Introduction

One of the most obvious features of phase variation is the appearance of a minority of colonies or colony sectors displaying a different aspect. Phase variation or phenotypic switch is used by several bacterial species to generate intra-population diversity that increases bacterial fitness and is important in niche adaptation, or to escape host defences. Phase variation allows that the expression of a given phenotype is either ON or OFF; these events are usually reversible (ON ↔ OFF) but may be irreversible (ON → OFF or OFF → ON), and result from genetic or epigenetic alterations at specific loci. In contrast to spontaneous mutations, which occur at a frequency of approximately 10−8 to 10−6 mutations per growing cell per generation, phase variation occurs at frequencies higher than 10−5 switches per cell per generation and always affects the same phenotype(s). Phase variation has been described for many different bacterial genera belonging to diverse taxonomic groups and displaying different ecological behaviours (pathogens, saprophytes, symbionts) and can regulated various phenotypes, such as motility, synthesis of pili, expression of capsule, production of antifungal metabolites (Table 1).

Table 1 Phase and antigenic variation mechanisms leading to genome modifications

Related to phase variation, antigenic variation refers to the expression of a number of alternative forms of an antigen on the cell surface (such as lipoproteins, polysaccharides, type IV pili); this generates within a clonal population individual cells that are antigenetically distinct, allowing bacterial pathogens to escape the host immune system. The term antigenic variation is sometimes used in a broader sense, including the alternative expression (ON ↔ OFF) of genes specifying antigenic identity of a cell surface structure, but here these cases will be classified in phase variation. At the molecular level, some cases of antigenic variation share common features with phase variation mechanisms.

This review will focus on phase and antigenic variation mechanisms implying genome modifications; epigenetic mechanisms such as differential methylation of promoter sequences will not be developed and the reader is referred to other reviews (Henderson et al. 1999; van der Woude and Baumler 2004; van den Broek et al. 2005a). The aims of the present review are: (i) to present an overview of the molecular mechanisms underlying phase and antigenic variation, (ii) to emphasize on the diversity of phenotypes regulated by these mechanisms, and (iii) to discuss the ecological relevance of variant appearance within a given population.

Gene conversion

Gene conversion, highly documented for antigenic variation, involves a recombination event between a silent copy of a gene and another copy that is expressed, and leads to the formation of a new chimeric gene. When several copies of the silent gene are present, numerous chimeric sequences can be theoretically generated, allowing to express various forms of an antigen. There is no common mechanism for gene conversion, and in some cases, proteins of the homologous recombination pathway can be implicated.

Surface proteins in Borrelia

Borrelia burgdorferi, the causal agent of Lyme disease, can generate variants expressing different antigenic forms of the VlsE surface lipoprotein. A linear 28 kb-plasmid contains one functional copy of vlsE and fifteen silent vls cassettes. Segments of the silent vls cassettes can recombine with the central region of the vlsE gene, generating antigenic diversity in the mammalian host. The donor vls cassette and the extremities of the vslE gene remain unchanged during this recombination (Zhang and Norris 1998). By this process, B. burgdorferi escapes the humoral immune system (Bankhead and Chaconas 2007). Although the precise mechanism underlying this conversion event remains to be elucidated, proteins mediating homologous recombination like RecA are thought to be involved (Liveris et al. 2004). A similar mechanism occurs for antigenic variation of the Borrelia hermsii Vmp lipoprotein (Plasterk et al. 1985; Restrepo et al. 1994).

Outer membrane proteins in Helicobacter pylori

The outer membrane protein BabA mediates attachment of H. pylori, a gastric pathogen, to the Lewis B blood group antigen on gastric epithelium. H. pylori cells no longer expressing the BabA protein but expressing the BabB protein appear at a frequency of about 10−3, following a RecA-dependent recombination event between the genes encoding these two proteins (Pride and Blaser 2002; Solnick et al. 2004). Although the function of BabB has not been elucidated, this gene conversion event is thought to provide a control of adhesion allowing to adapt to different niches within the stomach and to escape the host immune system (Solnick et al. 2004).

Pili in Neisseria gonorrhoeae

Development of symptoms associated with gonorrhoeae is correlated with the ability of N. gonorrhoeae cells to attach and colonize mucous membranes via their pili, despite urinary flux. Various molecular mechanisms, including slipped-strand mispairing, account for the apparition of variants in vivo and in vitro displaying modified surface properties, that are affected in opacity proteins, LPS or pili synthesis (for a review, van der Woude and Baumler 2004). Type IV pili are composed of identical subunits of pilin, a 18-kDa protein encoded by pilE. N. gonorrhoeae strains carry one or two functional pilE loci and four to six pilS loci (each locus contains several variable silent copies of pilS), lacking promoter sequences and the 5′ portion of the gene (Haas and Meyer 1986; Segal et al. 1986) (Fig. 1). Each silent copy of pilS can exchange portions of variable regions with a pilE locus; in this well-documented conversion event involving proteins of the homologous recombination pathway, the pilE locus is altered whereas the pilS donor locus remains unchanged (Howell-Adams and Seifert 2000). Pilin antigenic variation exhibits the highest reported frequency of any pathogenic gene conversion system (0.13 recombination events per cell) and can account for the extensive pilin variation detected during human infection (Criss et al. 2005). Among the numerous sequences encoding pilin that can be theoretically generated, two particular forms of pilins can appear: the soluble pilin (S pilin) secreted in the extracellular medium originates from the appearance of an amber mutation in pilE and the L pilin accumulating within the periplasm originates from the transfer of several silent copies of pilS into pilE. In both cases, variant cells are non-piliated (Haas et al. 1987). Recombination between pilS and pilE loci is increased in an iron-deficient medium, suggesting that iron limitation signals for recombinational events triggering pilin antigenic variation, allowing N. gonorrhoeae to colonise different sites (Serkin and Seifert 2000). By modifying its antigenic properties, N. gonorrhoeae escapes the human immune system and is able to survive in its unique host.

