the mutant cells are likely, at least in part, due to reduced binding to Asf1a, a histone chaperone that binds Mcm2 and is important for nucleosome disassembly at bivalent chromatin domains containing repressive H3K27me3 and active H3K4me3 modifications during differentiation. Mcm2 localizes at transcription starting sites and the binding of Mcm2 at gene promoters is disrupted in both Mcm2-2A ES cells and neural precursor cells (NPCs). Reduced Mcm2 binding at bivalent chromatin domains in Mcm2-2A ES cells correlates with decreased chromatin accessibility at corresponding sites in NPCs. Together, our studies reveal a novel function of Mcm2 in ES cell differentiation, likely through manipulating chromatin landscapes at bivalent chromatin domains.


Introduction
Embryonic stem (ES) cells (ESCs) are pluripotent cells that possess both the ability to self-renew and the potential to differentiate into lineage-specific cell types (De Los Angeles et al., 2015). The ability of ESCs to differentiate into specific lineage cell type both in vivo and in vitro opens exciting opportunities to study the events regulating the earliest stages of lineage specification during development (Keller, 2005). During mouse ESC differentiation, pluripotency genes such as Pou5f1, Sox2, and Nanog are silenced, whereas lineage-specific genes are up-regulated (Sha and Boyer, 2008). These dynamic changes in gene expression are regulated by transcription factors as well as chromatin factors (Young, 2011). For instance, the silencing of pluripotency genes (Pou5f1, Sox2, and Nanog) is associated with both a dramatic reduction of tri-methylation of histone H3 lysine 4 (H3K4me3), a histone modification associated with active gene transcription, and an increase in H3K27me3, a repressive histone modification (Atlasi and Stunnenberg, 2017;Mikkelsen et al., 2007). In contrast, the promoters of lineage-specific genes, which are marked by both repressive H3K27me3 and active H3K4me3 marks, needs to be resolved for silencing and/or induction of lineage-specific genes during differentiation (Bernstein et al., 2006;Harikumar and Meshorer, 2015). Although significant processes have been made to understand these dynamic changes in chromatin states during differentiation, the regulation of these dynamic changes remains underexplored.
Histone chaperones, a group of proteins best known for their roles in the assembly of DNA into nucleosomes following DNA replication and gene transcription, have been found to play multiple roles in stem cell maintenance and differentiation. For instance, chromatin assembly factor 1 (CAF-1), a histone chaperone involved in the deposition of newly synthesized H3-H4 onto replicating DNA, is important for cell fate maintenance (Ishiuchi et al., 2015;Smith and Stillman, 1989). Depletion of subunits of the CAF-1 complex in embryonic fibroblasts results in increased reprogramming efficiency into iPSC cells (Cheloufi et al., 2015). In ESCs, CAF-1 depletion leads to an increase in the percentage of 2C-like cells (2-cell-stage embryos), as well as defects in differentiation (Cheng et al., 2019;Ishiuchi et al., 2015). The role of CAF-1 in differentiation and the 2C-like state is linked to the function of CAF-1 in DNA replication-coupled nucleosome assembly. In contrast, Asf1a, a histone chaperone involved in the delivery of newly synthesized H3-H4 to CAF-1 in replication-coupled nucleosome assembly, as well as to HIRA, a histone chaperone involved in replication-independent (RI) nucleosome assembly (De Koning et al., 2007;English et al., 2006;Tagami et al., 2004), regulates the disassembly of nucleosomes at bivalent chromatin domains for the induction of lineage-specific genes (Gao et al., 2018). Moreover, histone chaperone CAF-1 promotes the formation of H3K27me3mediated silencing at pluripotent genes (Cheng et al., 2019), whereas HIRA facilitates the PRC2 silencing complex at developmental loci during differentiation (Banaszynski et al., 2013;Ray-Gallet et al., 2002). The H3.3-HIRA pathway also safeguards identities of differentiated cells, indicating its bimodal role in cell fate transition (Fang et al., 2018). Therefore, histone chaperones involved in deposition of newly synthesized H3-H4 play multiple roles in ESC differentiation and maintenance.
