Newly Discovered Archaebacterial Flap Endonucleases Show a Structure-Specific Mechanism for DNA Substrate Binding and Catalysis Resembling Human Flap Endonuclease-1*

Mammalian flap endonuclease-1 (FEN-1) is a structure-specific metalloenzyme that acts in processing of both the Okazaki fragments during lagging strand DNA synthesis and flap intermediates during DNA damage repair. We identified and cloned three open reading frames encoding a flap endonuclease fromArchaeglobus fulgidus, Methanococcus jannaschii, and Pyrococcus furiosus, respectively. The deduced FEN-1 protein sequences share approximately 75% similarity with the human FEN-1 nuclease in the conserved nuclease domains, and extensive biochemical experiments indicate that the substrate specificities and catalytic activities of these enzymes have overall similarities with those of the human enzyme. Thus, FEN-1 enzymes and likely reaction mechanisms are conserved across the eukaryotic and archaeal kingdoms. Detailed comparative analysis, however, reveals subtle differences among these four enzymes including distinctive substrate specificity, tolerance of the archaebacterial enzymes for acidic pHs and elevated temperatures, and variations in the metal-ion dependence of substrate cleavage. Although the archaebacterial enzymes were inactive at temperatures below 30 °C, DNA binding occurred at temperatures as low as 4 °C and with or without metal ions. Thus, these archaeal enzymes may provide a means to dissect the specific binding and catalytic mechanisms of the entire FEN-1 family of structure-specific nucleases.

DNA replication, recombination, and repair, which are essential for maintaining human genome stability, all depend on the activity of enzymes possessing 5Ј3 3Ј exonuclease activity. In repair, this activity is essential for damaged fragment excision and recombinational mismatch correction, whereas in replication this activity plays a critical role in the efficient processing of Okazaki fragments during lagging strand DNA synthesis (1)(2)(3)(4)(5)(6). In eukaryotic organisms this activity is provided by FEN-1 1 (FiveЈ ExoNuclease or Flap EndoNuclease), a 40-kDa Mg 2ϩ -dependent metallonuclease that recognizes and cleaves branched nucleic acid structures containing a singlestranded 5Ј flap (the junction where the two strands of duplex DNA adjoin a single-stranded arm) regardless of sequence and without the need for accessory proteins (7)(8)(9)(10)(11)(12)(13). The enzyme is unable to cleave bubble substrates, 3Ј single-stranded flaps, heterologous loops, and Holliday junctions; however, acting as an exonuclease, FEN-1 will affect hydrolysis of doublestranded DNA substrates containing a nick, gap, or a 3Ј overhang, albeit at reduced efficiency (9,14). The endonuclease activity of the enzyme is independent of the 5Ј flap length, cleaving a 5Ј flap as small as one nucleotide, whereas the endonuclease and exonuclease activities are insensitive to the chemical nature of the substrate, cleaving both DNA and RNA (15). Although FEN-1 is able to cleave a pseudo-Y structure (see Table I, substrate 3), the catalytic efficiency of both the endo-and exonuclease activities is significantly stimulated by the presence of an upstream primer (9,10,15,16). Recently, FEN-1 was proposed to act in the removal of apurinic/apyrimidinic sites via participation in an alternative base-excision repair pathway involving DNA polymerase, HAP-1, DNA ligase, and the polymerase processivity factor, PCNA (17). The polymerase acting in this pathway was originally postulated to be polymerase ␤; however, recent evidence suggests either pol ␦ or pol ⑀ provides this activity (18). The involvement of the processivity factor in the pathway likely reflects an interaction between PCNA and FEN-1 as complementary biochemical and genetic experiments have convincingly demonstrated that FEN-1 activity is stimulated by specific interactions between these two proteins (19 -23).
