Cooperation of the N-terminal Helicase and C-terminal Endonuclease Activities of Archaeal Hef Protein in Processing Stalled Replication Forks*

Blockage of replication fork progression often occurs during DNA replication, and repairing and restarting stalled replication forks are essential events in all organisms for the maintenance of genome integrity. The repair system employs processing enzymes to restore the stalled fork. In Archaea Hef is a well conserved protein that specifically cleaves nicked, flapped, and fork-structured DNAs. This enzyme contains two distinct domains that are similar to the DEAH helicase family and XPF nuclease superfamily proteins. Analyses of truncated mutant proteins consisting of each domain revealed that the C-terminal nuclease domain independently recognized and incised fork-structured DNA. The N-terminal helicase domain also specifically unwound fork-structured DNA and Holliday junction DNA in the presence of ATP. Moreover, the endonuclease activity of the whole Hef protein was clearly stimulated by ATP hydrolysis catalyzed by the N-terminal domain. These enzymatic properties suggest that Hef efficiently resolves stalled replication forks by two steps, which are branch point transfer to the 5′-end of the nascent lagging strand by the N-terminal helicase followed by template strand incision for leading strand synthesis by the C-terminal endonuclease.

It is essential for living organisms to replicate and transfer the genetic information in the genome precisely from parents to offspring. Interference with accurate DNA replication by a variety of intrinsic causes and extrinsic agents leads to serious diseases, including cancer. To maintain the accuracy of DNA replication, several repair systems developed in living organisms during evolution. The proofreading activity of DNA polymerases directly works for accurate DNA strand synthesis (for review, see Ref. 1). In addition, several repair systems operate at the damaged points and are controlled by the cell cycle checkpoint system (for review, see Refs. 2 and 3). The well known repair systems, such as nucleotide excision repair, base excision repair, mismatch repair, homologous recombination repair, and end joining are conserved across prokaryotes and eukaryotes (3). In the cells the DNA replication fork often stops at nucleotide lesions, backbone gaps, regions with secondary DNA structures, protein-bound regions, and other anomalous regions during DNA replication (for review, see Ref. 4 -6). To overcome the obstruction of the fork progression, cells utilize replication-coupled repair systems. Multiple pathways have beenproposedfortherepairofreplicationforks.DNApolymerasedependent translesion synthesis and bypass or collapse/restart of the stalled fork are two major pathways (for review, see Refs. 4 -13). In the latter case a Holliday junction (HJ) 1 intermediate is produced by regression of the fork and annealing of the two nascent strands, and this intermediate is resolved by reprogression after bypass synthesis at the lesion site or by cleavage at the junction point with HJ resolvases. HJ structures formed from stalled replication forks in the cells have been observed by electron microscopy (14,15). Alternatively, the stalled fork can be directly cleaved by a structure-specific endonuclease. When the DNA strands of the HJ or fork structure are cleaved by these endonucleases, the fork can be reconstructed by homologous recombination starting from the free DNA ends. The relationships between replication and recombination have mostly been analyzed in Escherichia coli (16,17). The Mus81 protein, which is conserved in all eukaryotic organisms and shares sequence similarity with the XPF (mammals) and Rad1 (yeast) endonucleases, is a structure-specific endonuclease that works as a heterodimer with Eme1 (human and Schizosaccharomyces pombe) or Mms4 (Saccharomyces cerevisiae) (18 -25). This protein complex has recently been in the spotlight in the research field of DNA repair and recombination, and there has been much debate about its in vivo function (26,27). Recent reports on the substrate specificity of the purified Mus81 complex suggested that this endonuclease directly cleaves stalled replication forks near the junction point to facilitate the resumption of the fork progression after recombinational repair (21)(22)(23)(24). Yeast mus81 mutants are sensitive to agents such as methylmethane sulfonate and camptothecin, which can stall replication forks, but not to ionizing radiation and bleomycin, which introduce DNA double strand breaks (21, 28 -31). These genetic analyses also support the function of the Mus81 complex in * This work was supported in part by the Japan New Energy and Industrial Technology Development Organization and by a grant-in-aid from the Ministry of Education, Science, and Sports of Japan (to Y. I. and H. S.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  repairing stalled replication forks. On the other hand, the human and yeast Mus81⅐Eme1 complexes also cleave HJs in vitro, and the yeast protein has substrate preference for a nicked HJ (32,33). From these observations, one of the proposed functions of Mus81 has been to resolve the intermediate in meiotic recombination to generate crossovers (for review, see Ref. 34).
