Phase variation in bacteria is often considered a random process that has evolved to facilitate immune evasion in a host. Here, alternative biological roles for this process are presented and discussed, incorporating recent studies on nonpathogenic and commensal bacterial species. Furthermore, the integration of phase variation into bacterial regulatory networks and the relevance of this for considering phase variation as a random process are reviewed. Novel approaches are needed to study phase variation and its biological roles, but the insights obtained can contribute significantly to our understanding of the dynamic behaviour of bacterial populations and their interactions with the environment.
A single bacterial species may be challenged to survive in environments as diverse as the soil and the gastrointestinal (GI) tract of a host. To grow and thrive the bacteria have evolved complex regulatory networks that allow them to adapt by changing their phenotype in response to signals from a changing environment. However, two processes that result in a phenotypic change in clonal population are traditionally not considered to be a responsive process. These processes are phase variation and antigenic variation.
Phase variation is a process that results in differential expression of one or more genes and results in two subpopulations within a clonal population: one lacking or having a decreased level of expression of the phase variable gene(s) and the other subpopulation expressing the gene fully. Abrogating expression can alter the antigenic properties of the bacterium. This process is distinct from antigenic variation however in which a bacterial structure, proteinaceous or otherwise, is consistently produced, but in different antigenic forms. In specific cases, phase variation can lead to antigenic variation, for example if phase variation affects expression of a lipopolysaccharide (LPS) modifying enzyme. A key feature of phase variation is that the ‘On’ and ‘Off’ phenotypes are interchangeable. Thus, a cell with gene expression in the ‘Off’ phase, that is lacking expression, retains its ability to switch to ‘On’ and vice versa. The frequency with which switching occurs can vary widely in different systems, ranging from as frequently as one cell in ten per generation to as infrequent as one in ten thousand. The occurrence of phase variation thus results in a heterogenic and dynamically changing phenotype of a bacterial population. A change in the ratio of the two subpopulations could in principle be achieved either through selection for or against a subpopulation, or through regulation of gene expression or of the switch frequency (Fig. 1).
Recent studies indicate that two dogmas of phase variation should be re-examined if we are to understand the dynamic behaviour of bacterial populations. A frequently cited hypothesis for the biological role of phase variation is that it allows the bacterium to evade the host's immune system. This minireview examines the need for additional or alternative hypotheses for the biological significance of phase variation, emphasizing recent developments. Second, the random nature of phase variation is examined in the context of regulatory networks. The event that results in a cell switching from an expressing to nonexpressing state and vice versa is random in the sense that no prediction can be made about which cell in a population will undergo the switch. However, a large body of work now shows that environmental signals and intercellular regulatory networks can be integrated with or superimposed on some of the phase variation mechanisms. These are briefly summarized, along with whether environmental regulation affects the way we should view the random nature of phase variation is discussed.
This review will show that phase variation should be considered as an integral part of a wide variety of bacterial lifestyles. Therefore, it will become apparent that examining how and when phase variation occurs can provide valuable insight into the population dynamics of bacteria and the strategies they use to adapt to life in a diverse set of challenging environments.
Phase variation may be more than a means to avoid the immune system
In medical microbiology, and specifically in the field of host–pathogen interactions, phase variation is considered an important bacterial tool to assist in evasion of innate and acquired immune mechanisms during colonization or infection of an animal host. The concept is that the subpopulations that are generated will present different sets of surface antigens, and the immune system will select against the subpopulation(s) that it recognizes (Fig. 1b). Thus, subpopulation A can evade the immune response generated against subpopulation B. This hypothesis is supported for antigenic variation through the strong correlation between the emergence of new antigenic phenotypes and the emergence of adaptive immune response supports (reviewed in Deitsch , 1997). However, for phase variation, the supporting data and this argument are less convincing: concurrent with switching to an ‘Off’ phase of the immmunogenic antigen that allows the bacterium to ‘hide’ from the immune system, the bacterium looses the biological function of this antigen. Thus, functional redundancy is required to cope with the loss of function as a result of phase variable expression (van der Woude & Baumler, 2004). How often this occurs is not clear. Alternatively, the loss of function can have a beneficial effect for the bacterium, as is illustrated in a few examples in this review. It is important to realize that only a few studies have directly addressed the role of phase variation in immune evasion or otherwise.