Most of the in vivo functional analyses of the FEN-1 gene have been carried out in the yeast Saccharomyces cerevisiae and Schizosaccharomyces pombe. The S. cerevisiae FEN-1 homologue is named RAD27 (or RTH1, YKL510), whereas in S. pombe it is named rad2 (24 -27). Mutations in RAD27 result in marked sensitivity to alkylating agents, modest sensitivity to ultraviolet radiation, and chromosome instability. Studies in S. pombe further indicate that rad2 is likely critical for the efficient removal of UV-induced photo-dimers, and deletion mutants display a temperature-sensitive growth defect and accumulate in S phase, apparently owing to a block in DNA replication (25)(26)(27). Simple repetitive DNA stability also requires FEN-1, as a RAD27 null mutation in S. cerevisiae leads to an increase in the mutation rate of simple repetitive DNA by as much as 280-fold. Epistasis analysis suggests this is consistent with RAD27 playing a role in the MSH2-MSH6-MLH1-PMS1 post-replication mismatch repair pathway, which is implicated as a causative agent of colorectal cancer (28). However, Kolodner's group has shown that the majority of these mutations are due to sequence duplications flanked by short repeats and suggest that RAD27 is critical for processing replication intermediates. In the absence of RAD27, replication intermediates were processed via a Rad51-and Rad52-dependent double-strand break repair mechanism and, to a lesser extent, by a mutagenic repair pathway leading to the observed duplications (29). In addition, RAD27 has an active role in preventing trinucleotide repeat expansion and contraction, as the same rad27 mutant showed evidence of length-dependent destabilization of CTG tracts and a marked increase in their expansion frequency. These observations suggest that FEN-1 mutants in humans may lead to genetic diseases such as myotonic dystrophy, Huntington's disease, several ataxias, fragile X syndrome, and a number of cancers, all of which result from trinucleotide repeat expansion and sequence duplications (29 -31).
The FEN-1 enzymes are functionally related to several smaller bacteriophage 5Ј33Ј exonucleases such as T5 5Ј exonuclease and T4 RNase H (32)(33)(34) as well as to the larger eukaryotic nucleotide excision repair enzymes such as XPG, which also acts in the transcription-coupled repair of oxidative base damage (10,35). In eubacteria such as Escherichia coli and Thermus aquaticus, Okazaki processing is provided by the DNA polymerase I (PolI) 5Ј33Ј exonuclease domain (2,36,37). These bacterial and phage enzymes share two areas of limited sequence homology with FEN-1, which are termed the N (Nterminal) and I (intermediate) regions, with the residue similarities concentrated around seven conserved acidic residues (14, 38 -40). Based on crystal structures of T4 RNase H and T5 exonuclease (33,34) and mutagenesis data (38 -40), these residues have been proposed to bind two Mg 2ϩ ions that are required for affecting DNA hydrolysis; however, the role each metal plays in the catalytic cycle, which is subtly different for each enzyme, is not well understood.
Although the sequence similarity between FEN-1 and prokaryotic and viral 5Ј33Ј exonucleases is low, FEN-1s within the eukaryotic kingdom are highly conserved at the amino acid level, with the human and S. cerevisiae proteins being 60% identical and 78% similar. Recently, we surveyed the genomes of three archaebacteria, Archaeglobus fulgidus, Methanococcus jannaschii, and Pyrococcus furiosus, and found that unlike all viral and eubacterial systems so far examined, these extreme thermophiles encode proteins that are highly homologous to the eukaryotic FEN-1 enzymes (41). Here, we report the cloning, overexpression, and purification of all three enzymes. Biochemical analyses of these archaebacterial enzymes allowed comparison of the properties of these nucleases with their eukaryotic counterparts, and the results identify both key functional similarities and informative distinctions between the two groups. The common biochemical properties of the archaebacterial and human FEN-1 enzymes demonstrate that these two groups of enzymes likely employ similar catalytic mechanisms. This conservation of FEN-1 enzymes between the eukaryotic and archaeal kingdoms supports the fundamental role of FEN-1 in cell biology and provides the basis for detailed comparative studies of these FEN-1 enzymes to define the structural chemistry underlying their structure-specific nuclease activities.

EXPERIMENTAL PROCEDURES
Materials-Restriction enzymes and bovine serum albumin were obtained from New England Biolabs, and chromatography materials were obtained from Amersham Pharmacia Biotech. Protein concentrations were determined using either the calculated ⑀ 280 value or with the Bio-Rad protein assay kit (Bio-Rad). All buffers, salts, and reagents were of the highest available purity and were purchased from Fisher Scientific. Oligonucleotides for polymerase chain reaction (PCR) were purchased from Life Technologies, Inc., and those used in the biochemical assays were synthesized on an Applied Biosystems, Inc. DNA synthesizer in the City of Hope National Cancer Center core facility.
