Replication stress induces amplification of early replicating loci in the human genome

Fang Ji; Hongwei Liao; Haoyu Tang; Liujian Ouyang; Menglin Qiu; Yan Chen; Yingfei Huang; Jixuan Liu; Wen Li; Zhihua Chen; Ian David Hickson; Songmin Ying

doi:10.15302/vita.2026.01.0002

Vita > Correspondence > DOI: 10.15302/vita.2026.01.0002

Vita Published:

Replication stress induces amplification of early replicating loci in the human genome

Fang Ji ¹^,²^,³^,^# ,
Hongwei Liao ¹^,²^,^# ,
Haoyu Tang ¹^,²^,^# ,
Liujian Ouyang ² ,
Menglin Qiu ¹^,² ,
Yan Chen ¹^,² ,
Yingfei Huang ² ,
Jixuan Liu ¹^,² ,
Wen Li ² ,
Zhihua Chen ² ,
Ian David Hickson ⁴ ,
Songmin Ying ¹^,²^,³

Author information +

¹. Department of Pharmacy, Center for Regeneration and Aging Medicine, the Fourth Affiliated Hospital of School of Medicine, and International School of Medicine, International Institutes of Medicine, Zhejiang University, Zhejiang-Denmark Joint Laboratory of Regeneration and Aging Medicine, Yiwu, Zhejiang, China.

². Key Laboratory of Respiratory Disease of Zhejiang Province, Department of Respiratory and Critical Care Medicine, Second Affiliated Hospital of Zhejiang University School of Medicine, Hangzhou, Zhejiang, China.

³. Department of Pharmacology, Zhejiang University School of Medicine, Hangzhou, Zhejiang, China.

⁴. Center for Chromosome Stability, Department of Cellular and Molecular Medicine, University of Copenhagen, Copenhagen, Denmark.

iandh@sund.ku.dk

yings@zju.edu.cn

Show less

History +

Vita () Cite this article

PDF (1142KB)

登录浏览全文

4963

注册一个新账户忘记密码

Dear Editor,

DNA replication stress, marked by slowed or stalled replication forks, is a major source of genomic instability in cancer and drives tumor evolution. The molecular pathways that stabilize, repair or restart individual replication forks are well characterized, yet little is known about how the DNA replication program responds to replication stress at a genome-wide scale. Here, we implemented high-throughput sequencing of nascent DNA synthesis to map loci where replication restarted after release from global replication stress. Results showed replication restart occurred most efficiently in naturally early-replicating regions. Strikingly, we also observed nascent DNA synthesis near early-firing replication origins, even in late S-phase cells recovering from replication stress, suggesting which presumably involved break-induced replication (BIR) of already duplicated regions. Such aberrant re-replication events correlated with significant copy number amplifications, suggesting a novel mechanism for pathogenic gene amplification.

To investigate genome-wide recovery from replication stress, we mapped nascent DNA synthesis using EdUseq^1–3. Briefly, U2OS cells were treated with high concentrations of hydroxyurea (HU) or aphidicolin (APH), two well-established replication stress inducers, to inhibit ongoing replication. Cells were then released into EdU-containing medium to label newly synthesized DNA. EdU-labeled DNA was biotinylated, captured and sequenced (Supplementary Fig. S1a). FACS confirmed that prolonged HU or APH treatment stalled S-phase progression (Supplementary Fig. S1b). Surprisingly, the observed restart of DNA synthesis following global replication stress was enriched at well-defined regions instead of being distributed uniformly throughout the genome (Fig. 1a; Supplementary Fig. S1c, d). EdUseq peaks were highly reproducible (Supplementary Fig. S2a–d) and, importantly, showed progressively increased signal intensity with longer durations of replication stress (Supplementary Fig. S2e), suggesting that a process other than conventional replication might be occurring.

We identified 75,404 signal peaks after a 12-hour (h) exposure to HU-induced replication stress, and these peaks were distributed across all chromosomes (Supplementary Fig. S2f). These replication restart peaks were distributed primarily within introns (39.39%) and intergenic regions (37.42%) (Supplementary Fig. S3a). The sequences also displayed several significant motifs (Supplementary Fig. S2g), potentially conferring resistance to irreversible fork inactivation. Interestingly, gene ontology analysis revealed enrichment of cell growth-related pathways, suggesting that these regions may be functionally protected during global replication stress (Supplementary Fig. S3b).

