Creation and Removal of Embedded Ribonucleotides in Chromosomal DNA during Mammalian Okazaki Fragment Processing*

Mammalian RNase HI has been shown to specifically cleave the initiator RNA of Okazaki fragments at the RNA-DNA junction, leaving a single ribonucleotide attached to the 5′-end of the downstream DNA segment. This monoribonucleotide can then be removed by the mammalian 5′- to 3′-exo-/endonuclease, a RAD2 homolog-1 (RTH-1) class nuclease, also known as flap endonuclease-1 (FEN-1). Although FEN-1/RTH-1 nuclease often requires an upstream primer for efficient activity, the presence of an upstream primer is usually inhibitory or neutral for removal of this 5′-monoribonucleotide. Using model Okazaki fragment substrates, we found that DNA ligase I can seal a 5′-monoribonucleotide into DNA. When both ligase and FEN-1/RTH-1 were present simultaneously, some of the 5′-monoribonucleotides were ligated into DNA, while others were released. Thus, a 5′-monoribonucleotide, particularly one that is made resistant to FEN-1/RTH-1-directed cleavage by extension of an inhibitory upstream primer, can be ligated into the chromosome, despite the presence of FEN-1/RTH-1 nuclease. DNA ligase I was able to seal different monoribonucleotides into the DNA for all substrates tested, with an efficiency of 1–13% that of ligating DNA. These embedded monoribonucleotides can be removed by the combined action of RNase HI, cutting on the 5′-side, and FEN-1/RTH-1 nuclease, cleaving on the 3′-side. After FEN-1/RTH-1 action and extension by polymerization, DNA ligase I can join the entirely DNA strands to complete repair.

exo-/endonuclease, called RTH-1 or FEN-1 (3)(4)(5)(6)(7). RNase HI makes a structure-specific cleavage, releasing the initiator RNA as an intact segment but leaving a single ribonucleotide on the 5Ј-end of the downstream DNA. FEN-1/RTH-1 nuclease can then remove this monoribonucleotide (8). Although FEN-1/RTH-1 cleavage often requires an upstream primer for stimulation, the presence of an upstream primer is sometimes neutral or even inhibitory, especially for removal of these monoribonucleotides (9). If an upstream primer approaches before FEN-1/RTH-1 action and inhibition at the particular junction is significant, Okazaki fragment processing may be halted, leaving a nick in the chromosome, just upstream of a single ribonucleotide. In such a situation, processing might still be completed via the endonucleolytic activity of FEN-1/RTH-1 nuclease (9). However, if DNA ligase is able to seal such monoribonucleotides into duplex DNA, upstream primer inhibition of FEN-1/RTH-1 provides an opportunity for the ligation to take place. Here, we explore the consequences of such a reaction with respect to the desired efficient joining of all Okazaki fragments.
Three distinct eukaryotic DNA ligases have been distinguished based on catalytic, physical, and immunologic properties (1,10). Recently, a fourth DNA ligase was discovered by sequence homology to other ligases, although its full characterization is still pending (11). All DNA ligases catalyze the formation of phosphodiester bonds between adjacent 5Ј-phosphoryl and 3Ј-hydroxyl termini at single strand nicks in duplex DNA (12). DNA ligases from bacteriophages T4 and T7 and all eukaryotes use ATP as a coenzyme for bond energy. Eukaryotic ligases catalyze three distinct reactions: formation of a ligaseadenylate complex with release of PP i , transfer of the adenylyl group to the 5Ј-phosphate, and attack of the 3Ј-hydroxyl on the activated 5Ј-phosphate, to form the bond, with release of AMP (1).
DNA ligase II is not induced with cellular proliferation (13), suggesting a role in DNA repair (14). It is structurally related to DNA ligase III, and it is presently unclear whether the two are encoded by separate genes or are the result of differential mRNA processing or post-translational modification (15,16). DNA ligase III complexes with the XRCC-1 repair gene product (17)(18)(19). XRCC-1 minus Chinese hamster ovary cell mutants are hypersensitive to alkylating agents and ionizing radiation, are defective in repair of single strand breaks, and have hyperfrequent spontaneous sister chromatid exchanges (20). In addition, a multiprotein complex of DNA polymerase ⑀, FEN-1/ RTH-1 1 nuclease, and DNA ligase III has been isolated from calf thymus and shown to function in vitro in recombination and repair (21). Little is known about DNA ligase IV except that it has a unique C terminus (11).
Of the four eukaryotic DNA ligases, DNA ligase I is the enzyme believed to be involved in DNA replication due to many different lines of evidence. Thus, its potential ability to seal monoribonucleotides into DNA is of particular importance. DNA ligase I is essential for completion of lagging strand synthesis in the SV40 system (22)(23)(24)(25). DNA ligase III is not a substitute (24). The activity of DNA ligase I is higher in proliferating than in quiescent cells (13). Homozygous knockouts in murine embryonic stem cells show that DNA ligase I is essential (26). Compared with DNA ligases II and III, DNA ligase I has higher fidelity of DNA joining and is particularly sensitive to mismatches at the 3Ј-end of the upstream primer (16). The homologs CDC9 in Saccharomyces cerevisiae and CDC17 in Schizosaccharomyces pombe (27,28) are required for replication, repair, and recombination (29). The human DNA ligase I cDNA complements the CDC9 mutant (30). DNA ligase I is also implicated in at least two human diseases, which help clarify the in vivo role of this enzyme. A fibroblast cell line derived from a patient with missense mutations in both alleles of the DNA ligase I gene shows defective Okazaki joining (31). Bloom's syndrome cells, which have decreased or abnormal DNA ligase I activity (32)(33)(34)(35)(36), are also defective in DNA replication (37).