Cloning of Archaebacterial FEN-1s-Archaebacterial FEN-1 homologues were identified with the NCBI BLAST search engine (http:// www.ncbi.nlm.nih.gov/BLAST/) using the human FEN-1 amino acid sequence. The M. jannaschii gene (mFEN-1), which was used as a template for PCR, was obtained from ATCC (construct AMJIM93), whereas the P. furiosus and A. fulgidus genes (pFEN-1 and aFEN-1, respectively) were cloned via PCR using genomic DNA generously supplied by F. Robb (University of Maryland) and R. Robson (University of Reading), respectively. For all three genes, the PCR primers were designed such that an NdeI site was introduced at the translation start site and a BamHI site was introduced in the 3Ј-untranslated region. These PCR products were digested with NdeI and BamHI, and the resulting product was ligated into pET-15B (Novagen), which incorporates a 6 ϫ histidine tail at the N terminus of the protein. In each case, the DNA sequence was verified at the Scripps Research Institute core sequencing facility.
FEN-1 Overexpression and Purification-All of the plasmids containing the archaebacterial FEN-1 genes were transformed into the E. coli strain BL21(DE3) (Novagen), and protein overexpression was induced in log phase cells by adding isopropylthiogalactopyranoside to a final concentration of 0.6 mM. Following growth for an additional 6 h, the cells were pelleted, taken up in buffer H1 (10 mM Tris, pH 7.5, 150 mM NaCl, 10 mM imidazole), sonicated briefly to resuspend the cells, lysed by heating the suspension at 75°C for 45 min, and then cooled rapidly to 0°C in an ice water bath. This protocol, which takes advantage of the archaebacterial protein's thermostability, not only lysed the cells but also precipitated the majority of the contaminating mesophilic proteins. The resulting solution was centrifuged at 25,000 ϫ g, and the supernatant was loaded onto a 25-ml Ni-NTA-Sepharose column pre-equilibrated in buffer H1. After washing with 200 ml of H1, the FEN-1 proteins were eluted with H1 containing 500 mM imidazole and then extensively dialyzed against buffer F1 (10 mM Tris, pH 7.5, 150 mM NaCl, 1 mM EDTA). Although these two steps provided protein that was approximately 95% pure, the protein was further purified by passage over a Superdex-75 fast protein liquid chromatography column.
DNA Binding Assays-Binding assays were performed using the method of Harrington and Lieber (42) with some modifications. Briefly, the indicated amounts of FEN-1 were mixed with 10 fmol of labeled DNA in a final volume of 13 l containing 50 mM Tris (pH 8.0), 10 mM NaCl, 5 mM EDTA, 10% glycerol, and 50 g/ml bovine serum albumin. After a 5-min incubation period on ice, each binding reaction was loaded onto a 5% polyacrylamide gel containing 0.5 ϫ TBE and electrophoresed for 90 min at 125V at 4°C; the dried gels were then exposed to Kodak film for imaging.
Endonuclease Assays-Endonuclease assays were generally performed using substrates 2, 3, and 8 according to the method of Harrington and Lieber (9). In a final volume of 13 l, varying amounts of FEN-1 and 1.54 pmol of labeled DNA were incubated at different temperatures for 30 min before the reaction was quenched with an equal volume of stop solution (10 mM EDTA, 95% deionized formamide, and 0.008% bromophenol blue and xylene cyanol). Reactions were electrophoresed through denaturing 15% polyacrylamide gels, and the relative amounts of starting material and product were quantitated using the IPLabGel system (Stratagene) running MacBAS image analysis software. Most reactions were performed in standard assay buffer (10 mM Tris (pH 8.0), 10 mM MgCl 2 , and 50 g/ml bovine serum albumin); however, in a series of experiments the effect of different divalent metals and pH levels was studied by varying the standard buffer. For divalent metals, MgCl 2 was omitted, and different metal ions were used at a final concentration of 10 mM. To study the influence of pH, buffers containing differing amounts of Tris, glycine, and sodium acetate were used at a final concentration of 10 mM to obtain a wide range of pH levels at 25°C as described previously (9).
Exonuclease Assays-Exonuclease activities were assayed using a nicked duplex (substrate 7) under conditions identical to those described for the endonuclease assays. The precise positions of DNA cleavage in both the exonuclease and endonuclease experiments were obtained by partial digestion of the 5Ј 32 P-labeled template strand (oligo 3) using the 3Ј-5Ј exonuclease activity of Klenow fragment (New England BioLabs).
