Ubiquitin and SUMO pathways in DNA replication and replication-coupled repair

Abstract

Accurate and efficient DNA replication constitutes the most effective safeguard against genome instability. Numerous aspects of replication initiation, elongation, and termination are tightly regulated by post-translational modifications. In this review, we summarize recent advances in elucidating pathways regulated by ubiquitin and the small ubiquitin-like modifier, SUMO, and compare insights gained in yeast with those obtained in vertebrate systems. These reversible modifications play critical roles in both DNA replication and replication-coupled repair processes. When active replisomes encounter obstacles such as nucleotide depletion, DNA secondary structures, or base lesions that impede fork progression, multiple genome surveillance pathways are activated to coordinate the replication stress response. Stalled replication forks undergo remodeling and reversal, thereby stabilizing the fork and facilitating replication restart. In parallel, diverse tolerance mechanisms have evolved to enable lesion bypass or replication traverse, which transiently alters the replication machinery yet permits continuation of DNA synthesis. At the core of these processes are the DNA damage tolerance and Fanconi anemia pathways, whose components collaborate to prevent under-replication during S phase and beyond. Furthermore, ubiquitin and SUMO signaling act synergistically through the activity of SUMO-targeted ubiquitin ligases. These enzymes sequester damaged replication forks at the nuclear periphery and promote recombination-mediated restart under stringent spatiotemporal control of the replication checkpoint. Failure of these mechanisms forces the cell to engage in a final, “do-or-die” attempt to initiate DNA synthesis during mitosis, a process that is also orchestrated by ubiquitin signaling.

Introduction

Ubiquitin, small ubiquitin-like modifier (SUMO) and SUMO-targeted ubiquitin ligases

This review is intended for experts in DNA replication and repair who possess a solid understanding of ubiquitin and ubiquitin-like modifiers, and are interested in the dynamic regulation of active and stalled replisomes. We focus on DNA synthesis during unperturbed replication and under replication stress, when progression of leading- or lagging-strand synthesis is impeded, and highlight mechanisms that safeguard against under-replication. In this introduction, we provide a brief overview of these modifications and refer the novice reader to more in-depth reviews.

Ubiquitination is a dynamic post-translational modification (PTM) that plays a central role in orchestrating the cellular response to replication stress and DNA damage (Jentsch et al. 1987). It regulates the timing and specificity of DNA repair processes and controls their spatial organization within distinct nuclear compartments. Ubiquitination is catalyzed by an E1 ubiquitin-activating enzyme, which transfers ubiquitin to an E2 ubiquitin-conjugating enzyme, which works in concert with an E3 ubiquitin ligase that is essential for substrate recognition and ligation (Pickart 2001). E3 enzymes are therefore critical to determine substrate specificity (Mirsanaye et al. 2021). This regulatory mechanism is closely coordinated with deubiquitination, ensuring the reversibility and fine-tuning of ubiquitin-dependent signals (Wilkinson 1997). Deubiquitinating enzymes (DUBs) counteract ubiquitin ligases by removing covalently attached ubiquitin moieties from target proteins. The largest family of DUBs is comprised of cysteine proteases known as ubiquitin-specific proteases (USPs; Clague et al. 2019). The mechanism of de-ubiquitination is not confined to ubiquitin C-terminal hydrolysis, and seven structurally diverse protein families have been identified in humans and five in yeast (Suresh et al. 2020). DUBs regulate the duration and reversibility of ubiquitin-dependent signals and are essential for recycling ubiquitin and maintaining the cellular ubiquitin pool. DUBs themselves are subject to complex regulation through phosphorylation, ubiquitination, SUMOylation, and interactions with specific protein partners. Together, these modifications are crucial for maintaining genome integrity.

SUMO proteins also function as PTMs and are conjugated to target proteins via a cascade of enzymatic steps analogous to ubiquitination, involving an E1 SUMO-activating enzyme, an E2 SUMO-conjugating enzyme, and an E3 SUMO ligase (Hay 2005; Zhao 2018). In yeast, a single ~10 kDa SUMO isoform is expressed (Johnson and Hochstrasser 1997; Li and Hochstrasser 1999). In humans, four SUMO protein isoforms exist: SUMO-1, SUMO-2, SUMO-3, and SUMO-4. SUMO-2/3 are the isoforms predominantly used for polymeric chain formation. SUMO can be deconjugated by SUMO specific proteases (SENPs), a family of cysteine specific proteases (Claessens and Vertegaal 2024). Lastly, SUMOylation and ubiquitination pathways intersect with a specialized class of enzymes known as SUMO-targeted ubiquitin ligases (STUbLs). STUbLs specifically recognize SUMOylated substrates via SUMO-interacting motifs (SIMs) and attach ubiquitin to these proteins, marking them for proteasomal degradation or modulating their activity (Prudden et al. 2007; Nie and Boddy 2016). These enzymes play critical roles in DNA repair, replication, and chromatin organization (Chang et al. 2021; Thu 2024).

The role of ubiquitination in maintaining genomic integrity during normal DNA replication

Origin licensing and replication initiation

DNA replication in eukaryotic cells begins in G1 phase with the formation of the pre-replicative complex (pre-RC; Lam et al. 2025). The origin recognition complex (ORC), comprising six conserved subunits (ORC1–6), binds to replication origins in an ATP-dependent manner. In multicellular organisms, ORC recruitment is facilitated by ORCA (LRWD1; Figure 1(A)), which binds to trimethylated histone H4K20 and modifies chromatin structure (Sahu et al. 2023) Although not well explored, the E3 ubiquitin ligase RFWD3, best known for its role in the DNA damage response (Gong and Chen 2011; Liu et al. 2011; Elia, Wang, et al. 2015; Dubois et al. 2017; Feeney et al. 2017; Lin et al. 2018; Inano et al. 2020) ubiquitinates ORCA to stabilize its interaction with ORC (Shen et al. 2012; Hsu et al. 2020). Following ORC recruitment, other licensing factors such as CDC6 and CDT1 assist in loading the MCM2–7 helicase, forming the pre-RC (Figure 1(A)). To prevent re-replication, cells tightly control the levels of these licensing factors. The CRL4^Cdt2 E3 ubiquitin ligase targets CDT1, CDC6, and other replication proteins for degradation, but only when they are bound to proliferating cell nuclear antigen (PCNA), which is loaded on DNA after origin firing (Panagopoulos et al. 2020; not shown in Figure 1). CDT1 that escapes degradation inhibits fork progression by binding to the MCM2–7 complex and blocking helicase activity (Ratnayeke et al. 2023). CRL4^Cdt2 can monoubiquitinate PCNA under unperturbed conditions to counteract endogenous replication stress (Terai et al. 2010). Overexpression of CDT2 correlates with poor prognosis in cancer (Mazian et al. 2022).

Figure 1.

Ubiquitin regulated steps during DNA replication in human cells. (A) Pre-replicative complex formation and origin licensing. ORCA facilitates ORC1–6 complex recruitment. RFWD3, an E3 ubiquitin ligase, directly interacts with ORC1 and ORC2 subunits of the complex and ubiquitinates ORCA, thereby stabilizing the pre-RC at the origin. CDC6 and CDT1 are recruited to the origin via the ORC1–6 complex, and together they function as loaders of the MCM2–7 double hexamer. The double hexamer adopts an inactive, open conformation. ORC facilitates the passage of DNA between MCM2 and MCM5, a process that possibly promotes the closure of the complex. CDT1 and CDC6 are subsequently polyubiquitinated by CRL4^Cdt2 (not shown) and degraded by the proteasome. (B) Replication origin firing. During the G1/S phase transition, the MCM2–7 double hexamer becomes phosphorylated (red squares) by CDK2 and DDK (not shown). This phosphorylation in concert with MCM10 promotes the recruitment of GINS and CDC45 to each MCM complex within the double hexamer, resulting in the formation of a “double” CMG helicase. OBI1, an E3 ubiquitin ligase, multi-monoubiquitinates ORC3 and ORC5, which facilitates origin firing. Activation of the CMG helicase causes the double MCM2–7 hexamer to dissociate and the two CMG complexes bypass each other before they move bidirectionally along the DNA. The directionality of CMG progression is indicated by pink arrows. (C) Replication termination. Two CMG complexes converge (the directionality of CMG progression is indicated by pink arrows), which trigger CMG removal from DNA. MCM10 and phosphorylations on MCM2–7 are not shown for simplicity. Pathway 1 is shown on the left: the MCM7 subunit of MCM2–7 complex is polyubiquitinated by K63-linked chains by CRL2^Lrr1 in higher eukaryotes. Polyubiquitinated MCM7 is a target of p97 segregase, which helps to dissociate CMG from the DNA. Pathway 2, shown on the right, depicts the facte of CMG complexes that are not unloaded (e.g., at stalled forks: when CMG remains chromatin bound, phosphorylated (red squares) TRAIP polyubiquitinates MCM7 or MCM3 at mitotic entry to facilitate p97-dependent CMG dissociation.

Origin firing and replication elongation

Upon S phase entry, CDKs and the DDK complex (CDC7–DBF4) phosphorylate MCM2–7 subunits, promoting the assembly of the active CMG helicase (CDC45–MCM–GINS; Figure 1(B)). MCM10 facilitates interactions between the MCM complex, CDC45, and GINS and contributes to DNA unwinding (Baxley and Bielinsky 2017; Lõoke et al. 2017). OBI1 (RNF219) was recently identified as a regulator of origin selection in humans, catalyzing multi-monoubiquitination of ORC subunits ORC3 and ORC5 (Coulombe et al. 2019). Additionally, in yeast, Mcm3 is subject to SUMOylation, which regulates MCM2–7 complex abundance and thereby DNA replication dynamics (Quan et al. 2022). Therefore, even in budding yeast where replication origins are characterized by a consensus sequence (Nieduszynski et al. 2006), PTM of pre-RC components by ubiquitin and SUMO is critical in regulating firing, and evolutionarily more highly conserved than DNA sequence motifs. Not all licensed origins are activated during S phase, and in yeast this is partly regulated by ORC hyper-SUMOylation (Regan-Mochrie et al. 2022). Pre-RCs that remain on chromatin can serve as back-up or dormant origins that can ensure genome duplication in the event of replication fork collapse (Shima and Pederson 2017). Stem cells utilize this mechanism to shield themselves from DNA under-replication (Ge et al. 2015).

