ABSTRACT
Phage T5 exonuclease is a 5' -> 3' exodeoxyribonuclease that also exhibits endonucleolytic activity on flap structures (branched duplex DNA containing a free single-stranded 5'-end). Oligonucleotides were used to construct duplexes with either blunt ends, 5'-overhangs, 3'-overhangs, a flap or a forked end (pseudo-Y). The binding of T5 exonuclease to various structures was investigated using native electrophoretic mobility shift assays (EMSA) in the absence of the essential divalent metal cofactor. Binding of T5 exonuclease to either blunt-ended duplexes or single-stranded oligonucleotides could not be detected by EMSA. However, duplexes with 5'-overhangs, flaps and pseudo-Y structures showed decreased mobility with added T5 exonuclease. On binding to DNA the wild-type enzyme was rendered partially resistant to proteolysis, yielding a biologically active 31.5 kDa fragment. However, the protein-DNA complex remained susceptible to inactivation by p-hydroxymercuribenzoate (PHMB, a cysteine-specific modifying agent), suggesting that neither cysteine is intimately associated with substrate binding. Replacement of both cysteine residues of the molecule with serine did not greatly alter the catalytic or binding characteristics of the protein but did render it highly resistant to inhibition by PHMB.
Bacteriophage T5 exonuclease (T5-exo) was originally purified from phage-infected cells by Paul and Lehman (1 ) and later shown to be a 5' -> 3' exonuclease by Frenkel and Richardson (2 ). The D15 gene encodes this nuclease (3 ) and it is required for DNA replication (2 ). The wild-type T5-exo is a 33 kDa protein which has been overexpressed, purified to homogeneity and has an absolute requirement for divalent cations (Mn2+ or Mg2+) for activity.
Bacteriophages such as T4 and T7 express proteins that are related to T5-exo. These phage-encoded proteins share many biological activities and sequence similarities with the N-terminal domains of the eubacterial DNA polymerase I group (4 -6 ). The 5' -> 3' exonucleases are also responsible for processing of Okazaki fragments (an RNase H activity) formed during DNA replication (7 ). Recent findings suggest that the 5' -> 3' exonuclease domain of Escherichia coli DNA polymerase I (ECPolI) is also involved in RecA-independent recombination processes (8 ) and for initiation of replication at OriK sites (9 ).
Recent reports have shown that several 5'-nucleases, such as DNA polymerase I, T7 exonuclease, Thermus aquaticus polymerase (TaqPolI) and FEN-1 (a mammalian homologue, also called DNase IV; 10 ), possess structure-specific endonuclease activity as well as the more obvious exonuclease. Strand-displaced or bifurcated structures are cleaved in an endonucleolytic manner at or near the bifurcation site (11 ). The crystal structure of T5-exo reveals a helical arch that may be involved in DNA binding and a mechanism in which DNA could thread through such an archway has been proposed (12 ). Further evidence for a threading mechanism has been provided for another member of this class of nucleases, the calf homologue of FEN-1, in elegant work carried out by Bambara and co-workers (13 ).
In this study we set out to investigate the DNA binding properties of T5-exo and to re-examine an earlier report that this exonuclease is inhibited by mercury-containing compounds (14 ). The latter observation was intriguing, as cysteine residues are not conserved in this class of enzymes (5 ). Indeed, the 5' -> 3' exonuclease domains of Haemophilus influenzae and T.aquaticus polymerase completely lack cysteine residues. Yet, if cysteine residues are not conserved how can agents such as p-hydroxymecruribenzoate (PHMB) inhibit this particular enzyme so effectively? We set out to determine the mode of mercury(II) inhibition of T5-exo and to investigate whether either of the two cysteine residues present in the protein are essential for biological activity.