Kinetics-FEN-1 cleavage kinetics were performed under standard conditions as described in the endonuclease assay by using varying concentrations of DNA substrates and constant amounts of FEN-1 (hFEN-1, 30 ng; aFEN-1, 200 ng; mFEN-1, 300 ng; and pFEN-1, 400 ng). For the hFEN-1 enzyme reaction kinetics were performed at 30°C, whereas the archaebacterial enzymes were assayed at 55°C, the highest temperature at which our standard substrate could be used. Reactions were initiated by combining standard reaction buffer, substrate, and enzyme in that order. Samples were mixed and incubated for 10 min, and the initial velocity was calculated by measuring products and substrate intensity on the gel by using IPLab Gel and converted using the , where t ϭ time in seconds, I 1 ϭ product intensity, I 0 ϭ final substrate concentration, and substrate concentration is expressed in nM. This calculation modifies the original equation established by Petruska et al. (43) so that velocity is expressed as converted substrate concentration/time as opposed to converted percent substrate/time. These velocity units are more appropriate for our kinetic analysis. V max and K m values were calculated by directly fitting the data to Michaelis-Menten equation using Quickgraph software (Quickgraph, CA).

RESULTS
Cloning and Purification of Archaebacterial FEN-1 Proteins-By using data provided by three microbial genome-sequencing projects (44 -46), three putative open reading frames were identified that shared a significant degree of homology to the mammalian FEN-1 family of structure-specific endonucleases (Fig. 1). Multisequence automated alignment established sequence similarity and conservation of all key residues implicated by mutagenesis in substrate binding and catalysis (37). As a first step toward the characterization of the enzymatic properties of these archaebacterial nucleases, a PCR-based approach was used to clone the genes from the thermophilic organisms P. furiosus, M. jannaschii, and A. fulgidus, such that the feasibility of protein overexpression in E. coli could be determined. Overexpression of the archaebacterial nucleases, in the T7-polymerase containing bacterium BL21(DE3), typi- cally led to FEN-1 being produced at approximately 5% of total cellular protein. Each protein was purified to near homogeneity using methods that first exploited the enhanced thermal stability of the archaebacterial proteins. Thus, before subjecting each protein to Ni 2ϩ affinity and conventional gel filtration chromatography, the solubilized cell extract was heated at 75°C for 45 min to precipitate mesophilic proteins that were removed by centrifugation. For all three archaebacterial FEN-1 proteins, these procedures yielded protein that was greater than 95% pure (Fig. 2). However, mFEN-1 always copurified with small amounts of ϳ30-kDa and ϳ10-kDa fragments. Nterminal sequencing of these fragments revealed that they resulted from proteolytic cleavage of the enzyme between Arg 91 and Lys 92 and suggests that this region may be devoid of any stable secondary structure. Consistent with results obtained for hFEN-1, gel permeation chromatography through a fast protein liquid chromatography Sepharose S-100 column indicated that each archaeal protein behaved as a monomer at concentrations near 100 M.
FEN-1 Binding to Several DNA Substrates-The structurespecific cleavage of mammalian flap endonucleases to several DNA substrates has recently been reviewed by Lieber (16), and among the more than 14 substrates tested, only flap structures, pseudo-Y structures, gaps, nicks, and 3Ј overhangs were substrates for the enzyme. To test the substrate binding preferences of the archaebacterial FEN-1s, we employed seven different substrates (listed as 1-7 in Table I) and compared the results with hFEN-1. After incubating different amounts of each FEN-1 enzyme with the various DNA substrates in the absence of Mg 2ϩ , binding was visualized on a non-denaturing polyacrylamide gel. All three archaebacterial enzymes had binding specificities that are similar to the human enzyme (Fig.  3). Thus, all the archaebacterial FEN-1s bind to flap, pseudo Y, 3Ј overhang, and nicked DNA structures with affinities closely matching hFEN-1. Further, like the human enzyme, the archaebacterial FEN-1s bound weakly to 5Ј overhangs and showed no apparent affinity toward either single-stranded or duplex DNA.