RNF4’s role in replication initiation and restart

RNF4 is a well-known STUbL involved in DNA double-strand break (DSB) repair where it controls the turnover of MDC1 and RPA and facilitates loading of RAD51 onto damaged DNA (Chang et al. 2023). However, it also plays a role in DNA replication. Chromatin surrounding active replisomes is rich in SUMO and low in ubiquitin, whereas post-replicative chromatin displays the opposite pattern, likely facilitating removal of replication factors by ubiquitin-dependent degradation (Abbas 2021; Martín-Rufo et al. 2022). The concept of “SUMO clouds” was originally introduced by Psakhye and Jentsch in 2012 in the context of DNA repair (Psakhye and Jentsch 2012). SUMO was compared to a glue that stabilizes and enhances the function of larger protein complexes. Conceptually similar is the SUMOylation of replisomes. However, SUMO also allows for STUbL-mediated remodeling of forks, and this is emerging as a critical regulation in replication fork restart (Chang et al. 2021). The environment at active forks is maintained by USP7, which opposes the ubiquitination of SUMO and SUMOylated proteins (Lecona et al. 2016; Lecona and Fernandez-Capetillo 2016). Prolonged exposure to hydroxyurea (HU) for over 16 h, resulting in fork collapse, renders RNF4 essential for firing dormant origins near stalled forks (Ellis et al. 2021). This is further corroborated by the finding that RNF4 promotes origin firing in MCM10-deficient cells that suffer chronic replication stress (Oram et al. 2024). However, in the absence of RNF4, the activation of origins near stalled forks can be promoted by the co-depletion of the Bloom syndrome (BLM) helicase (Ellis et al. 2021). There is no evidence that BLM has an inhibitory effect on origin firing; rather, the degradation of BLM at collapsed forks may aid in the release of factors that activate dormant origins. RNF4 also targets Fanconi anemia (FA) proteins, including FANCA and FANCD2/FANCI (D2–I), to regulate the cellular response to DNA damage that triggers replication fork stalling (Gibbs-Seymour et al. 2015; Xie et al. 2015)We will cover this pathway in more detail later in this review. FA proteins are recruited early after fork arrest with the D2–I heterodimer forming a ring that encircles double-to-single-stranded DNA junctions. D2–I removal is necessary for the resumption of DNA synthesis (Gibbs-Seymour et al. 2015). Thus, RNF4 controls the turnover of several target proteins to promote replication fork restart. Not surprisingly, tissue-specific knockout of RNF4 in mice leads to genomic instability, albeit without affecting tumor susceptibility (Her et al. 2024). Combined knockout of RNF4 (reducing ubiquitin recycling by protein turnover) and the SUMO-deubiquitinase USP7 (enhancing ubiquitin conjugation to SUMO and other target proteins) in p53-deficient cells causes defective DNA replication by inhibiting ubiquitin recycling, which ultimately depletes ubiquitin pools. These cells are sensitive to bortezomib, which ablates ubiquitin-dependent RPA phosphorylation and subsequent ATR activation in response to genotoxic agents (Chang et al. 2023). As such, RNF4 not only controls the timely turnover of chromatin-binding proteins but also the recycling of ubiquitin and SUMO.

Replication termination and CMG disassembly

Termination of DNA replication involves disassembly of the CMG helicase, primarily through K48-linked polyubiquitination of MCM7 (Moreno et al. 2014; Figure 1(C)). In yeast, this is mediated by SCF^Dia2, a member of the Skp1-Cullin1-F-box family of E3 ubiquitin ligases (Skaar et al. 2014), while in metazoans this function is carried out by CRL2^Lrr1, a Cullin-Ring ligase and “cousin” of SCF complexes (Maric et al. 2014; Dewar et al. 2017; Sonneville et al. 2017; Jones et al. 2024; pathway 1 in Figure 1(C)). The AAA+ ATPase p97 (VCP/CDC48 in yeast) recognizes ubiquitinated CMG and mediates its unloading (Moreno et al. 2014). Replisome unloading is facilitated by UBXN7, which stabilizes the interaction between CMG and p97 (Tarcan et al. 2022; Kochenova et al. 2022; not shown in Figure 1(C)). Structural studies revealed that Lrr1’s leucine-rich repeat domain interacts with the zinc finger domains of MCM3 and MCM5, where lagging strand DNA exits the CMG complex. This steric interaction prevents premature CMG ubiquitination and disassembly (Jenkyn-Bedford et al. 2021; Le et al. 2021; Zhou et al. 2021). Loss of Lrr1 impairs replisome disassembly and triggers ATR-mediated checkpoint activation (Fan et al. 2021). An alternative CMG disassembly pathway operates in mitosis or under replication stress (pathway 2 in Figure 1(C)) involving the E3 ubiquitin ligase TRAIP (also known as RNF206), which catalyzes K6- and K63-linked polyubiquitination of MCM7 (Deng et al. 2019; Villa et al. 2021; Poovathumkadavil et al. 2025). Although TRAIP associates with the replisome during S phase, its activity is regulated by mitosis-specific PTMs rather than replisome binding alone (Poovathumkadavil et al. 2025). In Xenopus egg extract, TRAIP activity in S phase is required for DNA:protein crosslink removal (Larsen et al. 2019). The enzyme has also been implicated in resolving transcription-replication conflicts through ubiquitination of RNA polymerase II (Scaramuzza et al. 2023). SUMOylation of TRAIP facilitates its stabilization and localization to the nucleus (Park et al. 2016). While the SUMO E3 ligase that targets TRAIP has not been identified, TRAIP has been shown to be activated by CDK1-dependent phosphorylation (Can et al. 2024; Poovathumkadavil et al. 2025). Whether TRAIP promotes the degradation of the CMG complex it is associated with, or a nearby CMG in trans is still open for investigation (Wu et al. 2021). In yeast, the helicases Rrm3 and Pif1 help disassemble persistent CMG complexes that remain on chromatin as cells exit S phase (Polo Rivera et al. 2024). In vertebrates, an alternative pathway involves RNF8, which polyubiquitinates MCM7 at K145 during late S phase. RNF168 and BRCA1 contribute to this regulation upon replication termination (Sun et al. 2024), consistent with a previous report that implicated BRCA1 in the unloading of the CMG helicase from forks that stalled at a DNA interstrand cross-link (ICL) (Long et al. 2014). In summary, what these findings suggest is that eukaryotic cells have a system in place that senses the difference between temporarily stalled forks, permanently arrested replisomes, and termination.

PCNA ubiquitination and SUMOylation under unperturbed conditions

Under unperturbed conditions, PCNA is ubiquitinated at K164, a modification that is evolutionarily highly conserved and facilitates the bypass of DNA lesions to promote efficient DNA replication in yeast (Daigaku et al. 2017) and non-transformed human cells (Thakar et al. 2020; Leung et al. 2023). PCNA is also SUMOylated at K164 (and K127 in budding yeast; Figure 2) to suppress “illegitimate” sister chromatid recombination at replication forks. In yeast, PCNA K164 is modified by Ubc9, a SUMO-conjugating enzyme, during S phase to inhibit unwanted homologous recombination (HR) between sister chromatids (Hoege et al. 2002; Stelter and Ulrich 2003). SUMOylation of PCNA at K164 recruits the Srs2 helicase, which suppresses recombination-based salvage pathways by displacing RAD51 filaments essential for HR (Papouli et al. 2005; Pfander et al. 2005). When Srs2 is inactivated or when Elg1 removes PCNA from the primer–template junction, the likelihood of single-stranded breaks increases (Arbel et al. 2020; Fan et al. 2023). In human cells, a functional homolog of Srs2, the PCNA-Associated Recombination Inhibitor (PARI), shares a similar domain organization, including a SUMO-interacting motif and a PCNA-interacting domain. PARI also acts to displace RAD51 and suppresses inappropriate HR (Moldovan et al. 2012). Impairment of PCNA SUMOylation leads to elevated DSB formation in human cells. Although the full role of SUMOylated PCNA remains unclear, current evidence suggests it may limit breakage at stalled replication forks (Gali et al. 2012). Importantly, PCNA ubiquitination and SUMOylation only occur when the protein is chromatin-bound. Loading and unloading of PCNA are regulated by the replication factor C (RFC) complex. Several alternative RFC complexes exist, all of which share a common RFC2–5 core (Arbel et al. 2021; Kang et al. 2024). One such variant is the Elg1–RFC complex in yeast, which unloads unmodified and SUMOylated PCNA and facilitates the recruitment of DUBs (Parnas et al. 2010; Kubota et al. 2013). Notably, Elg1 interacts with both the Usp1–Uaf1 complex and PCNA and is essential for Usp1 DUB activity (Lee et al. 2010). The Elg1 homolog in mammalian cells, ATAD5, promotes unloading of both mono- and polyubiquitinated PCNA from chromatin (Kang et al. 2019). ATAD5-RFC also recruits USP1–UAF1, thereby coordinating the processes of PCNA deubiquitination and chromatin unloading in an evolutionarily conserved manner (Ryu et al. 2024).

Figure 2.

