Genes encoding R-M systems are present on most bacterial and archaeal genomes. Their prevalence indicates how important R-M systems are to prokaryotes. The well-known function role of R-M systems is to protect the host cell from invading foreign DNAs (Murphy et al., 2013). Canonical R-M systems contain enzymes carrying out two activities: a restriction endonuclease, which binds to specific recognition sites and hydrolyzes DNA when the sequence is unmethylated, and a corresponding DNA methyltransferase, which methylates DNA on the same sites recognized by their cognate endonuclease. Some DNA methyltransferases, such as M.EcoGII, could methylate not only residues in DNA but also those in DNA:RNA-hybrid oligonucleotide duplexes (Murray et al., 2018).

Because of their ability to distinguish “self” from “non-self,” the R-M systems are considered to provide a primitive immune system (Vasu and Nagaraja, 2013). Consequently, R-M systems are thought to be essential for host colonization by pathogenic bacteria (Ershova et al., 2015).

R-M systems are classified into four main types based on factors such as the subunit composition of its REase and MTase complexes; ATP and cofactor requirements; recognition site structure; and DNA-cleavage site (Murphy et al., 2013; Ershova et al., 2015). Typical Type I R-M systems consist of three subunits: two modification subunits and one S-subunit. The S-subunit is encoded by hsdS as a specificity subunit, which can specifically bind to recognition sites. The two modification subunits are encoded by hsdM as methyltransferase (M) and by hsdR as restriction endonuclease (R). The MTase in Type I can modify both strands of its substrate DNA. There are two target recognition domains (TRD) in the S-subunit, and each interacts with half of the bipartite recognition site (Murray, 2000; Loenen et al., 2014). Functional restriction activity requires a pentamer comprised of R₂M₂S.

Type II R-M systems are the most prevalent and simplest. They generally function as two individual proteins. The REase cleaves the target DNA at defined positions within or close to its recognition site, and the MTase protects host DNA by methylation. They are also the best-investigated group of R-M systems because Type II REases are key enzymes in genetic engineering. Type III R-M systems comprise two subunits: Restriction (Res) and Modification (Mod) enzymes, encoded by the res and mod genes, respectively. Mod binds to and methylates substrate DNA, while Res functions as a DNA restriction endonuclease. Interestingly, Mod can function independently of Res, whereas Res has no activity without Mod (Dryden et al., 2001). Finally, Type IV R-M systems, distinctive from the other three types, hydrolyze modified DNA. Type IV systems have the methyltransferase and endonuclease activity combined in a single enzyme, which exclusively cleaves modified DNA (Vasu and Nagaraja, 2013).

In addition to their traditionally understood role, the R-M systems have been found to be involved in epigenetic regulation in some bacteria (Table 1). Several studies have reported that methyltransferases in Type III R-M systems have sequence features that are consistent with phase-variable expression (Srikhanta et al., 2005; Fox et al., 2007; Li and Zhang, 2019; Anton and Roberts, 2021). Phase variation is a strategy to generate phenotypic diversity in a bacterial population in the absence of selection. It involves reversible, high-frequency ON/OFF switching of gene expression. Phase variation is often mediated by the presence of highly mutagenic simple tandem DNA repeats, also known as simple sequence repeats (SSRs). The SSRs are often located either within the ORF of genes encoding variant proteins or in their promoter region. Recent research has identified phase variably expressed DNA methyltransferases that act as epigenetic regulators in several pathogenic bacteria (Tan et al., 2016; Lyko, 2018; Han et al., 2022). Many virulence factor genes in bacteria display phase-variable expression, such as pili (Srikhanta et al., 2017), iron-binding proteins (Tauseef et al., 2011), lipopolysaccharide biosynthesis genes (Lüneberg et al., 1998), and outer-membrane proteins (Owen et al., 1996; Green et al., 2019). Phase variation results in genetically and phenotypically diverse populations, which is important in pathogenesis as it provides rapid adaptation to changes brought by the host environment and immune responses (Atack et al., 2018). The DNA methyltransferases that are involved in phase variation themselves are often subject to phase-variable expression (Atack et al., 2015; Tan et al., 2016). It is conceivable that a methyltransferase that is involved in phase variation can cause even more variation possibilities once the methyltransferase itself goes through a phase variation. The numerical possibility of phase variation multiplies by an amplitude when the variable methyltransferase itself is phase varied. The existence of phase-variable methyltransferase raises the possible roles for phase-variable R-M systems in pathogenesis. Using the flagellin A (flaA) gene in Helicobacter pylori as a model, we know that the ModH5 (a Type III MTase) modulates flaA promoter activity in a methylation-dependent manner. The phase-variable switching of ModH5 expression plays a role in regulating Helicobacter pylori phenotypes (Srikhanta et al., 2017).

