The importance of being DNA

Most organisms store their genetic information in genomes comprised of DNA, which is more resistant to strand cleavage than is RNA, which contains a reactive 2' hydroxyl on the ribose ring. To maintain this chemical identity, DNA polymerases effectively exclude ribonucleoside triphosphates (rNTPs) from being incorporated during DNA synthesis.¹ However, this rNTP exclusion is not absolute, and the concentrations of rNTPs in cells are much higher than are the concentrations of dNTPs,^2,3 implying that some rNTPs may be stably incorporated into DNA. Indeed, we have found that during DNA synthesis in vitro using dNTP and rNTP concentrations present in vivo, yeast replicative DNA polymerases incorporate rNTPs into DNA to the extent that they may be the most common non-canonical nucleotides introduced into eukaryotic genomes.⁴ More recently, we provided evidence that rNTPs are incorporated into DNA during replication in vivo by DNA polymerase ε (Pol ε).⁵ We also found that an RNase H2-dependent repair process normally removes the rNMPs from DNA, and that failure to perform this repair results in replication stress and deletions of 2–5 base pairs in short repetitive sequences. These rNMP-dependent deletions represent a “replication fidelity” issue, in this case involving sugar discrimination rather than the much more extensively studied discrimination against incorporation of incorrect bases. The two recent studies of rNMP incorporation into DNA raise several interesting questions (numbered in Fig. 1).

Figure 1. Implications of recent studies of rNMP incorporation during replication. Depicted is one model of the eukaryotic replication fork, with the leading strand replicated by DNA polymerase ε and the lagging strand replicated by DNA polymerases α and δ. Numbers 1–10 correspond to points in the text. Triangular symbols for DNA depicted for points 3, 4 and 7 represent unpaired template bases.

(1) Do the proofreading exonucleases of eukaryotic replicative DNA polymerases δ and ε, which are implicated in lagging and leading strand replication, respectively,⁶ excise rNMPs incorporated into the nascent strand during ongoing replication? If so, then defects in proofreading could increase rNMP-dependent genome instability. (2) What proteins operate in conjunction with 3-subunit RNase H2 to remove rNMPs incorporated by DNA polymerases? Current candidates include FEN1 and PCNA,⁷ but other proteins likely participate. It is almost axiomatic in the DNA repair field that multiple, partially redundant repair pathways can remove the same lesion from DNA, especially if the lesion is abundant (e.g., abasic sites, 8-oxo-guanine).⁸ (3) Thus, in addition to RNase H2-dependent repair, do other pathways also remove rNMPs incorporated by DNA polymerases? If so, this may partly explain the observation that deletion of RNase H2 does not strongly inhibit growth, even in yeast strains encoding a mutant Pol ε derivative that has increased rNMP incorporation activity.⁵

How are unrepaired rNMPs converted into short deletions? Since the deletions predominantly occur in repetitive sequences, the mechanism(s) likely involves misaligned DNA strands stabilized by correct base pairing. (4) Such misalignments could arise during the next round of replication of templates containing unrepaired rNMPs, perhaps because replicative polymerases have some difficulty incorporating dNTPs when bypassing an rNMP in a DNA template.⁴ Relevant here may be the fact that deletion rates in Pol ε mutant strains lacking RNase H2 depend on the orientation of the mutational reporter gene relative to the nearest origin of replication.⁵ This orientation bias in mutagenesis is consistent with a model wherein rNMP incorporation occurs during leading strand replication by Pol ε.⁶ (5) It remains to be determined the extent to which rNMPs are incorporated by Pol α and Pol δ during lagging strand replication in cells, and if Pol γ incorporates rNMPs during replication or repair of the mitochondrial DNA genome. (6) If strand slippage with rNMP-containing templates occurs during replication, the resulting deletion loop mismatches may be subject to DNA mismatch repair (e.g., dependent on MutSβ),⁹ an unexplored possibility currently under investigation. If the deletion rate is independent of mismatch repair, then perhaps mismatch repair simply does not repair rNMP-containing mismatches, an issue that also has yet to be explored. Alternatively, the mutagenic DNA transaction may occur outside the context of normal DNA replication. (7) Such a process could involve synthesis by polymerases specialized for translesion synthesis (TLS). The ability of several of the highly specialized DNA polymerases to incorporate rNMPs into DNA, or to bypass rNMPs in DNA templates, remains to be determined. (8) Mutations could also result from aberrant repair of single-strand or double-strand breaks generated by spontaneous hydrolysis or processing of unrepaired rNMPs, a hypothesis that has yet to be tested. Breaks resulting from unrepaired rNMPs in DNA offer one possible explanation for the slow S phase progression of cells lacking RNase H2.⁵