Immunocytochemistry and immunofluorescence show that DNA ligase I is localized to the nuclei of intact cells (38). It is found in replication factories with other replication proteins during S phase but is diffuse during other phases. A 13-amino acid nuclear localization signal has been identified in the Nterminal regulatory domain of the protein, with an adjacent 115-amino acid sequence required to direct the enzyme to the sites of DNA replication. Thus, activity in vivo may be controlled by subnuclear compartmentalization (39). Since DNA ligase I is present at the sites of Okazaki fragment processing, it has the opportunity to seal monoribonucleotides into the nascent DNA.
Substrate specificities of different DNA ligases vary widely (1,12). DNA ligases II and III can join oligo(dT)-poly(rA). Ligases I and III can join oligo(rA)-poly(dT). DNA ligase I and T4 DNA ligase can join blunt ends (10,40). T4 DNA ligase can join oligo(rA)-poly(dT) but not the similar oligo(rU)-poly(dA), and Escherichia coli DNA ligase joins oligo(dT)-poly(dA) but not oligo(dA)-poly(dT). T4 and E. coli DNA ligases and DNA ligase I can all link the 3Ј-OH of RNA to the 5Ј-P of DNA, but ligation of upstream DNA to RNA had not been shown for any of them. In fact, for the E. coli enzyme, the latter reaction has been shown not to occur (1). Given these variations, ligase activities on natural substrates clearly cannot be predicted based on action with homopolymers; yet, most studies thus far have been done with homopolymers, largely as a way to distinguish the enzymes from each other during purifications (1).
In the current report, we find that DNA ligase I indeed has the ability to ligate monoribonucleotides into double-stranded DNA. It can even do so in the presence of FEN-1/RTH-1 nuclease. The embedded RNA is susceptible to alkaline digestion, while the ligation product of a control substrate consisting entirely of DNA is not. Experiments with highly purified recombinant enzyme clearly demonstrate that DNA ligase I possesses both DNA ligation and monoribonucleotide ligation activities. Since DNA ligase I can ligate monoribonucleotides, the cell must have a mechanism to remove these nucleotides. We have shown here that RNase HI and FEN-1/RTH-1 nuclease can carry out this repair function. Enzyme Purification-Calf FEN-1/RTH-1 nuclease was purified through hydroxyapatite chromatography as described previously (3,9). To remove contaminating RNase H activity, this preparation was further purified. First, the nuclease was eluted from an 8-ml heparin-Sepharose column developed with an 80-ml gradient from 100 to 750 mM KCl. This pool was then subjected to chromatography on 1 ml of CM-Sepharose using a 10-ml gradient from 100 to 750 mM KCl. Final specific activity was 194,000 units/mg, with 1 unit defined as the amount of nuclease required to release 5 pmol of [ 32 P]TMP from 5Ј-[ 32 P]dT 16 -dA 2000 in 15 min at 37°C.
Calf thymus RNase HI was purified by the procedure of Eder and Walder (41). Details of the purification have been described in Turchi et al. (8) and Huang et al. (9), except that the heparin-Sepharose column chromatography was carried out directly following phenyl-Sepharose chromatography, and only Mono S was employed after blue Sepharose chromatography. The resultant active fractions were dialyzed into final dialysis buffer, which contained 50 mM NaCl, 50 mM Tris-HCl (pH 7.5), 5 mM DTT, 2 mM EDTA/EGTA, and 50% glycerol. The enzyme remains stable when stored at Ϫ80°C for several months. The final pool of RNase HI had a specific activity of 150,000 units/mg as measured on a poly-([ 3 H]rA) 38 -137 (dT) 16 substrate according to Eder and Walder (41).
DNA ligase I was purified according to the procedure of Tomkinson et al. (42). A Sephacryl S-200HR (Pharmacia) column was used in place of the Ultrogel AcA 34 column. Final specific activity was 50,000 units/mg, with 1 unit defined as the amount of protein that converts 5 pmol of terminal phosphate residues to calf intestinal phosphatase-resistant form in 15 min at 25°C. Ligase activity was followed throughout the purification as in Tomkinson et al. (42), monitoring for the ability to convert phosphomonoesters to calf intestinal phosphatase-resistant diesters. Ligase activity was also followed using our standard DNA control substrates and our standard monoribonucleotide substrates on 10% polyacrylamide, 7 M urea gels (43). We could then observe both DNA to DNA ligations and embedding of a monoribonucleotide, using Phospho-rImager (Molecular Dynamics) quantification. Protein was determined by the method of Bradford (44), using protein assay dye from Bio-Rad. Thus, we could observe the coelution of protein, DNA ligase activity by two methods, and junction monoribonucleotide ligation activity. DNA ligase-adenylate formation was also assayed as in Tomkinson et al. (42) via 7.5% SDS-polyacrylamide gel electrophoresis. An aliquot of this assay was run beside an aliquot directly from the enzyme preparation. The preparation ran as a single band by silver stain at the expected molecular weight and to the same position as the adenylated enzyme as detected by autoradiography.