In addition to the R-M systems, some methyltransferases exist with no association with any restriction enzymes and are designated as “solitary” or “orphan” methylases (Casadesús and Low, 2006). MTases from R-M systems are distinct from orphan MTases. MTases from R-M systems can cleave unmethylated foreign DNA, whereas orphan MTases cannot. In general, MTases in R-M systems are poorly conserved, whereas orphan MTases are as conserved within a genus as any average gene (Seshasayee et al., 2012). Despite the differences, orphan MTases may have been derived from the R-M systems through loss of function in the REase of an R-M system. An R-M system in which the restriction endonuclease lost its activity but its cognate DNA methyltransferase retained its activity is functionally equivalent to orphan methyltransferases (Casadesús and Low, 2006). Some studies have found that orphan methyltransferases may have originated through horizontal gene transfer, with only the MTase part of an R-M system being transferred (Oliveira et al., 2014). Orphan MTases are present in diverse bacterial and archaeal phyla and show motif specificities and methylation patterns that are consistent with functions in gene regulation and DNA replication (Blow et al., 2016). In addition, there is a theory that orphan MTases might have evolved from R-M systems to fight parasitism caused by selfish or rogue R-M systems. However, there is currently a lack of evidence to support this theory (Murphy et al., 2013).

There are three well-studied conserved orphan methyltransferases: Dam, Ccrm, and Dcm, as shown in Table 1. Among them, the Dam of γ Proteobacteria and the CcrM of Caulobacter crescentus are the most widely studied (Chen et al., 2014). Dam is an orphan MTase first found in Escherichia coli and is involved in methyl-directed mismatch DNA repair and regulation of chromosomal replication (Marinus and Morris, 1973; Urig et al., 2002; Marinus and Casadesús, 2009). Its homologs are present in various enteric bacteria, including Yersinia spp., Vibrio cholerae, Salmonella app., Haemophilus influenzae, and other genera (Barras and Marinus, 1989; Torreblanca and Casadesús, 1996; Zaleski and Piekarowicz, 2004; Giacomodonato et al., 2009; Banas et al., 2011). Dam methylation can influence the expression of virulence factors and, thereby, the pathogenicity of some bacteria (Pucciarelli et al., 2002; Watson et al., 2004; Ershova et al., 2015). Modified live attenuated S. enterica serovar Typhimurium that harbor loss-of-function mutations in dam are capable of eliciting protection against a diversity of Salmonella and are well-tolerated when applied as modified live vaccines in poultry, mice, calves, and sheep (Heithoff et al., 2015). Although Dam is not essential for most bacteria, it is necessary for the survival of V. cholerae (Torreblanca and Casadesús, 1996; Robinson et al., 2005; Taylor et al., 2005; Casadesús and Low, 2006; Val et al., 2014).

Dam is a 32 kDa monomeric protein that catalyzes the transfer of the methyl group from AdoMet to the N⁶ position of the adenine residue in 5′GATC3′ sequences. Dam has similar efficiency for both hemi- and unmethylated templates and has been shown to be involved in chromosome replication and segregation, DNA mismatch repair, regulation of transposition events, phase variation, and bacterial conjugation processes (Sánchez-Romero et al., 2015). In fact, all DNA methyltransferases can either methylate a half-methylated DNA strand or simultaneously methylate two unmethylated strands of DNA (Urig et al., 2002). Notably, Dam is a processive enzyme and could continuously methylate adenine in 5′GATC3′ sites, with about 55 sites being methylated per binding event (Brooks et al., 1983; Urig et al., 2002; Bheemanaik et al., 2006; Marinus and Casadesús, 2009). That explains the disproportionally low ratio of Dam to 5′GATC3′ sites.

