INTRODUCTION

Bacterial conjugation is defined as the transmission of genetic material from a bacterium to another cell by a process requiring intimate cell contact. DNA is transmitted unidirectionally as a single strand in 5 to 3 polarity. The site of initiation lays within the origin of transfer (oriT) at the nic site, a phosphodiester group within a specific dinucleotide sequence subject to cleavage in the course to prepare the DNA for transfer. Proteins involved in this process belong to the DNA processing functions of a transfer system and one of the key enzymes catalyzing the cleavage reaction, known as relaxase, is a site- and strand-specific DNA strand transferase (for review articles, see Lanka and Wilkins (1995) and Pansegrau and Lanka (1996)).

Initiation of DNA processing has been studied intensively in vivo and in vitro using plasmids of the incompatibility group P as a model system (Pansegrau et al., 1994a). The generation of the single DNA strand to be transferred requires the assembly of a relaxosome, the initiation complex of DNA transmission. The complex nucleoprotein structure is formed in a cascade-like process, the first step of which is the binding of the TraJ protein to the 19-bp¹ inverted sequence repetition in oriT (Fig. 1A). TraJ binds asymmetrically by recognizing a 10-bp palindromic sequence (srj) in the right half of the imperfect inverted repeat (Ziegelin et al., 1989). In the second assembly step the relaxase (TraI) binds to TraJ·srj probably by protein-protein interaction between TraI and TraJ and recognition of sri by TraI; sri consists of a 6-bp sequence in the nic region in between the right end of the 19-bp repeat and nic. The nic region was characterized by its necessity for cleavage at oriT and by its sequence conservation among distantly related conjugative DNA transfer systems, particularly by its similarity to the T-DNA border sequences of Ti and Ri plasmids (Pansegrau and Lanka, 1991, 1996; Waters et al., 1991; Pansegrau et al., 1994a). The ternary complex of TraJ, TraI, and negatively supercoiled oriT DNA was sufficient to demonstrate site- and strand-specific cleavage at nic (Pansegrau et al., 1990a) and covalent attachment of the 5 moiety of the T-strand to TraI (Pansegrau et al., 1990a; Waters et al., 1991).

Two coexisting forms of the complex were observed, containing oriT DNA cleaved and uncleaved, indicating the maintenance of a balanced equilibrium. Since the degree of the negative superhelicity of the input DNA was retained in the complexed, cleaved DNA, TraI was proposed to interact non-covalently with the 3 end of the T-strand, holding the ends at nic of the T-strand tightly together. Two additional proteins encoded by the relaxase and leader operon, TraH and TraK, are known to participate in relaxosome formation and to influence cleavage at oriT. TraH, an acidic oligomeric protein, stabilizes relaxosomes by protein-protein interaction by forming a complex with both, TraJ and TraI (Pansegrau and Lanka, 1996); TraH does not bind to DNA by itself. TraK, is a basic tetrameric protein, that interacts specifically with oriT by wrapping srk, a 180-bp region downstream of nic, around a core of TraK (Ziegelin et al., 1992). Binding of TraK in vivo and in vitro increases the fraction of plasmid molecules that can be captured cleaved at nic (Fürste et al., 1989; Waters et al., 1991). The induction of additional supercoiling by the TraK·srk structure may influence the local oriT topology in a way that helps to expose the nic region single-stranded facilitating interaction of TraI with its target site. These data are in agreement with the finding that TraI specifically cleaved and joined single-stranded nic region oligonucleotides (Pansegrau et al., 1993).

The TraI-mediated cleavage reaction at nic of oriT consists of a transesterification initiated by nucleophilic attack of the phosphodiester moiety at nic by the hydroxyl group of TraI tyrosine 22 (Pansegrau et al., 1993). A new phosphodiester bond is formed between TraI tyrosine 22 and the deoxycytidyl-5-phosphoryl at nic resulting in the covalent attachment of the protein to the DNA (Fig. 1, A and B); the second reaction product ends with a deoxyguanosyl residue carrying a free 3 hydroxyl. The equilibrium between joined and cleaved products of single-stranded substrates in a TraI-mediated reaction indicates the reversibility of the reaction and suggests that the relaxase is not only involved in initiation of conjugative DNA transfer but also in termination.

Currently, two versions of the DNA transmission mode are discussed: (i) unwinding of the cleaved strand, the T-strand, and transfer of just one unit length-equivalent to the plasmid, or (ii) unwinding of the T-strand in the presence of elongation at the free 3 hydroxyl. The latter corresponds to a rolling circle type of replication (RCR) generating T-strands of greater than unit length. Hence, recircularization of a one unit lengthplasmid strand would require a second cut or direct reaction between the reoccurring oriT sequence and the 5 bound TraI molecule to release a fully restored monomeric circular T-strand. The process is called ``second cleavage'' mechanism (Wilkins and Lanka, 1993). Genetic evidence for a RCR-like mechanism was recently presented for two plasmid systems, the conjugative plasmid F and the small mobilizable plasmid RSF1010 (Gao et al., 1994; Rao and Meyer, 1994). The assay system measures RecA-independent, intramolecular recombination at oriT sites during mobilization (Erickson and Meyer, 1993).

