The plant-growth promoting rhizobacterium Azospirillum lipoferum strain 4B generates in vitro a stable phase variant designated 4VI at frequencies of 10−4 to 10−3 per cell per generation. Variant 4VI displays pleitropic modifications, such as the loss of swimming motility and the inability to assimilate certain sugars compared to the wild type. The mechanism underlying phase variation is unknown. To determine whether RecA-mediated processes are involved in phase variation, the recA gene of A. lipoferum 4B was cloned and sequenced and a recA mutant (termed 4BrecA) was constructed by allelic exchange. Strain 4BrecA showed increased sensitivity to UV and MMS compared with 4B and impaired recombinase activity. The ability to generate variants in vitro was not altered; the variants from 4BrecA exhibited all morphological and biochemical features characteristic of the variant generated by strain 4B. However, the frequency of variants generated by 4BrecA was increased by up to 10-fold. So, in contrast with many studies showing the abolition or a large reduction of the frequency of phase variation in recA mutants, this study describes an enhancement of phase variation in the absence of a functional recA.
nutrient agar supplemented with 0.0005% bromothymol blue
triphenyl tetrazolium chloride
Bacterial populations endure fluctuating conditions in many kinds of environment. Phase variation is one adaptive process by which bacteria undergo frequent and sometimes reversible multiple phenotypic changes resulting from genetic alteration at specific loci of their genomes. Various genomic reorganization promoting phase variation have been described (see  for review). For example, recombinational deletion events occur for switching of type IV pili in Neisseria gonorrhoeae and for loss of a pathogenicity island in Yersinia pestis; DNA inversion results in the differential expression of surface layer proteins in Campylobacter fetus. Amplification was shown for toxin genes of Vibrio cholerae, and for switching from a smooth pathogenic to a rough non-pathogenic form of the fungal pathogen Pseudomonas tolaasii. In all these examples, the occurrence of phase variation relies on a functional recA gene [7–11]. In contrast, other mechanisms of phase variation are epigenetic, as the primary DNA sequence is not altered, such as methylation-controlled expression of the pap operon in Escherichia coli or translational control by slipped strand mispairing for opacity proteins of Neisseria meningitidis.
Azospirilla are predominantly root surface-colonizing bacteria . Plant growth promotion by Azospirillum seems to be mainly due to production of phytohormones. The most abundant phytohormone produced is the auxin indole-3-acetic acid, allowing an increase in the number of lateral roots and root hairs, which results in higher nutrient and water uptake by the inoculated plant (see  for review).
Azospirillum lipoferum 4B, a strain isolated from the rice rhizosphere, generates in vitro at high frequencies (10−4 to 10−3 per cell per generation) a non-reversible variant form designated 4VI. Variant colonies (4VI) are readily distinguishable from wild type colonies by the differential absorption of dyes incorporated into the growth medium. The variant 4VI exhibits pleiotropic modifications; assimilation of several carbohydrates differs between 4B and 4VI; in contrast to 4B, 4VI cannot reduce triphenyl tetrazolium chloride (TTC), cannot bind some dyes  and has lost the ability to swim, as it lacks a polar flagellum and constitutively expresses mechanosensing lateral flagella . Variant 4VI keeps its ability to promote plant growth (unpublished results). Emergence of a non-swimming variant is thought to be important for colonization of the plant root by the bacteria .
The mechanism underlying these non-reversible changes in A. lipoferum 4B is unknown. To determine the role of recA in phase variation in A. lipoferum, we cloned and characterized the recA gene of A. lipoferum 4B. A recA mutant was generated by allelic exchange and its ability to generate variants in vitro was studied and compared to that of the wild type strain.
2Materials and methods
2.1Media and bacterial growth conditions
Strains and plasmids used in this study are presented in Table 1. Strains of E. coli were grown in Luria–Bertani (LB)  broth or agar at 37 °C. A. lipoferum strains were grown in TY broth or agar  at 28 °C. Where appropriate, antibiotics were added at the following final concentrations: 100 μg ml−1 ampicillin (Ap), 40 μg ml−1 colimycin (Col), 25 μg ml−1 chloramphenicol (Cm), 40 μg ml−1 kanamycin (Km), 10 μg ml−1 tetracycline (Tet).
