Post-replicative chromatin accessibility predicts cell fate change

Summary

It has long been hypothesized that DNA replication is important in reconfiguring the chromatin landscape during cell identity changes in development, disease, and reprogramming. There is now a large body of work showing that DNA replication indeed alters chromatin structure and composition, but a function for these changes has remained elusive. Using replication-coupled ATAC-seq in differentiating embryonic stem cells and reprogramming mouse embryonic fibroblasts, we profiled replicated and unreplicated chromatin and observed de novo chromatin opening specifically in the replicated fraction. These opening events created an accessibility landscape similar to that seen in later time points, and binding of lineage-specific transcription factors was enriched in these regions. Opening of these regions was impaired when replication was inhibited during early reprogramming. This work bridges the gap between replication-induced structural chromatin changes and functional consequences by demonstrating that replication facilitates a “window of opportunity” that advances the chromatin landscape during cell identity change.

Graphical abstract

Highlights

• •Lineage-specific regulatory elements show early accessibility in replicated chromatin
• •Replicated chromatin is globally advanced in the first S-phase of reprogramming
• •Chromatin opening is impaired when replication is inhibited
• •Proliferation rate contributes to reprogramming potential in induced pluripotency

Knudsen and colleagues use repli-ATAC-seq to compare replicated and unreplicated chromatin in two models of cell identity change. They find that lineage-specific elements are accessible earlier in replicated chromatin. Chromatin opening is impaired when DNA replication is inhibited, and proliferation history correlates with reprogramming potential. Collectively, this demonstrates that DNA replication generates accessibility at key regulatory regions during cell fate change.

Introduction

The foundation of cell identity and function is its pattern of gene expression. The master regulators of cell identity are therefore transcription factors, proteins that bind specific DNA sequences to control transcription. As a result, alterations to the transcription factors expressed in a cell can drive alterations in cell identify. In development and during differentiation, transcription factor activity organizes spatiotemporal expression patterns (Spitz and Furlong, 2012). The finding that ectopic expression of transcription factors can manipulate cell fate (Davis et al., 1987; Takahashi and Yamanaka, 2006; Xie et al., 2004) created an entire field devoted to manipulating and reprogramming cellular identity and has led to the discovery of many transcription factor “cocktails” that drive cell fate change (Wang et al., 2021). The mechanisms underlying transcription factor-mediated cell state changes are key to understanding how cells change their identity.

Cell fate change starts with transcription factor binding, and although some transcription factors can bind DNA wrapped around nucleosomes (Zaret, 2020), binding is generally inhibited by chromatin. The most reliable chromatin-based readout of gene expression is open chromatin profiling, where DNA’s relative accessibility to the transcription machinery is mapped genome-wide using nuclease or transposase assays (Furey, 2012). Transcriptionally active chromatin is more open and accessible than repressed chromatin; active and repressed chromatin also features differing DNA methylation, histone post-translational modification (PTM), and nucleosome density profiles (Millán-Zambrano et al., 2022; Smith et al., 2024). Establishing a new gene expression profile during cell identity transitions requires modulating chromatin, both activating repressed regions and repressing active ones.

One of the most profound and extensively studied cell state changes is the entry and exit from pluripotency, a cell state characterized by the potential to differentiate into any somatic cell type. This state exists transiently in the epiblast of the early embryo but can be captured indefinitely in vitro, in pluripotent or embryonic stem cell (PSC/ESC) cultures. Pluripotency becomes progressively restricted in development, with naive pluripotency existing pre-implantation and primed pluripotency existing immediately prior to the initiation of germ layer differentiation. This transition involves significant epigenome remodeling, where chromatin shifts from a largely open, euchromatic state to one where heterochromatin begins to form, restricting differentiation toward the somatic lineages (Argelaguet et al., 2019). In contrast, the transition from somatic cell types into early embryonic-like pluripotent states requires substantial removal of repressive chromatin. Despite the profound changes in chromatin architecture during differentiation and reprogramming, the underlying cellular processes that elicit these changes remain poorly characterized.

An attractive mechanism to achieve large-scale epigenome remodeling is the massive disruption to chromatin and DNA methylation posed by DNA replication (Stewart-Morgan et al., 2020; 2023). Early work showed that chromatin becomes particularly nuclease-sensitive after replication (Seale, 1975; Worcel et al., 1978), likely a consequence of the hyperacetylated state of H3-H4 tetramers synthesized and assembled onto sister chromatids post-replication. This is a phenomenon observed in bulk chromatin; we and others have seen that accessibility immediately after replication decreases rather than increases at regulatory elements, at least in cell populations not undergoing cell identity change (Ramachandran and Henikoff, 2016; Stewart-Morgan et al., 2019). Simultaneously, the deposition of newly synthesized histones dilutes histone PTM levels (Alabert et al., 2015), potentially weakening repression within transcriptionally silenced domains (Stewart-Morgan et al., 2020). In support of a role for replication, many cell identity transitions require the cell cycle and feature increased proliferation (Perera et al., 2022; Soufi and Dalton, 2016; Wong et al., 2023). Moreover, manipulation of replication-coupled chromatin assembly has profound effects on stem cell self-renewal and induced cell fate change (Cheloufi et al., 2015; Franklin et al., 2022; 2025; Ishiuchi et al., 2015). Based on these observations, DNA replication might create a “window of opportunity” during cell identity change: the short interval post-replication where chromatin is hyperacetylated, repressive chromatin marks are diluted, and DNA is partly hemi-methylated could be exploited by key transcription factors to bind otherwise-inaccessible target loci and precipitate transcription program changes, initiating a stable cell fate change.