The gene A protein of phage X174, acting in RCR as a cleaving-joining protein, provides an intriguing economic possibility of DNA strand transfer (Kornberg and Baker, 1992). The gene A protein contains two tyrosines, interspaced by three amino acids, which alternate in catalyzing the cleaving-joining reaction (van Mansfeld et al., 1986; Hanai and Wang, 1993). The cleaving-joining domain of TraI protein consists of at least three motifs: I, II, and III (Fig. 1B) (Pansegrau et al., 1994b; Balzer et al., 1994). Motif I is thought to be involved in target binding, motifs I and III are part of the protein's active center. Active site characterization indicated that both relaxases, RP4 TraI and RSF1010 MobA, rely on only one active site tyrosine for DNA strand transfer (Pansegrau et al., 1993; Scherzinger et al., 1993). Thus, one question to be answered is how many relaxase molecules are involved in one round of transfer?

Here we report the mechanistic dissection of the relaxase catalyzed cleaving-joining reaction by the use of immobilized TraI protein. 3-Biotinylated oligodeoxynucleotides and randomly biotinylated TraI and TraI mutant proteins were synthesized and were bound to magnetic beads via streptavidine. The compounds served to develop a novel assay system for studying the properties of relaxases, particularly to analyze target binding and catalytic activity in one and the same experiment. Immobilized oligonucleotides were applied to demonstrate that the monomeric form of relaxase is insufficient for second cleavage and hence for termination of conjugative transfer DNA replication. Implications for the mechanism of conjugative DNA processing and additional possible applications of the method will be discussed.

EXPERIMENTAL PROCEDURES

TraI, TraJ, TraK, and mutant derivatives of TraI were purified as described previously (Ziegelin et al., 1989, 1992; Pansegrau et al., 1990a, 1994b).

Synthetic oligodeoxyribonucleotides were purified by ion exchange chromatography using a PE TSK DEAE-NPR column (4.6 × 35 mm, Perkin-Elmer) under the conditions recommended by the manufacturer. In the following, the term ``oligonucleotide'' refers to oligodeoxyribonucleotides, unless stated otherwise. Oligonucleotides were labeled at their 5 ends by incubation with [-³²P]ATP in the presence of T4 polynucleotide kinase (New England Biolabs) as described previously (Sambrook et al., 1989). The oligonucleotides used for the second cleavage assay were: 28-mer, CTACTTCACCTATCCTGCCCGGCTGACG; 26-mer, CGGGTGGGCCTACTTCACCTATCCTG; 30-mer, CGGGTGGGCCTACTTCACCTATCCTGCCCG. The construction of the RP4 oriT-containing plasmid pJF142n, which is based on the vector plasmid pBR329, has been described previously (Fürste et al., 1989). Superhelical form I plasmid DNA was isolated by CsCl gradient centrifugation. The DNA was stabilized by extraction with phenol and chloroform.

Prior to biotinylation, aliquots of TraI-preparations were subjected to gel filtration on Superose 12 using Pharmacia FPLC equipment. 1 mg of protein in a volume of 250 µl (20 mM Tris-HCl, pH 7.6, 500 mM NaCl, 0.1 mM EDTA, 50% (w/v) glycerol) was loaded onto the column (10 × 300 mm) equilibrated with buffer B (50 mM sodium borate, pH 9.5, 500 mM NaCl, 0.025% Brij 58). Gel filtration was carried out at a flow rate of 0.5 ml/min. Fractions (0.5 ml) containing TraI or TraI derivatives were adjusted to a final concentration of 50% glycerol in buffer B and stored at 20 °C until use.

For biotinylation, the protein (250 µg) was brought to a volume of 600 µl by addition of buffer B. Sodium hydrogen carbonate (pH 8.0, final concentration: 100 mM) and sulfosuccinimidyl-6-(biotinamido)-hexanoate (Calbiochem, 30 nmol) were added. Thus, the molar ratio of protein and biotinyl cross-linker was 0.1. Following incubation for 16 h at 4 °C biotinylation was stopped by addition of Tris-HCl (pH 7.6, final concentration: 100 mM), and the samples were concentrated to a final volume of 200 µl by spin filtration in a Centricon 30 cartridge (Amicon).

The biotinylated protein was separated from by-products by gel filtration on Superose 12. The column was equilibrated with buffer C (20 mM Tris-HCl, pH 7.6, 500 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol, 10% (w/v) glycerol) and the flow rate was 0.5 ml/min. The pooled peak fractions (1 ml) were adjusted to 50% (w/v) glycerol in buffer C and stored at 20 °C.

