Calf 5' to 3' exo/endonuclease must slide from a 5' end of the substrate to perform structure-specific cleavage.

Calf 5' to 3' exo/endonuclease, the counterpart of the human FEN-1 and yeast RTH-1 nucleases, performs structure-specific cleavage of both RNA and DNA and is implicated in Okazaki fragment processing and DNA repair. The substrate for endonuclease activity is a primer annealed to a template but with a 5' unannealed tail. The results presented here demonstrate that the nuclease must enter the 5' end of the unannealed tail and then slide to the region of hybridization where the cleavage occurs. The presence of bound protein or a primer at any point on the single-stranded tail prevents cleavage. However, biotinylation of a nucleotide at the 5' end or internal to the tail does not prevent cleavage. The sliding process is bidirectional. If the nuclease slides onto the tail, later binding of a primer to the tail traps the nuclease between the primer binding site and the cleavage site, preventing the nuclease from departing from the 5' end. A model for 5' entry, sliding, and cleavage is presented. The possible role of this unusual mechanism in Okazaki fragment processing, DNA repair, and protection of the replication fork from inappropriate endonucleolytic cleavage is presented.

A 5Ј to 3Ј exonuclease purified from calf has been shown to cooperate functionally with calf DNA polymerase ⑀ to perform nick translation (Siegal et al., 1992). In this reaction, exonucleolytic activity on a downstream primer required synthesis from an upstream primer. The exonuclease was later found to be much more active on a nicked compared with a gapped double-stranded DNA substrate . This suggests that stimulation of nuclease activity by DNA synthesis resulted from continuous generation of the favored substrate.
The calf 5Ј to 3Ј exonuclease has been implicated in processing of Okazaki fragments . A DNA template, primed with an oligoribonucleotide that had been extended at the 3Ј end with DNA, was used as a model Okazaki fragment. Calf RNase H1 removed the initiator RNA with a single cut at the 5Ј side of the ribonucleotide at the RNA-DNA junction. The 5Ј to 3Ј exonuclease then could remove the remaining ribonucleotide, creating a 5Ј DNA. Extension of an upstream primer by polymerization in the presence of DNA ligase I allowed the two primers to be joined, as expected on the lagging replication fork strand in vivo.
In addition to the above activities, the 5Ј to 3Ј exonuclease could also cleave endonucleolytically . The substrate for this reaction was a DNA primer-template, having a downstream primer with a noncomplementary, unannealed region at the 5Ј end. Additionally, an upstream primer was annealed directly adjacent to the first annealed nucleotide of the downstream primer. Endonucleolytic cleavage removed the unannealed region as an intact segment. Cleavage occurred either at the point of annealing to make a nicked substrate or one nucleotide downstream to make a 1-nucleotide gap. The 5Ј to 3Ј exonuclease domains of Escherichia coli DNA polymerase I and Thermus aquaticus (Taq) DNA polymerase (Lyamichev et al., 1993) are functionally homologous to the calf nuclease. The homologous nuclease has also been purified from murine cells and designated cca nuclease (Goulian et al., 1990) and flap endonuclease (FEN-1) (Harrington and Lieber, 1994). The corresponding HeLa cell nuclease has been purified (Ishimi et al., 1988;Waga et al., 1994;Robins et al., 1994;Murray et al., 1994), and called maturation factor I (Waga et al., 1994) and DNase IV (Robins et al., 1994). The homolog of this mammalian protein has been identified in Saccharomyces cerevisiae and named the RTH-1 nuclease (Sommers et al., 1995;Johnson et al., 1995). Knockout mutants are temperature-sensitive for DNA replication and repair of methylmethane sulfonate lesions (Sommers et al., 1995). RTH-1 nuclease was also found to be necessary for the stability of DNA repeats in yeast, suggesting that mutation of the human counterpart could impair the mismatch repair pathway that protects from colorectal cancer (Johnson et al., 1995).
In this report, we explore the substrate structure requirements for endonuclease activity of the calf nuclease. We show that the nuclease must slide over the 5Ј end of the unannealed tail of the primer and then along the single-stranded tail to the point of cleavage. The sliding action is bidirectional. The presence of a primer or protein on the tail prevents entry of the nuclease or traps a nuclease that is already on the tail. Cleavage is observed on tails with covalent adducts on nucleotides either at the 5Ј end or within the tail. A model is proposed that has implications for the biological role of this unusual nuclease.