As a test of the “window of opportunity” theory of DNA replication in cell identity change, we used replication-coupled ATAC-seq (repli-ATAC-seq) (Stewart-Morgan et al., 2019) to investigate the accessibility landscape of replicated chromatin in two models of cell fate change that feature increased proliferation. We find that during both naive-to-primed differentiation of ESCs and in fibroblast reprogramming to induced PSCs (iPSCs), the first round of post-replication chromatin maturation exhibits chromatin changes that become fixed in bulk chromatin in subsequent cell cycles. By profiling accessibility in fast- and slow-dividing subpopulations of reprogramming fibroblasts, we further show that proliferation heterogeneity can predict reprogramming potential early in induced pluripotency.

Results

Replication supports de novo chromatin opening during the transition from naive to primed pluripotency

To assess whether replication enhances the remodeling of chromatin accessibility in cell identity change, we profiled replication-dependent changes in the transition from naive to primed pluripotency.

Differentiation of naive ESCs to primed epiblast-like cells (EpiLCs) is achieved efficiently within 48 h following the removal of ESCs from naive conditions and their introduction to primed conditions (Buecker et al., 2014; Hayashi et al., 2011) (Figure 1A). While naive cells are highly proliferative prior to differentiation, they exhibited a steady increase in the fraction of cells in S-phase (5-ethynyl-2′-deoxyuridine [EdU] positive) as early as 9 h following the shift to primed conditions (Figures 1B and S1A).

Figure 1.

A subset of peaks precociously gains enrichment post-replication during naive-to-primed differentiation

(A) Overview of EpiLC induction strategy.

(B) Proportion of EdU+ cells across the time course of EpiLC induction, quantitated by flow cytometry. Line, mean of three independent experiments.

(C) Schematic of repli-ATAC-seq strategy.

(D) Replicated and unreplicated repli-ATAC- and ATAC-seq Reads Per Kilobase per Million mapped reads (RPKM)-normalized signal across EpiLC differentiation (0h, yellow; 9h, orange; 48h, red) at the Nanog locus (left) and Sox9 locus (right).

(E) Principal-component analysis of chromatin accessibility variance across consensus peaks in replicated (green) and unreplicated (gray) datasets.

(F) Central: Venn diagram of overlap between peaks called in unreplicated 0h ATAC-seq (“Naive”) and unreplicated 48h ATAC-seq (“Primed”). (Left) Boxplots of replicated repli-ATAC-seq (top) (Fisher’s median test p values for 0h vs. 9h = 1.3e−131 and 9h vs. 48h = “0”; R precision limit <1e−300) and unreplicated ATAC-seq (bottom) (Fisher’s median test p values for 0h vs. 9h = “0” and 9h vs. 48h = “0”; R precision limit <1e−300) signal over unique naive peaks. (Right) Boxplots of replicated repli-ATAC-seq (top) (Fisher’s median test p values for 0h vs. 9h = 2.6e−304 and 9h vs. 48h = “0”; R precision limit <1e−300) and unreplicated ATAC-seq (bottom) (Fisher’s median test p values for 0h vs. 9h = 5.8e−166 and 9h vs. 48h = “0”; R precision limit <1e−300) signal over unique primed peaks. Horizontal black line: median; black dot: mean; boxes: 25^th–75^th percentiles; vertical black lines: 1.5× interquartile range.

(G) Average accessibility in RPKM over the identified precocious peaks (as defined in Figure S1H) at 0h, 9h, and 48h in unreplicated (left) and replicated (right) chromatin.

(H) Top 5 motifs enriched significantly in precocious peaks relative to all opening peaks (p value = 0.01 for all).

(I) Venn diagrams and fractions of precocious peaks (490 regions) and all opening sites (23,195 regions) that overlap with OTX2 binding (43,563 binding sites) in primed conditions (Buecker et al., 2014).

See also Figure S1.

We performed repli-ATAC-seq before (0h), during (9h) and after (48h) the replication burst upon induction of differentiation of naive ESCs (Figures 1B and S1A). At each time point, we profiled the chromatin landscape in replicated and unreplicated chromatin (Figure 1C). We generated “replicated” repli-ATAC-seq libraries by chasing a 10-min EdU pulse for 2 h, an interval we previously showed was sufficient to restore transcription and transcription factor occupancy on replicated chromatin (Stewart-Morgan et al., 2019). As a reference, we generated conventional, “unreplicated” ATAC-seq libraries using EdU-negative chromatin at each time point (see methods). After sequencing and quality filtering, we identified a total of 228,226 consensus peaks across the replicated and unreplicated chromatin datasets. We additionally profiled nascent chromatin with repli-ATAC-seq (Figure S1B) and observed the same loss of signal observed previously in homeostatic ESCs (Stewart-Morgan et al., 2019) (Figure S1C).

