A site-specific DNA lesion causes a transient stalling of DNA replication
To reconstitute DNA lesion replication, we replicated a plasmid DNA that carried a site-specific lesion in NPE (A). The lesion chosen is an apurinic/apyrimidinic (AP) site, which is abundant (16
) and known to stall many purified DNA polymerases (17
). It is non-instructive, so error-free and error-prone replication products can be unambiguously distinguished. Normally, the AP lesion is rapidly repaired in cytosol, prior to the initiation of replication (B). Even depletion of AP endonuclease I (xAPE I) failed to provide significant protection of the AP lesion, most likely due to the presence of another AP endonuclease and/or other base repair pathways in the extract (data not shown). We thus protected the AP lesion with a dominant negative mutant of the human AP endonuclease I (dnAPE) that cannot cleave but still binds tightly to it (18
). When dnAPE was included in the reaction, the AP site was efficiently protected (>98% by this assay) from repair (B). As the AP endonuclease has no significant affinity for single-strand AP sites [(18
) and data not shown], the mutant protein would fall off the unwound DNA and not by itself pose a hindrance to DNA polymerases.
Figure 1. Establishment of the AP lesion replication system. (A) Experimental design for AP lesion replication. A plasmid DNA that carries a synthetic AP site (Δ) is first incubated with the dominant negative human APE (dnAPE) to protect the AP lesion and (more ...)
The AP DNA was replicated in NPE supplemented with either dnAPE or buffer (for the replication reactions in this study, dnAPE was included to protect the AP lesion unless otherwise indicated). In the absence of dnAPE (AP lesion repaired), the DNA was gradually replicated and converted to supercoiled plasmids (C). In the presence of dnAPE (AP lesion protected), the supercoiled replication product was still generated, but there was a transient accumulation of the relaxed form. For example, at the 20′ time point, the supercoiled form and the relaxed form were present at similar levels. By 60’, most of the relaxed form was converted into the supercoiled form. The effect was specific for the AP DNA as the non-AP DNA was not affected by dnAPE (D). (The slight excess of the relaxed products in the presence of dnAPE was most likely due to the unavoidable basal level of random AP lesions, formed either by spontaneous base loss or as the intermediates of base excision repair of damaged bases in the extract.)
These data suggested that the lesion-free strand was replicated normally and gave rise to the supercoiled product. The AP-carrying strand, in contrast, was temporarily stalled and DNA synthesis restarted downstream, by either the same fork or the opposing fork, forming a gap at or near the lesion (A). To test this hypothesis, we isolated the relaxed and supercoiled products of the 20’ time point from the agarose gel and digested the DNA with various restriction enzymes. As shown in B, the supercoiled product was completely digested by all of the enzymes, but the relaxed product was completely digested by only a subset of the enzymes. Many of the enzymes could not digest well the relaxed product even though they completely digested pET28a plasmid, which was used as the carrier for DNA purification and served as the internal control for restriction digestion. When the digestion pattern was plotted on the plasmid, the enzymes that failed to digest the relaxed product were found to have sites within a small region immediately downstream of the AP site [with the exception of ClaI (see subsequently)]. This observation strongly suggested that most of the relaxed product carried a gap between the EcoRI site immediately 5′ to the AP site and the PvuII site 257 nt downstream of the AP site.
Figure 2. Restriction enzyme mapping of the replication intermediate. (A) Probable products of the lesion-carrying strand and lesion-free strand after replication. (B) The relaxed and supercoiled products of the 20 min time point were gel-purified, digested with (more ...)
The stalling occurs on the lesion-carrying strand
The digestion by ClaI, which has a site 886 nt upstream of the lesion, did not conform to the above pattern. While ClaI completely digested the supercoiled product, it was very inefficient in digesting the relaxed product (B). An examination of the sequence revealed that this particular ClaI site overlaps with a GATC dam methylation site and methylation is known to block ClaI digestion. After one round of replication, the two daughter molecules would be hemi-methylated, but at different adenines within the ClaI site (A), and might therefore be differentially digested by ClaI. To test this hypothesis, we used a DNA polymerase to copy the two strands of the NdeI fragment that contains the ClaI site. The two hemi-methylated products were then digested by ClaI. As shown in B, the product copied from the lesion-free strand was digested, but the product from the lesion strand was not. This observation showed that the supercoiled replication product (ClaI sensitive) was exclusively derived from the lesion-free strand (which also suggested that the AP site was not repaired before replication). The gap, on the other hand, was present on the replication product derived from the AP strand (ClaI resistant).
Effect of ClaI hemi-methylation. (A) The predicted methylation pattern of the ClaI site in Δ:C after one round of replication. (B) ClaI digestion of the two hemi-methylated, ClaI-containing NdeI fragments copied from Δ:C.
The stalling occurs one nucleotide before the AP site
We next determined exactly where the AP lesion stalled replication. The strategy was to first add a tail (dC or dT) to the 3′ end of the stalled strand of the gel-purified relaxed DNA (20 min time point) with terminal deoxynucleotidetransferase (TdT) and then use a primer complementary to this tail and another primer further upstream to amplify the intervening region (A). The PCR product was cloned into a vector and introduced into E. coli by transformation. The plasmid DNA was isolated from the transformants and the inserts were sequenced. As shown in B, most of the clones from the dT tailing reaction (17/21) ended in GAGCT … T, and most of the clones from the dC tailing reaction (20/26) ended in GAGCTC … C. Combining the two sets of data, it became clear that the major stalling site was one nucleotide before the AP site, indicating that the replication of the lesion per se was kinetically slow. In addition, this experiment and the one above provided further evidence that replication stalling was caused by the lesion rather than the steric hindrance of the dominant negative mutant APE. A steric hindrance would be extremely unlikely to stall just one template strand and one nucleotide before the lesion.
