Characterization of IntA, a Bidirectional Site-Specific Recombinase Required for Conjugative Transfer of the Symbiotic Plasmid of Rhizobium etli CFN42

Abstract

Site-specific recombination occurs at short specific sequences, mediated by the cognate recombinases. IntA is a recombinase from Rhizobium etli CFN42 and belongs to the tyrosine recombinase family. It allows cointegration of plasmid p42a and the symbiotic plasmid via site-specific recombination between attachment regions (attA and attD) located in each replicon. Cointegration is needed for conjugative transfer of the symbiotic plasmid. To characterize this system, two plasmids harboring the corresponding attachment sites and intA were constructed. Introduction of these plasmids into R. etli revealed IntA-dependent recombination events occurring at high frequency. Interestingly, IntA promotes not only integration, but also excision events, albeit at a lower frequency. Thus, R. etli IntA appears to be a bidirectional recombinase. IntA was purified and used to set up electrophoretic mobility shift assays with linear fragments containing attA and attD. IntA-dependent retarded complexes were observed only with fragments containing either attA or attD. Specific retarded complexes, as well as normal in vivo recombination abilities, were seen even in derivatives harboring only a minimal attachment region (comprising the 5-bp central region flanked by 9- to 11-bp inverted repeats). DNase I-footprinting assays with IntA revealed specific protection of these zones. Mutations that disrupt the integrity of the 9- to 11-bp inverted repeats abolish both specific binding and recombination ability, while mutations in the 5-bp central region severely reduce both binding and recombination. These results show that IntA is a bidirectional recombinase that binds to att regions without requiring neighboring sequences as enhancers of recombination.

INTRODUCTION

Site-specific recombinases are a set of DNA-breaking and -rejoining enzymes that play a pivotal role in bacterial genome plasticity. All of them perform recombination between DNA segments, independently of RecA, by recognizing and binding to short (<50-bp) DNA sequences. Based on amino acid sequence alignments, the presence of characteristic amino acids in the active site, and catalytic mechanisms, site-specific recombinases have been grouped into two families: the tyrosine family and the serine family. These two families are unrelated to each other, with different protein structures and reaction mechanisms (1). Most site-specific recombinases require additional host factors for efficient catalysis.

The tyrosine recombinase family is the one most represented in bacterial genomes. A recent survey identified over 1,300 gene sequences belonging to the family (2). Tyrosine recombinases catalyze recombination between substrates that share limited sequence identity. The sequence identity usually extends over the short strand exchange region and flanking recombinase-binding sites (inverted repeats). DNA homology within the 6- to 8-bp region between the strand cleavage sites, called the overlap, spacer, or crossover region, is critical for the recombination reaction in most (3, 4), but not all (5), cases studied.

Members of the tyrosine recombinase family catalyze a variety of sequence-specific DNA rearrangements in biological systems, including the integration and excision of phage genomes, such as the phage λ integrase (6), the yeast Flp recombinase (7), the phage P1 Cre recombinase (8, 9), and the Escherichia coli XerC/XerD recombinases (10, 11), into and out of their bacterial hosts. Although the tyrosine recombinase family is the best understood among the recombinases, both structurally and biochemically, it is still difficult to predict the function (i.e., integration versus excision) based solely on the primary sequence. Moreover, it is a structurally diverse family, where over 56 subfamilies (containing at least four elements each) have been identified (2). Additionally, although tyrosine recombinases have been found in almost every sequenced bacterium, functional characterization has been concentrated mostly in the Enterobacteriaceae.

For instance, in alphaproteobacteria, a very diverse and ecologically important bacterial class, only two representatives of the tyrosine family and another from the serine family have been characterized. Among these, phage 16-3 is a temperate phage of Sinorhizobium meliloti 41 that integrates its genome with high efficiency into the host chromosome by site-specific recombination between the attachment regions attB and attP using the tyrosine recombinase Int (12). In Mesorhizobium loti R7A, integration of the symbiosis island (an integrative conjugative element) into the Phe-tRNA gene is catalyzed by the tyrosine recombinase IntS (13); interestingly, excision of the symbiosis island requires, besides IntS, the recombination directionality factor RdfS (13). For the serine family, a site-specific recombinase (RinQ) can be found as part of the plasmid multimer resolution system, whose target site is a locus that participates in the incompatibility with p42d of Rhizobium etli (14).

R. etli CFN42 is a soil alphaproteobacterium able to induce nitrogen-fixing nodules on the roots of bean plants. The strain contains six plasmids (p42a to p42f), whose sizes range from 185 to 643 kb. Plasmid p42d is the symbiotic plasmid (pSym), carrying most of the information required for nodulation and nitrogen fixation; p42a is self-transmissible at high frequency and is indispensable for conjugative transfer of the pSym (15, 16, 17). The requirement for p42a for conjugative transfer of the pSym in an otherwise wild-type strain was striking, given the existence on the pSym of a full traACDG-virB1-virB11 system functional for conjugation (18, 19). This system, however, is kept transcriptionally silent, under all conditions tested, by the action of the strong repressor RctA (19, 20). Operation of the pSym conjugative system was detected only upon inactivation of the repressor RctA or constitutive expression of the rctB gene (19, 20).