Recombinant human DNA ligase I was generously provided by Dr. Alan Tomkinson. The enzyme was expressed in baculovirus-infected insect cells (45) and purified to greater than 90% homogeneity, essentially as in Tomkinson et al. (42). The recombinant enzyme has a specific activity of 2.5 units/mg, with 1 unit defined as the amount of protein that converts 1 nmol of terminal phosphate residues to a phosphatase-resistant form in 15 min at 16°C (45). Storage buffer contained 50 mM Tris-HCl, pH 7.5, 300 mM NaCl, 1 mM EDTA, 0.5 mM DTT, and 10% glycerol, and the enzyme is stable at Ϫ80°C. 2 In this report, all discussion of DNA ligase I refers to the enzyme purified from calf thymus unless specifically designated otherwise.
Substrates-The sequences of the primers and the structures of the substrates used in this study are described in Table I and in the figures. T4 polynucleotide kinase was used to 5Ј-phosphorylate the downstream primers as specified using [ 32 P]ATP according to the manufacturer's protocol. Substrates designed to assay ligation of junction monoribonucleotides by DNA ligase I were constructed by annealing the appropriate 5Ј-end-labeled synthetic monoribonucleotide-DNA segment and corresponding upstream primer to the template and then isolating via 12% native gel electrophoresis (43). The resulting substrates, each having only a nick between primers, were eluted from the gel using elution buffer (0.5 M ammonium acetate, 0.1% sodium dodecyl sulfate, 0.1 mM EDTA), ethanol-precipitated, and resuspended in 1 ϫ annealing buffer (50 mM Tris, pH 8, 10 mM magnesium acetate, 50 mM NaCl, 1 mM DTT). To form a nicked substrate when downstream primer 10 or 9 was used, corresponding upstream primer 1 or 4 was annealed first and extended with 10 units of Sequenase by 1 or 3 nucleotides, respectively, as indicated in Table I. Control substrates, consisting entirely of DNA, were made similarly. These substrates are the same as the nick substrates of a previous study examining FEN-1/RTH-1 junction cleavage activity (9).
For studying the repair of embedded monoribonucleotides, the above nick substrates were first ligated in bulk by T4 DNA ligase. Ligated product was isolated from starting material and then reannealed to template to serve as substrate for repair. The bulk ligation was done in 66 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 1 mM DTT, 1 mM ATP, 100 fmol of substrate, and 10 units of enzyme in a final volume of 250 l at 37°C overnight. Isolation was via 10% denaturing gel, followed by electroelution using a UEA Electroeluter (International Biotechnologies Inc., New Haven, CT), and ethanol precipitation.
Enzyme Assays-The DNA ligase I junction monoribonucleotide assay was performed in FEN-1/RTH-1 buffer containing 60 mM BisTris (pH 7.0), 5% glycerol, 0.1 mg/ml bovine serum albumin, 5 mM ␤-mercaptoethanol, 10 mM MgCl 2 , 10 mM ATP, and 80 fmol of substrate in a volume of 192 l. Reactions were initiated by the addition of 24 units of enzyme and incubated at 37°C. Reactions were stopped at the appropriate times by adding 25 l of sample to an equal volume of 2 ϫ formamide dye (98% formamide and 10 mM EDTA (pH 8.0) with 0.01% (w/v) each xylene cyanol and bromphenol blue) and heating at 95°C for 5 min. Zero time controls were removed from the reaction mixture before the addition of enzyme, as were T4 ligase controls. One unit of T4 DNA ligase was added to these positive controls, which were incubated for 15 min at 37°C. Products were separated by 10% polyacrylamide, 7 M urea gel electrophoresis (43), visualized by autoradiography using a Dupont Cronex Lightning Plus intensifying screen at Ϫ80°C, and analyzed via PhosphorImager.
Experiments with recombinant human DNA ligase I were performed as described above, with 70 fmol of substrate in a final volume of 168 l. From this mixture, 24-l aliquots were drawn for the addition of 1 l of water, 1 l of T4 DNA ligase, and the indicated amounts of recombinant human DNA ligase I, respectively. All reactions were brought to a final volume of 25 l with water, incubated at 37°C for 60 min, terminated, and processed as above.
The DNA ligase I, FEN-1/RTH-1 competition assay was also performed under the same buffer conditions as the above assays, with 50 fmol of substrate in a final volume of 115 l. From this mixture, 23-l aliquots were drawn. To separate aliquots were added 1 l of water, 1 unit of T4 DNA ligase, 4 units of DNA ligase I, and 0.2 units of FEN-1/RTH-1 nuclease. To a fifth 23-l aliquot, 4 units of DNA ligase I and 0.2 units of FEN-1/RTH-1 were added simultaneously. Water was added to bring the volume of each reaction to 25 l. Reactions were incubated at 37°C for 30 min and were terminated and processed as above, except the gel used was 18% polyacrylamide, 7 M urea. A snake venom phosphodiesterase ladder of an unrelated DNA was run on the gel simultaneously as a size marker.
