RusA proteins from the extreme thermophile Aquifex aeolicus and lactococcal phage r1t resolve Holliday junctions



The RusA protein of Escherichia coli is a DNA structure-specific endonuclease that resolves Holliday junction intermediates formed during DNA replication, recombination and repair by introducing symmetrically paired incisions 5 to CC dinucleotides. It is encoded by the defective prophage DLP12, which raises the possibility that it may be of bacteriophage origin. We show that rusA-like sequences are indeed often associated with prophage sequences in the genomes of several bacterial species. They are also found in many bacteriophages, including Lactococcus lactis phage r1t. However, rusA is also present in the chromosome of the hyperthermophilic bacterium Aquifex aeolicus. In this case, there is no obvious association of rusA with prophage-like sequences. Given the ancient lineage of Aquifex aeolicus, this observation provides the first indication that RusA may be of bacterial origin. The RusA proteins of A. aeolicus and bacteriophage r1t were purified and shown to resolve Holliday junctions. The r1t enzyme also promotes DNA repair in strains lacking the RuvABC resolvase. Both enzymes cleave junctions in a sequence-dependent manner, but the A. aeolicus RusA shows a different sequence preference (3 to TG) from the E. coli protein (5 to CC), and the r1t RusA has relaxed sequence dependence, requiring only a single cytosine.


The RusA Holliday junction resolvase of Escherichia coli is a member of a growing family of structure-specific endonucleases associated with DNA replication, recombination and repair (Sharples et al., 1999; Garcia et al., 2000; Lilley, 2000;Lilley and White, 2000; 2001; Sharples, 2001). It is encoded by the quiescent rusA gene of the defective prophage DLP12 located within the chromosome of E. coli and was discovered when insertions of either IS2 or IS10 upstream of the coding sequence were found to activate its expression and promote DNA repair in mutants lacking the RuvABC resolvase (Mandal et al., 1993; Mahdi et al., 1996). RusA has the hallmarks of a true resolvase. It is a small metal ion-dependent enzyme composed of two identical 14 kDa subunits (Chan et al., 1997; Giraud-Panis and Lilley, 1998). It binds a Holliday junction with high affinity and resolves it to nicked duplex products by introducing paired incisions 5′ to CC dinucleotides located symmetrically about the four-way branched junction structure (Chan et al., 1997). RusA also appears to act alone, which distinguishes it from the prototypic RuvC resolvase. The latter appears to function in vivo only in a ‘resolvasome’ complex with the RuvAB proteins. RuvAB catalyse branch migration and provide RuvC with the support it needs to locate junctions at specific sequences required for strand cleavage (Mandal et al., 1993; Davies and West, 1998; van Gool et al., 1998; Zerbib et al., 1998).

The small size of RusA, its ability to act alone and to retain activity when fused to a nuclear localization sequence make it an attractive candidate for probing Holliday junction processing in vivo. Its ability to function without the direct aid of a branch migration activity is of particular advantage. This has been exploited to show that replication forks arrested by UV-induced lesions in the template DNA form Holliday junctions via RecG- mediated regression of the fork (McGlynn and Lloyd, 2000; McGlynn et al., 2001; Singleton et al., 2001). In fission yeast, it has provided evidence for the branch migration of Holliday junctions by the Rqh1 protein, a RecQ-like DNA helicase (Doe et al., 2000). RusA also complements the meiotic recombination defect of fission yeast cells lacking the Mus81-associated endonuclease (Boddy et al., 2001). This has been used to support the idea that Mus81 is part of a eukaryotic Holliday junction resolvase (Boddy et al., 2001; Chen et al., 2001; Kaliraman et al., 2001). Whether this resolvase is related to the RuvABC-like activity detected in mammalian cells has yet to be established (Constantinou et al., 2001).