Oligonucleotide site-directed mutagenesis was carried out on a single-stranded M13 derivative carrying the cloned T5 D15 exonuclease gene (4 ). The phosphorothioate-based high efficiency mutagenesis procedure (15 ) was used to alter the Cys115 codon from TGT to TCT, which results in a substitution of Cys by Ser. This was achieved using the primer CS115 (dTGT AGT TTT AGA
The mutated genes were subcloned essentially as described (4 ,17 ). Protein expressed from a gene carrying the C115S mutation was designated CS115-exo; similarly, 2CS-exo contained both Cys -> Ser mutations. The wild-type and mutant proteins were purified essentially as described (4 ), except that an initial urea solubilization stage was included for the mutant proteins, as they were found to be insoluble on overexpression. The proteins were bound to a MacroQ (BioRad) ion exchange column and renatured by elution of the protein from the column using a NaCl gradient in a Tris-HCl buffer which contained no urea. Activity of the mutant protein was monitored and used as the main criterion for pooling of the column fractions. Protein concentration was determined using the Bradford assay (18 ).
Three oligonucleotides were combined to form the flap structure described by Lieber and Harrington (19 ). The flap strand (5'-GATGTCAAGCAGTCCTAACTTTGAGGCAGAGTCC-3') and adjacent strand (5'-CACGTTGACTACCGTC-3') were annealed to a complimentary bridging strand (5'-GGACTCTGCCTCAAGACGGTAGTCAACGTG-3'). The terminal 20 5'-nucleotides of the flap structure are single stranded. The oligonucleotide 5'-CTTGAGGCAGAGTCC-3' was annealed to either the bridging strand, forming a structure with a 15 nt 5' single-stranded region, or to another oligonucleotide (5'-GGACTCTGCCTCAGG-3'), to form a 15 base pair duplex (Fig. 1 A). The oligonucleotides shown in Figure 1 A were 5'-end-labelled with [32P]ATP (Amersham International) under standard conditions and purified from a 7 M urea-15% acrylamide gel (length 40 cm) essentially as described (20 ). The unlabelled oligonucleotides were also purified from a 7 M urea-15% acrylamide gel, after visualization by UV shadowing.
Reactions containing 13.5 nM unlabelled complementary oligonucleotides and 450 pM labelled oligonucleotides in 25 mM potassium glycinate, pH 9.3, 100 mM KCl were heated to 80oC for 5 min and cooled for 1 h to room temperature in order to anneal the oligonucleotides. Reactions were then diluted to contain 4.5 nM unlabelled oligonucleotide, 150 pM labelled oligonucleotide in 25 mM potassium glycinate, 100 mM KCl, 1 mM EDTA, 5% glycerol, 1 mM DTT, 0.1 mg/ml acetylated BSA and T5-exo at the concentration stated. The enzyme was diluted in 25 mM potassium glycinate, 50% glycerol, 1 mM EDTA, 1 mM DTT, 0.1 mg/ml acetylated BSA and this buffer was added to the reactions that contained no T5-exo. The enzyme and oligonucleotides, in a final reaction volume of 10 [mu]l, were incubated on ice for 10 min and analysed on a 17% native acrylamide gel, buffered in 50 mM Tris-Bicine, pH 8.3, 1 mM EDTA, 1 mM DTT at 4oC for 2 h at 15 V/cm. The gel was visualized and results quantified with a Molecular Imager (BioRad) and Molecular Dynamics software. Reactions in which the influence of PHMB on binding were investigated contained neither DTT nor EDTA. The enzyme was diluted in 25 mM potassium glycinate, 30% glycerol, 0.1 mg/ml acetylated BSA and PHMB at the concentration stated, incubated on ice for 15 min and an aliquot added to the reaction mix described above.
The release of acid-soluble nucleotides from high molecular weight (Type XIV; Sigma) DNA was monitored by UV spectroscopy as described, (4 ) except that the assay contained DNA at 670 [mu]g/ml in 600 [mu]l 25 mM potassium glycinate, pH 9.3, 10 mM MgCl2 and 1 [mu]g protein.
Inhibition with a mercury derivative was investigated by preincubating the enzymes at room temperature for 10 min in 15 mM potassium glycinate buffer containing 0.1-5 mM PHMB (Aldrich) and 5 [mu]l of this preincubation reaction, containing 1 [mu]g enzyme, was added to the nuclease assay as above. Similar assays were also carried out using 5 mM iodoacetic acid in the preincubation step instead of PHMB.