Fig. 1
figure 1

Pilin antigenic variation in Neisseria gonorrhoeae by gene conversion. pilE and pilS loci are composed of a constant 5′ region (which is shorter for pilS), a semi-variable region (SV), a hypervariable loop (HVL) flanked by two 30-pb conserved sequences, cys1 and cys2 and a hypervariable tail (HVT). DNA exchange occurs between a silent pilS locus and the pilE locus of the donor chromosome at a short region of homology (small black boxes). The recombination events are indicated by the crosses. The initial model accounting for this gene conversion event, involves duplication of the pilE locus following pilE/pilS recombination, which would then lead to the excision of a pilE-pilS hybrid on a closed-circular piece of DNA. This episome would then recombine with pilE via two crossover events, leading to pilE gene variation, as depicted here (Howell-Adams and Seifert 2000). A recent finding demonstrating the involvement of the RecBCD recombination pathway suggests that pilE gene variation proceeds via a double-chain-break repair model that utilizes the RecBCD enzyme (Hill et al. 2007). Adapted from (Howell-Adams and Seifert 2000)

Site-specific inversion

In this process, recombinases recognize short inverted repeat sequences located upstream and downstream of the element to be inverted, an element that usually contains a promoter. The inversion event acts as a switch triggering the expression of a gene initially silent (OFF → IN mode) or impeding the expression of a gene initially expressed (IN → OFF mode).

Polysaccharides and outer membrane proteins in Bacteroides fragilis

Genome sequencing of Bacteroides fragilis, a bacterium of the intestine microflora that is sometimes pathogen, and several studies evidenced various regions undergoing inversion (Krinos et al. 2001; Cerdeño-Tarraga et al. 2005). At least eight different regions involved in polysaccharides biosynthesis undergo inversion-mediated antigenic variation, allowing B. fragilis to synthesize different types of capsules. Mpi, the site-specific recombinase involved in these inversions, not only inverts the promoters of loci involved in polysaccharide biosynthesis but can act on other promoters of genes encoding products of unknown functions (Krinos et al. 2001). Other invertases were evidenced with the genome of B. fragilis, some achieving the inversion of promoters related to proteins of the outer membrane (Weinacht et al. 2004; Cerdeño-Tarraga et al. 2005). If DNA inversion usually controls the expression of a single class of molecules, this mechanism in B. fragilis seems to be dedicated to controlling the expression of different classes of molecules, allowing this putative opportunistic pathogen to escape the immune system and to colonize new sites (Cerdeño-Tarraga et al. 2005).

Surface Layer Proteins (SLP) of Campylobacter fetus

Campylobacter fetus, an opportunistic human pathogen and the major cause of ovine abortion, can generate variants at a frequency of 10−3 affected in the size and antigenicity of SLP. Campylobacter fetus possesses eight SLP cassettes, encoding proteins from 97 to 149 kDa. All these cassettes share a 600-pb sequence homology beginning 74 pb before the start codon but only one of the copies (sapA) has a functional promoter (Dworkin and Blaser 1996). SLP variation occurs through the inversion of a DNA fragment containing the sapA promoter (Fig. 2a); this inversion event moves the sapA promoter upstream one of the SLP cassette, allowing the exclusive expression of this cassette. Although the size of the inverted fragment is variable, the frequency of inversion does not seem to be dependent upon the size of the inverted fragment (Dworkin and Blaser 1996). Unlike most of the inversion systems that do not rely on RecA activity, the major inversion pathway of SLP in C. fetus is RecA dependent as a recA-strain exhibits a reduced frequency of SLP variability; alternative lower-frequency, RecA-independent inversion mechanisms exist (Dworkin and Blaser 1997; Ray et al. 2000). Although SLP appear essential for colonization and/or translocation to the placenta of ewes, they are not required to mediate fetal injury (Grogono-Thomas et al. 2000). SLP switching delays the host antibody response allowing pathogen persistence in an immunologically hostile environment (Grogono-Thomas et al. 2003).

