Reconstitution of Translesion Synthesis Reveals a Mechanism of Eukaryotic DNA Replication Restart

Abstract

Leading-strand template aberrations cause helicase-polymerase uncoupling and impede replication fork progression, the details of how uncoupled forks are restarted are uncertain. Using purified proteins from Saccharomyces cerevisiae, we have reconstituted translesion synthesis (TLS) mediated restart of a eukaryotic replisome following collision with a CPD. We find that TLS functions “on the fly” to promote resumption of rapid replication fork rates, despite lesion bypass occurring uncoupled from the CMG helicase.

Surprisingly, the main lagging-strand polymerase, Pol δ, binds the leading strand upon uncoupling and inhibits TLS. Pol δ is also crucial for efficient recoupling of leading-strand synthesis to CMG following lesion bypass. PCNA monoubiquitination positively regulates TLS to overcome Pol δ inhibition. We reveal that these mechanisms of negative and positive regulation also operate on the lagging-strand. Our observations have implications for both fork restart and the division of labour during leading-strand synthesis generally.

The eukaryotic replisome is responsible for accurate and timely genome duplication. The replicative polymerases (Pols), ε and δ, perform bulk synthesis of the nascent leading and lagging strands, respectively^1–3. However, template aberrations, including DNA damage, can cause these Pols to stall⁴. To maintain DNA replication on a damaged template, DNA damage tolerance (DDT) mechanisms are required⁵. One key DDT pathway is translesion synthesis (TLS), which utilises a specialised TLS Pol to directly synthesise across the lesion before canonical replication resumes⁶. Compared to replicative Pols, TLS Pols display greatly reduced fidelity and processivity^6,7. Their contribution to nascent-strand synthesis must therefore be tightly controlled. It is currently unclear how TLS Pol access is negatively regulated when synthesis transiently stalls but TLS is not required.

PCNA monoubiquitination by the E2–E3 ubiquitin ligase complex, Rad6–Rad18, is proposed to positively regulate TLS⁸. Cellular and biochemical studies support a role for PCNA monoubiquitination in TLS Pol recruitment and stimulation of lesion bypass^9–13. However, data contradicting these findings have also been reported^14–17. Although these studies have been a matter of debate¹⁸, additional reports suggest that PCNA monoubiquitination is not strictly required for TLS^19–24. In UV-irradiated DT40 cells, PCNA monoubiquitination was not required to maintain normal fork rates but was essential for post-replicative gap filling²⁵. It is unknown if this differential requirement for PCNA monoubiquitination exists broadly in eukaryotes.

TLS of a lagging-strand lesion likely only occurs via gap filling as initiation of the next Okazaki fragment enables the continuation of synthesis.

Leading-strand TLS could also occur by gap filling following re-priming. Alternatively, it could maintain continuous leading-strand synthesis; termed “on the fly” TLS²⁵. Which of these mechanisms is predominant has been debated for many years. Observations of daughter-strand gaps in UV-irradiated yeast²⁶ and the absence of an effect on viability when TLS is delayed to G2-M^27,28 provide support for post-replicative TLS. However, these reports do not demonstrate this is the default mechanism. Indeed, Daigaku et al. observe maximal PCNA ubiquitination in S phase and identify strong S phase checkpoint activation when TLS is delayed to G2-M²⁷. TLS could therefore operate on the fly, but how this would coordinate with the replisome is unknown.

Redundancy and competing pathways pose significant challenges to the characterisation of the two models of TLS in vivo. Meanwhile, in vitro investigations have lacked the context of a eukaryotic replisome and produced conflicting results^13,14. Recently, origin-dependent eukaryotic DNA replication was reconstituted using purified budding yeast proteins^29,30. The initial response of the replisome to DNA damage was characterised^31,32. A leading-strand lesion caused stalling of synthesis by Pol ε and “uncoupling” from CMG (Cdc45-MCM-GINS) helicase activity. Template unwinding and lagging-strand synthesis persisted downstream of the lesion at a reduced rate and leading-strand re-priming was extremely inefficient. In contrast, lagging-strand lesions stalled synthesis of a single Okazaki fragment, generating a single-stranded (ss) DNA gap.

Using this system, we characterise both proposed models of TLS by locating the lesion in either the leading-strand or lagging-strand template. The mechanistic details we delineate have implications for how TLS operates alongside the replisome, fork restart, and the division of labour between Pol ε and Pol δ.

Results

Pol η facilitates lagging-strand TLS in a gap-filling manner

To investigate if TLS enables gap filling opposite a lagging-strand lesion in a reconstituted DNA replication system, we utilised the efficiency of Pol η (Extended Data Fig. 1a) in bypassing a common UV photoproduct, the cyclobutane pyrimidine dimer (CPD)⁶. DNA replication was established on a linearised 9.7 kb plasmid containing a lagging-strand CPD ~6.7 kb from the origin of replication, using the previously detailed system^29,30,32. Briefly, MCM double hexamers are loaded onto origins and activated to form two replisomes. The rightward fork encounters the CPD, while the leftward fork generates a short 1.3 kb product (run off) (Fig. 1a). Okazaki fragment maturation was promoted using Fen1 and Ligase³³ (Cdc9) (Extended Data Fig. 1a), to generate an ~1.7 kb stall product downstream of the CPD and an ~8 kb product from upstream fragments (Fig. 1a).

Fig. 1. Reconstitution of lagging- and leading-strand TLS.

(a) Schematic of the 9.7 kb ARS306 lagging-strand CPD template. The origin of replication (Ori), CPD site, and leading- (red) and lagging-strand (blue) replication products produced in the presence of Fen1 and Ligase are illustrated. A dashed line indicates lesion bypass and gap filling. Below is shown the position of the BamHI and SwaI restriction sites used for urea polyacrylamide gel analysis, in addition to the stall (165 nt) and bypass (187 nt) products generated by cleavage.

(b) Standard replication assay on the lagging-strand CPD template and undamaged equivalent in the presence of Fen1 and Ligase (Cdc9) with increasing concentrations of Pol η. Urea polyacrylamide gel samples were treated with BamHI and SwaI to generate the stall and bypass products shown in (A) (used where indicated in subsequent figures). The ‘cut lead’ product was generated by cleavage of the leading strand by BamHI and SwaI.

(c) Schematic of the 9.7 kb ARS306 leading-strand CPD template. Leading- (red) and lagging-strand (blue) replication products are illustrated with the full-length extended bypass product shown as a dashed line.

(d) Schematic representation of an uncoupled fork product present in the native agarose gel.

(e) Standard replication assay on the leading-strand CPD template and undamaged equivalent with increasing concentrations of Pol η. Denaturing gel samples were treated with SmaI to reduce heterogeneity in product length caused by variation in the site of initiation (used where indicated in subsequent figures).

(f) Lane profiles of native gel lanes 7 and 12 shown in (e).

Uncropped versions of Fig. 1 gels are available in Source Data Fig. 1.

In the absence of Pol η we observed a diffuse CPD-dependent stall product, but were unable to resolve the reciprocal ~8 kb product (Fig. 1b, denaturing, and Extended Data Fig. 1b, compare lanes 1 and 7). Priming site heterogeneity and incomplete fragment processing likely compromised the detection of a defined stall product. To more directly observe lagging-strand TLS, reaction products were cleaved with the restriction endonucleases SwaI and BamHI, mapping upstream and downstream of the CPD, respectively.