Ubiquitinated and SUMOylated PCNA residues in budding yeast and human cells. Post-translational modifications in human and yeast PCNA include lysine residues known to undergo ubiquitination only (shown in orange – K107, K242 in yeast, and K117, K248 in humans), both SUMOylation and ubiquitination (shown in red – K164 in yeast and humans), SUMOylation only (shown in lime green – K127 in yeast, and K254 in human). Modification sites have been mapped onto the crystal structures available in the RCSB Protein Data Bank: human PCNA (PDB ID: 1VYM, Kontopidis et al. 2005) and yeast PCNA (PDB ID: 1PLQ, Krishna et al. 1994).

Genetic screens in budding yeast revealed that the interaction signature of a PCNA K164R mutant most strongly correlates with the interaction profiles of mutant lagging strand genes. Based on these findings, Becker and coworkers proposed a role for ubiquitinated PCNA in sensing problems during Okazaki fragment processing (Becker et al. 2015). Lagging strand synthesis is a discontinuous process that initiates with the synthesis of an RNA primer and a short DNA oligonucleotide by DNA polymerase alpha (Pol-α)-primase. This initial primer-template structure is recognized and extended by DNA polymerase delta (Pol-δ). As Pol-δ synthesizes DNA, it displaces the 5′ end of the preceding fragment, resulting in a flap structure. This flap is subsequently processed by the endonuclease FEN1 (Sun et al. 2023). The maturation of Okazaki fragments involves the removal of these RNA–DNA primers and the joining of adjacent fragments to create a continuous DNA strand. PCNA is not only central to coordinating these events, but is ubiquitinated at K164 (Figure 2) when flap processing is disrupted (Becker et al. 2015). Similarly, in mammalian cells, ubiquitination of PCNA at K164 (Figure 2) serves to protect lagging strand synthesis from disruptions during Okazaki fragment maturation (Thakar et al. 2020). Human PCNA K164R mutant cell lines cannot fill ssDNA gaps, thereby disrupting Okazaki fragment maturation and timely unloading of PCNA. Consistent with these observations, loss of PCNA ubiquitination increases PARylation, a known response to Okazaki fragment processing defects (Hanzlikova et al. 2018). Employing substrate-trapping methodologies such as TULIP and TULIP2, which are designed to identify substrates of E3 ubiquitin ligases, Sala-Lloret and coworkers demonstrated that the BRCA1/BARD1 heterodimer mediates ubiquitination of PCNA at K164 during S phase under physiological conditions. Notably, no evidence was found for RAD18-dependent PCNA ubiquitination under these circumstances, suggesting that RAD18 activation is strictly linked to replication stress and DNA damage. This observation is consistent with findings that BRCA1-deficient cells exhibit reduced levels of ubiquitinated PCNA (Salas-Lloret et al. 2019; 2024) and that enhanced PCNA ubiquitination promotes resistance of BRCA-deficient cells to PARP inhibition (Thakar et al. 2020). This insight is clinically relevant and fits in the larger context of gap repair, which will be reviewed in the next section. Its significance is further underscored by the finding that replication gaps are an intrinsic feature of BRCA deficiency (Panzarino et al. 2021).

Lastly, an additional E3 ligase, RNF168, ubiquitinates PCNA independently of and in cooperation with RAD18. RNF168 is recruited to replication factories in a manner distinct from its canonical role in DSB repair, which involves ATM and RNF8 (Panier and Durocher 2009). RNF168 interacts with ubiquitinated PCNA through a degenerate PCNA-interacting peptide motif located at its C-terminus, as well as through an ubiquitin interacting domain, thereby contributing to the maintenance of replication fidelity under normal conditions (Yang et al. 2024).

Replication stress and fork recovery mechanisms

DNA damage tolerance pathways repair replication gaps

During DNA replication, active forks can encounter various obstacles, such as DNA lesions or nucleotide shortages that hinder their progression. The continued unwinding of the DNA duplex by the CMG helicase leads to the accumulation of ssDNA that is rapidly coated by RPA. RPA is critical for the activation of the replication stress response. It recruits ATRIP and ATR kinase (Cortez et al. 2001; Saldivar et al. 2017). ATR targets include CHK1 and p53 to trigger cell cycle arrest and inhibition of origin firing. In parallel, RPA also aids in activating PCNA-dependent DNA damage tolerance pathways (DDT). RPA recruits the E2 conjugating enzyme RAD6 and E3 ligase RAD18 to promote PCNA monoubiquitination at K164. RAD18 translocates along the RPA–ssDNA filament and is directed to PCNA trapped at the primer–template junction of the stalled fork, a process that occurs independently of ATP hydrolysis (Davies et al. 2008; Li et al. 2020). The E2 complex Ubc13/Mms2 in conjunction with the E3 ligase activity of Rad5 in yeast or the mammalian Rad5 homologs HLTF and SHPRH catalyze K63-linked polyubiquitin chain formation. Whereas monoubiquitination of PCNA facilitates translesion synthesis (TLS), which uses error-prone DNA polymerases to replicate over template strand lesions, polyubiquitination promotes template switching (TS), which utilizes the nascent strand of the sister chromatid as a template strand to circumvent the lesion (Vanoli et al. 2010; Leung et al. 2018).

TLS displays distinct features depending on whether the DNA lesion occurs on the leading or lagging strand, due to the asymmetric dynamics of the replication fork. On the leading strand, TLS polymerases can bypass lesions “on-the-fly,” substituting damaged bases and allowing (DNA polymerase epsilon) Pol-ε to resume synthesis beyond the lesion (Guilliam and Yeeles 2020; Figure 3(A)). On the lagging strand, however, a new Okazaki fragment can be initiated downstream of a lesion (Figure 3(B)). That this will ultimately cause the formation of a persistent ssDNA gap that interferes with timely Okazaki fragment processing was not directly evident, but has emerged as the field has more closely interrogated specific challenges posed by lagging strand synthesis. In vitro experiments provide evidence that Pol-δ dissociates upon encountering a lesion, and TLS polymerases are not immediately recruited, necessitating the synthesis of a new primer downstream of the damaged site, resulting in post-replicative gap filling (Hedglin and Benkovic 2017). Notably, high concentrations of Pol-δ can negatively regulate TLS by occupying the leading strand and preventing the recruitment of DNA polymerase eta (Pol-η), a TLS polymerase, to the lesion site. Therefore, “on-the-fly” TLS cannot occur and a gap forms over the lesion upon repriming. The ability to study DNA replication with purified proteins and/or cellular extracts in vitro has provided an opportunity to closely dissect the dynamics of Okazaki fragment initiation under normal conditions and genotoxic stress. Surprisingly, when nucleotides are limiting and cells lack a functional replication checkpoint, the lagging strand initiation machinery makes repeated attempts to begin synthesis of a new Okazaki fragment, creating structures that prompt PCNA loading by RFC (Canal et al. 2025). This leads to the sequestration of both PCNA and RFC, and this “depletion” from the nucleoplasmic pool causes replication fork instability, which in turn inhibits the resumption of DNA synthesis once nucleotide pools have been restored (Bertolin et al. 2025; Canal et al. 2025). Therefore, the replication checkpoint actively suppresses PCNA loading and synthesis of short Okazaki fragments, which sequesters replication factors and leaves nascent DNA vulnerable to degradation and prevents replication fork restart (Canal et al. 2025). This is interesting, because it provides an alternative to the long-standing model that the replication checkpoint primarily prevents RPA exhaustion (Toledo et al. 2013; Bertolin et al. 2025); albeit both models are not mutually exclusive. The idea that PCNA accumulation on the lagging strand causes loss of fork protection is not completely novel, as it had been discussed previously (Thakar and Moldovan 2021). Inefficient PCNA unloading enables degradation of nascent DNA by the endonuclease/helicase DNA2 in human cells that express a mutant form of PCNA that cannot be ubiquitinated at K164 (Thakar et al. 2020). Together with data in yeast, K164 ubiquitination of PCNA (Figure 2) is therefore important for timely unloading in response to lagging strand lesions or Okazaki fragment maturation defects (Yu et al. 2014). This is consistent with the finding that Δrad27 (FEN1) and Δelg1 in yeast accumulate ubiquitinated PCNA (Becker et al. 2015). To make matters more complex, the timely deubiquitination of PCNA during normal Okazaki fragment maturation also contributes to PCNA unloading and the dissociation of the Okazaki fragment processing machinery (Zamarreño et al. 2024). Interestingly, PCNA is also ubiquitinated in response to LIG1 deficiency and when mismatch repair enzymes are overexpressed (Das-Bradoo et al. 2010; Nguyen et al. 2013; Su et al. 2021; Medina-Rivera et al. 2023). In the former scenario, ubiquitin is attached to a non-canonical residue, K107 (Figure 2). Additional non-canonical ubiquitination sites have been mapped in yeast (Becker et al. 2018) and human cells (Elia et al. 2015; Lau et al. 2015; Figure 2). Although it is not clear how non-canonical ubiquitination attachment is selected, it has been proposed that a “DNA damage code” for PCNA ubiquitination exists (Das-Bradoo et al. 2010).

Figure 3.