Fig. 2
figure 2

Phase and antigenic variation by site-specific inversion. (a) Antigenic variation of Surface Layer Proteins (SLP) of Campylobacter fetus. The 5′ conserved region and variable regions of the SLP gene cassettes are represented respectively by small stripped box and thick arrows. Only one SLP gene (sapA) has a functional promoter (over bent arrow). DNA inversion takes place between two oppositely oriented cassettes flanked by a 6.2-kb DNA element, following DNA exchange within the 5′ conserved region. Inversion of DNA segments containing the sapA promoter (over bent arrow) allows expression of alternated SLP gene cassettes (mRNA depicted as thin dashed arrows). Two inversion events and their resulting new genotypes are presented: (1) inversion of the 6.2-kb element alone; (2) inversion of the 6.2-kb element and one SLP gene cassette. For clarity, only three SLP cassettes have been drawn. Adapted from (Dworkin and Blaser 1997). (b) Phase variation of type 1 pili in E. coli. The relative positions of the promoters (over bent arrows), genes (thick arrows), and inverted repeats IRR and IRL (triangles) are shown. The invertible DNA element is framed by the inverted repeats. IRR and IRL are located within the binding sites for the recombinases FimB and FimE. Binding sites for other regulatory proteins (such as IHF and Lrp) are not shown. Adapted from (van der Woude and Baumler 2004). (c) Phase variation of the H1 and H2 flagellins of Salmonella enterica serovar Typhimurium. A 955-bp fragment flanked by repeated sequences (hixL and hixR) gets inverted by the action of the Hin recombinase. This fragment bears the gene encoding the Hin recombinase and a promoter (over bent arrow) allowing the transcription of fljB (encoding flagellin H1) and of fljA (encoding a negative regulator for fliC expression). In the ONH1/OFFH2 position, fljB is cotranscribed with fljA. FljA binds to fliC (encoding flagellin H2) mRNA, inhibiting its translation and triggering its degradation. After the inversion event (OFFH1/ONH2 position), fljB and fljA are no longer expressed, allowing fliC mRNA to be translated. mRNA are depicted as thin dashed arrows. Adapted from (Yamamoto and Kutsukake 2006)

Type I pili of Escherichia coli

Escherichia coli can generate, at a frequency of 10−3 per cell and per generation, cells lacking pili (Abraham et al. 1985). The mechanism underlying this phenotypic switch relies on the inversion of part of the fim operon encoding the type I pili, which are essential for colonization and attachment to eukaryotic cells. The fimA gene, encoding the major subunit of type I pili, can either be transcribed (ON position) or be silent (OFF position). Upstream of fimA lies an invertible element containing a promoter (Fig. 2b); the inversion of this element modifies the promoter orientation and abolishes fimA transcription (OFF position). The inversion is mediated by two site-specific recombinases: FimB and FimE (Klemm 1986). Whereas FimB allows the inversion in both ways, FimE can only mediate the transition from the ON position to the OFF position (Gally et al. 1993). Other regulatory proteins are involved, such as H-NS, Integration Host Factor (IHF), Leucine responsive protein (Lrp), RpoS (Blomfield et al. 1997; Blomfield 2001). pH and osmolarity also influence this invertible switch (Schwan et al. 2002).

Root colonization in Pseudomonas fluorescens

While searching for P. fluorescens WCS365 mutants deficient in root colonization, the role of a recombinase displaying homologies with site-specific recombinases of the λ integrase family was evidenced. A strain inactivated in this locus (sss) colonizes less efficiently than the parental strain the roots of several plants (Dekkers et al. 1998).

During root colonization of alfalfa by Pseudomonas fluorescens F113, two types of variants (F and S) with increased motility appear; they preferentially colonize root apex and display unusual long flagella, due to the enhanced synthesis of flagellin. The type F variant also displays an increased synthesis of siderophores but no longer synthesizes antifungal biocontrol compounds such as hydrogen cyanide and exoproteases (Sanchez-Contreras et al. 2002). When alfalfa roots are inoculated with the parental strain, 10% of the root-inoculated cells become variants whereas less than 1% of variants are recovered after inoculation with a P. fluorescens F113 sss mutant (Sanchez-Contreras et al. 2002). A P. fluorescens F113 no longer expressing XerD, a second site-specific recombinase, generates fewer variants than the parental strain (Martinez-Granero et al. 2005). Conversely, strains overexpressing sss or xerD produce more variants than the wild type in vivo and in vitro.

The sss and xerD genes are induced by a non-diffusible compound of the plant root (Martinez-Granero et al. 2005). Phase variation could thus play an important role during root colonization and would be mediated by site-specific recombinases induced at the root site.

Flagellins in Salmonella enterica serovar Typhimurium

Salmonella enterica serovar Typhimurium cells can express two types of flagellins, H1 or H2, the transition from one serotype to another occurring at a frequency of 10−5 to 10−3 per cell and per generation (Stocker 1949). A 955-bp fragment bearing a promoter and flanked by repeated sequences (hixL and hixR) gets inverted by the action of the Hin recombinase, whose gene is located within the inverted fragment (Silverman et al. 1979; Heichman and Johnson 1990). More than a simple transition from an ON position to an OFF position, this system allows the transition from the ONH1/OFFH2 position to an ONH2/OFFH1 position (Fig. 2c). In the ONH1/OFFH2 position, the gene encoding flagellin H1 (fljB) is cotranscribed with fljA encoding a negative regulator for fliC expression. FljA binds to fliC mRNA (encoding flagellin H2), inhibiting its translation and triggering its degradation (Bonifield and Hughes 2003; Yamamoto and Kutsukake 2006). After the inversion event, fljB and fljA are no longer expressed, allowing fliC mRNA to be translated (Silverman et al. 1979; Bonifield and Hughes 2003; Yamamoto and Kutsukake 2006). In strain LT2, this inversion can also be mediated by Fin, another DNA invertase located within a resident prophage (Kutsukake et al. 2006). It was recently shown that deletion of luxS, a gene responsible for the synthesis of the quorum-sensing signaling molecule autoinducer 2, polarizes flagellar phase variation toward expression of flagellin H1 (Karavolos et al. 2008).