The cell cycle-regulated DNA MTase family (CcrM) constitutes a second important group of orphan methyltransferase (Skerker and Laub, 2004; Kozdon et al., 2013; Ershova et al., 2015). It plays an essential role in the life cycle of C. crescentus and is highly conserved in all α Proteobacteria, except for Rickettsiales and Magnetococcales (Brilli et al., 2010; Mouammine and Collier, 2018). In Caulobacter, CcrM is an essential cell component and plays a crucial role in cell cycle regulation (Casadesús and Low, 2006). However, it is not present for the entire life cycle and is only expressed right before cell division. Evidence suggests that CcrM participates in the cell-cycle regulation of C. crescentus through regulating the expression of cell-division genes (Zweiger et al., 1994; Stephens et al., 1996; Reisenauer and Shapiro, 2002; Fioravanti et al., 2013). Interestingly, the culture conditions seem to determine the dependence on CcrM methyltransferase in Caulobacter (Gonzalez and Collier, 2013; Adhikari and Curtis, 2016). In some α Proteobacteria, such as Brucella abortus, C. crescentus, and Agrobacterium tumefaciens, CcrM methyltransferase is indispensable (Stephens et al., 1996; Robertson et al., 2000). However, it is not vital for cell growth in other α Proteobacteria, like in Brecundimonas subvibrioides (Robertson et al., 2000; Curtis and Brun, 2014).

CcrM is a functional monomer but may form a dimer at physiologic concentration (Shier et al., 2001; Skerker and Laub, 2004; Kozdon et al., 2013; Casadesús, 2016; Horton et al., 2019). Unlike Dam, CcrM has a distinct preference for hemimethylated DNA substrates (Bheemanaik et al., 2006; Albu et al., 2012). It binds to DNA, and only one 5′GANTC3′ (N represents any base) site can be methylated and modified before the enzyme detaches from the DNA (Marinus and Casadesús, 2009).

Dcm is another orphan MTase that is mainly found in enterobacteria such as E. coli. Dcm is a 51KD protein that methylates cytosine in a position that is rarely modified in bacteria but commonly in eukaryotes. As a C⁵-cytosine methyltransferase, Dcm can methylate the second cytosine in 5′CCAGG 3′ and 5′CCTGG 3′ sites (Militello et al., 2014). Methylated 5′CAG3′ sequences are mutational hotspots. After deamination, 5-methylcytosine is thymine and is not removed by the uracil-N-glycosylase, so GC to AT mutations are common at 5′ CCAGG 3′ sites. Dcm participates in several cellular processes, but it has been shown that Dcm is not necessary for the survival of E. coli (Baba et al., 2008).

Currently, there are 27 bacterial DNA MTases that have been crystallized with three-dimensional structures determined.¹ Their target sequences are summarized in Table 2 (Roberts et al., 2015). Most of these DNA MTases belong to Type I or II R-M systems, and structures for orphan methyltransferases and Type III R-M systems are relatively rare. In general, MTases are bilobed structures folded into two domains: one is the catalytically active region responsible for the transfer of methyl groups, and the other is a smaller region responsible for recognizing methylation sites on DNA (Bergerat et al., 1991; Bheemanaik et al., 2006), Figure 2. There is a common structural core in the larger catalytic region, consisting of a six-stranded parallel β-sheet with a seventh strand inserted in an antiparallel fashion between the fifth and sixth strands (Bheemanaik et al., 2006). This larger catalytic region can be divided into two subdomains. One of the subdomains creates a S-adenosylmethionine (AdoMet) binding site, and the other subdomain is the binding site for the extrahelical target base. The small domains are very diverse in amino acid sequence, size and structure, because of selection for their DNA binding specificity (Bheemanaik et al., 2006). In all cases, the substrate to be methylated is bound or expected to bind in a pocket adjacent to the AdoMet binding site.