The presence of resolvases in all forms of life underlines the importance of these enzymes in genome maintenance and transmission (White et al., 1997; Sharples, 2001). Sequence comparisons have revealed little if any homology between the different classes of resolvases identified to date, indicating that enzymes with such activity have evolved independently on several occasions. However, certain relationships have been reported. Thus, the bacterial RuvC protein shares a conserved catalytic motif with yeast mitochondrial (CceI, Ydc2) and vaccinia virus resolvases. Indeed, the structure of the RuvC active site superimposes well on known and predicted folds in these enzymes (Ariyoshi et al., 1994; Aravind et al., 2000; Lilley and White, 2000; Ceschini et al., 2001). Similar methods have established that the archaeal resolvase Hjc is related to type II restriction endonucleases (Aravind et al., 2000; Kvaratskhelia et al., 2000; Bond et al., 2001; Nishino et al., 2001). The overall shape of Hjc also re-sembles the bacteriophage resolvase T4 endonuclease VII, although there is no obvious sequence homology (Raaijmakers et al., 1999; Bond et al., 2001).

RusA does not share sequence motifs or predicted protein folds with any of the other resolvases or for that matter with any nuclease, which suggests that it has evolved independently (Aravind et al., 2000). Studies of RusA to date have been limited to the enzyme encoded by the DLP12 prophage of E. coli. In this paper, we describe the organization of rusA-like sequences in a variety of genomes and show that the proteins encoded by such sequences in the hyperthermophilic bacterium Aquifex aeolicus and Lactococcus lactis bacteriophage r1t function as Holliday junction resolvases. In phylogenetic terms, A aeolicus is very deeply rooted (Deckert et al., 1998) and is also unique among bacteria in lacking RuvABC (Sharples et al., 1999). RusA may therefore have arisen early on in evolution.


Genome organization of rusA sequences

Database searches revealed that open reading frames (ORFs) with significant homology to the E. coli rusA gene occur in the genomes of several bacteriophages. They are also frequently associated with cryptic prophage sequences in the chromosomes of a variety of bacteria (Sharples et al., 1999; Bolt et al., 2000; Sharples, 2001). Figure 1A shows examples of the organization of these rusA-like genes relative to flanking sequences. All encode proteins that show >40% similarity to E. coli RusA. A number, including those in E. coli K-12 and O157:H7, Neisseria gonorrhoeae and bacteriophages 82, HK022, HK97 and HK620, are located in regions that show similarity to the ninR region of phage λ. The location of rusA in these cases generally correlates with that of the rap gene, which encodes an unrelated endonuclease from lambdoid phages that also targets Holliday junctions (Sharples et al., 1998). Both rusA and rap are located downstream of genes such as ninB, dnaC, O and P involved in DNA metabolism and upstream of lysis genes that resemble λQ, R and S. This arrangement suggests that the ninR region specifies factors critical for DNA processing and that RusA and Rap may constitute alternative endonucleases for processing joint molecule intermediates formed during DNA replication, recombination and repair.

Figure 1.

RusA genes and proteins.

A. Molecular organization of rusA in phage and bacterial genomes. Genes are represented by arrows that indicate their size and transcriptional orientation. Related genes are shown with the arrows shaded in the same pattern. Others are labelled with the names of homologous genes or, in the case of phage sequences, with the orf numbers given in the database entry. Potential genes with no matches in the database are unlabelled. Dashed lines indicate abbreviated sections. Eco, Escherichia coli; 82, lambdoid phage 82; Spa, Salmonella paratyphi; HK, Hong Kong phages 022 and 97; Ngo, Neisseria gonorrhoeae; r1t, Lactococcus lactis phage r1t; Efa, Enterococcus faecalis, Spy, Streptococcus pyogenes, Bsu, Bacillus subtilis skin cryptic prophage; PVL, Staphylococcus aureus phage φPVL; Aae, Aquifex aeolicus. The N. gonorrhoeae genome has three copies of rusA, only one of which is shown. Likewise, only one of two rusA homologues in both E. faecalis and S. pyogenes is shown.

B. Alignment of RusA proteins from E. coli, A. aeolicus and phage r1t. Conserved residues are in bold, and residues essential for catalysis by the E. coli enzyme are marked with a grey dot.

Gram-positive bacteria lack the characteristic bacterial Holliday junction resolvase, RuvC, although they do have the RuvAB branch migration system (Sharples et al., 1999). However, RusA-like sequences associated with remnant phage sequences occur in most Bacillus, Strep-tococcus, Staphylococcus, Enterococcus and Lactococcus species. These may substitute for RuvC in genetic recombination and DNA repair.