Archaea, the third domain of life (35), is distinct from both Bacteria and Eukaryotes. Whereas Archaea belong to the prokaryotes in terms of cellular ultrastructure, they share many similarities with eukaryotes in the information processing pathways, including DNA replication, transcription, and translation (36), and the archaeal processes provide a useful model system to understand the more complex mechanisms of their eukaryote equivalents. Several proteins involved in homologous recombination are also conserved between eukaryotes and Archaea. For example, RadA and RadB, which play a central role in the initiation steps of homologous recombination (37)(38)(39), have sequences more similar to that of eukaryotic Rad51 than bacterial RecA (40,41). Furthermore, the Pyrococcus furiosus RPA, which is composed of three subunits like the eukaryotic RPA but differs from the bacterial single-stranded DNA binding protein, clearly stimulated a RadA-mediated strand exchange reaction (42). With regard to the HJ processing, we identified an archaeal HJ resolvase and named it Hjc (43). Hjc is an Archaea-specific protein, which lacks sequence and three-dimensional structural similarity to any other known HJ resolvase (44,45). To understand the molecular mechanism of the HJ processing in Archaea, we searched for proteins related to Hjc and found a novel endonuclease activity in P. furiosus. Identification of the gene corresponding to this activity revealed that the encoded protein consists of two distinct domains that are similar to the DEAH helicase family and the XPF nuclease superfamily. Biochemical characterization of each purified domain showed that these proteins have a specific affinity for branched DNA structures, including the replication fork. Actually, the N-terminal domain possesses an ATPase activity that was dramatically stimulated by fork-structured DNAs. The C-terminal domain has an endonuclease activity that specifically cleaved nicked, flapped, and fork-structured DNAs. Therefore, we designated this protein as Hef (helicaseassociated endonuclease for fork-structured DNA) (46,47). Hef is well conserved in Archaea and may be a prototypical enzyme for the eukaryotic XPF/Rad1/Mus81 nuclease family, working in important repair processes. To understand the function of Hef in more detail, we purified the full-length Hef protein to homogeneity and characterized its helicase and nuclease activities in this study. These two activities seem to work together for very efficient fork processing.

EXPERIMENTAL PROCEDURES
Synthetic DNA Substrates-The duplex (D), nicked (N), 3Ј-flap (3F), 5Ј-flap (5F), and pseudo-X (PX)-structured DNAs and the Holliday junction DNA (HSL) were prepared by annealing of the synthetic oligonucleotides as described earlier (48). The various fork-structured DNAs without a mobile region (FI), with a single-stranded region on the nascent strand of lagging synthesis (FG, FG4) or leading synthesis (FGE), containing a mobile region (FM) and containing a homologous region to facilitate the formation of a Holliday junction (FH) were also prepared by annealing the oligonucleotides. The nucleotide sequences and the combination of the synthetic oligonucleotide DNAs for the various substrate structures are shown in Table I.
Construction of Expression Vectors for the hef Gene-The hef gene was amplified by PCR directly from P. furiosus genomic DNA using the primers, 5Ј-GCACTACCATGGTATTAAGGAGAGACTTAA-3Ј and 5Ј-GTGCGTCGACCTACTCCTCATCCTCTATAT-3Ј. PyroBEST DNA polymerase (TakaRa) was used to maintain the accuracy of amplification. To place the translational initiation codon ATG at the NcoI site for the expression vector of pTV118 (Takara), the recognition sequences were included in the forward primer. In the reverse primer the SalI site was placed just after the termination codon. The PCR products were digested with NcoI-SalI, and the proper fragment was inserted into the corresponding sites of pTV118. The resultant plasmid was designated as pTVHef, and its nucleotide sequence was confirmed by DNA sequencing. The full-length Hef protein can be produced without any tag by pTVHef.