In our initial study,⁴ we demonstrated that the probability of rNMP incorporation in vitro varies over a wide range, depending on the identity of the DNA polymerase and the base (whether A, T/U, G or C), the local sequence context and the dNTP:rNTP ratio. (5) It follows that circumstances that alter dNTP:rNTP ratios in vivo should alter the probability of rNTP incorporation. Such circumstances include environmental stress and mutations in genes that regulate dNTP pools, such as the genes encoding ribonucleotide reductase. In fact, mutations in the gene encoding the large subunit of ribonucleotide reductase have recently been identified that alter the concentrations of specific dNTPs,¹⁰ with rNTP concentrations being largely unaffected. Such mutants offer the potential to regulate the types, amounts and sites of rNMP incorporation into the genome, in order to determine if there is any biological significance to variations in rNMP incorporation probability.⁴ This idea relates not only to replication, but also to DNA repair, perhaps most especially to non-homologous end joining of double strand DNA breaks during the G₁ phase of the cell cycle or to base and/or nucleotide excision repair in quiescent cells. This is because, unlike proliferating cells that contain higher dNTP concentrations to support replication, cells that are non-cycling or have not yet entered S phase are likely to have substantially lower dNTP:rNTP ratios.

With respect to preventing and editing base substitution, insertion and deletion mismatches, Pol ε is one of the most accurate polymerases known. Although discrimination against these errors need not necessarily be related to sugar discrimination, we were nonetheless surprised to see that wild-type Pol ε, i.e., the proofreading-proficient, four-subunit holoenzyme, incorporates one rNMP per 1,250 dNMPs in vitro when using physiologically relevant dNTP and rNTP concentrations.

(9) Even though these rNMPs are efficiently removed by RNase H2-dependent repair in vivo,⁵ this high level of initial rNMP incorporation suggests that the transient presence of rNMPs in the nascent leading strand may have signaling functions, perhaps promoted by changes in DNA helical parameters resulting from an rNMP in DNA. Possible functions include signaling for (1) cell cycle checkpoint control, (2) strand discrimination during mismatch repair, perhaps especially for leading strand replication errors, (3) loading (or delaying loading) of nucleosomes behind the replication fork, (4) chromatin remodeling that alters gene expression and/or (5) other DNA transactions, such as mating type switching or meiotic recombination (reviewed in reference 4).

(10) Loss of RNase H2-dependent repair of rNMPs incorporated during replication also has potential human health implications. Mutations in the genes encoding each of the three subunits of human RNase H2,⁷ are associated with Acardi Goutiéres Syndrome (AGS), a rare and severe autoimmune disease. At least hypothetically, the cellular consequences of loss of RNase H2-dependent repair of ribonucleotides incorporated into DNA by polymerases could be a contributing factor to AGS, or perhaps less severe instances of autoimmunity. Another possibility is implied by the microsatellite instability (MSI) used as a biomarker for tumors with mutations in genes that inactivate mismatch repair. The 2–5 base pair deletions in repetitive DNA sequences resulting from deletion of RNase H2 in yeast is a special type of MSI, suggesting that genes encoding proteins involved in RNase H2-dependent repair could be tumor suppressor genes, perhaps characterized by what has been called “MSI-low” in the cancer research field.¹¹

Nick McElhinny SA, Kumar D, Clark AB, Watt DL, Watts BE, Lundström EB, et al. Genome instability due to ribonucleotide incorporation into DNA. Nat Chem Biol. 2010;6:774–81. doi: 10.1038/nchembio.424.