Ribonucleotide Incorporation by Eukaryotic B-family Replicases and its Implications for Genome Stability

Abstract

Our current view of how DNA-based genomes are efficiently and accurately replicated continues to evolve as new details emerge on the presence of ribonucleotides in DNA. Ribonucleotides are incorporated during eukaryotic DNA replication at rates that make them the most common non-canonical nucleotide placed into the nuclear genome, they are efficiently repaired, and their removal impacts genome integrity. This review focuses on three aspects of this subject, the incorporation of ribonucleotides into the eukaryotic nuclear genome during replication by B-family DNA replicases, how these ribonucleotides are removed, and the consequences of their presence or removal on genome stability and disease.

1. INTRODUCTION

In the early history of life on earth, genetic information was thought to be encoded in RNA (1). Now, the central dogma of modern molecular biology states that genetic information flows from DNA to RNA to protein. In reviewing this dogma (2), Crick noted the well-known loop depicted in Figure 1A, which involves the transfer of genetic information from one DNA molecule to another via DNA replication. To the extent that this loop avoids use of ribonucleoside triphosphates (rNTPs) as precursors for DNA replication, DNA-based organisms have an advantage over RNA-based organisms because DNA is a more stable storage medium for genetic information than is RNA. This is because ribonucleotides contain a reactive 2′-hydroxyl on the ribose ring that sensitizes the DNA sugar-phosphate backbone to hydrolysis (Figure 1B, (3)). In addition, ribonucleotides in DNA alter nucleic acid geometry (e.g., see (4–7) and references therein), thereby potentially influencing its information content and potentially altering cellular DNA transactions. For these reasons, a great deal of research has recently been conducted on ribonucleotide incorporation into DNA. This leads to the subject of this review, which considers ribonucleotide incorporation into DNA, how these ribonucleotides are processed and the consequences of these processes. We address this subject with a strong focus on recent studies of ribonucleotides incorporated into DNA during eukaryotic nuclear DNA replication by the B-family of DNA polymerases, DNA polymerases α, δ, ε and ζ. Readers are also encouraged to consult other recent reviews that consider the causes and consequences of ribonucleotides in DNA, including ribonucleotides incorporated by other DNA polymerases and their effects on various cellular transactions, including translesion DNA synthesis, DNA repair and transcription (8–14).

Figure 1. Ribonucleotide incorporation into DNA.

A. The central dogma of molecular biology states that genetic information flows from DNA to RNA to protein. DNA replication is depicted by the essential loop, and it is during this process that ribonucleotides (red R) can be incorporated into DNA during synthesis.

B. The 2′-oxygen on the ribose ring sensitizes the DNA backbone to hydrolysis under alkaline conditions. This results in cleavage and the production of 5′-hydroxl and 2′−3′-cyclic phosphate DNA ends.

C. The structural basis for ribonucleotide exclusion for the bacteriophage RB69 B-family replicase. As shown, when the incoming nucleotide is a dNTP, a steric clash is not formed between the aromatic sidechain of tyrosine 416 (purple) and the incoming C2′ position on the sugar. In contrast, when the incoming ribonucleotide is a rNTP, a steric clash is created between the sidechain of Y416 and the 2′-OH on the ribose moiety.

D. An amino acid alignment of residues in Motif A from the B-family replicases from bacteriophage RB69 and the S. cerevisiae Pols α, δ, ε and ζ. The strictly conserved steric gate tyrosine (Y) is shown in red.

2. RIBONUCLEOTIDE INCORPORATION AND PROOFREADING

2.1. Structural biology and biochemistry of ribonucleotide incorporation into DNA

That rNTPs can be incorporated into DNA has been appreciated since shortly after the structure of DNA was described, when Arthur Kornberg and his colleagues first purified E. coli DNA polymerase I (Pol I) and described its properties (15), including its ability to incorporate rNTPs. Subsequent structural studies demonstrated that when doing so, Pol I discriminates against rNTP incorporation though interactions of amino acid side chains within its polymerase active site that strongly limit rNTP incorporation, but do so incompletely (16). Particularly relevant is tyrosine 762 at the polymerase active site, whose extracyclic oxygen atom interacts with the 2′-oxygen of an incoming rNTP to prevent its incorporation. This “steric exclusion” reduces rNTP incorporation by 10⁵ to 10⁶-fold, such that Pol I is primarily considered to be a DNA polymerase. An equivalent tyrosine that acts to exclude a ribonucleotide during incorporation is a common feature of replicases, including the bacteriophage RB69 polymerase (Figure 1C).

In addition to this steric gate tyrosine residue, there are adjacent amino acids that modulate its position and are conserved among many different DNA polymerases (Figure 1D, (17)). This includes the eukaryotic B-family DNA polymerases (Pols) α, δ, ε and ζ. Their role is to replicate the eukaryotic nuclear DNA genome (see below). Like E. coli Pol I, these four eukaryotic nuclear DNA replicases effectively discriminate against rNTP incorporation in vitro, such that they too are primarily DNA polymerases. Nonetheless, rNTP concentrations are much higher than dNTP concentrations in cells (18, 19), and rNTPs compete with dNTPs for incorporation by DNA polymerases. As a result, at the concentrations of the eight NTPs measured in budding yeast cells (18), rNTPs are incorporated into DNA in vitro at rates that can easily be measured. For example, at physiological concentrations of dNTPs and rNTPs measured in Saccharomyces cerevisiae (Table 1), yeast Pols α, δ and ε incorporate one rNTP for every 650, 5,000 and 1250 dNTPs, respectively (Table 2, (18)). Similar results have been obtained with B-family polymerases in higher eukaryotes, including human (20, 21). In addition, amino acid substitutions in the steric gate tyrosine or in adjacent amino acids further reduce discrimination against ribonucleotides in the four eukaryotic B-family polymerases (Table 2), making these variant polymerases ideal tools for studying replication enzymology in vivo (see below).