A further assay was carried out in which the protein was first treated with an excess of high molecular weight DNA in an attempt to protect the enzyme from inactivation by PHMB. The 10 [mu]l preincubation reactions contained 2 [mu]g wild-type T5-exo incubated with 2 [mu]g Type XIV fish sperm DNA for 5 min at room temperature. PHMB was then added to a concentration of 1 mM and the reactions incubated for a further 5 min at room temperature before addition of 5 [mu]l preincubation mix to the nuclease reaction.
Wild-type T5-exo (2 [mu]g) was incubated with 7 [mu]g Type XIV DNA from herring testes in a 10 [mu]l reaction volume, buffered in 25 mM potassium glycinate, pH 9.3, for 5 min at 37oC. Proteolysis was effected by addition of 1 [mu]l fresh solution of modified sequencing grade trypsin (Promega) or proteinase K. The reactions were stopped after incubation at 37oC for 5 min by addition of 2 [mu]l 2 mg/ml PMSF stock solution in ethanol and analysed on an SDS-PAGE substrate gel.
A discontinuous SDS-PAGE gel containing 20 [mu]g/ml Type XIV DNA from herring testes in the resolving gel was prepared essentially as described (21 ). The products of partial proteolysis were separated on this gel, the protein renatured in situ and MgCl2 added to a concentration of 10 mM. After staining with ethidium bromide, exonuclease activity was visualized as a shadow against a fluorescent background when viewed on a UV transilluminator. The gel was then Coomassie stained.
T5-exo treated with trypsin in the presence of DNA was subjected to SDS-PAGE and electroblotted onto a PVDF membrane (22 ). The membrane was stained briefly to locate the protein band, which was N-terminal sequenced on an Applied Biosystems Model 476A amino acid sequencer.
Labelled flap structures were formed by heating 6 fmol labelled flap strand and 170 fmol unlabelled adjacent and bridging strands per reaction to 80oC in a volume of 10 [mu]l and allowing them to cool slowly to room temperature. Exonuclease was diluted in 25 mM potassium glycinate, pH 9.3, 10% glycerol, 1 mg/ml BSA, 1 mM DTT, 1 mM EDTA. The flap structure was subjected to treatment with 2 pmol exonuclease in a reaction volume of 10 [mu]l containing 25 mM potassium glycinate, pH 9.3, 100 mM KCl, 0.5 mg/ml BSA and 10 mM MgCl2 for 2 min at 37oC. The marker lane was produced by digesting labelled flap strand with 0.5 and 2 ng snake venom phosphodiesterase for 3 min at 37oC and combining the products of the reactions. All the reactions were stopped by addition to 95% formamide, 15 mM EDTA stop mix. The products of the reactions were separated on a 7 M urea-15% polyacrylamide gel (length 50 cm) run in 1* TBE at 45 W for 2 h and visualized using a BioRad Molecular Imager phosphorimager.
The gel retardation assay showed that T5-exo (at a protein concentration of 120 nM) binds to pseudo-Y substrate, flap structure and structures with a 3'- or 5'-single-stranded region in the absence of Mg2+ (Fig. 1 ). An exonuclease concentration of 1.2 nM is sufficient to detect retardation of the pseudo-Y substrate (Fig. 4 ). However, even at a concentration of 3 [mu]M T5-exo fails to form a detectable complex with the single-stranded oligonucleotide. At a concentration 100 times higher than that required to show binding to the pseudo-Y substrate T5-exo did not appear to bind a duplex substrate.
The C115-S (CS115-exo) and double C115S, C265S (2CS-exo) mutants accumulated as insoluble aggregates upon overexpression. The insoluble protein was refolded after urea solubilization and the specific activity of the mutant enzymes was compared with that of the wild-type protein. These results are shown in Figure 2 A. The wild-type enzyme, CS115-exo and 2CS-exo reactions all have very similar progression curves. Assays were also carried out in which the DNA concentrations used were either doubled or halved, with no detectable change in the initial reaction velocity (results not shown), thus the initial reaction velocity was likely to be approaching Vmax. The structure-specific endonuclease activity of the wild-type and mutant proteins was compared (Fig. 2 B). A flap structure consisting of a duplex DNA containing a free single-stranded 5'-end was digested with the exonucleases and the reaction products were analysed on a denaturing 15% polyacrylamide gel. There were no significant differences in the sites of cleavage of the substrate by the mutant and wild-type enzymes (Fig. 2 C).