Activity of Archaebacterial FEN-1 Enzymes Requires Elevated Temperatures-The archaebacteria from which the FEN-1 homologues were obtained are thermophilic and thus grow optimally at elevated temperatures. M. jannaschii and A. fulgidus, being moderately thermophilic, grow optimally at 83 and 85°C, respectively, whereas P. furiosus, which is an extreme thermophile, has an optimal growth temperature that is greater than 95°C (44 -46). Given these optimal growth temperatures, enzymes isolated from these organisms often show activity profiles that peak at much higher temperatures than enzymes isolated from mesophilic sources. To assay both the endonucleolytic and exonucleolytic properties of these enzymes, we conducted assays using temperatures only as high as 55°C, as higher temperatures denatured our doublestranded substrates. Whereas hFEN-1 has optimal endonuclease activity near 30°C, all three archaebacterial enzymes only begin to show this activity at temperatures above 35°C (Fig. 4). At 45°C, when the human enzyme is almost entirely inactive, both aFEN-1 and mFEN-1 show a large increase in activity,  q specifies that respective substrate was used in the indicated assay. * 5Ј-labeled end. ** Boxed area indicates two G-C pairs vs. two A-T pairs in the same location in flap-1.  Table I as 1-7 were subjected to non-denaturing gel electrophoresis in the absence of FEN-1 and illustrated at left. Binding of human FEN-1 and A. fulgidus FEN-1 to each of the seven DNA substrates is shown at right. 200 ng of hFEN-1 and 500 ng of aFEN-1 were used in each binding assay, and care was taken to ensure no Mg 2ϩ or Mn 2ϩ was present. Similar results were obtained for pFEN-1 and mFEN-1 (data not shown). reaching a level that is roughly 10% of hFEN-1 enzyme's optimal activity, whereas pFEN-1 requires even higher temperatures for activity, reaching approximately 5% of the hFEN-1 optimal activity at 55°C. The products of the endonuclease reaction varied from enzyme to enzyme. Using flap substrate-1 and as illustrated in Fig. 4, both hFEN-1 and aFEN-1 gave roughly equal amounts of cleavage products 19 and 21 nucleotides in length, whereas mFEN-1 and pFEN-1 primarily gave a product either 21 or 19 nucleotides long, respectively.
The temperature dependence of the exonuclease activity of the archaebacterial FEN-1 enzymes closely paralleled that observed for the endonuclease activity with the exception that both aFEN-1 and mFEN-1 are active at 30°C. Although the initial exonucleolytic product of both hFEN-1 and mFEN-1 are mononucleotides, dinucleotides, and trinucleotides, both aFEN-1 and pFEN-1 only digest mononucleotides (Fig. 5).
Metal-Ion Dependence of Archaebacterial FEN-1 Enzymes-As all the FEN-1 enzymes characterized to date show an absolute requirement for divalent metal ions, we have characterized the ability of Mg 2ϩ , Mn 2ϩ , Cu 2ϩ , Co 2ϩ , Ni 2ϩ , Ca 2ϩ , Fe 2ϩ , and Zn 2ϩ to activate the archaebacterial enzymes. As with the human enzyme, endonucleolytic cleavage of flap substrate-1 was only supported by Mg 2ϩ and Mn 2ϩ ; however, the cleavage site preferences for each enzyme was altered in the presence of Mn 2ϩ . Instead of giving an equal mixture of 19-and 21-nt-long products by aFEN-1, all other three enzymes showed a preference for a 19-nt product when Mn 2ϩ is present (Fig. 6).
Archaebacterial Enzymes Tolerate Acidic pH Values-The endonucleolytic activity of the archaebacterial and human FEN-1 enzymes was assayed as a function of pH, over the range 4.5-10.0. For hFEN-1, the reaction was performed at 30°C, whereas the archaebacterial enzymes were assayed at 55°C. The archaebacterial enzymes are more tolerant of acidic pH by approximately one pH unit (Table II). In addition, the ratio of cleavage products (19 versus 21 nucleotides) differed depending on the enzyme assayed. This ratio showed very little pH dependence for mFEN-1 and aFEN-1, whereas for hFEN-1 the amount of the 21-nt product decreased with increasing pH, and for pFEN-1 it increased (data not shown).