Table 1.

Nucleotide pools in cycling S. cerevisiae cells

dNTP	Concentration (μM)	rNTP	Concentration (μM)	rNTP fold excess
dA	16	rA	3,000	190
dC	14	rC	500	36
dG	12	rG	700	58
dT	30	rU	1,250	57

Table 2:

Ribonucleotide discrimination factors for various yeast DNA polymerases

Yeast B-family Polymerase	Ribonucleotide incorporation (rNTP per dNTPs)	Reference
Pol α - wt	1/650	(18)
Pol α - L868M	1/40	(73)
Pol δ - wt	1/5,000	(18)
Pol δ - L612M	1/300	(73)
Pol ε - wt	1/1,250	(18)
Pol ε – M644G	1/91	(51)
Pol ε – M644L	1/3,000	(51)
Pol ζ - wt	1/200–300	(101)

Why such differences in the rates of ribonucleotide incorporation by the B-family members? While more work deserves to be performed in the future to understand these differences, one possibility that may contribute is the effect that ribonucleotides have on nucleic acid conformation. For example, there is evidence to suggest that ribonucleotides may also be incorporated in order to provide a signaling function through their impact on DNA structure. Oligonucleotides containing ribonucleotides in various positions in otherwise normal DNA adopt an A-type conformation (4). This conformation differs somewhat from the B-type conformation of DNA observed in the crystal structures of many DNA polymerases, which may account for some of the discrimination against ribonucleotide incorporation. Structural studies have demonstrated other effects of ribonucleotides in duplex DNA on helix parameters, including sugar pucker when the sugar adopts a C3′-endo or closely related conformation (5–7, 22). Local structural effects of the presence of ribonucleotides in DNA have been examined through measurements of DNA elasticity using Atomic Force Microscopy (AFM), molecular dynamics simulations and nuclear magnetic resonance spectroscopy (23, 24). A single ribonucleotide in a nucleosomal template causes local distortions in duplex unwinding and impacts the interaction of the ribonucleotides and surrounding DNA nucleotides with histones (25). These alterations may have important effects on diverse processes that rely heavily on proper DNA helix geometry, including replication, repair, transcription or cellular signaling processes.

It also remains to be determined what accounts for the differences among DNA polymerases in the amount of each of the four rNTPs incorporated. As but one example, initial experiments suggest that purified yeast Pol α incorporates more ATP than CTP, while the opposite is true for Pol ε (18). Studies performed in the future may reveal whether such differences are relevant to the positive or negative roles of newly incorporated ribonucleotides by the four B-family polymerases (see below).

Although this review is focused on the main source of ribonucleotides in DNA through their incorporation by the B-family replicases, other sources do exist. These include genomic ribonucleotides that are the result of incomplete Okazaki fragment maturation (OFM) (26), or DNA modified by oxidative DNA damage to form ribonucleotide derivatives (27, 28). Still another parameter to be considered, but is yet to be extensively investigated, is bypass of template ribonucleotides by the B-family polymerases. We do know that bypass occurs with varying efficiency and decreases as the number of template ribonucleotides is increased from one to four (21, 29–31). The exception is Pol ζ, which efficiently synthesizes DNA using templates containing up to four consecutive ribonucleotides (31).

2.2. Nucleotide concentrations

Discrimination against ribonucleotide incorporation by the replicases is strongly influenced by the relative concentrations of ribonucleotide and deoxyribonucleotides within a cell. This varies during the cell cycle, with deoxyribonucleotide concentrations transiently increasing during S phase compared to G₂, M, and G₁ phases (32), thereby providing a more favorable environment for deoxyribonucleotide incorporation into DNA by the replicases. There have been a variety of recent advances in measurement of nucleotide concentrations within a cell, including HPLC and mass spectrometry (33, 34). Ribonucleotides are greatly in excess of deoxyribonucleotides in both yeast (Table 1; (18)) and human cells (19), and tight regulation of dNTP pool concentrations is required to minimize replication errors and maintain proper growth in budding yeast (35–37). Nuclear nucleotide concentrations vary greatly across species, tissue types, cell cycle stages, and in normal versus cancer cells. For example, rNTP:dNTP ratios are variable in mouse tissues, with a lower ratio in dividing embryonic tissues and a higher ratio in differentiated skeletal cells (38). Similar results are seen in human cells, with a lower rNTP:dNTP ratio in logarithmically growing Balb/3T3 fibroblasts and a higher ratio in G₁-synchronized fibroblasts (33). Thus, it is generally true that a higher concentration of rNTPs and a much lower dNTP concentration in non-cycling cells (39) may drive incorporation during DNA repair reactions as compared to dividing cells, in which replication-incorporated ribonucleotides could be expected to be the major source of ribonucleotides in the nuclear genome. Nuclear nucleotide pool imbalances have been observed in cancer cells and contribute to replication infidelity (40). Finally, nucleotide pools are a critical determinant of genome stability in mitochondria, where they govern ribonucleotide incorporation (41, 42).