The requirement for p42a for transfer of the pSym was uncovered by sequence analysis and characterization of conjugative products. Sequence analysis revealed a 53-bp region that is 90% identical between the pSym and p42a, including a 5-bp central region flanked by 9- to 11-bp inverted repeats, reminiscent of bacterial and phage attachment sites (21). A gene encoding an integrase-like protein belonging to the tyrosine recombinase family (intA) was localized downstream of the attachment site on p42a (21). We have shown previously that cointegration of the pSym and p42a, mediated by site-specific recombination between the attachment-like sites present on the pSym (attD) and p42a (attA), participates in conjugative transfer of p42d (21). The pSym-p42a cointegrate is able to perform conjugative transfer. In most cases, resolution occurs at the same site, regenerating the wild-type symbiotic and p42a plasmids (21).

Although our previous studies clearly demonstrated the role of IntA in catalyzing cointegration reactions in vivo, several important questions were left unanswered. Among these, the issue of directionality of IntA for the recombination activity (i.e., cointegration versus excision), as well as the determination of the minimal regions participating in IntA binding and recombination, were still unknown.

In this study, we demonstrate that (i) IntA is able to catalyze both cointegration and excision reactions in vivo, although the balance is strongly skewed toward cointegration; (ii) the minimal regions needed for specific binding and recombination activity of IntA on the attA and attD regions are comprised of 9- to 11-bp inverted regions plus a 5-bp central region; and (iii) integrity of the inverted repeats is absolutely required for IntA binding and recombination, while the central sector modulates the efficiency of these activities.

(This research was conducted by R. Hernández-Tamayo in partial fulfillment of the requirements for a Ph.D. in Ciencias Biomédicas from the Universidad Nacional Autónoma de México, Cuernavaca, México, 2013.)

MATERIALS AND METHODS

Bacterial strains, plasmids, and growth conditions.

The strains and plasmids used in this work are shown in Table 1. Rhizobium strains were grown at 30°C in PY (peptone-yeast extract) medium (23). E. coli strains were grown at 37°C in LB (lysogeny broth) medium (26). When needed, antibiotics were added at the following concentrations (in micrograms per milliliter): nalidixic acid, 20; spectinomycin, 100; kanamycin, 30; erythromycin, 100; carbenicillin, 100; and tetracycline, 10 for E. coli or 5 for Rhizobium. Media containing sucrose were prepared by adding appropriate volumes of a filter-sterilized, concentrated sucrose solution (50% [wt/vol]) to autoclaved PY medium.

Table 1.

Bacterial strains and plasmids

Strain or plasmid	Relevant features	Reference or source
R. etli strains
CFN42	Wild type	22
CE2	CFN42 Rifr derivative	23
CE3	CFN42 Strr derivative	23
CFN2001	CE2 derivative lacking p42a and p42d	21
CFNX107	CFN2001 recA::ΩSpr Smr	21
CFNX663	CE3 recA Rifr/p42d::Tn5mob;p42a::intA::ΩSpr Smr	21
CFNX750	CFNX107/pRD01	This study
CFNX751	CFN2001/pRD01	This study
CFNX752	CFNX750/pRD02	This study
CFNX753	CFNX663/pRD18	This study
A. tumefaciens strain
UIA143	C58 recA derivative lacking pTi	24
E. coli strains
S17.1	thi pro recA hsdR hsdM RP4-2-Tc::Mu-Km::Tn7	25
DH5α	λ− ϕ80dlacZΔM15 Δ(lacZYA-argF)U169 recA1 endA1 hsdR17(rK− mK−) supE44 thi-1 gyrA relA1	26
BL21(DE3)/pLysS	F− ompT hsdSB(rB− mB−) gal dcm (DE3) pLysS (Camr)	Novagen
Plasmids
pK18 mob sacB	Suicide vector; Kanr	27
pBBRMCS3	Broad-host-range cloning vector; Tcr	28
pMalc2X	Protein maltose fusion and expression vector	29
pRK404	Broad-host-range cloning vector; Tcr	30
pRD01	pBBRMCS3::intA-attA	This study
pRD02	pK18 mob sacB::attD (253 bp)	This study
pRD03	pK18 mob sacB::attD (198 bp)	This study
pRD04	pK18 mob sacB::attD (203 bp)	This study
pRD05	pK18 mob sacB::attD (150 bp)	This study
pRD06	pK18 mob sacB::attD (85 bp)	This study
pRD07	pK18 mob sacB::attD (65 bp)	This study
pRD08	pK18 mob sacB::attD (53 bp)	This study
pRD09	pK18 mob sacB::attD (23 bp)	This study
pRD10	pK18 mob sacB::attA (467 bp)	This study
pRD11	pK18 mob sacB::attA (287 bp)	This study
pRD12	pK18 mob sacB::attA (237 bp)	This study
pRD13	pK18 mob sacB::attA (187 bp)	This study
pRD14	pK18 mob sacB::attA (98 bp)	This study
pRD15	pK18 mob sacB::attA (89 bp)	This study
pRD16	pK18 mob sacB::attA (53 bp)	This study
pRD17	pMalc2x::intA gene	This study
pRD18	pRD17 fused with pRK404	This study
pRD19	pK18 mob sacB::attD (MutPalinA)	This study
pRD20	pK18 mob sacB::attD (MutPalinB)	This study
pRD21	pK18 mob sacB::attD (MutPalinC)	This study
pRD22	pK18 mob sacB::attF (CFN42)	This study

Molecular and microbiological procedures.