In addition to histone chaperones involved in the deposition of newly synthesized histones, we and others have also uncovered a group of proteins that function in recycling parental histone H3-H4 following DNA replication. Pole3 and Pole4, two subunits of leading strand DNA polymerase ε, interact with H3-H4 and facilitate the transfer of parental histones to leading strands of DNA replication forks in both yeast and mouse ESCs (Li et al., 2020;Yu et al., 2018). On the other hand, Mcm2, a subunit of the minichromosome maintenance proteins 2-7 (Mcm2-7) complex that plays an essential role in DNA replication as the replicative helicase (Tye, 1999), contains a histone binding domain (HBD) (Huang et al., 2015). Mutations at HBDs (Y81A and Y90A) of  in both yeast and mouse ESCs leads to defects in the transfer of parental histones to lagging strands of DNA replication forks Li et al., 2020;Petryk et al., 2018). Moreover, Mcm2 interacts with Asf1 that is bridged by H3-H4 proteins (Groth et al., 2007). Besides its functions in DNA replication and histone deposition, Mcm2 also interacts with the carboxyl-terminal domain of RNA Pol II in Xenopus oocytes, and the Mcm2-7 complex is required for RNA Pol II-mediated transcription at some settings in mammalian cells (Snyder et al., 2009;Yankulov et al., 1999), indicating a possible role of Mcm2 in gene transcription.
Mouse ESCs with deletion of Pole3 and Pole4 or mutations at the HBD of Mcm2 grow normally and maintain stemness. As parental histones with their histone modifications are the blueprint for the recapitulation of the epigenetic landscape during cell division (Corpet and Almouzni, 2009), we investigated the roles of these parental histone chaperone proteins during differentiation of mouse ESCs. We found that Mcm2, relying on its HBD, promotes mouse ESC differentiation. The Mcm2-2A mutant ESCs show defects both in silencing of pluripotent genes and induction of lineage-specific genes during differentiation. The defects in induction of lineage-specific genes are associated with reduced Asf1a binding in Mcm2-2A mutant cells. Mcm2 localizes at transcription starting sites (TSS) and that this localization is dramatically reduced in Mcm2-2A cells, which correlates with reduced chromatin accessibility at bivalent chromatin domains during differentiation. Together, these studies reveal a novel role for Mcm2 and its ability to bind H3-H4 in the differentiation of mouse ESCs.  . It is known that CAF-1, Asf1a, and HIRA function in stem cell differentiation through distinct mechanisms (Banaszynski et al., 2013;Cheloufi et al., 2015;Cheng et al., 2019;Gao et al., 2018). We therefore asked whether histone chaperones involved in parental histone transfer are also important in this process. To do this, we first monitored the formation of embryoid bodies (EBs), which mimics the formation of three germ layers in vitro, in Pole3 KO, Pole4 KO, Mcm2-2A single and double mutant mouse ESCs (Li et al., 2020;Figure 1-figure supplement 1A). Briefly, three-dimensional colonies (EBs) were formed in hanging drops without leukemia inhibiting factor (LIF) for 3 days. Then, EBs were cultured in suspension and collected at different times during differentiation for the evaluation of morphology and expression of selected genes involved in stemness and lineage specification ( Figure 1A). We observed that all mutants exhibited reduced EB size at day 9 compared to wild type ( Depletion of CAF-1 results in an increase in totipotent 2C-like cells, which are marked by reduced expression of Pou5f1 and increased expression of the endogenous retrovirus MERVL (Ishiuchi et al., 2015). We thus assessed whether Mcm2-2A mutants altered cellular plasticity in ESCs by analyzing the expression of Pou5f1 and MERVL-Gag using immunofluorescence. We found that Mcm2-2A mutant and WT cells showed similarly low frequencies of 2C-like cells ( Figure 1D and E), indicating that ESC plasticity is not altered in Mcm2-2A mutant cells. Together, these results suggest that the ability of Mcm2 to bind histone H3-H4 plays an important role during mouse ESC differentiation.