When inducing primed differentiation in naive cells, we observed the expected downregulation of the naive marker NANOG and upregulation of the primed marker OCT6 (Figure S1D). Additionally, changes in our unreplicated ATAC-seq between naive (0h) and primed (48h) showed a positive Pearson correlation (0.75) with previously mapped global gene expression changes between these time points (Figure S1E) (Yang et al., 2019). Consistently, we observed loss of accessibility at the Nanog locus and concurrent gain of accessibility at the Sox9 locus (primed marker) (Figure 1D). Principal-component analysis (PCA) separated our datasets along two main axes (Figure 1E). The first principal component (PC1, 82.6% of variance) covered the shift in cell identity from pre-implantation (Figure S1F) to post-implantation (Figure S1G). The second principal component segregated 9h from 0h and 48h (9.3% of variance, PC2). Taken together, our time course effectively captures the transition from naive to primed pluripotency.

To investigate post-replicative accessibility at cell-specific loci, we focused on quantifying ATAC-seq signal enrichment at peaks found uniquely in either unreplicated 0h (naive, n = 59,989) or 48h (primed, n = 53,357) datasets (Figure 1F, central Venn diagram). If differentiation utilizes replication for chromatin remodeling, replicated chromatin at the 9h intermediate time point should be distinct from the unreplicated chromatin landscape. However, we observed the same trends in replicated and unreplicated chromatin over both 0h unique and 48h unique peaks, suggesting the replicated landscape largely mirrors the steady state (Figure 1F, boxplots). To identify whether there are peaks where replicated chromatin dynamics differ from that of unreplicated chromatin, we conducted targeted differential enrichment analysis. Focusing on the unique 48h peaks, we filtered for regions significantly enriched between 0h and 9h in replicated chromatin, while insignificant at the same time points in unreplicated chromatin (Figure S1H). This identified a subpopulation of 490 peaks that showed increased accessibility in replicated but not unreplicated 9h chromatin, suggestive of a more differentiated landscape post-replication at these loci (Figure 1G). These “precocious” peaks were in proximity to genes directly related to post-implantation epiblast, such as primitive streak and embryonic ectoderm (Figure S1I; Table S1). Motif analysis showed that DNA recognition sites for key regulators of primed pluripotency such as OCT4, SMAD4, and BRN1/OCT6 were enriched in precocious regions relative to all opening regions (Figure 1H) (Buecker et al., 2014; Waisman et al., 2024; Zhao et al., 2024). Last, 38% of precocious peaks overlapped with binding of the primed specific transcription factor OTX2 (Buecker et al., 2014) compared to 26% of all opening peaks (Figure 1I), suggesting replication is providing a window of opportunity for opening chromatin regions specific to epiblast maturation. Targeted filtering for the reciprocal pattern of peaks, closed in replicated chromatin and open in unreplicated chromatin (Figure S1J), yielded 1,697 statistically significant peaks, but these showed minimal difference in signal (Figure S1K). Taken together, this suggests that replication contributes to chromatin opening rather than chromatin closing, helping to establish a new chromatin landscape during the naive to primed transition.

During differentiation from naive to primed pluripotency, cells experience an endogenous push to further increase proliferation from an already-high baseline (Figures 1B and S1A) (Morgani et al., 2017). While the system is well-characterized, efficient, and robust, the high level of proliferation in the starting population can mask replication-dependent chromatin changes. The unreplicated fraction of chromatin will include chromatin replicated prior to or just after the EdU label, causing a noisy background for extraction of replication-dependent changes in accessibility and underestimation of replication’s effects. We, therefore, turned to iPSC reprogramming of mouse embryonic fibroblasts (MEFs), where cells experience an exogenous push to rapidly shift from a low-to a high-proliferation state.

Replicated chromatin accessibility predicts iPSC progression

Overexpression of the Yamanaka factors OCT4, KLF4, SOX2 and cMYC (OKSM) (Stadtfeld et al., 2010; Takahashi and Yamanaka, 2006) reprograms MEFs to iPSCs (Figure 2A). Contrary to the naive-to-primed transition characterized above, this process is extremely inefficient and thought to be stochastic, with approximately 1% of starting MEFs undergoing reprogramming to iPSCs within 14 days of OKSM induction in our cell-intrinsic transgenic rTA/OKSM system (Stadtfeld et al., 2010). However, we found that upon OKSM induction, cells uniformly increased their proliferative rate, with a higher proportion of cells in S-phase observed as early as 12 h post-OKSM induction (Figures 2B and S2A).

Figure 2.

The first S-phase during OKSM reprogramming advances the replicated chromatin landscape

(A) Overview of iPSC induction strategy.

(B) Proportion of EdU+ cells across the time course of initial MEF reprogramming, quantitated by flow cytometry. Line, mean of three independent experiments.