In at least 14 genera, each containing species associated with animal infections or colonization, phase variable expression of one or more genes has been identified and the mechanism characterized (van der Woude & Baumler, 2004). Considering this wide distribution, it is difficult to argue against an apparent importance of phase and antigenic variation for host–bacterium interactions. The role of phase variation in relation to the adaptive immune response was recently reviewed, and the reader is referred to the review and references therein for a detailed discussion of the experiments (van der Woude & Baumler, 2004). The challenge remains to understand how a finite number of variations of expression of only a subset of surface structures can suffice to avoid the immune system and facilitate chronic infection and mucosal colonization. A recent study suggests that this may be achieved if phase variation modulates the immune response. Helicobacter pylori can modify its LPS, including with one or both Lewis antigens, Lex and Ley, and expression of the responsible genes futA and futB phase varies. Expression of monomeric Lex and Ley facilitates H. pylori binding to DC-SIGN on dendritic cells. This in turn affects the activation cascade of T cells, specifically blocking the T helper cell (Th1) response (Bergman , 2004). As immune responses are diverse in a human population, it is proposed that a heterogenic bacterial population as a result of phase variation, in combination with host-mediated selection for specific phenotypes, results in a Th1/Th2 response that can be optimized to allow bacterial colonization of different individuals (Bergman , 2004). In contrast, in a mouse model of infection, phase variation of the Vsa surface antigen in Mycoplasma pulmonis was found to occur less frequently in Rag−/− mice than in wild-type or iNOS−/− mice, suggesting an immune-mediated selection drives appearance of phase variants in this system (Denison , 2005).
Infection by different Salmonella enterica serovars leads to serotype-specific immunity that is based on serovar-specific expression of O-antigens in the LPS. Serovars also share antigens like fimbriae, yet coexistence of serovars occurs. In this case, fimbrial phase variation was shown to be required to facilitate coexistence of multiple serovars in one host (Norris & Baumler, 1999; Nicholson & Baumler, 2001). Thus, phase variation can yield bacterial population benefits that extend beyond the clonal population.
If frequency of occurrence implies the importance of phase variation for host–bacterium interactions, then the hypothesis is strongly supported by the recent genome analysis of Bacteroides fragilis. This is a commensal of the human GI tract as well as opportunistic pathogen (Kuwahara , 2004; Cerdeno-Tarraga , 2005), and can be isolated from polluted waters (Allsop & Stickler, 1985). The genome analysis identified over 30 enzymes that may cause site-specific DNA inversion (Cerdeno-Tarraga , 2005) and 31 invertible DNA regions (Kuwahara , 2004), which are elements that facilitate phase and antigenic variation in other species (van der Woude & Baumler, 2004). In Bacteroides fragilis they appear to be involved in ON/OFF switching and in generating hybrid proteins, shuffling domains and rearranging operon structures (Kuwahara , 2004; Cerdeno-Tarraga , 2005). Many of these genes encode for probable surface antigens, which could support the hypothesis for a role in immune evasion. An alternate hypothesis which does not involve the immune system is that the variable surface structures may facilitate colonization of different sites in the host (Cerdeno-Tarraga , 2005). Indeed, differential expression of phase variable MR/P fimbriae in Proteus mirabilis contributes to specific organ colonization (Li , 2002).
Some of the putative phase variable genes in B. fragilis have no apparent means to facilitate immune evasion or colonization. Neither do numerous phase variable genes in other pathogens (Saunders , 2000), and phase variation is now also described in nonpathogenic species and in species that do not appear to be host associated. Thus, there are compelling reasons to examine the existence of additional roles for phase variation. Some hypotheses are discussed and are mostly based on the knowledge of the function of these phase variable genes. Interestingly, most of these hypotheses have not been tested directly.
An encompassing view on the role of phase variation is that the generation of diverse subpopulations enhances the chance that at least one can overcome a stressful challenge, in essence a ‘hedge-betting’ strategy. In Escherichia coli, an unusual form of variation may indeed facilitate survival in the face of specifically oxidative or disulfide-mediated stress. Variation in aphC leads to interconversion between two forms of the AphC enzyme, specifically between a peroxidase and an apparent glutathione-glutaredoxin reductase. Both forms appear to occur in the absence of stress, and interconversion occurs at frequencies similar to normal phase or antigenic variation rates (Ritz , 2001). Whether a similar role in stress survival can be attributed to some genes under the control of a classic ON/OFF phase variation is less apparent.