Purification of the Recombinant Hef Protein-E. coli JM109 carrying pTVHef was grown on LB agar medium containing 100 g/ml ampicillin at 37°C for 1 day. The cells (11 g) were harvested from the agar plates, suspended in buffer A (50 mM Tris-HCl, pH 8.0, 0.1 mM EDTA, 0.5 mM DTT, 0.2 M NaCl, proteinase inhibitor (Complete TM , Roche Applied Science)), and disrupted by a French press. The supernatant was treated at 80°C for 20 min, and the heat-resistant fraction was obtained by centrifugation. The heat-resistant fraction was fractionated on a HiTrap heparin column (Amersham Biosciences), which was equilibrated with buffer B (50 mM Tris-HCl, pH 8.0, 0.1 mM EDTA, 0.5 mM DTT, 10% glycerol) containing 0.2 M NaCl. The Hef protein was eluted at 0.5-0.6 M NaCl. The fraction was diluted with two volumes of buffer B and loaded onto a Mono S HR 5/5 column (Amersham Biosciences). The fractions that eluted at 0.36 -0.44 M NaCl were pooled and fractionated on a Superdex 200 PC 3.2/30 column (Amersham Biosciences) equilibrated with buffer B containing 0.3 M NaCl. The fractions that eluted at about 1.2 ml of retention volume were quickly frozen in liquid nitrogen and were stored at Ϫ140°C. The concentration of purified Hef protein was calculated by measuring the absorbance at 280 nm. The theoretical molar absorption coefficient of the molecule was calculated based on its tryptophan and tyrosine content, as described previously (49). The molar extinction coefficient of Hef is 50,920 M Ϫ1 cm Ϫ1 . Two truncated proteins, N546 (1-546) and C547Ј (547-763), were prepared as described previously (46).
Endonuclease Assays-The purified full-length Hef and C547Ј proteins were incubated with 10 nM concentrations of the various substrate structures labeled with 32 P at the 5Ј-or 3Ј-end in cleavage buffer (10 mM Tris-HCl, pH 8.8, 10 mM MgCl 2 , 1 mM DTT, 100 mM KCl) at 60°C for 30 min. The products were mixed with the same volume of loading buffer (98% deionized formamide, 1 mM EDTA, 0.1% xylene cyanol) and analyzed by 15% denaturing polyacrylamide gel electrophoresis in 1ϫ Trisbuffered EDTA followed by autoradiography.
Western Blot Analysis-P. furiosus cells (1 g) were disrupted by sonication in 15 ml of buffer B containing proteinase inhibitor (Complete TM ), and the extract was obtained by centrifugation. The P. furiosus cell extract (800 ng) and the purified Hef proteins (1 ng) were separated by 12% SDS-PAGE, blotted onto polyvinylidene difluoride membranes, and reacted with anti-N546 or anti-C547Ј antiserum. The bands were detected by using an enhanced chemiluminescence system (Amersham Biosciences) according to the supplier's recommendations.
Gel-retardation Assay-Purified N546 protein was incubated with 2 nM 5Ј-32 P-labeled duplex (D) or fork (FI) in binding buffer (20 mM triethanolamine-HCl, pH 7.5, 10 mM MgCl 2 , 1 mM DTT, 0.1 mg/ml bovine serum albumin, 0.5 mM ATP␥S) at 37°C for 20 min. Then glutaraldehyde was added to a final concentration of 0.2%, and the reactions were incubated for further 5 min at room temperature for fixation of the DNA-protein complexes. The products were separated by 6% polyacrylamide gel electrophoresis in 1ϫTAE buffer (40 mM Trisacetate, 1 mM EDTA, pH 8.0) and were visualized by autoradiography.