(C) Replicated and unreplicated repli-ATAC- and ATAC-seq RPKM-normalized signal across iPSC induction (12h, turquoise; 36h, green) at the Thy1 locus (top) and Tfcp2l1 locus (bottom).

(D) Principal-component analysis of chromatin accessibility variance across all consensus peaks in replicated (green) and unreplicated (gray) datasets.

(E) Central: Venn diagram of overlap between peaks called in unreplicated 6h ATAC-seq and unreplicated 36h ATAC-seq in induced MEFs. (Left) Boxplots of replicated repli-ATAC-seq (top) (Fisher’s median test p values for 6h vs. 12h = “0” and 12h vs. 36h = “0” R precision limit <1e−300) and unreplicated ATAC-seq (bottom) (Fisher’s median test p values for 6h vs. 12h = “0” and 6h vs. 36h = “0”; R precision limit <1e−300) signal over 6h unique peaks. (Right) Boxplots of replicated repli-ATAC-seq (top) (Fisher’s median test p values for 6h vs. 12h = 6.1e−86 and 12h vs. 36h = 4.1e−43) and unreplicated ATAC-seq (bottom) (Fisher’s median test p values for 6h vs. 12h = “0” and 12h vs. 36h = “0”; R precision limit <1e−300) signal over 36h unique peaks. Horizontal black line: median; black dot: mean; boxes: 25^th–75^th percentiles; vertical black lines: 1.5× interquartile range.

(F) Motif enrichment in de novo opening and closing peaks, as defined in Figure 2E. Top 10 motifs for each peak set are shown. Dot size reflects the prevalence of the motif in the target peak set relative to a representative, random background peak set.

(G) Overlap of OCT4 (65,526 binding sites), SOX2 (53,634 binding sites), and KLF4 (26,785 binding sites) binding in MEFs induced to reprogram for 48 h (Chronis et al., 2017) with sites enriched in 12h replicated (52,23 regions) or unreplicated (8,343 regions) chromatin. The number of overlapping peaks were in all cases normalized to the respective 12h sample to calculate the percentage overlap.

(H) Principal-component analysis of chromatin accessibility variance across all time course consensus peaks in the original replicated (green) and unreplicated (gray) 6h and 12h repli-ATAC-seq datasets (also shown in D), plus control (black) and aphidicolin (red)-treated samples from ATAC-seq of reprogramming with blocked replication (see also Figure S2E).

(I) Overlap of OCT4, SOX2, and KLF4 binding sites in MEFs induced to reprogram for 48 h with chromatin sites specific for control (6,208 regions) and aphidicolin-treated (14,993 regions) ATAC-seq samples. Number of binding sites for the respective transcription factors is given in (G). The numbers of overlapping peaks were in all cases normalized to the respective control/aphidicolin sample to calculate the percentage overlap.

See also Figure S2.

In this time course, we profiled MEFs induced to reprogram before upregulation of proliferation (6h), during their first S-phase post-induction (12h), and in the middle of their initial proliferative burst (36h) (Figures 2B and S2A). We generated replicated and unreplicated accessibility profiles using the same strategy as in the naive to primed differentiation time course (Figure 1C). We confirmed that changes in our unreplicated ATAC-seq between our earliest time point of 6h and latest time point of 36h showed a positive Pearson correlation (0.35, p < 2e−16) with previously mapped global gene expression changes between 0h and 48h of reprogramming (Figure S2B) (Chronis et al., 2017). The correlation is decreased by a group of peaks/genes where accessibility precedes transcriptional changes, in accordance with previous findings that epigenetic changes, primarily opening of chromatin (H3K4me2), precede transcriptional changes early in reprogramming (Koche et al., 2011). Between 12 and 36h, we observed the expected decreased accessibility at the MEF-specific Thy1 locus and increased accessibility at known pluripotency loci, such as Tfcp2l1 (Figure 2C). The replicated signal at 12h and 36h at the Tfcp2l1 locus was comparable, raising the possibility that the 12h replicated chromatin landscape is more advanced than the corresponding unreplicated fraction.

PCA revealed that 12h replicated chromatin clustered with all 36h datapoints, while 12h unreplicated datasets clustered with the 6h datapoints (Figure 2D). This suggests that in the first S-phase of fibroblast reprogramming, replicated chromatin has a distinct accessibility profile compared to unreplicated chromatin in the same cells, and that this profile is similar to the accessibility seen in more progressed MEF intermediates. Contrasting our 6h and 36h peaks, we did not identify any clear Gene Ontology (GO) term signatures, likely because gene expression changes in early stages of reprogramming are stochastic and highly variable (Buganim et al., 2012).

We wondered whether the distinctive chromatin landscape seen in replicated chromatin at 12h reflected de novo chromatin opening or closing. To assess this, we scored whether the loci contributing positively to PC1 were opening or closing across the time course and found that it is specifically peaks gaining accessibility that move samples toward positive PC1 values (Figure S2C). This suggests that, as we saw in the transition from naive to primed pluripotency, the post-replicative window is being utilized specifically to open loci for transcriptional regulation, rather than rendering them inaccessible.