EXPERIMENTAL PROCEDURES
Materials-Unlabeled deoxyribonucleotides and cloned E. coli singlestranded binding protein were purchased from Pharmacia Biotech Inc. Radiolabeled nucleotides (3000 mCi/mmol) were purchased from Du-Pont NEN. Biotinylated oligonucleotides were synthesized by the University of Rochester Core Nucleic Acid Laboratory. Other synthetic DNA oligonucleotides were obtained from Genosys (The Woodlands, TX). Sequenase (version 2.0) and T4 polynucleotide kinase were from U. S. Biochemical Corp. Proteinase K was obtained from Boehringer Mannheim. All other reagents were from Sigma unless otherwise noted. The calf 5Ј to 3Ј exonuclease was purified through hydroxyapatite chromatography as described previously . Final specific activity was 180,000 units/mg, with one unit defined as the amount of exonuclease required to release 1 pmol of [ 32 P]TMP from 5Ј-[ 32 P] dT 16 ⅐dA 2000 in 15 min at 37°C.
DNA Substrates-According to the manufacturer's protocol, T4 polynucleotide kinase was used to 5Ј phosphorylate the specified oligonucleotides with either ␥-32 P-radiolabeled ATP or unlabeled ATP. Oligomer sequences are listed in Table I, and primer-template substrates were constructed as described in the figure legends. Annealed downstream primers were sometimes radiolabeled by the addition of a single nucleotide at the 3Ј terminus using [␣-32 P]dCTP and 10 units of Sequenase. Substrates were 3Ј end-labeled to compare rates of 5Ј to 3Ј exonucleolytic degradation of the phosphorylated versus nonphosphorylated downstream primers and in experiments using biotinylated substrates; 5Ј-labeled substrates were used in all other experiments. Poly(dA)-5Ј[ 32 P]dT 16 was prepared as described by Ishimi et al. (1988).
Enzyme Assays-The standard nuclease reaction was described previously  and contained 60 mM BisTris, pH 7.0, 5% glycerol, 100 mg/ml bovine serum albumin, 5 mM ␤-mercaptoethanol, and 5 fmol of substrate in a volume of 25 l. Reactions were initiated by the addition of 2.25 units of nuclease and incubation at 37°C. For nuclease prebinding experiments, substrate and exonuclease were preincubated for 5 min on ice in a reaction mixture lacking Mg 2ϩ ; these reactions were initiated by the addition of MgCl 2 to 5 mM. To terminate reactions, 25 l of 90% formamide (v/v) with bromphenol blue and xylene cyanole markers was added. Electrophoresis through a denaturing urea-polyacrylamide gel was used to separate the reaction products by the method of Sambrook et al. (1989). The protocol for exonuclease assays using poly(dA)-5Ј[ 32 P]dT 16 as the substrate has been described previously (Turchi and Bambara, 1993).
Gel Shift Assays-The substrates used for gel shifts with SSB 1 are described in Fig. 3. SSB was added with 5 fmol of substrate in the reaction buffer and allowed to bind for 5 min at 4°C. To assess the strand displacement activity of SSB, samples were incubated at 37°C for a time that corresponded to the longest time point used during the endonuclease assay. Proteinase K (14.3 g) and 20 pmol of cold downstream primer were added to digest the SSB and prevent reannealing of labeled strands. Displaced strands were resolved from remaining intact substrate by nondenaturing polyacrylamide gel electrophoresis (8% gel).
Gel shifts with calf 5Ј to 3Ј exo/endonuclease were also employed. For these experiments, the initial reaction mixtures lacked Mg 2ϩ , and the amount of substrate and total volume were doubled to 10 fmol and 50 l, respectively. Each of the four different substrates in Fig. 3 were incubated with enzyme either before or after addition of primers D2, D3, or D4 or a noncomplementary primer. The nuclease was bound at 4°C for 5 min. The blocking primers, added in a 100-fold molar excess, were annealed by placing the reaction at 37°C for 2 min and then at room temperature for 10 min. Half of the reaction solution was mixed with 5 ϫ loading buffer (50% glycerol with bromphenol blue and xylene cyanole markers). These samples were analyzed by separation on a 7.5% native gel followed by autoradiography. To the remaining sample, MgCl 2 was added to 5 mM, and endonuclease activity was assayed as described above.