A rusA-like sequence is also present in the genome of the extreme thermophile A. aeolicus, a eubacterial species of ancient lineage (Deckert et al., 1998). How-ever, there is no evidence in this case for association with phage-like sequences. It is flanked by rnhB and rplS on one side and lplA on the other. These genes code for RNase HII, ribosomal protein L19 and lipoate protein ligase, respectively, which are typical bacterial enzymes. Given the ancient lineage, this typically eubacterial organization provides the first indication that rusA may have arisen early during the evolution of bacteria. It may be significant that A. aeolicus is unique among eubacteria in lacking both RuvAB and RuvC (Sharples et al., 1999). It may therefore rely on RusA to resolve Holliday junctions.

Bacteriophage r1t RusA promotes DNA repair

Figure 1B shows the sequence for E. coli RusA aligned with that predicted for the RusA proteins from A. aeolicus and L. lactis bacteriophage r1t. Several residues critical for catalysis by E. coli RusA have been identified (Bolt et al., 1999; 2000). These are conserved in Aae and r1t RusA, which suggests that both proteins may be functional. To test this directly, we investigated their ability to promote DNA repair in E. coli cells lacking the RuvABC Holliday junction resolvase. Previous studies established that multicopy plasmids encoding E. coli RusA are able to promote survival of UV-sensitive ruv mutants (Mahdi et al., 1996). The pXNA plasmid construct, which carries an 8.1 kb DNA fragment from phage r1t that includes rusA cloned in a BlueScript vector (van Sinderen et al., 1996), was therefore introduced into closely related wild-type and ruv mutant strains. Figure 2A shows that the r1t RusA clone promotes efficient survival of both ruvA and ruvC strains irradiated with UV light and has little or no effect on a ruv+ control. Indeed, it proved more effective than the E. coli RusA construct, pAM151, which tends to have a negative effect on the survival of a ruv+ strain as a result of overexpression of RusA (Mahdi et al., 1996).

Figure 2.

DNA repair in E. coli promoted by RusA.

A. Survival of UV-irradiated cells carrying plasmids encoding E. coli (pAM151) or phage r1t (pXNA) RusA or vector controls (pT7-7 or pBS-SK– respectively). The strains used and symbols are shown.

B. Restriction map of the phage r1t rusA (orf14) region cloned in pGS886. The size (kb) of each restriction fragment is indicated. ORFs are numbered according to the database entry for phage r1t (van Sinderen et al., 1996).

C. Structure of plasmids carrying fragments of the phage r1t genome encoding rusA and their effect on survival of UV-irradiated cells. The vector plasmid is indicated in each. The arrows show the direction of transcription from the vector lac promoter. The columns of numbers on the right show the fraction of N2057 (ruvA) and AB1157 (ruv+) cells irradiated on the surface of LB agar plates with 30 J m−2 UV light that survive to form colonies.

D. Translation of the r1t rusA sequence. The 3′ end of orf13 and the 5′ end of rusA are shown. Possible ATG and GTG initiation codons for translation of rusA are in bold and numbered 1 and 2 respectively. Potential ribosome binding sites (RBS) associated with these initiation codons are also indicated. The C-terminal sequence of the Orf13 product is shown. The translation of rusA is shown from the GTG initiation codon, which is consistent with the determined sequence of the first eight N-terminal amino acids of the purified protein.

To confirm that the increased survival promoted by pXNA results from the rusA gene, we tested various fragments cloned from the pXNA insert (Fig. 2C). All constructs carrying rusA promote efficient survival of ruvA strain N2057 when the coding sequence runs counter to the vector lac promoter (clones pGS886, 887, 890 and 892). Furthermore, they have no negative effect on survival when introduced into the wild-type strain, AB1157. However, a construct carrying rusA under the control of the lac promoter has a negative effect on AB1157, and this is associated with reduced ability to promote survival of N2057 (clone pGS893). The negative effect in AB1157 is eliminated in constructs deleted for a 100 bp sequence upstream of rusA that interrupts orf13. This also eliminates any ability to promote survival of N2057 (clones pGS878 and pGS879). The 5′ end of the rusA region reveals two possible initiation codons, each associated with a potential ribosome binding site (RBS) for translation of rusA transcripts (Fig. 2D). N-terminal sequencing of the protein purified using pGS893 (see below) established that translation initiates at the second site (GTG; data not shown). This GTG codon lies downstream of the BclI restriction site used to construct pGS878 and pGS879 (Fig. 2A and B). However, the associated potential RBS is a poor match to the consensus (AGGAGG), which might prevent pGS878 and pGS879 from ex- pressing sufficient RusA to promote repair. We therefore inserted the coding sequence for r1t rusA beginning with the GTG start codon into pT7-7, such that transcription from the vector promoter would link rusA to a strong RBS. The resulting construct, pGS906, was able to promote the survival of strain N2057 and had a slight negative effect on the survival of strain AB1157 (Fig. 2C). These data demonstrate that expression of rusA alone is sufficient to promote DNA repair in the absence of RuvABC.