2.3. Proofreading of newly incorporated ribonucleotides

Evidence suggests that once a rNMP is incorporated, the efficiency of extension of a DNA primer containing a newly incorporated rNMP is reduced. This property could provide an opportunity for proofreading by an exonuclease. Pol α and Pol ζ are naturally devoid of 3′-exonuclease activity and therefore they themselves cannot exonucleolytically proofread newly incorporated ribonucleotides. This may partly explain why they incorporate more ribonucleotides than do Pols δ and ε. In contrast, Pol δ and Pol ε both have associated 3′-exonuclease activities in their catalytic subunits. This activity could remove ribonucleotides if they fray and become single-stranded to become substrates for exonucleolytic removal by Pol ε and/or Pol δ. An initial study of yeast and human Pol δ showed little if any proofreading of ribonucleotides during DNA synthesis in vitro (20). However, small differences in loss of 2–5 base pairs in repetitive DNA sequences observed in ribonucleotide excision repair (RER)-deficient yeast strains lacking the exonuclease activity of Pol ε (30) suggest that proofreading of ribonucleotides incorporated during leading-strand DNA replication may occur, albeit much less efficiently than for proofreading of single base mismatches that result in base substitution errors (30). Less efficient proofreading of ribonucleotides than mismatches is perhaps expected, since a ribonucleotide at the primer terminus contains a 2′-oxygen on the sugar rather than a mismatch that contains aberrant base-base hydrogen bonding. Nonetheless, theoretically and just as occurs for single-base mismatches, proofreading of ribonucleotides could occur intrinsically, i.e., without intervening dissociation as the mistake transits from the polymerase active site, which has evolved to bind to correct base pair, to the exonuclease active site, which has evolved to bind single strand DNA (see (43) and references therein). Alternatively, proofreading could occur extrinsically (44–47), when a newly incorporated mistake results in enzyme dissociation, allowing a new polymerase (like Pol δ) to bind the mismatch directly in the exonuclease active site and remove it. This suggests that a more complete understanding of potential proofreading of ribonucleotides incorporated during DNA replication deserves further investigation in the future.

3. RIBONUCLEOTIDE REMOVAL

3.1. Ribonucleotide Excision Repair

The major pathway by which single ribonucleotides are excised from DNA is Ribonucleotide Excision Repair (RER). The first step of RER involves incision on the 5′ side of the ribonucleotide by RNase H2, a conserved heterotrimeric enzyme able to recognize and cleave a single ribonucleotide in a DNA substrate as well as longer RNA/DNA hybrids (48). Following the demonstration that purified RNase H2 could cleave double-stranded oligonucleotide probes containing a single ribonucleotide (49), Rydberg and Game subsequently showed that RNase H2 and FEN1 are required for removal of a single ribonucleotide-containing DNA substrate in cell extracts (50). This was followed by a study in 2010 (51) demonstrating that ribonucleotides incorporated into the nuclear genome by yeast Pol ε are efficiently removed by a DNA repair process requiring RNase H2. Using the physiological concentrations of dNTPs and rNTPs measured in yeast (18), Peter Burgers and colleagues then reconstituted RER using purified yeast proteins. They demonstrated that following RNase H2 incision on the 5′ side of a ribonucleotide, the subsequent steps of strand-displacement synthesis and ligation are very similar to those that occur during Okazaki fragment maturation (Figure 2A) (52). Accordingly, they coined the term Ribonucleotide Excision Repair (RER) for this process. It has been demonstrated that both Pol δ and Pol ε are capable of strand displacement synthesis during RER in vitro (52, 53). Whether or not the ribonucleotide incorporation propensities of polymerases, or the fidelity of downstream events such as ligation (54), are altered during strand displacement synthesis as compared to the bulk of nuclear DNA synthesis remains to be investigated.

Figure 2. Ribonucleotide removal and biological consequences associated with these processes.

A. Ribonucleotide excision repair (RER) is initiated by RNase H2 cleavage on the 5′ side of a single ribonucleotide in duplex DNA. Strand displacement synthesis by Pol δ generates a flap that is processed by the Fen1 nuclease to release the ribonucleotide-containing segment of DNA. The resulting nick can then be sealed by DNA ligase 1.

B. In the absence of RER, Top1 is able to cleave on the 3′ side of a ribonucleotide to generate a nick that can be reversed during Top1-mediated ligation. If not resealed, 2′−3′-cyclization occurs (shown as a red triangle). Top1 cleavage in DNA 2–5 bp upstream (5′) of the first incision generates a small gap. If this cleavage occurs in non-repetitive DNA, then error-free DNA repair involving Top1 proteolysis, DNA end processing, DNA synthesis and ligation by DNA ligase 1 can occur. When the ribonucleotide is located in repetitive DNA, then mutagenic repair and loss of a repeat unit can occur due to DNA slippage, re-alignment and Top1-ligation across the gap.

C. In addition to 2–5 bp deletion mutagenesis, several negative biological consequences are associated with Top1-processing of ribonucleotides in DNA. These include genome instability in the form of replication stress, checkpoint activation, DNA breaks and chromosomal rearrangements, some of which may result from premature release of toxic DNA repair intermediates that arise following Top1 incision at unrepaired ribonucleotides.

D. Ribonucleotides provide a provide a positive signaling function during DNA MMR by acting as a strand-discrimination signal for repair of errors made during leading strand synthesis. A DNA mismatch is first recognized and bound by the MutSα heterodimer. RNase H2 cleavage at a ribonucleotide incorporated by Pol ε into the same strand generates a nick that can be used as an entry point for MutLα, which can then translocate along the DNA and incise close to the mismatch to facilitate excision of a DNA patch containing the mismatch, DNA resynthesis and ligation.