Cleavage of the stall product would generate a 165 nt product with the 5' end defined by SwaI and the 3' end by stalling at the CPD. TLS and extension of the nascent lagging strand past the BamHI site would generate a 187 nt bypass product (Fig.1a). Because BamHI cleaves DNA in a staggered manner, digested leading strands (191 nt) were resolved from bypass products by analysis on urea polyacrylamide gels.

In the absence of Pol η, a 165 nt CPD-dependent product was produced with no detectable bypass, confirming synthesis of an Okazaki fragment had stalled at the lesion (Fig. 1b, urea polyacrylamide, compare lanes 1 and 7). Pol η promoted conversion of the stall product to a 187 nt bypass product in a concentration-dependent manner (Fig. 1b, urea polyacrylamide lanes 7-12). Furthermore, the diffuse stall product in the denaturing gel was less prominent in reactions containing Pol η, indicating gap filling and ligation (Fig. 1b and Extended Data Fig. 1b, denaturing lanes 7-12). Collectively, these results demonstrate Pol η promotes TLS at lagging-strand gaps.

Leading-strand TLS functions on the fly

Leading-strand CPDs cause helicase-polymerase uncoupling (Fig. 1d), slowing fork progression³². To analyse Pol η function on the leading strand we linearised the plasmid to place the CPD 3 kb from the origin in the leading-strand template (Fig. 1c). Unless stated in figures, leading-strand experiments were performed without Fen1 and Ligase. On an undamaged template Pol η did not noticeably affect replication and full-length products (Fig. 1e, native, lanes 1-6) containing leading strands of ~8.2 kb (FL-lead) (Fig. 1e, denaturing, lanes 1-6 and Extended Data Fig. 1c) were generated. Consistent with our previous work³², in the absence of Pol η, replication of the CPD template predominantly generated uncoupled forks (Fig. 1d and e, native, lane 7) containing leading strands of ~3 kb (Stall) (Fig. 1e, denaturing, lane 7 and Extended Data Fig. 1d).

Pol η promoted full-length product formation on the CPD template and concomitantly reduced uncoupled forks, with full-length products being the major reaction species in the presence of 16 nM Pol η (Fig. 1e, native and f). Analysis of nascent leading strands revealed a Pol η-dependent increase in FL-lead (Fig. 1e, denaturing, lanes 7-12), which was exclusively associated with full-length products (Extended Data Fig. 1e). These data demonstrate Pol η promotes on the fly TLS across a leading-strand CPD. This prevents and, or, alleviates fork stalling, accelerating their arrival at the end of the template to generate full-length products. Since TLS increased fork rate downstream of damage, and coupling of leading-strand polymerisation to CMG is a key factor in determining this rate^31,34, our data indicate Pol η-mediated CPD-bypass enables maintenance or re-establishment of CMG-coupled leading-strand synthesis.

On the fly TLS operates uncoupled from CMG

The restoration of rapid fork rates by on the fly TLS (Fig. 1e and f) could be via a concerted mechanism facilitating the swift exchange of Pol ε and Pol η.

However, CMG continues to advance at a reduced rate after stalling of leading-strand synthesis³¹, potentially meaning TLS operates uncoupled from CMG, therefore necessitating a mechanism to efficiently recouple synthesis to CMG.

To differentiate between these two possibilities, we used oligonucleotides mapping downstream of the CPD that promote recoupling of leading-strand synthesis without impacting replication of an undamaged template³², to compete with Pol η. A concerted mechanism should limit ssDNA exposure on the leading strand, preventing oligonucleotide binding so that FL-lead products are generated (Fig. 2a, concerted TLS). Alternatively, if uncoupling occurs before TLS ssDNA will be exposed, permitting oligonucleotide binding and extension, generating a discontinuous leading-strand product (Fig. 2a, uncoupled TLS).

Fig. 2. On the fly TLS occurs uncoupled from CMG.

(a) Diagram illustrating the rationale for the oligonucleotide competition assay and the two possible outcomes. CMGE (CMG + Pol ε).

(b) Schematic of the possible replication products generated from the oligonucleotide competition assay shown in (c). For simplicity, reciprocal lagging-strand products are not shown.

(c) Time course oligonucleotide competition assay in the absence or presence of 5 nM Pol η.

(d) Schematic of the possible replication products generated from the oligonucleotide competition assay shown in (e). For simplicity, reciprocal lagging-strand products are not shown.

(e) Oligonucleotide competition assay using oligonucleotides mapping to various locations downstream of the CPD (as indicated).

Uncropped versions of Fig. 2 gels are available in Source Data Fig. 2.

In reactions containing a scrambled oligonucleotide, Pol η stimulated production of FL-lead (Fig. 2b and c, lanes 1-6). These products were abolished when an oligonucleotide mapping 21 nt downstream of the CPD was used instead (Fig. 2c, lanes 7-12). Here, the major leading-strand products comprised stall and an ~5.2 kb product, corresponding to oligonucleotide-mediated restart (Oligo restart) (Fig. 2b and c, lanes 7-12). This indicates at least 21 nt of ssDNA were exposed prior to CPD bypass. Moreover, restart products were apparent by 20 min (Fig. 2c, lanes 7 and 10), whereas Pol η-mediated FL-lead was not visible until 40 min (Fig. 2c, lanes 4 and 5), further indicating that uncoupling occurs before TLS. These observations are inconsistent with a concerted mechanism and instead suggest on the fly TLS functions ‘behind’ uncoupled CMG.

To understand the extent of uncoupling before TLS, we tested oligonucleotides mapping 265 nt and 1210 nt downstream of the CPD (Fig. 2d and e). Again, both oligonucleotides substantially reduced Pol η–dependent FL-lead production (Fig. 2e, lanes 5-8). However, a small amount of FL-lead was visible, suggesting some TLS-mediated recoupling had occurred before the oligonucleotide-binding site was unwound (Fig. 2e, lanes 6-8). In each reaction containing re-priming oligonucleotides and Pol η, the leading strand was extended across the CPD, presumably until it reached the 5' end of the oligonucleotide (Fig. 2e and Extended Data Fig. 2a). These observations indicated that extensive uncoupling had occurred before TLS. To confirm Pol η can operate uncoupled from CMG we performed a pulse-chase experiment. Pol η was either present during the pulse, added with the chase or 10 min later (Extended Data Fig. 2b). Pol η extended stalled leading strands even when added 10 min after the chase when extensive uncoupling had occurred. Indeed, uncoupled products, generated when uncoupled CMG reaches the end of the template, were extended after Pol η addition (Extended Data Fig. 2c, native, lanes 10-12). Although delayed addition of Pol η supported efficient TLS, FL-lead was less prominent compared to when Pol η was present in the pulse, indicating that recoupling of leading-strand synthesis to CMG was occurring less efficiently. These results demonstrate on the fly TLS operates behind CMG to restart uncoupled forks, rather than by preventing their formation.