ATPase Assay-Purified N546 or Hef protein (each at 50 nM) was incubated with 1 mM ATP and [␣-32 P]ATP (25 pCi/l) in cleavage buffer in the presence or absence of DNA with various structures (800 nM) at 60°C for 0, 5, 10, 15, 20, 40, and 60 min. Aliquots of reactions were analyzed by thin layer chromatography. The amounts of ATP and ADP were quantified from the autoradiograms using a laser-excited image analyzer (BAS5000, Fujifilm).
Effect of C547Ј on the Arrested Replication Forks in E. coli-To produce the C547Ј and C547Ј mutant (C547Јm) proteins in E. coli, the genes encoding C547Ј and C547Јm were each inserted into the pBAD24 vector under the control of the araC promoter. As for C547Јm, a mutant with no endonuclease activity due to one amino acid substitution, D583A, at the catalytic center (47), was used. The fork-blocked strains (50) harboring pBAD24-C547Ј, pBAD24-C547Јm, or the pBAD24 vector alone were grown logarithmically in LB medium containing 1% glucose and were collected at the indicated time points after the induction of C547Ј or C547Јm production by changing the medium to LB plus 0.2% arabinose. The collected cells were embedded into agarose plugs and lysed in basically the same way as described previously (51) and then subjected to electrophoresis in a 1.0% agarose gel following PvuII restriction enzyme digestion. Southern hybridization analysis was performed as described previously (52).

RESULTS
Production and Purification of the Hef Protein-The fulllength hef gene was directly amplified from P. furious genomic DNA and cloned into the pTV118 expression vector. The resultant plasmid, pTVHef, was introduced into E. coli JM109, and the hef gene was expressed in the cells. However, the transformants grew slowly in the liquid medium, and the production A, full-length Hef or C547Ј mutant protein (each at 20 nM) was incubated with 10 nM 5Ј-32 P-labeled nicked (N), 5Ј-flap (5F), 3Ј-flap (3F), pseudo X (PX), or fork structured (FI) DNAs at 60°C for 30 min. The reaction products were analyzed by 15% denaturing PAGE and detected by autoradiography. The structure of each substrate is shown above the corresponding lanes, and the asterisks in the fork structures indicate the 32 P-labeled position. F, C, and Ϫ indicate the reactions with full-length Hef, C547Ј, and no enzyme, respectively. The GA sequence ladders of each labeled oligonucleotide generated by Maxam-Gilbert method were loaded alongside to provide markers (GA). B, the white and black arrowheads indicate the incision sites of the full-length Hef and C547Ј in each 5Ј-32 P-labeled substrate DNA. The sizes of the arrowheads represent the cleavage efficiency. level of Hef protein was very low. Furthermore, the protein rapidly degraded in the cells, although various cultivation conditions were tested. When E. coli carrying pTVHef was grown on LB-agar plates and the cells were directly collected from the plates, the yield and the stability of Hef protein in E. coli were substantially improved. The recombinant Hef protein was stable against heat treatment at 80°C for 20 min. After sequential column chromatography, including gel-filtration for the removal of the degraded products, the Hef protein was purified to homogeneity (Fig. 1A). Western blot analyses using domainspecific anti-N546 and anti-C547Ј antisera showed that the molecular weight of the recombinant Hef protein was consistent with that of the protein produced in P. furiosus cells (Fig.  1B). The slight difference of the band mobility between recombinant Hef and native Hef in Fig. 1B is probably due to the difference of salt concentration in the sample solutions.