Streptavidinconjugated magnetic beads (Dynabeads M-280, Dynal AS, Oslo, Norway) were washed and resuspended in buffer D (20 mM Tris-HCl, pH 8.0, 1 M NaCl, 1 mM EDTA, 0.1% Triton X-100) and biotinylated TraI derivatives were added at a ratio of 10 pmol of protein/6.7 × 10⁶ beads in a final volume of 50 µl. Following incubation for 30 min at 25 °C, the beads were washed three times and resuspended in buffer E (20 mM Tris-HCl, pH 8.8, 5 mM MgCl₂, 50 mM NaCl, 0.1 mM EDTA, 0.1% Triton X-100).

TraI-conjugated beads (equivalent to 10 pmol of TraI) were incubated with 0.5 pmol of the labeled oligonucleotide and 50 pmol of an unlabeled competitor oligonucleotide in buffer E (30 min, 37 °C, final volume: 20 µl). Following magnet separation, the supernatant was removed (non-binding fraction of oligonucleotides) and the beads were washed once in 20 µl of buffer E. The beads were then washed in buffer E + 1% SDS (20 µl, binding fraction of oligonucleotides). Supernatant and SDS wash were analyzed by electrophoresis in 20% polyacrylamide gels containing 8 M urea. The reaction products were quantified by autoradiography of the gels with the storage phosphor technology (Johnston et al., 1990).

Oligonucleotides (28-mer, 300 pmol) were 3-biotinylated by incubation with biotinyl-16-ddUTP (Boehringer Mannheim, 1 nmol) in the presence of terminal deoxynucleotidyl transferase (Amersham Corp., 10 units) and under the conditions recommended by the manufacturer of the enzyme (1 h, 37 °C, final volume: 50 µl). The reaction was stopped by addition of EDTA (pH 8.2, final concentration: 50 mM), and the oligonucleotides were ethanol-precipitated. Oligonucleotides were dissolved in buffer A (20 mM triethyl ammonium acetate, pH 7.0, 0.1 mM EDTA) and separated from nonincorporated nucleotides by reversed phase high pressure liquid chromatography on a C₁₈ column (Waters, Delta-Pak HPI, 3.9 × 150 mm) applying a linear gradient of acetonitrile (10-30% in buffer A, 1 ml/min, total volume 20 ml). Fractions containing the biotinyl-oligonucleotides (biotinyl-29-mer) were lyophilized, and the compounds were redissolved in 10 mM Tris-HCl, pH 8.0, 1 mM EDTA (final concentration: 2 µM).

For coupling of biotinylated oligonucleotides, Dynabeads M-280 were washed twice and resuspended in the original volume of buffer D. The biotinylated oligonucleotide was added at a ratio of 40 pmol/1.34 × 10⁸ beads in a final volume of 250 µl. Following incubation for 30 min at 25 °C with shaking, the beads were washed three times in buffer E and resuspended in 200 µl of buffer E. Immobilized TraI-oligonucleotide adducts were formed by addition of 8 pmol of TraI (10 µl) and incubation for 1 h at 37 °C with shaking. TraI-oligonucleotide adducts were captured by addition of 0.5 M EDTA (pH 8.2) to a final concentration of 50 mM. After 5 min, 5 M NaCl was added to a final concentration of 1 M. The beads were washed two times in 200 µl buffer D. Aliquots (50 µl) were washed twice in buffer E and were resuspended in 25 µl of buffer E with or without MgCl₂. Oligonucleotides (40 pmol) labeled with ³²P at their 5 ends were added (final volume: 50 µl), and the reactions were incubated for 1 h at 37 °C with shaking. Reactions were stopped by washing the beads three times in buffer D. Reaction products were released from the Dynabeads by incubation for 10 min at 90 °C in 90% (w/v) formamide, 2 mM EDTA. Following magnet separation, the supernatants were directly applied to electrophoresis in a 20% (w/v) polyacrylamide gel containing 8 M urea. Reaction products were quantified by autoradiography of gels with the storage phosphor technology (Johnston et al., 1990).

RESULTS

TraI is known to interact specifically with single-stranded oligonucleotides carrying the nic region (sri/nic) of the transfer origin (T-strand) (Fig. 1A). The reaction consists in the cleavage of the oriT oligonucleotide at nic by transesterification via the active-site tyrosine (Tyr-22) of TraI resulting in a covalent protein-oligonucleotide adduct. The remaining part of the oligonucleotide, the sri oligonucleotide carries a free terminal 3 hydroxyl (Pansegrau et al., 1993). To examine the first step of this interaction, recognition of sri and binding by TraI, we have developed an assay that involves immobilized protein, taking advantage of the use of magnetic beads as a matrix (Fig. 2). The set of substrates used in the binding assay consists of 5-labeled oligonucleotides: the T-strand of oriT containing 15 nucleotides upstream of the nic site, carrying sri, and six nucleotides downstream; a T-strand oligonucleotide that was known to be inert to cleavage, containing a G A transition at the 1 position of nic; the 15-mer cleavage product; and two non-cleavable substrates as controls (Fig. 3). Substrate oligonucleotides were incubated with the immobilized relaxase at a molar ratio of 0.05. Additionally, the assay mixtures contained a 100-fold molar excess of competitor oligonucleotides over the labeled substrate oligonucleotide, resulting in an overall molar ratio of oligonucleotides and relaxase of about 5. The competitor oligonucleotide was included in the assay since TraI was found to have a considerable general affinity for ssDNA. This might be explained by the high pI (10.78) of the protein (Pansegrau et al., 1994a). To evaluate the enzymatic activity of immobilized TraI, the cleavage ability was checked. It was found that TraI attached to the beads was still perfectly active and specific, indicating that recognition and binding are unaffected by the chemical immobilization procedure (see Fig. 5, lane a).