Mcm2-2A mutation compromises mouse ESC differentiation into neural lineages
To further study the function of Mcm2 and Pole4 during ESC differentiation, we differentiated WT, Mcm2-2A, and Pole4 KO single and double mutant ESCs into neural lineage cells as previously described (Gao et al., 2018). We then compared expression levels of two pluripotency genes (Pou5f1 and Nanog) and two neural lineage-specific genes (Sox21 and Pax6) in WT, Mcm2 Because Mcm2-2A mutant ESCs showed consistent defects in differentiation in both EB formation and neural differentiation assays, we focused the remaining studies on this mutant. First, we monitored morphology changes and silencing of the EGFP reporter gene driven by the Pou5f1 distal enhancer during neural differentiation in WT and Mcm2-2A cells (Hotta et al., 2009). The silencing of EGFP reporter and the cell morphologies were comparable between WT and Mcm2-2A cells from day 2 to day 6 ( Figure    Mcm2 and Asf1a function in the same pathway for the induction of lineage-specific genes Mcm2 interacts with histone chaperone Asf1, which is bridged by H3-H4 (Huang et al., 2015). Consequently, the Mcm2-Asf1 interaction is reduced in Mcm2-2A cells. We have shown that deletion of Asf1a, but not Asf1b, in mouse ESCs impairs the induction of lineage-specific genes (Gao et al., 2018). To understand how Mcm2 These results indicate that Mcm2 likely have two roles during differentiation, silencing of pluripotent genes and induction of lineage-specific genes, the latter of which is mediated, at least in part, through Mcm2-H3-H4-Asf1a interaction. This idea was further supported by transcriptome analysis described below. independent experiments. Statistical analysis in C and E was performed by two-tailed unpaired Student's t test with p values marked on the graphs (N.S., no significant difference).
The online version of this article includes the following source data and figure supplement(s) for figure 1: Source data 1. Relative mRNA level of pluripotency and lineage-specific genes during embryoid body (EB) formation in wild type (WT) and Mcm2     Source data 2. Whole SDS-PAGE images and uncropped blots represented in Figure 2D.
Source data 3. Whole SDS-PAGE images and uncropped blots represented in Figure 2E.        . These results support the idea that Mcm2 and Asf1a function together for the induction of lineage-specific genes expression during differentiation. Collectively, these results suggest that Mcm2's function in promoting ESC differentiation is partially dependent on Asf1a. However, Mcm2 and Asf1a also have independent roles in gene regulation, with Mcm2, but not Asf1a, participating in the silencing of pluripotency genes during differentiation.

Mcm2-2A mutation disrupts the epigenetic landscape in differentiated cells
During differentiation, histone modification landscapes are rewired (Mikkelsen et al., 2007). For instance, bivalent domains that are enriched with both active markers (H3K4me3) and silencing markers (H3K27me3), which are associated with lineage-specific genes, are resolved either through the removal of repressive marker H3K27me3 for gene activation or removal of active marker H3K4me3 for gene silencing (Voigt et al., 2013). In addition, pluripotency genes are silenced through gain of H3K27me3 and loss of H3K4me3 at promoters (Bernstein et al., 2006;Harikumar and Meshorer, 2015). Therefore, we analyzed the impact of the Mcm2-2A mutation on the total levels of H3K27me3 and H3K4me3 during neural differentiation. We observed that in both WT and Mcm2