The assay of RNase HI, FEN-1/RTH-1 repair of embedded monoribonucleotides was also done similarly to the above. As with DNA ligase I, we found that RNase HI activity is efficient in FEN-1/RTH-1 buffer. Thus, FEN-1/RTH-1 buffer conditions (minus the ATP, which serves only as a substrate for ligase) were used, with 50 fmol of substrate in a final volume of 40 l. From this mixture, three 8-l aliquots were drawn for the addition of 1 l of water, 1 unit of T4 DNA ligase, and 0.2 units of FEN-1/RTH-1 nuclease, respectively. To the 16-l remainder were added 0.6 units of RNase HI. Following incubation for 30 min at 37°C, this latter aliquot was divided into two 8-l aliquots, one of which received 0.2 units of FEN-1/RTH-1 nuclease. This gave a 1:1000 RNase HI to FEN-1/RTH-1 molar ratio, which has been shown to be optimal for these two enzymes to work together (9). Each reaction was brought to a final volume of 10 l with water and was incubated for an additional 30 min at 37°C. They were stopped and processed as above. Sometimes additional reactions were performed, which included Sequenase extension by one or two nucleotides of the upstream fragment remaining after RNase HI cleavage, so that FEN-1/RTH-1 could remove the exonucleolytically resistant monoribonucleotide by an endonucleolytic mechanism.
Alkaline digestion assays were also performed in FEN-1/RTH-1 buffer (with ATP) with 37.5 fmol of substrate in a final volume of 90 l. From this mixture, 24-l aliquots were drawn for the addition of 1 l of water and 1 unit of T4 DNA ligase, respectively. The reaction was initiated by the addition of 7 units of DNA ligase I to the remainder, and all tubes were incubated for 30 min at 37°C. From the DNA ligase I reaction, 25-and 8-l aliquots were drawn. To the latter was added 1 l of 0.25 M NaOH, followed by 30 min of incubation at 65°C. Alkaline digestion was quenched by the addition of 1 l of 0.25 M acetic acid. Reactions were terminated and processed as above using a 10% polyacrylamide, 7 M urea gel. RNase H digestion assays were done as with alkaline digestion except that 0.3 units of RNase HI were added to the 8-l aliquot instead of NaOH, incubation was at 37°C, and no acetic acid was added.

DNA Ligase I Can Join an Upstream DNA Primer to a
Downstream Monoribonucleotide-DNA Segment-Downstream primers 5-10 were each annealed to template T1, along with their respective upstream primers, to form a nick (see Table I). The 5Ј-most nucleotide of each of these downstream primers is a ribonucleotide. Thus, these substrates should be suitable for ligation, but only if DNA ligase I can join an upstream DNA segment to a downstream monoribonucleotide. Fig. 1A shows that DNA ligase I can indeed embed monoribonucleotides. In this case, upstream primer 1ϩ was ligated to downstream primer 10, which includes a 5Ј-U. Quantification using Phos-phorImager analysis indicates that the percentage of ligation reached 8.58% in 30 min, under the conditions employed. For this substrate, ligation by T4 DNA ligase was less efficient than by DNA ligase I. For most substrates, T4 DNA ligase I was more efficient and was typically used to provide a positive control. DNA ligase I ligated another 5Ј-U, on primer 8, with similar efficiency (8 -10%), while a 5Ј-rA, a 5Ј-rG, and two 5Ј-rC nucleotides, on primers 5, 6, 7, and 9 respectively, were ligated less well (1-2%) (data not shown, but see other figures for ligations of other monoribonucleotides). Although these experiments clearly indicate that ligation of monoribonucleotides is possible, at least at certain junctions, it is not nearly as effi- TCGAGCTTTAATTGGGAGTGATTTCCCTTGTTTTCGAACGTACGGACGTCCAGCTGAGATCTCCTAGGGGCCCAT cient as DNA to DNA ligations. Control substrates were created to correspond to the monoribonucleotide substrates, having the exact same sequence but consisting entirely of DNA. Fig. 1B shows ligation of downstream primer 12, with a 5Ј-dG, to its upstream primer, 2. Although the corresponding 5Ј-rG ligated quite inefficiently, as discussed above, ligation of this 5Ј-dG, and indeed all tested DNA-DNA ligations, clearly proceeded very well. At 30 min, the percentage of ligation of a typical DNA substrate is approximately 75%. Fig. 1C shows a short time course for ligation of the 5Ј-dT on primer 14. For this and all other tested DNA-DNA ligations, 10% ligation, the maximum seen for any monoribonucleotide, is reached and passed by 2 min. Experiments were repeated with all six monoribonucleotide substrates and with four control substrates 3-7 times, and results were always similar. A Monoribonucleotide Embedded by DNA Ligase I Can Be Digested by Alkaline Hydrolysis and by RNase HI-Since DNA ligase I had not previously been shown to join the 3Ј-OH of a DNA segment to the 5Ј-P of an RNA, we wished to rule out the possibility that the small percentage of ligation observed above was due to a contamination of fully DNA oligomers in our RNA-terminated synthetic substrates. To do so, we attempted to digest away the ligated product with base. Fig. 2A shows that alkali could digest the DNA ligase I-ligated product of an RNA-containing substrate. Lane 3 represents an aliquot of the reaction products run in lane 2 but subjected to alkaline hydrolysis. The ligated product present in lane 2 is very faint due to the relative inefficiency of ligation of these monoribonucleotides. However, this faint band is not present at all in lane 3.