Activity of the 5Ј to 3Ј Exo/Endonuclease on Primers Either
Fully Annealed or with Tails-Rates of exonucleolytic versus endonucleolytic cleavage were compared on substrates with either a fully annealed downstream primer or with a 5Ј tail (Fig. 1). Measurement of these activities was also performed on substrates in which the 5Ј terminus of the downstream primer was either hydroxyl-terminated or phosphorylated. Overall, endonucleolytic cleavage of tailed primers was as much as 4-fold more efficient than exonucleolytic cleavage of fully annealed primers. Cleavage of tailed primers was slightly more rapid in the presence of a hydroxyl group at the 5Ј end versus phosphate (8.6 and 6.8% of initial substrate cleaved per min, respectively). In contrast, the 5Ј moiety present on the fully annealed downstream primer had a substantial effect on its efficiency as a substrate. The phosphorylated substrate was digested at an initial rate of 2.2% cleaved per min. However, essentially no degradation (0.17% cleaved per min) occurred when the 5Ј end was hydroxyl-terminated. Rates became nonlinear over the duration of the time course, indicating a loss of enzyme activity.
Primers Bound to the 5Ј Tail Inhibit Cleavage by 5Ј to 3Ј Exo/Endonuclease- Fig. 2A shows that the presence of a primer annealed to the 5Ј tail blocks cleavage. Harrington and Lieber (1994), working with the human enzyme FEN-1, reported this effect with a substrate having a tail that was completely double-stranded. The tail of our substrate was 25 nucleotides in length with the blocking primer annealed to the 5Ј-most 17 nucleotides of the tail, leaving a single-stranded gap of 8 nucleotides between the 3Ј end of the blocking primer and the expected site of cleavage. Thus, the blocking primer does not need to be juxtaposed to the cleavage site. Similar results were obtained when examining the endonucleolytic activities of the nuclease domains of E. coli DNA polymerase I and Taq polymerase by Lyamichev et al. (1993). The authors proposed that these bacterial enzymes must track or slide along the single-stranded tail to the point of cleavage.
To determine whether the calf nuclease slides along the 5Ј 1 The abbreviation used is: SSB, single-stranded binding protein. (20-mer) CCGGGGACTCTCTAGAGTCG B3 (21-mer) GGGCGAATTCGAGCTCGGTAC B4 (23-mer) ACGGCCAGTGAATTAACTAACAG tail, we employed a substrate with a 73-nucleotide tail. The longer single-stranded region allowed us to position primers at three distinct locations along the tail: primer B2 was annealed at the 5Ј end, primer B3 was annealed in the middle, and primer B4 was juxtaposed to the site of cleavage. Fig. 2B shows a time course experiment in which the nuclease was able to cleave off the 73-nucleotide tail in the absence of a blocking primer (lanes 1-6). When the 5Ј end of the tail was blocked with Primer B2, essentially no cleavage was observed (Fig. 2B,. In this case, the single-stranded region of the tail between the 3Ј end of the blocking primer and the site of cleavage was 53 nucleotides compared with only 8 nucleotides with the first substrate tested. It would seem that the nuclease cannot enter in the middle of the tail, even if a long single-stranded region is available. Primer B2 forms a duplex up to the very 5Ј end of the tail. Its blocking capability implies that either the nuclease must enter at the 5Ј end of the single-stranded region of the tail or must recognize the 5Ј end of the tail before it can enter at another location. Distinguishing a Sliding versus Looping Mechanism for Cleavage-An alternative to the sliding or tracking of the enzyme along a single-stranded tail is that the enzyme need only bind or associate with the 5Ј end of the tail to be activated. In the activated state, it could then bind directly to the point of cleavage and commence catalysis (see Fig. 8A). If after activation it remained bound to the 5Ј end of the tail, it could carry the 5Ј end back to the point of cleavage. The 5Ј end of the tail would then be looped back to the cleavage point just prior to release of the cleaved tail. If the cleavage mechanism involved an activation or looping process, primers B3 and B4 should not have the same inhibitory effect as primer B2. However, cleavage of substrates having these primers was inhibited (see Fig.  3A, lanes 5 and 7) just as efficiently as cleavage of the substrate with primer B2. Additionally, the presence of a primer with no complementarity to the tail (Fig. 3A, lane 1) did not stop cleavage. These results support a mechanism in which the nuclease enters at the 5Ј end and requires the entire length of the tail to be single-stranded to allow the enzyme a passage to the site of cleavage (see Fig. 8B).