Mcm2 is enriched at actively transcribed regions in both ESCs and NPCs
To further explore the mechanisms underlying the transcriptome and epigenetic changes in Mcm2  We also noticed the relatively low signal-to-noise ratio for the Mcm2 and Flag-Mcm2 CUT&RUN peaks. Once loaded on chromatin during G1 phase of the cell cycle, it is known that the CMG helicase travels along with DNA replication forks (Prioleau and MacAlpine, 2016). On the other hand, Mcm2 CUT&RUN experiments were performed using asynchronous ESCs. Therefore, the relatively low signal-to-noise ratio of Mcm2 CUT&RUN peaks likely reflect the dynamic nature of chromatin binding of the CMG helicase during the cell cycle progression. Nonetheless, because of low signal-to-noise ratio, cautions should be made for the interpretation of Mcm2 CUT&RUN results described below.
To gain insight into these Mcm2 CUT&RUN peaks, we aligned and clustered these Mcm2 CUT&RUN peaks based on their overlaps with H3K4me3 and H3K27me3 CUT&RUN signals and ATAC-seq peaks in ESCs ( Figure 5A). We observed that the majority of Mcm2 CUT&RUN peaks were enriched with H3K4me3 CUT&RUN signals and ATAC-seq peaks. A small number of Mcm2 CUT&RUN peaks were found at bivalent chromatin domains (H3K4me3+ and H3K27me3+), with far fewer peaks co-localizing with H3K27me3 markers (H3K4me3-, H3K27me3+) in ESCs ( Figure 5A, left). However, when taking into consideration of total number of H3K4me3, H3K27me3, and bivalent domains in ESCs, we observed that over 26% bivalent domains had at least one Mcm2 peaks nearby compared with 21% H3K4me3 peaks having a Mcm2 peak ( Figure 5B). We also performed similar analysis on 2686 of Mcm2 peaks identified in WT NPCs and observed that almost all the Mcm2 peaks co-localized with H3K4me3 and ATAC-seq peaks (H3K4me3+, H3K27me3-) ( Figure 5A Figure 5E). Together, these results suggest that alterations in chromatin binding of Mcm2-2A mutant proteins at ESCs likely contribute to its altered chromatin binding in NPCs.
We also asked whether the changes in Mcm2 binding in ESCs and NPCs correlated with changes in gene expression in ESCs and in NPCs. We found that the average level of Mcm2-2A density at the TSS of down-regulated genes in ESCs was significantly lower than that in WT ESCs, but Mcm2 density at up-regulated genes was similar between WT and Mcm2-2A mutant ESCs ( Figure 5F). Similar results down-regulated and up-regulated genes in Mcm2-2A NPCs (Figure 3A, right). The Y-axis represents the log2 ratio of CUT&RUN density (reads per kilobase per million reads [RPKM]). The p values were calculated using Wilcoxon signed-rank test. The average of two independent replicates is shown.
The online version of this article includes the following source data and figure supplement(s) for figure 4: Figure 4A.   were obtained in NPCs ( Figure 5G). In addition, Mcm2 CUT&RUN density in Mcm2-2A mutant ESCs was reduced at the down-regulated genes in NPCs ( Figure 5H), as exemplified at the promoter of neural lineage gene Pax6, where Mcm2 CUT&RUN density was down-regulated in mutant NPCs compared to WT NPCs ( Figure 5I). This finding suggests that the reduced association of Mcm2 with chromatin in Mcm2