DNA damage tolerance pathways in human cells. The replication forks in this figure depict the replisome progressing left to right, with the leading strand above the lagging strand. For leading strand replication the black DNA indicates the 3′−5′ oriented leading strand template and the gray DNA indicates the 5′−3′ nascent strand. (A) Mechanism of “on-the-fly” TLS on the leading strand. During replication, DNA Pol-ε encounters a DNA lesion, is unable to proceed and becomes uncoupled from the CMG helicase. In response, PCNA is monoubiquitinated at K164 by the RAD6–RAD18 E2–E3 ligase complex. This modification facilitates a polymerase switch from the replicative Pol-ε to a TLS polymerase, depending on the type of lesion encountered. The TLS polymerase inserts a nucleotide opposite the lesion and extends the DNA strand, therefore bypassing the damage. Once the lesion is replicated, PCNA is deubiquitinated, allowing Pol-ε to reassociate with PCNA and resume DNA synthesis. (B) Mechanism of gap formation and post-replicative filling by TLS. Gap formation during DNA replication depends on the location of the lesion relative to leading or lagging strand synthesis. In the left panel, the lesion is on the leading strand template; in the right panel, it is on the lagging strand template. Due to the physical properties of CMG helicase mediated DNA unwinding and ssDNA formation, PRIMPOL reprimes the leading strand, whereas Pol-α reprimes the lagging strand. When replicative polymerases stall at a DNA lesion, PCNA is monoubiquitinated at K164 by the RAD6–RAD18 E2–E3 ligase complex. However, in the depicted scenario an immediate polymerase switch does not occur. Despite polymerase stalling, the CMG helicase continues to unwind DNA, leading to the accumulation of ssDNA, which is quickly coated by RPA. PRIMPOL activity is tightly regulated by BRCA2 and MCM10, which prevent its association with ssDNA. However, upon depletion of these regulators, or through other unknown mechanisms, PRIMPOL reprimes DNA synthesis downstream of the lesion on the leading strand template. On the lagging strand template, Pol-α reprimes a new Okazaki fragment past the lesion. On both strands, repriming gives rise to ssDNA gaps. These gaps are filled post-replicatively in by TLS polymerases in a PCNA-dependent manner. (C) Template switching. DNA lesions may be repaired through a less mutagenic, error-free pathway that uses the sister chromatid as a template. PCNA is first monoubiquitinated at K164 by the RAD6–RAD18 complex. Subsequently, UBC13, together with either HLTF or SHPRH, extends the modification by adding K63-linked polyubiquitin chains. ZRANB3 is recruited by polyubiquitnated PCNA to stimulate fork reversal. SMARCAL1 is a helicase with DNA annealing activity that is directed by RPA (not shown) to act on DNA on the leading strand. HLTF is the third enzyme promoting fork reversal, which is a key step in initiating template switching.

Several independent lines of evidence suggest that TLS-mediated gap filling does not always occur at the replication fork, but post-replicatively (Figure 3(B)). In yeast, the Y-family DNA polymerase Rev1 that acts in concert with DNA polymerase zeta (Pol-ζ) is primarily expressed in G2 phase (Waters and Walker 2006), and gap filling by TS (Figure 32(C)) is therefore predominant in S phase, whereas TLS is highly active in G2 phase. Indeed, Karras and Jentsch showed that both TLS and TS effectively fill gaps when artificially confined to G2 phase in yeast (Karras and Jentsch 2010). A similar S- and G2-specific division of labor for DDT pathways was described for human cells where monoubiquitinated PCNA primarily promoted TLS in G2 phase and polyubiquitinated PCNA facilitated TS during S phase (Tirman et al. 2021). This explains why many RPA-marked ssDNA regions do not coincide with sites of active replication using incorporation of 5′ethynyl-2’deoxyuridine (EdU) as a readout. These regions have been termed “post-replicative repair territories”, where repriming has generated gaps that TLS, TS, or other salvage pathways later repair (Wong et al. 2020). Live-cell imaging has enabled detailed investigation of protein dynamics at stalled replication forks. Upon nucleotide depletion induced by HU, the majority of chromatin-bound PCNA is rapidly displaced. In contrast, RPA accumulates at ssDNA regions for up to 1.5 h. ATR checkpoint activation mitigates excessive DNA unwinding and prevents RPA exhaustion (Dyankova-Danovska et al. 2025). Interestingly, despite PCNA removal, RAD18 is still recruited to sites of DNA damage, suggesting that a residual pool of PCNA remains and can be monoubiquitinated at HU-stalled forks (Chang et al. 2006).

To bypass DNA lesions, the replicative polymerases Pol-δ and Pol-ε must be replaced by TLS polymerases (Figure 3(A)), primarily from the Y-family, such as Pol-η, Polι-, Pol-κ, Pol-λ, and REV1, as well as the B-family polymerase Pol-ζ, which retains two subunits of Pol-δ (Gianluca and Korzhnev 2024). Two models have been proposed to explain the mechanism underlying the polymerase switch: one posits a sequential exchange between replicative and TLS polymerases, while the other suggests the formation of a multi-enzyme “mega-complex” similar to a tool belt in which both TLS and processive polymerases are simultaneously tethered to PCNA, coordinated by the scaffolding function of REV1 (Freudenthal et al. 2010; Dyankova-Danovska et al. 2025). This issue remains an active area of investigation.

Regulation of RAD18

In human cells, the recruitment and activity of RAD18 at stalled replication forks are tightly regulated by a network of PTMs and protein–protein interactions. PARP10, an ADP ribosyltransferase that catalyzes mono-ADP ribosylation, promotes PCNA ubiquitination by interacting with and modifying RAD18. PARP10 is recruited to ssDNA gaps through its PCNA-interacting motif. Depletion of PARP10 leads to the accumulation of replication-associated gaps, a process mediated by MRE11 (Khatib et al. 2024). Additionally, REV1 facilitates RAD18 accumulation at stalled forks through direct interaction with ubiquitinated RAD18 (Wang et al. 2016). RAD18 is also subject to negative regulation at stalled forks through the ATR kinase pathway. Formation of ssDNA activates the ATR signaling cascade, which in turn phosphorylates RAD18 at S403, a site adjacent to its PCNA-interacting peptide (PIP) box. This phosphorylation disrupts the RAD18–PCNA interaction (Chen et al. 2024). Besides RAD18 and BRCA1/BARD1, the E3 ubiquitin ligase RFWD3 has been implicated in DDT, as its depletion results in decreased PCNA mono- and polyubiquitination (Gallina et al. 2021).

In addition to its role in replication-associated repair, RAD18 is also recruited to DSBs in a manner dependent on the E3 ubiquitin ligase RNF8 and the E2 enzyme UBC13. Notably, RAD18’s function at DSBs is independent of PCNA and TLS activity. Instead, RAD18 is thought to promote HR by facilitating the recruitment of RAD51or inhibition of 53BP1, although the exact mechanism remains to be fully elucidated (Huang et al. 2009; Nambiar et al. 2019). Moreover, this effect appears to be cell-type specific (Rogers et al. 2025). RAD18 also acts as a scaffold for the recruitment of SLF1, SLF2, and the SMC5/6 complex, which are important for maintaining chromatin stability at damaged sites (Räschle et al., 2015; Ryder et al., 2024). Pathogenic variants in SLF2 and SMC5 cause Atelis syndrome (Grange et al. 2022), a rare disease in the class of genome instability syndromes.

PCNA deubiquitination

Deubiquitination of PCNA is essential for the resumption of high-fidelity DNA synthesis and for preventing excessive mutagenesis by TLS polymerases. Additionally, maintaining a pool of unmodified PCNA is critical, as its depletion impairs subsequent rounds of DNA replication. In human cells, USP1 is a key DUB responsible for the removal of monoubiquitin from PCNA (Huang et al. 2006). USP1 requires the presence of a cofactor, USP1-Associated Factor (UAF1), for its enzymatic activity, as USP1 is catalytically inactive in its absence (Cohn et al. 2009). Notably, UAF1 contains two SUMO-like domains, which display sequence and structural similarity with human SUMO2, and are required for its regulation of PCNA ubiquitination (Yang et al. 2011). Functionally, USP1 plays a pivotal role in replication fork protection by negatively regulating PCNA ubiquitination. Notably, USP1 is overexpressed in BRCA1-deficient cells, where it significantly contributes to fork stabilization by limiting the engagement of TLS polymerases (Lim et al. 2018; Simoneau et al. 2023). USP1 activity is also modulated by autocleavage in response to UV-induced DNA damage, leading to the accumulation of monoubiquitinated PCNA (Huang et al. 2006). In addition to USP1, USP7 has been implicated in the deubiquitination of monoubiquitinated PCNA (Kashiwaba et al. 2015; Masuda et al. 2019). Interestingly, USP1 suppresses UV-induced mutagenesis, whereas USP7 suppresses oxidative stress-induced mutagenesis (Kashiwaba et al. 2015). To what extent DUBs affect not only the balance between monoubiquitinated and unmodified PCNA, but also the ratio of mono- to polyubiquitinated PCNA is still under investigation and may actively contribute to DDT pathway choice.

PRIMPOL activity fills replication gaps

PRIMPOL is a DDT protein with both primase and polymerase activities (García-Gómez et al. 2013; Kobayashi et al. 2016; Sale 2016). It plays a key role in the repair of replication-associated gaps, functioning during S and G2 phase through distinct mechanisms. In S phase, gap repair is more dependent on TLS, whereas in G2 phase, repriming becomes the dominant pathway. The decision between TLS and repriming is modulated by multiple cellular factors. Notably, cells appear to favor repriming in contexts where fork reversal is compromised, such as in the absence of SMARCAL1 or following PARP inhibition (Tirman et al. 2021). PRIMPOL initiates repriming downstream of a leading strand template lesion, generating an ssDNA gap that is subsequently filled post-replicatively by TLS (Figure 3(B)). PRIMPOL also contributes to cisplatin resistance in BRCA-deficient cells. ATR pathway activation induces elevated repriming activity by PRIMPOL through CHK1 phosphorylation, thereby limiting reliance on fork reversal-based mechanisms (Quinet et al. 2020; Mehta et al., 2022). Furthermore, MCM10 bound to the MCM2–7 complex recruits BRCA2 to arrest fork progression following DNA damage. This interaction suppresses PRIMPOL-mediated repriming, and safeguards the formation of ssDNA gaps (Kang et al. 2021; Figure 3(B)).

