In order to identify critical residues involved in the anti-RR and/or the excisive recombination activity of the TorI RDF, we designed a tester strain that can report both activities. Strain LCB995 contains a torA'-'lacZ fusion as well as a cat cassette inserted into a non-coding region of the KplE1 prophage [20,27]. When this reporter strain was plated onto MacConkey lactose plates in the presence of 10 mM TMAO to induce the tor operon promoter, the colonies formed turned red, as the β‑galactosidase was produced [27]. In contrast, when this strain was transformed with a multicopy plasmid encoding a wild-type version of the torI gene (pJFi plasmid), which allows inducible TorI production in the presence of 1 mM IPTG, the resulting colonies remained white on the same medium due to the anti-RR activity of TorI. On the other hand, the RDF activity of TorI was monitored with the same tester strain containing pJFi plated onto chloramphenicol (Cm) containing plates in the presence of 1 mM IPTG. As described before [20], expression of the torI gene is sufficient to promote KplE1 excision, and thus renders the cells chloramphenicol sensitive. Since colonies on agar plate arose from single cells, colonies are not forming from cells that have excised the KplE1 prophage. We thus have set up an experimental procedure to screen a mutant library of TorI that may be affected in two different activities (Figure 1).

The mutant library of torI alleles was generated by extensive random PCR-mutagenesis of the torI gene (201 base pairs) carried by plasmid pJFi (see the Experimental Section). About 3,000 colonies were screened on reporter (MacConkey plates containing lactose and TMAO) or selective media (LB plates containing chloramphenicol), and colonies that presented a red color on MacConkey lactose plates containing TMAO or that could grow in the presence of chloramphenicol, were selected. A final set of 42 clones reproducibly altered in one or both TorI activities was selected for further investigation (Figure 1). To make sure that the observed phenotypes were not due to chromosomal mutations, all plasmids were extracted and used to transform the tester strain. Direct Sanger sequencing of the plasmids carrying the mutated alleles allowed the identification of the mutations. In total we isolated 18 distinct mutations, of which 16 were due to a point mutation leading to a single amino-acid substitution (Table 1). Overall, the mutagenesis showed no obvious mutational bias with mutations dispersed all along the primary sequence and a wide diversity in the mutations that we obtained with 15 transitions, four transversions, one deletion and two extensions due to a frameshift on the stop codon. No redundant mutations were observed, in total 20 different positions out of 66 were affected, most of them were changes in residues located on the surface of the protein (with the exception of Ile₁₆, Trp₄₇ and Phe₅₆), and only one substitution has impacted the overall structure of the protein (L5P, Table 1 and text below). Thereafter, pJFi* designates a plasmid carrying a mutated allele, and TorI* indicates a mutant protein.

A preliminary in vivo quantification of TorI activities was performed (Table 1). The anti-RR activity of the TorI mutants was estimated by monitoring β-galactosidase activities of strain LCB995 containing pJFi*. This activity reports the expression of the torA'-'lacZ fusion in the presence of TMAO. On the other hand, the RDF activity of TorI mutants was measured in the same strain by counting colonies able to grow on chloramphenicol plates relative to colonies counted on ampicillin plates. Results are expressed as the ratio of ampicillin-resistant/chloramphenicol-resistant colonies, and reflect the ability of the TorI mutants to promotes KplE1 excision. We identified two classes of mutants: one class contained mutants affected in the anti-RR activity only, whereas the second class was composed of mutants affected in both activities. As a result, we did not isolate mutants affected in the RDF activity only. In recent studies [20,26], we also designed several mutants by site-directed mutagenesis which were included in Table 1 (^d mark) and were further analyzed together with the mutants generated by random mutagenesis.

To our surprise, all the mutations we obtained impaired the anti-RR activity of TorI, and these mutations map all over the surface of the protein (Figure 2A). In a previous work, we showed that TorI inhibited the transcriptional activator TorR through a direct protein-protein interaction without preventing TorR binding to its specific DNA targets, and we hypothesized that TorI was binding to a region of TorR that is important for RNA polymerase recruitment [24]. All the surface mutations that we obtained altered the anti-RR activity, suggesting that more than one face of the protein is involved in this function and that one face of the protein could bind to TorR and another to RNA polymerase. A vast majority of the mutants isolated affect surface residues, which are probably involved in protein‑protein interactions. One may have expected a more subtle effect on the anti-RR activity, however, our screen failed to select those.

In contrast to the anti-RR activity, only a subset of mutations altered the RDF activity of TorI and most of these mutations map on the surface of the protein although the mutations are scattered along the primary sequence (Figure 2B). By this approach, we identified two regions of the protein involved in the excisionase function: one centered on helix 1 and another one comprising the helix 2 and the wing motif. The wing-helix is an atypical helix-turn-helix motif involved in DNA-binding commonly found in transcription factors and particularly in the OmpR family of response regulators [28,29]. Based on the work done on the lambda Xis protein, two separate functions are [22] required for the excisionase activity: (i) Xis interacts with DNA to position itself on the X sites on the attR recombination region, (ii) Xis interacts with Int to direct a proper positioning of the integrase for the excisive reaction [15]. In a previous work, residues involved in DNA binding were identified by NMR titration with a short DNA sequence and mapped onto the wing-helix motif [20]. Therefore, two positions of the TorI wing-helix motif were mutated (Tyr₂₈ and Arg₄₅) and the resulting proteins proved to be impaired in KplE1 excision, although Tyr₂₈ mutants were more affected than Arg₄₅ ones, suggesting a role for this region in DNA binding activity.