Incubation of the wild-type protein with 0.1 mM PHMB was shown to inhibit binding to a pseudo-Y substrate, but 5 mM PHMB did not alter binding of the 2CS-exo mutant to the same substrate (Fig. 3 ). After incubation of the wild-type T5-exo with 1 mM PHMB it was shown that in order to bind ~50% of the substrate 40 times more enzyme was required than when the enzyme was not incubated with PHMB. In the absence of the mercury compound approximately three times more 2CS-exo than wild-type exonuclease was required to bind 50% of the substrate (Fig. 4 A). The apparent Kd of the wild-type exonuclease, measured using the EMSA (23 ), was 5 nM; the Kd of the 2CS-exo mutant was 15 nM.
The influence of the sulfhydryl-specific modifying agents PHMB and iodoacetic acid on exonuclease activity were studied using the spectrophotometric assay. Preincubation with up to 5 mM PHMB did not significantly reduce activity of 2CS-exo, caused an approximate 70% reduction in CS115-exo activity and abolished wild-type activity altogether (Fig. 6 A). Concentrations of PHMB as low as 0.1 mM in the preincubations reduced wild-type reaction initial velocity to virtually zero (Fig. 6 B). Addition of DTT to a final concentration of 3 mM in the reaction restored ~50% activity to wild-type T5-exo after treatment with PHMB (Fig. 6 C). Preincubation of the 2CS-exo mutant with 5 mM iodoacetic acid had no affect on nuclease activity, but the same preincubation conditions reduced the level of wild-type activity by 50% (results not shown). Preincubation of the wild-type T5-exo with an excess of high molecular weight DNA before addition of the mercury derivative did not protect the enzyme from inactivation (results not shown).
We first found that T5-exo is capable of binding to DNA in the absence of cofactor when we undertook differential proteolysis experiments. We examined the cleavage fragments produced by the action of proteases on T5-exo, both in the presence and absence of DNA. Lysine and arginine residues are subject to attack by trypsin and there are 37 potential target sites for trypsin in the primary sequence. Reaction of T5-exo with trypsin produced fragments too small to be identified on 10% SDS-PAGE. Theoretically, tryptic digestion of T5-exo would produce >25 fragments with molecular weights in the range 2926-500 Da. Some fragments of ~25 kDa were identified on SDS-PAGE analysis, but no fragment <31 kDa showed activity on substrate gels (Fig. 5 ). However, a T5-exo-DNA complex was significantly protected from protease digestion and was converted rapidly to a slightly faster migrating metastable 32 kDa intermediate (as determined by SDS-PAGE). Thus it appears that DNA binding masks many trypsin-sensitive cleavage sites (Lys/Arg). A similar result was obtained when proteinase K was substituted for trypsin. The simplest explanation of this is that the bound DNA contacts all but the N-terminal protease-sensitive surface-exposed sites. Alternatively, DNA binding may impart greater rigidity to the protein, rendering it less susceptible to attack.
Using a substrate gel it was shown that the 32 kDa fragment retained exonucleolytic activity. It can clearly be seen that a trypsin concentration of 0.4 [mu]g/ml has no apparent affect on activity when the enzyme is incubated with DNA, but without DNA the exonuclease activity is almost completely lost. Sequencing of the 32 kDa fragment revealed that the first 17 amino acids had been removed from the enzyme by treatment with trypsin. This would yield a product of 31.5 kDa, which is in agreement with the size estimated from SDS-PAGE. A small amount of this fragment was purified by liquid chromatography and was shown to be fully active (data not shown). We conclude from this that the first 17 amino acids of the polypeptide are not required for catalysis. This was expected, as sequence similarities do not appear to extend to the N-termini of the 5'-nuclease family (5 ). The first 18 amino acids have also been shown to be disordered and distant from the active site by X-ray crystallography (12 ).