Trapping the Nuclease on the Tail with Primers-The even lanes in Fig. 3A show reactions in which the substrate was preincubated with nuclease prior to the annealing of a blocking FIG. 1. Measurement of rates for exonucleolytic and endonucleolytic degradation of either 5 phosphorylated or nonphosphorylated downstream primers. Rates of exonucleolytic and endonucleolytic activity were determined using substrates pictured in the graph inset. The exonuclease substrate (I) consisted of primers U2 and D1 annealed directly adjacent to each other on template T1. See Table  I for the key to oligonucleotide sequences in this and subsequent figures. The endonuclease substrate (II) had an unannealed 5Ј tail of 12 nucleotides that formed when primer D2 was annealed downstream of U2 on T1. Downstream primers D1 and D2 were 3Ј end 32 P-labeled. Assays were performed as described under "Experimental Procedures" with 5 fmol of substrate and 2.25 units of calf 5Ј to 3Ј nuclease per reaction. In the graph, plots of exonucleolytic activity (circles) were derived from calculations of percent substrate degraded by a single nucleotide; endonucleolytic activities (triangles) were based on the amount of substrate cleaved to products 18 or 19 nucleotides long from release of the tail as a 12-or 13-nucleotide fragment. These products were resolved in a denaturing 12% polyacrylamide gel and quantitated using a Molecular Dynamics PhosphorImager and ImageQuant software. Closed symbols indicate activity on substrates having a 5Ј phosphorylated downstream primer; open symbols represent values for nonphosphorylated substrates. The 5Ј phosphate was unlabeled and introduced by polynucleotide kinase. Each time point represents the mean of three experiments; the error bars are S.E.

FIG. 2.
Blocking of endonucleolytic cleavage with primers that hybridize to the single-stranded 5 tail. Assays measuring cleavage of an endonuclease substrate (D3:U2:T1) with a 25-nucleotide 5Ј tail are depicted in A. Blocking primer B1 anneals to the 5Ј end of D3 and leaves a single-stranded gap of 8 nucleotides between itself and the point at which the tail anneals to the template. A time course with substrate lacking a blocking primer is shown in lanes 1-6; substrate with the blocking primer is in lanes 7-12. B shows endonuclease activity on a substrate with a 73-nucleotide tail (D4:U2:T2). Primer B2 hybridizes to the 5Ј tail at a position that generates a single-stranded gap of 53 nucleotides. Lanes 1-6 show a time course with substrate lacking a blocking primer; lanes 7-12 have substrate with the blocking primer. Substrate structures and oligonucleotide composition are shown above the appropriate lanes. Downstream primers D3 and D4 were 5Ј radiolabeled, and reaction conditions were as described under "Experimental Procedures." Size markers at the left indicate product length in nucleotides.
primer and the addition of Mg 2ϩ . Once again, a noncomplementary primer does not inhibit the nuclease (Fig. 3A, lane 2). By prebinding the enzyme before annealing primers B2 or B3, a small amount of cleavage product was seen. At higher concentrations of B2 or B3, the amount of product remained the same (data not shown). This suggested that the primer does not compete with enzyme for the same binding site. On the other hand, the addition of primer B4 after preincubation does not result in any product. Because primer B4 binds directly adjacent to the cleavage site, it might compete with the nuclease, preventing access to this site.
A portion of each of the above reactions was taken prior to addition of Mg 2ϩ and fractionated on a native polyacrylamide gel (Fig. 3B). The migration position of the labeled substrate is shown in lane 2 of Fig. 3B. Addition of the nuclease shifts the mobility of the substrate in the absence of a blocking primer (Fig. 3B, lane 3) or in the presence of the noncomplementary primer (Fig. 3B, lane 1). When primer B2, B3, or B4 was annealed prior to addition of the nuclease, no gel shift was observed. Gel shifts were observed, however, when the nuclease was preincubated with substrate before annealing primers B2 and B3. In fact, under these conditions, more substrate is shifted than that observed without any blocking primer. We interpret this result as a trapping of the nuclease on the substrate, with it caught on the single-stranded region between the blocking primer and the cleavage site. This result also indicates that movement of the nuclease on the tail is bidirectional, with nuclease molecules continuously sliding onto and off of the 5Ј end of the tail.
Primer B4 is not able to trap the nuclease, possibly because it does not leave any single-stranded space for the enzyme to occupy. This interpretation implies that the binding of primer B4 forces bound nucleases to move in the direction of the 5Ј end of the tail, where they fall off. Alternatively, the primer induces direct dissociation of the nuclease from the region of the tail that is complementary to primer B4. Interestingly, a 53-nucleotide region of the tail between the 5Ј end and the primer remained available to bind the nuclease after primer B4 was annealed, but no shift was observed. This suggests that the nuclease has greater affinity for the cleavage site than for various positions along the tail.