The primary sequences of MTases share a set of conserved motifs (I-X) and a variable target-site- recognition domain located near the C-terminus (Lauster et al., 1989; Pósfai et al., 1989). Together, they are responsible for three basic functions: (i) AdoMet binding, (ii) sequence-specific DNA binding, and (iii) catalysis of methyl group transfer. Briefly, motif I accommodates the methionine moiety of AdoMet, which is conserved among all AdoMet-dependent enzymes as a main binding region with AdoMet. In fact, all DNA methyltransferases that use S-adenosylmethionine as methyl donor are remarkably conserved in motif I.

Motifs II and III are also involved in AdoMet binding, but are less conserved than motif I. Several conserved charged residues in motifs I–III have been shown to have substantial effects on AdoMet binding (Ahmad and Rao, 1996). In motif I, substituting certain glycine residues with aspartic acid or arginine residues abolishes AdoMet binding. However, charged residues are not the only determining factor in AdoMet binding. Hydrophobic side chains in motifs I–III have been shown to stabilize AdoMet binding as well (Roth et al., 1998).

Motif IV, which is critical for catalysis, is also known as the DPPY motif because it contains a consensus sequence of (S, N/D)PP(Y/W/F) (Bheemanaik et al., 2006; Horton et al., 2006, 2019). The prolyl dipeptide on Motif IV is considered to play an important role in the transfer of methyl group to the target base (Smith et al., 1990; Cheng and Roberts, 2001). This sequence constitutes an active pocket that can accommodate the target base. DNA methyltransferase can modify the target base when it enters the active pocket through base flipping (Bheemanaik et al., 2006). Aromatic side chains in motif V have been found to interact with AdoMet (Schluckebier et al., 1995). Motifs VI, VII, and VIII form a DNA-binding cleft, whereas motifs I, IV, and X form a binding pocket for AdoMet’s methionine moiety (Cheng and Roberts, 2001).

In general, the mechanism by which DNA methyltransferase recognizes methylation modification sites mainly depends on the surface between the enzyme and the major groove of the substrate double-stranded DNA. The enzyme interacts with deoxyribose and phosphate on the backbone through Van der Waals forces (Roberts and Cheng, 1998; Cheng and Roberts, 2001). Although there are a large number of conserved sequences and homologous structures between different DNA methyltransferases, the interactions between DNA recognition sites and methyltransferase show many variations. The DNA backbone is often severely distorted when a methyltransferase binds to the recognition site.

Methyltransferase Dam contains two domains: a seven-stranded catalytic domain (residues 1–56 and 145–270) harboring the binding sites for AdoMet/S-Adenosyl-L-homocysteine (AdoHcy or SAH) and a DNA binding domain (residues 57–144) consisting of a five-helix bundle and a β-hairpin loop (residues 118–139) (Horton et al., 2006). Dam uses a mechanism called base-flipping to methylate the target adenine. Side chains of the residues in the binding site directly interact with the phosphate groups of DNA backbone. Dam has four conserved residues (R95, N126, N132, and R137) that interact with three consecutive phosphate groups flanking the fourth GATC base-pair of the non-target strand. The methylation target, the adenine of the second base-pair in GATC, is flipped out from the DNA helix. The specific interactions with the remaining bases of the site occur in the DNA major groove. The N-terminal K9 is responsible for recognizing the guanidine of the first base-pair. Contacts to the non-target strand in the second half of the GATC site are established by R124 with the fourth base pair and by L122 and P134 to the third base-pair. The aromatic ring of Y119 intercalates into the DNA between the second and third base-pairs, which is essential for base-flipping to occur (Horton et al., 2006), Figure 3A.

The dimeric CcrM methyltransferase uses a different mechanism to recognize the 5′ GAN⁶ATTC 3′ methylation site (Woodcock et al., 2017; Reich et al., 2018). Four loops of CcrM are involved in recognition of methylation sites. However, almost all the phosphates on the non-target base chain come into contact with Loop-2B (residues 31–61), which recognizes the first three bases of the target sequence. The P45 in Loop-2B is inserted between the second and third base pairs. Loop-45 (residues 119–133) can be inserted into the DNA double helix from a minor groove, where P125 and F126 recognize the fifth base in the target site to provide interactions with the two pyrimidines, T⁴ and C⁵. The shorter Loop-3C (residues 92–94) and Loop-6E (residues 172–194) supply additional interactions for thymine T⁴ (Tyr93 and His94), target adenine A² (Thr191 and Lys193), and the DNA backbone phosphate groups flanking guanine G1 (Arg179 and Lys187) (Horton et al., 2019), Figure 3B.