Mcm2 facilitates chromatin accessibility during mouse ESC differentiation
The landscape of chromatin accessibility dynamically changes during the development (Trevino et al., 2020).  Figure 6A). Moreover, the average level of ATAC-seq signals at the TSS of down-regulated genes in Mcm2-2A mutant NPCs was significantly lower than WT NPCs. The opposite was true for up-regulated genes ( Figure 6B, Figure 6-figure supplement 2A). These results are consistent with the idea that chromatin accessibility, as detected by ATAC-seq, is linked to gene transcription. Next, we explored the correlation between Mcm2 binding and ATAC-seq density. Based on Figure 5D, we first separated Mcm2-2A CUT&RUN peaks in ESCs and in NPCs into three categories (increased, reduced, and non-differential Mcm2 peaks compared to WT Mcm2 ESCs and NPCs, respectively) and calculated ATAC-seq density at each group of Mcm2 peaks. We observed a dramatic reduction of ATAC-seq density at the reduced Mcm2 peak group in Mcm2-2A mutant NPCs, along with a slight increase of ATAC-seq density at increased Mcm2 peak groups ( Figure 6C and D). Of note, ATAC-seq signals at increased Mcm2 peaks were very low ( Figure 6C and D). In ESCs, ATAC-seq density at down-regulated Mcm2 peaks in Mcm2-2A mutant cells was also reduced, but to a far lesser extent than in NPCs (Figure 6-figure supplement 2B, C). These results support the idea that Mcm2 is important for chromatin accessibility in both ESCs and potentially in NPCs.
Since Mcm2 (Figure 6-figure supplement 2D), ATAC-seq signals were slightly reduced in Group 2 (H3K4me3+, H3K27me3-) ( Figure 6-figure supplement 2E, F), suggesting that the reduced Mcm2 binding does not affect chromatin accessibility dramatically. In NPCs, Mcm2 CUT&RUN density in Mcm2-2A mutant cells was reduced in both Group 1 and Group 2, with Group 1 showing a larger reduction than Group 2 ( Figure 6-figure supplement 2G, H). Importantly, ATAC-seq signals in Mcm2-2A NPCs at Group 1 and Group 2 Mcm2 peaks were significantly reduced compared to WT NPCs, with a larger reduction in Group 1 than in Group 2, whereas ATAC-seq signals in Mcm2-2A NPCs were increased slightly at Group 3 and Group 4 Mcm2 peaks ( Figure 6E and F). As Mcm2 Group 1 peaks co-localize with bivalent chromatin domains (H3K4me3+ and H3K27me3+), these results indicate that a reduction of Mcm2 binding in Mcm2-2A mutant ESCs, while having little effects on chromatin accessibility in ESCs, perturbs chromatin changes at bivalent chromatin domains during differentiation. Together, these results support the idea that Mcm2 binding at bivalent chromatin domains in ESCs is important for the resolution of these regions for subsequent gene activation during differentiation.