Phase variation of flagellins contributes to virulence in a murine typhoid infection model, as cells locked into expressing fliC appear to be more virulent and to survive better in the host than mutant cells locked into expressing fljB (Ikeda et al. 2001).

Insertion–excision

Iron oxidation in Acidithiobacillus ferrooxidans

Acidithiobacillus ferrooxidans (formerly Thiobacillus ferrooxidans) is able to use iron and other sulfur compounds as energy source and is mainly studied for its ability to solubilize metals in mining operations, a process known as biolixiviation. Variants displaying the capacity to spread rapidly on solid medium (LCS “large spreading colony”) were observed in vitro on medium containing both ferrous iron and thiosulfate as available energy sources. Those variants have lost the capacity to oxidize FeII but have retained the ability to oxidize sulphur, and could revert to the parental phenotypes (Schrader and Holmes 1988). The appearance of variants is correlated with the insertion of an IS in resB impeding the transcription of resB and resC, genes involved in the maturation of type C cytochromes essential for FeII oxidation (Cabrejos et al. 1999). Whether one or several transposition events occur or whether other modifications take place, remains to be determined, as well as the event permitting reversion.

Lipopolysaccharides in Legionella pneumophila

Cells of L. pneumophila, the causal agent of legionellosis, are entirely covered with lipopolysaccharides (LPS) which synthesis requires a locus composed of 25 genes. From a virulent strain, an avirulent form lacking flagella and LPS was isolated (Lüneberg et al. 1998). During the transition from wild-type to variant, a 30-kb region is excised from the chromosome and replicates like a high-copy number plasmid in the variant (Lüneberg et al. 2001). During reversion, the 30-kb region is inserted back into the chromosome at the initial location. Interestingly, the excised region does not bear any genes involved in LPS biosynthesis but contains ORFs displaying homologies with phage ORFs and genes involved in recombination (recE, recT, rusA). Products encoded by these genes could be responsible of this excision event that is RecA-independant; the excised region is likely to contain a gene encoding a global regulator acting on the different phenotypes altered by phase variation (Lüneberg et al. 2001).

The transition from wild-type to variant as well as the reversion was evidenced in vivo in guinea pig and factors from the host increase the frequency of phase variation (Lüneberg et al. 1998). Variants, although less competitive for infection of the host, could be adapted to the aquatic environment where L. pneumophila usually lives (Lüneberg et al. 2001).

EPS synthesis in Pseudoalteromonas atlantica

Pseudoalteromonas atlantica is able to colonise various marine environments and several components such as flagella, proteins of the outer membrane, EPS were shown to be involved in P. atlantica adhesion. Different colony morphologies can be distinguished on solid medium; type M colonies are large, opaque and smooth, type T colonies are intermediate in size, translucent, shiny and smooth whereas type C colonies are small and wrinkled due to the absence of EPS secretion. Type C colonies are unstable and can revert to types M and T (Bartlett et al. 1988). The mechanism underlying this phenotypic switch relies on the insertion of IS492 into a gene involved in EPS biosynthesis. A precise excision of IS492, occurring at the high frequency of 10−3 to 10−2 per cell and per generation, was reported allowing reversion to the wild-type phenotype (EPS+). IS492 belongs to the IS110 family of atypical insertion elements as no IR sequence can be detected at its extremities, and transposition requires the MooV transposase (Higgins et al. 2007). After excision, IS492 would remain in the genome in a circular form that may be an intermediate in transposition or a terminal product of excision (Perkins-Balding et al. 1999).

EPS synthesis appears necessary when P. atlantica develops biofilms on algae or on sand, but would become dispensable when bacteria go back to the marine environment in the planktonic form (Perkins-Balding et al. 1999).

Cell surface properties in Shigella flexneri

Bacteria belonging to the genus Shigella are responsible of shigellosis, a dysentery syndrome. Phase variation was evidenced in vitro at a frequency of 10−4 per cell per generation in S. flexneri. Variants are non-invasive, avirulent, form opaque orange colonies and do no longer express Ipa surface polypeptides. The appearance of variants was correlated with the insertion of IS1SFO into virF, an invasion plasmid-encoded positive regulator of ipa gene expression (Mills et al. 1992). Reversion to a virulent phenotype was observed within human hosts and corresponds to the excision of the insertion element. IS1SFO insertion into virF is thought to stabilize the invasion plasmid outside the host, a plasmid that is easily cured as virulence genes are not essential for survival outside the host; indeed, opaque variants with essentially unaltered invasion plasmid have a selective advantage compared to individuals with cured or deleted invasion plasmids that can no longer exploit the host environment (Mills et al. 1992).