EcoP15I belongs to the type III R-M family. EcoP15I, like CcrM, also causes the backbone of its target DNA fragment to twist after binding (Gupta et al., 2015). EcoP15I binding target 5′ GANTC 3′ sequence (N represents any base). EcoP15I consists of two methylation (Mod) and one (or two) restriction (Res) subunits, resulting in a Mod2Res1 or Mod2Res2 complex. The Mod subunits are responsible for DNA recognition and methylation, whereas the Res subunits are responsible for ATP hydrolysis and cleavage. Cleavage only occurs if two recognition sites (5′CAGCAG3′) are in an inverted-repeat orientation, arranged either as “head-to-head” or “tail-to-tail” (Gupta et al., 2015), Figure 3C.

C⁵-cytosine methyltransferases catalyze cytosine methylation through intermediates in which the DNA is drastically remodeled. The target cytosine is usually buried in a DNA double helix, which hinders the catalytic reaction. In order to carry out the catalytic reaction, a distortion in the DNA occurs, and the target cytosine residue extruded from the DNA helix and plunged into the active site pocket of the enzyme (Sankpal and Rao, 2002). This base flipping was found in all members of the monomeric type II bacterial methyltransferases, like M.HhaI and M.HaeIII (Matthews et al., 2016). The targeting mechanism is particularly intriguing in the case of M.HaeIII, a bacterial C⁵-cytosine methyltransferase that not only extrudes the substrate cytosine but also induces frameshifted base pairing and the formation of a large gap in the duplex DNA recognition site (Didovyk and Verdine, 2012). The target recognition region (TRD) of M.HaeIII contacts most nucleotides in the recognition site directly, thereby stabilizing their internal helical conformation. Then, residue Ile-221 is inserted between the second and third base pair of the recognition sequence (5′GGCC3′), weakening the base stacking force between them and resulting in the extrusion of the target cytosine. Finally, with concomitant abandonment of pairing for the G and C, underwinding (caused by negative supercoiling) of the DNA double helix located near the target site causes the target base to be flipped out of the DNA double helix, Figure 3D.

Similarly, in Dam, the aromatic ring of tyrosine at 119 can insert between the second and third base pairs of 5′GATC3′, which is considered to be a necessary condition for base flipping (Horton et al., 2006). It is worth noting that the angle of the base flip is different for different MTases. In Dam and CcrM, the target base flips about 90°, while in M.HhaI and EcoP15I, the target base flips about 180° (Gupta et al., 2015). In addition, all four nucleotides in the target site of CcrM can be flipped to different degrees at the same time (Horton et al., 2019). The mechanism needs further investigation.

Over the past two decades, R-M systems have been found not only to protect the integrity of the genome but also to participate in gene expression regulation (Seib et al., 2011; Blakeway et al., 2014; Kwiatek et al., 2015; Anjum et al., 2016; Tan et al., 2016; Srikhanta et al., 2017). Most R-M systems involved in the regulation of gene expression belong to type I and III, with a few belonging to type II, Figure 4A.

Phase variation, the reversible generation of variants of surface antigens, is a survival strategy that is frequently found in pathogenic bacteria (Wion and Casadesús, 2006). Genes that can undergo phase variation and change the expression of other genes are called phase-variable regulons (Tan et al., 2016). Phase-variable regulons can influence methylation patterns of the bacterial genome, which may further change the expression of other genes (Loenen et al., 2014). Phase-variable regulons can produce proteins with different specificities or directly switch gene expression through a phase-switching mechanism and then change the expression of other genes through epigenetic mechanisms. In fact, DNA methylation can regulate gene expression without changing gene sequence, thereby promoting phenotypic heterogeneity in bacterial populations and the formation of bacterial lineages (Cota et al., 2015; García-Pastor et al., 2018), Figure 5. It is worth noting that the coding genes of these R-M systems contain simple sequence repeats (SSR) that are easy to distinguish. These simple repetitive sequences can be a repetitive nucleotide or an inverted repetitive sequence (inverted sequences, IS) (Atack et al., 2018). Pathogens that have adapted to a host can change the number of simple repeats in the open reading frame or shuffle the inverted repeats to generate phase variation (Dupont et al., 2009; Atack et al., 2018). For example, in H. pylori, some DNA methyltransferase genes encoding type II R-M systems contain SSRs that are related to bacterial colonization and pathogenicity (Lin et al., 2001; Ando et al., 2010; Gauntlett et al., 2014).