In fact, since base cleavage occurs on the 3Ј-side of the ribonucleotide, a new band has appeared in lane 3. This band runs slightly higher than expected for a 26-mer, but it almost certainly represents the product of alkaline cleavage. This is because it emerges only in this lane, and the mobility of the entire lane is shifted slightly up, because of the high salt concentration in the alkaline digestion. Significantly, no corresponding band appears in the base-treated sample containing the fully DNA substrate (lane 6). A single nucleotide appears at the bottom of lane 3 (slightly shifted up), resulting from alkaline cleavage of the 5Ј-monoribonucleotide from unligated starting substrate. The single nucleotide band, which seems to appear in lane 2 is a tail from lane 3, a phenomenon that we have consistently seen as an effect of the high salt concentration. We considered that it may have resulted from a small FEN-1/ RTH-1 contamination in our DNA ligase I preparation, but, if so, it would also be expected to appear in lane 5. The ligated product of the DNA substrate was not digestible. Ligated bands of approximately the same intensity appear in both lanes 5 and 6, and there is no sign of either a 26-mer or single nucleotide product. Similar results were seen for all sub-  Table I. Downstream primers were 5Ј-radiolabeled, and reaction conditions were as described under "Experimental Procedures." The upper band represents ligated product, and the lower band represents starting material. strates tested (data not shown).
RNase HI has been shown to cleave on the 5Ј-side of a single embedded monoribonucleotide for all four bases in otherwise identical 14-mer primers annealed to complementary DNA (41,46). Based on this observation, we attempted to provide further confirmation of the presence of a ribonucleotide in ligated products by using RNase HI to digest them. Since this cleavage specificity had been tested only for a small number of substrates with very similar sequences, we were unsure whether RNase HI digestion would work on our ligated products, having their own unique sequence environment and lengths of 51-60 nucleotides. Fig. 2B shows that RNase HI, in fact, could digest the product of an RNA-containing substrate, whether produced by DNA ligase I or by T4 DNA ligase, but could not digest the corresponding DNA to DNA product. For the RNA substrate, the faint ligated product present in lane 3 disappears entirely in lane 4 when treated with RNase HI. Additionally, the RNAcontaining product in the T4 control lane has significantly decreased in lane 1. For the DNA substrate, the ligated product in lanes T4 and 7 remains in lanes 5 and 8. Unfortunately, since RNase HI cleaves on the 5Ј-side of the sealed monoribonucleotide, the cleavage product simply returns back into the starting material, and no new band appears in lane 1 or 4 as it did in lane 3 of Fig. 2A. Additionally, no single nucleotide band appears, since the ribonucleotide was not cleaved from the starting material. Similar results were seen for all substrates tested (data not shown). The results presented in Fig. 2 clearly indicate that DNA ligase I is able to join a 5Ј-monoribonucleotide to an upstream primer, because both alkaline and RNase HI were able to later cleave the ligated product.
DNA Ligase I and FEN-1/RTH-1 Nuclease Both Have Activity When Added to Substrate Simultaneously-After demonstrating the ability of DNA ligase I to seal in monoribonucleotides, we wished to verify that this activity is maintained in the presence of FEN-1/RTH-1 nuclease. We predicted that, since FEN-1/RTH-1-directed cleavage of monoribonucleotides is usually inhibited by the presence of an upstream primer, DNA ligase I would have the opportunity to act. Alternatively, given the relative inefficiency of DNA ligase I with monoribonucleotides, FEN-1/RTH-1 nuclease might successfully compete with the DNA ligase I, despite inhibition by the upstream primer. Surprisingly, an intermediate result was observed, whereby both ligation and FEN-1/RTH-1 cleavage occurred (Fig. 3). Experiments like that shown in Fig. 1A had indicated that substantial junction monoribonucleotide ligation activity was achieved by 30 min of incubation. The upstream primer inhibition of all tested substrates was previously characterized by Huang et al. (9). Lane 4 of Fig. 3 demonstrates that FEN-1/ RTH-1 alone is able to cleave the 5Ј-rA from the substrate created with downstream primer 5 and upstream primer 1 very efficiently in 30 min, despite the upstream primer inhibition of this substrate. Similarly, DNA ligase I alone embeds the monoribonucleotide, as shown in lane 5. In lane 6, both a ligated product and a cleavage product appear, when DNA ligase I and FEN-1/RTH-1 were added simultaneously. These results suggest that DNA ligase I can embed monoribonucleotides in vivo, when FEN-1/RTH-1 is present. Neither enzyme prevents the action of the other. Each appears to be independently equilibrating with the available substrate molecules to carry out their respective reactions. Some inhibition of each activity is evident, but this may be the result of competition for binding to the same substrate. The results suggest that monoribonucleotides will, at least occasionally, be incorporated into nascent DNA strands in vivo, requiring a mechanism for removal.