DNA Substrate Sequence and Upstream Primer Influence on FEN-1 Activity-The effects of DNA substrate sequence and upstream primer on the activities and cleavage sites of murine  Table  I) Table  I) was prepared as described under "Experimental Procedures." The reaction was carried out in the same conditions as in Fig. 3.   FEN-1 were documented by Harrington and Lieber (9). To examine these effects in the human and archaebacterial systems, endonuclease assays were undertaken using a pseudo-Y and a flap substrate that was GC rich at the junction of doublestranded and single-stranded DNA. At 30°C, hFEN-1 cleaved both substrates with the pseudo-Y substrate giving 21-nucleotide-long products, and the GC-rich flap structure giving primarily 20 nucleotide products. The archaebacterial enzymes showed somewhat different reaction products at 55°C. Although mFEN-1 was able to cleave the GC-rich flap substrate, giving an equal mixture of 19-and 20-nt-long products, both aFEN-1 and pFEN-1 showed very little activity toward this molecule. All three enzymes cleaved the pseudo-Y substrate, giving reaction products primarily 21 nucleotides in length; however, as was found for hFEN-1, the activity of all three archaebacterial enzymes were reduced (Fig. 7). This illustrates the importance of the primer strand as a stimulator of FEN-1 catalytic activity. FEN-1 Cleavage Kinetics-We have determined the kinetic constants for endonucleolytic cleavage by the archaebacterial FEN-1 enzymes at 55°C using our kinetic assay and compared them with results obtained for the human enzyme at 30°C (Fig. 8). The V max for the human enzyme was roughly 16-, 5-, and 36-fold greater than those obtained for aFEN-1, mFEN-1, and pFEN-1, respectively, regardless of whether or not cleavage led to a 19-or 21-nucleotide product (Table III). The K m values for the archaebacterial enzymes are similar to or slightly lower than those obtained for the human enzyme, confirming that the archaebacterial enzymes have a high affinity for the flap junction even at suboptimal temperatures. Interestingly, and in contrast to the results for the other three enzymes, the V max values for pFEN-1 vary significantly as a function of the cleavage product. V max for cleavage giving a 19-mer is 3-fold higher than that for the cleavage leading to a 21-mer. This result is consistent with the cleavage site preference of this enzyme, which was also observed in the previous experiments. The cleavage site preferences of the various FEN-1 enzymes likely arise from subtle, yet unique structural differences at the active site of each nuclease.

DISCUSSION
Starting from putative FEN-1 gene sequences identified in sequence data bases, we have cloned, overexpressed, purified, and biochemically characterized the FEN-1 enzymes from A. fulgidus, M. jannaschii, and P. furiosus. These enzymes are strikingly similar to the eukaryotic members of the FEN-1 family and, as is true for the eukaryotic enzymes, exist as independent proteins (41). This finding is consistent with the observation that most archaebacterial DNA repair and replication proteins are closely related to their eukaryotic counterparts (44 -46) and explains the limited sequence homology between the archaebacterial FEN-1s and the prokaryotic and viral 5Ј-3Ј exonucleases. Importantly, key residues in hFEN-1 that are critical for substrate binding and catalysis, as implicated by site-directed mutagenesis (38 -40), are absolutely conserved in all three archaebacterial enzymes (Fig. 1). Moreover, the C-terminal domain, which is absent in the eubacterial and viral endonucleases and has been implicated in mediating the interaction between FEN-1 and PCNA (21)(22)(23), is present in the archaebacterial enzymes. In eukaryotes, the FEN-1⅐PCNA complex is likely to be relevant in vivo for the processing of Okazaki fragments. PCNA, which is known to play a key role in eukaryotic replication, likely recruits FEN-1 to branched nucleic acid structures near the replication fork, thereby stimulating its activity 10 -50-fold (19,20). Recently, the role of FEN-1 in an alternate long-patch base-excision repair pathway that removes apurinic/apyrimidinic sites was demonstrated, and the efficiency of the pathway was improved if PCNA was present (17,18), again demonstrating the importance of the FEN-1-PCNA interaction for mediating critical transactions with nucleic acids. The role of the FEN-1-PCNA interaction in archaebacteria is unknown; however, putative PCNA homologues exist in each of the organisms discussed here (44,45), 2 suggesting that this crucial interaction is conserved in archaebacteria and likely plays a role in DNA replication and repair pathways of these organisms.