Effect of Single-stranded Binding Protein on Endonucleolytic Activity-Single-stranded DNA is generally coated with protein in vivo (Kornberg and Baker, 1992). SSB from E. coli can bind single-stranded segments as short as 4 nucleotides (Ruyechan and Wetmur, 1976) and reportedly functions to stimulate both DNA replication and repair. We therefore examined the effect that SSB has on the endonucleolytic capacity of the calf exo/endonuclease. In an experiment depicted in Fig.  4, endonucleolytic cleavage of substrates with 5Ј tails either 12 or 25 nucleotides in length was tested. Cleavage of the shorter FIG. 3. Preincubation of nuclease with substrate prior to annealing of a blocking primer permits endonucleolytic cleavage by trapping the enzyme on the 5 tail. A, endonucleolytic activity was assessed on the D4:U2:T2 substrate having one of three different blocking primers, B2, B3, or B4, hybridized respectively at distances leaving a gap of 53, 31, or 0 nucleotides from the first annealed nucleotide on the downstream primer beyond the 73-nucleotide tail. In the even numbered lanes, enzyme was preincubated with the substrate in the absence of a blocking primer (lanes 4, 6, and 8) or the presence of dT 16 , a noncomplementary primer (NC in lane 2) as described under "Experimental Procedures." No prebinding (odd lanes) indicates that the enzyme was added after the addition and annealing of the blocking primers. Products were resolved on a denaturing 10% polyacrylamide gel and visualized by autoradiography. B, native 7.5% polyacrylamide gel analysis was used to detect enzyme-substrate complexes. Preincubation and primers are indicated as in A. In lane 2, migration of substrate in the absence of enzyme (ϪE) and blocking primer is indicated by I. Preincubation with enzyme results in a mobility shift of substrate without blocking primer to position II. Substrate with blocking primer migrates to III and is shifted to IV when bound by enzyme.

FIG. 4. Effect of single-stranded binding protein on endonucleolytic cleavage.
A, analysis of endonucleolytic activity for reactions performed in the presence (even lanes) or absence (odd lanes) of 1.5 g of E. coli DNA SSB is shown. Preincubation of substrate before the addition of nuclease is detailed under "Experimental Procedures." In lanes 1 and 2, the substrate had a short 12-nucleotide 5Ј tail (D2:U2: T1), whereas substrate in lanes 3 and 4 (D3:U2:T1) had a longer 25-nucleotide tail. Separation and detection of cleavage products was performed as described in the legend to Fig. 2. Indicators for nucleotide lengths are given on both sides of the autoradiograph. B, reaction mixtures in A were sampled before the addition of enzyme and treated as described under "Experimental Procedures." The substrate in lanes 1, 2, and 3 was D2:U2:T1 with a 12-nucleotide tail; the substrate with the 25-nucleotide tail (D3:U2:T1) was in lanes 4, 5, and 6. Samples incubated without SSB were run in lanes 1 and 4. In lanes 2 and 5, substrates were incubated with SSB. Lanes 3 and 6 are aliquots from SSB-incubated substrates treated with proteinase K and excess unlabeled downstream primer. Numerals II and IIЈ indicate mobilities of intact substrates D2:U2:T1 and D3:U2:T1, respectively. DNA-SSB complexes are labeled III and IIIЈ. Unannealed D2 and D3 downstream primers have mobilities I and IЈ, respectively. tailed substrate was not affected by SSB in the reaction (Fig. 4,  lane 2), whereas a significant inhibition resulted if the substrate had a 25-nucleotide tail. These results show that protein bound to the tail can inhibit cleavage. Possibly the nuclease can compete successfully with the SSB for binding if the tail is sufficiently short.
SSB can unwind duplexed DNA through cooperative binding of additional SSB molecules. In our experiment, it was possible that SSB displaced the oligomer with the 25-nucleotide tail from the substrate, disrupting the recognition site for cleavage. We therefore sampled each of the reactions before the addition of nuclease and performed gel shift analysis on a native gel. As seen in lanes 2 and 5 of Fig. 4B, the mobility of the different substrates was shifted by SSB. When samples containing SSB were treated with proteinase K, along with a 1000-fold excess of unlabeled primer strand, the SSB was digested away, and any displaced labeled strands were unable to reanneal. Lane 3 of Fig. 4B shows that almost all of the short tailed substrate was still present as an annealed primer-template after proteinase K treatment, but less than half of the longer tailed substrate could be recovered as primer-template in lane 6 of Fig.  4B. The displacement of the tailed strand from the second substrate would decrease the concentration of substrate in the primer-template configuration and result in a smaller amount of cleavage product. Nevertheless, inhibition of cleavage was very effective and could not be accounted for solely by this strand displacement phenomenon.