DNA manipulations and molecular techniques were done using established procedures (31). Extraction of DNA from agarose gels was done with a GeneJet extraction kit (Fermentas), and plasmids were isolated with a High Pure Plasmid Isolation kit (Roche). All oligonucleotides used (Table 2) were synthesized at the Unidad de Síntesis de Oligonucleótidos of the Instituto de Biotecnología, Universidad Nacional Autónoma de México. DNA fragments suitable for cloning were produced by PCR amplifications, using Taq polymerase High Fidelity (Invitrogen). Amplification consisted of 30 cycles of 1 min at 94°C, 1 min at variable temperature (depending on the primer combination), and 1 to 3 min at 68°C. PCR products were extracted with phenol, precipitated in ethanol, resuspended in Tris-EDTA buffer, and digested with the appropriate restriction enzyme(s) to generate the required ends of the fragments. PCR products were purified before cloning by band slicing. Primer combinations employed to generate attA or attD fragments for cloning and electrophoretic mobility shift analyses are shown in Table S1 in the supplemental material. Fragments of 23 bp, containing either wild-type or mutant versions of att sequences, were generated by annealing custom-made oligonucleotides (see Table S1) and purified using a GeneJet extraction kit (Fermentas); double-stranded fragments were digested with the required enzymes and ligated into suitably digested pRD02. For ligations, T4 polynucleotide ligase (Fermentas) was used. Plasmid transformation of E. coli was done using CaCl₂-competent cells. All plasmid constructions were verified by restriction analysis and PCR and, in most cases, by DNA sequencing.

Table 2.

Oligonucleotides used in this work

Name	Sequencea	Locationb
UpXbaintA	GCTCTAGAATGATAAAAGCCCGCAAAT	1339 p42a
LwPstintA	TGCACTGCAGTTAAATAAGTCGTGCGGC	2589 p42a
pcaDup	CTTGGATCCGCCGGCCTGCTGCTCGATCTGCTCT	72348 p42e
pcaDlw	GATGGATCCTCGACTTGGCCGCTTGGGTGAGA	72731 p42e
UpattA1	CTGGATCCCAAAGGTCGCTCCTGAAT	2252 p42a
UpattA2	CTGGATCCTCTGTCGGCTCATTTCGC	2432 p42a
UpattA3	CTGGATCCGAGATTTTTATACGTCATTAAC	2492 p42a
UpattA4	CTGGATCCAACGGAAGCGACGTCAG	2532 p42a
UpattA5	CTGGATCCGATTGATAAAAGCCCGCAAAT	2595 p42a
LwattA1	CGAATTCATTTTCCCGAGAATAACATG	2687 p42a
LwattA2	CGAATTCACTTCCGATAAGCAGTACTTA	2617 p42a
53bpUpA	CTGGATCCTCCGATAAGTAC	2595 p42a
53bpLwA	CGAATTCTGATTTGCGGGCT	2647 p42a
UpattD1	CTGGATCCTTGCGATTGAGAGTCCGGTCA	5655 p42d
UpattD2	CTGGATCCATTTCCGGGGCGAATCCGCC	5705 p42d
UpattD3	CTGGATCCTCGCTTCCGATAAGCATTACT	5765 p42d
LwattD1	CGAATTCCGGCGTTTGCATCTCGTTA	5875 p42d
LwattD2	CGAATTCGAAGCCAATATGGCAGTACATTC	5823 p42d
LwattD3	CGAATTCCGATTGCTATAACGACGAAAA	5765 p42d
53bpUpD	CTGGATCCTCCGATAAGTAA	5743 p42d
53bpLwD	CGAATTCTGATTTGCGGGCG	5796 p42d
Up23bp	CGGATCCTCCGATAAGCATTACTTATCGGACTTAAGT	5743 p42d
Lw23bp	GCCTAGGAGGCTATTCGTAATGAATAGCCTGAATTCA	5765 p42d
UpMutA	CGAATTCGATAGTAAGCATTACTTATCGGAGGATCCG	5743 p42d
LwMutA	GCTTAAGCTATCATTCGTAATGAATAGCCTCCTAGGC	5765 p42d
UpMutB	CGAATTCTCCGATAAGTGACCCTTATCGGAGGATCCG	5743 p42d
LwMutB	GCTTAAGATTCTATTCACTGGGAATAGCCTCCTAGGC	5765 p42d
UpMutC	CGAATTCTCCGATAAGCATTACTTAGAATGGGATCCG	5743 p42d
LwMutC	GCTTAAGAGGCTATTCGTAATGAATCTTACCCTAGGC	5765 p42d
UpattF	CGAATTCTCCGATAAGCAGCCCTTATCGGAGGATCCG	101656 p42f
LwattF	GCTTAAGAGGCTATTCGTCGGGAATAGCCTCCTAGGC	101678 p42f

^aAll oligonucleotides are shown in the 5′-to-3′ direction. Nonencoded bases introduced as clamps are shown in italics. Restriction sites are underlined; the first two oligonucleotides carry XbaI and PstI, respectively, while the rest carry either EcoRI (GAATTC) or BamHI (GGATCC) sites. Novel bases changed by mutation are in boldface italics.

^bThe location is indicated by the first 5′ nucleotide and the replicon where the sequence is located. Accession numbers are p42a (NC_007762), p42d (NC_004041), p42e (NC_007765), and p42f (NC_007766) of R. etli.