Fork reversal as a mechanism of fork protection

Fork reversal is a critical protective mechanism that prevents replication fork collapse upon encountering DNA lesions. This process involves the reannealing of the parental DNA strands and the annealing of newly synthesized (nascent) strands, resulting in the formation of a four-way, Holliday junction-like structure (Figure 3(C)). Fork reversal is facilitated by a set of specialized proteins including SMARCAL1, ZRANB3, HLTF, and RAD51. Reversal stabilizes stalled forks and enables several downstream outcomes: (1) it allows replication to resume on the “switched” template strand and thereby circumventing a template lesion (Figure 3(C) on the right), (2) it promotes the activation of nearby dormant replication origins, or (3) it prepares the fork structure for repair pathways that require duplexed DNA, such as homologous recombination. For a more detailed discussion on the action of fork reversal enzymes, readers are referred to a recent review on this subject (Adolph and Cortez 2024).

ZRANB3 is recruited to stalled replication forks in a manner that depends on the polyubiquitination of PCNA (Ciccia et al. 2012). This recruitment is mediated by its NPL4 zinc finger domain, which specifically recognizes and binds to K63-linked polyubiquitin chains on PCNA. These polyubiquitin chains are catalyzed by the E3 ubiquitin ligases SHPRH and HLTF (Mailand et al. 2013; Figure 3(C)). However, RFWD3 (FANCW) has also been implicated in this process and is necessary to recruit ZRANB3 (Moore et al. 2023; not shown in Figure 3(C)). In addition to modifying PCNA, RFWD3 targets SMARCAL1 for ubiquitination, a modification that facilitates SMARCAL1 dissociation from RPA-coated ssDNA. This regulation prevents excessive fork remodeling and thereby protects genome stability (Yates et al. 2024). Lastly, HLTF functions in polyubiquitination of PCNA and fork reversal (Adolph and Cortez 2024). Its action is tightly regulated by the replication checkpoint (Bertolin et al. 2025). In its absence, the HLTF helicase processes unprotected nascent DNA and leads to hyperaccumulation of ssDNA and subsequent fork collapse (Bertolin et al. 2025). This is an example of a “dual-faced regulator” that turns into a threat under adverse conditions. Similarly, RNF4 has been implicated in fork recovery and DNA repair (see next section; Chang et al. 2021). However, in ATR-deficient cells, the processing of stalled forks into DSBs is promoted by RNF4 function (Ragland et al. 2013), revealing the multi-faceted nature of these enzymes.

Another well-known target of RNF4 is DNA topoisomerase II alpha (TOP2A), a critical enzyme involved in replication fork reversal, where it functions to resolve positive supercoils. The regulation of TOP2A is mediated through PTM and targeted degradation. Specifically, the SUMO E3 ligase ZATT facilitates SUMOylation of TOP2A, which is subsequently recognized by RNF4. RNF4 targets SUMOylated TOP2A for proteasomal degradation, thereby modulating the extent of fork reversal. This degradation event prevents the recruitment of the DNA translocase (PLK1)-interacting checkpoint helicase (PICH) to collapsed forks, effectively limiting PICH-mediated excessive fork remodeling (Tian et al. 2021; Ding et al. 2022).

As alluded to above, fork reversal requires a substantial amount of fine-tuning. In BRCA2-deficient cells, replication stress triggers a pathological form of fork reversal characterized by defective RAD51 loading, which leads to accumulation of ssDNA. If not properly protected, this ssDNA becomes vulnerable to degradation by nucleases such as MRE11 and EXO1. Nonetheless, in the absence of BRCA2, these forks can be rescued through break-induced replication (BIR), mediated by MUS81 and POLD3 (Lemaçon et al. 2017). As we will describe later in this review, in humans, BIR is regulated by both ubiquitin and SUMO. In contrast to BRCA2 mutants, BRCA1-deficient cells exhibit enhanced replication fork recovery facilitated by PCNA polyubiquitination, which is mediated by RAD18 and the E2 enzyme UBC13. This process requires additional factors, including RNF168 and PALB2 (Cybulla et al. 2024).

Relocation of collapsed forks to the nuclear core complex is regulated by ATR and SUMO

SUMOylation dynamics play a crucial role in the cellular response to replication fork collapse (Chang et al. 2021). The balance between SUMO conjugation and deconjugation modulates the ability to arrest or restart HR-mediated repair. When replication forks persistently stall or collapse, a specialized repair program is initiated that often involves relocation of the damaged forks to nuclear pore complexes (NPCs; Figure 4), a process highly dependent on SUMO signaling in yeast. In fission yeast, the SUMO E3 ligase Pli1 protects nascent DNA from excessive resection and mediates fork relocation to the NPC. However, Pli1 also delays HR-dependent DNA synthesis until forks have relocated and the SUMO protease Ulp1 removes SUMO modifications, allowing repair synthesis to proceed (Kramarz et al. 2020). At expanded CAG repeats in budding yeast, the SUMO ligase Mms21 modifies proteins such as RPA, Rad59, and Rad52, creating cumulative SUMO signals that are recognized by the SIM domain of Slx5, part of the Slx5/Slx8 heterodimer, the analog of human RNF4 (Whalen et al. 2020). The first evidence that a STUbL regulates recruitment and repair of broken DNA at NPCs was uncovered in 2008 by a landmark study using budding yeast and a combination of imaging, genetic and biochemical interaction studies (Nagai et al. 2008). Subsequently, polySUMOylation emerged as a key regulator (Horigome et al. 2016). It wasn’t until recently that the concept of recruiting stalled forks to the NPC could be extended to mammalian systems, albeit it remains unclear if SUMO and RNF4 have similar roles in higher eukaryotes as their counterparts in yeast. Using a novel experimental approach to identify proteins associated with DNA replication, which combines sucrose gradient fractionation of DNA-protein complexes and a label-free mass spectrometry (MS) algorithm, iPOND2-DRIPPER, effectively distinguishes replisome-bound proteins from those on post-replicative chromatin by exploiting the distinct density of replisomes. Compared to other methods that pair replisome capture with quantitative MS, iPOND2-DRIPPER achieved up to a 300-fold enrichment of known replication factors (Rivard et al. 2024). The method also maps site-specific ubiquitination events without requiring an enrichment of di-glycine containing peptides. After aphidicolin treatment, NPC proteins were markedly enriched at stalled forks. The timely association and dissociation from the NPC was dependent on ATR kinase activity. p97 regulated the ubiquitin-dependent turnover of proteins involved in fork relocalization rather than dismantling the replisome itself. In addition, RPA was displaced from stalled forks and RAD51, which directs recombination-mediated restart, was loaded onto DNA. RAD51 levels were tightly controlled by p97, consistent with observations in budding yeast that showed that Rad51 was not binding to stalled forks until they localized to NPCs (Whalen et al. 2020; Rivard et al. 2024). Independently, two studies in fission yeast describe similar dynamics (Kramarz et al. 2020; Schirmeisen et al. 2024). Therefore, at least in budding and fission yeast, a model is emerging in which SUMO-modified RPA and other proteins facilitate fork relocalization to NPCs where localized STUbL activity mediates removal of RPA enabling RAD51 nucleofilament formation under tight spatiotemporal control.

Figure 4.

Localization of stalled replication forks to the nuclear periphery in yeast. Following replication fork stalling (pink panel, bottom left), the CMG helicase continues to unwind DNA, leading to the accumulation of ssDNA, which is coated by RPA (blue panel on the left). SUMOylation of RPA (gray panel) is required for the relocalization of stalled forks to the NPC under the control of the checkpoint kinase ATR. STUbLs play a role in RPA turnover by recognizing SUMOylated RPA and adding ubiquitin, which targets RPA for Cdc48/p97-dependent displacement from DNA (not shown). Fork reversal through template switching (gray panel on the right) stabilizes the replisome by loading RAD51 onto nascent DNA. There is no direct evidence that fork reversal occurs in the nuclear periphery. Displacement of RPA (blue panel on the right) is thought to aid in reannealing of nascent DNA to their respective template strands (not shown) while RAD51 shields nascent DNA from degradation. RAD51 is then required for recombination-mediated restart (green panel, bottom right).

FA and replication traverse of DNA interstrand crosslinks

The FA pathway

The FA pathway plays a central role in the repair of ICLs, a highly cytotoxic form of DNA damage that engages multiple DNA repair mechanisms downstream of D2–I ubiquitination, including nucleotide excision repair, TLS, and HR. The toxicity of ICLs is exploited in several anticancer chemotherapeutic agents, such as cisplatin and psoralen. A more comprehensive discussion of ICLs in the context of cancer therapy can be found elsewhere (Wang and Gautier 2010; Huang and Li 2013; Bhattacharjee and Nandi 2017; Niraj et al. 2019).