Structure-specific endonucleolytic DNA cleavage catalysed by 5' -> 3' exonucleases is a relatively recently characterized phenomenon. However, as long ago as 1982, Lundquist and Olivera proposed that the 5' -> 3' exonuclease domain of DNA polymerase I (PolI) could process displaced 5'-ends during nick translation synthesis (24 ). Dahlberg and co-workers showed that strand- displaced structures (bifurcations or flaps) could be cleaved efficiently by an endonucleolytic activity present in several prokaryotic 5' -> 3' exonucleases (11 ). Flap endonuclease I (FEN-1) is a eukaryotic protein that is homologous to the prokaryotic 5' -> 3' exonucleases. Work on FEN-1 provides compelling evidence that this protein loads on to the free 5'-end of the flap structure by a threading mechanism and then translocates to the duplex region (13 ), as had been proposed in the case of the eubacterial 5' -> 3' exonucleases (11 ). Cleavage takes place at this site. Circumstantial evidence supporting a threading model in the homologous T5-exo has emerged. The crystal structure of this latter enzyme shows the presence of an unusual helical arch motif, which forms a hole in the protein. The hole is made up from an inverted V-shaped (bent) helix. This is positioned above a globular domain comprised of two binding sites for divalent metal ions (12 ) contained within a concave surface smothered with a preponderance of positively charged amino acids.
A model for DNA binding was proposed in which the single-stranded arm of a flap substrate threads through the helical archway. Most of the amino acid residues conserved in the multiple sequence alignment (5 ) are positioned in or very close to the bottom of this arch. The positively charged concave surface could be involved in electrostatic interactions with the nucleic acid phosphodiester backbone. The binding of various model DNA structures to T5-exo in the absence of divalent metal ions was analysed using a gel retardation assay. We were able to demonstrate binding to a substrate with a free 5'- or 3'- end, but not to a single-stranded substrate or purely duplex DNA (Fig. 1 ). T5-exo digests this single-stranded substrate (result not shown; 4 ). Thus the absence of binding in the retardation assay may be due to the lack of a divalent cation, which would imply that alternative binding mechanisms apply to different substrates depending upon their structure. Alternatively, rapid threading of the single-stranded DNA through the helical arch of the exonuclease could explain the lack of a positive band shift (12 ). However, whilst the single-stranded 5'-arm of the pseudo-Y and flap structures would also pass through the helical arch, the enzyme presumably becomes immobilized at the junction between the single-stranded and double-stranded regions of the substrate.
The binding of substrates consisting of a single-stranded region upstream of a duplex (flap, pseudo-Y and 5'-overhang) by T5-exo and failure to bind strongly to purely double-stranded or single-stranded substrate is in accordance with the results determined for the mammalian nuclease FEN-1 (25 ). However, FEN-1 was reported to bind to a 3'-overhang substrate with an apparently much lower affinity than that reported for T5-exo here. Using an identical flap substrate to that utilized by Lieber and Harrington for their work with FEN-1, we found that T5-exo and FEN-1 cleave predominantly in the same positions, 1 nt either side of the predicted duplex-5'-overhang junction (26 ), underlining the similarity between these prokaryotic and eukaryotic enzymes. The initial exonucleolytic product produced by T5-exo is a trinucleotide (Fig. 2 ).
Inhibition by PHMB of an unidentified nuclease activity present in bacteriophage T5-infected E.coli was reported anecdotally by Rogers and Rhodes in 1976 (14 ), long before the cloned and purified D15 gene product was characterized (4 ). The former report mentioned that the major exonuclease activity in phage- infected cell extracts could be 95% inhibited by 3 mM PHMB. The influence of PHMB on purified T5-exo activity was investigated using a spectrophotometric activity assay. Modification of the cysteine residues in the wild-type T5-exo with PHMB was found to abolish activity, though this was not due to an irreversible disruption of the enzyme structure, as activity could be regained (Fig. 6 C) by addition of the sulfhydryl-containing compound DTT, which reduces the mercury-sulphur bond.
As cysteine residues are not conserved in this class of enzymes and they are located far from the active site (5 ,12 ,27 ), we sought to clarify their role in the enzyme structure-function relationship. Site-directed mutagenesis was used to produce a single and double mutant with either C115S or both Cys -> Ser substitutions. The wild-type and both cysteine mutant exonucleases, in the absence of PHMB, had approximately equal reaction velocities under the conditions used in the spectrophotometric activity assay. Neither doubling nor halving the concentration of DNA substrate used in these reactions significantly altered the outcome of the assay. This shows that the cysteine residues are not required for normal exonuclease activity under conditions where substrate is present at saturating concentrations. The wild-type and both mutated exonucleases were also able to cleave flap structures in a manner analogous to that reported for the FEN-1 nuclease, hence the cysteine residues are not required for recognition of structure-specific cleavage sites (26 ).