Discussion
We and others have previously shown that Mcm2, Pole3, and Pole4, three replisome components first known for their roles in DNA replication, function in the transfer of parental H3-H4 following DNA replication in yeast and mouse ESCs Li et al., 2020;Petryk et al., 2018;Xu et al., 2022;Yu et al., 2018). Remarkably, mouse ESCs with deletion of Pole3, Pole4, or with mutations at the histone binding motif of Mcm2 (Mcm2-2A) largely grow normally. Here, we found that these mutant cells all exhibit defects in differentiation, revealing a novel role of these proteins in ES differentiation.
Mcm2 is a subunit of MCM helicase consisting of Mcm2-7. The MCM helicase is loaded on chromatin at the G1/S transition and serves as the core of the CMG replicative helicase to unwind doublestranded DNA for DNA synthesis during the S phase of the cell cycle (Tye, 1999). The N-terminus of Mcm2 contains a conserved histone binding motif that interacts with H3-H4 (Huang et al., 2015). Mutations at this histone binding motif (Mcm2-2A) that impair Mcm2's ability to bind H3-H4 lead to a dramatic enrichment of parental H3-H4 at leading strands compared to lagging strands of DNA replication forks during early S phase of the cell cycle Li et al., 2020;Petryk et al., 2018). Using two different in vitro differentiation assays, we observed that Mcm2-2A mutant mouse ESCs, while growing normally, showed dramatic defects during differentiation. Furthermore, we found that the Mcm2-2A mutation induces global changes in gene expression, chromatin accessibility, and histone modifications during differentiation. Together, these studies reveal a novel role of Mcm2, through its interaction with H3-H4, during differentiation.
Previously, we have shown that deletion of two histone chaperones involved in deposition of newly synthesized H3-H4, Asf1a and the p150 subunit of the CAF-1 complex, also impairs differentiation of mouse ESCs (Cheng et al., 2019;Gao et al., 2018). Compared to CAF-1 p150 KO and Asf1a KO cells, the Mcm2-2A mutation exhibited distinct defects in differentiation. p150 KO mutant ESCs show defects in both the silencing of pluripotent genes and induction of lineage-specific genes during differentiation, with defects in the former much more pronounced (Cheng et al., 2019). In contrast, Asf1a KO ESCs minimally impact the silencing of pluripotent genes and dramatically affect the induction of lineage-specific genes (Gao et al., 2018). Silencing of pluripotent genes and induction of lineagespecific genes are both defective in Mcm2-2A cells, with the defects in the latter more pronounced. These results suggest that CAF-1, Mcm2, and Asf1a likely perform non-overlapping functions during differentiation.
First, our results indicate that Mcm2 and Asf1a function in the same pathway for the induction of lineage-specific genes. For instance, we found that Mcm2-2A Asf1a KO double mutant cells showed similar defects in induction of lineage-specific gene as Mcm2-2A and Asf1a KO single mutant alone, suggesting that Mcm2 and Asf1a function in the same pathway for the induction of lineage-specific genes during differentiation. Second, transcriptome analysis by RNA-seq indicate that a large fraction    (Huang et al., 2015). Furthermore, we found that Asf1a is important to resolve bivalent chromatin domains through its ability for nucleosome disassembly (Gao et al., 2018). Consistent with this idea, chromatin accessibility at bivalent chromatin domains is reduced the most in Mcm2-2A mutant cells during differentiation, and this reduction is linked to the reduced Mcm2 at bivalent chromatin domains in ESCs, indicating that in addition to its role in parental histone transfer, Mcm2 also has a role in dynamic changes in chromatin during cell fate transition. Therefore, we suggest that Mcm2 and Asf1a function together at bivalent chromatin domains for the induction of lineage-specific genes during differentiation.
Second, our results indicate that Mcm2-2A mutant ESCs also show defects in differentiation in Asf1a-independent manner. For instance, silencing of pluripotent genes such as Pou5f1 and Nanog was defective in Mcm2 (Huang et al., 2015;Li et al., 2020;Petryk et al., 2018) in ESCs. Therefore, the defects in silencing of pluripotent genes in Mcm2-2A mutant cells may be due to defects in the recycling of parental H3 during differentiation. Consistent with this idea, mutations at Pole3 and Pole4, two other genes involved in recycling both parental H3.3 and H3.1 in mouse ESCs following DNA replication (Xu et al., 2022), also result in defects in differentiation. Currently, the fate of parental histones at different chromatin domains during mouse ESC differentiation is largely unknown. It is possible that the ability of Mcm2 to recycle both parental H3.1 and H3.3 contributes to dynamic changes in chromatin states during differentiation. Alternatively, it is possible that Mcm2, via its ability to bind H3-H4, regulates gene expression directly during differentiation. Supporting this idea, we have shown that Mcm2 localizes at gene promoters of actively transcribed genes in both ESCs and NPCs. Furthermore, MCM2-7 proteins are associated with RNA Pol II in Xenopus and HeLa cells (Snyder et al., 2009;Yankulov et al., 1999), and Mcm2 and Mcm5 are required for Pol II-mediated transcription (Snyder et al., 2009). Finally, it is possible that defects in silencing of pluripotent genes in Mcm2-2A mutant cells are associated with firing of dormant origins. It is known that excessive amounts of MCM2-7 complexes than active origins are loaded on chromatin during G1 phase of the cell cycle. Dormant origins provide a first line of defense for the genome under replication stress. ESCs contain more dormant origins than progenitor cells such as NPCs (Ge et al., 2015). Consistent with this idea, we detected 13742 and 2686 Mcm2 CUT&RUN peaks, respectively, in WT ESCs and NPCs. Furthermore, activation of dormant origins is impaired in Mcm2 depleted cells (Ibarra et al., 2008), and that partial depletion of Mcm4 and Mcm5 does not affect ESCs self-renewal but impairs their differentiation, including toward the neural lineage (Ge et al., 2015). However, unlike depletion of MCM subunits in these studies, Mcm2-2A mutation in yeast and mouse ESCs has little impacts on the response to replication stress. Therefore, it is unlikely that defects in differentiation in Mcm2-2A mutant cells are largely due to its impact on firing of dormant origins. Nonetheless, future studies are needed to address whether Mcm2's role in differentiation, in particular its role in silencing of pluripotent genes during differentiation is linked to its ability to bind H3-H4.
In summary, our results indicate that Mcm2, through its interaction with H3-H4, likely regulates differentiation through multiple mechanisms, one of which involves in the resolution of bivalent chromatin domains through its interaction with Asf1a bridged by H3-H4.