To further characterize replication restart peaks, we divided the genome into 50 kb non-overlapping domains⁴ and defined 1,658 regions with a > 7-fold increase in EdU signal (vs control) as ‘restart hotspots’ (RHs). Size-matched random sites (control sites, CSs) were generated for comparison. RHs showed higher CpG island density and GC content than CSs (Supplementary Fig. S3c–e), indicating a sequence basis for efficient restart after replication stress. To rule out cancer-specific effects, RHs in non-transformed human bronchial epithelial cells (HBE) largely overlapped and strongly correlated with those in U2OS cells (Supplementary Fig. S4a–d). Additionally, RHs were also detected in mouse embryonic fibroblasts (MEF), and aligned with syntenic regions in human cells (Supplementary Fig. S4a, b). Collectively, these results revealed a set of RHs that are likely more resilient during replication stress and are conserved across mammalian species.

To identify features common to RHs, we examined their replication timing using Repli-Seq (Supplementary Fig. S4e). This showed that RHs were predominantly located in early replicating genomic regions (Fig. 1b; Supplementary Fig. S5a) and consistently harbored a high density of replication origins as indicated by ORC1 occupancy in G1 phase (Fig. 1b; Supplementary Fig. S5b). Next, we aligned the RH peaks to previously mapped early-firing replication origins in U2OS cells¹. This revealed that RHs predominantly colocalize with early firing origins (Supplementary Fig. S5c, d). These data suggested that the identified RHs might be specific for a particular stage of S-phase. To investigate this, we mapped sites of replication restart in flow cytometry-sorted S-phase subfractions (Supplementary Fig. S4f). We observed that 78% of the RHs in unsorted cells corresponded to sites normally replicated in early S-phase (Fig. 1c). Because HU stalls ongoing replication forks rather than blocking the firing of new origins¹, RHs at early origins should mainly arise from the restart of arrested forks that had barely progressed from the origin site, instead of from additional origin firing (Supplementary Fig. S6a). These findings demonstrate that newly established replication forks in early S-phase cells are particularly capable of resuming DNA synthesis after global replication stress. Although exhibiting weaker intensities, replication restart signals that correspond to normally early-replicating regions were also detected in sorted late S-phase cells (Fig. 1c). This opened up the intriguing possibility that these peaks may represent some form of re-replication. To exclude contamination of late S-phase fractions by early S-phase cells, we performed EdUseq using late S-phase cells acquired after 2 sequential rounds of sorting, and the putative re-replication signal persisted (Supplementary Fig. S6b–d). Importantly, this phenomenon was observed with different replication stress inducers and in different cell lines (Fig. 1d; Supplementary Fig. S7a, b), precluding the possibility of agent- or cell type-specific effects.

For quantitative analysis, we designated 3,576 regions that were normally early replicating but yet were apparently conducting nascent DNA synthesis in late S-phase following replication stress as potential DNA re-replication sites (reRSs). The same number of size-matched CSs were analyzed for comparison.

Replication stress is known to induce fork breakage and activate BIR⁵. We considered whether BIR might account for the generation of reRSs following replication fork breakage. To investigate this, we examined the effects of PIF1 or SLX4 knockdown, two canonical BIR factors, on EdU incorporation at reRSs. We observed that knockdown of either protein significantly decreased the intensity of EdU signals at reRSs (Fig. 1e, f; Supplementary Fig. S7c, d). Additionally, re-replication signals increased with prolonged HU treatment, consistent with enhanced fork collapse and subsequent BIR (Supplementary Fig. S8). These data suggest that BIR contributes to re-replication of early-replicated regions in late S phase.

Next, we analyzed the epigenetic landscape around reRSs to identify features that promote re-replication. ReRSs correlated with chromatin accessibility, as shown by ATAC-seq and enrichment of H3K27ac (Supplementary Fig. S9a, b), and displayed significantly higher ATAC-seq and H3K27ac signals than CSs (Supplementary Fig. S9c, d). Analysis of genome-wide histone modification datasets revealed enrichment of active transcription markers H3K4me3⁶ and H3K36me3⁷ (Supplementary Fig. S9e, f), alongside reduced levels of the repressive marker H3K9me3⁷ at reRSs (Supplementary Fig. S9g). RNA-seq further confirmed that reRSs are highly accessible and transcriptionally active (Supplementary Fig. S9h). On one hand, active transcription may lead to frequent replication–transcription conflicts that can be further exacerbated by additional replication perturbations to the extent of fork collapse, as has been observed that highly transcribed regions are more prone to replication stress and DNA breakage^8,9, thereby initiating BIR and re-replication. On the other hand, active transcription may help maintain an open chromatin configuration that is more permissive for the loading of DNA replication or repair factors, thereby facilitating re-replication.