Concentrations of PHMB as high as 5 mM had little affect on the double Cys -> Ser (2CS-exo) mutant (Fig. 6 B), whereas incubation of the wild-type T5-exo with 0.1 mM PHMB abolished activity. As 2CS-exo was active after treatment with high concentrations of PHMB, this demonstrated that the affects seen were due to interaction of PHMB with the enzymes and not that PHMB-DNA interactions were preventing catalysis. Iodoacetic acid, another reagent for selective cysteine modification, was similarly able to inhibit wild-type T5-exo.
We investigated the affect of PHMB on the DNA binding properties of wild-type and mutant T5 exonucleases. The mercury compound was not added directly to the assay reaction because it was found to lead to dissociation of the oligonucleotides used to form the substrate at higher concentrations (result not shown; 28 ). Instead, the enzymes were preincubated with PHMB at concentrations of between 0.1 and 5 mM. A fraction of this incubation was added to the assay reaction; the concentration of the mercury compound in the assay was <0.5 mM (Fig. 3 ). This concentration of PHMB did not alter the amount of unannealed labelled oligonucleotide in the reactions (labelled single-stranded in Fig. 3 ) and so did not disrupt the substrate. Thus any change in binding was due to direct interaction of PHMB with the protein during preincubation. The results show that modification of the wild-type enzyme with PHMB interferes with binding to the DNA substrate used in the EMSA. In contrast, 2CS-exo is insensitive to preincubation with 5 mM PHMB.
This result might suggest that PHMB exerts its affect by blocking the DNA binding site on T5-exo, however, structural studies have shown that neither cysteine is located on the same face as the active site and proposed DNA binding region. If cysteines do occupy sites near the DNA binding region then it should be possible to protect the nuclease from inactivation by first binding DNA. Under the same conditions used to form a nuclease-DNA complex in the trypsin protection experiments above it was found that wild-type T5-exo still became inactivated by PHMB (results not shown). This suggests that modification occurs at a site not occupied by DNA, i.e. the cysteines are not likely to be intimately associated with the DNA binding site of the enzyme. The mercury compound must disrupt the structure of the enzyme, interfering with DNA binding. Inactivation of an enzyme by chemical modification at a site remote from the active centre has been observed and several similar cases of allosteric inactivation have been reported in the literature; for example Cys148 of E.coli lac permease and Cys395 of the [beta]-chain of E.coli glycine tRNA synthetase, which when chemically modified yield inactive proteins. However, when site-directed mutagenesis was employed to substitute the supposed critical residues with other amino acids the mutants remained fully active (reviewed in 29 ).
In conclusion, it is clear from these studies that T5-exo, in the absence of divalent metal cofactor, is able to bind strongly to DNA structures consisting of a single-stranded arm adjacent to a duplex region. Neither blunt-ended nor single-stranded substrates showed significant retardation in the EMSA system used to analyse binding. The biological significance of binding to a 3'-overhang is unclear, though it may be possible that the 5'-end at the single-stranded junction in such substrates is more liable to breathing than a blunt-ended homologue. Cysteines are not required for endonuclease or flap endonuclease activity and wild-type and mutant enzymes all showed remarkably similar cleavage specificities to eukaryotic FEN-1. These results support the threading model proposed for 5'-nuclease binding and imply that conserved mechanisms of structure-specific DNA cleavage and recognition operate within the 5'-nuclease family.
We thank the Biotechnology and Biological Sciences Research Council for a studentship to S.J.G. and thank Arthur Osyczka and Robert Trzos for technical assistance with the preparation of mutant exonuclease genes. This work benefited from the use of the Daresbury Laboratory's SEQNET service. We also thank Dr A.Moir (Krebs Institute, University of Sheffield) for protein sequencing.
*To whom correspondence should be addressed. Tel: +44 114 271 2327; Fax: +44 114 271 2882; Email; j.r.sayers@sheffield.ac.uk
REFERENCES