We also tested various clones encoding the A. aeolicus RusA. However, none is able to promote the survival of strain N2057 at either 37°C or 42°C (data not shown). Aquifex aeolicus grows optimally at 95°C, and enzymes from this organism may not function at temperatures that allow growth of E. coli.

Binding and cleavage of Holliday junctions

Phage r1t RusA was purified as described in Experi-mental procedures. The A. aeolicus protein was more problematic. We made a number of different constructs, but most failed to express RusA, including a clone de-signed to produce RusA with a hexahistidine tag at the N-terminus. The only clone that expressed detectable levels of RusA was a fusion construct in which maltose-binding protein (MBP) was fused to the N-terminus of RusA. We used this clone to purify MBP–RusA (Experimental procedures). Enzymes fused to MBP can often be liberated by factor Xa-catalysed proteolysis. In our hands, this was not possible in the case of Aae MBP–RusA, and the fusion protein was therefore used for biochemical studies.

Escherichia coli RusA binds with high affinity to Holliday junctions. However, in the absence of Mg2+, bandshift assays reveal a ladder of complexes consistent with RusA binding not only at the branch but also to the duplex arms (Fig. 3A, lanes i–j) (Chan et al., 1998). A similar ladder of complexes is observed with the r1t RusA (Fig. 3A, lanes b–g). In contrast, E. coli RusA forms a single well-defined complex in the presence of Mg2+ (Fig. 3B, lanes b–d) (Bolt et al., 1999). A complex with the same mobility is detected with r1t RusA (Fig. 3B, lanes e–g). In these two cases, binding in Mg2+ also leads to some junction cleavage (Fig. 3B, lanes b–g; see below). The Aae MBP–RusA binds junction DNA in the presence of Mg2+ (Fig. 3B, lanes h–j). However, a stable complex is detected only if binding is conducted at 37°C or higher temperatures (data not shown). As might be expected, the complex formed shows reduced mobility because of the additional 42.7 kDa from MBP.

Figure 3.

Holliday junction binding by RusA proteins.

A. Bandshift assay showing binding of phage r1t and E. coli RusA to Holliday junction J3 in the presence of 5 mM EDTA. Reactions contained 0.5 nM 32P-labelled J3 and no RusA protein (lanes a and h) or r1t RusA at 7, 14, 28, 56, 112 or 224 nM (lanes b–g) or E. coli RusA at 10 or 100 nM (lanes i and j).

B. Bandshift assay showing binding of RusA to J3 in the presence of 2 mM Mg2+. Reactions contained 0.5 nM 32P-labelled J3 and no RusA (lane a) or E. coli RusA at 10, 20 or 50 nM (lanes b–d) or r1t RusA at 45, 90 or 180 nM (lanes e–g) or Aae MBP–RusA at 10, 225 or 450 nM (lanes h–j). Binding conditions were standard except that the Aae MBP–RusA reactions were incubated at 37°C before loading on the gel.

To investigate resolvase activity, the RusA enzymes were incubated with a synthetic junction substrate (J3) in which CC dinucleotide targets for cleavage by E. coli RusA are located within a 3 bp mobile core at the branch point (Chan et al., 1997). The r1t RusA gives a single product indistinguishable from the nicked linear duplex product of resolution by E. coli RusA (Fig. 4A). Similar results were obtained using a different junction (J11) that has CC dinucleotide targets located within a mobile core of 11 bp (Fig. 4B). Additional trace products consistent with the removal of a single arm of the junction are also visible with J11, although not with J3. However, these are also seen with the E. coli protein, which indicates that such aberrant cleavage events are not specific to r1t RusA. Furthermore, the activity of r1t RusA is reduced at high concentrations of the protein. This effect is also characteristic of E. coli RusA (Chan et al., 1997).