Budding yeast RNase H2 is comprised of three subunits encoded by the RNH201, 202 and 203 genes (RNASEH2A, H2B and H2C in humans), all of which are required for catalytic activity (reviewed in (48, 55).) Although dispensable for viability in yeast, RNase H2 is essential in mice, with loss of either the H2A or H2C subunits resulting in p53-dependent embryonic lethality (56–58). RNase H2 is critical for removal of both single ribonucleotides incorporated by the DNA polymerases during replication, as well as other RNA-containing structures such as R-loops (reviewed in (59). By expressing the RNase H2-Ribonucelotide Excision-Defective (RED) mutant in mice, it was demonstrated that it is the failure to remove single ribonucleotides that is the cause of embryonic lethality (56). In yeast, RNase H activity is particularly important under conditions when dNTPs are limiting, as was demonstrated by Cerritelli et al. (2019) using an RNR1 mutant that impacts Ribonucleotide Reductase (Rnr) activity (59).

Genomic DNA isolated from RNase H2-deficient yeast strains is associated with increased alkali sensitivity, as the presence of a ribonucleotide renders the DNA backbone susceptible to hydrolysis under alkaline conditions. This can be exacerbated genetically in yeast by expressing a mutant form (pol2-M644G) of the leading strand replicase, Pol ε, which renders the enzyme less able to discriminate against ribonucleotide incorporation during synthesis (51). Genomic DNA isolated from a pol2-M644G rnh201Δ strain is extremely sensitive to alkaline hydrolysis, as demonstrated by the appearance of small DNA fragments on an alkaline-agarose gel (51) that correspond to nascent leading strand DNA (60). Alkaline hydrolysis has now been used to demonstrate the presence of genomic ribonucleotides in multiple model systems when they are RNase H2-deficient, including mouse (57) and Archaea (61). Additional approaches, such as single cell gel electrophoresis (62) using a modified comet assay where lysed cells are treated with RNaseHII to induce cleavage at ribonucleotides and generate DNA breaks have also been informative.

3.2. Unrepaired ribonucleotides as a biomarker of replicase action

Variant of Pols α, δ and ε containing amino acid substitutions in the polymerase active site have been useful in assigning their respective roles in replicating the two strands of nuclear DNA. These efforts began with studies of the mutational specificity of three variants with amino acids substituted for residues immediately adjacent to the steric gate tyrosine in the B-family replicases, i.e., Pol α L868M, Pol δ L612M and Pol ε M644G. Each polymerase was found to have a specific and distinctive mutation signature in vivo, and whole genome sequencing of strains harboring these variants revealed the strand-specific types and locations of substitutions and insertion/deletion (indel) errors made by each. The results (63–68) suggested that Pol ε makes mistakes located primarily in the leading DNA strand during replication, while Pols α and δ primarily make mistakes during replication of the lagging strand. This idea has since been applied to the ability of these and similar variants to incorporate rNMPs into DNA in yeast strains lacking RER. Because the rate of ribonucleotide incorporation (Table 2) exceeds the rates of making point mutations by 100- to 1,000-fold, and because loss of RER does not result in severe growth defects in yeast, ribonucleotide incorporation has been used as a sensitive biomarker of the roles of Pols α, δ and ε in eukaryotic nuclear DNA replication. In 2015, four groups developed whole genome sequencing strategies to map ribonucleotide incorporation into RER-defective yeast genomes (69–72). The sensitivity of these strategies has been improved over time (reviewed in (43)), and results from these approaches have led to the following conclusions.

3.2.1. Roles of Pols α, δ and ε in normal DNA replication

In an RER-defective Saccharomyces cerevisiae strain encoding the pol2-M644G variant of Pol ε, ribonucleotides are primarily incorporated into the leading DNA strand (Figure 3, (60, 69, 73–75). In RER-defective Pol α and Pol δ strains harboring mutations that elevate ribonucleotide incorporation (pol1-L868M or Y869A and pol3-L612M/G), ribonucleotides are primarily incorporated into the lagging strand (69, 73–75). These results are consistent with the conclusions derived from the mutagenesis studies mentioned above. They strongly suggest that Pol ε is the primary leading strand replicase and that Pols α and δ are the primary lagging strand replicases (Figure 3).

Figure 3. The use of ribonucleotide mapping to characterize the stages of replication.

A. Mapping of ribonucleotide incorporation by the DNA polymerases during replication initiation provided insight into how leading strand synthesis is initiated. At a replication origin, the first lagging strand Okazaki fragment initiates leading strand synthesis, meaning that Pol δ synthesizes a short stretch of leading strand DNA before Pol ε takes over. ACS: ARS (autonomously replicating sequence) consensus sequence.

B. The bulk of synthesis follows the canonical division of labor, with Pol ε acting as the primary leading strand replicase and Pol δ as the primary lagging strand polymerase.

C. During replication termination, Pol δ takes over from Pol ε for the last few Kbps of leading strand replication. Figure 2A in (43).

Higher resolution inspection of ribonucleotide incorporation (74, 75) indicates that yeast Pols α and δ incorporate ribonucleotides into both DNA strands at origins of replication. This leads to the model in Figure 3A, which suggests that during synthesis of the initial Okazaki fragment at replication origins, replication is initiated by Pols α and δ. This occurs until a “collision-release” process permits Pol δ to depart (76), thereby allowing Pol ε to take over to continuously replicate the leading strand (Figure 3B). Evidence for this model is further supported by results in a reconstituted replication system that includes purified yeast proteins (77). Similar ribonucleotide incorporation patterns were found in S. pombe, which has more diffusive origin activation across the genome that is similar to humans (78–81). Zhou et al. (75) also provided ribonucleotide incorporation data suggesting that Pol δ synthesizes both strands during replication termination, when two replication forks collide (Figure 3C).