Pol δ, but not Pol ε, inhibits lagging and leading-strand TLS

TLS Pols display poor fidelity compared to replicative Pols, potentially driving mutagenesis if not regulated⁷. We investigated how access of Pol η to stalled nascent strands is controlled in our experimental system, which might provide insights into how TLS Pol access to nascent strands is restricted when replicative Pols stall but TLS is not required. Reasoning that non-replisome associated replicative Pols might compete with Pol η for access, we examined the influence of Pol δ availability on lagging-strand TLS by performing a Pol δ titration with a fixed concentration of Pol η. Okazaki fragment maturation was supported across the Pol δ titration range (Extended Data Fig. 3a and b), however CPD bypass was highly sensitive to Pol δ concentration (Fig. 3a). At low concentrations a prominent bypass product and no stall product were observed (Fig. 3a, lanes 1-3). Higher Pol δ concentrations increased the stall product, with almost no bypass visible at 10 nM Pol δ (Fig. 3a lanes 4-6). The same result was observed in experiments lacking Fen1 and Ligase (Extended Data Fig. 3c).

Fig. 3. Pol δ, but not Pol ε, inhibits lagging and leading-strand TLS.

(a) Standard replication assay on the lagging-strand CPD template in the presence of 5nM Pol η, Fen1, and Ligase, and increasing concentrations of Pol δ.

(b) Standard replication assay on the leading-strand CPD template in the presence of 5 nM Pol η, with increasing concentrations of Pol ε or Pol δ.

(c) Mapping of the 3' end of the stalled nascent leading strand in the absence or presence of Pol η and Pol δ. Urea polyacrylamide gel samples were treated with SwaI to generate a stall product and resolved alongside a sequencing ladder. The region surrounding the stalled 3' end of the nascent leading strand is shown to the right of the gel with position of the stall and CPD shown in red.

Uncropped versions of Fig. 3 gels are available in Source Data Fig. 3.

Next, the ability of the principal leading-strand polymerase, Pol ε^1–3, to modulate leading-strand TLS was assessed. In contrast to Pol δ on the lagging strand, increasing concentrations of Pol ε had no appreciable influence on leading-strand TLS (Fig. 3b). We considered Pol δ might also target stalled leading strands upon release by Pol ε. Indeed, titration of Pol δ into reactions on the leading-strand CPD template revealed a dramatic concentration-dependent inhibition of CPD bypass but did not affect replication of an undamaged template (Fig. 3b, lanes 6-10, and Extended Data Fig. 3d). Increase of Pol δ above 2.5 nM caused accumulation of uncoupled forks (Fig. 3b, native lanes 6-10) and stalled leading strands (Fig. 3b, denaturing lanes 6-10), with FL-lead barely detectable above 10 nM Pol δ (Fig. 3b, lanes 9-10). Similar results were observed when additional Pol ε or Pol δ were added with Pol η to pre-stalled leading strands in pulse-chase experiments (Extended Data Fig. 3e and f).

To confirm Pol δ blocked CPD bypass, the 3' end of the stalled leading strand was mapped by SwaI digestion and resolution of the cleavage products on urea polyacrylamide gels. In the absence of Pol η leading strands stalled at the base immediately preceding the CPD (Fig. 3c, urea polyacrylamide, lanes 1 and 3). When Pol η was added in the absence of Pol δ no stall was detected (Fig. 3c, lane 2). Despite efficient bypass in this reaction, full-length products were underrepresented in the denaturing gel compared to previous reactions containing Pol δ (Fig. 3c, denaturing lane 2). Instead a smear of products was evident above the stall position, indicating incomplete leading-strand synthesis after bypass. When Pol δ was added to reactions containing Pol η, leading-strand synthesis again stalled at the base immediately preceding the CPD (Fig. 3c, lane 4).

These results demonstrate the principal lagging-strand Pol, Pol δ, acts as a ‘first responder’ to uncoupling by binding the nascent leading strand and inhibiting TLS. This suggests the nascent leading strand is released from CMG–Pol ε (CMGE) after stalling.

Pol δ recouples leading-strand synthesis to CMGE

Inefficient extension of bypassed leading strands in reactions lacking Pol δ indicated a requirement for Pol δ in leading-strand synthesis following bypass (Fig. 3c, lane 2). We reasoned it might recouple bypassed leading strands to CMGE, similar to its involvement in initiating leading-strand synthesis^35–37.

Synthesis by Pol ε must be coupled to CMG for maximum fork rates²⁹, therefore failure to recouple synthesis after bypass will delay the resolution of uncoupled forks.

To test this, we performed a Pol η titration in the absence or presence of Pol δ on the leading-strand CPD template. In the absence of Pol δ, Pol η promoted CPD bypass, however, most bypass products were not extended to FL-lead and instead resolved as a smear above stall (Fig. 4a denaturing, lanes 1-6). Slow uncoupled extension of the leading strand after bypass likely generated this smear, therefore termed an uncoupled extension product.

Fig. 4. Pol δ recouples leading-strand synthesis to CMG following TLS.

(a) Standard replication assay on the leading-strand CPD template in the absence or presence of Pol δ, with increasing concentrations of Pol η.

(b) Lane profiles of denaturing gel lanes 6 and 12 shown in (a). FL: full-length; UE: uncoupled extension; S: stall; RO: run-off; OF: Okazaki fragments.

(c) Reaction scheme for the pulse chase experiment shown in (d).

(d) Pulse chase experiment on the leading-strand CPD template in the absence of Pol δ, or with Pol δ (5 nM) added either at the start of the chase or 5 min into the chase. In all cases, 5 nM Pol η was added at the start of the chase.

(e-g) Two-dimensional gels of the 20 min time-points in (d).

Uncropped versions of Fig. 4 gels are available in Source Data Fig. 4.

Native gel analysis revealed a Pol η-dependent retardation of uncoupled forks (Fig. 4a, native, compare lanes 1 and 2), consistent with limited uncoupled extension downstream of the CPD. Pol δ again inhibited CPD bypass (Fig. 4a denaturing, lanes 7-9) with uncoupled forks the predominant reaction species at low Pol η concentrations (Fig. 4a, lanes 7-9). At higher Pol η concentrations, full-length products became more prominent and FL-lead was apparent in the denaturing gel (Fig. 4a, denaturing, lanes 10-12). At 16 nM Pol η a considerably less intense stall product was evident in the absence of Pol δ, but this did not correspond to an increase in FL-lead due to the accumulation of uncoupled extension products (Fig. 4a, denaturing compare lanes 6 and 12, and 4b). This suggests Pol δ plays a crucial role in extending leading strands after CPD bypass to restart rapid fork progression.

To further investigate recoupling a pulse-chase experiment was performed without Pol δ. Pol η addition was delayed to the chase, enabling CMG to advance further along the template before bypass, thereby making recoupling more challenging (Extended Data Fig. 4a). Extended Data Fig. 4b shows that, although all leading strands bypassed the CPD within 10 min of Pol η addition, few full-length products were generated. Uncoupled extension products were gradually elongated, the majority of which were associated with uncoupled forks (Extended Data Fig. 4c and d), consistent with slow leading-strand synthesis after CPD bypass. Uncoupled products migrated more slowly as a diffuse population in the native gel, further confirming CPD bypass and limited downstream extension had occurred.