RNase HI and FEN-1/RTH-1 Nuclease Can Repair Monoribonucleotides Embedded into DNA-The observation that
RNase HI is able to cleave on the 5Ј-side of embedded monoribonucleotides, together with its known role in initiator RNA removal, suggest that RNase HI is present at the sites of ribonucleotide ligation errors and should be capable of participating in a repair mechanism. We hypothesized that embedded monoribonucleotides are removed by the combined action of RNase HI and FEN-1/RTH-1 nuclease. To test this hypoth-esis, we had to start with a ligated substrate so that the RNase HI cleavage product would not simply disappear back into starting material. Fig. 4 illustrates that an embedded monoribonucleotide, in this case rA, can be removed by RNase HI and FEN-1/RTH-1 nuclease. Lane T4 is a control, which showed that the starting substrate used was completely ligated, since no change in the starting material occurred upon the addition of T4 DNA ligase. Lane 2 shows that FEN-1/RTH-1 nuclease alone also has no effect on the substrate. RNase HI must act first, preparing the substrate for FEN-1/RTH-1 action. RNase HI can cleave on the 5Ј-side of the embedded monoribonucleotide, as seen in lane 3 (see also Fig. 2B). The product of RNase HI action was shown to be the appropriate sized 35-mer by migration with respect to a snake venom phosphodiesterasegenerated standard ladder. The length was confirmed by comigration with unligated downstream primer (not shown).
Lane 4 shows that after RNase HI has prepared the substrate by cleavage on the 5Ј-side of the single ribonucleotide, FEN-1/RTH-1 can remove the ribonucleotide by cleavage on its 3Ј-side. Although the FEN-1/RTH-1 cleavage shown in lane 4 proceeded efficiently, it was difficult to achieve with certain substrates (not shown), presumably because the nuclease is Lane T4 shows that T4 DNA ligase alone has no effect on the already ligated starting substrate. Lane 1 shows starting material with no enzyme added. Lane 2 shows that FEN-1/RTH-1 alone also has no effect on starting material. Lane 3 shows that RNase HI cleaves on the 5Ј-side of the embedded monoribonucleotide to release a 35-mer. Lane 4 shows that, following RNase HI action, FEN-1/RTH-1 nuclease is able to remove the monoribonucleotide by cleavage on its 3Ј-side. Lane 5 shows that, if the upstream primer is extended, FEN-1/RTH-1 nuclease can remove the ribonucleotide by an endonucleolytic mechanism as well. Substrate structure and oligonucleotide composition are shown above the lanes. For this experiment, the monoribonucleotide was radiolabeled, and reaction conditions were as described under "Experimental Procedures." The upper band represents ligated starting material, the middle band represents RNase HI cleavage product, and the lower bands represent FEN-1/RTH-1 cleavage products. inhibited by the presence of an upstream primer following RNase HI action. We therefore extended the upstream primer with Sequenase by one or two nucleotides, unannealing the 5Ј-end of the downstream primer and allowing FEN-1/RTH-1 to act by an endonucleolytic mechanism, as shown in lane 5. For the substrate shown, endonucleolytic cleavage was no more efficient than exonucleolytic cleavage. However, the endonucleolytic mechanism enabled monoribonucleotide removal to be completed for all six ribonucleotide-containing substrates, some of which could not be processed at all exonucleolytically (not shown). Cleavage positions were altered, as seen in lane 5, so that trimers and dimers were observed as well as monomers. This alteration in specificity during endonucleolytic cleavage has been described (3) and occurred here just as expected.
After removal of the ribonucleotide by the action of FEN-1/ RTH-1 nuclease, synthesis from the upstream primer will form a nick. We have previously shown that on nicked DNA the combined action of FEN-1/RTH-1 nuclease, a DNA polymerase, and DNA ligase I results either in immediate ligation or in nick translation for several nucleotides followed by ligation (47). Consistent observations were made for the system used in this report as well (data not shown).

Highly Purified, Recombinant Human DNA Ligase I Exhibits both DNA Ligation and Junction Monoribonucleotide Ligation
Activities-We wished to confirm that ribonucleotide ligation activity is present in DNA ligase I and was not the result of some minor contaminant present in our preparation. DNA ligation and junction monoribonucleotide ligation appear to elute together from double-stranded DNA cellulose and Mono Q, the last two columns in the DNA ligase I purification procedure. However, because of the inefficiency of RNA ligation, especially when assayed from a purification fraction, we could not verify that the peaks of both activities were exactly coincident. Consequently, we took the alternate approach of assaying for both activities in highly purified recombinant DNA ligase I. The enzyme was made in a baculovirus expression system and purified by the method of Tomkinson (42,45) to greater than 90% homogeneity. A dilution series was performed so that both the ribonucleotide and DNA-DNA ligation activities of the human enzyme could be compared with those same activities in the calf enzyme. Fig. 5 demonstrates that the human DNA ligase I has both activities, since it could produce a ligated product from both a fully DNA substrate and a substrate with a downstream 5Ј-monoribonucleotide. PhosphorImager quantitation indicates that 82.8% ligation was achieved for DNA ligation in lane 10, at the same dilution producing 2.5% RNA ligation in lane 4. These are similar to the relative ligation efficiencies achieved by our purified calf DNA ligase I for these same substrates. DISCUSSION Removal of the initiator RNA of Okazaki fragments is performed by RNase HI and FEN-1/RTH-1 nuclease (2). RNase HI makes a specific cleavage between the two 3Ј-most ribonucleotides of the RNA primer, leaving a single ribonucleotide on the 5Ј-end of the downstream DNA segment. FEN-1/RTH-1 nuclease can then remove this monoribonucleotide (8). Recently, we determined that the presence of an upstream primer is sometimes neutral or even inhibitory for FEN-1/RTH-1 cleavage, especially for removal of these monoribonucleotides (9). We proposed that, if an upstream fragment is extended to form a nick that inhibits FEN-1/RTH-1 nuclease, DNA ligase I may be available and capable of sealing that 5Ј-junction monoribonucleotide into the nascent DNA. In this report, we demonstrate that DNA ligase I can indeed carry out that reaction. We also show that the combined action of RNase HI and FEN-1/ RTH-1 nuclease can remove the resultant embedded ribonucle-otide. We propose that this process is used to eliminate inappropriate ribonucleotides from the chromosome in vivo.