The Hef Protein Forms a Homodimer at the C-terminal Domain-The XPF nuclease family members are known to form and act as heterodimers (27). To investigate whether Hef forms a dimer, gel-filtration analyses were carried out for the fulllength and truncated mutant Hef proteins. The full-length Hef protein was eluted at a position corresponding to about 170 kDa, clearly showing the formation of a homodimer ( Fig. 2A). Then the region responsible for the dimer formation was analyzed by using two mutant proteins. The two truncated proteins, N546 (1-546) and C547Ј (547-763), were subjected to the same column chromatography under the same conditions. The elution peaks of N546 and C547Ј corresponded to 62 and 41 kDa, respectively, indicating that the N546 and C547Ј proteins exist as a monomer and a dimer, respectively (Fig. 2B). These results show that Hef forms a stable dimer and the dimer interface region exists in the C-terminal endonuclease domain, as revealed by our crystallographic analysis (47) but not in the N-terminal helicase domain. When the mixture of the N546 and C547Ј proteins was subjected to gel-filtration chromatography, these proteins eluted individually (Fig. 2B). The elution profiles of the N546 were slightly different between the mixture with C547Ј and N546 alone (lower and upper panels, respectively in Fig. 2B). This difference probably depends on the different amounts of N546, which were loaded onto the gel-filtration column. To adjust the molecular ratio of N546 and C547Ј to 1:1 in the mixture, 1.4 g of N546 was mixed with 5 g of C547Ј. On the other hand, 11 g of protein was applied to the column to analyze the N546 by itself. These results indicate that the N-terminal and C-terminal regions exist as distinct domains and do not associate with each other in the Hef protein.
Substrate Specificity of the Hef Endonuclease Activity-To investigate the substrate specificity of the Hef endonuclease activity, DNAs with various structures were labeled with 32 P and were reacted with the Hef or C547Ј protein. The Hef protein specifically cleaved nicked (N), flapped (5F, 3F, and PX), and fork-structured (FI) DNAs at the 5Ј side of the nicked position (Fig. 3). Hef and C547Ј could not cleave pseudo-Y structured DNA (PY, a fork-like structure with only the template DNA strands), in contrast to the XPF/ERCC1 endonuclease, which specifically cleaved this structure. Hef and C547Ј recognized and cleaved the same DNA structures, suggesting that the substrate specificity for the cleavage ability of the Hef protein is contained in the C-terminal domain. Interestingly, the major bands produced by Hef and C547Ј were detected with a one-or two-nucleotide difference in the denaturing gel electrophoresis (discussed below).
Effects of Various Fork-structured DNAs on the Hef ATPase Activity-Previously, we reported that the N-terminal domain of Hef, N546, has ATPase activity, which is drastically en- hanced by fork-structured DNAs (46). This result supports the idea that Hef works at a stalled replication fork. In this study, we further characterized the helicase-like domain of Hef. As shown in Fig. 4A, N546 specifically bound to fork-structured DNA. The ATPase activity of N546 was enhanced in the presence of fork-structured DNAs about 4 -12 times more than in the reaction with normal duplex DNA. Among the fork-structured DNAs used in this study, a fork with an immobile branch point in which a single-stranded region is fixed in the template for lagging strand synthesis (FG) enhanced the ATPase activity of N546 about 2-fold higher than that by fork DNAs without a single-stranded region (FI) or with a single-stranded region on the nascent leading strand (FGE) DNA. When fork-structured DNAs containing a homologous region, by which the branch point can migrate (FM and FH), were added to the reaction, the most prominent enhancement of the ATPase activity was observed with the N546 protein. The DNA structure dependence of the ATPase derived from the full-length Hef protein was somewhat similar to that of the N546 protein (Fig. 4B).