Plasmid transfer from E. coli to Rhizobium was done by biparental mating, using E. coli S17.1 harboring the appropriate plasmid as a donor and specific R. etli strains as recipients. The strains were grown in liquid medium to stationary phase, washed twice with PY medium, mixed in a donor/recipient ratio of 1:2 on PY plates, and incubated at 30°C overnight. After incubation, the cells were resuspended in 10 mM MgSO₄-0.01% Tween 40, and dilutions were plated on PY medium containing nalidixic acid, kanamycin, and tetracycline. Cointegration frequencies were evaluated independently at least three times and are expressed as the number of transconjugants per recipient cell ± standard deviation. To determine cointegrate excision frequencies, we used the sacB gene (27) in one of the plasmids employed to generate the cointegrate, thus allowing positive selection for cointegrate excision. To that end, appropriate R. etli strains harboring the desired cointegrate were grown overnight in PY medium, and suitable dilutions of each culture were plated onto PY medium containing sucrose at 12.5% (wt/vol). Excision frequencies were evaluated at least three independent times and are expressed as the number of sucrose-resistant colonies per total number of cells ± the standard deviation.

Plasmid profiles and hybridization.

Rhizobium plasmids were visualized by the Eckhardt procedure (32). Gels were transferred onto Hybond N+ membranes (Amersham) using the manufacturer's protocol and cross-linked using a UV cross-linker unit (Stratagene). Hybridizations were performed overnight using [α-³²P]dCTP-labeled probes (Megaprime kit; Amersham) under high-stringency conditions (65°C in rapid-Hyb buffer; Amersham). Hybridization signals were detected with a PhosphorImager (Molecular Dynamics).

Expression and purification of IntA.

The short form of IntA (see Results) was overexpressed in E. coli strain BL21(DE3)(pLysS) (Novagen) as an N-terminal maltose-binding protein (MBP)-tagged protein. For this purpose, intA was amplified by PCR from R. etli CFN42 DNA using the UpXbaintA (built-in XbaI site) and LwPstintA (built-in PstI site) primers (Table 2), digested with the indicated enzymes, and ligated into suitably digested pMALc2x (29). The resulting construct, pMAL-c2x::intA, was verified by sequencing. The E. coli BL21(DE3) strain was transformed with the pMAL-c2x::intA recombinant plasmid (pRD17) and grown overnight in LB medium plus carbenicillin at 30°C. The overnight culture was diluted 1:100 in fresh LB medium carrying carbenicillin and grown at 30°C to an A₆₀₀ of 0.6 to 0.7; expression was induced by the addition of 0.1 mM IPTG (isopropyl β-d-thiogalactoside) for 2 h at 30°C. Bacteria were harvested by centrifugation at 4,100 × g for 10 min at 4°C, and the pellets were suspended in column buffer (20 ml of 20 mM Tris-HCl, pH 7.4, 200 mM NaCl, 1 mM EDTA). Sonication was done using an Ultra Cell (Sanyo) at an output control of 25 W by continuous pulses (interstimulus interval, 10 s) interrupted by 5-s breaks on ice. Soluble and insoluble fractions of the E. coli lysate were separated by centrifugation at 8,200 × g for 15 min at 4°C. The MBP-IntA fusion protein was purified with an amylose affinity column (New England BioLabs). After loading and washing, MBP-IntA was eluted with a buffer containing 10 mM maltose. Proteins were detected by SDS-PAGE after staining with Coomassie blue. Fractions containing MBP-IntA were concentrated by Amicon (Millipore) filtration, pooled, and kept in 30% glycerol at −20°C.

Electrophoretic mobility shift assays (EMSAs).

Regions corresponding to attA and attD prepared by PCR (see “Molecular and microbiological procedures” above) were separated on 1.5% agarose gels and purified by band slicing. Reaction mixtures for DNA mobility shift assays contained 0.10 pmol of DNA and variable amounts of the MBP-IntA protein in a final volume of 10 μl. MBP-IntA was incubated with the desired fragments for 30 min at room temperature in binding buffer (20 mM Tris-HCl [pH 8.5], 10% glycerol, 50 mM KCl, 3 mM MgCl₂, 0.5 mg of bovine serum albumin/ml). To evaluate binding specificity, a 383-bp fragment from the R. etli pcaD gene was added to the binding reaction mixture. The mixtures were loaded on native 8% polyacrylamide gels prepared with TB-EDTA (40 mM Tris base, 40 mM boric acid, 1 mM EDTA) and subjected to electrophoresis for 1.5 h at 60 V. DNA was visualized with ethidium bromide.

DNA footprinting.

To obtain labeled attA and attD fragments for DNase I footprinting, the primers of interest were 5′ end labeled with T4 polynucleotide kinase in the presence of [γ-³²P]ATP (3,000 Ci/mmol) and used to amplify attA and attD. An amount of probe equivalent to about 100,000 cpm was preincubated at room temperature with increasing concentrations of MBP-IntA in binding buffer (20 mM Tris-HCl [pH 8.5], 10% glycerol, 50 mM KCl, 3 mM MgCl₂, 0.5 mg of bovine serum albumin/ml). After 20 min, 0.003 U of DNase I (Roche, Nutley, NJ) in dilution buffer (8 mM Tris-HCl [pH 7.9], 40 mM MgSO₄, 4 mM CaCl₂, 40 mM KCl, 2 mM EDTA [pH 8.0], 24% glycerol) was added to the mixture and incubated at room temperature for 2 min. Reactions were stopped by addition of 500 μl DNase I stop solution (570 mM ammonium acetate, 80% ethanol, 0.5 mg/ml tRNA). Digested DNA samples were isolated by phenol-chloroform extraction and ethanol precipitation. Pellets containing DNA were air dried and then resuspended in formamide-containing loading dye. The samples were heated at 95°C for 3 min and loaded into an 8% polyacrylamide gel containing 7 M urea. The gels were vacuum dried and visualized with a PhosphorImager (Molecular Dynamics). Sequencing reactions were included as size markers. Nucleotide sequence determination was performed by the dideoxy chain termination method (33) with a TaqTrack sequencing kit (Promega).