Biofilms in Staphylococcus epidermidis/aureus

The virulence of the human pathogen S. epidermidis found on mucous membranes and skin, can be attributed partly to the formation of biofilms. Biofilm development is correlated with the production of the EPS PIA (Polysaccharide Intercellular Adhesin), which synthesis requires the ica operon composed of four genes (icaAicaD). From S. epidermidis cells able to form a biofilm, variants unable to develop a biofilm can be generated at a frequency of approximately 10−5 per cell per generation (Ziebuhr et al. 1997). Wild-type colonies are black on medium containing Congo red whereas small-colony variants are red as they no longer produce the PIA adhesin. An insertion event of IS256 at different sites within the ica operon (predominantly in the icaC gene) was evidenced for about 30% of variants (Ziebuhr et al. 1999). Reversion, occurring at a low frequency (below 10−8), is generally accompanied by the precise excision of IS256 and results in the formation of an episome (Ziebuhr et al. 1999; Loessner et al. 2002). Variants with reduced ica expression were associated with IS256 insertions in rsbU, a positive regulator of the stress response regulator σB, and in sarA encoding a staphylococcal accessory regulator (Conlon et al. 2004). Some strains devoid of IS256 can generate variants although at a lower frequency, suggesting that other mechanisms could be involved in the appearance of variants in S. epidermidis (Conlon et al. 2004). Variants could detach from the biofilm and disseminate into novel habitats. Reversion to the wild-type phenotype could then allow S. epidermidis to form biofilms into new environments (Ziebuhr et al. 1999).

In Staphylococcus aureus, variants unable to form biofilms also display IS256 insertions into the icaC and sarA loci (Kiem et al. 2004). The absence of the σB transcription factor dramatically increases the rate of switching to the biofilm-negative phenotype. IS256-mediated biofilm switching is reversible, as revertants could emerge from biofilm-negative σB mutants (Valle et al. 2007).

Duplication

Virulence in Pseudomonas tolaasii

Pseudomonas tolaasii, the causal agent of brown blotch disease of the mushroom Agaricus bisporus, degrades fungi tissues by the action of an extracellular toxin, tolaasin, displaying biosurfactant and ion channel-forming properties (Grewal et al. 1995). Aged colonies of P. tolaasii often display sectors that when isolated, form distinct colonies; wild-type cells (designated 1116S) show domed, opaque, nonfluorescent colonies and are pathogen whereas variant cells (designated 1116R) present flat, translucent, and fluorescent colonies but are no longer pathogen. Moreover, variants display a stronger chimiotactic response and enhanced motility, and reversion could be observed (Grewal et al. 1995).

The pheN regulatory gene, whose deduced product displays homology to both the sensor and regulator domains of the conserved family of two component bacterial sensor regulator proteins, undergoes a 661-bp duplication during phenotypic variation (Han et al. 1997). PheN both acts as an activator for expression of tolaasin and opacity proteins and as a repressor of chemotaxis and of the synthesis of a fluorescent pigment. The duplication event introduces a frameshift mutation in the predicted pheN ORF; this results in the formation of two new non-functional ORFs: a truncated ORF containing only the sensor domain and a second ORF lacking 204 amino acids of the N-terminus of PheN and hence the sensor domain (Han et al. 1997). A recA strain generates 3-fold less variants than the wild-type strain; the few variants obtained from a recA strain show no duplication within pheN but would arise from a punctual mutation in pheN (Sinha et al. 2000). Reversion to the wild-type occurs via a precise deletion of the 661-pb region and is RecA-independent as similar frequencies of revertants are observed for 1116R and 1116RrecA (Sinha et al. 2000).

Based on differential phenotypic features of 1116S and 1116R, it was proposed that these two forms are adapted to different environmental niches. The wild-type (1116S) would penetrate and proliferate within fungi tissues, inducing their degradation. The variant type (1116R), with increased motility and chemotaxis but no synthesis of virulence factors, would be more adapted to telluric life; by reverting to the 1116S form, the variant would become virulent again and could infest a new host (Grewal et al. 1995). However, the phenotypic switch from the 1116S from to the 1116R form has been demonstrated only in vitro.

Capsule synthesis in Streptococcus pneumoniae

Streptococcus pneumoniae is a human pathogen, responsible of otitis, pneumonia and meningitis. Small colonies lacking a capsule can be isolated from a biofilm of S. pneumoniae serotype 3 at a frequency of 0.2%. Duplications (from 11 pb up to 239 pb) in the first gene of the capsule biosynthesis pathway (cap3A) were evidenced (Waite et al. 2001). Reversion to capsulated cells occurs via the precise excision of the duplication; the frequency of reversion depends on the length of the duplication as for duplications of 10 pb and 100 pb, excision occurs at a frequency of respectively 10−5 and 10−3 (Waite et al. 2001). Duplications in other genes of the capsule biosynthesis were evidenced for other serotypes of S. pneumoniae (Waite et al. 2003), but the implication of proteins of the homologous recombination pathway, such as RecA, has not been studied.

The ability to regulate capsule biosynthesis is advantageous when invading eukaryotic cells; indeed, adherence and invasion would be 200-fold less efficient for a capsulated strain than for a non-capsulated strain (Ring and Tuomanen 2000). But after cellular invasion, the capsule prevents S. pneumoniae from being eliminated by phagocytosis.