The most abundant source of ribonucleotides in nuclear DNA are those incorporated into the RNA-DNA primer used to initiate lagging strand DNA synthesis by the Pol α-primase. Although a significant number of these ribonucleotides are removed during Okazaki fragment maturation following Pol δ-dependent strand displacement synthesis (82), RNase H2 is not able to excise the 3′ terminal ribonucleotide of the RNA primer (49, 83), and there is additional evidence that some of these ribonucleotides remain. This includes the fact that both mismatch mutagenesis and genomic ribonucleotide incorporation are elevated in Pol α variants (L868M, Y869A) with reduced fidelity for incorporation of the correct base as well as the correct sugar (68, 69, 72), suggestive of retention of Pol α -synthesized DNA representing ~1.5% of the yeast genome following lagging strand maturation (72).

Recent evidence from high-throughput sequencing approaches suggests that ribonucleotide incorporation into DNA by the replicases is non-random, and that there are signatures and patterns associated with ribonucleotide incorporation into the genome. Analysis of RNase H2-deficient yeast strains revealed that rC and rG are more frequently incorporated than rU and rA (69, 71, 84). Furthermore, the deoxyribonucleotide incorporated by the DNA polymerase immediately upstream of the ribonucleotide appears to influence ribonucleotide distribution, with dA most often located upstream of the most frequently incorporated ribonucleotides, rC and rG (84). This preference is not directly related to nucleotide pool concentrations, as one would expect rA to be the most abundant ribonucleotide based on the large rATP:dATP ratio (Table 1). This may be due to high discrimination against rA incorporation by the replicases (18), or related to the immediate upstream DNA sequence which may influence ribonucleotide incorporation or extension from a 3′-ribonucleotide, a substrate known to impede DNA polymerase extension (21, 29).

3.2.2. Break Induced Replication (BIR) and Replication Restart.

BIR is a specialized form of recombination-based replication for repair of a DNA DSB formed when the DNA replication fork collapses. Fork collapse leaves one end of the DNA to initiate homologous recombination (HR) by strand invasion as a means of bypassing the lesion that stalled the fork. This involves a migrating displacement loop (D-loop) and DNA synthesis is initiated at the 3′ invaded end (reviewed in (85). Until recently, polymerase usage during BIR was unclear. This was investigated by mapping genomic ribonucleotides by HydEn-seq in RNase H2-deficient yeast strains expressing polymerase variants with increased propensity for ribonucleotide incorporation (86). The use of ribonucleotide incorporation as a biomarker for polymerase action provided evidence that Pol δ synthesizes both the leading and lagging strands during BIR (see Figure 2B–C in (43)), and that this synthesis is largely independent of Pol ε. Consistent with this finding, while Pol ε depletion from cells had no effect on BIR, depletion of Pol δ reduced BIR (86).

Replication restart is another specialized form of DNA replication that occurs after disruption of the replication fork, which leads to replication fork collapse, followed by restart by HR. During replication fork restart in S. pombe, it was demonstrated, through analysis of ribonucleotides incorporated by mutant replicative polymerases, that Pol δ performs both lagging and leading strand synthesis (70) in a process similar to BIR. A recent, more detailed analysis showed that Pol δ continues to synthesize HR-restarted replication forks for up to 30 kb (87).

3.3. Topoisomerase 1-mediated removal of ribonucleotides

Just as for other non-canonical nucleotides in DNA, ribonucleotides incorporated into DNA during replication can be processed by more than one pathway. One of these processes is removal by topoisomerase 1 (Figure 2B). In 1997, Sekiguchi and Shuman (88) provided evidence that vaccinia topoisomerase 1 (Top1), the enzyme well known to have a critical role in removing positive and negative supercoils that accumulate during transcription and DNA replication, possessed ribonuclease activity (89). Interestingly, further evidence that ribonucleotides remaining in DNA in RNase H2-deficient yeast strains are removed by Top1 is a decrease in deletion mutations involving loss of 2- to 5-base pairs in repetitive DNA sequences when comparing the rnh201Δ top1Δ strain to the rnh201Δ TOP1⁺ strain (see below, and (51, 60, 90). The majority of this removal is likely to be non-mutagenic, because the number of alkali-sensitive sites (60) appears to be much greater than the rate of 2–5 base pair deletions in the pol2-M644G strain containing a high density of ribonucleotides in the leading strand (51, 60).

3.4. Other pathways for removing genomic ribonucleotides

Still other pathways for removal of ribonucleotides from DNA have been identified. These include removal of ribonucleotides incorporated by the primase activity of Pol α that initiates replication. During Okazaki fragment maturation of the lagging DNA strand, the RNA/DNA portion of the preexisting downstream fragment can be displaced by Pol δ to generate a flap that is cleaved by the Fen1 or DNA2 nucleases (82). Removal of damaged ribonucleotides can also involve additional removal pathways. For example, oxidized and abasic ribonucleotides, while not recognized and removed by RNase H2-dependent RER, can be processed by the APE1 endonuclease (28, 91). And ribonucleotides found in longer RNA-DNA hybrids, such as R-loops formed during transcription, can be removed by both RNases H1 and H2 (reviewed in (55). Recently, the DEAD-box RNA helicase, DDX3X, was shown to have RNase H2-like activity able to support reconstituted RER reactions, and silencing of DDX3X resulted in accumulation of unrepaired genomic ribonucleotides (92). Interestingly, DDX3X was able to support RER reactions with Pol δ, Pol β and Pol λ, suggesting that in human cells, DDX3X may promote a unique pathway of repair that can involve both the replicative and the DNA repair polymerases.