Figure 4.

Holliday junction cleavage by RusA.

A. Cleavage of J3 by E. coli and phage r1t RusA proteins. Reactions contained 0.5 nM 32P-labelled J3 and no RusA (lanes a and h) or 14, 28, 56, 112, 224 or 448 nM phage r1t RusA (lanes b–g) or 10 or 100 nM E. coli RusA (lanes i and j).

B. Cleavage of J11 by E. coli and phage r1t RusA proteins. Reactions contained 0.5 nM 32P-labelled J11 and no RusA (lanes a and h) or 14, 28, 56, 112, 224 or 448 nM phage r1t RusA (lanes b–g) or 10 or 100 nM nM E. coli RusA (lanes i and j).

C. Cleavage of J3 by Aae MBP–RusA. Reactions contained 0.5 nM J3 32P-labelled in strand 1 and no RusA (lane a), 100 nM E. coli RusA (lane b), 450 nM Aae MBP–RusA (lane c) or 180 nM phage r1t RusA (lane d). The reaction with Aae MBP–RusA was incubated at 60°C. Otherwise, standard reaction conditions were used. Reaction products in (A–C) were analysed on native gels.

As A. aeolicus is thermophilic, assays with the Aae MBP–RusA were conducted over a range of temperatures up to 65°C, the maximum that can be used without dissociating the substrate DNA. No activity could be detected on J3 at 30–45°C (data not shown), but substantial cleavage was seen at 60°C (Fig. 4C, lane c). The product generated migrates more slowly than the nicked linear duplex generated by the E. coli (Fig. 4C, lane b) and r1t (Fig. 4C, lane d) enzymes, suggesting some difference in the nature of the cleavage. Activity is also detected on other mobile junctions cleaved by the E. coli protein. However, very little resolution activity is detected using a static junction (see below).

Sequence specificity of junction cleavage

The E. coli RusA protein resolves Holliday junctions by introducing paired incisions directed almost exclusively 5′ to CC dinucleotides located symmetrically about the branch point. To determine whether the r1t and Aae enzymes catalyse a similar sequence-specific reaction, we mapped the strand cleavage sites. The results are presented in Fig. 5. Figure 5A and B shows that the two enzymes cleave J11 at sites that differ from those cleaved by the E. coli protein. The cleavage sites were mapped by reference to sequencing ladders of each strand, and the results are summarized in Fig. 5C. The r1t RusA cleaves J11 at one major site on each of the four strands, unlike the E. coli protein, which cleaves predominantly at symmetrical sites in strands 2 and 4, as shown previously (Chan et al., 1997). The r1t RusA incisions on strands 2 and 4 are orientated symmetrically but are located between the CC dinucleotides, i.e. 1 bp away from the sites of cleavage by E. coli RusA. The main incisions in strands 1 and 3 are also located symmetrically and occur 5′ of a CT. The E. coli enzyme cleaves very inefficiently at these sites. Dual incisions in either strands 1 and 3 or strands 2 and 4 account for the efficient resolution of the J11 Holliday junction structure by r1t RusA.

Figure 5.

Mapping RusA cleavage sites.

A and B. Sequencing gels showing cleavage of strands 1–4 in Holliday junction J11. Reactions contained 0.7 nM J11 32P-labelled in the strand indicated above the lanes and 100 nM of the indicated RusA protein. Reactions with Aae MBP–RusA were incubated at 60°C. Lanes headed m are Maxam–Gilbert A+G sequencing ladders for each strand.

C. Location of RusA cleavage sites on J3, J11 and J0. The DNA sequences at the branch point of each junction are shown. Strands are numbered as for (A) and (B) Homologous cores in J11 (11 bp) and J3 (3 bp) are boxed. The sites cleaved are marked with arrows, and the RusA protein is identified by shading. Reaction conditions for mapping cleavages in J3 and J0 were as for J11.