The oligonucleotide cleaving-joining reaction mediated by the TraI relaxase is known to be an equilibrium reaction (Pansegrau et al., 1990a, 1993, 1994b). In fact, the process can be subdivided in at least three independent equilibria: (i) binding and release of the substrate oligonucleotide; (ii) cleaving and joining of the bound oligonucleotide and (iii) release and binding of the 5 terminal moiety of the cleaved oligonucleotide (Fig. 2). To dissect these processes, we initially planned to omit Mg²⁺ ions from the reaction to prevent cleaving-joining and hence to study the binding equilibrium separately. However, when we omitted the Mg²⁺ ions, the characteristics of the oligonucleotide binding reaction changed considerably. Binding was not specific for sri/nic-containing oligonucleotides anymore (data not shown), suggesting that Mg²⁺ ions, which are not required for relaxosome formation (Pansegrau et al., 1990a) are nevertheless involved in specific recognition of sri when it is offered in form of an oligonucleotide. In contrast, binding of the wild type 21-mer was highly specific in the presence of Mg²⁺ ions (Fig. 4). Even a single nucleotide exchange at the nic 1 position (G A) resulted in an at least 10-fold reduction of affinity (Fig. 4, second column). We therefore decided to study the binding reaction only in the presence of Mg²⁺, since it obviously provided a better approximation to the in vivo situation.

The Covalently Attached Oligonucleotide Determines the Affinity of the Relaxase for a 3 Terminal sri Site

The affinity of immobilized TraI to the oligonucleotides tested was highest for the wild type 21-mer of oriT (Fig. 4, first column). The TraI binding specificity for sri became obvious by comparing the 30-fold lower affinity of the control oligonucleotide (competitor) with that of the 21-mer wild type (Fig. 4, first and fifth columns). The complementary sequence of the 21-mer wild type binds 3-fold better to TraI than the competitor, which has a similar AT/GC content as the 21-mer of oriT (Fig. 4, fourth and fifth columns). The cleavage product, the 15-mer, of the TraI-catalyzed reaction that corresponds to the 3 end of the T-strand also shows a rather low affinity for TraI (Fig. 4, third column), unless TraI is already attached covalently to the 5 moiety of the T-strand (Fig. 4, first column). This 5-fold increased affinity for the 15-mer to the relaxase indicates that the protein-bound 5 terminus of the T-strand plays an important role in the recognition/binding and probably in the adjustment of the 3 end of the T-strand in the active center of the enzyme.

The relaxase contains at least three motifs that are conserved among other relaxases and proteins predicted to possess relaxase activity (Pansegrau and Lanka, 1991; Ilyina and Koonin, 1992; Pansegrau et al., 1994b; Balzer et al., 1994). The motifs arranged within the N-terminal protein portion form the cleaving-joining domain of the enzymes (Fig. 1B). Motif I belongs to the catalytic center, which contains the active-site tyrosine Tyr-22 (Pansegrau et al., 1993, 1994b). Mutational analyses of conserved residues of the motifs suggested participation of motifs II and III in binding of sri and catalysis of cleaving-joining, respectively (Pansegrau et al., 1994b; Balzer et al., 1994). To evaluate the contribution of each of the motifs to sri binding, we examined a selected set of immobilized TraI mutant proteins with the binding assay (Figs. 1B, 2, 5, and 6A). Each of the motifs contains at least one crucial residue, which, if substituted, reduces the TraI affinity for sri binding strongly, at least 10-fold, i.e., Y22L, H70S, S74A, and H116S (Fig. 5, lanes b/i, c/j, d/k, and g/n), indicating that these residues are indeed involved in recognition and binding of sri.

The enzymatic activities of relaxase include in addition to the cleaving-joining activity a topoisomerase I-like activity, which was first detected in the presence of the TraI mutant protein S74A (Pansegrau et al., 1994b). Enhanced topoisomerase activity of TraI mutants has been related to an impaired interaction of the protein with its target site, sri (Pansegrau et al., 1994b). Thus, three enzymatic properties of the immobilized TraI mutant proteins were assayed to answer the question whether binding and catalytic activity are correlated: (i) specific sri oligonucleotide cleavage, (ii) cleavage at nic of oriT on the double-stranded DNA substrate, and (iii) enhanced topoisomerase activity.