RESULTS

In vivo IntA-dependent cointegration and cointegrate resolution.

To evaluate if IntA is able to catalyze both cointegration and resolution reactions in vivo, a simplified two-plasmid system was designed. The system employed as hosts R. etli strains lacking both p42a and p42d and, when required, without an active recA gene, as well. One of the plasmids used (pRD01) contains a broad-host-range origin of replication and an RP4-oriT sequence and harbors a 1.32-kb fragment possessing both the intA gene and the attA sequence. Upon introduction by conjugation of pRD01 into the desired hosts, the plasmid becomes the only source of both IntA and attA. The second plasmid used (pRD02) comprised a nonhomologous plasmid carrying an origin of replication that is active in E. coli but inactive in R. etli (i.e., a suicide plasmid in R. etli), an RP4-oriT sequence, and a 0.26-kb fragment harboring attD. Note that, besides the RP4-oriT sequence, these two plasmids share only att regions. In this system, stable transconjugants containing pRD02 should only be obtained by integration catalyzed by IntA.

Stable transconjugants containing pRD02 were obtained only in the presence of pRD01 (Table 3, top two rows); in contrast, no transconjugants were obtained whenever the mobilized plasmid lacked attD (Table 3, bottom two rows). Note that similar frequencies were obtained irrespective of the recA status of the host strain, indicating that most of the events leading to stabilization of pRD02 were recA independent. To ascertain if stabilization of pRD02 requires integration, plasmid profiles of selected transconjugant strains were analyzed by the in-gel lysis technique of Eckhardt (32). Figure 1A shows that all transconjugants bearing pRD02 present a novel plasmid band, whose size is compatible with integration of pRD02 into pRD01. To verify that this novel plasmid band is due to cointegration, plasmid profiles were subjected to Southern blot hybridization, using as a probe either pK18mob (Fig. 1B) or the intA gene (Fig. 1C). As expected, the novel plasmid band harbors both sequences, confirming cointegration between pRD01 and pRD02. Further proof of cointegration in these strains was obtained through PCR. For every transconjugant strain analyzed, no amplified products were observed when PCRs were set up with primers flanking the att regions in pRD01 or pRD02; in contrast, amplified products were obtained upon combination of the forward primer flanking one att insert with the reverse primer of the other att insert (data no shown). These results indicate that cointegrates were formed by IntA-mediated recombination between attD and attA.

Table 3.

Frequencies of in vivo IntA-dependent cointegration and excision in R. etli

Strain	Relevant genotype	Frequency (10−5) of:
		Cointegrationa	Excisionb
CFNX750::pRD02	recA mutant intA+ attA+ attD+	2.34 ± 1.43	0.28 ± 0.14
CFNX751::pRD02	recA+ intA+ attA+ attD+	7.6 ± 1.52	ND
CFNX750::pK18 mob sacB	recA mutant intA+ attA+	≤0.002	NA
CFNX751::pK18 mob sacB	recA+ intA+ attA+	≤0.002	NA

^aCointegration frequencies are expressed as the number of transconjugants per recipient cell ± standard deviation for at least three independent determinations.

^bExcision frequencies were evaluated at least three independent times and are expressed as the number of sucrose-resistant colonies per total cells ± standard deviation. ND, not done; NA, not applicable.

Fig 1.

Analyses of in vivo cointegration and excision events. (A and D) Plasmid profiles of selected strains stained with ethidium bromide. (B, C, E, and F) Southern blots of the corresponding plasmid profiles revealed by autoradiography using ³²P-labeled intA (B and E) or pK18mob sacB (C and F) as a probe. (A to C) Cointegration events. Lane 1, CFN42; lane 2, CFNX107; lane 3, CFNX750; lanes 4 to 8, putative cointegrates pRD01::pRD02. (D to F) Excision events. Lane 1, CFN42; lane 2, CFNX107; lane 3, CFNX750; lane 4, CFNX752; lanes 5 to 9, putative excisants.

Cointegrates obtained in this section also offer a convenient way to evaluate if IntA can catalyze excision. Since pRD02 also contains a Bacillus subtilis sacB gene (a sucrose-dependent, conditionally lethal marker in Gram-negative cells), excision of pRD02 from these cointegrates can be evaluated by sucrose selection. Putative excisants were obtained at a 10-fold-lower frequency than transconjugants (Table 3). Analysis of the plasmid profile in these excisants revealed the loss of the cointegrate plasmid band, with a plasmid band corresponding in size to pRD01 (Fig. 1D), which also hybridizes with an intA probe (Fig. 1E), appearing instead. Complete loss of pRD02 from the excisants was corroborated by the absence of hybridization with the pRD02 plasmid backbone (Fig. 1F). These results indicate that R. etli IntA is a “bidirectional” recombinase, able to promote not only integration, but also excision events, albeit at a lower frequency.

IntA binds specifically to the attA and attD sites.

The R. etli IntA open reading frame (ORF) (accession no. AF538364) has two potential UUG start codons, localized 11 codons apart. Since it is unknown which of the two start codons is employed in vivo, appropriate oligonucleotides were used to amplify both the long and the short forms of IntA. Given the insolubility of the native protein, both forms were fused to MBP and purified as described in Materials and Methods. No difference was found between the two forms for in vivo recombination activity and DNA binding (data not shown); therefore, we report here only the data obtained with the short form of IntA.