Finally, staged pulse-chase experiments were conducted to test if Pol δ could rescue uncoupled leading strands when added after TLS. Pol δ was either omitted from the reaction, added with Pol η in the chase, or 5 min later (Fig. 4c). In the absence of Pol δ recoupling was defective (Fig. 4d, lanes 1-4). When Pol η and Pol δ were added concomitantly TLS was inhibited (Fig. 4d, lanes 5-8). However, when Pol δ was added after Pol η, stall products were extended to FL-lead (Fig. 4d, lanes 9-12) associated with full-length products (Fig. 4e-g). This illustrates Pol δ synthesises the leading strand after TLS and this is critical to recouple leading-strand synthesis to CMGE.

Monoubiquitination of PCNA stimulates on the fly TLS

Our findings that Pol δ controls access of Pol η to blocked leading strands (Fig. 3) and is crucial for recoupling after bypass (Fig. 4) indicate it might perform a general role in preventing access of TLS Pols when uncoupling occurs independently of DNA damage. Here, Pol δ availability must be sufficient to outcompete TLS Pols, however; this would also inhibit TLS upon damage-induced uncoupling. Therefore, a mechanism should exist to positively regulate TLS when uncoupled leading strands cannot be rescued by Pol δ alone.

We considered PCNA monoubiquitination by Rad6–Rad18 might promote on the fly TLS when Pol δ is inhibitory. PCNA monoubiquitination was reconstituted in the replication assay by addition of Rad6–Rad18, ubiquitin and the ubiquitin-activating enzyme Uba1 (Fig. 5a and Extended Data Fig. 1a). Omission of Mcm10, essential for CMG activation³⁰, abolished ubiquitination, demonstrating a dependency on DNA replication. Rad6–Rad18 conjugates ubiquitin to PCNA K164¹¹. Accordingly, no ubiquitination was observed when K164 was substituted for arginine (PCNA^K164R) (Fig. 5b and Extended Data Fig. 1a). Little difference in monoubiquitination was observed between an undamaged and CPD template in the absence of Okazaki fragment maturation (Extended Data Fig. 5a, lanes 1 and 3). However, yeast lacking Fen1 exhibit elevated damage-independent PCNA monoubiquitination³⁸, suggesting ubiquitination at unligated Okazaki fragments could mask a damage-specific effect. Indeed, addition of Fen1 and Ligase reduced PCNA monoubiquitination to barely detectable levels on both templates (Extended Data Fig. 5a). The small fraction of PCNA retained and ubiquitinated at the CPD is therefore likely beyond the detection limit of the assay. Nevertheless, these experiments demonstrate the ubiquitination machinery is active in replication assays, dependent on CMG activation, requires PCNA K164, and suggest specificity for PCNA retained at unligated 3ʹ ends. Finally, the ubiquitination machinery did not noticeably affect unperturbed replication (Extended Data Fig. 5b).

Fig. 5. PCNA monoubiquitination stimulates on the fly TLS.

(a) Western blot of PCNA from standard 60 min replication reactions on the leading-strand CPD template. Individual ubiquitination machinery components and MCM10 were omitted from reactions as indicated.

(b) Western blot of PCNA from standard time course reactions performed on the leading-strand CPD template with either wild-type PCNA or PCNA^K164R in the presence of ubiquitin, Uba1, and Rad6–Rad18.

(c) Standard replication reaction time course performed with wild-type PCNA in the absence or presence of Uba1. Reactions contained Pol η, Pol δ, ubiquitin, Rad6–Rad18, Fen1 and Ligase, in addition to standard replication proteins.

(d) Quantification of the percentage of bypass in the absence or presence of uba1 as performed in (c). Data are plotted as the means of three independent experiments with error bars representing the SEM.

(e) Standard replication reaction time course as performed in (c) but with PCNA^K164R.

(f) Quantification of the percentage of bypass in the absence or presence of uba1 as performed in (e). Data are plotted as the means of three independent experiments with error bars representing the SEM.

Uncropped versions of Fig. 5 gels are available in Source Data Fig. 5.

To assess if PCNA ubiquitination stimulated on the fly TLS, we performed a time course on the leading-strand CPD template with the ubiquitination machinery, Fen1, Ligase, and an inhibitory Pol δ concentration (Extended Data Fig. 5c). Omission of Uba1 reduced the amount of fully ligated material produced, suggesting a stimulation of bypass in the complete reaction (Extended Data Fig. 5d). To quantify TLS, reaction products were processed with BamHI and SwaI, and the bypass percentage calculated (Fig. 5c and d). Strikingly, bypass products were more prominent at later time points when Uba1 was present (Fig. 5c, lanes 3-5 and 8-10), with quantification revealing an ~2.5-fold increase by 45 min (Fig. 5d). This effect was abolished when Rad6–Rad18 was omitted (Extended Data Fig. 5e and f). However, despite the increased levels of bypass, Pol δ was still required for efficient recoupling (Extended Data Fig. 5g).

To confirm stimulation of TLS was specific to PCNA-K164 ubiquitination, we performed replication assays with PCNA^K164R, which supported unperturbed DNA replication and ubiquitination-independent TLS like PCNA (Extended Data Fig. 6a and b). Surprisingly, a higher concentration of both proteins was required to support TLS than leading-strand synthesis.

This was not due to competition by Pol δ at low PCNA concentrations because the same behaviour was observed in its absence (Extended Data Fig. 6c). This also confirms TLS is dependent on PCNA and suggests it is reloaded at the stalled leading strand before bypass. In contrast to PCNA, Fig. 5e and f show there was no Uba1-dependent stimulation of bypass with PCNA^K164R suggesting stimulation of TLS functions through ubiquitination of PCNA-K164.

Compared to PCNA, reactions performed with PCNA^K164R displayed higher background signal in urea polyacrylamide gels (Fig. 5c and e), indicating incomplete Okazaki fragment maturation. Indeed, comparison of fragment processing in reactions containing PCNA or PCNA^K164R revealed a defect with the mutant protein (Extended Data Fig. 6d), consistent with observations of a potential defect in Okazaki fragment maturation in PCNA^K164R mutant yeast³⁸.

Consequently, we also monitored ubiquitination-dependent on the fly TLS in the absence of Fen1 and Ligase. Uba1 again stimulated bypass only in reactions containing PCNA (Extended Data Fig. 7a-d). Collectively, these experiments reveal that PCNA monoubiquitination on K164 stimulates on the fly TLS by Pol η when Pol δ is inhibitory to bypass. This mechanism therefore promotes TLS when Pol δ is unable to recouple leading-strand synthesis alone, in this case due to the CPD.

PCNA monoubiquitination promotes lagging-strand TLS

Finally, we investigated if PCNA monoubiquitination stimulated gap-filling TLS on lagging strands. Experiments were conducted in the same manner as those performed on the leading-strand template using an inhibitory concentration of Pol δ (Extended Data Fig. 8a). In complete reactions ~2.5-fold more bypass was observed by 25 min compared to reactions lacking Uba1 (Fig. 6a, b and Extended Data Fig. 8b). No stimulation of TLS occurred with PCNA^K164R (Fig. 6c and 6d). These data illustrate that both on the fly and gap-filling TLS are stimulated by PCNA monoubiquitination, providing a mechanism for TLS Pols to outcompete Pol δ to permit continued DNA synthesis on damaged templates.

Fig. 6. PCNA monoubiquitination promotes lagging-strand TLS.

(a) Standard replication reaction time course performed with wild-type PCNA in the absence or presence of Uba1 on the lagging-strand CPD template. Reactions contained Pol η, Pol δ, ubiquitin, Rad6–Rad18, Fen1 and Ligase, in addition to standard replication proteins.