Analyses of the supernatants of the binding experiment showed that single-stranded oligonucleotide cleavage cannot be detected or is reduced to a relative activity of less then 0.1 with four TraI mutant proteins, i.e. Y22L, H70S, D111S, and H116S (Fig. 5, lanes b/i, c/j, e/l, and g/n; Fig. 6B). As shown in the SDS wash, three of these four mutant proteins (Y22L, H70S, and H116S) appear to bind the 21-mer sri/nic substrate at least 7-10-fold less efficiently than wild type TraI (Fig. 5, lanes a/h, e/l, and f/m). The SDS wash has been tested to remove all of the tightly, non-covalently bound material from immobilized TraI. The radioactivity of the SDS-washed beads was found to range around background activity. Hence, the combined binding and cleavage data of D111S suggest that the processes of binding and cleavage can be uncoupled from each other, at least partially. The opposite effect to that seen with D111S was found with S74A, which still gave reasonable good cleavage as compared with wild type TraI, but apparently it bound about 10-fold less tightly than wild type TraI or D111S (Fig. 5, lanes a/h, d/k, and e/l). These two examples, high affinity/little cleavage and low affinity/good cleavage, both provide evidence that recognition/binding and cleavage of sri/nic by TraI are independent processes.

Data obtained with immobilized TraI are in good agreement to those obtained with TraI in solution, since under both conditions no cleavage was found with TraI Y22L and H116S (Fig. 5, lanes b/i and g/n) (Pansegrau et al., 1994b; Balzer et al., 1994). At least for TraI Y22L this is the expected result, because Tyr-22 is the active-site residue the replacement of which by leucine rendered the enzyme inactive.

Recently we have reported that a certain TraI mutant (S74A) in the presence of the accessory protein TraJ exhibits enhanced topoisomerase activity on supercoiled plasmid substrates. The topoisomerase activity is suppressed in the presence of TraK, a protein known to function as a DNA chaperone, thereby reducing the superhelical stress that lasts on the relaxosome (Ziegelin et al., 1992; Pansegrau et al., 1994b). We therefore suggested that an impaired binding of TraI S74A to sri results in an occasional release of the 3 terminus at nic when the plasmid is in the cleaved state, thus leading to the formation of partially relaxed topoisomers.

Applying the oligonucleotide binding assay, TraI S74A, which displays a more than 140-fold enhanced topoisomerase activity (Fig. 6C), behaves consistently with this hypothesis. TraI S74A displays a very low affinity to the 21-mer nic region oligonucleotide and an even more reduced affinity to the 15-mer cleavage product (Fig. 6A). In conjunction with that, a high cleavage activity on superhelical oriT plasmid DNA in the presence of TraK (Fig. 6B) and on the 21-mer nic region oligonucleotide was observed (Fig. 5, lane d). Remarkably, also another mutant in relaxase motif II (TraI H70S) exhibiting equally low affinity to the nic region oligonucleotide and to the 15-mer cleavage product is about 50-fold enhanced in topoisomerase activity above the wild type (Fig. 6C). TraI H70S is almost undisturbed in oriT plasmid cleavage activity (Fig. 6B). The other mutants (D111S, D113S, and H116S) show a moderate enhancement of topoisomerase activity in the range of 10-30-fold (Fig. 6C). Two mutants (S74A and D113S) display another interesting phenotype; TraI S74A and D113S cleave superhelical oriT plasmid DNA only in the presence of both accessory proteins TraJ and TraK. In the absence of TraK, cleavage can barely be detected (Fig. 6B).

One of the long-standing questions in DNA transfer by bacterial conjugation addresses the problem of termination of DNA transmission to a recipient cell, which usually yields a copy identical to that of the transferring plasmid in the donor cell. The intriguing question is whether the recircularization process involves: (i) direct joining of the relaxase-bound 5 moiety of the T-strand with the free 3 hydroxyl of its terminus, implying that just the T-strand of unit length is transmitted between donor and recipient cell; or (ii) the joining reaction taking place on a restored nic region of a substrate that exceeds unit length and was generated by RCR. The latter recircularization mode requires a second cleavage reaction that generates the 3 terminus of the T-strand.

The magnetic bead technique in combination with the nic-specific oligonucleotide cleavage reaction provides an ideal approach to mimic the in vivo termination reaction (Fig. 7). In this experiment the immobilized oligonucleotide was reacted with TraI protein to yield the intermediate, the TraI-oligonucleotide adduct. One advantage of the matrix technique is the possibility to remove the cleavage product, the 17-mer, and unreacted TraI protein very easily by extensive washing. The purified immobilized TraI-oligonucleotide adduct was reacted either with the 26-mer sri oligonucleotide to demonstrate the reversibility of the cleavage reaction, or with a 30-mer oligonucleotide containing sri/nic to proof the second cleavage reaction (Figs. 7 and 8). The sizes of the 5 labeled oligonucleotides were chosen in such a way that the expected reaction products are 38 nucleotides in length. Sufficient reaction product, biotinyl-38-mer, was obtained with the 26-mer, demonstrating that the reverse reaction, the joining of the 5 and the 3 moieties of the T-strand, took place (Fig. 8, lanes e). However, significant amounts of reaction product were not detected by incubating the 30-mer sri/nic oligonucleotide with the TraI-12-mer adduct, suggesting that the TraI monomer is not sufficient to catalyze second cleavage (Fig. 8, lanes c). The weak bands (Fig. 8, lane c) correspond to the input 30-mer oligonucleotide and to an unspecific joining product (biotinyl-42-mer) of the input 30-mer and the covalently attached 12-mer. The occurrence of very small amounts of the expected second cleavage product (biotinyl-38-mer) might be explained by the tendency of TraI to aggregate. Unreacted TraI can be removed only after extensive washing of the beads under high-salt conditions. Therefore, the possibility that traces of unreacted TraI remain attached to the beads and that these might account for production of the observed traces of biotinyl-38-mer appears to be the most likely explanation.

