2.1. Background of Phase Variation

All living organisms are faced with the constant challenge of maintaining fitness in order to survive and reproduce, and this is no less true for bacterial species. Bacteria are under constant onslaught from fluctuations in their local environment, infection from bacteriophages, and (in the case of pathogenic bacteria) attack from their infected host. Although bacteria possess robust mechanisms of classical gene regulation that allow them to respond to extracellular changes (e.g., Bacterial Two-Component Systems), these alone may be unable to cope with the constant barrage of fluctuating pressures they face. These selective pressures are often focused on bacterial external proteins which form the first line of contact with the outside environment and this has led to development of what have been termed “contingency loci” ^[14][15][14,15]. Contingency loci are hypermutable genes that generate genetic and phenotypic variation allowing bacterial populations to survive unpredictable pressures. This hypermutability is conferred by the phenomena of Phase Variation (PV) and antigenic variation.

PV is a reversible gene expression switch that can alter expression between an ON and an OFF state and occurs through several genetic and epigenetic mechanisms [16]. It is characterized by high frequencies, usually exceeding 1 × 10⁻⁵ variants per total number of cells ^[17][18] [17,18] which is orders of magnitude above the typical frequencies of spontaneous mutations (10⁻⁶ to 10⁻⁸ per cell per generation) [18]. Depending on the method of calculation, the frequency of PV may describe not only rate of the PV mechanisms but also the growth of the phase variants themselves. Antigenic variation is related to PV and occurs through similar mechanisms. However, rather than alternating between an ON and OFF state, antigenic variation mechanisms generate variations in the sequence of surface proteins resulting in the expression of different forms and structures of the antigenic proteins on the cell surface ^[17][18][19][17,18,19]. 

As mentioned above, genes subject to PV often encode for cell-surface associated features such as adhesins, liposaccharide synthesis enzymes, and pili ^[20][21][22][20,21,22] but can also encode for virulence factors and secreted proteins such as iron acquisition machinery ^[23][24][23,24]. The collection of phase variable loci in a bacterial species is referred to as the “phasome” ^[16] [16] and generally includes genes which are involved in bottlenecks experienced by the bacterial population. This is most clearly seen in pathogenic bacteria which undergo constant challenge from host immunity during the infection process. For example, PV mediated shutdown of liposaccharide synthesis genes in the invasive pathogen Haemophilus influenzae confers protection against neutrophil-mediated immune clearance but is detrimental in other environments ^[22][25][26][22,25,26]. In another example, PV in Salmonella typhimurium flagellae can modulate their antigenic properties and allow for evasion from host immunity [27].

It is likely that the original role of PV was as a mechanism of innate immunity against bacteria’s greatest enemy: bacteriophages [28]. Although bacteriophages exist in exaggerated abundance relative to their bacterial hosts, their host range is often limited to just a few specific strains of a given bacterial species [29]. Thus, there is a constant cyclical arms race between bacteria and bacteriophages in order to stay one step ahead of each other [30], and PV plays an important role in both sides of this war. An example can once again be found in liposaccharide synthesis genes of H. influenzae in which PV can result in a switch from a sensitive to resistant phenotype against the HP1c1 phage ^[31] [31]. On the other hand, PV in the Escherichia coli phage Mu causes a switch in expression between two sets of tail fibers resulting in modulation of the host specificity ^[32][33] [32,33] with similar phenomena identified in other phages [34].

Considering the above information, it can be inferred that genetic loci susceptible to PV would be found in abundance amongst bacterial species that experience population bottlenecks. Typically, such bottlenecks often occur during the infectious process which imposes limits onto the bacterial population size. These bottlenecks reduce genetic diversity at a time when variation is most beneficial, and PV offers a solution to this hurdle and indeed, several pathogenic bacteria have been documented to undergo PV [18].

While PV is, by definition, a stochastic process, it occurs through several discrete mechanisms. Broadly speaking, mechanisms of generating PV can be discriminated into genetic and epigenetic mechanisms ^[16] [16] both of which will be addressed in Section 2.2 and Section 2.3 respectively.