(b) Quantification of the percentage of bypass in the absence or presence of Uba1 as performed in (a). Data are plotted as the means of three independent experiments with error bars representing the SEM.

(c) Standard replication reaction time course as performed in (A) but with PCNA^K164R.

(d) Quantification of the percentage of bypass in the absence or presence of Uba1 as performed in (c). Data are plotted as the means of three independent experiments with error bars representing the SEM.

Uncropped versions of Fig. 6 gels are available in Source Data Fig. 6.

Discussion

We have reconstituted regulated TLS in the context of a eukaryotic replisome, demonstrating that Pol η facilitates on the fly bypass of a leading-strand CPD to restart uncoupled replication forks. On the lagging strand Pol η mediates TLS to permit gap filling. In both cases, lesion bypass is subject to negative and positive regulation by Pol δ and PCNA monoubiquitination, respectively. Pol δ is also crucial for efficient restart of uncoupled replication forks by recoupling the nascent leading strand to CMGE following TLS. These observations allow a more informed model of leading- and lagging-strand TLS to be proposed.

A model for leading and lagging-strand TLS

We propose the models displayed in Fig. 7 for leading and lagging-strand TLS during replication. Upon encountering a leading-strand CPD, synthesis stalls and uncoupling occurs. CMG progression and lagging-strand synthesis continue³², whilst Pol δ binds the stalled leading strand and inhibits access of Pol η (Fig. 7, leading strand i). Rad6–Rad18 are recruited to exposed RPA-coated ssDNA and facilitate monoubiquitination of PCNAM^39,40, promoting Pol η recruitment and TLS (Fig. 7, leading strand ii). Following bypass, Pol δ synthesises the nascent leading strand until it recouples to CMGE and canonical replication resumes (Fig. 7, leading strand iii). When a lagging-strand CPD is encountered, replication of both strands continues, generating a ssDNA gap and stalled Okazaki fragment (Fig. 7, lagging strand i). Pol δ inhibits recruitment of Pol η until Rad6–Rad18 bind the ssDNA gap and facilitate PCNA monoubiquitination (Fig. 7, lagging strand ii). Following TLS, Pol δ fills in the ssDNA gap before Okazaki fragment maturation (Fig. 7, lagging strand iii).

Fig. 7. Model of leading- and lagging-strand TLS by Pol η.

Implications for the default mode of leading-strand TLS

Our observations show leading and lagging-strand TLS are mechanistically similar, except that on the fly bypass requires recoupling, whereas lagging-strand gap filling is completed by Okazaki fragment maturation. Whether leading-strand TLS occurs on the fly, or by gap filling, may therefore be dictated by the efficiency of re-priming. Yeast Pol α is inefficient at leading-strand re-priming due to negative regulation by RPA^32,35. This suggests leading-strand TLS likely operates on the fly under normal conditions when DNA lesions are seldom encountered. However, high lesion density may deplete RPA and promote re-priming and gap-filling^32,41,42. Accordingly, UV-irradiation of yeast with doses predicted to impede most replication forks, generated ssDNA gaps in both nascent strands²⁶. Additional factors may affect the efficiency of leading-strand re-priming in vivo and therefore the balance between on the fly and gap filling TLS. Higher eukaryotes possess an additional primase, PrimPol, that re-primes leading-strand synthesis^43,44.

Consequently, gap-filling TLS on the leading strand may be favoured in higher eukaryotes.

Pol δ – the first responder to cessation of leading-strand synthesis

Pol δ is likely utilised for recoupling (Fig. 4) due to its faster rate of synthesis than the rate of uncoupled CMG progression^29,34, and its higher affinity than free Pol ε for PCNA⁴⁵. Our observation that Pol δ binds leading strands upon stalling, suggests it acts as a “first-responder” to transient uncoupling of leading-strand synthesis. Spontaneous dissociation of the flexible catalytic domain of CMG-bound Pol ε from the leading strand may deposit PCNA to enable mismatch repair and nucleosome assembly^46,47. Here, Pol δ likely recouples synthesis to CMGE and prevents recruitment of error prone TLS Pols, which is supported by its higher in vivo abundance⁴⁸. Likewise, damage-independent uncoupling factors, such as moderate nucleotide deprivation or repetitive template sequences, may be resolved by Pol δ-mediated recoupling. Intriguingly, work in DT40 cells suggests that Pol δ may support a subset of TLS events alone^49,50.

Monoubiquitination of PCNA is not strictly required for TLS

When Pol δ is unable to recouple leading strands or complete Okazaki fragment synthesis due to template lesions TLS must be promoted to enable bypass. PCNA monoubiquitination by Rad6–Rad18 promoted both leading and lagging-strand TLS in this scenario (Fig. 5 and 6), but was not strictly required. Bypass efficiency in its absence depended on the relative concentrations of Pol η and Pol δ, consistent with an increase in UV sensitivity when Pol η expression was reduced in PCNA^K164R cells²².

Our finding that PCNA monoubiquitination enables TLS Pols to outcompete Pol δ is supported by reports that TLS was less dependent on PCNA monoubiquitination when the Pol δ–PCNA interaction was compromised⁵¹. Additional levels of regulation including increased cellular and sub-cellular concentrations of TLS Pols and post-translation modifications may also promote TLS^6,21,52,53. Evidence also suggests Rev1 fulfils a non-catalytic role, independently of PCNA monoubiquitination, to coordinate TLS Pols⁵⁴.

The switch back to Pol δ

Following bypass Pol δ must actively re-engage PCNA to facilitate recoupling or fragment maturation. Our experiments did not include de-ubiquitinating enzymes or the PCNA unloader, Elg1^55,56. Nevertheless, no recoupling defect in the presence of PCNA monoubiquitination was evident. This contrasts with primer extension studies where pre-ubiquitinated PCNA was refractory to exchange from Pol η to Pol δ⁵⁷. It is possible that PCNA reloading following TLS permitted Pol δ recruitment in our system. However, Pol η acts distributively and rapidly dissociates from PCNA after bypass⁵⁸, potentially allowing Pol δ to actively re-engage PCNA. Deubiquitination may be critical here to prevent unwarranted recruitment of TLS Pols upon subsequent spontaneous uncoupling. The dynamics of PCNA ubiquitination-deubiquitination and their effect on DNA replication will be an interesting topic of further investigation.

The reconstitution of regulated TLS provides valuable insights into the mechanisms of lesion bypass and the restart of uncoupled replication forks generally. Moreover, this system offers a platform for future mechanistic studies of DNA damage tolerance and fork restart.

Online Methods

Experimental model and subject details

Proteins were purified from Saccharomyces cerevisiae strains (genotype: MATa ade2-1 ura3-1 his3-11,15 trp1-1 leu2-3,112 can1-100 bar1::Hyg pep4::KanMX) harbouring integrated constructs for overexpression of the protein of interest. Synthetic gene constructs used for integration were codon optimised for high-level protein expression^30,59. Strain and plasmid details are given in the Supplementary tables 1-3.

Proteins were also purified from Escherichia coli RosettaTM 2(DE3) cells (Novagen) (genotype: F– ompT hsdSB(rB– mB–) gal dcm (DE3) pRARE2 (CamR)) transformed with plasmids for overexpression of the desired protein.