There are two different kinds of phase variation that have been observed in bacterial R-M systems. Most Type I R-M systems have been shown to carry out phase variation via homologous recombination, Figure 4B. Type I systems are encoded by hsdR, hsdM, hsdS, and 4.3% of hsdR, 2% of hsdM, and 7.9% of hsdS contain simple repetitive sequences which can regulate phase variation (Seib et al., 2020). Regulation of Type I systems can either increase or decrease the number of simple repeats on the coding genes in hsdM and hsdS. Phase variation can also be regulated through the recombination of the inverted repeats in hsdS to generate phase mutations. The number of simple repeats can regulate the expression of hsdM, which leads to the ON or OFF state of DNA MTase. The recombination of inverted repeats or the change in the number of simple repeats in the open-reading frame of hsdS leads to the production of HsdS with different specificities. Ultimately, the site at which methylation occurs changes (Manso et al., 2014).

In H. influenzae, the open reading frame of hsdM, which encodes the type I phase variation regulator, contains simple repeats of a 5′GACGA3′ sequence. Changing these repeats can regulate the expression of hsdM and the sensitivity of H. influenzae to phage infection (Zaleski and Piekarowicz, 2004; Atack et al., 2018). In Neisseria gonorrhoeae, hsdS, which encodes the Type I regulator of phase variation, contains two open-reading frames, hsdSgoAV1 and hsdSgoAV2. There are multiple G residues at the 3′ end of hsdSgoAV1 (polyG tract). Both hsdSgoAV1 and hsdSgoAV2 are expressed when hsdSgoAV1 contains seven G residues at its 3′ end, whereas only hsdSgoAV1 is expressed when there are six G residues. Therefore, the methylation and/or restriction sites can undergo phase variation (Adamczyk-Poplawska et al., 2011).

Another example is the Type I phase variation regulator SpnD39III in Pneumococcus, which contains an hsdS with two target-recognition regions (TRD) and two inverted repeats (IRs) of SpnD39III. The TRD and IR sequences can be shuffled, resulting in six different versions of hsdS (SpnIII39A-F) with different specificities (Manso et al., 2014). Expression of genes involved in capsule synthesis is down-regulated if SpnIII39B is produced. Pneumococcus with SpnIII39A has changes in the expression of genes involved in stress response and nutrient acquisition. However, no significant changes in gene expression have been detected with SpnIII39C or D (Manso et al., 2014).

In Campylobacter jejuni, there is a type II R-M system, Cj0031, containing polyG that could participate in regulating other genes. Campylobacter jejuni regulates the ON or OFF expression of Cj0031 by changing the number of G residues. This regulation can directly lead to the expression or absence of DNA methyltransferase in the system, which affects the methylation status of 5′CCYGA3′ sites on its genomic DNA. Some studies have found that when cj0031 is deleted or shut off by phase variation, the expression or absence of genes related to adhesion and invasion of host cells, such as capA, cadF, and flpAde, are down-regulated, whereas peb1A, a gene encoding periplasmic-binding proteins associated with ABC transporters, is up-regulated (Anjum et al., 2016).