We next assessed whether reRSs were associated with genomic instability. ReRSs were enriched for γH2AX and RAD51, two canonical DNA damage response/repair proteins (Fig. 1g, h), which might reflect the breakage of replication forks and thus BIR at these sites. Furthermore, we examined whether reRSs were associated with copy number variations (CNVs), the anticipated consequence of re-replication. Whole genome sequencing of U2OS cells revealed that reRSs exhibited a significantly higher level of gain CNVs and a correspondingly lower frequency of loss CNVs compared to CSs (Supplementary Fig. S10a, b). Importantly, CNV profiles obtained from patient-derived datasets showed that osteosarcoma-specific and pan-cancer-associated gain CNVs were close to reRSs (Fig. 1i, j), whereas loss CNVs were not (Supplementary Fig. S10c, d). These findings support the occurrence of re-replication at reRSs and reveal a new pathway of replication stress-induced genomic instability. Interestingly, reRSs were also more evolutionarily conserved than CSs, as indicated by increased constrained elements¹⁰ (Supplementary Fig. S10e, f), consistent with the lower frequency of CNVs associated with sequence loss.

In this study, we sequenced replication restart sites after global inhibition of DNA synthesis by HU or APH to define genome-wide patterns of replication recovery. Surprisingly, we identified conserved genomic regions that preferentially restart replication after release from global replication stress. The enrichment of RHs in normally early replicating regions suggests that newly fired forks at these sites are particularly resilient and serve a defined biological purpose. We propose that this priority is to preserve the integrity of gene-rich regions that are well-documented to be clustered in early replicating loci¹¹. RHs may therefore represent an evolved program that resists severe replication stress, protecting essential genes involved in cell growth and metabolism under adverse conditions.

In many published studies, cell synchronization in early S-phase is achieved by HU or APH treatment. Our data clearly indicate that this regimen is not innocuous, as cells may undergo aberrant re-replication at certain loci during the experiment. In contrast, thymidine synchronization appears far less injurious (our unpublished data) and may be preferable for obtaining cells without additional genome rearrangements.

Previous analyses conducted in cells lacking RAD51¹² or BRCA2¹³, which stabilize stalled forks and protect nascent DNA, have shown that early replicating loci are prone to undergo re-replication via BIR. Our data indicate that this phenomenon is not limited to cells lacking fork protection factors but also occurs in wild-type cells subjected to prolonged HU or APH treatment in S phase. Future studies may further delineate this process by employing DNA fiber analysis, which can be maneuvered to reveal finer details of the dynamics of re-replication.

A model for how BIR arising from unprotected forks might generate regions of gene amplification has been proposed¹². In brief, this model posits that a lack of fork protection leads to fork breakage at early replicating loci where there is a high density of origins. This broken fork then undergoes BIR later in S-phase, and at a time after a converging fork has already passed the site of the broken fork. When the BIR fork is activated to conduct new DNA synthesis, it serves only to copy already replicated DNA, generating a localized region of gene amplification specifically in an early replicating locus. We suggest that HU/APH-treated cells can generate sites of gene amplification by a similar mechanism. Our data identified reRSs as hotspots of genomic instability, although the extent to which reRSs are translated into, e.g., CNVs may be modulated by the underlying genetic background. Notably, osteosarcoma-specific gain CNVs were more closely associated with reRSs than pan-cancer-associated gain CNVs, indicating that cell type-specific factors, such as competence and preference of checkpoints and DNA repair pathways, could influence the outcome of re-replication. These findings emphasize that both intrinsic genomic features and cellular context are critical for understanding mechanisms underlying CNV formation.

[1]	Macheret, M. & Halazonetis, T. D. Nat. Protoc.14, 51–67 (2019).

[2]	Ji, F. et al. Cell Res. 30, 1009–1023 (2020).

[3]	Macheret, M. et al. Cell Res. 30, 997–1008 (2020).

[4]	Brison, O. et al. Nat. Commun. 10, 5693 (2019).

[5]	Costantino, L. et al. Science 343, 88–91 (2014).

[6]	Easwaran, H. et al. Genome Res. 22, 837–849 (2012).

[7]	Perez, G. et al. Nucleic Acids Res. 53, D1243–D1249 (2025).

[8]	Hamperl, S. et al. Cell 170, 774–786 (2017).

[9]	García-Muse, T. & Aguilera, A. Nat. Rev. Mol. Cell Biol. 17, 553–563 (2016).

[10]	Reid, D. A. et al. Science 372, 91–94 (2021).

[11]	Tubbs, A. et al. Cell 174, 1127–1142 (2018).

[12]	Bhowmick, R. et al. Mol. Cell 82, 3366–3381 (2022).

[13]	Groelly, F. J. et al. Mol. Cell 82, 3382–3397 (2022).

RIGHTS & PERMISSIONS

The Author(s) 2026. Published by Higher Education Press. This is an Open Access article distributed under the terms of the CC BY license (https://creativecommons.org/licenses/by/4.0/).

Cite this article

Download citation ▾

Ji, F. et al. Replication stress induces amplification of early replicating loci in the human genome Vita https://doi.org/10.15302/vita.2026.01.0002 ()