It is not difficult to imagine that AphC interconversion may contribute to survival under different stress conditions, but many phase variable genes have no obvious role in stress survival. This is true for example for phase variable expression of DNA restriction/modification (R/M) systems, which occurs in numerous species (Dybvig , 1998; De Bolle , 2000; De Vries , 2002). One biological reason that has been proposed is that the OFF phase of R/M systems facilitates foreign DNA uptake and thus contributes to genetic diversity. Conversely, the ON phase may prevent uptake and integration of foreign DNA that could be detrimental to the isolate. Phase variable expression may be required to balance these two advantages among individuals in the population, presumably leading to an optimal population phenotype as a result of environmental selective pressures and selective growth. However, this may not apply to variable expression of an R/M system in M. pulmonis, since this is not a naturally competent species (Dybvig , 1998). R/M should also protect against phage infection, which could ensure survival of the individual cell if lysis occurs as a result, or unwanted genetic elements are introduced. However, lysogenic phage infection could benefit the population if the phage encodes a virulence factor that increases fitness (Brussow , 2004), or lytic infection could be beneficial if the lysed subpopulation benefits the survivors by providing nutrients. The latter has been described for bacterial populations in the stationary growth phase (Finkel & Kolter, 2001). Phase variable expression could potentially be a way to balance these negative and positive effects within a clonal population.
DNA modification, specifically DNA methylation, can however also directly control gene expression. Specifically, in E. coli the constitutively expressed maintenance methylase Dam controls phase variation of the pap-family of fimbrial operons and the outer membrane protein Ag43 (Henderson , 1997; Wallecha , 2002; Hernday , 2004a). Srikhanta . (2005) examined whether a DNA modification enzyme of an R/M system can also function to control gene expression, hypothesizing that phase variable R/M expression may lead to co-ordinated phase variable expression of multiple genes. This was tested with expression of the mod gene in Haemophilus influenzae that is controlled by phase variation. Gene expression in a population expressing mod was compared with one containing an inactivated mod gene, and differential expression of seven genes was identified. As yet, there is no molecular mechanism for this correlation, but it could be similar to the principle underlying Dam-dependent phase variation (Hernday , 2004a). However, the identification of stress-associated genes in what the authors have termed the ‘phasevarion’ could also indicate that mod expression is associated with a general stress response (Srikhanta , 2005). Regardless of mechanism, if this intriguing hypothesis of a regulon controlled by the phase variable expression of a DNA modification gene holds true, it will be of interest to determine the significance and the occurrence among the different species.
Very few studies have focused on phase variation in species that are not associated with an animal host, but as illustrated by a recent study on the soil bacterium Bacillus subtilis, this can suggest an alternative role for phase variation. In this case, the indicated role may be universally applicable: facilitating a switch between a planktonic and biofilm lifestyle. Wild-type strains of B. subtilis have the capacity to swarm, but laboratory strains show no swarming behaviour (Kearns , 2004). The molecular basis was examined and could be attributed to a change in the nucleotide repeat numbers in the coding region of swrA, which is essential for swarming, that renders the gene product nonfunctional. The frequency of the reversible nucleotide insertion and excision suggests that phase variation of the swarming phenotype occurs (Kearns , 2004). Kearns et al. suggest that a swarming phenotype may be more advantageous during growth on a solid surface than in liquid, and thus the environment would dictate which subpopulation thrives. However, a regulatory mechanism that senses a solid surface is superimposed on the phase variation of swrA (Calvio , 2005), and thus it may not be that simple. Interestingly, in this species the expression of a gene involved in biofilm formation may also be phase variable (Kearns , 2005). Based on these initial studies, phase variation in B. subtilis may be considered a random event directing a planktonic or surface-associated lifestyle, but one that is integrated in a regulatory network.
Many other phase variable surface antigens are involved in attachment to abiotic or biotic surfaces. In Escherichia coli fimbriae and motility are essential, and the outer membrane protein Ag43 has been implicated in biofilm formation (Pratt & Kolter, 1998; Danese , 2000; van der Woude & Baumler, 2004). Ag43 and most fimbriae phase vary, and expressions of the latter may be linked to each other (Schembri & Klemm, 2001; Wallecha , 2003; Holden & Gally, 2004), suggesting phase variation may affect biofilm formation. In Neisseria meningitidis expression of the capsule is phase variable (Hammerschmidt , 1996) and biofilm formation is inhibited by the capsule (Yi , 2004), again suggesting a role for phase variation in biofilm formation and dispersal. This phase variation may however have multiple roles, as in this species, the capsule is essential for serum resistance (Vogel , 1997), but inhibits invasion (Hammerschmidt , 1996).
Most of the data interpretations are driven by the assumption that one subpopulation will be advantageous and dominate in a given environment. However, it is also possible that each subpopulation plays a distinct role and that together they increase the fitness of the population, which should then lead to a (stable) mixed phenotype in specific environments. This could be a particularly interesting hypothesis in the latter examples of this section. Experiments that directly address the role of phase variation in swarming, biofilm formation and dispersal in these and other species could provide new insights into the dynamic behaviour of bacterial populations.