DISCUSSION

The enzymology of conjugative DNA relaxases has been studied extensively in vivo and in vitro (Lanka and Wilkins, 1995; Pansegrau and Lanka, 1996). Isolation of IncP-type relaxosomes from bacterial cells allowed mapping of the cleavage site (nic) within the transfer origin and demonstrated the covalent association of one relaxosome component, the relaxase TraI, with the 5 terminus of the cleaved T-strand (Pansegrau et al., 1990b). Overproduction and purification of the RP4 relaxase and of the relaxosomal accessory proteins TraH, TraJ, and TraK led to the reconstitution of IncP relaxosomes in vitro (Pansegrau et al., 1990a). In vitro experiments with the purified relaxase showed that the enzyme by itself specifically cleaves and joins DNA single strands containing sri/nic. After cleavage, the tyrosine residue 22 of TraI forms the covalent linkage to the T-strand 5 terminus via a phosphodiester bridge (Pansegrau et al., 1993). Generating specific mutants within the three conserved domains of the TraI relaxase, and testing of these mutants in vivo and in vitro, allowed us to assign distinct activities to relaxase motifs I, II, and III (Fig. 1B) (Pansegrau et al., 1994b; Balzer et al., 1994).

By immobilization of either reaction partner, the relaxase or the substrate oligonucleotide, to super-paramagnetic beads, we have extended the spectrum of available methods for examining activities of DNA relaxases in vitro. Two possibilities for biotinyl modifications of proteins are available: (i) in vivo by the synthesis of a naturally biotinated N-terminal fusion peptide of the corresponding protein (Cronan, 1990), and (ii) in vitro by randomly reacting succinimide derivatives of biotin with primary amino groups in the protein. We have chosen the chemical method, because it is known from our earlier work that N-terminal alterations disturb the catalytic activity of TraI. Limited biotinylation of TraI at a theoretical molar ratio of about 10 biotinyl residues per TraI monomer did not interfere notably with any testable activity of the relaxase, i.e. cleaving-joining of sri/nic oligonucleotides, cleavage of superhelical oriT plasmid DNA in the presence of TraJ (and TraK), and topoisomerase I-like activity on superhelical oriT plasmid DNA in the presence of TraJ.

The immobilized TraI protein binds sri/nic oligonucleotides with high specificity. Even a single nucleotide exchange at a crucial position within sri leads to a 10-fold loss of the relaxase's affinity to the respective oligonucleotide (Fig. 4, second column), demonstrating the potential of our binding assay to detect the specific interaction between the relaxase and its substrate.

Immobilization of TraI to the magnetic beads allowed us to dissect functionally ssDNA binding and ssDNA cleaving-joining. Two mutants in motif II TraI H70S and S74A display drastic reductions in binding of sri/nic oligonucleotides, but these mutants are almost undisturbed in the cleavage of superhelical dsDNA in the presence of TraJ and TraK. In both mutants these properties are correlated with an enhanced topoisomerase activity. The enhanced topoisomerase activity of motif II mutants can be explained by the hypothesis that an impaired binding of sri should result in an occasional release of the 3 end at nic by the relaxosome when the DNA is in the cleaved state (Pansegrau et al., 1994b). Following spontaneous relaxation of the plasmid DNA, the relaxase could reseal the cleaved strand. Thus, particularly after prolonged incubation (e.g., 180 min) of superhelical oriT plasmids in the presence of TraI S74A and TraJ, partially relaxed topoisomers accumulate spontaneously in the reaction mixture. In the case of TraI S74A, nearly all the input supercoiled DNA is converted to the covalently closed relaxed form I⁰, indicating efficient cleaving and joining. Since form I⁰ DNA is inert to cleavage, it only appears that TraI S74A cleaves supercoiled oriT DNA in the absence of TraK very poorly (Fig. 6B). This interpretation is supported by the finding that capturing of relaxed intermediates by adding protein denaturants after a short incubation period (e.g. 45 min) results in yields of form II DNA comparable to those obtained with the wild type TraI protein (Pansegrau et al., 1994b). Hence, our experiments confirm that motif II is engaged in the tight binding of the 3 terminus (sri) by the relaxase but most probably is not directly involved in the catalytic cleaving-joining activity of the enzyme.