Because all of the longer tailed substrate was shifted by SSB and yet some cleavage product could still be detected (Fig. 4A,  lane 4), the nuclease may either have some capacity to act on tails bound by protein or be able to displace SSB even when it is bound to the longer substrate. With the shorter tailed substrate, although a majority of the substrate was bound by SSB, we were unable to obtain a complete shift. A time course experiment measuring cleavage in both the presence and absence SSB revealed very similar rates; this also supports the conclusion that the nuclease can cleave a substrate when SSB is bound to or equilibrating with the tail.
Effects of Biotin Adducts and Streptavidin on Endonucleolytic Activity-To avoid the complication of strand displace-ment associated with SSB, we attempted to inhibit the nuclease by using substrates with primers having a biotinylated base at a selected position, either at the 5Ј end or middle of the downstream primer. Quantitative biotinylation of the substrate is evident because the biotinylation changes the gel mobility of the substrate. Streptavidin is a 60-kDa protein that recognizes and tightly binds biotin (Green and Joynson, 1970). Substrates were incubated with streptavidin in 25-fold molar excess, resulting in quantitative gel shifts of the biotinylated substrates (data not shown). This demonstrated that protein was bound to all of the biotinylated oligonucleotides. In Fig. 5, the nuclease is shown to cleave 5Ј tails with biotin adducted at either the 5Ј terminus or at the midpoint of the tail (lanes 9 -12 and 17-20, respectively) or on the upstream primer (data not shown). This demonstrates that the nuclease can traverse a nucleotide in which the mass of the base has been increased by about a factor of four by the biotin and linking moiety.
When streptavidin was allowed to bind to the biotinylated substrates prior to addition of the nuclease, no cleavage was observed with substrates that were biotinylated at either of the two downstream primer sites (Fig. 5, lanes 13-16 and 21-24). Cleavage still occurred on the substrate with a biotinylated upstream primer (data not shown). These results indicate that although an adduct larger than the base can be passed, when the adduct is protein sized, the sliding mechanism is thwarted.
Analysis of the Upstream Primer-dependent Specificity of Endonucleolytic Cleavage-Previously, we observed that a primer placed immediately upstream of the annealed portion of the downstream primer resulted in an increase in either exonuclease or endonuclease activity. Here we have attempted to determine whether the 3Ј OH group on the upstream primer is a structural feature that is essential for the stimulation to occur. Substrates were prepared having either a hydroxyl or hydrogen present on the 3Ј carbon of the upstream primer. As illustrated in Fig. 6, the lack of a hydroxyl group does not significantly reduce cleavage after the first annealed nucleotide. What we observe instead is the complete absence of the minor product, a 12-nucleotide fragment resulting from some cleavage occurring 5Ј of the first annealed nucleotide. This surprising result shows that although the 3Ј OH of the up- FIG. 5. Effects of biotin adducts and streptavidin binding on endonucleolytic activity. Time course experiments on substrates that were preincubated with or without streptavidin are shown. Substrate structure, nucleotide position of biotin substitutions, and complexes formed with streptavidin are depicted above the autoradiograph. Biotin (b) was covalently bound to the 1st and 7th nucleotides from the 5Ј end of the downstream primer. Streptavidin, in a 25-fold molar excess, was preincubated with 5 fmol of substrate for 5 min at 37°C in 1 ϫ binding buffer (12% glycerol, 12 mM HEPES-NaOH, pH 7.9, 4 mM Tris-Cl, pH 7.9, 60 mM KCl, 1 mM EDTA, and 1 mM dithiothreitol) (lanes 5-8, 13-16, and 21-24). Endonuclease reactions were then initiated by the addition of purified calf 5Ј to 3Ј exonuclease according to "Experimental Procedures." Lanes 1-8, nonbiotinylated downstream primer; lanes 9 -16, biotin-substituted primer at position 1; lanes 17-24, biotin-substituted primer at position 7. stream primer is not essential to stimulate cleavage, it influences the positioning of the cut. Stimulation of activity may lie in the ability of the enzyme to sense uninterrupted base pairing associated with nick-like structures.
Because the 3Ј hydroxyl was not responsible for stimulation dependent on nicked structure, we next examined the importance of base pairing of the upstream primer near the cleavage site. In Fig. 7,the left panel shows the structure and sequence of the substrates tested. The right panel illustrates the cleavage products generated after digestion with the calf nuclease for 20 min at 37°C. In the absence of upstream primer (Fig. 7, right, lane 2), little activity was observed. The major and minor products obtained from the substrate with a fully annealed upstream primer correspond to the positions described earlier.