The functionality of the MBP-IntA fusion protein was ascertained by in vivo assays. To this end, we employed R. etli CFNX663, which contains the symbiotic plasmid (p42d) marked with Tn5 and also carries inactivating mutations in both recA and intA. Inactivation of these two functions renders a strain unable to donate its pSym by conjugation due to the inability to form the p42a-pSym cointegrates necessary for transfer. This strain should recover conjugative ability upon complementation with an active intA.

To evaluate if the MBP-IntA hybrid can complement conjugative capacity, a plasmid harboring the hybrid (pRD17) was fused with pRK404, giving rise to pRD18, which is able to replicate in R. etli (see Materials and Methods), and introduced into strain CFNX663. As expected, strain CFNX663 was unable to transfer the pSym to Agrobacterium tumefaciens strain UIA143, but all the CFNX633 derivatives complemented with the MBP-intA hybrid regained conjugative pSym transfer at a frequency indistinguishable from the one afforded by the wild-type intA (see Fig. S1 in the supplemental material). These results indicate that fusion of IntA to MBP does not affect its in vivo recombination ability.

The MBP-tagged IntA derivatives were purified to homogeneity by amylose affinity chromatography (see Materials and Methods) (see Fig. S1B in the supplemental material) and used to set up EMSAs. Linear DNA fragments carrying the regions attA (467 bp) and attD (253 bp) were mixed with a nonspecific DNA fragment of R. etli (pcaD; 383 bp). Increased amounts of MBP-IntA, up to a 7-fold molar excess with respect to DNA, were added to these mixtures. As shown in Fig. 2, well-defined complexes were detected even at the smallest amount of IntA used. As expected for specific binding of IntA to att-containing fragments, increased amounts of IntA led to a gradual recruitment of attA and attD fragments into retarded complexes. No retardation of the pcaD fragment was observed, even at the largest amount of IntA employed. These results clearly indicate specific binding of MBP-IntA to attA and attD.

Fig 2.

IntA interacts specifically with attA and attD. EMSAs were set up with equal molar proportions of specific (attA and attD) and nonspecific (pcaD) DNA fragments, mixed in the absence (lane 1) or presence (lanes 2 to 8) of MBP-tagged IntA, analyzed on an 8% polyacrylamide gel, and stained with ethidium bromide. Lane 1, no IntA; lanes 2 to 8, increasing amounts of IntA (molar ratios [DNA/protein] were 1:0.5, 1:0.75, 1:1, 1:1.25, 1:1.5, 1:2, and 1:3, for lanes 2 to 8, respectively).

The palindromic regions in attA and attD are required for specific binding of IntA.

To determine the minimal regions needed for specific binding of IntA, shortened derivatives of the attA and attD regions were produced by PCR (Fig. 3A). All shortened fragments of the attA and attD regions were subjected to IntA EMSA analyses.

Fig 3.

The palindromic sequences in attA and attD are required for specific binding of IntA. (A) Shortened derivatives of attA and attD regions, which were analyzed by EMSA; derivatives marked with a plus showed IntA-dependent retarded complexes, while those labeled with a minus lacked any detectable complex. The boxed regions in the topmost diagrams correspond to the 53-bp homology between attA and attD; the arrows indicate the palindromic sector present in each of these regions. The vertical dashed lines show the retention of specific sequence characteristics in each derivative. (B and C) EMSA analyses of shortened derivatives of the attA and attD regions, respectively. Lanes marked with a minus lack MBP-tagged IntA, while lanes with a plus contain increasing amounts of MBP-tagged IntA (DNA/IntA molar ratios ranged from 1:1 to 1:6). (D) EMSA analysis of a synthetic 23-bp palindromic sequence with increasing amounts of MBP-tagged IntA.

Clear IntA-DNA complexes were detected for all the shortened derivatives harboring the corresponding palindromic region, for both attA (Fig. 3B) and attD (Fig. 3C). The IntA-protected complexes for the palindrome-bearing, shortened derivatives were produced with the same stoichiometry as the full-length fragments. Interestingly, fragments lacking the corresponding palindrome (attA, 98 bp, and attD, 65 bp) (Fig. 3B and C, respectively) failed to produce IntA-protected complexes. These results indicate that the palindromic sequence in each att region is indispensable for IntA binding.

To verify if the palindromic regions are sufficient for binding, a synthetic 23-bp palindrome was produced by annealing of custom-made oligonucleotides (see Materials and Methods). As shown in Fig. 3D, IntA-dependent complexes were readily detected with this synthetic palindrome, even at the smallest IntA amount employed, indicating that the palindromic region is sufficient to achieve specific binding of IntA.

IntA binds mainly to the palindromic sector in att regions.

To ascertain more precisely the sequence bound by IntA, DNA-footprinting assays with IntA in the presence of DNase I (Fig. 4) were carried out for both att regions. Consistent with the results presented thus far, a 34-bp sequence motif in attD was protected from DNase attack in the presence of IntA, even at the lowest protein concentration tested (Fig. 4, left). This 34-bp sector fully encompasses the palindromic sequence. In a similar way, a 33-bp sequence in attA, spanning the corresponding palindrome, was also protected by IntA (Fig. 4, right). Protection of both palindromic sequences by IntA is fully consistent with the results described above.

Fig 4.

