An end for mismatch repair

DNA mismatch repair (MMR) is an important postreplicative repair mechanism that removes DNA polymerization errors and is responsible for increasing the fidelity of genome replication by orders of magnitude. Defects in MMR lead to the most common form of hereditary colon cancer and a variety of sporadic cancers (1–3). One of the main features of MMR is its ability to distinguish old and new DNA strands, but when a mismatch is found, how does MMR determine which sequence is the correct one? One of the first papers to describe MMR suggested that the MMR repair complex could have a special relationship with the replication apparatus (4), and that idea seems to be a central feature of strand discrimination in eukaryotes. In a study published in PNAS, Nick McElhinny et al. (5) provide evidence that the close relationship between MMR and replication leads to preferential repair of mismatches near initiation sites for DNA synthesis.

The discovery of the structure of DNA led to a simple prediction for its accurate replication: Each strand serves as a template for synthesizing a complementary strand. The antiparallel nature of the double helix and the finding that all DNA polymerases function only in a 5′ to 3′ direction complicated the picture, however. We now understand that, at any given point in a chromosome, both strands of DNA are replicated at the same time, with the leading strand being replicated continuously in the 5′ to 3′ direction and the lagging strand being replicated discontinuously as Okazaki fragments (Fig. 1). In Escherichia coli, genome duplication requires only two polymerases: a primase that initiates the synthesis of new fragments and a DNA polymerase that does the bulk of replication. Eukaryotes typically have three polymerases (6, 7): Pol α-primase, a complex that synthesizes the primers necessary for replication initiation, and the main replicative polymerases Pol δ and Pol ε. It was suggested as early as 1990 that Pol δ and Pol ε were responsible for replication of different strands (8); recent work from Kunkel and colleagues (9, 10) has demonstrated that the leading strand in yeast is synthesized predominantly by Pol ε and the lagging strand by Pol δ, although an alternative replication fork with Pol δ on the leading strand may exist in certain cases (6, 7). Many DNA polymerases have an associated 3′ to 5′ exonuclease “proofreading” activity that is activated if an incorrect base is added; this proofreading activity greatly increases the fidelity of replication. Both Pol δ and Pol ε are high-fidelity DNA polymerases with associated proofreading activity, but Pol α has no exonuclease activity, and consequently has a lower fidelity (7).

Fig. 1.

DNA synthesis and strand polarity at the replication fork. On the lagging strand, as the DNA helix is opened up, Pol α-primase first synthesizes an RNA fragment of about 10 nt (red) and then extends that with 20–30 nt of DNA (orange). Pol δ then extends the primer a length of 200–300 nt (green) until it reaches the already synthesized fragment downstream. Joining of the Okazaki fragments involves additional enzymes, such as FEN1 and DNA ligase. Synthesis of the leading strand is likely attributable to Pol ε (blue). Potential mismatches created by Pol α and Pol δ are indicated.

Even with proofreading, a number of replication errors remain uncorrected, and those errors provide a substrate for MMR. Central components of MMR are the protein complexes that recognize mismatches (i.e., MutS homodimer in E. coli, MutSα and MutSβ heterodimers in eukaryotes) and a complex that facilitates downstream processing steps (i.e., MutL homodimer in E. coli, primarily the MutLα heterodimer in eukaryotes). Once a mismatch is detected, a crucial issue is how the incorrect sequence is determined. In E. coli, newly replicated DNA is transiently unmethylated and a dedicated endonuclease that recognizes hemimethylated DNA, MutH, is activated on mismatch recognition. This results in cleavage of the newly replicated and unmethylated, and thus presumed incorrect, DNA strand. The nicked strand containing the mismatch is then removed by exonucleases, and the DNA is synthesized once again (1–3). In eukaryotes, DNA methylation is not used to distinguish old and new strands, and a central issue is how strand discrimination occurs. It has been known for some time that a DNA nick or end can serve as a strand discrimination signal, both in vitro and in vivo (1–3), but the source of the nicks and ends has remained elusive. A recent paper in PNAS from the Modrich laboratory (11) showed in vitro that, through association with the sliding clamp of the replication machinery, MutLα specifically introduces nicks on the newly replicated strand. A distinguishing feature of the lagging strand is the occurrence of the frequent ends that initiate each Okazaki fragment. The current paper (5) addresses whether these ends might also serve as a strand discrimination signal.