4. BIOLOGICAL CONSEQUENCES OF RIBONUCLEOTIDES IN DNA

Ribonucleotides incorporated by Pols α, δ and ε can have biological consequences that in certain cases are clearly negative (Figure 2C), but in other cases may provide a selective advantage (Figure 2D).

4.1. Negative consequences

4.1.1. 2 – 5 bp deletion mutagenesis.

As mentioned above, RNase H2-deficient budding yeast have a strongly increased rate of 2–5 base pair deletions in repetitive DNA sequences (Figure 4A). This observation was first made using specific reporter genes (51, 60, 73, 90, 93), and has since been reported across the entire yeast genome (94). The rate of these deletions strongly decreases upon deletion of the gene encoding Top1 (60, 73, 90, 94). These data suggest a model (Figure 2B) wherein ribonucleotides incorporated into short repetitive sequences during DNA replication are incised twice by Top 1 and then religated, thus generating short deletions mutations. The rate of these deletions is much higher in a yeast strain encoding the pol2-M644G mutant of Pol ε than in strains harboring the pol1-L868M variant of Pol α, or the pol3-L612M mutant of Pol δ (73). This asymmetry in the mutagenic potential of newly incorporated ribonucleotides deserves more attention in the future. It is likely related to the different functions of the three major polymerases in continuous replication of the leading strand versus discontinuous replication of the lagging DNA strand. Future studies may also provide information on additional genomic parameters that impact the rates of these deletion mutations, including transcription and chromatin status as well as specialized regions that are rich in repetitive sequences, and in the origin of certain diseases (see below).

Figure 4. The absence of ribonucleotide removal by RNase H2-dependent RER can lead to point mutations and larger forms of genome instability.

A. Top1-dependent 2–5 bp deletion mutagenesis can occur following Top1 incision at a ribonucleotide incorporated into repetitive DNA during leading strand synthesis by Pol ε. The steps of this pathway are described in detail in Figure 2B.

B. DNA breaks are produced when Top1 incises unrepaired ribonucleotides. In the absence of RER, Top1 cleavage at a ribonucleotide generates a DNA nick (single strand break; SSB) containing a 2′−3′-cyclic phosphate end (red triangle). If this is followed by cleavage by Top1 on the opposite DNA strand in close proximity to the first cleavage, this leads to formation of a double strand break (DSB). The presence of a covalently-linked Top1 is prone to intermolecular ligation with other DNA ends. Alternatively, the DSB can be repaired by homologous recombination (HR) involving the Rad51 and Rad52 proteins.

C. DNA breaks produced during Top1-processing of unrepaired ribonucleotides can lead to recombination and larger structural changes to the chromosomes. Single strand breaks (SSBs) can result from spontaneous hydrolysis or Top1-cleavage at a ribonucleotide. Double strand breaks (DSBs) can be generated directly or indirectly following Top1 cleavage events or conversion of a Top1-induced SSB to a DSB. Genome rearrangements that include gross chromosomal rearrangements (GCRs), loss-of-heterozygosity (LOH) and chromosomal translocations have been observed in yeast and mouse cells.

4.1.2. Larger genomic rearrangements

In addition to short deletion mutations, larger scale genomic rearrangements also occur as a result of recombination in RNase H2-deficient yeast and mouse cells (Figure 4C) (57, 95–98). A major source of these events is the recombination that occurs following cleavage at unrepaired ribonucleotides by Top1. In a pol2-M644G rnh201Δ yeast strain containing a high density of unrepaired ribonucleotides in the leading strand, Top1 cleavage at ribonucleotides can lead to the formation of DNA double-strand breaks (DSBs) that are repaired via Rad52-dependent homologous recombination (HR) (Figure 4B, (93, 99)). The formation of γH2AX foci and micronuclei in mouse RNase H2-deficient embryos is also indicative of DSB formation (57).

4.1.3. Ribonucleotide incorporation by the fourth B-family member, DNA polymerase ζ

Pol ζ is an error-prone polymerase involved in translesion DNA synthesis via its ability to bypass lesions that stall replication forks. In doing so, Pol ζ is responsible for the majority of DNA damage-induced mutagenesis in eukaryotes (reviewed in (100). Pol ζ incorporates ribonucleotides during synthesis, is able to extend from a ribonucleotide present at the 3′ terminus of a primer, and efficiently bypasses ribonucleotides present in template DNA (31, 101). More work is warranted on this polymerase, including on its contribution to genome instability across the nuclear genome.

4.1.4. Ribonucleotides incorporated during mitochondrial replication

Although this review is focused on ribonucleotides incorporated into DNA during nuclear replication, ribonucleotides are also inserted into mitochondrial DNA during synthesis by DNA Pol γ (see (69, 102) and references therein). Ribonucleotides in mitochondrial DNA have been mapped and quantified (42). In contrast to the efficient removal of genomic ribonucleotides by RER that occurs in the nucleus, there does not appear to be a ribonucleotide repair pathway in the mitochondria, and therefore the identity and quantity of ribonucleotides in mitochondrial DNA is primarily determined by the rNTP:dNTP ratio in both yeast (103) and mammalian (38) cells. Ribonucleotide content in mitochondrial DNA increase with age, as a possible reflection of the decreased number of proliferating cells in an aging population of cells. Because proliferating cells have higher dNTP concentrations than do quiescent cells, fewer proliferating cells means that dNTP availability is quite low and therefore mitochondrial ribonucleotide content is higher.