2.2. Genetic Mechanisms of PV

There are three genetic mechanisms of PV which shall be discussed in the following chapters: Variation in length of DNA Short Sequence Repeats (SSRs) ^[35][36][37][35,36,37], DNA inversion [38], and DNA recombination ^[39][40][39,40].

2.2.1. Variation in Length of DNA Short Sequence Repeats (SSRs)

SSRs are homo- or hetero-nucleotide repeats in DNA that are highly prone to insertion/deletion (indel) errors due to Slipped-Strand Mispairings (SSMs) during DNA replication ^[35][36][37][35,36,37]. SSRs can be as complex as repeating units of tetranucleotides or as simple as a straight homonucleotide run. Indels in SSRs can result in frameshifts that largely have an ON↔OFF effect on protein function or gene expression (by resulting in abrupt termination of translation or inhibition of RNA polymerase binding, respectively Figure 1A) but can have an alternative gradation effect on gene expression as well. For example, alterations in the length of a dinucleotide TA10 tract in the promoter regions of the divergently transcribed hifA and hifB genes controlling fimbriae expression in H. influenzae can either significantly affect hif expression (TA10→TA9) or only moderately affect it (TA10→TA11) [41]. The evolution of the mutability of SSR tracts is largely driven by a combination of environmental and molecular drivers. The environmental drivers include factors such as the aforementioned population bottlenecks arising during infection processes. These bottleneck conditions exert a primary selective pressure for phenotypes that can survive them, e.g., a population that can shut down the expression of a surface protein that is targeted by host immunity. The necessity to survive this recurrent primary selection serves as a secondary layer of selection for plasticity of the gene itself.

Figure 1. Genetic mechanisms of Phase Variation. A cartoon depicting the three main genetic mechanisms of Phase variation (PV). (A) Slipped-Strand Mispairing events within Short Sequence Repeats (SSR) result in expression (green tick mark) of truncated dysfunctional proteins (if SSR is in the CDS) or inhibition (red cross) of transcription by preventing RNA polymerase/transcription factor binding or by other mechanisms. For example, an interesting method of PV-mediated transcriptional control is shown by Danne et al. who demonstrate SSR alterations upstream of the pilA locus of Streptococcus gallolyticus can destabilize a premature transcription-terminating stem loop [61]. (B) Site-specific inversion is carried out by recombinases that recognize inverted repeat regions (Inverted Repeat Left/Right IRL/IRR) and flip the DNA sequence in between them. If a promoter region (e.g., p^B) lies within the sequence flanked by the inverted repeats this leads to shut down of gene expression. (C) RecA-mediated DNA recombination of N. gonorrhoea pilS into pilE results in the formation of new pilE variants. Both pilS and pilE contain variable regions (depicted in green and orange, respectively) interspersed with conserved regions (white) while pilE has a further 5’ conserved region (dark orange) and a promoter to initiate transcription.