As expected from the optimal growth temperature of A. fulgidus, M. jannaschii, and P. furiosus, the enzymatic activity of the archaebacterial enzymes was considerably higher at 55°C than at 30°C. Several theories have been proposed to account for the differences in thermal stabilities of thermophilic and mesophilic enzymes (47,48). One hypothesis suggests that thermophilic proteins can be stabilized relative to their mesophilic counterparts through an increase in the number of salt bridges. Analysis of the amino acid sequence of the human and archaebacterial FEN-1 enzymes suggests that salt bridges may play a role in stabilizing these thermophilic enzymes, as the number of charged residues increases as the thermophilicity of the organism from which the enzyme is isolated also increases. Thus, the mesophilic hFEN-1 contains a total of 100 charged residues, whereas the extremely thermophilic pFEN-1 enzyme has 113. Intermediate to these two extremes are the M. jannaschii and A. fulgidus enzymes, which possess 108 and 106 charged residues, respectively. This analysis of the number of charged residues for the human enzyme excludes the charged region C-terminal to the PCNA binding domain (Fig. 1), which is absent in the archaebacterial enzymes and likely functions as a nuclear localization signal (10, 49 -51).
Although the catalytic activity of the archaebacterial enzymes was severely diminished at temperatures below 30°C, all three enzymes were able to interact with flaps, pseudo-Ys, and 3Ј overhangs, in the absence or presence of divalent metal, at temperatures as low as 0°C. Although additional experi-  Table I  ments will address the change in binding affinity as a function of temperature, these results demonstrate that the decreased activity of the archaebacterial FEN-1 enzymes at temperatures below 30°C does not result from a loss of their ability to bind substrate. Instead, as the temperature is raised, the enzyme likely attains enough flexibility to accommodate the subsequent hydrolysis of the DNA backbone. Small structural changes in the enzyme that arise from the increased temperature may be required to bring the required catalytic residues into a conformation necessary for substrate hydrolysis. Additionally, the increased temperature may affect the inherent dynamic properties of the enzymes, allowing sufficient mobility for catalysis. In parallel to their resistance to heat lability, the enzymes from these thermophilic organisms are also more tolerant of acidic pH levels (Table II) than are their eukaryotic counterparts. Perhaps, a conserved metal binding residue(s) in the archaebacterial enzyme has a reduced pK a relative to the human enzyme permitting efficient metal binding, and thus catalysis, at reduced pH levels. Structural and biochemical studies of both mesophilic and thermophilic FEN-1 enzymes alone and in complex with DNA may elucidate the role of these factors in determining the temperature and pH-dependent catalytic properties of the FEN-1 enzymes.
As found for all of the 5Ј-3Ј structure-specific nucleases assayed to date, only Mg 2ϩ and Mn 2ϩ can support catalytic activity in the archaebacterial enzymes. The cleavage preferences, defined by whether or not a 19-or 21-nucleotide product was the primary reaction product, was sensitive to the nature of the metal cofactor. In particular, Mn 2ϩ -catalyzed cleavage generally gave 19-nucleotide-long reaction products, as opposed to mixtures of 19-and 21-mers when Mg 2ϩ was used. A possible explanation of this result would be that Mn 2ϩ can still perform the required chemistry for the hydrolysis reaction; however, its larger atomic radius may restrict access of the 19-nt-long product site giving rise to the 21-nt-long product.
Alternatively, Mn 2ϩ may slightly alter the active site conformation and/or water structure such that binding and catalysis at the site 3Ј of the flap junction (and giving rise to the 21-mer) is disfavored. Interestingly, aFEN-1 shows no change in cleavage site preference in the presence of Mn 2ϩ , suggesting that the structural requirements in the aFEN-1 metal coordination site are not as strict as in the other enzymes. It has been proposed for the FEN-1 and the eubacterial and viral 5Ј-3Ј nucleases that several conserved acidic residues ligate a pair of Mg 2ϩ ions that subsequently interact with substrate DNA near the flap junction either directly or through tightly bound water molecules (38,39). It has thus been suggested that FEN-1 may employ a two-metal ion mechanism, as described for 3Ј-5Ј exonuclease domain of E. coli DNA polymerase I (52,53). This mechanism describes a reaction in which nucleophilic attack on the phosphorous atom at the flap junction is carried out by a hydroxide ion that is activated by one divalent metal, whereas the expected pentacoordinate transition state and the leaving oxyanion are stabilized by a second divalent metal ion that is 3.9 Å from the first. The crystal structures of several viral and eubacterial 5Ј-3Ј nucleases (33,34,54), however, demonstrated that the distance between the two bound metal ions is much larger than 3.9 Å, and a substantial conformational change, induced by DNA binding, would be required to accommodate a two-metal-ion mechanism. Thus, in the eubacterial and viral enzymes, it is possible that one Mg 2ϩ is catalytic whereas the other may have a structural role. To date, no structural information for a FEN-1 enzyme is available, and both the geometry and number of the metal liganding residues, which have currently been suggested by homology modeling with the weakly related bacterial and viral nucleases (16), remain to be elucidated. The availability of three archaebacterial enzymes in pure form should facilitate structural analysis and help clarify the specific mechanisms by which FEN-1 utilizes divalent metal ions to affect DNA structure-specific hydrolysis.