Table 1. Bacterial strains and plasmids used in this study
Strains and plasmids
Source or reference
Strain isolated from rice rhizosphere (Camargue, France), Colr
pSUP106 carrying a 9 kb genomic fragment of A. lipoferum 4B containing the recA gene
pBluescript carrying a 3.1 kb genomic fragment of A. lipoferum 4B containing the recA gene
pR1.9 with a Kmr cassette inserted into SphI site of recA
pSUP202 carrying a 4.4 kb fragment from pR1.11, Kmr
Bacterial conjugations were performed by triparental matings using pRK2013 as a helper . Donor E. coli and recipient A. lipoferum strains were grown overnight with gentle shaking, respectively in LB and TY media containing appropriate antibiotics. Cells were collected by centrifugation, gently washed and resuspended in 10 mM MgSO4. These suspensions were mixed to achieve a donor:recipient ratio of 1:10 or 1:5 and were placed on top of a nitrocellulose filter laid onto the surface of a TY plate. Plates were incubated at 28 °C for 20 h. The filters were then removed and shaken in 1 ml of glycerol:TY (in a 1:1 ratio) to recover the cells. Transconjugants were selected on the minimal nitrogen free basal medium supplemented with 1 g l−1 of NH4Cl (NFBm) containing appropriate antibiotics and incubated at 28 °C .
Preparations of plasmid DNA or total genomic DNA, restriction digestions, ligations, Southern blotting and hybridization were carried out according to established protocols . Transformation of E. coli cells was carried out by electroporation with a Gene Pulser apparatus (Bio-Rad Laboratories) according to the manufacturer's instructions. Sequencing was performed with a MegaBACE 1000 (Amersham Pharmacia Biotech) capillary sequencer, and results were analysed with the SEQUENCE ANALYSER 2.1 software (Amersham Pharmacia Biotech). Searches for similarity to known DNA and protein sequences in databases were performed using the Blast algorithms . Similar DNA and aminoacid sequences were aligned using CLUSTAL W . RNA secondary structures and their folding energies were computed using the http://www.bioinfo.rpi.edu/applications/mfold site .
2.3Cloning of the A. lipoferum recA gene
PCR amplification of a 524 bp internal fragment of the recA gene of A. lipoferum 4B was done using degenerate oligonucleotides F1193 (GGNAAGACCACGCTGGCNCT) and F1194 (TTGACGCGNGTNTGGTTGCC) based on several aligned recA sequences of GenBank. DNA amplifications by PCR (50 μl) were performed according to the Taq polymerase manufacturer (Gibco BRL-Life Technologies). One hundred ng template of A. lipoferum 4B DNA was used per reaction tube. The amplification cycle consisted of an initial 5 min at 95 °C, 35 cycles of 45 s at 95 °C, 1 min at 60 °C, 1 min at 72 °C, followed by a final 7 min extension at 72 °C. The amplicon was sequenced allowing to design new specific oligonucleotides: F1337 (GACTTGGAGATCGAGCGGT) and F1338 (CCATCGCCCAGGCCCAGAA). These primers amplified a specific internal fragment of 315 bp of the recA gene of A. lipoferum 4B, but not the E. coli recA gene. They were then used to screen a genomic library of strain 4B constructed in plasmid pSUP106 . The amplification cycle consisted of an initial 5 min at 95 °C, then 40 cycles of 1 min at 95 °C, 1 min at 64 °C, 1 min at 72 °C, followed by a final extension of 7 min at 72 °C. A positive clone (designated pR1.8) was obtained and analysed by Southern hybridization after digestion with either AvaI, BglII, HindIII, PvuII, SalI, ScaI or SphI using as a probe the PCR fragment of 315 bp randomly labelled by [α-32P]dCTP (using the kit Random Primed DNA Labelling, Boehringer Mannheim). A positive 3.1 kb PvuII fragment was ligated into pBlueScript cut by EcoRV and this plasmid was designated pR1.9.