Deletion

Motility and sugar assimilation in Azospirillum

Azospirillum is a plant growth-promoting rhizobacterium associated with roots of monocots, such as wheat, corn, and rice. Azospirillum lipoferum 4B generates in vitro a stable variant, 4VI, at frequencies of 10−4 to 10−3 per cell per generation. Pleiotropic modifications are observed such as the loss of swimming motility, the incapacity to bind some dyes and the inability to assimilate certain sugars (Alexandre and Bally 1999). Interestingly, the frequency of variants generated by a recA mutant was increased up to 10-fold, a result contrasting with many studies that showed the abolition or a large reduction of the frequency of phase variation in recA mutants (Vial et al. 2004). Recently, the appearance of variants was correlated with the loss of a 750-kb replicon (Vial et al. 2006). A. lipoferum 4T, a non-swimming strain displaying all the features of the 4VI variant and retaining the ability to efficiently colonize rice roots, was isolated from the same rice rhizosphere than A. lipoferum 4B, suggesting that A. lipoferum 4T could in fact be a variant of strain 4B generated within the soil (Alexandre et al. 1996); this non-swimming A. lipoferum 4T strain also lacks the 750-kb replicon (Vial et al. 2006). Loss of swimming ability in variant 4VI and in strain 4T is directly linked to enhanced swarming motility (Alexandre et al. 1999). Thus, the non-swimming Azospirillum strains are expected to keep the ability to move along plant roots but the adaptive significance of phase variation in Azospirillum remains to be established.

Strains of two other Azospirillum species, i.e. A. brasilense and A. irakense, were shown to generate variants, displaying the same phenotypic features as 4VI variant but various stabilities; large-scale genomic rearrangements during phase variation were demonstrated for two additional strains (Vial et al. 2006).

Capsule synthesis in Haemophilus influenzae

The type b capsule of the human pathogen H. influenzae is a key virulence factor. Loss of capsule at a frequency of 0.1–0.3% was evidenced in vitro and in vivo (Hoiseth et al. 1985). Most of the H. influenzae strains possesses an 18-kb tandem duplication of genes involved in type b capsule expression, flanking the bexA gene essential for export of capsular polysaccharide. During phase variation, one of these two copies is lost and bexA is disrupted, impeding the capsule synthesis (Hoiseth et al. 1986; Kroll et al. 1988). The high-frequency loss of type b capsule expression is probably due to rec-dependent recombination between the two copies of the 18-kb tandem repeat, as no deletion occurs in a recA mutated strain (Hoiseth et al. 1986).The presence of a capsule would be deleterious for an efficient colonization of epithelial cells. As a duplicated region is more prone to undergo recombination events, the 18-kb tandem repeat could be easily lost when H. influenzae no longer needs its capsule.

Capsule synthesis in Vibrio cholerae

Vibrio cholerae strain O139, responsible of a cholera outbreak in India and Bangladesh in 1992, expresses various virulence factors such as cholera toxin and capsule, the latter conferring opacity to the colonies. Opaque colonies, after a transfer into liquid medium and a new plating, can generate translucent colonies at a frequency of 3.10−3 (Smirnova et al. 1996). The new features of variants, i.e. loss of capsule and purine auxotrophy, are due to the induction of a prophage localized near genes involved in capsule (cap) and in purine (pur) biosynthesis. The imperfect excision of this prophage would provoke a deletion affecting the cap and pur genes, and might also affect a gene implicated in regulation of virulence (Smirnova et al. 1996). Variants appear sensitive to serum bactericidal activity and display decreased expression of virulence factors; as no reversion event was evidenced either in vitro or in vivo, the ecological significance of this phenomenon remains unclear.

Virulence in Yersinia pestis

Yersinia pestis, the causal agent of bubonic and pneumonic plague, displays pigmented colonies when grown in agar medium containing haemin or Congo red (Pgm + phenotype). Spontaneous non-pigmented avirulent mutants arise at a frequency of 10−5, with no reversion. Loss of the Pgm + phenotype is accompanied by the spontaneous deletion of a 102-kb region flanked by a repetitive element (Fetherston et al. 1992); this 102-kb region is composed of an iron acquisition segment linked to a pigmentation segment (Buchrieser et al. 1998). Smaller deletions within this island as well as mutation could also cause the non-pigmented phenotype (Buchrieser et al. 1998).

Slipped-strand mispairing

Phase variation may be accomplished via frequent and reversible changes in the lengths of short DNA sequence repeats, generally composed of stretches of polypurines and/or polypyrimidines. The gain or loss of repeat units involves a mechanism of slipped-strand mispairing (SMM), a RecA-dependent process occurring during chromosomal replication, DNA repair and recombination processes that require DNA synthesis. When occurring in the coding sequence of the gene, changes in the number of DNA repeats result in translational frameshift mutations, thereby switching the expression of the encoded protein ON or OFF; gain or loss of repeat units can also affect transcription initiation by modifying the relative positioning of the RNA-polymerase-binding sites within the promoter or transcription termination.