Figure 4. Determination of the stalling sites of DNA replication. (A) Experimental strategy to clone the stalled intermediates. The arrows illustrate the primers used for PCR. The templates for tailing and PCR were the gel-purified relaxed DNA from the 20 min time (more ...)
Both error-free and error-prone mechanisms are used to replicate the AP lesion
The replication stalling was only temporary and the AP lesion was eventually replicated over, leading to the accumulation of completely replicated, supercoiled product. If the AP lesion was replicated by copying the information from the lesion-free sister chromatid (error-free mechanism), the correct nucleotide would be expected at the position opposite the lesion. In contrast, if the AP lesion was replicated by translesion DNA polymerases (error-prone mechanism), then incorrect nucleotides would be expected. Furthermore, depending on what translesion polymerases were recruited, all 4 nt might be used at random or some nucleotides might be used in preference. To distinguish among these possibilities, we purified the final supercoiled replication products (after 75 min of replication in NPE) from an agarose gel. The DNA was treated with DpnI (to digest any residual unreplicated, fully methylated DNA; the plasmid contains 18 DpnI sites) and ClaI (to digest the product from the lesion-free template strand) and then introduced into E. coli
by transformation. As a control, DpnI and ClaI were found to have efficiently digested the pET28a carrier DNA, as shown by both DNA agarose gel staining and transformation assay (A and B). The plasmids were isolated from the transformants and sequenced. (E. coli
could accurately repair the AP lesion. In a control experiment, 34 transformants from the AP DNA were examined and all were found to be correctly repaired.) In this experiment, among the 44 Δ:C replication products sequenced, 28 had a C, 13 an A, 1 a G and 1 a T inserted opposite the AP site. (C, middle column). A was an incorrect nucleotide, clearly the product of error-prone lesion replication. C was the correct nucleotide, suggesting that the AP lesion might also be replicated by an error-free mechanism, but the error-prone mechanism could not be ruled out because Rev1, a translesion DNA polymerase, is known to insert a C opposite an AP site (19
). To resolve this uncertainty, we performed a similar analysis on the replication products from a DNA that carried an AP lesion opposite a G (Δ:G). If a true error-free mechanism had been used, then more Gs (and correspondingly fewer Cs) would now be found opposite the AP site. This was indeed the case. In this experiment, among the 39 Δ:G replication products sequenced, 16 had a G, 14 an A, 8 a C and 1 a T inserted opposite the AP site (C, right column). We repeated these experiments and calculated the average ratios of each nucleotide inserted opposite the AP site on Δ:C and Δ:G replication products. As shown in D, the ratios were very different between Δ:C and Δ:G replication products and deviated dramatically from the expected ratios of random insertion. Together, these data strongly suggested that both an error-free mechanism (inserting C for Δ:C and G for Δ:G) and an error-prone mechanism (inserting A and C but rarely T and G for both substrates) were used to replicate the AP lesion.
Figure 5. Determination of the nucleotides opposite the AP site in the final replication products (after 75 min of incubation in NPE). (A) Restriction digestion of the gel-purified supercoiled final replication products (detected by 32P) and the control pET28a (more ...)
These data provided the first biochemical evidence for the existence of error-free lesion replication, but a mundane explanation was that the correct nucleotide was inserted on DNA whose AP lesion had been repaired before replication. While this seemed very unlikely as AP sites were efficiently protected, we nevertheless attempted to determine if the correct nucleotide was inserted on Δ:G replication products that still carried the AP lesion [95% of the AP strand replication products still carried the AP lesion (Supplementary Figure S2)]. The AP lesion in Δ:G was embedded within a KpnI site, and the nicking of the lesion by AP endonuclease rendered the DNA completely resistant to KpnI digestion (Supplementary Figure S3A). As illustrated in A, we digested the purified replication products with DpnI (to remove any un-replicated DNA; 4 of the 18 DpnI sites in the plasmid lie between ClaI and KpnI), ClaI (to remove the product of the lesion-free strand), APE (to nick the AP site) and finally KpnI (to remove all DNA with an intact KpnI site, including the putative pre-repaired DNA and their replication products). The digested DNA was then used as the template for PCR with two primers that bracketed the ClaI and KpnI sites. The correct nucleotide could only be recovered on PCR products amplified from the newly replicated strand of the DNA that still carried the AP lesion. As a control for digestion efficiency, the DNA purified from a replication reaction containing the normal plasmid pBS-Trx (used in AP DNA construction) did not generate any PCR product after digestion with all four enzymes (Supplementary Figure S3B). In contrast, the DNA purified from the Δ:G replication reaction generated the expected PCR product even after digestion with all four enzymes. This PCR product was subcloned into pUC19 and the DNA isolated from the transformants was analyzed by sequencing. As shown in B, 17 out 50 had the correct nucleotide G, and the remaining clones had mostly A and C but rarely T. In contrast, when the Δ:T DNA was used as the substrate, T was now frequently found opposite the AP lesion, but G became rare (C). Together, the results from these and the above experiments demonstrated that both error-prone and error-free mechanisms were used to replicate the AP lesion.
Figure 6. Analysis of the replication products that still carried the AP lesion (after 75 min of incubation in NPE). (A) Six potential types of DNA and their sensitivity to various enzymes. The lesion-carrying DNA would be nicked on the AP strand but intact on (more ...)