The substrates used in the experiments presented here have structures that should appear during Okazaki fragment processing in vivo. For example, in Fig. 1A, they simulate the intermediates expected after an upstream primer has been extended to form a nick with the 5Ј-monoribonucleotide-terminated downstream primer. Since DNA ligase I is present at the replication fork in vivo (38,39), we anticipated that it would act on this substrate. However, prior to actual testing, we could not predict ligation efficiency. We expected that the monoribonucleotide might disrupt the helical structure of a DNA duplex, such that DNA ligase I would no longer recognize and seal the substrate or would do so inefficiently. On the other hand, we also considered that the effect on helical structure of a single ribonucleotide amid long stretches of DNA might be minor. Alternatively, the mere presence of the 2Ј-hydroxyl on the ribonucleotide might have inhibited or prevented ligation. We found that purified DNA ligase I can join the primers, but much less efficiently than it carries out corresponding DNA to DNA ligations (Fig. 1).
Furthermore, the efficiency of junction monoribonucleotide ligation is quite dependent on sequence context. The two most effectively ligated substrates both contained a downstream 5Ј-rU and reached peak ligation of almost 10% in 30 min of incubation. We observed that the other three ribonucleotides, rA, rC, and rG, could also be ligated, although less efficiently. Perhaps the sequence context around a junction determines how much the monoribonucleotide disrupts the helix, which in turn affects substrate recognition and activity of the enzyme. RNA-DNA junctions occur at frequencies that suggest near random distribution on the genome (48,49). This suggests that junction monoribonucleotides will be embedded with a wide range of efficiencies.
We also considered that the ability to seal in a ribonucleotide might vary with the sequence of the upstream primer. Possibly, upstream primers that inhibit FEN-1/RTH-1 nuclease would also inhibit the ligation reaction. This would imply that certain upstream primer structures have a general inhibitory effect on enzyme action at these nicks. However, this was not the case. Removal of most monoribonucleotides by FEN-1/RTH-1 nuclease is upstream primer-inhibited, including removal of the rU on primer 8, but removal of the rU on primer 10 is upstream primer-stimulated (9). Nevertheless, ligation of both rU substrates occurred approximately equally well.
We wished to confirm that the observed ligated products contained RNA, so we subjected them to both alkaline and RNase HI digestion (Fig. 2). Alkaline hydrolysis has the advantage of cleaving on the 3Ј-side of the ribonucleotide, releasing the upstream primer with the labeled ribonucleotide attached, so that the appearance of a new band can be observed, accompanying the disappearance of the ligated product. However, the salts present during alkaline hydrolysis cause anomolous migration, which makes assignment of bands imprecise. RNase HI, like alkali, digested all ligated products of RNA-containing substrates. Since the product of DNA ligase I was always of relatively weak intensity, we also examined the product of T4 DNA ligase. RNase HI clearly digested both products, showing that both ligases incorporated RNA. Neither alkali nor RNase HI had an effect on any of the ligated products of the corresponding DNA substrates.
Several results suggest that DNA ligase I contains junction monoribonucleotide ligation activity and that it is not due to a contaminating enzyme in our preparation, such as an RNA ligase. We almost exactly followed the procedure of Tomkinson et al. (42) to obtain highly purified DNA ligase I from calf thymus. Furthermore, nearly homogenously pure human DNA ligase I from a baculovirus expression system also exhibits the ability to ligate 5Ј-monoribonucleotides (Fig. 5). Although the calf and human enzymes were purified from very different sources, the ratio of their DNA and RNA ligation activities are similar. These results support intrinsic RNA ligation capability.
To determine whether embedding of ribonucleotides is likely to occur during replication in vivo, we wished to verify that this reaction could be carried out in the presence of FEN-1/RTH-1 nuclease. Our results showed that both ligation and cleavage reactions occur when both enzymes are present (Fig. 3). The enzyme concentrations that we used in our experiments are not necessarily the physiological levels of these enzymes in vivo. However, in vivo, the enzymes involved are almost certainly assembled into a replication complex present at sites of Okazaki fragment processing. Therefore, the overall physiological concentration may not be relevant. The structure of this complex could strongly influence the relative activities of these enzymes and their order of function. It might also prevent full-length extension of the upstream primer until RNA removal has been completed. Unfortunately, simulating precisely how these enzymes could be acting in a replication complex is beyond current technology. Our current reconstitution reactions do not show evidence of timing and coordination. Instead, results demonstrate that embedding of ribonucleotides can occur at a frequency that would require an efficient repair process.