The DNA structural requirements for binding and catalysis by the archaebacterial FEN-1 enzymes, in some respects, are similar to those observed for eukaryotic members of the family (42). Like human FEN-1, substrates missing an adjacent strand (i.e., pseudo-Ys) were substrates for the enzymes; however, the rate of cleavage was reduced. In addition, when a pseudo-Y substrate was used, all three enzymes only gave a single, 21-nt-long reaction product, demonstrating that the upstream primer influences the ability of all the FEN-1 enzymes to cleave at nucleotides 5Ј of the flap junction. It is  (Table III) were determined using direct linear plots with Quickgraph software (Quickgraph Co.). E, hFEN-1; q, mFEN-1; Ⅺ, aFEN-1; and OE, pFEN-1. possible that interactions between this strand and the enzyme stabilize a pre-incision complex that disfavors melting of the DNA duplex, and thus cleavage rates at the site 3Ј of the flap junction are reduced. The structure-specific nature of FEN-1-catalyzed DNA hydrolysis implies that all substrates that possess the correct structure, regardless of the sequence of the component bases, are hydrolyzed at similar rates. Given that the primary role of this nuclease in eukaryotes is to assist in the removal of Okazaki fragments during lagging strand DNA synthesis, sequence-biased cleavage would likely be detrimental to the cell as certain sequences would not be hydrolyzed effectively by the enzyme. Thus, in murine FEN-1, when assayed with flap substrates containing different permutations of DNA bases near the flap junction, cleavage was equally efficient (42). We have repeated these experiments with the archaebacterial enzymes using two flap substrates, one that is more AT rich near the junction and one that is more GC rich near the junction. Whereas hFEN-1 and mFEN-1 showed no preference for either substrate, aFEN-1-and pFEN-1-mediated cleavage of the GCrich substrate was substantially reduced. Thus, for these two enzymes, changing two bases in the vicinity of the flap junction from AT to GC greatly reduced the overall activity of the enzymes. Examination of the aligned amino acid sequences of the four enzymes ( Fig. 1) reveals that in aFEN-1 and pFEN-1, but not in hFEN-1 and mFEN-1, an extra 9 amino acids are inserted between amino acids 198 and 206 (GKNVYVE for pFEN-1). This extra sequence motif may play a role in this unique property of the aFEN-1 and pFEN-1 nucleases.
Based on the above knowledge, we have set up "optimal" conditions to determine the kinetic parameters. At 55°C, the archaebacterial FEN-1s have only reached 1 ⁄16, 1 ⁄5, and 1 ⁄36 of the hFEN-1 nuclease activity, respectively. However, the archaebacterial enzymes have similar Michaelis constants (K m ), demonstrating that all the archaeal enzymes bind to the flap DNA substrate in a manner similar to the human enzyme even at a suboptimal temperature and with or without metal ions. These results are consistent with the result obtained in the direct binding assay (Fig. 3). A V max difference in pFEN-1 for 19-and 21-nt products once again confirmed that this enzyme has a cleavage site preference as we observed in the other experiments.
In summary, starting from three putative archaebacterial FEN-1 genes we have cloned, expressed, and purified the proteins and determined that each protein does in fact possess FEN-1 structure-specific nuclease activity. The availability of these three archaebacterial enzymes in pure form and in large amounts should facilitate biochemical and structural studies that define the factors that make the archaebacterial enzymes tolerant of high temperatures and also help decipher the detailed catalytic mechanisms that allow this critical enzyme to participate in DNA replication and repair. As the archaeal enzymes maintain affinity for DNA at low temperatures and because their catalytic activities are severely diminished even in the presence of metal ions at these low temperatures, these newly characterized enzymes will allow the substrate recognition and catalysis steps common to all FEN-1 family members to be examined separately.