To confirm this finding, we investigated post-replicative accessibility at the most dynamic loci, focusing on peaks found uniquely in either unreplicated 6h or 36h datasets as we did with our naive-to-primed differentiation datasets (Figure 2E, central Venn diagram). At unique 36h peaks, which represent regions that open early in reprogramming, replicated chromatin 12h post-induction had signal comparable to that seen in the 36h datasets, while unreplicated chromatin showed a gradual opening from 6h to 36h (Figure 2E, right), consistent with our PCA (Figure 2D) and the hypothesis that replication is utilized for chromatin opening. The similarity between unreplicated and replicated chromatin at 36h suggests that the trends we observed at 12h are particularly prevalent in the first S-phase in MEF reprogramming.

At unique 6h peaks, which represent regions that close during early reprogramming, replicated and unreplicated chromatin showed the same gradual trend, decreasing in signal across our time course (Figure 2E, left). Consistent with our naive-to-primed differentiation time course, this indicates that although replication transiently ablates accessibility at regulatory elements (Figure S2D) as previously reported (Ramachandran and Henikoff, 2016; Stewart-Morgan et al., 2019), the cell seemingly does not exploit this to stably repress loci during cell identity changes.

Motif analysis showed a clear pluripotency signature at opening regions, with enrichment of recognition sites for OCT4, NANOG, KLFs, and SOXs (Figure 2F). In contrast, regions that close early in reprogramming were enriched for MEF-specific recognition sites such as AP-1 (Figure 2F) (Vierbuchen et al., 2017). To evaluate the biological importance of the regions that precociously open in the replicated 12h chromatin, we assessed to what extent these regions overlap with actual binding of OCT4, SOX2, and KLF4 in MEFs induced to reprogram for 48 h (Figure 2G) (Chronis et al., 2017). For the 3 factors, we found 90%, 84%, and 55% overlap, respectively, while overlaps with the 12h unreplicated regions were 23%, 21%, and 18%, respectively (Figure 2G).

To formally test a causative link between replication and chromatin opening during cell state transitions, we induced MEFs to reprogram while blocking replication by addition of the DNA polymerase inhibitor aphidicolin and assayed reprogramming cells with conventional ATAC-seq (Figures S2E–S2G). Blocking replication during early reprogramming impaired chromatin opening (Figure 2H), specifically at sites bound by OCT4, SOX2, and KLF4 (Figure 2I). We conclude that during early reprogramming, replication is utilized for opening chromatin at regulatory loci crucial for cell state transition.

Proliferative heterogeneity underlies early iPSC reprogramming efficiency

In profiling replicated and unreplicated chromatin, we more readily distinguished replication-dependent accessibility changes in our OKSM system compared to naive-to-primed differentiation. We, therefore, used OKSM reprogramming to further investigate how proliferation could facilitate cell identity change.

To better understand the link between proliferation and reprogramming potential, we employed CellTrace labeling to track the proliferation rate of individual cells (Figures 3A, S3A, and S3B). Consistent with a functional role for proliferation in reprogramming, induced MEFs divided more than uninduced MEFs (Figure S3C). By dividing CellTrace signal into quartiles, we could delineate the cells that divided the most and the least as cells reached the end of the proliferative burst, around 60 h post-induction (Figure 3B). We isolated cells from each quartile and performed standard ATAC-seq to profile accessibility in these proliferative subpopulations.

Figure 3.

Proliferation predicts reprogramming potential in iPSC induction

(A) Overview of strategy to monitor cell proliferation during MEF reprogramming.

(B) Sorting strategy to isolate 60h induced MEFs based on proliferative rate.

(C) Principal-component analysis of chromatin accessibility variance across all consensus peaks in Q1–Q4 ATAC-seq at 60h post-iPSC induction, showing PC2 and PC3, representing 18.2% of total variance.

(D) Top 12 MGI Expression GO enrichment for peaks significantly more accessible in Q1 (highly proliferative fraction). Theiler stages (TSs) corresponding to pre-implantation development are highlighted with bold font.

(E) Top 10 MGI Expression GO enrichment for peaks significantly more accessible in Q4 (least proliferative fraction). TSs corresponding to mesenchyme are highlighted with bold font.

(F) Venn diagram of overlap between peaks called in Q1, Q2, Q3, and Q4 ATAC-seq.

(G) Distribution of Q1 and Q4 unique peaks in replication timing domains. Replication timing data from Hiratani et al. (2010).

(H) Boxplots of SSEA1 ATAC-seq (left) (Fisher’s median test p value = 7.2e−7) and THY1 ATAC-seq (right) (Fisher’s median test p value = 3.5e−49) signal over unique Q1 and Q4 peaks (as defined in F). Horizontal black line: median; boxes: 25^th–75^th percentiles; vertical dashed lines: 1.5× interquartile range. SSEA1 and THY1 ATAC-seq data from Knaupp et al. (2017).

See also Figure S3.