In H. influenzae, there are multiple modA alleles encoding DNA methyltransferases that are involved in phase variation. These alleles include modA1, modA2, modA4, modA5, modA9, and modA10 (Atack et al., 2015). These alleles contain some simple repetitive sequences, and their expression can be adjusted to ON or OFF (Srikhanta et al., 2005; Atack et al., 2015), Figure 4B. For example, the modA1 open-reading frame contains several 5′AGCC3′ SSRs. The number of these repeats determines whether the expression of modA1 is ON or OFF. The modA1 expression level can affect the expression levels of sixteen other genes (Srikhanta et al., 2005). The bacteria can form a robust biofilm when modA2 is expressed. The level of high molecular protein (HMP) in the cell decreases when the expression of modA4 is turned off, which enhances escape from macrophages (Atack et al., 2015).

In H. pylori, there are 19 mod alleles encoding regulators of Type III phase variation (Srikhanta et al., 2010, 2017; Atack et al., 2018). Applying single molecule real-time (SMRT^®) sequencing technology and methylome analysis (Figure 5 inset), it has been found that there is a 5′GACC3′ ModH5 methylation site in the promoter of flaA, which encodes the main component of the flagellar filament (Srikhanta et al., 2017). It has been suggested that methylation at the promoter region of flaA could directly regulate its expression. Similarly, a regulator of Type III phase variation is also reported in pathogenic bacteria Neisseria gonorrheae (Srikhanta et al., 2009), Neisseria meningitidis, Mannella hemolyticus (Srikhanta et al., 2010), and Moraxella catarrhalis (Blakeway et al., 2018). SSRs in the open reading frame of mod have been found to produce different regulators of Type III phase variation that affect the expression of other genes.

The Dam-regulated pap operon is a classic example of regulation by positive feedback (Van der Woude et al., 1996). Pap expression is regulated in a complex manner that involves PapB, PapI, Lrp (Leucine-responsive regulatory protein), and Dam. There are five promoters (P1–P5) in the regulatory region of the pap operon (Løbner-Olesen et al., 1992; Marinus and Casadesús, 2009). There are two main sets of Lrp binding sites, with each set containing a Dam methylation modification site 5′GATC3′. The methylation sites are named as GATC-I and GATC-II (van der Woude et al., 1992). Once Lrp binds to GATC-II, it prevents Dam from methylating GATC-II, resulting in a non-methylated state and blocking RNA polymerase σ70 from binding. In the meantime, GATC-I without Lrp binding is in a methylated state and the transcription of the pap operon is turned off (Pap-Off state). Conversely, when Lrp binds to GATC-I, resulting its being in an unmethylated state and GATC-II being in a methylated state, the transcription of the pap operon is turned on (Pap-On state) (Nou et al., 1995), Figure 6. It has been shown that the binding of PapI to Lrp in vitro reduces the affinity of Lrp for the first set of binding sites by half but increases the affinity of Lrp to the second set of binding sites by four times (Nou et al., 1995). Thus, when PapI expression reaches a certain level, it promotes the transfer of Lrp to the second set of binding sites, shifting from a Pap-On to a Pap-Off state. It has been reported that the frequency of is about 100 times higher than Pap-Off to Pap-On shifting (Blyn et al., 1990). On top of Lrp binding, the entire regulatory region of the pap operon can be bound by H-NS (a histone-like nucleoid structuring protein) (White-Ziegler et al., 1998), phosphorylated CpxR (CpxR-P) (Hernday et al., 2004), and RimJ (White-Ziegler et al., 2002) to affect the methylation status of the region.

Another example of gene regulation is opvAB, which is a cytoplasmic membrane protein gene that can shorten the length of the lipopolysaccharide O-antigen of Salmonella enterica (Cota et al., 2012). The expression of opvAB is turned on when the length of the lipopolysaccharide O antigen needs to become shorter in order to escape infection by bacterial phages, which attach to the lipopolysaccharide O antigen. However, shortened lipopolysaccharide O antigen can reduce the ability to infect its host (Cota et al., 2015). In response to those needs, the expression of opvAB can be turned off in the absence of phage to restore the length of the O antigen.