The motif I mutant TraI Y22L, as expected, is inert in cleaving-joining of any substrate DNA. However, it was quite unexpected that this mutant is also strongly reduced in its affinity to the sri/nic oligonucleotide (Fig. 6A). Possible explanations for the observed effect are as follows. (i) The Tyr to Leu exchange disturbs the secondary structure of the protein in such a way that it does not bind the substrate oligonucleotide. However, this explanation is rather unlikely since TraI Y22L can be applied to assemble stable relaxosomes on superhelical oriT DNA (Balzer et al., 1994). (ii) Tight binding of the oligonucleotide involving motif II takes place only following entry of the oligonucleotide into the active site of the relaxase. This step could involve the interaction of the nucleophilic hydroxyl group of Tyr-22, which is not present in TraI Y22L, with the DNA backbone at the nic site. Hence, tight binding is not observed in TraI Y22L. The same explanation could apply for the low affinity of TraI H116S to substrate oligonucleotides (Fig. 6A). His-116 is thought to be involved either in activation of the Tyr-22 hydroxyl group by proton abstraction (Pansegrau et al., 1994b), or it could mediate the proton transfer reaction to the 3 terminus that creates the leaving 3 hydroxyl group in the cleavage reaction. In both variants of the reaction mechanism, His-116 either contacts the DNA backbone directly or it determines the properties of an amino acid residue that contacts the DNA backbone. Both types of interactions could be crucial for positioning of the DNA within the active center of the relaxase.

Two other motif III mutants, TraI D111S and D113S are almost undisturbed in tight binding of sri/nic oligonucleotides (Fig. 6A). Nevertheless, TraI D111S is severely reduced in cleaving-joining of ss- and dsDNA (Fig. 5, lanes e/l; Fig. 6B), demonstrating that the Asp 111 residue, although not required for positioning of the substrate within the active center of the relaxase, must be involved in the cleaving-joining reaction. Thus, the TraI D111S mutant confirms our hypothesis that motif III is involved in the catalytic activity of the relaxase. TraI D113S displays an even more differentiated phenotype: TraI D113S is nearly undisturbed in binding and cleaving of oligonucleotides (Fig. 5, lanes f/m, Fig. 6A) but cleavage of form I oriT DNA is dependent not only on the presence of TraJ but also on TraK. In the corresponding reaction with the wild type TraI protein only TraJ is essential. TraK is not essential but enhances the yield of cleaved plasmid DNA that can be captured by treatment of relaxosomes with protein denaturants (Pansegrau et al., 1994b). In contrast to the effect with the TraI S74A mutant (see above), the TraI D113S phenotype cannot be due to the conversion of the input form I DNA to form I⁰, since the combination TraJ/TraI D113S displays comparably low topoisomerase activity (Fig. 6C). Since TraK is thought to assist TraJ in facilitating local melting of the sri/nic region in form I oriT DNA (Ziegelin et al., 1992; Pansegrau et al., 1994b), it is likely that another function of the relaxase motif III could consist in invading and stabilizing the melted nic-region. This function might be disturbed in the TraI D113S mutant when only TraJ is present.

Interestingly, Mg²⁺ ions are not only involved in the cleaving-joining reactions but also in specific recognition of a single-stranded substrate DNA (data not shown). This result seems to contradict former data obtained by in vitro reconstitution studies of IncP relaxosomes that indicated that Mg²⁺ ions are not required for site-specific assembly of relaxosomes (Pansegrau et al., 1990a). However, reconstitution of relaxosomes on superhelical dsDNA involves not only the relaxase but also the accessory proteins TraJ and TraH. TraJ has been shown to bind specifically to srj also in the absence of Mg²⁺ ions (Ziegelin et al., 1989) and therefore formation of the TraJ·srj complex might be sufficient to direct the relaxase specifically to sri/nic on dsDNA even in the absence of Mg²⁺. Protein-protein interactions of the TraH protein with TraI and TraJ might then result in a relaxosome, sufficiently stable to be detected by gel electrophoresis (Pansegrau et al., 1990a).

Another interesting result was obtained by incubating the immobilized TraI-12-mer oligonucleotide adduct with the labeled 26-mer sri oligonucleotide (Fig. 7). In the presence of Mg²⁺ ions, this oligonucleotide is efficiently joined to the immobilized 12-mer, demonstrating the catalytic activity of the relaxase-oligonucleotide adduct (Fig. 8, lane e). However, in the absence of Mg²⁺ the oligonucleotide is bound very tightly, although non-covalently, by the immobilized TraI-oligonucleotide adduct (Fig. 8, lane f). The complex between the TraI adduct and the added 26-mer is stable even in the presence of 1 M NaCl, suggesting that the 26-mer and the TraI adduct are captured as a stable non-covalent intermediate. The association is specific for oligonucleotides with sri at the 3 terminus since the corresponding 30-mer containing sri/nic displays no affinity to the TraI adduct under the same conditions (Fig. 8, lane d). Another oligonucleotide with a 3 terminal sri region does not show significant affinity for an immobilized TraI protein lacking the covalently attached T-strand 5 terminus (Fig. 4, third column). Therefore, this result supports our hypothesis that the presence of the covalently attached oligonucleotide moiety at Tyr-22 modulates the affinity of the relaxase for the 3 terminus of the T-strand.