Materials availability statement
Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Zhiguo Zhang (zz2401@cumc.columbia.edu).
To generate pluripotency EGFP reporter mouse E14 WT and Mcm2-2A ESC line, the lentivirusbased EGFP reporter vector PL-SIN-EOS-C(3+)-EGFP plasmid (21318, Addgene) was used for infection. After selection, single cells were seeded and individual clones were then isolated, expanded, and confirmed under fluorescence microscope.
To generate mouse E14 WT and Mcm2-2A ESC lines expression Mcm2 or Asf1a ectopically, the lentivirus-based vector pWPXL empty vector, pWPXL-Mcm2, or pWPXL-Asf1a plasmids were used for infection. After selection, single cells were seeded and individual clones were then isolated, expanded, and confirmed using Western blot.

EB assay
Mouse ESCs were disaggregated and suspended in ESC medium without LIF. EBs were formed using the hanging drop method (300 cells per drop) on dish lids for 3 days. EBs were then collected and cultured in 10 cm low attachment Petri dish in ESC culture medium without LIF, and the medium was changed every other day. Samples were collected at the indicated time points for analysis of gene expression.

Immunofluorescence
Cells were seeded on coverslip coated with 1% gelatin, and then fixed in 4% of formaldehyde for 15 min at room temperature (RT). After washing with PBS, fixed cells were permeabilized with 0.1% Triton X-100 in PBS (PBST) for 10 min and blocked for 1 hr with 5% normal goat serum (NGS) in PBST at RT. Cells were incubated with primary antibodies diluted in 1% NGS in PBST overnight at 4°C. Cells were then washed with PBST and incubated with fluorophore-labeled secondary antibodies for 1 hr at RT. DNA were stained with DAPI. Images were captured by Nikon 80i Fluorescence Microscope.

RT-PCR analysis
Total RNA was isolated from 1×106 cells using RNeasy Plus kit (74104, Qiagen); 0.5 μg of total RNA were used for cDNA synthesis with random hexamers (18080-051, Invitrogen). Quantitative PCR was performed in 12 μL reactions containing 0.1 μM primers and SYBR Green PCR Master Mix (Bio-Rad Laboratories). Primers used are listed below.

Neural differentiation assay
The neural differentiation of ESCs was performed as previously described with some modifications (Gao et al., 2018;Ying et al., 2003). Briefly, ESCs (day 1) were seeded at high density (2×10 6 cells/60 mm dish) onto gelatin-coated dish in standard mES medium with LIF for 24 hr. The differentiation was initiated by seeding ESCs (day 2) in N2B27 medium at a density of 3×10 5 cells per 60 mm dish. The N2B27 medium was changed every day up to day 6. Cells were cultured continuously in N2B27 medium supplied with EGF (10 ng/ml, R&D) and FGF-2 (10 ng/ml, R&D) for 3 more days with medium changed every day. Neural precursors were collected at day 9 for analysis.