2.4Construction of a recA mutant
The suicide delivery vector required for allelic exchange mutagenesis was constructed as follows. A 1164 bp fragment encoding a kanamycin-resistance gene (nptII) was amplified from Tn5 using oligonucleotides F1212 (GCAGGTAGCATGCAGTGGGCTT) and F1213 (GAAGAACTCCAGCATGCGATCC); this fragment was blunted (using the SureClone kit, Amersham), inserted into pR1.9 previously digested with SphI (which has a single site in the recA gene), and filled with the Klenow subunit of DNA polymerase I. This insertion resulted in a recA::Km construct which was composed of 534 bp and 576 bp of recA flanking the 5′ and 3′ ends of the Km fragment, respectively. The resulting plasmid (pR1.11) was then digested with XbaI and ClaI; the released fragment was excised, filled with Klenow subunit. The resulting blunt-ended recA::Km fragment was ligated into the vector pSUP202 digested with EcoRV to produce pR1.61. Plasmid pR1.61 was then used to transform E. coli DH10B. Conjugal transfer of pR1.61 between this transformant E. coli and A. lipoferum 4B was performed. Co-integrates were selected by their resistance to Km. Since pSUP202 is suicidal in Azopirillum, Km-resistant exconjugants can only be generated by recombinational events between the cloned DNA fragment and its genomic homologue. To discriminate between single or double cross-over, Km-resistant exconjugants were checked for loss of the vector pSUP202 by ampicillin and chloramphenicol sensitivity. Allelic replacement of recA with recA::Km was confirmed by PCR and Southern hybridization analysis.
2.5UV and MMS sensitivity
Cells from an exponential liquid culture in TY broth with antibiotics where appropriate were harvested and diluted serially in NaCl 0.8%, and the suspensions were plated onto TY media containing 0.005% methyl methane sulfonate (MMS, Sigma) or TY media only for the UV experiment. After drying for 10 min, the TY plates were UV-irradiated (312 nm, 5 J/s/m2, UV Transilluminator 2000, Bio-Rad) for different times (up to 90 s). In both cases, the plates were incubated for 2 days at 28 °C in the dark. Relative survival was calculated by comparing the number of c.f.u. in the non-treated sample. The UV experiment was done in triplicate; for 4B and 4BrecA, the experiment was repeated twice with similar results. The MMS experiment was done in triplicate.
The recombinase activity of the recA mutant was compared with that of the wild type. Fragments of 1–4 kb of A. lipoferum 4B genomic DNA digested with HindIII and EcoRV were purified and ligated into the mobilizable suicide vector pSUP202 (digested at HindIII and EcoRV sites both present in the gene conferring tetracycline resistance). The resulting plasmids were introduced into E. coli S17.1 by transformation, selecting for Cm and Ap resistance, and Tet sensitivity. Ten clones were randomly selected and these constructs were introduced into 4B and 4BrecA by conjugation; exconjugants where a single recombination event had occurred were selected on NFBm containing Col, Cm and Ap. The recombination frequency was then calculated by comparing the number of exconjugants to the total number of recipient cells. This experiment was repeated 2 times.
2.7Phenotypic characterization of variants
Variant colonies are readily distinguishable from wild type colonies by the differential absorption of dyes added to the growth medium. Thus variants from 4BrecA were analysed by plating bacteria onto nutrient agar (Difco) supplemented with 0.0005% (w/v) bromothymol blue (NAB medium). TTC reduction was tested on nutrient agar containing 0.004% (w/v) TTC. Binding of the dye aniline blue was tested on nutrient agar containing 0.001% (w/v) aniline blue. Binding of Congo red was tested on nutrient agar containing 0.025% (w/v) Congo red. Morphology of colonies on NAB following dye binding was examined after 5 days of incubation at 28 °C. Acid production from sugars was tested as described previously . Bacterial motility was examined by placing a 10 μl drop from exponentially growing cultures between a microscope slide and a coverslip, and observing through a phase-contrast microscope (Zeiss). For swimming ability, bacteria from exponentially growing cultures were applied as a small drop (10 μl) in a TY semi-solid agar plate (0.5% agar w/v) and incubated for 2 days at 28 °C.
2.8Frequency and kinetics of appearance of variants
100 μl of a suspension obtained from a single colony of 4B or 4BrecA were serially diluted and spread onto NAB plates in order to calculate the total number of cells in the colony (N) and to determine the number of generations (n=log2N). The proportion of variants was then estimated (number of variants out of total number of bacteria counted on plates). Phase variation frequency was calculated as the proportion of variants per number of generations and using 15 individual colonies for each strain, according to the method of Stocker .