While the CellTrace ATAC-seq datasets (n consensus peaks = 84,466) were affected by inter-replicate variance (Figure S3D), we did observe a separation of the datasets based on their proliferative quartile along principal component 3 (PC3, 5.3% of total variance) (Figure 3C). Contrasting Q1, the most proliferative subpopulation, and Q4, the least proliferative subpopulation, we identified 118 and 1,000 peaks, respectively, that were significantly enriched. GO analysis of these peaks revealed proximity of Q1 peaks to genes expressed during pre-implantation development, including Theiler stages 2, 4, and 5, reflecting enhanced reprogramming in this subpopulation (Figure 3D). In contrast, Q4 peaks were solely enriched for terms related to development beyond 11.5 days after conception, many of them related to mesenchyme and a fibroblast cell identity (Figure 3E). Compared to less-proliferative cells in the same dish, the most proliferative MEFs after 60 h of OKSM expression have, therefore, opened chromatin at loci linked to pluripotency.

To strengthen numbers while maintaining a Q1/Q4 binary, we next defined loci called as peaks exclusively in either Q1 (n = 3,946) or Q4 (n = 3,231) compared to other quartiles (Figure 3F). These peak sets were distributed differently among replication timing domains, with Q1 peaks more often in early-replicating regions compared to Q4 peaks (Figure 3G). This is consistent with earlier work showing that changes in euchromatin, which replicates early in S-phase, is the site of most chromatin remodeling early in OKSM reprogramming (Koche et al., 2011). We further analyzed accessibility at Q1 and Q4 peaks in previously published ATAC-seq datasets of SSEA1+ and THY1+ MEFs (Knaupp et al., 2017). Three days post-OKSM induction, all eventual iPSCs will originate from a small SSEA1+ population, whereas THY1+ cells at this timepoint cannot reprogram (Polo et al., 2012). We found that day 3 SSEA1 ATAC-seq showed specific enrichment at Q1 peaks, while day 3 THY1 ATAC-seq showed enrichment for Q4 peaks (Figure 3H). Taken together, these results demonstrate that the rate of cell proliferation early in reprogramming correlates with a cell’s propensity for later induction of pluripotency.

Discussion

The relationship between proliferation and cell state changes has long been debated, but has remained difficult to disentangle. By isolating and characterizing the replicated chromatin landscape, we have been able to focus on accessibility changes in the wake of the replication fork that are obscured in assays of bulk/steady-state chromatin. Our results establish a role specifically for replication, not just active cell division, in cell identity change.

How does replication influence lineage choice? Comparing our results with previous characterizations of replicated chromatin suggests several, non-exclusive mechanisms through which replication could facilitate cellular transformation. The unique hyperacetylated state of nucleosomes post-replication could create a local environment more amenable to transcription factor binding. This mechanism could underpin previous observations that the lineage-specific transcription factor GATA1 binds target loci specifically in S-phase at the onset of erythropoiesis (Pop et al., 2010). It could also help reconcile the seeming paradox that increased proliferation is expected to dilute transcription factor concentrations but is also correlated with improved reprogramming efficiency. In future work, it will be important to examine the role of replication in cell fate change from more slowly-dividing cells and compare it to the rapidly cycling systems we have used here.

A proximity ligation-based assay has detected opportunistic transcription factor binding after replication in differentiating embryonic and hematopoietic stem cells (Petruk et al., 2017a, 2017b; 2017a), but we did not observe global accessibility increases in replicated chromatin in either this work or our previous study of homeostatic mouse ESCs (Stewart-Morgan et al., 2019). Globally increased accessibility has also not been observed in analogous studies in other models (Fennessy and Owen-Hughes, 2016; Gutierrez et al., 2019; Ramachandran and Henikoff, 2016; Vasseur et al., 2016). However, a recent study that utilized long-read sequencing identified under-wrapping of DNA on nascent chromatin fibers (Ostrowski et al., 2024), suggesting a unique nucleosome structure on replicated DNA that might facilitate transcription factor binding. Future studies manipulating acetylation on naive histones will help elucidate the contribution of this transient state to transcription factor-driven cell identity change.

It could additionally be that transcription factors, especially exogenously expressed factors in experimental reprogramming contexts, are blocked from many of their target loci on pre-replicated chromatin by repressive epigenetic marks. DNA replication dilutes these epigenetic marks by generating hemi-methylation of DNA and distributing methylated parental histones between sister chromatids (Alabert et al., 2015; Stewart-Morgan et al., 2023). This likely lowers the barrier for gene activation in otherwise repressive environments. A number of pioneer factors are known to be mitotic bookmarkers, and their established occupancy through one S-phase could accelerate transcriptional change (Perera et al., 2022; Wong et al., 2023). We envision this being especially relevant to binding events in H3K9me3 domains. H3K9me3 is refractory to all transcription factor binding, including pioneer transcription factors (Becker et al., 2016; Nicetto and Zaret, 2019), and experimental disruption of H3K9me3 improves reprogramming efficiency in multiple contexts (Becker et al., 2017; Chen et al., 2013; Hirai and Kikyo, 2014; Sridharan et al., 2013). This opportunistic binding post-replication would allow transcription factors to gain a foothold and initiate transcription changes, ultimately changing cell identity.