Opacity proteins in Neisseria

The opacity proteins (Opa) of N. gonorrhoeae and N. meningitidis, which mediate bacterial adhesion and invasion of host tissues, undergo both antigenic variation through gene conversion and phase variation through SSM (Stern et al. 1986; Stern and Meyer 1987; Murphy et al. 1989). Neisseria gonorrhoeae can host 11–12 opa loci, while N. meningitidis contains 3–4 loci; bacterial cells either display the Opa state, a single Opa protein or express multiple Opa proteins simultaneously. The region encoding the signal sequence of Opa contains a repetitive pentamer sequence (5′-CTCTT-3′), which is subjected to slipped-strand mispairing. Depending on the number of repeats that are present, the translational reading frame of the opa gene may be shifted. Hence, each opa gene has several ON configurations (i.e. 6, 9, 12 repeats), in which the ATG initiation codon is in frame with the remaining coding sequence; in other configurations (i.e. 4, 8 repeats), the initiation codon is out of frame with the rest of the coding sequence, leading to aberrant or truncated proteins. By altering the expression of individual Opa proteins by SSM and by generating new assortments of Opa proteins by gene conversion, a great variety of combinations with diverse host-receptor specificities can be generated, allowing bacterial populations to adapt to new niches within the same or different hosts.

The expression of other surface components (capsule, outer membrane proteins) as well as other phenotypes can be regulated by SSM in Neisseria (Table 1). Moreover, systematic searches of DNA repeats in complete genomes revealed that up to 65 candidate genes can be found in a strain of N. meningitidis, showing its high potential for SSM-mediated phase variation (Saunders et al. 2000).

Pili (LKP fimbriae) in Haemophilus influenza

One of the virulene factors implicated in the colonization and pathogenesis of H. influenza are long thick pili (LKP fimbriae). The expression of H. influenzae pili is subjected to reversible phase variation and several observations suggest that expression of pili is beneficial during the early stages of infection but might be disadvantageous in establishing systemic disease. Phase variation is controlled by SSM at the transcriptional level of two divergently orientated genes, hifA and hifB, encoding respectively the major fimbrial subunit and the fimbrial chaperone. The hifA and hifB overlapping promoter regions contain repetitive TA units; variation in the number of units changes the normally strictly constrained spacing between the −35 and −10 sequences and results in bidirectional control of transcription initiation (van Ham et al. 1993).

Motility in Pseudomonas putida

Frameshift mutations in a poly(G) track at the 5′ region of the flhB flagellar gene of Pseudomonas putida DOT-T1E are responsible for motility switch. The fhlB sequence obtained from cells grown on soft-agar plates contain a string of 8 or 11 G’s, allowing the synthesis of a functional protein of respectively 380 and 381 amino-acids; when cells are grown in liquid or solid medium, the fhlB sequences contain 7, 9 or 10 G’s, leading to truncated proteins (Segura et al. 2004). These frameshift mutations would allow the cells to save energy under conditions where motility is not necessary. This strain is a toluene-resistant isolate; mutations in genes encoding the flagellar export apparatus, such as fhlB, lead to hypersensitivity to toluene shocks and growth conditions favoring a functional fhlB gene (such as growth on soft-agar) result in increased innate tolerance to a sudden toluene shock. So, the presence of intact flagellar machinery is critical for innate solvent tolerance but whether FhlB is directly responsible for solvent tolerance remains to be demonstrated.

Multiple events affecting a single gene

Adhesins in Mycoplasma gallisepticum/mycoides

Mycoplasma gallisepticum, an avian pathogen causing respiratory infections in chicken and turkey, can generate variants in vitro at a frequency ranging from 2 × 10−4 to 5 × 10−2 per cell per generation. Those variants have lost the ability to adhere to red blood cells, and no longer express two proteins displaying homology with cytadhesins, GapA and CrmA. Molecular analysis of several variants reveals an amber mutation at the beginning of the gapA gene, that also affects expression of the downstream gene, crmA. Revertants recovering an intact gapA gene can be observed (Winner et al. 2003). In Mycoplasma mycoides subsp. mycoides, a similar mechanism by an amber mutation was also reported for ptsG (a gene encoding a putative glucose PTS permease) (Gaurivaud et al. 2004). Although the occurrence of this phenomenon has not yet been described in vivo, surface protein variation in mycoplasma could be a strategy for host adaptation during the infection.

gacA and gacS mutations in Pseudomonas spp.

From a collection of Pseudomonas strains isolated from the maize rhizophere and displaying antagonist activity towards fungal phytopathogens, the appearance of variants was evidenced in vitro for 43 strains with a frequency ranging from 1.5 × 10−4 to 9.0 × 10−2 depending on the strain (van den Broek et al. 2003). Whereas wild-type colonies are opaque, variants colonies are translucent, do no longer display biocontrol activity, and are unable to produce exoenzymes (lipase, protease, etc.).

In one of these strains, Pseudomonas sp. PCL1171, the appearance of variants at a frequency of 6.4 × 10−5 is correlated with point mutation, insertion or deletion events within gacA and gacS, genes encoding a two-component regulatory system involved in the production of secondary metabolites and exoenzymes in Pseudomonas spp. (van den Broek et al. 2003; van den Broek et al. 2005c). Reversion due to recovery of a functional gacA or gacS gene is observed at a high frequency in vitro; mutants locked in the variant state can be obtained by transposon mutagenesis within gacS and complementation with a functional gacS allows reversion to the wild-type phenotype (van den Broek et al. 2005c). Moreover, for a mutS mutant of strain Pseudomonas spp. PCL1171, the frequency of variants appearance is increased by 1,000-fold, probably due to accumulation of mutations within gacA and gacS (van den Broek et al. 2005b). A strain overexpressing RpoS, the sigma factor involved in stress response, produces ten times more variants than the wild-type strain. RpoS is thought to directly or indirectly repress mutS transcription; thus when cells reach stationary phase or when exposed to stress (i.e. when RpoS is expressed), mutations would accumulate within gacA and gacS (van den Broek et al. 2005b). A gacS mutated strain displays a shortened lag phase compared to a wild-type strain. Then, variants would be more competitive and could adapt more easily to the heterogenous and challenging rhizophere ecosystem (van den Broek et al. 2005c).