Regulatory networks and random phase variation
The term ‘contingency genes’ is often adopted to describe the class of genes that are expressed in a phase variable manner. ‘Contingency’ is defined as ‘a future event, which is possible but can not be predicted with certainty’ by the Oxford dictionary, and thus incorporates an essential part of phase variation: the inability to predict which cell in a population will undergo an alteration in gene expression. Phase variation is also referred to as a ‘random’ event, which reflects this stochastic occurrence. The apparent need for a stochastic event implies that the benefit cannot be obtained by mediating change through environmental signalling and regulation. This could be the case for example if the time required to incorporate a regulatory signal would be detrimental to the chance of success of the bacterium, or if there is a lack of a suitable environmental signal. Regardless, owing to its random nature, phase variation has long been considered to be insensitive to environmental conditions. However, it is now clear that regulatory controls can be superimposed on and integrated with some of the phase variation mechanisms. Integration can result in a change in the switching frequency, in some cases to a degree where phase variation is abrogated. All cells are in the ‘Off’ phase and thus expression of the gene does not occur in the entire population. Superimposing a regulatory control that responds to environmental signals on the phase variation mechanism can also result in a phenotype lacking expression of the phase variable gene, but in this case the cell may still be in the ‘On’ phase (Fig. 1). This phenotype may resemble the ‘Off’ phase, but the expression state can be rapidly reversed in each cell to the ‘On’ phenotype by environmental signals. As illustrated below, control over the random occurrence of phase variation can take the form of intracellular signalling among operons, as well as external stimuli that can be incorporated through action of global regulators. The individual mechanisms are not elaborated upon in detail here, and the reader is referred to a number of reviews and the references therein (Henderson , 1999; Blomfield, 2002; Hernday , 2004a; van der Woude & Baumler, 2004).
Phase variation events of individual fimbrial operons in E. coli are linked, forming an interdependent, co-ordinated network of phase variable gene expression within one cell (Morschhäuser , 1994; van der Woude & Low, 1994) (reviewed in Holden & Gally, 2004) (Fig. 2). Co-ordinated expression among the pap-regulatory family of fimbrial operons like daa and sfa can occur as a result of crosstalk between homologues of the regulatory proteins PapB and PapI that are essential for expression and phase variation of these operons (Morschhäuser , 1994; van der Woude & Low, 1994). In addition, PapB and the PapB paralogue SfaB repress recombination-mediated phase variation of the type 1 encoding fim operon (Xia , 2000; Holden , 2001), but the paralogue DaaA does not (Holden , 2001). Thus, one stochastic event can set in motion a series of events that influence the frequency of occurrence of other stochastic events (Fig. 2a). The network can also include genes that are not directly controlled by a phase variation mechanism. MrpJ is encoded by the phase variable mrp fimbrial operon in P. mirabilis, but represses flagellar expression (Li , 2001). Thus, the phenotype of flagellar expression can be phase variable as a result of phase variation of the fimbrial operon.
Based on our current knowledge of the phase variation mechanisms, it appears that the ability to superimpose or integrate environmental signals will vary with the phase variation mechanism. Phase variation of fimbriae in E. coli by an epigenetic mechanism and by site-specific recombination is subject to extensive regulation by environmental conditions (reviewed in van der Woude & Baumler, 2004). The DNA-methylation-dependent phase variation mechanisms that control the pap fimbrial operon are affected by Cpx-dependent regulation, which signals envelope stress (Hernday , 2004b). Other regulatory signals act epistatic to the switching event, including repression by low temperature, which is mediated by H-NS, and carbon source availability, which is signalled by activation of cAMP-CAP (reviewed in Hernday , 2004a). By incorporating this environmental regulation into the signalling network, it becomes apparent that this environmental regulation of pap expression can influence the frequency of phase variation of for example fim, encoding type 1 fimbriae (Figs 2b and c).