FA is a rare hereditary disorder caused by mutations in one of 23 genes encoding components of the FA pathway (Moreno et al. 2021; Harrison et al. 2025). Patients exhibit stunted growth, increased susceptibility to various cancers, including leukemia, breast cancer, and head and neck cancers (Alter 2003; Rosenberg et al. 2003; Webster et al. 2022; Engel et al. 2024), as well as bone marrow failure. In addition to hematological and oncological manifestations, FA is often associated with developmental abnormalities such as short stature and malformations of the skin, thumbs, and upper limbs (Shimamura and Alter 2010). Beyond ICL repair, FA pathway proteins contribute broadly to genome maintenance, including checkpoint activation; for instance, FANCM/FAAP24 facilitates ATR signaling by recruiting RPA to forks that stall at ICLs (Huang et al. 2010). The FA core complex consists of 14 proteins, including FANCA, FANCB, FANCC, FANCD, FANCF, FANCG, FANCL, and FANCM, along with several auxiliary factors such as FAAP10 (MHF2), FAAP16 (MHF1), FAAP20, FAAP24, and FAAP100 (Castella et al. 2015; Harrison et al. 2025). Among these, FANCL serves as an E3 ubiquitin ligase, necessary for the monoubiquitination of FANCD2 and FANCI (Meetei et al. 2003), whereas FANCM functions as a DNA translocase (Blackford et al. 2012; Abbouche et al. 2024). The core complex does not act on free FANCD2 but specifically targets the D2–I heterodimer. Once monoubiquitinated, FANCD2 remains bound to FANCI and clamps onto chromatin at the site of DNA damage to facilitate repair (Tan et al. 2020). The D2–I heterodimer coordinates several subsequent steps: (1) endonucleolytic cleavage adjacent to the crosslink on one DNA strand, giving rise to a DSB, (2) TLS polymerases that bypass the remaining lesion on the other strand, and (3) HR effectors that repair the DSB. For a more detailed and mechanistic review of these steps, we refer the reader to previous publications (Zhang and Walter 2014; Haynes et al. 2020). Additionally, the diverse and specific functions of FANCD2 are thoroughly discussed in an earlier review (Ishiai 2021). Here, we aim to highlight the processing of replication forks that stall at ICLs and the roles of D2–I in fork protection as it relates to mitotic DNA synthesis (MiDAS).

Replication traverse

ICLs block replication by arresting the progression of the CMG helicase, thereby halting DNA unwinding. Theoretically, an ICL constitutes the ultimate barrier for replication. Surprisingly, however, similar to lesions on a template strand, cells have the ability to tolerate ICLs, bypass them through a process known as replication traverse, and repair them post-replicatively (Williams et al. 2013; Zhang et al., 2021; Ahmed et al. 2024).

The hallmark paper that first described the process of replication fork traverse was published in 2013 (Huang et al. 2013). The Seidman group developed a single-molecule imaging technique that visualizes ICLs (generated through digoxygenin-tagged trimethyl psoralen) at replication forks in vertebrate and mammalian cells (Huang et al. 2013). Their findings demonstrated that ICLs are often (~70% of the time) traversed rather than excised and repaired in S phase (Figure 5). The traverse mechanism is dependent on several key factors, most notably the FANCM/MHF complex, which binds both DNA and components of the replisome, including phosphorylated MCM2 (Huang et al. 2019, 2013). Once recruited to the fork, FANCM facilitates replisome remodeling, promotes fork traverse across the ICL, and GINS is released from the CMG complex in an ATR-dependent manner (Huang et al. 2019). DNA synthesis then resumes distal to the ICL in the original direction of replication and requires the translocase function of FANCM. To accomplish this, FANCM interacts with PCNA through its PIP-box. Mutations in the PIP-box impair ICL traverse and prevent FANCD2 monoubiquitination (Rohleder et al. 2016). Recent evidence suggests that FANCM is primarily needed during late stages of S phase when heterochromatin is replicated. During early S phase in euchromatic regions, the non-catalytic replisome factor DONSON facilitates the restart of DNA synthesis after GINS removal (Zhang et al. 2020). The traverse mechanism also requires PRIMPOL which is recruited to ICLs in an RPA-dependent manner (González-Acosta et al. 2021; not shown in Figure 5).

Figure 5.

Different models of ICL repair initiation in human cells. The replisome stalls upon encountering an ICL. The majority of the time (~70%), the replication fork traverses the ICL (top) with the assistance of either the FANCM translocase or the DONSON protein. In the process of replication fork traverse GINS disassembles from CMG. Less frequently (~15%), a single replication fork (middle) or two converging replication forks (~10%; bottom) encounter the ICL simultaneously, initiating repair (Zhang et al. 2020).

ICL Repair

The exact mechanism by which ICL repair is initiated during S phase remains debated, as different models propose varying requirements for fork convergence and the substrate used for initial incision (Zhang and Walter 2014). Some studies suggest that a single stalled fork is sufficient to initiate ICL repair (Nakanishi et al. 2011), while others indicate that two opposing forks must converge at the ICL to create a suitable substrate for repair (Zhang et al. 2015; Figure 5). In support of the latter model, a single fork collision does not promote the unloading of the CMG complex, rendering the fork unsuitable as a substrate for incision-making enzymes and inhibiting the binding of downstream ICL proteins to chromatin. Furthermore, double-fork convergence may serve as a safety mechanism to prevent excessive CMG unloading, which could lead to fork collapse, as there is no means to reload CMG (Long et al. 2014; Zhang et al. 2015).

FANCM, along with assisting proteins, such as FAAP24, MHF1, and MHF2, is thought to act as a sensor that recognizes stalled forks at the ICL. This is based on evidence showing that FAAP24 interacts with FANCM, targeting and loading FANCM onto branched DNA structures in vitro (Ciccia et al. 2007). Both FANCM and FAAP24 are necessary for FANCD2 monoubiquitination, likely through the recruitment of FA core complex proteins (Ciccia et al. 2007; Kim et al. 2008; Sobeck et al. 2009). FAAP100 (FANCX) associates with FANCB and FANCL, and its absence prevents FANCD2 and FANCI ubiquitination (Harrison et al. 2025).

Downstream of D2–I ubiquitnation, multiple nucleases are involved in ICL repair, as it requires the coordination of multiple DNA repair mechanisms (Andreassen and Ren 2009). A detailed analysis of nucleases, their specific substrates, and their potential roles in ICL repair have been discussed in other reviews (Sengerová et al. 2011; Zhang and Walter 2014). The XPF (FANCQ)-ERCC1 heterodimer, a structure-specific endonuclease, cleaves synthetic ICL substrates, an activity that requires SLX4 (FANCP) scaffolding and FANCD2 monoubiquitination (Kim et al. 2011). Moreover, FANCD2 monoubiquitination is essential for the recruitment of SLX4 and, consequently, XPF-ERCC1 to the ICL (Klein Douwel et al. 2014). Following the incision and unhooking of ICLs, ICL repair requires TLS polymerases. These polymerases generate a substrate suitable for HR-mediated repair (Ho and Schärer 2010; Roy and Schärer 2016). The final step in ICL repair is HR, which resolves replication-associated DSBs generated during the processing of crosslinks, thereby maintaining genomic stability.

FANCD2 monoubiquitination

FANCD2 monoubiquitination is critical for ICL repair, as lesions persist when this process is disrupted by mutations at K561 (Garcia-Higuera et al. 2001). However, the exact role of the D2–I complex in ICL repair is complicated due to FANCD2’s multiple roles and functions. Currently, the most widely accepted model suggests that the D2–I complex acts as a scaffold, facilitating the recruitment of other proteins, especially nucleases, while coordinating their enzymatic activities. It also provides protection for stalled replication forks and promotes their recovery. FANCD2 forms a heterodimeric complex with FANCI that adopts a clamp-like conformation around DNA (Joo et al. 2011; Alcón et al. 2020; Rennie et al. 2020; Tan et al. 2020; Wang et al. 2020; 2021). In vitro biochemical studies suggest that the D2–I complex exists in multiple conformational states that are regulated by PTMs and interactions with specific DNA structures. According to current models, the D2–I complex initially forms a loosely closed “open” clamp that encircles and diffuses along DNA. Upon encountering a dsDNA-ssDNA junction, such as those present at stalled replication forks, the complex undergoes a conformational change to a tightly “closed” state. This transition is stabilized by monoubiquitination of FANCD2, effectively “locking” the D2–I complex onto DNA. The clamp-closing step is also dependent on ATR-mediated phosphorylation of both FANCD2 and FANCI. The D2–I complex binds directly to stalled replication forks at ICLs via a large central cavity that accommodates dsDNA, with assistance from the FANCD2 tower domain. Biochemical data indicate that the D2–I complex has a higher binding affinity for ICL-containing fork structures than for ICL-free DNA (Liang et al. 2016). This specificity is supported by in vitro evidence showing that branched or duplex DNA structures, rather than ssDNA or chromatinized DNA, preferentially stimulate FANCD2 monoubiquitination within the D2–I complex (Longerich et al. 2014). Similarly, studies in mouse models confirm that D2–I exhibits strong binding affinity for dsDNA-ssDNA junctions that resemble ICL-stalled forks (Joo et al. 2011), underscoring its role in lesion recognition during replication.

In its unmodified state, the D2–I complex is unable to bind ubiquitin, as the critical monoubiquitination sites (FANCD2-K561 and FANCI-K523) are buried at the interface of the heterodimer. Cryo-EM structural analyses have confirmed that these sites are inaccessible in the free, unbound conformation (Joo et al. 2011; Wang et al. 2021). This supports a model in which the D2–I complex is recruited to stalled forks prior to ubiquitination. Upon localization, the FA core complex induces conformational remodeling of D2–I, closing the clamp around DNA and separating the D2–I interface. This conformational change exposes the lysine residues for monoubiquitination (Wang et al. 2021). Ubiquitination then stabilizes the complex on DNA but does not itself mediate direct interactions with downstream repair factors (Alcón et al. 2020; Tan et al. 2020; Wang et al. 2020). Rather, monoubiquitination reinforces the retention of D2–I at ICLss, effectively “coating” damaged DNA and coordinating downstream repair processes while protecting replication forks. Notably, several ICL repair proteins, including SLX4 and REV1, possess ubiquitin-binding domains, implying a more complex and potentially indirect regulatory interaction with the ubiquitinated D2–I complex (Moldovan et al. 2010; Lachaud et al. 2014).