To follow the proportion of variants during growth, individual colonies of 4B, 4BrecA, 4BrecA with pSUP106 and 4BrecA complemented with pR1.8 were grown in TY broth at 28 °C with appropriate antibiotics when needed. Samples were taken at seven levels of absorbance (580 nm), serially diluted and plated onto NAB medium. After a 5-day incubation at 28 °C, colonies were counted and the proportion of variants was evaluated on a minimum of 1000 colonies for each strain at each absorbance. The same experiment was done three times independently for each strain.
3.1Characterization of the recA gene of A. lipoferum 4B
Using a PCR-based approach, the recA gene of A. lipoferum 4B was obtained from a genomic library and a 3.1 kb fragment containing recA subcloned in pBluescript, yielding pR1.9. About 2.9 kb of the pR1.9 plasmid were sequenced (GenBank Accession Number AY422794) and revealed a (G + C) content of 65%, a typical value for Azospirillum DNA. The fragment contains one entire open reading frame and a partial one. The entire open reading frame of 1110 bp (including the stop codon) encodes a putative protein of 369 aminoacids, with a predicted molecular weight of 39 404 Da. The deduced aminoacid sequence shares 71% and 72% identity with RecA of respectively Agrobacterium tumefaciens and Rhodopseudomonas palustris[28,29] (Fig. 1). Aminoacid residues associated with functional activities, which include co-protease activity, ATP fixation and DNA binding were all highly conserved in the RecA protein of A. lipoferum (Fig. 1). Two putative promoters sequences GAACAN6GTAC were found 70 bp and 133 bp upstream from the ATG initiation codon of the recA gene of A. lipoferum. This sequence found in the promoter region of recA of several α-Proteobacteria was proposed to be an equivalent of the SOS box (i.e. the LexA binding site) and is implicated in recA expression in R. palustris. The typical sequence of the SOS box as found in many bacteria including E. coli is CTGN10CAG, but this sequence has never been found in α-Proteobacteria . At the 3′ end of the recA gene and overlapping the stop codon, lies a 29 nucleotides GC-rich hairpin-forming sequence with a folding energy of −147.5 kcal mol−1 followed by six thymines, which could function as a transcriptional terminator. Such stem-loop structures have been found downstream of Xenorhabdus bovienii and X. nematophilus recA gene .
Upstream of recA and in the same orientation, a partial open reading frame of 507 bp coding for a putative sensor of a two-component regulatory system was found. The deduced aminoacid sequence shares 49% and 46% identity with sensory histidine kinases of respectively Brucella suis and Mesorhizobium loti[33,34]. No putative open reading frame was detected in the 1044 bp downstream of recA. It is noteworthy that no recX gene was found in close vicinity of recA; recX encodes a small protein involved in SOS response and is implicated in regulation of recA expression in several bacteria including E. coli.
3.2Construction of A. lipoferum recA mutant
To assess whether recA is important in phase variation of A. lipoferum 4B, we sought to inactivate recA. Using insertional mutagenesis by rescue of a marker from a suicide plasmid (in this case pSUP202), we introduced a kanamycin resistance gene (Km) at the SphI site located 534 bp from the start of recA and thus generated strain 4BrecA. PCR analysis (using primers F1337 and F1338) confirmed that a double recombination event had indeed occurred in strain 4BrecA; a 320 pb PCR fragment was amplified from DNA of the parental strain 4B whilst a 1.4 kb amplicon was obtained from strain 4BrecA, which corresponds to the insertion of the 1.1 kb Km resistance fragment. Furthermore, Southern hybridization analysis confirmed that the disrupted recA allele was successfully integrated in place of the recA allele present in single copy in A. lipoferum 4B genome (data not shown).
3.3Phenotypic characterization of the A. lipoferum recA mutant
Since recA mutants are expected to be more sensitive to DNA-damaging agents than their isogenic parental strains, 4B, 4BrecA and 4BrecA complemented with pR1.8 were first tested for their sensitivity to UV irradiation and MMS sensitivity. The three strains were exposed to UV light for increasing time periods and their survival was monitored by plating. After a 15-s exposure, the 4BrecA strain showed only 5% survival whereas the parental strain and strain 4BrecA(pR1.8) retained respectively 40% and 28% survival (Fig. 2). After a 30-s exposure, only 1%, 36% and 18% survival were observed respectively for 4BrecA, 4B and 4BrecA (pR1.8). 4B and 4BrecA were also plated on medium containing 0.005% MMS; survival was 0.75% for 4BrecA versus 75% for 4B. Thus, 4BrecA is clearly more sensitive to UV light and MMS than the parental strain and this is consistent with a role of recA in DNA repair.