DNase I protection of the attA and attD regions by IntA. Increasing amounts of MBP-tagged IntA were mixed with ³²P-end-labeled DNA fragments corresponding to attD (253-bp) (left) or attA (287 bp) (right) and treated with DNase I. DNA/IntA molar ratios ranged from 1:1 to 1:6; controls containing the specified fragment plus IntA alone or the fragment plus DNase I alone are also shown. Samples were subjected to electrophoresis on an 8% polyacrylamide sequencing gel and detected on a PhosphorImager. The brackets encompass the protected region for each fragment; nucleotides depicted in boldface correspond to the palindromic sequence. The lanes labeled G, A, T, and C correspond to sequencing reactions with the same fragments run in parallel.

In vivo recombination abilities of shortened derivatives of attA and attD.

To evaluate if shortening of both att regions affects in vivo recombination abilities, every reduced fragment (Fig. 3A) was cloned separately into a mobilizable suicide plasmid, thus generating families of plasmids (pRD03 to pRD08, attD; pRD11 to pRD16, attA). Introduction of these plasmids into an R. etli strain harboring pRD01 (intA and attA) allows easy evaluation of in vivo cointegration ability.

As shown in Table 4, no significant reduction in cointegration ability was found for shortened derivatives of attD, with the exception of pRD07, which was unable to generate cointegrates; this plasmid contains a 65-bp fragment that is the only one lacking the palindromic sequence and hence is unable to bind IntA in vitro. Similar results were found for variants of the attA region (Table 4). Every variant of attA that contains the palindromic sequence cointegrates readily; the variant that lacks the palindromic sector was unable to cointegrate. As expected from these data, a plasmid that carries a synthetic 23-bp palindromic sector (pRD09) promotes intA-dependent cointegration at frequencies comparable to those found with larger fragments.

Table 4.

Frequencies in R. etli of in vivo cointegration and excision with variant att regions

Plasmid with strain CFNX750	Relevant genotype (size [bp])	Frequency (10−5) of:
		Cointegrationa	Excisionb
pRD03	intA+ attA+ attD (198)	3.7 ± 0.89	ND
pRD04	intA+ attA+ attD (203)	1.9 ± 0.87	ND
pRD05	intA+ attA+ attD (150)	7.7 ± 0.72	0.49 ± 0.32
pRD06	intA+ attA+ attD (85)	6.1 ± 0.75	0.47 ± 0.21
pRD07	intA+ attA+ attD (65; no palindrome)	≤0.002	NA
pRD08	intA+ attA+ attD (53)	2.2 ± 0.45	0.56 ± 0.19
pRD11	intA+ attA+ attA (287)	4.0 ± 0.45	0.34 ± 0.12
pRD12	intA+ attA+ attA (237)	4.9 ± 0.75	ND
pRD13	intA+ attA+ attA (187)	2.2 ± 0.26	0.22 ± 0.026
pRD14	intA+ attA+ attA (98; no palindrome)	≤0.002	NA
pRD15	intA+ attA+ attA (89)	5.6 ± 0.47	ND
pRD16	intA+ attA+ attA (53)	2.5 ± 0.72	0.53 ± 0.28
pRD09	intA+ attA+ palindromic region (23)	7.4 ± 1.7	0.43 ± 0.08
pRD19	intA+ attA+ MutPalinA	≤0.002	NA
pRD20	intA+ attA+ MutPalinB	0.071 ± 0.0063	0.43 ± 0.046
pRD21	intA+ attA+ MutPalinC	≤0.002	NA
pRD22	intA+ attA+ attF+	0.077 ± 0.0072	0.54 ± 0.35

^aCointegration frequencies are expressed as the number of transconjugants per recipient cell ± standard deviation for at least three independent determinations.

^bExcision frequencies were evaluated at least three independent times and are expressed as the number of sucrose-resistant colonies per total number of cells ± standard deviation. ND, not done; NA, not applicable.

Cointegrates generated with these plasmids also allowed us to explore the excision ability of IntA, using the methodology described above. As shown in Table 4, fragments that retain in vivo cointegration ability were also as efficient as the original fragments in promoting in vivo excision. The positive correlation between in vitro IntA binding and in vivo recombination ability, as well as the finding that the palindromic sequence suffices for binding of IntA and recombination ability, support the proposal that the palindromic sequence is the minimal region needed for the action of IntA.

Mutations in the palindromic sequence abolish specific binding and recombination ability.

The conserved 23-bp palindromic sequence is comprised of two 9-bp inverted repeats separated by a 5-bp sequence. To evaluate the roles of the inverted repeats and central region in the binding of IntA and recombination ability, we constructed mutant versions of the palindromic sequence in which five contiguous changes were made in the left (MutPalinA) or right (MutPalinC) inverted repeat, thus disrupting their complementarity; a 5-bp change was also constructed in the central region (MutPalinB).

Analyses of IntA binding to these mutant fragments by EMSA revealed that mutations that disrupt the integrity of the 9-bp inverted repeats abolish both specific binding (Fig. 5) and recombination ability (Table 4, plasmids pRD19 and pRD21). In contrast, a mutation that disrupts the 5-bp central region reduces both binding (Fig. 5) and recombination (as much as 100-fold; compare plasmids pRD09 and pRD20 in Table 4). Sequence analysis of the R. etli CFN42 genome revealed the presence of a palindromic sector on plasmid p42f highly similar to attA and attD. The inverted repeats in this motif are identical to those in both attA and attD, but the central region differs from the one in attA in only 2 out of 5 bases (Fig. 5). To evaluate if the palindromic sequence on p42f is active in IntA-mediated recombination, this sequence was subjected to IntA binding experiments (Fig. 5) and in vivo recombination (Table 4, pRD22). IntA recognizes the palindromic sequence in p42f, albeit with reduced affinity compared to the palindromic sequence in attA and attD. In vivo recombination mediated by IntA also occurred with the attF palindrome, but at a 100-fold-lower efficiency than for the attA and attD motifs. These results demonstrate that integrity of the inverted repeats is absolutely required for IntA binding and recombination, while the central sector modulates the efficiency of these activities.