Replication templates

The ZN3 ARS306 template used in this report is identical to that described previously³². Integration of a CPD containing oligonucleotide was performed as previously detailed³². Briefly, gapped ZN3 was generated by nicking at two Nt.BbvCI sites and heating to release a short oligonucleotide flanked by the sites, which was subsequently competed away. The gapped plasmid was purified on a Sepharose 4B (Sigma-Aldrich 4B200) gel filtration matrix, before a short oligonucleotide containing the CPD was annealed to the gap and ligated (see Supplementary table 1 for sequence).

Undamaged and CPD containing plasmids were subject to CsCl gradient purification and linearised with either AhdI, to generate a leading-strand CPD/undamaged template, or AleI, to generate a lagging-strand CPD/undamaged template, as described previously³². All experiments were performed on undamaged or CPD containing ZN3 plasmids linearised to generate a leading or lagging-strand template (as indicated in individual figures).

Yeast expression strain construction

Strains unique to this study (yJY31, yJY33, yTG1, yTG3, yTG4) were generated by transforming the parental strain yJF1⁶⁰ with the respective linearised expression vector following standard techniques (Supplementary table 2 and 3). For yJY31 (Fen1-3xFLAG expression), a C-terminal CBP tag was replaced with a 3xFLAG tag by transformation with a linear PCR product generated from pBP83 using primers JY112 and JY128. Codon optimised expression sequences^30,59 were synthesised by Invitrogen GeneArt Synthesis and cloned as described in Supplementary table 3 (for details of original expression vectors see^60,61). The Uba1 gene was PCR amplified from genomic DNA using primers to generate an N-terminal 2xFLAG tag (see Supplementary table 1 for primer sequences), before cloning into pRS303.

Additional information about the yeast expression strains and plasmids originating from this study can be found in Supplementary tables 2 and 3.

PCNA^K164R mutant cloning

PCNA^K164R was generated by site-directed mutagenesis of vJY19 using the primers listed in the Supplementary table 1. Sequencing analysis (Source BioScience) confirmed successful clones and the absence of additional unwanted mutations prior to transformation and protein expression.

Protein purification

The majority of proteins were purified as previously detailed^29,30,32. An overview of the purification strategies and tags used for these proteins can be found in Supplementary table 4. An in-depth description of the purification of those proteins unique to this study, Pol η, Rad6–Rad18, Uba1, Fen1, and Ligase, is given below. Note that PCNA^K164R was purified from E. coli in an identical manner to the wild-type protein²⁹. Pol η, Rad6–Rad18, Uba1, Fen1, and Ligase were all expressed in budding yeast. Cells were grown to 2-3 × 10⁷ cells per ml at 30°C in YP + 2% raffinose before induction of protein expression by supplementation with galactose to 2%. Cells were grown for a further 3 hours, harvested and resuspended in buffer (same as resuspension buffer described for each protein below), prior to dropwise freezing in liquid nitrogen. Lysis was performed manually using a pestle and mortar chilled with liquid nitrogen. The resulting powder was stored at -80°C until purification.

Pol η purification

Cell powder from 10 L of yTG1 culture was diluted 2:1 in Buffer A (40 mM Tris-HCl pH 7.5, 0.02% NP-40-S, 1 mM DTT, 10% glycerol) + 200 mM NaCl + 3 mM CaCl₂ + protease inhibitors. The cell lysate was clarified by centrifugation (235,000g, 45 min, 4°C) before addition of 2 ml Calmodulin-Sepharose 4B (GE Healthcare) to the soluble extract and incubation for 90 min at 4°C. Resin was collected in a disposable gravity flow column (Bio-Rad) and washed with Buffer A + 200 mM NaCl + 3 mM CaCl₂ + protease inhibitors. A second wash was performed with Buffer A + 100mM NaCl + 2 mM CaCl₂. Proteins were eluted in Buffer A + 100mM NaCl + 2 mM EDTA + 2 mM EGTA. Eluate was applied to a 1 ml HiTrap Heparin column equilibrated in Buffer A + 100 mM NaCl + 0.5 mM EDTA. The column was washed with the same buffer and proteins were eluted with a 30 ml gradient to 1 M NaCl. Peak fractions were pooled and dialysed against Dialysis Buffer (25 mM HEPES-KOH pH 7.5, 300 mM KOAc, 10% glycerol, 1 mM DTT, 0.02% NP-40-S) + 0.5 mM EDTA. Following dialysis, aliquots were snap frozen in liquid nitrogen and stored at -80°C until use.

Rad6–Rad18 purification

Cell powder from 15 L of yTG3 culture was resuspended in Buffer A + 200 mM NaCl + protease inhibitors. Following clarification by centrifugation (235,000g, 45 min, 4°C), 2 ml Anti-FLAG M2 affinity gel (Sigma-Aldrich) was added to the soluble extract. Lysate was incubated with the resin for 60 min at 4°C. Resin was collected and washed with Buffer A + 200 mM NaCl. The column was washed again with Buffer A + 200 mM NaCl + 5 mM Mg(OAc)₂ + 0.5 mM ATP. A final wash was performed with Buffer A + 100 mM NaCl. FLAG-tagged proteins were eluted by incubation in 5 ml Buffer A + 100 mM NaCl + 0.25 mg/ml 3×FLAG peptide for 10 min on ice. The eluate was loaded on to a 1 ml HiTrap Heparin column equilibrated in Buffer A + 100 mM NaCl. The column was washed and the protein eluted by a 30 ml gradient to 1 M NaCl. Peak fractions were pooled and diluted 1:2 in Buffer A before applying onto a 1 ml MonoQ column equilibrated in Buffer A + 100 mM NaCl. Again, proteins were eluted by a 30 ml gradient to 1 M NaCl. Peak fractions were pooled, concentrated to ~ 400 μl, and loaded onto an S200 column equilibrated in Dialysis Buffer. Fractions containing Rad6–Rad18 were pooled, snap frozen in liquid nitrogen, and stored at -80°C.

Uba1 purification

Cell powder from 10 L of yTG4 culture was resuspended in Buffer A + 200 mM NaCl + 0.5 mM EDTA + protease inhibitors. The lysate was clarified by centrifugation (235,000g, 45 min, 4°C) and 2 ml Anti-FLAG M2 affinity gel (Sigma-Aldrich) was added to the supernatant. Following incubation for 60 min at 4°C, the resin was collected and washed with Buffer A + 100 mM NaCl. Resin was incubated with Buffer A + 100 mM NaCl + 0.25 mg/ml 3×FLAG peptide for 10 min on ice to elute FLAG-tagged proteins. Peak fractions were pooled and applied to a 1 ml MonoQ column equilibrated in Buffer A + 100 mM NaCl. Following washing, proteins were eluted by a 30 ml gradient to 1 M NaCl. Peak fractions were pooled and dialysed against Dialysis Buffer before snap freezing and storage at -80°C.