When the terminal 3Ј nucleotide of the upstream primer formed a 1-nucleotide 3Ј mismatched tail, the cut was primarily before the first annealed nucleotide (Fig. 7, right, lane 4). Unlike previous results, the 1-nucleotide gap between the annealed portions of the upstream and downstream primer formed by this 3Ј mismatch caused more rather than less cleavage. If the annealed regions of the upstream and downstream primers were directly adjacent but an extra mismatched nucleotide formed a 3Ј tail on the upstream primer, the downstream primer was cleaved after its first annealed nucleotide (Fig. 7, right, lane 5). Finally, an upstream primer with a 3Ј nucleotide tail and a 1-nucleotide gap before the downstream primer generated the longest cleavage product. Though this primer was only a single nucleotide longer than the primer used in lane 5 (Fig. 7, right), the cut was positioned 2 nucleotides downstream.

DISCUSSION
The calf 5Ј to 3Ј exo/endonuclease is a member of the RTH-1 class of eukaryotic nucleases that perform structure-specific cleavage of DNA and have important roles in DNA replication and repair. The substrate for endonuclease activity is a primer-template in which the primer has a noncomplementary, unannealed 5Ј end region. Here we show that in order to catalyze cleavage, the nuclease must slide over the 5Ј end of the unannealed tail, move down the length of the tail to the point of annealing with the template, and then carry out cleavage. The sliding process is bidirectional, such that nucleases can also slide off of the tail at the 5Ј end. Evidence is also presented that the nuclease binds with highest stability to the point of cleavage. The presence of biotin adducts on the tail do not affect either recognition of the 5Ј end or the sliding process. However, protein bound to the tail, either SSB or streptavidin, has an inhibitory effect on cleavage, presumably by blocking the movement of the nuclease to the site of cleavage. The steps taken by the nuclease to carry out this process are depicted in Fig. 8.
Biochemical and genetic evidence suggests that nucleases of this class are components of the DNA replication machinery. Studies of simian virus 40 DNA replication in vitro indicated the requirement of the human 5Ј to 3Ј exo/endonuclease in the removal of the RNA primer before joining of viral lagging strand DNA segments (Ishimi et al., 1988;Waga et al., 1994). The mouse nuclease was able to partially remove RNA from initiator RNA primers made and elongated with DNA by DNA polymerase ␣ in vitro (Goulian et al., 1990). Using a model Okazaki fragment substrate consisting of an initiator RNA elongated by DNA and annealed to a template, we showed that two nucleases can effect complete RNA removal . Mammalian RNase HI cleaves off the initiator RNA as an intact segment, leaving only a single ribonucleotide at the RNA-DNA junction Huang et al., 1994). The 5Ј to 3Ј exo/endonuclease can then remove the last ribonucleotide. Null mutants of the counterpart nuclease in S. cerevisiae, the RTH-1 nuclease, are temperature-sensitive for DNA replication (Sommers et al., 1995). The appearance of dumbbell-shaped cells with a nucleus at the isthmus as the terminal phenotype is characteristic of mutations in DNA polymerases ␦ FIG. 6. Role of the upstream primer 3 hydroxyl group in endonucleolytic cleavage. Substrates were prepared with both a 5Ј end-radiolabeled downstream primer (D2) and an upstream primer (either U1 or U2). Primer U1 was extended with a single dideoxycytosine triphosphate to generate a 3Ј end that terminated in a hydrogen instead of a hydroxyl group. In the presence of 100 mM dNTPs, Sequenase could not further extend U1 on T1 (lane 15). Synthesis occurred to the end of the template when U2 was the upstream primer (lane 16). Lanes 1-7 show the time course of endonuclease activity on substrate having a 3Ј hydrogen on the upstream primer. In lanes 8 -14, time course reactions contained substrate with U2 that had the 3Ј hydroxyl group.