The Aae RusA also cleaves J11 on strands 1 and 3. Two pairs of symmetrical cleavage sites were identified, which account for the resolution of the structure (Fig. 5B and C). The more minor cleavage site is located 5′ to the same CT targeted by the r1t enzyme. However, the major activity is 3′ to a G. This different sequence preference of A. aeolicus RusA was confirmed using J3 as a substrate. Cleavage in this case is 3′ to TG dinucleotides in strands 2 and 4 (Fig. 5C). These sites are outside the core of homology, and their locations relative to the branch point differ by 1 bp. Thus, although J3 is cleaved by Aae RusA, the two products are not regular nicked linear duplexes. The one detected by a 5′ end label in strands 1 or 4 has an extra nucleotide at the nick. This accounts for the slightly slower migration of the labelled cleavage product on a native gel (Fig. 4C). The other product, detected by labelling strand 2 or 3, lacks a nucleotide at the nick. It should therefore migrate faster than the product generated by E. coli RusA. This is indeed what we found (data not shown). It is significant that, although Aae RusA cleaves 3′ to a guanine in J11, it fails to cleave 3′ to either of the two guanines within the homologous core of J3. This indicates that either a single guanine is not sufficient for target site recognition or the scissile bond needs to be 3 bp or more from the branch point. The r1t RusA cleaves J3 at exactly the same sites as the E. coli enzyme.

From these data, we conclude that the r1t and Aae RusA proteins are both able to resolve Holliday junctions by catalysing a dual-strand incision reaction. However, it would appear the r1t enzyme is less strict in its sequence selectivity than the E. coli enzyme. We therefore tested whether it could catalyse strand incisions in a static substrate (J0) that has no dinucleotide sequences located symmetrically about the branch point. The E. coli enzyme has very weak activity 5′ to the CC in strand 3 (Fig. 5C). Significantly more activity is detected with the other RusA proteins, especially with the r1t enzyme. The sites cleaved by r1t RusA are also distributed through all four strands. These results are consistent with r1t RusA having somewhat relaxed sequence selectivity. However, the strand cleavage activity is very low compared with that on J3 and J11, and resolution products are detectable on native gels only with high levels of protein (data not shown). The Aae RusA also cleaved J0, but the activity was low and restricted to strands 2 and 4. However, cleavage is directed 3′ to TG in strands 2 and 4, which, given the activity on J3, supports the idea that a TG dinucleotide is important for target site recognition by Aae RusA.

Cleavage of branched DNA molecules by r1t RusA

The reduced sequence specificity of the r1t enzyme and its significant activity on J0 led us examine whether it might also cleave other DNA substrates that lack the four-way branched structure of a Holliday junction. Figure 6A shows that it binds well to a variety of such structures in the presence of Mg2+, forming a single well-defined complex in each case. It also cleaves these structures, producing a variety of products consistent with incision of the DNA at or near the branch point. However, the activity is poor compared with the very efficient resolution of a Holliday junction. This reinforces the idea that efficient catalysis by the subunits of RusA requires two identical cleavage sites located symmetrically about the branch point.

Figure 6.

Binding and cleavage of branched DNA molecules by phage r1t RusA.

A. Bandshift assay showing binding in the presence of 2 mM Mg2+. Reactions contained 0.5 nM of the structure indicated, each 32P labelled on one strand (*), and phage r1t RusA at 55, 110 or 220 nM.

B. Native gel showing cleavage of branched DNA by r1t RusA. Reactions contained 0.5 nM of the structure indicated, each 32P labelled on one strand (*), E. coli RusA at 10 or 100 nM and phage r1t RusA at 110 or 220 nM.


We have shown that sequences in the chromosome of the bacterium A. aeolicus and in L. lactis bacteriophage r1t homologous to the E. coli rusA gene encode proteins that most probably function as Holliday junction resolvases in these organisms. The proteins encoded by the rusA homologues were purified and shown to cleave Holliday junction substrates at specific sequences located symmetrically about the branch point. Such cleavage is characteristic of specific Holliday junction resolvases. In the case of r1t RusA, this conclusion is further supported by the fact that the enzyme promotes survival of UV- irradiated E. coli cells lacking the normal resolvase, RuvABC. However, its overexpression in wild-type cells has a marked negative effect, suggesting that it may also catalyse inappropriate cleavage of structures that are necessary for survival, such as replication intermediates. Indeed, our in vitro studies show that it has a somewhat relaxed sequence preference relative to the E. coli enzyme and cuts branched DNA structures that lack the symmetry of a Holliday junction. These properties may reflect its role in the life cycle of bacteriophage r1t, where they might provide the means for RusA to cleave branched structure before DNA packaging, as is the case with T4 endonuclease VII and T7 endonuclease I (Pottmeyer and Kemper, 1992; Giraud-Panis and Lilley, 1997; Declais et al., 2001). In recent studies (E. L. Bolt and R. G. Lloyd, manuscript submitted), we have shown that catalysis by the two subunits of E. coli RusA can be uncoupled. If the subunits of r1t RusA can be similarly uncoupled, the enzyme may be able to cleave branched structures via a ‘nick and counter-nick’ mechanism, as proposed for T4 endonuclease VII (Pottmeyer and Kemper, 1992). Such activity may explain the trace products seen in resolution assays with certain junction substrates (Fig. 4B).