Two classes of alternative models exist for the mechanism of termination of transfer DNA replication in bacterial conjugation: (i) a simple joining reaction requiring that the 3 hydroxyl terminus is not used as a primer for complementary strand synthesis in the donor (Fig. 9A); (ii) mechanisms that require second cleavage at the reconstituted nic site. This reaction could be exerted either by the 5-attached monomeric relaxase, requiring a tandem arrangement of active sites (Fig. 9B) or by a relaxase heterodimer. One of the subunits is covalently attached to the T-strand, whereas the second subunit provides a free active site that catalyzes the second cleavage reaction. The free 3 hydroxyl that is generated in this cleavage reaction in a separate step substitutes the covalently attached relaxase subunit to recircularize the T-strand (Fig. 9C). The second TraI molecule could already be present in the relaxosome, but in an inactive form, or TraI could be part of the membrane-associated DNA transport structure in the donor cell, which provides the exit for the TraI-piloted ssDNA to the recipient.

Our experiments with immobilized TraI-oligonucleotide adducts have shown that the TraI protein under the given conditions cannot catalyze a second cleavage reaction on a sri/nic oligonucleotide (Fig. 8, lane c). Since we know from gel filtration experiments that the TraI protein under the conditions applied here exists as a monomer (Pansegrau, 1991), the alternative given in Fig. 9B is unlikely to apply for TraI.

In fact, the mechanism shown in Fig. 9B is observed in certain ssDNA bacteriophages (e.g. X174) (van Mansfeld et al., 1986; Hanai and Wang, 1993). During viral ssDNA synthesis, a tandem arrangement of tyrosine residues within the gpA protein alternates in cleaving and circularizing the single-stranded phage genomes that are generated by RCR. Obviously, the tandem arrangement of active sites enables the gpA protein to reinitiate RCR after a full round of replication with high efficiency, allowing the synthesis of a large number of closed circular copies of the single-stranded phage genome. In contrast, bacterial conjugation in general results in generation of just a single copy of the DNA molecule that is destined for transfer, making efficient reinitiation mechanisms unnecessary or even undesired. Moreover, a tandem arrangement of active site-tyrosine residues corresponding to that of X174 gpA is absent from the amino acid sequence of TraI (Pansegrau et al., 1994b).

Experiments with the IncQ plasmids (R1162) applying tandem arrangements of oriT show that after conjugative transfer of the respective constructs a large proportion of the plasmid molecules in the recipient cell contain only a single copy of oriT and have lost the DNA located between the two copies (Meyer, 1989). This holds also true if one of the oriT copies is deficient in initiation but proficient in termination of transfer DNA replication, demonstrating that deletion of the DNA between the oriT copies is not due to tandem nicking during initiation but due to a second cleavage mechanism (Bhattacharjee et al., 1992). Similar results have been obtained in the F system (Gao et al., 1994). These findings provide evidence that at least the IncQ and IncFI relaxases are capable of terminating transfer DNA replication by a second cleavage mechanism.

Therefore, we postulate that the TraI relaxase during termination of transfer DNA replication acts in a dimeric form (Fig. 9C). Similar mechanisms are known from plasmids of Gram-positive bacteria, which replicate by a RCR mechanism. In these plasmids (e.g. pT181 and pC221) a dimer of the RCR initiator protein terminates replication after a full round of replication as depicted in Fig. 9C (Thomas et al., 1990). Following termination, a short oligonucleotide remains attached to one of the protein subunits, resulting in an inactive heterodimer (Rasooly et al., 1994b, 1994a). Inactivation of the RCR initiator protein by heterodimer formation is thought to be a means of copy number regulation, preventing over-replication by uncontrolled reinitiation. A similar mechanism could also work in conjugative transfer DNA replication, preventing unproductive transfer of multiple copies of the T-strand to the recipient cell.

The protein immobilization technique via (i) covalently associated oligonucleotides and (ii) limited biotinylated TraI protein is likely to be useful in studying TraI protein-protein interaction between the accessory proteins like TraJ and TraH. The two proteins are known to form stoichiometrically assembled complexes with each other and with the relaxase (Pansegrau, 1991). However, TraK protein, although it influences the cleaving-joining equilibrium at oriT as a DNA chaperone, was proposed to interact with TraI protein indirectly, since the binding site for TraK and the assembly site for the relaxosome are separable (Ziegelin et al., 1992). Thus, TraK will provide a control for investigating protein-protein interactions in the relaxosome and other components of the transfer system. One important example is the TraG protein, which is thought to have a crucial role in the interplay of DNA processing and transport functions (Waters et al., 1992; Lessl et al., 1993). Preliminary studies obtained by the magnetic bead technique indeed indicated complex formation between TraI and TraG supporting earlier genetic evidence.²