FIG. 7. Effects of upstream primers with various 3 structures on endonucleolytic cleavage site specificity. The left panel illustrates the structures of the substrates assayed in the right panel. For all substrates, downstream primer D5 was 5Ј radiolabeled and formed a 10-nucleotide 5Ј tail when annealed to template T2. Upstream primers annealed either directly adjacent to the annealed region of the downstream primer (U2 and U4) or one nucleotide away (U3 an U5). Additional structural features are as follows: ϪU, no upstream primer; U2, fully annealed primer; U3 and U4, 1-nucleotide 3Ј tail; U5, 3-nucleotide 3Ј tail. The right panel is an autoradiograph of reactions with the various substrates assayed for endonuclease activity and resolved on a denaturing 12% polyacrylamide gel. Reaction time was 20 min at 37°C. Lanes 1 and 3 show reactions with substrate U2. Lane 1 contained no enzyme as a control. Substrates with ϪU, U3, U4, and U5 were assayed in lanes 2, 4, 5, and 6, respectively. In lanes 2-6, reactions were carried out with 2.25 units of enzyme. and ⑀ (reviewed in Bambara and Huang (1995)). The RTH-1 mutation causes hyper-recombination, symptomatic of the presence of long-lived breaks in the chromosome (Sommers et al., 1995). These would be expected if joining of Okazaki fragments was delayed by the mutation.
The sensitivity of the RTH-1 mutants to damage by the agent methylmethane sulfonate (Sommers et al., 1995), which adducts to bases, suggests that the RTH-1 class of nucleases participate in repair of base damage. The demonstration that null mutants in RTH-1 increase the instability of simple repetitive DNA indicates a role for this nuclease in the mismatch repair pathway (Johnson et al., 1995). Our results show that biotinylation of nucleotides on a tail do not inhibit the movement of the nuclease to its site of cleavage. This suggests that in vivo the nuclease can traverse an unannealed, damaged segment of DNA and cut it away from the chromosome. Because the nuclease does not have to cut at the damaged nucleotide to accomplish this process, it should be able to remove even modified nucleotides that would not be substrates for cleavage attempted at the modified site.
Our results show that the nuclease cannot enter the tail strand except at the 5Ј end. Because the RTH-1 class nucleases are expected to be present at the replication fork, this property would be expected to protect the lagging strand template from unintentional cleavage. There is a risk of such cleavage because the extension of the leading strand primer continuously generates a structure with most of the characteristics of the substrate for RTH-1 endonuclease activity, at the point where the parental strands separate. In this structure, the lagging strand template would be the equivalent of the single-stranded tail in our model substrates, except that the lagging strand template does not have a nearby 5Ј end. The absence of this single feature prevents endonucleolytic cleavage.
Properties of the calf 5Ј to 3Ј exo/endonuclease suggest another means by which it can participate in the removal of the initiator RNA segments of Okazaki fragments. It could remove this RNA by endonucleolytic cleavage of the Okazaki fragment in the region of DNA just downstream of the initiator RNA. DNA polymerase ⑀ has been proposed to be responsible for completion of the extension of Okazaki fragments in vivo (reviewed in Bambara and Huang (1995)). This polymerase partially copurifies with DNA helicase E (Siegal et al., 1992). The Ku autoantigen is also a helicase that copurifies with human DNA polymerase ⑀ (Hwang et al., 1995), suggesting that helicase E is the calf Ku protein. Helicase E moves in a 3Ј to 5Ј direction on the template and so can displace primers ahead of the advancing polymerase. It is possible that the polymerase, helicase, and RTH-1 class nuclease collaborate in a pathway for removal of initiator RNAs of Okazaki fragments.
In summary, the calf 5Ј to 3Ј exo/endonuclease can cleave off the 5Ј unannealed tail of a primer annealed to a template by cutting near the point of complementarity and releasing the tail as an intact segment. Here we show that to catalyze this reaction, the nuclease must enter the tail at the 5Ј end and slide to the point of cleavage. This unique mechanism of action is consistent with a proposed role for the nuclease in removal of the initiator primers from Okazaki fragments during DNA replication and in cleaving off segments of damaged DNA during DNA repair. The yeast counterpart nuclease RTH-1 has been shown to be necessary to maintain stability of repetitive DNA and apparently is involved in the MSH2-MLH1-PMS1 mismatch repair pathway (Johnson et al., 1995). Mutations in the mammalian mismatch repair pathway are associated with colorectal cancers (Reenan and Kolodner, 1992;Prolla et al., 1994). This suggests that the human RTH-1 counterpart applies its unique cleavage specificity for repair surveillance that prevents colorectal cancer. FIG. 8. Models of nuclease binding to tailed substrates. A, looping model. The enzyme binds at the cleavage site but must also interact with the 5Ј terminus of the single-stranded tail for cleavage to occur. B, threading model. The enzyme must bind and enter at the 5Ј terminus of the single-stranded tail (I). Movement of the enzyme on the tail is bidirectional (II). Cleavage can occur when the enzyme slides down to the point at which the tail is annealed to a template (III).