We observed more precocious accessibility in OKSM-induced MEFs compared to differentiating ESCs. This may be technical, as ESCs are already highly proliferative prior to EpiLC induction. We postulate that this created more noise in our pluripotency datasets, where chromatin replicated just prior to or after our EdU pulse made up a substantial component of our EdU-negative chromatin controls. It is not surprising that we did not see global trends in our EpiLC differentiation given that the precocious peaks we did identify constituted only 0.2% of the consensus peak set. As the iPSC system represents induction, proceeding cell-intrinsically throughout the MEF culture, it is a more synchronous change than relying on the provision of cytokines to a differentiating culture, where variation in cellular responses to signaling results in heterogeneous differentiation. Thus, iPSCs provided a much cleaner model, with coordinated entry into S-phase very quickly post-induction. Biologically, the drastic transition between such distinct cell types as MEFs and iPSCs, compared to the normal progression of pluripotent cell types in development, represents vastly different degrees of cell state changes. While our observations with iPSCs provide a more robust demonstration of the role of replicated chromatin in altering cellular identity, the demonstration that these also occur in the more physiological and subtle transition through pluripotency suggests replication could be a general mechanism of differentiation.

By parsing induced MEFs by their proliferation history 60 h post-induction, we detected a more iPSC-like signature in the most proliferative MEFs compared to those that had divided the least. This is consistent with previous work that used serum withdrawal to arrest induced MEFs after defined numbers of divisions and found changes in euchromatic H3K4me2 in the most divided samples (Koche et al., 2011). The starting MEF population is heterogeneous (Shakiba et al., 2019; Singhal et al., 2016), including epigenetically (Pour et al., 2015). This heterogeneity likely extends to the differences in proliferation rate we observed between our repli-ATAC-seq and ATAC-seq independent experiments. Recent work in transdifferentiating fibroblasts has shown that proliferation history affects how cells respond to transcription factors and that early hyperproliferation predicts improved reprogramming (Wang et al., 2025). There, therefore, remains much to learn about the relationships between the epigenome, the cell cycle, and cell fate transitions.

Given that such a small percentage of MEFs ultimately become iPSCs, even our most proliferative ATAC-seq dataset still contained many MEFs that would ultimately fail to reprogram. Nevertheless, our findings add to the body of work linking proliferation speed and reprogramming potential. Previous work has shown that virtually all starting MEFs can reprogram in conditions that facilitate continuous growth (Hanna et al., 2009). A MEF subpopulation with a fast cell cycle has been identified as the progenitor of effectively all reprogrammed cells in the classic OKSM protocol (Guo et al., 2014), although this work assessed proliferation 4 days after induction, not from the start as we did in this study. Fast proliferation is a common feature of successful reprogramming, including direct reprogramming (Wang et al., 2025), and proliferation is required for high fidelity differentiation (Eminli et al., 2009; Perera et al., 2022; Wong et al., 2023). This fast cycling could afford transcription factors multiple opportunities to access their binding sites on replicated chromatin or could make these sites more accessible generally by diluting the repressive epigenome, which is known to be restored with slow kinetics (Alabert et al., 2015; Reverón-Gómez et al., 2018).

Methods

Experimental model and study participant details

Reprogramming MEF culture

MEFs were cultured in MEF media: DMEM (Gibco, 31966021) supplemented with 10% heat-inactivated fetal calf serum (FCS) (GE Hyclone, SV30160.03), 1× GlutaMAX (Gibco, 35050061), 1× non-essential amino acids (Gibco, 11140050), 55 μM beta-mercaptoethanol (Gibco, 21985023), and 1× penicillin-streptomycin (Gibco, 15140122) and incubated at 37°C with 5% oxygen. For repli-ATAC-seq, reprogramming was induced by seeding 6 × 10⁶ MEFs on 10-cm plates coated with 0.2% gelatin (Sigma-Aldrich, G9391) and culturing in reprogramming media: KnockOut DMEM (Gibco, 10829018) supplemented with 15% heat-inactivated FCS (GE Hyclone, SV30160.03), 1× GlutaMAX (Gibco, 35050061), 1× non-essential amino acids (Gibco, 11140050), 55 μM beta-mercaptoethanol (Gibco, 21985023), 1× penicillin-streptomycin (Gibco, 15140122), custom-made LIF, 2 μg/mL doxycycline (Sigma-Aldrich, D9891), and 50 μg/mL L-ascorbic acid (Sigma-Aldrich, A4544) incubated at 37°C with atmospheric oxygen levels. Cells were then labeled and harvested at the appropriate time point for repli-ATAC-seq/ATAC-seq library generation. Our replicates were distinct primary MEF cells originating from different embryos, all either at passage 2 or 3. While this results in a smaller group of significant regions from our analyses, it ensures that the effects we identify are robust and truly biological.