The evolutionary origin of RusA is unclear (Aravind et al., 2000). Our analysis of rusA homologues revealed that nearly all are found in bacteriophages or are associated with prophage or cryptic prophage sequences in bacterial chromosomes. The one exception is the A. aeolicus rusA, for which there is no evidence of a bacteriophage ancestry. It is therefore possible that rusA is an ancient bacterial gene that has subsequently been acquired by bacteriophages. This would certainly explain its presence in the E. coli chromosome where its activity would be superfluous given the presence of RuvABC.

We were able to study the A. aeolicus RusA only as a fusion to E. coli MBP. It resolves junctions by introducing paired strand incisions preferentially 3′ to a TG. It has no activity at the canonical site (5′ to CC) cleaved by E. coli RusA. How this altered sequence preference is determined is not clear. It is unlikely that the MBP tag is responsible. E. coli MBP–RusA and MBP–RuvC retain their usual sequence preferences (Shah et al., 1997; Giraud-Panis and Lilley, 1998). However, this possibility cannot be excluded, as we were unable to generate active, untagged Aae RusA, and the fused MBP may denature at the temperatures (60°C) required to detect catalysis. The amino acids essential for catalysis by E. coli RusA are conserved in the A. aeolicus enzyme (Fig. 1B) and are therefore unlikely to be involved in sequence recognition. Further analyses of the three RusA enzymes may increase our understanding of how sequence specificity is determined. Engineering of the A. aeolicus RusA may generate a robust resolvase for use at high temperatures.

Experimental procedures

Strains, media and general methods

The E. coli K-12 strains JM101 [F′ (F128) proAB+lacIqZΔM15] and AB1157 (ruv+rusA+), and the AB1157 derivatives N2057 (ruvA60::Tn10) and CS85 (ruvC53) have been described previously (Mahdi et al., 1996). pAM151 carries E. coli rusA cloned in pT7-7 (Mahdi et al., 1996). LB broth and agar media were used for bacterial culture. Strains carrying plasmids were grown in media containing 100 μg ml−1 ampicillin. Media recipes and procedures for measuring the survival of UV- irradiated cells have been described previously (Al-Deib et al., 1996). Survival values are the means of two or three independent experiments.

Cloning of rusA from A. aeolicus and phage r1t

Chromosomal DNA from A. aeolicus (Deckert et al., 1998) was obtained from the Department of Microbiology, University of Regensburg, Germany. The rusA sequence (ORF Aq 1953) was amplified by polymerase chain reaction (PCR) using the primers 5′-ATGGAAGAAATTGAGCTTTACCTTTC-3′ and 5′-CCGCTATTGACCTAGGAGTCGTGAG-3′, and cloned into pMAL-c2 (New England Biolabs) The resulting construct, pEB257, fuses the rusA coding sequence to the 3′ end of the MBP gene. The rusA gene of phage r1t (ORF14) was obtained from pXNA, which carries an 8 kb NheI–XbaI r1t genomic DNA inserted in pBlueScript SK– (van Sinderen et al., 1996). pGS886 is a HindIII deletion derivative of pXNA that retains a 3.7 kb insert of r1t sequences as shown in Fig. 2B. Various sections of the region carrying rusA were cloned in pUC18 and pUC19 to make the other constructs shown in Fig. 2C, using the restriction sites in the map annotated in Fig. 2B. pGS906 carries the phage r1t rusA coding sequence cloned in pT7-7. The rusA insert extends from the GTG initiation codon (Fig. 2D, initiation codon 2) to the stop codon and was generated by PCR using rusA-specific primers incorporating NdeI (5′rusA primer) and BamHI (3′rusA primer) sites to facilitate cloning into NdeI–BamHI-digested pT7-7.