Fen1 purification

Cell powder from 10 L of yJY31 culture was diluted 2:1 in Buffer A + 200 mM NaCl + 1 mM EDTA + protease inhibitors. Lysate was cleared by centrifugation (235,000g, 45 min, 4°C) and the supernatant supplemented with 1.5 ml Anti-FLAG M2 affinity gel (Sigma-Aldrich) before incubation for 60 min at 4°C. The resin was collected and washed with Buffer A + 200 mM NaCl + 1 mM EDTA + protease inhibitors. A second wash was performed with Buffer A + 100 mM NaCl and proteins were eluted with the same buffer + 0.25 mg/ml 3xFLAG peptide. Peak fractions were pooled and applied to a 1 ml HiTrap Heparin column equilibrated in Buffer A + 100 mM NaCl. The column was washed and proteins eluted by a 30 ml gradient to 1 M NaCl. Peak fractions were pooled and aliquots snap frozen in liquid nitrogen and stored at -80°C.

Ligase purification

Cell powder from 10 L of yJY33 culture was diluted 2:1 in Buffer A + 200 mM NaCl + protease inhibitors. Cell lysate was clarified by centrifugation (235,000g, 45 min, 4°C) and 1.5 ml Anti-FLAG M2 affinity gel (Sigma-Aldrich) was added to the soluble extract before incubation for 60 min at 4°C. Resin was collected and washed with Buffer A + 200 mM NaCl + protease inhibitors. Proteins were eluted following a 10 min incubation with 5 ml Buffer A + 200 mM NaCl + protease inhibitors + 0.25 mg/ml 3×FLAG peptide. The eluate was diluted to 100 mM NaCl by addition of Buffer A and loaded onto a 1 ml MonoQ column. The column was washed with Buffer A + 100 mM NaCl, before elution by a gradient to 700 mM NaCl. Peak fractions containing Ligase were pooled and dialysed against Dialysis Buffer before snap freezing and storage at -80°C.

Standard replication assays

Replication reactions were performed in broadly the same manner as those previously described³². MCM loading and phosphorylation were facilitated by incubating 5 nM linearised CPD-containing, or undamaged, ZN3 DNA template with 75 nM Cdt1/Mcm2-7, 45 nM Cdc6, 20 nM ORC, 50 nM DDK, 5 nM ATP, 25 mM HEPES-KOH (pH 7.6), 100 mM potassium glutamate, 10 mM Mg(OAc)₂, 0.02% NP-40-S, and 0.1 mg/ml BSA at 24°C for 10 min. The loading reaction was then supplemented with S-CDK to a final concentration of 150 nM and incubated for an additional 5 min. Subsequently, the loading reaction was diluted 4-fold with replication buffer to give final concentrations of 25 mM HEPES-KOH (pH 7.6), 250 mM potassium glutamate, 0.02% NP-40-S, 10 mM Mg(OAc)₂, 0.1 mg/ml BSA, 5 mM ATP, 200 μM CTP, 200 μM GTP, 200 μM UTP, 30 μM dATP, 30 μM dCTP, 30 μM dGTP, 30 μM dTTP, and 1 μCi [α-³²P]-dCTP. DNA replication was initiated by incubation at 30°C and addition of the following proteins (final concentrations given): 30 nM Dpb11, 210 nM GINS, 40 nM Cdc45, 10 nM Pol ε, 5 nM MCM10, 20 nM Ctf4, 150 nM RPA, 20 nM Csm3/Tof1, 10 nM Mrc1, 20 nM RFC, 20 nM PCNA or PCNA^K164R, 10 nM TopoI, 80 nM Pol α, 20 nM Sld3/7, and 20 nM Sld2. Note that the concentration of RPA was lowered to 60 nM for oligonucleotide-mediated replication restart experiments and 100 nM oligonucleotide was used. Reactions with Okazaki fragment processing additionally contained 5 nM Fen1 and 5 nM Ligase. Where present, Pol η (concentration stated in figure or figure legend) and Pol δ were added 4 min into the reaction. Pol δ was typically used at a final concentration of 2.5 nM unless otherwise stated in figures or figure legends.

Pulse chase experiments

Pulse chase reactions were performed in the same manner as standard reactions except that the concentration of dCTP was lowered to 5 μM for the pulse phase. 10 min into the reaction the concentration of dCTP was then increased to 250 μM for the chase phase.

PCNA monoubiquitination experiments

PCNA monoubiquitination reactions contained 200 nM Rad6–Rad18, 25 nM Uba1, and 250 nM ubiquitin, in addition to standard replication proteins. For all experiments, Rad6–Rad18 and ubiquitin were added at the start of the reaction. For minus/plus Uba1 experiments, Uba1 was added to the reaction after 4 min, along with Pol η and Pol δ. For experiments with Okazaki fragment maturation, Pol η and Pol δ were both used at a final concentration of 2.5 nM. For reactions on the leading-strand CPD template, in the presence of Okazaki fragment maturation, 1 μM ubiquitin was used.

Post-reaction sample processing

All reactions were quenched by addition of EDTA to 25 mM and deproteinised by proteinase K (8 U/ml, NEB) – SDS (0.1%) treatment at 37°C for 15 min.

Samples were then phenol-chloroform-isoamyl alcohol extracted (Sigma-Aldrich), before removal of unincorporated nucleotide using Illustra G-50 columns (GE Healthcare). Where indicated in individual figures, samples were digested with SmaI to reduce heterogeneity in product sizes prior to loading onto denaturing alkaline agarose gels, as previously described³². Analysis of samples by native, denaturing, and two-dimensional agarose gel electrophoresis was performed as detailed previously³².

Samples to be analysed by urea polyacrylamide gel electrophoresis were additionally cleaved by BamHI-HF and BlpI (NEB) in CutSmart buffer for 20 min at 37°C, and SwaI (NEB) in Buffer 3.1 for 20 min at 25°C. Note that BlpI was used as a control for SwaI cutting efficiency and was not used in the final analysis. Digests were stopped by addition of EDTA to 50 mM and subject to proteinase K-SDS treatment and phenol-chloroform extraction, as described above. Products were subsequently ethanol precipitated using standard techniques and resuspended in 10 mM Tris-HCl (pH 8), 1mM EDTA. Before loading onto urea polyacrylamide gels, samples were supplemented with an equal volume of loading dye (80% formamide, 0.05% SDS, 10 mM EDTA, 100 mM NaCl, 0.04% xylene cyanol, 0.04% bromophenol blue) and incubated at 95°C for 3 min. Products were then immediately resolved on a 40 cm × 20 cm 6% polyacrylamide (Bis-Acrylamide 19:1 – Fisher Scientific), 7 M Urea denaturing gel, run in 1× Tris-Borate-EDTA buffer (TBE) for 150 min at 40 W.

Generation of sequencing ladder

The sequencing ladder used to map the 3’ end of the stalled nascent leading-strand was generated using a USB Sequenase kit (Affymetrix) according to the manufacturers instructions. The ladder was synthesised by extension of a primer annealed to the undamaged ZN3 template, the 5’ end of which mapped to the SwaI cleavage site (for primer sequence see Supplementary table 1).

Gel processing

Urea polyacrylamide and native agarose gels were dried on to 3MM chromatography paper (GE Healthcare) immediately after running. Denaturing alkaline agarose gels were fixed by incubation for 2×15 min in 5% trichloroacetic acid solution at 4°C before drying. Gels were exposed on BAS-IP MS Storage Phosphor Screens (GE Healthcare) and visualised on a Typhoon phosphorimager (GE Healthcare) for use in quantification. For presentation, gels were additionally autoradiographed using Amersham Hyperfilm MP (GE Healthcare).