To prevent DNA replication, MEFs were cultured and induced to reprogram as above. Four hours after reprogramming induction, aphidicolin (5 μg/mL, Sigma-Aldrich, A0781) was added, and both treated and untreated controls were cultured for a further 10 h before being harvested for ATAC-seq generation. Cell lines used were derived in the Hochedlinger lab (Stadtfeld et al., 2010) and were Col1a1::tetON-OKSM heterozygous; R26-M2rtTA heterozygous; Pou5f1-GFP heterozygous on a B6-129 × 1/SvJ background. MEFs of this genotype have previously been shown to exhibit homogeneous and robust OKSM expression upon doxycycline induction (Stadtfeld et al., 2010). Cells used in all experiments were mycoplasma negative and had been passaged 2–3 times since thawing, 4–5 times in total.

ESC culture and EpiLC induction

ESCs were cultured on 0.1% gelatin in serum-free media containing N2B27 (Neurobasal medium (ThermoFisher 21103049) and DMEM:F12 (Fisher Scientific, 12-634-010) supplemented with N2 (made in house), B27 (Gibco, 17504001), L-glutamine (ThermoFisher, 25030024), and 100 μM 2-mercaptoethanol (Sigma-Aldrich, M6250) supplemented with 3 μM GSK-3 inhibitor (Axon Medchem, 1386), 1 μM MEK inhibitor (Sigma-Aldrich, PZ0162), and Leukemia Inhibitory Factor (made in-house). Twenty-four hours prior to EpiLC induction, media was exchanged with media containing 3 μM GSK-3 inhibitor, LIF, and 250 nM FGFR inhibitor PD17(PD173074 Sigma-Aldrich, P2499). For EpiLC induction, 10,000 cells/cm² were plated on fibronectin in N2B27 supplemented with FGF2 (10 ng/mL, Peprotech, 450-33) and Activin A (20 ng/mL, Peprotech 120-14E). The cell line used was the Hex-redStart:Hnf4α-GFP-reporter clone 44_7 (Anderson et al., 2017), an E14Tg2a subclone on a 129/Ola background originating from the Brickman lab. Cells used in all experiments were mycoplasma negative and had been passaged 2–3 times since thawing, 35–45 times in total.

DNA labeling

Repli-ATAC-seq samples were labeled with EdU as in Stewart-Morgan et al. (2019). Cells were labeled with medium containing 20 μM EdU (Invitrogen, A10044) for 10 min. Nascent repli-ATAC-seq samples were harvested and processed immediately. Mature repli-ATAC-seq samples were washed twice in 1× PBS and cultured for a further 1 h 50 min (total time since start of EdU label: 2 h) in media containing 10 μM thymidine (Sigma-Aldrich, T9250) before harvesting.

Method details

Further descriptions of all methodologies applied can be found in supplemental methods.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact: Kathleen Stewart-Morgan (kathleen@sund.ku.dk).

Materials availability

This study did not generate any new reagents.

Data and code availability

All data generated in this study follow FAIR data management principles and have been deposited at GEO with accession no. GSE296014. Details on data processing can be found in supplemental information.

Acknowledgments

We thank the staff of the genomics and flow cytometry platforms at BRIC and CPR/reNEW for their assistance with this project. We thank the Cheloufi lab for helpful discussion and comments on this manuscript. Work in J.M.B.’s laboratory was supported by Lundbeck Foundation (R198-2015-412, R370-2021-617, and R400-2022-769), Independent Research Fund Denmark (DFF-8020-00100B, DFF-0134-00022B, and DFF-2034-00025B), the Danish National Research Foundation (DNRF116), and the European Research Council (ERC, SENCE, 101097979). S.C. was supported by the U.S. National Institutes of Health (NIH) grant R35 GM151004. Research in A.G.’s laboratory was supported by the Novo Nordisk Foundation (NNF21OC0067425), the Lundbeck Foundation (R198-2015-269 and R313-2019-448), the European Research Council (ERC advanced grant no. 101142230), and Independent Research Fund Denmark (3101-00149A, 7016-00042B). Research in K.R.S-M’s laboratory is supported by the European Research Council (ERC, hypomethGENOME, 101161245), Independent Research Fund Denmark (1133-00003B), and the Novo Nordisk Foundation (NNF22OC0073038). The Novo Nordisk Foundation Center for Stem Cell Medicine (reNEW) is supported by the Novo Nordisk Foundation, grant number NNF21CC0073729, and previously NNF17CC0027852. Research at CPR is supported by the Novo Nordisk Foundation (grant NNF14CC0001).

Author contributions

Conceptualization and methodology, T.E.K., S.C., J.M.B., A.G., and K.R.S.-M.; investigation, T.E.K., N.R.-G., A.B., N.A., and K.R.S.-M.; formal analysis, T.E.K. and K.R.S.-M.; data curation and visualization, T.E.K. and K.R.S.-M.; writing – original draft and review and editing, T.E.K. and K.R.S.-M. with input from all other authors; funding acquisition, A.G. and J.M.B.

Declaration of interests

A.G. is co-founder and chief scientific officer (CSO) of Ankrin Therapeutics. A.G. is a member of the scientific advisory board of Molecular Cell.

Published: February 19, 2026