Extensive environmental regulation also occurs for expression of fim directly. The key event in phase variation of type I fimbriae is operon site-specific recombination, resulting in an inversion of a DNA sequence element that contains the main promoter for the fimbrial operon (McClain , 1993, reviewed in Blomfield, 2001). Most signals that influence expression, including temperature and growth medium (Gally , 1993; Blomfield, 2001), are known or presumed to affect expression of the operon-specific recombinases or the efficiency of recombination. A recent development is that sialic acid and N-acetyl-glucosaminidase (GlcNAc) have been identified as signals controlling type 1 phase variation (El-Labany , 2003). What makes this finding of special interest is that both sialic acid and GlcNAc are released as part of the host defense that is triggered by type I fimbriae. Thus, this sialic acid-dependent regulation may serve to signal to the bacterium that it is in an environment with an active (innate) immune and inflammatory response. The host response to the bacterium therefore may cause a change in bacterial gene expression that affects the bacterial adhesive potential and its antigenicity (Sohanpal , 2004). This bacterial response to the innate immune system could be part of a strategy that facilitates evasion of the adaptive immune system. Utilization of host signals by bacteria for their own benefit may be a common strategy among pathogens (Hornef , 2002).
In contrast to these systems, phase variation as a result of insertions and deletions at nucleotide repeat tracts does not seem to allow for integration of diverse environmental signals through global regulators. However, a change in the rate of switching in N. meningitidis by this mechanism is found in the presence of heterologous DNA (Alexander , 2004b), and additional loci affecting the phase variation rate have been identified by mutagenesis (Alexander , 2004a). Similarly, mutations in mismatch repair genes affect repeat stability in H. influenzae (Bayliss , 2002, 2004). Thus, these switching rates may vary under conditions that lead to DNA damage, but whether this reflects naturally occurring regulation is not clear.
Interestingly, a bias for specific regulatory mechanisms of phase variation appears to exist in different bacterial species (van der Woude & Baumler, 2004). For example, changes in nucleotide repeat tracts are predominantly identified as the mechanism in H. pylori, N. meningitidis and H. influenzae. These are rare in E. coli, where epigenetic and site-specific inversion-dependent mechanisms that can be regulated are predominant (van der Woude & Baumler, 2004). There is no apparent molecular basis for this specific biased distribution, since nucleotide repeats of different lengths that are introduced into the chromosome are subject to reversible insertion and excision of repeat units in E. coli at relevant frequencies (Torres-Cruz & Van der Woude, 2003). Furthermore, a first example of naturally occurring phase variation by this mechanism was identified recently in E. coli (Deszo , 2005). It is tempting to speculate that this apparent mechanistic bias in E. coli reflects a frequent need to regulate expression of the phase variable genes in the diverse environmental conditions this species may encounter. Additional work on other systems, like those of B. fragilis (Cerdeno-Tarraga , 2005), may determine if there is indeed a relevant correlation between the mechanism of phase variation and the environmental niche(s) occupied by the bacteria.
The mere existence of elaborate regulatory networks suggests that absolute random generation of diversity may not always be beneficial, and it should not be too surprising if more regulatory networks linking the expression of phase variable and nonphase variable genes are identified in the future. This also suggests that at least in some cases a more complex biological rationale for the occurrence of phase variation exists. Identifying signals that affect the expression of phase variable genes and operons and unravelling the regulatory networks can provide valuable insight into bacterial decision-making processes and help us understand the dynamic behaviour of bacterial populations.
This review has highlighted a lack of understanding of the biological significance of phase variable gene expression, indicating a need for continuing efforts in this area if we are to understand the interactions of bacterial populations and their environment. Phase variation is probably a much more frequent and more widely distributed feature than has been documented to date. Studies have mainly focussed on pathogens, and therefore the association with evasion of the immune system has been emphasized. Observations in nonpathogenic species of phase variable gene expression should be awarded the appropriate importance and pursued in greater detail, since it will further our understanding of phase variation and of the importance of heterogeneity in bacterial populations.
However, it must be noted that achieving a full understanding of the role of phase variation may not be easy. For example, observing a phenotypic change in a bacterial population does not identify the driving force. To understand and ultimately to predict these changes, it is important to distinguish between contributions of a passive ‘selection’ force like immune-mediated selection, tissue tropism or growth advantage and an active system of bacterial gene regulation as the driving force (Fig. 1). Thus, studying the regulatory networks will be invaluable. Furthermore, in order to understand the role of phase variation in host–bacteria interactions, novel strategies may be needed to separate effects resulting from the biological function from those affecting immunogenicity. Ultimately, we should strive to discover and characterize relationships between the phase variation mechanism, regulation of this mechanism and the signals involved, the role of the phase variable structure and the biological role of variable expression and how these benefit the bacterium. Finally, it is becoming increasingly likely that there is more than one answer concerning the role of phase variation. This realization in itself should continue to stimulate research in this fascinating area of microbiology.
Work in the author's lab is supported by the BBSRC and the European Union (Marie Curie IRG). I thank Gavin Thomas and James Chong for critical reading of the manuscript.