In this study (5), the authors measured forward mutation rates in the URA3 gene and sequenced large numbers of independent mutants. To study mismatch correction of errors created by different DNA polymerases, one has to know which DNA polymerase created the errors. Kunkel and colleagues (9, 10, 12) have accomplished this by isolating and characterizing DNA polymerase fidelity mutants. The increase in mutation rate for each of the mutants used here (six- to eightfold in the presence of MMR and much higher in its absence) is such that most mutations will be attributable to the mutated polymerase. With this method, for example, it is found that the distribution of Pol α errors within the URA3 gene is clearly nonrandom, leaving open the possibility that some regions are favored for primer initiation. The central question of this paper, however, is the efficiency of correction of Pol α errors compared with Pol δ errors. The analysis focuses on base pair substitution errors and reveals that Pol α errors are corrected by a factor of 5- to 12-fold more than Pol δ errors. There are several cases in which the same mismatch is created by either Pol α or by Pol δ, and Pol α errors are corrected by MMR with a higher efficiency in those cases as well.

Why should Pol α errors be corrected at a higher frequency than Pol δ errors? The authors suggest that the location of the Pol α errors near a DNA end favors their elimination. As illustrated in Fig. 1, a mismatch created by Pol α will always be closer to the 5′ end of DNA than a Pol δ error. Because MMR requires an end for excision, either a naturally occurring one or one generated by MutLα, Pol α errors should be in a favorable position for correction. The mismatch might also be removed in a reaction independent of excision; it is not known whether the sole exonuclease implicated in eukaryotic MMR (Exo1) can remove the RNA primer that initiates Okazaki fragments. If the mismatch has not been removed before the downstream Okazaki fragment is reached, MMR could stimulate a strand displacement reaction, as has been observed recently in cell extracts (13). Such a strand displacement model would also favor a Pol α-generated mismatch located near the 5′ end. A remarkable finding for the Pol α polymerase is that, even in the absence of MMR and unlike in vitro (12), it generates very few single base insertion or deletion (in/del) mutations, in striking contrast to Pol δ. It has recently been shown that Pol α errors can be subject to proofreading by the Pol δ exonuclease, and it might be that in/del mismatches in the short primer region are more likely to generate a distorted structure that would be recognized by Pol δ proofreading. Alternatively, such in/del mismatches could be less stable and more likely to be eliminated by strand displacement during joining of the Okazaki fragments. It should be noted that a corollary of the proposal that Pol α errors are corrected more efficiently because they are close to an end is that Pol δ errors are corrected less well because of a lack of a strand discrimination signal in some cases. In those events, does a Pol δ-generated mismatch fail to be corrected or is it possible that the mismatch could be corrected in the wrong direction?

The increased correction of Pol α errors by MMR is one more example of the fine-tuning of replication fidelity. Mismatches such as G-T that are easily extended by DNA polymerases and inefficiently proofread are recognized well by MMR, whereas those such as C-C that are not easily extended are not (2, 14). In/del

The 5′ ends of Okazaki fragments do, in fact, appear to be used for strand discrimination on the lagging strand.

mismatches, typically created by strand slippage in repeated sequences, are poorly recognized by proofreading but are a prime substrate for MMR. The current study estimates that 99.8% of single-base in/del mismatches are corrected by MMR, compared with only 97% of base-base mismatches (5). The significantly higher correction frequency of in/del mismatches is likely attributable to the fact that there are two complexes that recognize in/del mismatches (MutSα and MutSβ), whereas only MutSα plays a significant role in recognizing base-base mismatches. The biological significance of this differential correction is that in/del mutations are likely to create more severe phenotypes than base pair substitutions. Superimposed on top of the differential MMR efficiency of Pol α vs. Pol δ lagging-strand errors reported here, two different assays have indicated a greater activity for MutSα on the lagging strand than on the leading strand (15, 16). Although one explanation for the MutSα difference could be the greater density of DNA ends on the lagging strand, one of the assays additionally showed relatively more MutSβ activity on the leading strand than on the lagging strand (15). Determination of MMR correction with regard to Pol ε errors and individual measurements of MutSα and MutSβ correction activity, all of which are underway, should be especially revealing.

The recent papers in PNAS (5, 11) show the relationship of MMR with the replication apparatus to be critical in its strand discrimination activity. The paper by Pluciennik et al. (11) is particularly important in illustrating how strand discrimination can be achieved on the leading strand, and Nick McElhinny et al. (5) demonstrate that the 5′ ends of Okazaki fragments do, in fact, appear to be used for strand discrimination on the lagging strand. Although we are now much closer to understanding how eukaryotic MMR achieves strand discrimination when correcting replication errors, there remain additional complexities to be explored.