4.2. Positive consequences

4.2.1. Mating type switching

One positive signaling function for genomic ribonucleotides is during mating type switching in S. pombe. This event is initiated by replication-coupled recombination (reviewed in (104). At a specific site bound by a trans-acting DNA factor in the mat1 locus, leading strand replication is paused due fork stalling (105). Following this pause, ribonucleotides are incorporated into a lagging strand Okazaki fragment by the Primase-Pol α complex during site-specific priming at a specific distance from the stalled replisome, and priming and synthesis of a second Okazaki fragment follows suit. It is during ligation of these two Okazaki fragments that a dinucleotide imprint remains and is bound by an unknown protein. This protein-bound imprint acts as a barrier to the leading strand replication machinery in the next round of replication to trigger a replication-coupled recombination event that results in a mating type switch (106, 107). Given this precedent, it is possible that other site-specific ribonucleotide-dependent replication stalling events occur in the nuclear genome, providing a developmental signal in other biological contexts.

4.2.2. Ribonucleotide incorporation in the context of chromatin

Nucleosome assembly is inhibited by the presence of a single ribonucleotide in DNA (108, 109). Therefore, it is possible that ribonucleotides in DNA impact the chromatin landscape in multiple ways and thereby affect replication, transcription and DNA repair. Molecular dynamics simulations of a ribonucleotide at various positions within a nucleosome demonstrated impacts on nucleosome structure, dynamics and removal by RER (25). The structural impacts of a ribonucleotide on chromatin may impact RER and other processes that involved DNA binding and/or chromatin assembly and remodeling. (25). In addition, as mentioned above, the presence of a single ribonucleotide in DNA alters helix geometry (4–7) and the electrostatic potential of DNA (110), and could affect multiple signaling pathways that are sensitive to the local chromatin and DNA landscape. Repair of a ribonucleotide in the context of a nucleosome core particle (NCP) is dependent on its position, with ribonucleotides positioned in the central, inward-facing part of the NCP being much less accessible to RNase H2, although they are protected from spontaneous hydrolysis, In contrast, ribonucleotides facing outward in the NCP are more accessible to RNase H2, and potentially to other removal pathways (111).

4.2.3. Non Homologous End Joining (NHEJ) at DSBs

Ribonucleotides incorporated into DNA by the repair polymerases Pol μ and Tdt serve an important function during DSB repair by NHEJ. DNA overhangs present at a DSB are filled in by these polymerases to prepare them for ligation by Ligase 4. Both Pol μ and Terminal deoxynucleotidyl transferase (Tdt), a template-independent polymerase with an important role in V(D)J recombination, incorporate ribonucleotides into DNA very efficiently with almost no discrimination, and they are also able to extend a primer containing a 3′-ribonucleotide (112–114). Dale Ramsden and colleagues (115) demonstrated that when these transient substrate ribonucleotides are incised by RNase H2, they improve the efficiency of ligation by Ligase 4. Consequently, ribonucleotides incorporated into DNA by Pol μ and Tdt play an important function in promoting DNA repair by NHEJ, and can later be removed by RER.

4.2.4. Strand discrimination during mismatch repair (MMR)

Biochemical, genetic and genomic approaches have provided substantial evidence that the major leading strand polymerase, Pol ε, incorporates a significant number of ribonucleotides into DNA, suggesting one or more positive selective advantages for leading strand ribonucleotides. One possible role is during eukaryotic DNA mismatch repair (MMR), a multi-step pathway responsible for correction of various types of replication errors, including base-base mismatches and insertion/deletion mutations (reviewed in (116)). In the first step of MMR, the DNA mismatch is recognized and bound by the MutSα heterodimer (Msh2-Msh6 in yeast). RNase H2-incision at a ribonucleotide that has been incorporated during leading strand synthesis by Pol ε upstream of the mismatch generates a nick that may be used as an entry point for the MutLα heterodimer (Mlh1-Pms2) to load onto DNA (Figure 2D). This is followed by mismatch excision, DNA resynthesis and ligation to finalize the MMR reaction. Thus, nicks created during repair of ribonucleotides incorporated into leading strand DNA may act as one strand-discrimination factor for the correction of errors made in the newly synthesized strand. Consistent with this model, RER improves the efficiency of MMR in yeast and activates MMR in human and mouse nuclear extracts (117, 118). However, an RNase H2-deficient yeast strain (rnh201Δ) has a much lower spontaneous mutation rate (90) than a strain deficient in MMR (msh2Δ) (65), implying that the contribution of RNase H2 to MMR is small compared to nicking by MutLα. Initiation of repair of a small percentage of mismatches by RNase H2 that are not rapidly repaired via MutLα may occur outside of the context of DNA replication (116).

Although there is evidence that MMR is able to repair mismatched ribonucleotides (119), loss of MMR in budding yeast cells does not affect the high rate of ribonucleotide-dependent mutations observed in yeast strains lacking RER (90, 117). This provides evidence that correctly paired genomic ribonucleotides are not repaired by MMR.

5. RIBONUCLEOTIDE EFFECTS ON DISEASE

Several human diseases have now been linked to mutations in proteins critical for genomic ribonucleotide processing and removal from nuclear DNA. These disorders include those involved in autoimmunity and neurodegeneration, as well as cancer. The detailed mechanistic information provided from studies of ribonucleotide incorporation and removal in various model systems will hopefully inform future diagnostic and therapeutic efforts during the identification and treatment of the diseases discussed below.