A common feature between those variants of Pseudomonas spp. and variants of P. fluorescens F113 mentioned above (in the session “Site-specific inversion”) is the presence of mutations within gacA or gacS; it was thus suggested that gacA or gacS could be the targets of the site-specific recombinases (Sss and XerD), a hypothesis that remains to be demonstrated (Martinez-Granero et al. 2005).

phcA Mutations in Ralstonia solanacearum

Ralstonia solanacearum, the causal agent of bacterial wilting of numerous plants, enters the plant vascular system through natural lesions. Extracellular polysaccharides produced by R. solanacearum block xylem vessels, impeding transfer of water and minerals towards the aerial parts of the plant. The expression of numerous virulence factors, including EPS, plant cell-wall degrading enzymes, type-three secretion system, is under the control of the global regulator PhcA. A phenotypic conversion from a mucoid to a non-mucoid state was evidenced in vitro and in planta; these non-mucoid EPS variants display increased motility, reduced endoglucanase activity and reduced virulence but keep the ability to develop in planta. Mutations (deletions, duplications, point mutations or IS insertions) within phcA are correlated with the appearance of variants from different strains (Brumbley and Denny 1990; Poussier et al. 2003). Reversion to the wild-type phenotype, with recovery of a functional phcA, can be observed only in planta for variants originating from a 64-pb duplication or an IS insertion event (Poussier et al. 2003).

The avirulent variant would be more adapted to survive in soil, on plant debris or in starving conditions, and the increased motility could allow the colonisation of new niches. When encountering a new host plant, the variant would revert to a virulent state in order to effectively infect a new plant (Poussier et al. 2003).

Conclusions

Historically, phase and antigenic variation, principally investigated in bacterial pathogens of human and animals, was considered to help the bacterium to evade the host immune system. Indeed, many structures that are found to be affected by phase variation are on the cell surface, exposed to the immune system and to the environment (Table 1). However, variation occurs also for genes that are not related to changes in the cell surface properties. Recent studies demonstrated the occurrence of phase variation among plant-associated bacteria whether pathogenic or beneficial, where it affects various phenotypes such as virulence traits, biocontrol traits, root colonization or the ability to form biofilms. Even major metabolic capacities like in A. ferrooxidans can be controlled by phase variation, showing the impact of this phenomenon on the ecology of the bacterial species. It is also clear from this survey that phenotypic switch is a widespread strategy used by various bacterial genera belonging to different bacterial classes (Tables 1 and 2).

Table 2 Examples of phase and antigenic variations with uncharacterized mechanisms

Phenotypic switch usually affects the synthesis of one family of proteins but several phenotypes can be regulated by this process, like in B. fragilis or A. lipoferum. For some variants, phase variation could be a strategy to reduce the metabolic load, variant cells avoiding to synthesize factors that are dispensable under certain conditions. Thus, phase variation can be considered as a global strategy used by bacteria to survive environmental conditions or to colonize new environments by creating a heterogeneous population.

Phase variation implying genome modifications can be correlated to various mechanisms, some of which being specific to a bacterial species, such as insertion–excision of a plasmid in L. pneumophila (Lüneberg et al. 2001). In the case of point mutations, these can be considered as mutational mechanism occurring at specific loci (often encoding major regulators) at unusually high frequency. The switch to the parental phenotype is not always documented, and whether this is due to the incapacity to revert or to unsuitable conditions for reversion is not elucidated.

Many evidences of phenotypic switch came only from in vitro experiments, questioning the occurrence of this phenomenon under natural conditions and explaining why its functional significance is largely speculative. If epigenetic phase variation is often regulated by environmental factors such as temperature and medium composition, this aspect is less documented for genome modification-mediated phase variation. The putative role of host signals is also scarcely reported (Poussier et al. 2003; Martinez-Granero et al. 2005).

Finally, there are numerous examples of phenotypic switch for which the mechanism remains to be elucidated (Table 2). Indeed, unraveling the phenotypic diversification from microcolony-type biofilms of the emerging opportunistic pathogen Serratia marcescens (Koh et al. 2007), the hyperadhesion and enhanced ability to form biofilms of the opportunistic pathogen Pseudomonas aeruginosa (Déziel et al. 2001; Webb et al. 2004; Kirisits et al. 2005), and the phenotypic switch affecting response to environmental stress in the causative agent of melioidosis, Burkholderia pseudomallei (Chantratita et al. 2007) would also be important medical issues. Of ecological relevance for plant-bacteria interactions is also the higher ability to swim and swarm of Pseudomonas brassicacearum variants (Chabeaud et al. 2001; Achouak et al. 2004) and the increased chemotaxis and deficiency in xanthan gum production of the plant pathogen Xanthomonas campestris pv campestris (Kamoun and Kado 1990).