The N-terminal Domain of Hef Has a DNA Structure-specific Helicase-The amino acid sequence of the N-terminal domain of Hef is similar to those of the DEAH helicase family members and has an ATPase activity that is drastically stimulated by fork-structured DNAs, as described above. These observations suggested that the N-terminal domain of Hef may unwind and migrate the branch point of the stalled replication fork to a suitable position for incision by the C-terminal nuclease. To investigate whether the N-terminal domain of Hef has a DNA structure-specific helicase activity, the N546 protein was incubated with 32 P-labeled fork-structured DNA (FG4), Holliday junction DNA, and pseudo-Y-structured DNA in the presence of ATP (Fig. 5). N546 dissociated the nascent strands of the leading and/or lagging synthesis from the template in the forkstructured DNA (Fig. 5A). Moreover, N546 also unwound the synthetic Holliday junction DNA and produced a pseudo-Ystructured DNA as a product (Fig. 5B). When pseudo-Y-structured DNA was included in the reaction, N546 never unwound DNA strands like the RuvAB proteins (Fig. 5C), suggesting the strict specificity of the Hef helicase activity for fork-structured DNA and Holliday junction DNA. The helicase activity of N546 was strong with low concentrations of salt (0 -10 mM KCl) and high concentrations of ATP (10 mM ATP, Fig. 5, D and E). ATP hydrolysis is necessary for the N546 helicase activity because N546 did not unwind fork-structured DNA in the absence of ATP or in the presence of 10 mM ATP␥S (Fig. 5E).
ATP Hydrolysis Stimulates the Incision of Fork-structured DNA by Hef-Our biochemical analyses showed that both the N-terminal and C-terminal domains independently act on forkstructured DNAs, as described above, and therefore, it is possible that both domains of Hef cooperatively work to efficiently repair the stalled replication fork. To investigate whether the helicase activity coupled with ATP hydrolysis affects the incision efficiency of the fork-structured DNA, the endonuclease assay was carried out with 10 mM ATP or ATP␥S using synthetic fork-structured DNAs as substrates. When FM was used as a substrate, the cleavage activity of Hef with ATP was about 4 times higher than that with ATP␥S. On the other hand, neither ATP nor ATP␥S stimulated the cleavage activity of C547Ј (Fig. 6, A and B). These results suggest that the Nterminal domain binds to the fork-structured DNA and transfers its branch point to a suitable position for cleavage by the C-terminal domain. The major cleavage site in FM substrate by Hef was one or two nucleotides different from that generated by C547Ј on the denaturing gel (Fig. 6, A and B). The helicase domain may slightly shift the binding position of the nuclease domain onto the fork junction.
When FH, a fork-structured DNA with a mobile region capable of forming a Holliday junction, was used as a substrate, Hef cleaved only the template strand for the leading synthesis with either ATP or ATP␥S (Fig. 6C). The addition of ATP stimulated the Hef endonuclease activity, and the level of this activity with ATP was about four times higher than that with ATP␥S. Furthermore, the Hef endonuclease activity was about 30 times higher than that of C547Ј (Fig. 6, D and E). The FH DNA forms a stable Holliday junction (chicken-foot structure), because Hjc, the Holliday junction resolvase, cleaved the FH DNA symmetrically at two sites within the Holliday junction structure (Fig. 6C). It may be difficult for the Hef endonuclease domain to bind by itself without transfer of the branch point of FH DNA by the N-terminal helicases. Interestingly, the sites in the FH DNA cleaved by Hef in the presence of ATP were two nucleotides away from those cleaved by C547Ј or by Hef with ATP␥S.
To further investigate the cooperative work of the N-terminal helicase and the C-terminal endonuclease in the Hef protein, N546 and C547Ј were added to the reactions, and the full-length Hef. These results further support the idea that both the helicase and the endonuclease collaborate for the rearrangement of the replication fork structures.
C547Ј Cleaves Stalled Replication Forks in E. coli Cells-To investigate whether the Hef endonuclease works for the cleavage of the stalled replication fork in actual chromosomal DNA, an E. coli strain having an ectopic TerL, at which the replication fork is arrested, was used as the host cell (Fig. 8A), and the C547Ј protein was produced from the inducible expression plasmid in the E. coli cells. When the expression of the gene for C547Ј was induced, the amount of fork-structured DNA decreased, whereas a constant amount of the double-stranded DNA was detected. Furthermore, the amount of fork-structured DNA did not decrease when the gene for the mutant C547Ј protein, which lacks the endonuclease activity because of the substitution of the active residue by site-directed mutagenesis, was expressed under the same conditions (Fig. 8, B and C). This result indicates that the C547Ј protein produced in E. coli actually attacked and cleaved the stalled fork structure produced at the TerL sequence on the chromosomal DNA in the cells. Hef retains 5ϳ10% of the endonuclease activity at 37°C relative to that at 60°C in vitro, although Hef is originally from a hyperthermophilic organism (data not shown).