Monoubiquitination of FANCD2 is catalyzed by the FA core complex and the E2 ubiquitin-conjugating enzyme UBE2T/(FANCT) (Alpi et al. 2007). The role of FANCI ubiquitination, however, is less clearly defined. Some studies suggest that dual monoubiquitination of FANCD2 and FANCI is necessary to maintain the stability of the heterodimer on DNA (Smogorzewska et al. 2007), while others propose that FANCD2 ubiquitination alone is sufficient to stabilize DNA binding. In this latter model, FANCD2 ubiquitination promotes a structural rearrangement in the D2–I complex that enhances its DNA-binding affinity, with FANCI ubiquitination being dispensable for this function (Rennie et al. 2020). Another hypothesis proposes that FANCD2 can interact noncovalently with ubiquitinated FANCI via its CUE (coupling of ubiquitin conjugation to ER degradation) domain, shielding FANCD2 from proteasomal degradation (Rego et al. 2012). This is consistent with findings showing that the FANCD2 ubiquitin moiety is less accessible to the USP1–UAF1 deubiquitinase when D2–I is DNA-bound (Rennie et al. 2020). Conversely, other studies show that deubiquitination of FANCI by USP1 facilitates the recruitment of the FA core complex (Castella et al. 2015), indicating that FANCI ubiquitination and deubiquitination dynamically regulate the timing and function of D2–I activity.

Both RPA and ATR kinase activity are essential for FANCD2 monoubiquitination and full activation of the FA pathway (Andreassen et al. 2004; Smogorzewska et al. 2007). ATR is required for the recruitment of the FA core complex (Castella et al. 2015) and also phosphorylates FANCI at three conserved residues within a flexible domain (Ishiai 2021), priming the D2–I complex for ubiquitination (Sijacki et al. 2022). ATR-mediated phosphorylation promotes clamp closure (Sijacki et al. 2022) and prevents premature deubiquitination of FANCD2 (Rennie et al. 2020). In contrast, phosphorylation of FANCD2 by casein kinase 2 (CK2) negatively regulates its function by inhibiting both its recruitment to ICLs and its monoubiquitination (Lopez-Martinez et al. 2019). The phosphatase PP2A is required to remove this inhibitory phosphorylation at a conserved cluster (residues 882–898), thereby enabling proper chromatin loading of the D2–I complex (Yang et al. 2024).

FANCD2–FANCI control by PTMs

The recruitment and retention of the D2–I complex at sites of DNA damage are tightly regulated by a complex network of PTMs. Traditionally, removal of the D2–I complex from chromatin has been attributed to deubiquitination mediated by the USP1–UAF1 complex (Nijman et al. 2005). However, an alternative mechanism has been identified under conditions of CDK-driven replication stress, in which FANCD2 is polyubiquitinated by the E3 ligase FBLX12 in a CHK1-dependent manner. This modification targets FANCD2 for proteasomal degradation and facilitates replication fork progression (Brunner et al. 2023). In addition to ubiquitination, SUMOylation plays a critical role in regulating the D2–I complex under genotoxic stress. Upon exposure to UV light or ionizing radiation, FANCD2 and FANCI are SUMOylated in an ATR-dependent manner by the E3 ligases PIAS1 and PIAS4, in conjunction with the SUMO-conjugating enzyme UBC9. SUMOylated D2–I is subsequently recognized by the STUbL RNF4, which adds polyubiquitin chains to the complex. These chains serve as signals for extraction by the AAA+ ATPase p97, facilitating disassembly and turnover of the D2–I complex from chromatin (Gibbs-Seymour et al. 2015).

FA proteins have different roles in fork protection

FANCD2 is enriched at stalled replication forks where it plays a critical role in fork protection and stability (Lossaint et al. 2013). It functions to suppress nuclease activity and stabilize replication intermediates, thereby preventing fork collapse. In HR, FANCD2 is essential for the proper localization of CtIP to promote end resection (Unno et al. 2014). At stalled forks, FANCD2 directly binds to the DNA2 endonuclease, thus shielding nascent DNA (W. Liu et al. 2023), and also prevents degradation of stalled forks by MRE11 (Schlacher et al. 2012). Under replication conditions of nucleotide exhaustion, monoubiquitination of FANCD2 is required to prevent fork collapse and ensure fork stability (Schlacher et al. 2012). Monoubiquitinated FANCD2 also promotes the recruitment of BRCA2 to chromatin, facilitating RAD51 loading at appropriate DNA damage sites (Wang et al. 2004). The D2–I complex directly interacts with RAD51, stabilizing RAD51 nucleoprotein filaments and protecting the 5′ ends of DNA from resection by FA-associated nucleases, including FAN1. Notably, FANCD2 and RAD51 cooperatively function in fork protection independently of FANCI, guarding against degradation by MRE11, EXO1, and DNA2. In uveal melanoma, which exhibits reduced sister chromatid exchange, FANCD2 expression is significantly lower than in normal cells. Complementation with functional FANCD2 restores, highlighting its pivotal role in supporting RAD51 activity and genome stability (Gravells et al. 2013).

Beyond its role in DNA repair, FANCD2 is also involved in replisome surveillance. Upon replication stress, it binds to phosphorylated MCM2–7 and moderates helicase progression. In the absence of FANCD2, cells display extended tracts of ssDNA due to unrestrained MCM2–7 helicase activity (Lossaint et al. 2013). Upon mitomycin C (MMC)–induced DNA damage, monoubiquitinated FANCD2 helps to retain the MCM2–7 complex at dormant origins. This MCM2–7:D2–I assembly also interacts with RAD51, and loss of MCM7 impairs RAD51 localization and efficiency (Chen et al. 2017).

Collectively, these findings support a model in which FANCD2 functions as a molecular bridge between the replisome and key HR factors such as BRCA2 and RAD51, promoting efficient repair at stalled forks. Monoubiquitinated FANCD2 colocalizes with RAD51 (Taniguchi et al. 2002) and BRCA1/2 in response to DNA damage (Garcia-Higuera et al. 2001; Taniguchi et al. 2002), reinforcing its role in coordinating HR-mediated replication fork recovery.

TRAIP is a crucial regulator of ICL repair

In Xenopus egg extracts, the collision of converging replication forks with cisplatin-induced ICLs activates the FA pathway. This process begins with nuclease-mediated cleavage of the ICLs, which incises the phosphodiester backbone, resulting in ICL unhooking and the generation of DSBs (Zhang et al. 2015). In contrast, ICLs formed by agents such as psoralen and alkylating compounds, can be cleaved by the DNA glycosylase NEIL3, which generates abasic sites that are bypassed through TLS (Semlow et al. 2016). For efficient ICL repair upon fork convergence, a key requirement for FA pathway activation is the unloading of the CMG helicase by the p97 ATPase, a process triggered by MCM7 polyubiquitination by the E3 ligase TRAIP (Wu et al. 2019). Notably, CMG unloading is not required for TLS, suggesting that TRAIP-mediated polyubiquitination of MCM7 at the ICL is a decisive event that determines repair pathway choice. CMG removal allows replication forks to pause near the lesion site, enabling the recruitment of FANCM and FAAP24. Although the precise mechanism by which TRAIP is recruited and localized to active replisomes remains unclear, TRAIP’s E3 ligase activity is specifically activated during fork convergence at ICLs (Wu et al. 2019). Further evidence suggests that one of the converging replication forks may undergo fork reversal – a process dependent on CMG unloading but independent of D2–I (Amunugama et al. 2018). The functional importance of TRAIP in ICL repair is underscored by findings that TRAIP depletion sensitizes cells to MMC, highlighting its role in resolving MMC-induced damage (Hoffmann et al. 2016). Finally, short polyubiquitin chains on CMG are sufficient for NEIL3 recruitment, whereas longer chains target CMG for p97-dependent proteasomal degradation, which permits access of endonucleases to the ICL site. Importantly, during mitosis, TRAIP is capable of ubiquitinating residual CMG complexes, promoting their degradation (Deng et al. 2019). This mechanism ensures the clearance of replisomes that persist beyond replication termination into mitosis, thereby enabling MiDAS.

Replication completion and mitotic recovery

Mitotic DNA synthesis and its role in DNA repair

Despite ATR checkpoint activation during S phase, certain genomic regions can remain under-replicated without triggering a DNA damage response and therefore persist into mitosis. During mitosis, canonical DNA repair pathways, such as HR and NHEJ, are inactivated due to the mitotic phosphorylation of RNF8, RNF168, and 53BP1, precluding their chromatin association (Audrey et al. 2022). Under-replicated regions are resolved through a specialized pathway known as MiDAS (Figure 6). MiDAS is a conservative DNA synthesis process that uses the sister chromatid as a template. MiDAS shares mechanistic features with BIR, as was first observed in yeast, although the degree of similarity remains controversial (Malkova et al. 1996; Minocherhomji et al. 2015; Kramara et al. 2018; Bhowmick et al. 2023). Both pathways involve endonuclease activities, including MUS81–EME1, SLX4, and FANCD2 scaffolding, as well as Pol-δ subunit POLD3.

Figure 6.

Mitotic DNA synthesis in human cells. In response to replication fork stalling PCNA becomes monoubiquitinated at K164 by the RAD6–RAD18 complex, which facilitates recruitment of FANCD2. The D2–I complex, acting as an open clamp, surveys the DNA and binds to ds/ssDNA junctions. ATR-mediated phosphorylation of D2–I primes FANCD2 for monoubiquitination by the FA core complex. This modification stabilizes the D2–I clamp and induces a conformational switch to the closed state. Upon mitotic entry, TRAIP becomes phosphorylated, disassembling CMG by polyubiquitination-mediated proteasomal degradation. During mitosis, monoubiquitinated FANCD2 acts as a scaffold for SLX4, which recruits structure-specific endonucleases capable of introducing DNA breaks (MUS81-EME1/EME2 complex). These mitotic DNA breaks initiate MiDAS by pol-ζ in collaboration with the PIF1 helicase.