Fig 5.

Mutations in the palindromic sequence abolish IntA binding ability. (A) EMSA analyses with increasing amounts of MBP-tagged IntA (DNA/IntA molar ratios ranged from 1:1 to 1:6) of a wild-type palindrome (palindromic sequence, 23 bp) and mutants in the left (MutPalinA) or right (MutPalinC) arm of the palindromic sequence or in the central region (MutPalinB and attF p42f). (B) Alignment of different palindromic sequence variants; nucleotides that differ from attA are shown in boldface. (C) Quantification of retarded complex formation for the wild-type palindrome (■) and mutants in the central region (MutPalinB, ▲, and attF p42f, ●). Quantifications are based on three independent determinations ± standard deviations.

DISCUSSION

In this work, we characterize the function of the R. etli IntA site-specific recombinase through a combination of in vivo and in vitro assays. The results described here clearly show that IntA is able to catalyze both cointegration and excision events in vivo in a manner dependent on the presence of a characteristic 23-bp sequence (the att sector) harboring the two arms of a palindrome plus a divergent central region. Thus, R. etli IntA appears to be a bidirectional recombinase. Evaluations presented here regarding the frequency of integration (2 × 10⁻⁵) and excision (2 × 10⁻⁶) events differ only 10-fold. However, it should be kept in mind that the evaluation of integration frequency is in fact the sum of two independent processes: conjugative transfer and integration of the test plasmid. Frequency of conjugative transfer was determined separately at 1 × 10⁻³; thus, frequencies of integration shown in this paper should be on the order of 1 × 10⁻². Evaluations of the excision frequency are also a composite of excision itself and loss of the excised segment; however, the frequency of loss of the excised segment should be negligible, due to the lack of a replication origin. Thus, although it is true that IntA can catalyze both integration and excision, the balance between these activities is skewed toward integration by a factor of 4 orders of magnitude.

It should be stressed that experiments such as the one shown in Fig. 2 were designed to evaluate not only IntA binding, but also IntA-mediated recombination in vitro. IntA binding was readily demonstrated, but in vitro IntA recombination was not observed, despite designing segments containing the corresponding att in an asymmetric position. This inability persisted despite variations in the protein concentration used, the reaction pH, the reaction temperature, and the presence of different ions (Na⁺ and Mg²⁺) at variant concentrations (data not shown). Although it is formally possible that the suitable combination of parameters needed for in vitro recombination has not yet been found, it is also possible that IntA requires for this activity another protein that interacts with it. Experiments are under way to evaluate this possibility, looking for possible interacting proteins in pulldown experiments.

The conclusion that binding and recombination catalyzed by IntA requires only a 23-bp sequence motif (the att site) is based on in vivo and in vitro experiments that show that binding and recombination occur only with segments containing the att region and not in their absence. The fact that an artificial segment containing only the att site acts as efficiently as a larger segment further supports this conclusion. Moreover, this observation reveals that R. etli IntA, unlike other tyrosine recombinases (34), does not require other neighboring sequences acting as enhancers for its function. Consistent with this, DNase I-footprinting assays with IntA revealed protection of the att sites (Fig. 4); however, IntA also protects an additional 9-bp (attD) or 8-bp (attA) sequence. The fact that the sequences in the two additional pieces are dissimilar militates against a role of these extended sectors as determinants for specific binding. These extended protection sectors could be due to the well-known inability of DNase I to cut in close proximity to a bound protein, due to steric hindrance.

Besides its bidirectional action in vivo, another interesting characteristic of the R. etli IntA recombinase is that strict identity in the att sectors is not needed for recombination. Although the inverted repeats are identical between attA and attD, the central region differs in 1 nucleotide. Despite this difference, recombination frequencies in vivo were the same for attA-attD and attA-attA combinations (Table 4). Lack of strict identity in the central region is a characteristic shared with a recently described group of tyrosine recombinases (5).

Mutagenesis of the att region allowed us to explore the role of defined sequences in IntA binding and recombination. Mutations that disrupt the integrity of the 9-bp inverted repeats abolish both specific binding and recombination abilities, while mutations that change the 5-bp central region severely reduce both binding and recombination. Interestingly, reduced binding and recombination were also found with a natural variant of att (attF) differing from the sequence of attA only in the last 2 nucleotides in the central region (CC instead of TA). Systematic mutagenesis of the residues of the central region, either singly or in combination, is needed before assigning a role to residues in the central sector in resolution of the Holliday junction, as described for other tyrosine recombinases (4).

The characterization of the IntA recombinase presented here confirms our proposal that IntA participates in the high-frequency formation of cointegrates p42a-p42d (21). The fact that IntA presents weak excision activity in vivo would have as a consequence, over time, the accumulation of p42a-p42d cointegrates; this tendency, however, is opposed by resolution of these cointegrates through homologous recombination between repeated sequences shared between the plasmids (21). Our data also suggest the possibility of recombination between p42a and p42f; cointegrates between the two plasmids were recently detected in vivo (S. Brom, unpublished results). Besides its importance for understanding events of genome evolution in Rhizobium, the IntA system should be a novel tool to introduce new segments into the genome in defined locations, thus advancing genome engineering in this group of organisms. Experiments are under way to achieve this objective.