FACS analysis of GFP ratio
For GFP ratio analysis, exponentially growing mouse ESCs expressing EGFP reporter gene driven by Pou5f1 distal enhancer were collected and washed in PBS. Samples were analyzed by Attune NxT software of Attune flow cytometer (Thermo Fisher Scientific). Data were analyzed by FCS Express (version 7).

CUT&RUN and ATAC-seq analysis
CUT&RUN and ATAC libraries were constructed and sequenced using paired-end method on Illumina platforms. Adaptor sequences of all raw reads were removed by Cutadapt (Marcel, 2011) and reads <10 nt were removed. CUT&RUN and ATAC-seq data were then mapped to mouse (mm10) reference genome by Bowtie2 (Langmead and Salzberg, 2012). Multi-mapped reads were removed using SAMtools (MAPQ <40) (Li et al., 2009) and duplicate reads were removed using Sambamba software (Tarasov et al., 2015). Read coverage in a bin of 1 bp was calculated from filtered bam files by deepTools2 (Ramírez et al., 2016) and then normalized with total filtered reads number into reads per million. Genome-wide correlation was performed by deepTools2 (Ramírez et al., 2016) in a bin of 5000 bp. Peaks were called by MACS (Zhang et al., 2008) by parameters 'macs2 callpeak -g mm -f BAMPE -p 1e-04 --broad --broad-cutoff 1e-04 --llocal 10000 -nolambda' and the cutoff of peak was p=0.0001. Read density level surrounding gene promoters ([-3kb, 3kb] of TSS) or MCM2 peaks was calculated by featureCounts (Liao et al., 2014) and then normalized to reads per kilobase per million reads (RPKM). Heatmap clustering was performed by "ward.D2" (Murtagh and Legendre, 2014) method based on z score of log10(RPKM). To identify differential CUT&RUN peaks, peaks from both WT and mutant cells were first merged to a union pool and then read counts were calculated in the merged peaks by featureCounts (Liao et al., 2014). DESeq2 (Love et al., 2014) was then used to identify differential peaks by |log2 fold change|>1.

RNA-seq analysis
Total RNA from WT and Mcm2-2A ESCs and NPCs were isolated from 1×106 cells using RNeasy Plus Mini kit (74136, Qiagen). RNA-seq libraries were prepared and deep sequenced in Columbia University Genome Center. Two replicates for each sample were sequenced. RNA-seq libraries were sequenced using paired-end method on Illumina platforms. The paired-end reads of WT and Asf1a KO ESCs and NPCs were downloaded from GSE114424. Adaptor sequences of all raw reads were removed by Cutadapt (Marcel, 2011) and reads <10 nt were removed. RNA-seq data were mapped to the mouse (mm10) reference genome by STAR software (Dobin et al., 2013). Gene expression levels were first calculated by featureCounts (Liao et al., 2014) to obtain read counts and then normalized with total mapped reads into RPKM. Differential expressed genes were identified by DESeq2 (Love et al., 2014) by using adjusted p-value <0.01 and |log2 fold change|>1. GO enrichment analysis was performed by 'cluserProfiler' (Yu et al., 2012) in a level 4 of 'biological process' functions.

Statistical analyses
Data are presented as means ± SD. Differences between groups were evaluated using two-tailed unpaired Student's t test (noted in figure legends). Statistical analysis was performed in GraphPad Prism software (version 7). All tests were considered significant at p<0.05. For all the sequenced data analysis, the statistical test was performed using R software. Difference test between groups was evaluated by Wilcoxon signed-rank test. The center line is the medians of all data points, with the limits corresponding to the upper and the lower quartiles, respectively, and the whiskers representing the largest and smallest values. Where outliers were removed for plotting purposes, the removed data points were still used for statistical analyses.