The molecular trigger for MiDAS remains unclear, although it appears to occur at defined genomic loci prone to replication stress, such as common fragile sites (CFSs) and telomeres (Minocherhomji et al. 2015; Bhowmick et al. 2016; Özer et al. 2018). MiDAS is strongly induced by genotoxic agents, which stall replication forks and lead to persistent under-replication. Defective MiDAS results in ultra-fine bridges (UFBs) during anaphase and chromosome segregation errors (Minocherhomji et al. 2015) underscoring its importance in maintaining genomic stability (Naim et al. 2013). FANCD2 forms nuclear foci at CFSs under replication stress (Chan et al. 2009) and is a key regulator of MiDAS (Graber-Feesl et al. 2019; Xu et al. 2021). Core complex components such as FANCA and FANCL, which are required for FANCD2 ubiquitination, are also essential for MiDAS (Traband et al. 2023). Furthermore, SLX4 colocalizes with FANCD2 during MiDAS and acts as a recruitment platform for additional factors (Minocherhomji et al. 2015). SLX4 has UZB and SIM motifs that target it to DNA, however, it remains to be seen whether these domains are important for MiDAS (Gibbs-Seymour and Mailand 2015). RAD52 is required for MiDAS in cancer cells (Bhowmick et al. 2016) and recent studies suggest that RAD51 also plays a non-canonical role in this context. RAD51, classically known for HR, contributes to fork protection by promoting the reversal and restart of stalled forks. RAD51 is newly recruited in mitosis, where it facilitates MiDAS through fork protection rather than recombination (Wassing et al. 2021). Interestingly, RAD51-depleted cells exhibit MiDAS at loci distinct from canonical fragile sites (Bhowmick et al. 2016), suggesting locus- and context-specific roles. It is noteworthy that EdU incorporation during mitosis may also represent a continuation of S phase DNA synthesis that requires both RAD52 and RAD51 (Mocanu et al. 2022). However, the classic model for BIR requires CMG helicase unloading as a prerequisite for MiDAS. TRAIP E3 ligase activity mediates MCM7 polyubiquitination and subsequent CMG disassembly (Poovathumkadavil et al. 2025). In human U2OS cells, TRAIP is essential for the formation of mitotic FANCD2 and EdU foci, whereas CRL2^Lrr1 is dispensable (Sonneville et al. 2019). In Xenopus models, FANCD2 has been observed at forks where CMG remains present, possibly due to defective nuclease recruitment (Zhang et al. 2015). TRAIP depletion results in PICH-positive UFBs and increased 53BP1 bodies in the subsequent G1 phase, indicating unresolved replication intermediates (Sonneville et al. 2019). These data suggest that CMG unloading may facilitate access of MUS81–EME1/2 to stalled forks, independently of FANCD2 foci formation.

POLD3, a shared subunit of Pol-δ and Pol-ζ, is essential for MiDAS (Minocherhomji et al. 2015; Tumini et al. 2016). POLD3 is a homolog of Pol32 in yeast, which is required for BIR (Wu and Malkova 2021). PCNA monoubiquitination at K164 by RAD18 is therefore also critical (Figure 2), as it promotes the recruitment of translesion polymerases (Wu et al. 2023; Zhang et al. 2023). This parallels BIR in yeast, where Rad18-mediated PCNA ubiquitination is essential (Lydeard et al. 2010). However, ubiquitinated PCNA is not only required for initiating TLS, but also for the recruitment of PIF1 helicase (Li et al. 2021), and chromatin loading of FANCD2 (Leung et al. 2023) in human cells. FANCD2 has a PIP motif and mutations in the PIP box alleviate loading onto DNA (Howlett et al. 2009). However, human PCNA K164R mutants still express PCNA, and thus it is evident that the recruitment of FANCD2 to stalled replication forks requires PCNA ubiquitination at K164 (Figure 2). It has been proposed that the interaction between ubiquitinated PCNA and ubiquitinated FANCD2 is required to stably maintain ubiquitinated PCNA on chromatin throughout S and G2 phases into mitosis. Upon initiation of MiDAS, FANCD2 is displaced (Leung et al. 2023). Interestingly, depletion of ATAD5, a PCNA unloader, leads to increased MiDAS at telomeres and CFSs, suggesting that regulated PCNA dynamics are critical for this process (Kim et al. 2024). PCNA ubiquitination also promotes SLX4 recruitment to telomeres in ALT (alternative lengthening of telomeres) cells, supporting telomeric MiDAS (Kim et al. 2024; O’Sullivan and Greenberg 2025).

Sequencing analyses have revealed that MiDAS preferentially occurs at large, transcriptionally active genomic regions with few replication origins and replication timing in mid- to late-S phase (Macheret et al. 2020). Telomeric MiDAS occurs in both telomerase-positive and ALT cells and is RAD52-dependent but does not require MUS81–EXO1 activity (Özer et al. 2018). ALT cells, which exhibit higher levels of telomere fragility, show elevated MiDAS activity, particularly when replication fork protection is compromised (Min et al. 2017a). Although ubiquitination is critical for MiDAS, a possible role for SUMOylation has remained elusive. Interestingly, in both human and yeast ALT cells, telomere erosion triggers SUMOylation that has been linked to clustering telomeres to NPCs in yeast (Churikov et al. 2016) and ALT-associated PML bodies in humans (Min et al. 2017b). One model proposes that a polySUMO/polySIM condensate-like environment reminiscent of ALT-associated PML bodies can induce an ALT-like phenotype in telomerase-positive cells that exhibit MiDAS at telomere clusters in the presence of RAD52 and BLM helicase (Min et al. 2019). These findings are intriguing since the switch from telomere-positive to ALT has remained enigmatic but is clinically relevant to about 10% of cancers that show ALT characteristics and are difficult to treat (O’Sullivan and Greenberg 2025).

In BRCA2-deficient cells, MiDAS is preferentially activated at early S phase loci enriched in transcription-replication conflicts and R-loop structures (Groelly et al. 2022). Moreover, overexpression of cyclin E has been shown to induce DNA lesions that trigger MiDAS in response to mild replication stress, likely because cyclin E pushes cells prematurely into S phase with a deficit of licensed origins, which in turn cause under-replication (Audrey et al. 2024). MiDAS is currently an area of intense investigation and many of the details governing this pathway and variants of it are still to be discovered.

Conclusions and future directions

Ubiquitin and ubiquitin-like modifiers play critical roles in a broad array of biological processes. Here, we focused on recent advances in our understanding of how the reversible attachment of ubiquitin and SUMO – either as monomers or polymolecular chains – regulate DNA replication. First, it is important to note that active replisomes are heavily SUMOylated, whereas post-replicative chromatin or stalled forks are actively reconfigured, a process requiring ubiquitin-mediated removal of replication factors (Martín-Rufo et al. 2022). The SUMO–SIM interaction is crucial for these events, which are catalyzed by STUbLs that reside at nuclear pores (Chang et al. 2021). While these findings are well established in budding and fission yeast, the question remains whether they also hold in mammalian systems, highlighting a direction for future research. Furthermore, the replication field has experienced a renewed interest in the regulation of lagging strand synthesis, with an increasing appreciation for the intricate controls that ensure proper maturation of the approximately 30,000,000 Okazaki fragments needed to form a continuous lagging strand. Not surprisingly, eukaryotic cells have redundant surveillance mechanisms in place. The first is the replication checkpoint, which suppresses the sequestration of PCNA and RFC on lagging strand templates of stalled forks by an unknown mechanism (Bertolin et al. 2025; Canal et al. 2025); importantly, checkpoint activation depends on sufficiently high ubiquitin pools to enable RPA modification (Elia et al. 2015; Chang et al. 2023). The second mechanism involves PARP-dependent ADP-ribosylation (Hanzlikova et al. 2018; Elsborg et al. 2025), where PARP1 activity is modulated by ubiquitin-dependent protein degradation (Krastev et al. 2022; Tucker et al. 2025). Third, PCNA ubiquitination at K164 likely assists in flagging sites in distress (Becker et al. 2015), recruiting the chromatin assembly machinery, and facilitating protein turnover (Thakar et al. 2020). Shifting focus, a third important takeaway concerns the omnipresent role of the E3 ubiquitin ligase TRAIP during DNA replication, ICL repair, and MiDAS. As a PCNA interacting protein, TRAIP is part of the normal replisome but remains inactive until CMG complexes need to be unloaded during replication termination (Wu et al. 2021). TRAIP removes CMG during ICL repair, and at persistently stalled forks, where terminally stalled chromatin-bound CMG complexes are targeted for degradation to permit MiDAS (Sonneville et al. 2019). Interestingly, during protein–DNA crosslink repair, TRAIP targets the crosslinked protein but not CMG (Wu et al. 2021), a process that may involve regulatory action from DUBs such as USP37, which deubiquitinate MCM complexes (Bolhuis et al. 2025; Kochenova et al. 2025; Villa et al. 2025). Future experiments will be needed to fully uncover the regulatory network that determines TRAIP activity. Looking ahead, it is noteworthy that a largely obscure process described over a decade ago as “replication traverse” now draws new attention, as recent findings suggest the CMG helicase can accommodate bulky structures such as G-quadruplexes (Huang et al. 2013; Batra et al. 2025). Achieving a better mechanistic understanding of replication traverse is likely to reveal pathways that confer resistance to the crosslinking agents used in clinical cancer therapy. In parallel, several laboratories are dedicated to identifying targets of the more than 600 E3 ubiquitin ligases encoded in the human genome, supporting the emergence of targeted protein degradation technologies such as PROTAC, with significant clinical potential (Liu et al. 2023). Finally, as many oncogenes and tumor suppressors are SUMO substrates (Lee et al. 2017), continued profiling of tumor targets promises to reveal novel vulnerabilities in replication stress tolerance (Salas-Lloret and González-Prieto 2025).