Strains and plasmids used in this study are listed in Table 2. Strain LCB995 was constructed by transduction of the cat marker inserted in the KplE1 prophage between yfdO and yfdP (strain LCB970 [20]) into strain LCB620 carrying a torA'-'lacZ fusion [27]. Strains were grown in LB medium supplemented when necessary with ampicillin (50 µg.mL⁻¹), chloramphenicol (25 µg.mL⁻¹), TMAO (10% v/v) or IPTG (1 mM).

Error prone PCR was performed to generate mutated alleles of the torI gene. PCR was conducted with the primer pair torI_MunI (5'-TAC AAT TGC GGA GAT AGC ACT CAT GCA ACA C)/torI_BamHI (5'-TTG GAT CCT TAC CCA TTG GCG CGG CTT AAG AG), plasmid pJFi as a template and a classical Taq polymerase (GoTaq Promega, error rate 10⁻⁶). Three rounds of PCR were performed using as a template a 10⁶ dilution of the product generated by the previous round of PCR. After three rounds of PCR, products were purified, hydrolyzed with MunI and BamHI enzymes and ligated into the pJF119EH vector cut with EcoRI and BamHI. The plasmid library was then transformed into the tester strain LCB995.

IntS, TorI and IHF proteins were overproduced and purified to homogeneity as previously described [24,30]. TorI mutants were purified as the wild-type protein. All proteins were dialyzed in Tris-HCl buffer (40 mM, pH 7.6) containing 50 mM KCl and 10% glycerol. The protein concentrations were measured by densitometry with the wild-type TorI protein as a reference.

Strain LCB970 carrying torI encoding plasmids pJFi (7) was grown in LB medium until the OD₆₀₀ reached 0.5 units (0.5 × 10⁹ cells.mL⁻¹), and IPTG (1 mM) was added for 2 h at 37 °C under agitation. Culture dilutions were prepared and plated onto rich medium containing either 50 µg.mL⁻¹ ampicillin or 25 µg.mL⁻¹ chloramphenicol. Numeration of the colonies plated on both antibiotics was performed and the ratio of ampicillin-resistant/chloramphenicol-resistant colonies was calculated. Values represent the average of at least three independent determinations.

β-Galactosidase activities were measured on whole cells according to the method of Miller (1972); values represent the average of at least three determinations with a variation of no more than 10% from the mean.

Linear att sites were amplified by PCR with primer pairs attL-SpeI (5'-GAC TAG TTT CAA TCT GCT TAA CGG TGA GCA T)/attL-KpnI (5'-GGG GTA CCG CTA ATT GCA GGT TCG ATT CC) for attL (220 bp) and attR-XbaI (5'-GCT CTA GAG GTT TTA GGG ATA AAC ACA CAA GGA TG)/attR-IHF2 (5'-CTC TTA AGC CGC GCC AAT GG) for attR (135 bp), and then purified using Qiaquick PCR purification kit protocol (Qiagen). Reaction mixtures (25 µL) included linear att DNA sites (28 nM) in buffer containing 33 mM Tris-HCl pH 7.6, 33 mM KCl, 9 mM spermidine, 4 mM EDTA, 0.9 mg/mL⁻¹ acetylated BSA and 7 % glycerol. IHF (0.3 µM), IntS (0.6 µM), and TorI (2.1 µM) were added as indicated in the figures legends. The reactions were carried out in optimized conditions at 30 °C for 2 h at an IHF:IntS:TorI protein ratio of 1:3:7. Reaction products were purified (Qiaquick kit, Qiagen) and analyzed on a 2% agarose gel electrophoresis. The gel was then scanned and the data analyzed using AlphaView software (Protein simple).

EMSA were carried out using purified proteins and fluorescently Cy5-labelled attL DNA fragment that was amplified by PCR using MG1655 chromosomal DNA as a template with the primer pair attL‑pro (5'-AAT GGA TAT AAC GAG CCC CTC C)/attL-ter-Cy5 (5'-CAT CGA GAA GGC GGT ATG GTT TTT C). DNA and purified proteins were mixed together at different concentrations (as indicated in figure legends) in the presence of 4 mg.mL⁻¹ BSA and 0.5 mg.mL⁻¹ calf thymus DNA (CT-DNA) in binding buffer (40 mM Tris pH 7.6, 85 mM KCl, 19% glycerol). Reactions were incubated for 30 min at 30 °C. DNA-protein complexes were then separated using a 6% non‑denaturing polyacrylamide gel (37:1 acrylamide:bisacrylamide ratio). A pre-migration step (1 hour at 160 V) was carried out to reduce ionic charges which may have destabilized the DNA‑protein complex. Samples were then loaded and left to migrate at 80 V during 30 min and then at 160 V for 2 more hours in 0.5× TBE (50 mM Tris, 45 mM Boric acid, 0.5 mM EDTA) running buffer. The gel was scanned using a FLA5100 (Fuji) scanner, using excitation wavelength of 635 nm (800 V scanning intensity) and emission wavelength of 665 nm. Data was analyzed using Multi Gauge (Version 2.3) software [35].

TorI protein was pre-incubated 10 min at 25 °C in the presence or absence of IntS and/or TorR proteins. Then, homobifunctional sulhydryl reactive agent bis(maleimido)hexane (BMH, 1mM, Pierce) was added and the reaction continued for another 30 min. Samples were ran on a 7%–16% Tricine‑SDS PAGE, transferred onto a nitrocellulose membrane and revealed by immunodetection with TorI antiserum.