DISCUSSION
Hef is an archaeal structure-specific endonuclease originally identified from P. furiosus (46). Hef consists of two domains, the DEAH family helicase domain at the N-terminal region and the XPF/Rad1/Mus81 nuclease superfamily domain at the Cterminal region. In this study, using the full-length Hef protein we showed that the fork structure-specific endonuclease activity derived from the C-terminal domain was clearly stimulated by ATP. We also found that the purified N-terminal domain has an ATP-dependent helicase activity for fork structured and Holliday junction DNA. These results strongly suggest that the archaeal Hef protein works to restore the stalled replication fork via the cooperation between the N-terminal helicase and C-terminal endonuclease activities. Some genetic experiments are required to confirm our prediction about the function of Hef in vivo. However, a useful genetic system has not been established in Pyrococcus strains yet. Therefore, we used E. coli cells with the replication terminator sequence ter at an ectopic position, where the replication fork can be arrested. The amount of the Y-structure formed at the ter site clearly decreased when the expression of the Hef endonuclease gene was induced. This reduction was not observed when the mutant gene at the active site was expressed under the same conditions (Fig. 8). These results support the idea that the Hef endonuclease can collapse the arrested fork structure in the cells, probably by cleavage at the junction as observed in vitro.
The N-terminal domain of Hef carries the signature motifs of the DEAH helicase family members, which act in a large variety of cellular functions, especially in nucleic acid transactions in all organisms (53). The RecQ family helicases, which compose one group in the DEAH helicase superfamily, have been identified in bacteria and eukaryotes and are known to be involved in the restarting of stalled replication forks (for review, see Refs. 54 and 55). Indeed, the purified human BLM and WRN proteins and SGS1 of S. cerevisiae directly bind to Holliday junctions and promote branch migration in vitro (56 -58). Very recent work has shown that a fork structure with a leading strand gap is the best substrate of E. coli RecQ helicase among various branched DNAs (59). However, no gene encoding a structural homolog of RecQ family helicases has been found in archaeal genomes including P. furiosus, except for the Aeropyrum pernix and Methanosarcina species, and no biochemical analysis of an archaeal RecQ has been reported to The full-length Hef and C547Ј-truncated mutant proteins showed the same substrate specificity with or without ATP, suggesting that the substrate specificity for the cleavage reaction is independently determined by the C-terminal domain (Fig. 3). Moreover, gel-filtration analyses showed that the Nterminal and C-terminal domains did not associate with each other in solution (Fig. 2). These results indicate that the two  Template and nascent strands are shown by black and gray lines, respectively. The block indicates a lesion that causes a replication stall. A, when lagging strand synthesis is ahead of a leading strand that is arrested by encountering a lesion, the replication fork forms a 3Ј-single-stranded flap structure. The C-terminal domain of Hef protein (Hef-C) can resolve the fork and produces a gapped duplex DNA and a duplex DNA containing a 5Ј-single stranded tail. B, when lagging strand synthesis is behind a stalled leading strand, the branch point in the fork regresses to the 5Ј-end of the nascent lagging strand by the N-terminal helicase activity of Hef (Hef-N) with ATP hydrolysis. Then the C-terminal endonuclease of Hef (Hef-C) can incise the junction to generate a gapped duplex DNA and a duplex DNA containing a 3Ј-single-stranded tail. C, structural specificity of the Hef and Hjc proteins for recognition and cleavage. When replication fork is arrested, the Hef protein cleaves the arrested fork directly near the branch point. On the other hand, when the two nascent strands are unwound from the templates and reannealed with each other, Hjc recognizes the four-way junction and cleaves it symmetrically.