Overexpression and purification of RusA proteins

The MBP–RusA from A. aeolicus was purified from a 2 l culture of JM101 harbouring pEB257 grown at 37°C in LB broth supplemented with ampicillin (100 μg ml−1) and con-taining 0.2% (w/v) glucose to suppress maltose biosynthesis. At an A650 of 0.4, IPTG (1 mM final concentration) was added to induce expression of MBP–RusA, and incubation con-tinued for a further 2 h before harvesting the cells. All sub-sequent procedures were performed on ice, unless stated otherwise. The harvested cells were resuspended in 10 ml of buffer A (20 mM Tris, pH 7.4, 200 mM NaCl, 1 mM EDTA and 10% glycerol), broken by sonication and cell debris removed by centrifugation. The supernatant was diluted to 50 ml and loaded onto a 10 ml amylose column equilibrated with buffer A. Bound proteins were washed with 50 ml of buffer A and then eluted with 30 ml of buffer A containing 10 mM maltose. Fractions containing MBP–RusA were dialysed into buffer C (20 mM Tris, pH 8.0, 100 mM KCl, 1 mM EDTA and 10% glycerol), loaded on a 10 ml heparin column and bound proteins eluted with a linear gradient of 0.1–1.0 M KCl in buffer C. The MBP–RusA eluted in a single peak and appeared to be free from contaminants. Peak fractions (0.3–0.4 M KCl) were pooled (5 ml), dialysed into storage buffer (20 mM Tris, pH 8.0, 1 mM EDTA, 100 mM NaCl and 30% glycerol) and stored in small aliquots at −20°C.

For phage r1t RusA, broth cultures of a ΔrusA::kan deriva-tive of strain JM101 carrying pGS893 (Fig. 2C) were treated with IPTG to induce expression of the cloned rusA gene from the vector lac promoter, and the RusA protein was purified from the induced cells using the procedures established for E. coli RusA (Sharples et al., 1994). Protein concentrations were determined by the Bradford assay (Bradford, 1976) using bovine serum albumin (BSA) and E. coli RusA as standards and are expressed as mol of the monomeric species.

DNA substrates

All DNA substrates were prepared by annealing oligonucleotides, one of which was labelled at the 5′ end with 32P (Parsons et al., 1990). Holliday junction substrates J0, J3 and J11 have been described previously (Chan et al., 1997). Sequences of the oligonucleotides used for branched substrates 1–5 in Fig. 6 are given in Table 1.

Table 1. Sequences of branch substrate oligonucleotides.
SubstrateStrandSequence (5′–3′)
31(*)As substrate 1
 2As substrate 1
41(*)As substrate 2
 2As substrate 2
 3As substrate 3
51(*)As substrate 1
 2As substrate 1 strand 3

DNA cleavage and binding assays

For E. coli and phage r1t RusA cleavage assays, standard reactions (20 μl) were carried out as described and contained 32P-labelled DNA and RusA protein as indicated (Bolt et al., 1999). The same procedures were used for A. aeolicus RusA except that the standard reactions were incubated at 60°C. Reaction products were analysed by electrophoresis on native or denaturing polyacrylamide gels followed by phosphorimaging and autoradiography. Strand cleavage sites were mapped on sequencing gels by reference to a Maxam–Gilbert A+G sequencing ladder of the same labelled oligonucleotide. DNA-binding assays were performed as described and contained 5 mM EDTA or 2 mM MgCl2 as indicated (Bolt et al., 1999), except that binding reactions for A. aeolicus RusA were incubated at 37°C. Subsequent electrophoresis was at room temperature.


We are very grateful to Karl Stetter, Robert Huber and Thomas Hader for genomic DNA and cell biomass from Aquifex aeolicus, and Arjen Nauta for providing the pXNA clone of bacteriophage r1t. We also thank Carol Brown and Lynda Harris for excellent technical assistance, and Tom Meddows for helping to construct pEB257. This work was supported by a programme grant from the Medical Research Council to R.G.L. and G.J.S.