5.1. Autoimmune disorders

Aicardi Goutières syndrome (AGS) is a severe neuroinflammatory pediatric disorder with symptoms that closely resemble congenital viral infection. Mutations in a number of proteins involved in nucleic acid processing have been identified in cells isolated from AGS patients, including mutations in all three RNase H2 subunits (120). Studies using yeast, mouse and human cells have now been used to provide mechanistic information on how mutations in RNase H2 subunits identified in AGS patients affect the molecular functions of the enzyme and contribute to autoimmune disease (121–124). In one approach, AGS patient mutations were modeled onto the yeast RNase H2 subunits and mutation and recombination rates were measured (121). Nishimura et al. examined the stability and activity of recombinant human RNase H2 variants identified in AGS patients to identify effects on complex stability and enzymatic activity (125).

Another autoimmune disorder that shares phenotypes with AGS is Systemic Lupus Erythrematosis (SLE). In 2014, mutations in RNase H2 subunits that reduced complex stability or enzymatic activity were identified in cells from SLE patients (126), Both AGS and SLE patient cells have increased ribonucleotides present in genomic DNA, increased type 1 interferon production and activation of the DNA damage response (126).

5.2. Ataxia with Oculomotor Apraxia (AOA1)

AOA1 is an autosomal recessive neurological disorder linked to mutations in human APTX, a gene that encodes the conserved Aprataxin protein (127). Aprataxin is a hydrolase that removes 5′ adenylated DNA ends produced when DNA ligase engages a nick containing damaged termini during the final step of DNA replication and repair (128, 129). The first step of RER involves RNase H2 incision 5′ to a ribonucleotide, and a premature attempt by ligase to engage and seal this DNA nick aborts ligation and results in the production of a compound DNA lesion containing a bulky adenylated group (5′-AMP-RNA-DNA) (130). Aprataxin binds and catalyzes the reversal of this 5′-adenylated group, providing another opportunity for repair of the ribonucleotide by RER (130). Therefore, ribonucleotide-triggered abortive DNA ligation creates a toxic intermediate that requires Aprataxin for repair, and accumulation of DNA damage in AOA1 patient cells containing mutations in APTX may cause accumulation of ribonucleotide-triggered genome instability in the brain and contribute to neurodegeneration in AOA1 patients, especially when considering the high rNTP:dNTP ratio that is found in post-mitotic neurons.

5.3. Cancer

There are multiple lines of emerging evidence showing a connection between unrepaired ribonucleotides in DNA and the development of cancer. One major contributing factor may be the nucleotide pool imbalances identified in cancer cells that increase mutagenesis and genome instability, as well as mutations in the exonuclease domains of Pols δ and ε that may affect ribonucleotide incorporation and/or proofreading (40, 131–133). RER activation leads to intestinal tumorigenesis in mice, suggesting that ribonucleotide-dependent mutagenesis, DNA breaks and/or other forms of genome instability may contribute to the development of cancer (134). Loss of RNase H2 in the murine epidermis leads to spontaneous DNA damage, type 1 interferon signaling and the development of squamous cell carcinoma (135).

CRISPR screens of human cells treated with various chemotherapy drugs are providing valuable insight into the connection between loss of RNase H2 and cancer. Using a targeted approach, the RNase H2 genes were identified as causing sensitivity to poly (ADP-ribose) polymerase (PARP) inhibitors when inactivated (136). Top1-incised ribonucleotides produced in the absence of RER were shown to be DNA lesions to which PARP binds and becomes trapped upon treatment with PARP inhibitors, resulting in genome instability (136). RNase H2-deficiency was shown to be synthetic lethal with ATR checkpoint kinase inhibition in a genome-wide CRISPR screen (137). Treatment of RNase H2-deficient cells with a highly specific ATR inhibitor caused apoptosis or senescence, and reduced levels of RNASE H2 were identified in prostate cancer samples (137). Taken together, these results suggest that ATR inhibition may be an effective therapeutic approach to treating cancer patients with reduced levels of RNase H2. More recently, the synthetic lethality that exists between the APE2 nuclease and cells with BRCA1 or BRCA2 mutations was shown to be the result of the inability of APE2 to reverse the 3’ damage produced when Top1 incises an unrepaired ribonucleotide (138). Results from these studies highlight the possible therapeutic potential that may exist in inactivation of RNase H2 activity in cancer cells, or treatment of cells deficient in RNase H2 with appropriate chemotherapeutic drugs.

6. SUMMARY AND PERSPECTIVES

The literature on ribonucleotide incorporation and removal from DNA continues to rapidly emerge, as do the implications for human health and possible therapeutic potential. Clearly there is still much to learn about how all of the invaluable mechanistic properties identified in biochemical, structural and genetic studies performed in model systems extends to the human genome and nuclear biology. For example, it is possible that ribonucleotides have important unidentified physiological roles in mammalian cells, possibly as a developmental signal during embryogenesis. Modulation of nucleotide pools as a therapeutic strategy for treatment of such disorders as viral infection and cancer offers a potentially promising avenue of affecting the fidelity of DNA synthesis.

The use of the invaluable mechanistic information provided from studies in model organisms should inform cancer sequencing efforts regarding the origin of cancer mutation signatures that are the result of processing of unrepaired ribonucleotides in DNA. One technical challenge associated with detection of a mutation signature result from failure of RER is the fact that they involve deletions of between 2 and 5 bases in size in repetitive DNA sequences. Bioinformatically, this can be a greater technical challenge to detect than single base substitutions and additions or deletions. Future studies will be required to gain a full understanding of the physiological relevance of unrepaired ribonucleotides in nuclear genomes, including those incorporated by both replicases and DNA repair polymerases. We expect that the recent explosion of interest in genomic ribonucleotide incorporation and repair will continue to unveil surprising and informative mechanistic information and hopefully inform therapeutic strategies.