It is worth noting that the genomes of E. coli and other γ-Proteobacteria have more 5′GATC3′ sequences than expected by chance (Hénaut et al., 1996; Sobetzko et al., 2016). This strongly suggests that Dam methyltransferase may be involved, in regulating transcription of many other genes. Some studies have found that in E. coli, dam mutations could affect the expression of many genes related to aerobic respiration, stress and SOS response, and amino acid and nucleotide metabolism (Oshima et al., 2002; Adhikari and Curtis, 2016). Studies have shown that Dam methyltransferase is involved in regulating transcription of sci1 (Brunet et al., 2011), flu (van der Woude and Henderson, 2008), gtr (Broadbent et al., 2010; Sánchez-Romero and Casadesús, 2020) and std (García-Pastor et al., 2019) in E. coli and carA, dgoR, holA, nanA, ssaN, STM1290 and STM3276 (Sánchez-Romero and Casadesús, 2020) in S. enterica.

However, there is no obvious relationship between Dam-involved gene transcription regulation and the presence of 5′GATC3′ sequences in the promoter of the regulated gene (Horton et al., 2015). Although the expression of many genes changes in dam mutants, only a few of those genes contain a 5′GATC3′ sequence in their promoters (Seshasayee, 2007). Several factors must be considered. In addition to 5′GATC3′, Dam methyltransferase can also react with 5′GTYTA3′/5′TARAC3′. The regulatory regions of the pap, foo and clp genes contain these non-5′GATC3′ sites (Horton et al., 2015). After the deletion of dam in enterohemorrhagic E. coli (EHEC), the expression of the Tir protein and the virulence protein EspFU increased, but their mRNA levels did not change (Campellone et al., 2007). In Yersinia enterocolitica, overexpression of Dam methyltransferase changed the composition of O-antigen, but the mRNA level of the O-antigen gene cluster did not change (Fälker et al., 2007). Similarly, the involvement of Dam methyltransferase in regulating the expression of hild seem to be post-transcriptional (López-Garrido and Casadesús, 2010). Therefore, regulation of gene expression involving Dam can occur after transcription, a process which probably does not involve 5′GATC3′ sites on the genomic DNA. The mechanism of post-transcription regulation by Dam or other DNA methyltransferases is still unclear.

The genome of C. crescentus contains 4542 CcrM methylation modification sites (5′GANTC3′), of which 23% are located inside a cluster of genes (Kozdon et al., 2013). In addition, it has been reported that when CcrM in C. crescentus is deleted or overexpressed, 10% of the genes are misexpressed, and expression of 380 genes changes significantly (Gonzalez et al., 2014). The genes that were significantly affected by the level of CcrM contain at least one 5′GANTC3′ site in their respective promoters. It has been shown that CcrM participates in regulating the transcription of flaY, podJ, ftsN, mipZ and ctrA in C. crescentus (Adhikari and Curtis, 2016). The ctrA gene contains two promoters, P1 and P2, each containing a 5′GANTC3′ methylation modification site (Domian et al., 1999). However, the 5′GANTC3′ site in P2 does not seem to be involved in the regulation of ctrA (Reisenauer and Shapiro, 2002). CcrM regulates the gene ctrA through a corresponding transcriptional regulator GcrA (Reisenauer and Shapiro, 2002). During DNA replication, the replication fork passes through P1, which makes P1 hemimethylated. GcrA binds to the hemimethylated P1 and activates promoter P1, which transcribes ctrA, resulting in the production of a small quantity of CtrA protein (Holtzendorff et al., 2004). CtrA binds to a specific site on the promoter P2, leading to a rapid increase in the level of CtrA in the cell (Domian et al., 1999). The vast majority of CtrA comes from P2 promoter, which demonstrates that the P1 promoter is only active in the hemimethylated state (Reisenauer and Shapiro, 2002).

The timing and level of CtrA expression are critical to the cell cycle of C. crescentus. They coordinate the growth and division of pre-divisional cells. CtrA also regulates the expression of more than 90 genes (Laub et al., 2002). One of the 90 genes that CtrA regulates is ccrM. The surge of CtrA leads to an increase of the CcrM level in the cell. The promoter of ccrM contains two 5′GANTC3′ methylation sites. When chromosome DNA is replicated, two hemimethylated 5′GANTC3′ methylation sites are formed in the pre-divional cell, resulting the surge of CtrA and the increased expression of ccrM gene (Stephens et al., 1996; Reisenauer et al., 1999).