The molecular factors are intrinsic to SSR tracts and include the DNA replication and the Mismatch Repair (MMR) [42]. The discriminating factors of SSRs can be broadly delineated into two groups: the composition of the repeating nucleotide unit (i.e., a homonucleotide or a heteronucleotide repeat) and the tract length. These in turn are differentially affected by the DNA replication and MMR machinery. Amongst these proteins are the DNA polymerase enzymes which include the polymerase responsible for the construction of new DNA strands (DNA polymerase III) as well as the polymerase responsible for DNA repair (DNA polymerase I). Studies have shown that these polymerases have an inherent frequency of generating addition/deletion errors when constructing new DNA strands ^[43][42][43,44]. Following DNA replication, any errors are corrected by the MMR machinery which is a suite of Mut proteins that target and fix errors in a strand specific manner. Inactivation of components from either of these suites of proteins results in a hypermutable phenotype and can lead to SSR alteration e.g., [45]. Additionally, the hypermutable phenotype that results from loss of the MMR machinery is directly responsible for genetic variability of bacteria and mutator phenotypes play an important role in bacterial adaptation [46]. For example, both S. aureus and Pseudomonas aeruginosa isolated from the lungs of people suffering from cystic fibrosis are commonly associated with antibiotic resistance caused by hypermutability ^[47][48][49][47,48,49]. Interestingly, while both the MMR machinery and DNA polymerases are involved in SSR evolution, they do not appear to be fully redundant. Several studies have shown that MMR is more responsible for variability of homonucleotide SSRs, especially for those which exceed eight nucleotides in length, whereas DNA polymerase I is exclusively responsible for mutations in heteronucleotide SSRs ^[50][51][52][50,51,52]. This could have evolutionary implications for the mechanisms of generating SSRs. For example, H.influenza is enriched with tetra-nucleotide SSRs ^[51] [51] whose expansion/contraction is affected by DNA polymerase I. Furthermore, evidence suggests that the frequency of DNA polymerase I mediated errors differs between the leading and the lagging strands of newly synthesized DNA, implying that the direction of genes in the chromosome can also dictate the type of SSR that would evolve in them [53]. Lastly, an interesting study carried out by Lin et al. investigated the distribution of SSRs within the genomes of several bacterial species. They found that in many pathogenic species, SSRs were enriched towards the N-termini of protein coding sequences increasing the probability of frameshifts resulting in non-functional proteins ^[54][55][54,55]. This further suggests that bacteria have evolved SSRs in a manner to provide maximal PV.

2.3. Epigenetic Mechanisms of Phase Variation

An epigenetic trait has been defined as a heritable phenotype resulting from modified gene expression that is not due to any alterations in the DNA sequence of the chromosome ^[62][63][62,63]. In prokaryotes, DNA methylation occurs mainly at the nucleotide adenine although studies have shown that cytosine methylation can also occur ^[64][65][66][64,65,66]. DNA methylation usually occurs at specific target sites and is carried out either by methyltransferases that are part of dedicated Restriction–Modification (RM) systems or by orphan methyltransferases. A well-studied methylase responsible for bacterial epigenetic regulation of PV is the DNA Adenine Methyltransferase (DAM) which is an orphan methyltransferase of the gammaproteobacterial family that is specific for GATC sites [64]. Methylation of DNA represses transcription, and thus PV can result if there are GATC sites within a gene promoter which also binds transcription factors, causing mutually exclusive binding competition between the transcription factor(s) and DAM. If there are numerous GATC sites within a promoter region then the mutually exclusive competition can result in differential methylation patterns of the promoter region resulting in switching between an ON and OFF state. A paradigm of this sort of PV is established by a series of intriguing reports studying the pap operon of E. coli and the opvAB operon of Salmonella enterica ^{[21][67][68][69]}[21,67,68,69].

2.4. Combined Mechanisms of Phase Variation

There is growing evidence that shows that many bacterial species undertake a combined approach for PV to maximize the ability to generate rapid and diverse variation. This strategy involves generating PV through genetic mechanisms in genes of RM systems that can modify the transcriptome of the cell via epigenetic control. Such systems are referred to as “phasevarions” as they control phase-variable regulons ^[70] [70] and are immensely powerful weapons in the arsenal of pathogens.

The earliest phasevarions identified are controlled by Type III RM systems. PV occurs in SSRs in the mod gene resulting in ON↔OFF variation and altered methylation states ^[71][72][71,72]. Strikingly, analyses of known Type III system sequences indicate that at least 20% of these systems contain SSRs and could potentially be phasevarions [73]. Furthermore, mod genes are highly conserved, with variation occurring mainly in the DNA recognition domain. This allows mod genes to exist within the species as multiple alleles, each of which controls distinct phasevarions [16].

There is some evidence of a Type II RM regulated phasevarion detected in Campylobacter jejuni, and gene expression patterns were detectably different upon RNAseq analysis, though no direct link to any altered phenotype was reported [74].

PV in Type I systems largely occurs through DNA inversion in the hsdS gene, creating multiple allelic variants of the specificity protein of the Type I system resulting in different gene targets upon PV [75]. An example of a Type I RM phasevarions can be seen in variable capsular expression controlling virulence in Streptococcus pneumonia ^[16][76][16,76].