Western blotting

Monoubiquitinated PCNA from replication reactions was detected by western blotting using Anti-PCNA (Abcam ab70472; 1:2000 dilution) and HRP-conjugated Anti-mouse IgG (ThermoFisher A16078; 1:2000 dilution).

Quantification and statistical analysis

Quantification and data analysis were performed using ImageJ and Prism 7 software. For quantification of bypass percentage in PCNA monoubiquitination experiments, profiles of each lane of the urea polyacrylamide gel were generated in ImageJ. Straight lines were manually fit to the background baseline to allow the intensity of the bypass and stall products to be quantified. The percentage of bypass was calculated using the following equation: % bypass = (bypass ÷ (bypass + stall)) × 100 Data were plotted to show the mean and errors (SEM) from 3 independent experiments for each data set using Prism 7 software.

Extended Data

Extended Data Fig. 1. Pol η promotes TLS of lagging and leading-strand CPDs.

a, Purified Okazaki fragment processing and TLS proteins. b, Long exposure of the denaturing gel shown in Fig. 1b showing the diffuse ~1.7 kb stall product produced on the lagging-strand CPD template. c, Two-dimensional gel of the reaction performed in the absence of Pol η on the undamaged leading-strand template, shown in lane 1 of main text Fig. 1d. d, Two-dimensional gel of the reaction performed in the absence of Pol η on the leading-strand CPD template, shown in lane 7 of main text Fig. 1d. e, Two-dimensional gel of the reaction performed in the presence of 16 nM Pol η on the leading-strand CPD template, shown in lane 12 of main text Fig. 1d.

Extended Data Fig. 2. Leading-strand TLS occurs uncoupled from CMG.

a, Oligonucleotide competition assay performed in the absence or presence of Pol η. Reaction products were cleaved with SwaI to truncate stall products before resolution on a urea polyacrylamide gel. Addition of Pol η promotes extension of the stall product in the gap left behind from oligonucleotide-mediated recoupling. b, Reaction scheme for the pulse-chase experiment shown in (c). c, Pulse chase experiment on the leading-strand CPD template with 5 nM Pol η added 3 min into the pulse, at the start of the chase, or 10 min into the chase.

Extended Data Fig. 3. Pol δ, but not Pol ε, inhibits lagging and leading-strand TLS.

a, Pol δ titration into standard replication reactions on the undamaged leading-strand template containing Fen1 and Ligase. b, Denaturing gel of the reaction products from main text Fig. 3a. c, Standard replication reaction on the lagging-strand CPD template in the presence of 5 nM Pol η and increasing concentrations of Pol δ, as performed in Fig. 3a, but in the absence of Fen1 and Ligase. d, Pol δ titration into standard replication reactions on the undamaged leading-strand template. e, Reaction scheme for the pulse-chase experiment shown in (f). f, Pulse chase experiment on the leading-strand CPD template with 5 nM Pol η alone, or with 5 nM extra Pol ε, or Pol δ, added at the start of the chase.

Extended Data Fig. 4. Uncoupled replication forks display a recoupling defect in the absence of Pol δ.

a, Reaction scheme for the pulse-chase experiment shown in (b). b, Pulse chase experiment on the leading-strand CPD template in the absence of Pol δ and the absence or presence of 5 nM Pol η, added at the start of the chase. c, Two-dimensional gel of the 20 min time point shown in lane 6 of (b). d, Two-dimensional gel of the 20 min time point shown in lane 12 of (b).

Extended Data Fig. 5. PCNA monoubiquitination stimulates on the fly TLS.

a, Western blot of PCNA from standard 60 min replication reactions on the leading-strand CPD template, or undamaged equivalent, in the absence or presence of Fen1 and Ligase. All reactions contained ubiquitin, Uba1, and Rad6–Rad18 in addition to standard replication proteins. Denaturing gel of reaction products is shown below. b, Standard replication reaction time course on the undamaged template performed in the absence or presence of Rad6–Rad18, Uba1, and ubiquitin. c, Standard replication reaction time course on the leading-strand CPD template in the presence of 2.5 nM Pol η and 0.3 nM or 2.5 nM Pol δ. Samples were treated with BamHI and SwaI to generate bypass and stall products prior to resolution on the urea polyacrylamide gel. d, Denaturing gel of the reaction products from Fig. 5c. e, Replication reaction time course performed on the leading-strand CPD template in the absence or presence of Uba1 or Rad6–Rad18. Reactions contained 2.5 nM Pol η, 2.5 nM Pol δ, 1 μM ubiquitin, 5 nM Fen1, and 5 nM Ligase, in addition to standard replication proteins. Urea polyacrylamide gel samples were treated with BamHI and SwaI to generate quantifiable bypass and stall products. f, Quantification of the data in (e) showing the percentage of bypass in the absence or presence of uba1 or Rad6–18. g, Replication reaction time course performed on the leading-strand CPD template in the presence of PCNA monoubiquitination machinery (Rad6–Rad18, Uba1, and ubiquitin) and Pol η and the absence or presence of Pol δ.

Extended Data Fig. 6. Characterization of PCNA^K164R.

a, Standard replication reaction time course performed on the undamaged template with either wild type PCNA or PCNA^K164R. b, Standard replication reactions on the leading-strand CPD template containing increasing amounts of wild type PCNA or PCNA^K164R. Reactions contained 5 nM Pol η. c, Standard replication reactions on the leading-strand CPD template in the absence or presence of 2.5 nM Pol δ and increasing amounts of wild type PCNA. Reactions contained 5 nM Pol η. d, Standard replication reaction time course on the undamaged leading-strand template with either wild type PCNA or PCNA^K164R in the presence of Fen1 and Ligase.

Extended Data Fig. 7. PCNA monoubiquitination stimulates on the fly TLS in the absence of Fen1 and Ligase.

a, Standard replication reaction time course performed with wild-type PCNA in the absence or presence of Uba1. Reactions contained 10 nM Pol η, 10 nM Pol δ, 250 nM ubiquitin, and 200 nM Rad6–Rad18, in addition to standard replication proteins. Urea polyacrylamide gel samples were treated with BamHI and SwaI to generate quantifiable bypass and stall products. b, Quantification of the percentage of bypass in the absence or presence of Uba1 as performed in (b). Data are plotted as the means and s.e.m. of three independent experiments. c, Standard replication reaction time course performed with PCNA^K164R in the absence or presence of Uba1. Reactions contained 10 nM Pol η, 10 nM Pol δ, 250 nM ubiquitin, and 200 nM Rad6–Rad18, in addition to standard replication proteins. Urea polyacrylamide gel samples were treated with BamHI and SwaI to generate quantifiable bypass and stall products. d, Quantification of the percentage of bypass in the absence or presence of Uba1 as performed in (d). Data are plotted as the means and s.e.m. of three independent experiments.

Extended Data Fig. 8. PCNA monoubiquitination promotes lagging-strand TLS.

a, Standard replication reaction time course on the lagging-strand CPD template in the presence of 2.5 nM Pol η and 0.3125 nM or 2.5 nM Pol δ. Samples were treated with BamHI and SwaI to generate bypass and stall products prior to resolution on the urea polyacrylamide gel. b, Denaturing gel of the reaction products from Fig. 6a.

Reporting Summary statement.

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

Data availability statement

All data are provided in full in the results section and the Supplementary Information accompanying this paper. Unprocessed gels are available with the paper online.