Second, the recombinase efficiency of 4BrecA was compared with that of the wild type. Whereas a proportion of 5.5 × 10−8 recombinants was observed for strain 4B, no recombinant could be detected for 4BrecA in the conditions tested. So 4BrecA mutant is clearly impaired in the recombinase activity.
3.4Ability of 4BrecA to generate variants
When a single colony of 4BrecA was suspended in KCl buffer and plated onto NAB, two types of colonies developed. Most colonies were large and bound the dye (Fig. 3), as 4B does; a few colonies termed 4VIrecA were translucent (i.e. they did not bind the dye), and thus had the typical phenotype previously observed for the variant 4VI. Other physiological and biochemical traits shown to be linked to phase variation in A. lipoferum 4B were analysed on these two colonial forms (using 20 colonies of each form). First, swimming ability was analysed by microscope observation: whereas 4BrecA retained the typical swimming of spirilla, 4VIrecA (like 4VI) was unable to swim. Moreover, spreading through semi-solid agar was observed for 4B whereas no spreading was observed for the variant (data not shown). Analysis of sugar assimilation revealed that 4VIrecA could no longer utilize sorbitol, mannitol and glucose but grew on fructose (Table 2), as for 4VI. Finally, the ability to reduce TTC was lost for the 4VIrecA cells. Thus 4BrecA was capable to generate variants and those variants retained the same features as the variants formed by the parental strain 4B.
Table 2. Physiological and biochemical characteristics of A. lipoferum variants 4VI and 4VIrecA in comparison with the 4B and 4BrecA strainsa
a+ positive reaction; − negative reaction.
Binding of dye
Acid production from
To check the stability of 4VIrecA variants, single colonies of 4VIrecA were grown individually in TY broth, sub-cultured 4 times and plated onto NAB medium. All the colonies that grew were translucent so it was concluded that no reversion event had occurred, as previously observed for variant 4VI.
3.5Frequency and kinetics of appearance of variants
As 4BrecA was still able to generate variants, we were interested in evaluating the frequency of occurrence of variants. A frequency of 1.59 (±1.03) × 10−4 and 5.34 (±2.52) × 10−4 variants cell−1 generation−1 was found for 4B and 4BrecA, respectively.
Second, the kinetics of variant appearance was assessed for strains 4B, 4BrecA, 4BrecA(pSUP106) and 4BrecA(pR1.8) in TY broth (Table 3). Data clearly show that the proportion of variants was increased in strain 4BrecA by about 10-fold. For 4BrecA, the proportion of variants increased markedly during the early exponential phase, with a maximum reached at an absorbance of 0.6; in contrast the highest proportion of variants for strain 4B was reached in stationary phase. The proportion of variants for 4BrecA(pSUP106) (i.e. carrying the empty vector) was checked at two different absorbances and found identical to that of 4BrecA. The complemented strain 4BrecA(pR1.8) exhibited an intermediate phenotype between 4B and 4BrecA. Thus, not only a recA strain still generates variants but the frequency of appearance of variants is at least 10-fold larger than for the wild type strain.
Table 3. Kinetics of appearance of variants
Absorbance at 580 nm
Number of variants for 100 cellsa
aValues represent the means±standard deviation of data from three independent experiments. NT, not tested.
0.25 ± 0.03
1.43 ± 0.76
1.88 ± 0.63
0.21 ± 0.30
0.89 ± 0.64
13.91 ± 5.23
1.51 ± 0.45
1.62 ± 0.74
17.81 ± 6.52
17.95 ± 3.92
1.85 ± 0.16
1.31 ± 0.48
25.72 ± 7.15
4.47 ± 1.40
1.87 ± 0.34
25.87 ± 7.04
8.56 ± 1.00
2.62 ± 1.20
24.22 ± 7.71
9.12 ± 0.63
2.88 ± 1.06
19.60 ± 1.43
7.46 ± 0.87
Most physiological and ecological studies on Azospirillum have focused on A. brasilense and A. lipoferum, because of the importance of these species for plant growth promotion and crop inoculation . However, the implementation of standard genetic protocols with A. lipoferum is difficult, and consequently most genetic information on Azospirillum has been derived from A. brasilense only. Here, inactivation of the recA gene of A. lipoferum 4B was achieved by allelic exchange with recA::Km using a pSUP202-based suicide vector. This is the first report of an A. lipoferum insertion mutant. As many as 20 attempts were necessary to obtain a mutant in which the expected double recombination event had occured. We foresee that the recA-deficient strain described here will also be a valuable tool in future genetic studies of A. lipoferum 4B.