Hef protein domains are connected by a flexible linker and may independently recognize fork-structured DNA. In contrast to the substrate specificity, the cleavage efficiency of the fulllength Hef protein was enhanced by the addition of ATP (Fig.  6), although no effect was observed for the endonuclease activity of C547Ј. This observation suggests that ATP hydrolysis by the N-terminal helicase domain promotes structural changes in the replication forks to create the optimum structure for the C-terminal endonuclease.
Based on these results, we propose an active role of the Hef protein at stalled replication forks (Fig. 9). The essence of the model is to cut off the stalled leading strand synthesis. When leading strand synthesis is delayed behind the lagging strand, the endonuclease activity of Hef incises the template strand for leading synthesis directly at the branched point (Fig. 9A). In the opposite case in which leading synthesis goes ahead of lagging synthesis, the N-terminal helicase domain of Hef unwinds the nascent strand to regress the branch point to the 5Ј terminus of the newest nascent strand of the lagging synthesis and then the C-terminal endonuclease domain of Hef cleaves the template for leading synthesis (Fig. 9B). The strict cleavage specificity for the template DNA strand is quite suitable for Hef as a restoration enzyme for stalled replication forks. When both nascent strands anneal with each other and form a four-way junction (chicken-foot) structure, the N-terminal helicase domain also unwinds both strands and converts the fork structures for incision by the C-terminal nuclease domain of Hef (Fig. 9C). Cleavage of the junction of the stalled fork may be especially important when both of the leading and lagging strands are blocked. In hyperthermophilic organisms, which suffer from constant damage on both genomic DNA strands because of their extremely high temperature habitat, quick and effective repair systems may be essential for their survival. Interestingly, Archaea has two endonucleases, Hjc and Hef, which specifically cleave junction DNAs in vitro. Biochemical analyses suggested that Hjc resolves Holliday junctions formed at regressed replication forks and recombination intermediates, whereas Hef resolves replication forks (Fig. 6).
In E. coli, RuvC is the only endonuclease known to be involved in processing stalled replication forks. In eukaryotes, S. cerevisiae Mus81⅐Mms4 and S. pombe/human Mus81⅐Eme1 cleave replication forks and Holliday junctions in vitro (18 -20), and further analyses showed that the Mus81-associated endonuclease activity prefers 3Ј-flap and fork structures over Holliday junctions in vitro (21)(22)(23)(24)(25). On the other hand, human cell-free extracts contain a helicase-associated Holliday junction-specific endonuclease activity, which is different from the Mus81-associated endonuclease (60,61). It would be interesting to investigate whether these two endonuclease activities from human cells exhibit substrate specificities similar to those of Hef and Hjc from Archaea.
The archaeal Hef protein provides very interesting insight into the evolution of the XPF superfamily. Data base searching revealed that the most similar sequence to that of the Nterminal domain of Hef exists in Mph1 from S. cerevisiae. Mph1 is involved in DNA repair, and its homologs are widely found in yeast to human sources (62). More biochemical analyses of Mph1 are certainly needed. The N-terminal domains of Hef and XPF are also similar, although XPF lacks critical residues for the helicase function, and the C-terminal nuclease domains also share similarity between the two proteins (46,63). Therefore, these proteins may be derived from a common ancestor. Because there is no other XPF homolog in P. furiosus, Hef may act on the nucleotide excision repair pathway as an endonuclease/helicase in archaeal cells, as discussed previously by our group and others (46,64). Recently, we found another gene that encodes a protein with an Mph1-like DEAH helicase domain at its N terminus and an XPF/Mus81 nuclease domain at the C terminus in the human genome. This gene may be an ortholog of the archaeal Hef, and we are now characterizing the protein.
The distribution of XPF/Mus81 family proteins and their share of functions in living organisms are becoming quite interesting, and our archaeal study demonstrating the cooperative work of the helicase and the endonuclease at the stalled fork may shed light toward understanding a universal repair mechanism in living organisms. Practical genetic techniques have been developing for halophilic Archaea, which have hef and hjc gene homologs. Genetic studies using these strains are valuable to directly investigate the functions of these proteins in the living cells.