There had been prior evidence that RNase HI would cleave on the 5Ј-side of a single ribonucleotide embedded in doublestranded DNA (41,46). In fact, RNase HI has proven to be an excellent diagnostic enzyme for embedded RNA. Its effectiveness has had the added benefit of suggesting the pathway for removal of this RNA in vivo. RNase HI is highly active in the cell, induced even further during DNA synthesis (50), and an important component of mammalian lagging strand replication (24). These characteristics support a likely role in a repair process that takes place during replication.
We propose that removal of embedded ribonucleotides is achieved by the combined action of the two nucleases already known to be involved in Okazaki fragment processing, RNase HI, and FEN-1/RTH-1. We have demonstrated this cooperation in Fig. 4. These two enzymes may remove the embedded monoribonucleotide by either of two pathways in vivo, as depicted in Fig. 6. Both pathways begin with RNase HI cleavage on the 5Ј-side of the embedded monoribonucleotide. In the first pathway, FEN-1/RTH-1 can remove that ribonucleotide exonucleolytically by cleavage on its 3Ј-side before synthesis from the upstream primer. Then synthesis and nick translation would proceed normally, until DNA ligase I joins the two remaining DNA ends, forming complete double-stranded DNA as demonstrated earlier (47). This mechanism has the potential of a futile cycle, whereby the monoribonucleotide may be resealed before FEN-1/RTH-1 acts. However, once the ribonucleotide is removed, the reaction is irreversible. The second pathway envisions that displacement of the downstream primer occurs, catalyzed by a polymerase or helicase. This creates the substrate for FEN-1/RTH-1-directed endonucleolytic cleavage (2)(3)(4)(5) at the point of annealing of the displaced tail. In this case, religation of the terminal ribonucleotide would be avoided. Furthermore, efficiency of repair would be increased if displacement synthesis proceeds to a point where FEN-1/RTH-1 endonucleolytic cleavage is upstream primer-stimulated (51). Consistent with the possible requirement for displacement synthesis to stimulate FEN-1/RTH-1 cleavage, some of the tested substrates were difficult to process exonucleolytically, but all six tested substrates could be processed endonucleolytically after strand displacement using Sequenase.
When Eder et al. (41,46) first observed the ability of RNase HI to cleave on the 5Ј-side of embedded monoribonucleotides, they too hypothesized that this activity was a step in a repair mechanism. However, at that time the enzymatic process by which embedding could occur in vivo was not known. They speculated that single ribonucleotides might occasionally be incorporated into DNA during synthesis by a DNA polymerase because of the large pool of ribonucleoside triphosphates present in cells. We have proposed that DNA ligase I is a likely source of embedded monoribonucleotides.
Although monoribonucleotide ligation is inefficient com-FIG. 6. Model for repair of embedded monoribonucleotides and completion of Okazaki fragment processing. RNase HI must first cleave on the 5Ј-side of the embedded monoribonucleotide. Then, FEN-1/RTH-1 nuclease can remove it by either exonucleolytic or endonucleolytic cleavage on its 3Ј-side, before or after synthesis from the upstream primer. After removal, nick translation may occur, and the remaining DNA termini will be joined by DNA ligase I. This mechanism can produce a futile cycle if action of DNA ligase I precedes that of FEN-1/RTH-1.
pared with DNA to DNA ligation, it could still represent a major source of erroneous nucleotides in the genome. There are 30 -60 million Okazaki fragments generated per replication cycle in a human cell (52). If the upstream primer approaches the 5Ј-monoribonucleotide in only 1% of these fragments, and if only 1% of such cases are sealed, there would still be 3000 -6000 monoribonucleotides embedded every time a human cell replicates. This would clearly require an efficient repair process.
Eder et al. (41,46) could not demonstrate completion of repair because an appropriate enzyme that could cleave at the 3Ј-side of monoribonucleotides had not been identified. We show here not only that RNase HI can cut on the 5Ј-side of such embedded monoribonucleotides but that FEN-1/RTH-1 nuclease can complete removal by cleaving on the 3Ј-side. Our results suggest the enzymatic steps by which embedded monoribonucleotides are both formed and repaired in vivo.
We recently proposed a model for Okazaki fragment processing that required the arrival of an adjacent upstream primer only after FEN-1/RTH-1 cleavage of the junction ribonucleotide (9). In this report, we examine the events that would occur if lagging strand synthesis is mistimed, such that synthesis from an upstream fragment approaches to form a nicked structure inhibitory to cleavage of the 5Ј-ribonucleotide by FEN-1/RTH-1 nuclease. We previously showed that, in this situation, synthesis may continue, displacing the downstream fragment, including the monoribonucleotide, until FEN-1/RTH-1 nuclease is able to remove it endonucleolytically (9). Here we show that the presence of an adjacent upstream primer is conducive to damaging the chromosome by introduction of a monoribonucleotide. We have assigned a new or expanded activity to DNA ligase I, which we term junction monoribonucleotide ligation. We also showed that these embedded monoribonucleotides could be removed by activities of established replication proteins. Following monoribonucleotide removal, DNA replication can be appropriately completed.