The recA insertion mutant had significantly reduced ability to repair DNA lesions caused by exposure to UV radiation or an alkylating agent, as previously reported for A. brasilense Sp7  and other bacteria. As in X. bovienii, resistance to UV-induced damage was only partially restored to wild type level when complementing with a plasmid-borne functional copy of recA. Since it could be that the recA::Km mutant 4BrecA expresses the N-terminal half of the protein (which would lack the C-terminus), abnormal multimers containing truncated and wild type peptides might form and hinder the coprotease and recombination activities of RecA in the complemented strain . Recombinase activity is severely impaired in 4BrecA; as the proportion of recombination for strain 4B is 5.5 × 10−8, we cannot exclude that 4BrecA is still able to perform recombination but at a lower, undetectable rate.
Besides its role in homologous recombination, RecA seems to play an active role in modulating the expression of genes in many bacteria adapted to changing environments. In the phytopathogen Erwinia carotovora, pectin lyase (PnlA) production is induced by DNA-damaging agents and RecA proteolytic activity acts to relieve pnlA gene repression . Synthesis of nuclease, chitinases and lipases in Serratia marcescens is partly regulated by the SOS response but, in this case, RecA acts through the classical cleavage of LexA . Therefore, we cannot exclude such a role for RecA in A. lipoferum 4B, for example by cleaving a repressor that might be somehow involved in phase variation.
Several observations point to the occurrence of genomic rearrangements during phase variation of A. lipoferum 4B. First, the high rate of variant emergence (10−4 to 10−3 per cell per generation) is typical of genomic rearrangements rather than point mutations ; second, the random amplified polymorphic DNA (RAPD) profile of the variant 4VI differed from that of the wild type A. lipoferum 4B (our unpublished results). Different types of programmed genetic rearrangements involving homologous recombination (deletions, duplications, etc.) can take place during phase variation, and many are RecA-dependent. For example, phenotypic switching in P. tolaasii from a smooth pathogenic form to a rough non-pathogenic form is due to a reversible 661-bp duplication in the putative kinase domain of the regulatory locus pheN which occurs less frequently (∼5% of wild type level) in a recA deficient background . Loss of various physiological traits (including virulence) in Yersinia pestis is due to the RecA-dependent deletion of 102 kb of chromosomal DNA flanked by the insertion sequence IS100[3,8]. In Neisseria gonorrhoeae, pilus antigenic variation results from intragenic recombination between the pilin expression locus and incomplete pilin gene copies , and it occurs at a 100- to 1000-fold reduced rate in a recA mutant .
However, there are instances where DNA rearrangements associated with phase variation are catalysed by RecA-independent, site-specific recombination systems. In Pseudomonas fluorescens, phenotypic variants altered in motility (i.e. overexpressing flagellin) appear during root colonization, at a rate that is reduced following inactivation of a site-specific recombinase gene important for root colonization [42,43]. Similarly, deletion of 30 kb from the chromosome of Legionella pneumophila is the origin for LPS phase variation and occurs in a RecA-independent process . Indeed, inversions generally rely on site-specific recombinases , except in Campylobacter fetus where differential expression of surface layer proteins is due to recA-dependent DNA inversions . Despite its primary sequence conservation and ubiquity in bacteria, it is clear from these examples that phase variation originates from many different processes and that one cannot anticipate the role of RecA in a particular strain.
The most relevant finding of this study was that the mutant A. lipoferum 4BrecA not only retained the ability to generate variants but surprisingly the frequency of variants generated was enhanced compared to that of the wild type. Indeed, in all studies where the role of RecA has been evaluated, the authors saw either the abolition of variation or no effect on the frequency of variation. Thus, this study is the first to show that phase variation can increase in the absence of a functional recA. Moreover the phenotypic switch observed in 4B cannot be attributed to a recombination event performed via RecA. Further work will aim at understanding the mechanism underlying phase variation in A. lipoferum.
We are grateful to Jacqueline Haurat for technical assistance and Céline Lavire for helpful discussions. This research was supported by the Ministère de la Recherche et des Nouvelles Technologies.