Plasmid promiscuity: meeting the challenge of DNA immigration control

Bacterial plasmids are ubiquitous components of the genomes of naturally occurring bacteria and are typically transmissible between cells by the process of bacterial conjugation (see reviews in Thomas, 2000). A remarkable feature of conjugation is its promiscuity, allowing DNA transfer between phylogenetically remote bacteria and even from bacteria to plant and mammalian cells (Waters, 2001). Conjugative promiscuity is biologically important in dispersing the diverse cargoes of medically and environmentally significant genes carried on plasmids, as well as contributing to the horizontal transfer of ‘fitness islands’, consisting of clustered chromosomal loci that enable pathogens and symbionts to interact with their eukaryotic hosts (Finan, 2002). Conjugation is a bipartite process with a mechanism for effecting the cell-to-cell interactions essential for the formation of ‘mating pairs’ and a matching DNA processing system that causes transfer of a specific plasmid strand (the T-strand) to the recipient cell, where the complementary DNA strand is synthesized. Small natural plasmids may lack a complete conjugation system, but most are transmissible by the process of mobilization: here, the mobilizable plasmid encodes its own DNA processing system, but parasitizes the mating pair formation system of a conjugative plasmid co-resident in the donor cell. The promiscuity of conjugative transfer raises the question as to how plasmids cope with bacterial restriction endonucleases. These enzymes constitute a frequently encountered barrier to the acquisition of DNA identified as ‘foreign’ by lack of the imprinted modification pattern characterizing the DNA of the recipient cell as ‘self’. Restriction-modification (R-M) systems consist of a restriction endonuclease (REase) and a matching modification enzyme with DNA methyltransferase activity (MTase). The REase recognizes a specific nucleotide sequence as a prelude to DNA cleavage, whereas the cognate MTase confers protection from cleavage by methylating a defined adenine or cytosine residue within the specificity site. Importantly, these enzymes operate on double-stranded DNA. Hemimethylated DNA, as generated by replication of fully modified DNA, is protected from cleavage and is the substrate for a modification enzyme. Classical R-M systems are divided into three types according to enzyme composition, cofactor requirements, symmetry of recognition sequence and the DNA cleavage position (Redaschi and Bickle, 1996; Murray, 2002). Type I systems are complex enzymes exemplified by EcoKI. The main enzyme (R2M2S1) comprises heterooligomers of HsdR, HsdM and HsdS subunits conferring DNA sequence specificity (S), modification (M) and restriction endonuclease (R) activities respectively. The complex functions as an REase or MTase depending on the methylation status of the substrate DNA. A M2S1 complex with only methyltransferase activity has also been purified for EcoKI (Dryden et al., 1993). Type I specificity sites have an asymmetric, bipartite sequence of 13–15 bp. Restriction requires ATP hydrolysis and causes DNA cleavage at a location that may be distant from the specificity site. ­Classical type II systems, which include enzymes widely used in DNA manipulation, generally recognize a 4–8 nucleotide (nt) palindromic sequence with twofold rotational symmetry. Type II REases are separate enzymes from the cognate MTases and typically cleave at the ­specificity site. The restriction enzyme database (REBASE.neb.com) lists 50 type I and 3483 type II restriction enzymes. Both types of enzyme are distributed widely among Eubacteria and Archaea, with some strains containing several different chromosome-encoded enzymes (Milkman et al., 1999; Kong et al., 2000). The prevalence and diversity of R-M systems may reflect more than one ecological role. One widely cited view is that the systems evolve as cellular defence mechanisms against infecting foreign DNA identified as alien by lack of cognate methylation. This defensive role may be important in aiding a bacterial strain to colonize a new environment containing previously unencountered bacteriophages. However, any barrier may be transient because restriction enzymes are not absolute barriers, and replicons escaping destruction acquire modification (Korona et al., 1993). Another hypothesis is that restriction potentiates homologous recombination in natural populations of bacteria by cleaving large entrant molecules into recombinogenic fragments (Milkman et al., 1999). A more recent hypothesis is that type II R-M systems evolve as genomic parasites that force their maintenance on bacteria by killing segregant cells (Naito et al., 1995). Type I systems do not behave as selfish units, and some other selection pressure must be invoked to explain their evolution (O’Neill et al., 1997). The first plasmid antirestriction system to be discovered was the ArdA system of the IncI1 group of enterobacterial plasmids typified by ColIbP-9 (Delver et al., 1991; Read et al., 1992). Conserved ardA genes have been detected by functional tests in representatives of four other (B, FV, K and N) incompatibility groups of enterobacterial plasmids (Belogurov et al., 1992; Chilley and Wilkins, 1995). Sequenced ardA genes are predicted to encode small polypeptides of 166–169 amino acids, which are highly acidic with net negative charges ranging from −22 to −29. blast protein database searches indicate that the pMT1 plasmid of Yersinia pestis encodes a conserved ArdA homologue (Lindler et al., 1998). Significant alignments with ArdA are shown by putative antirestriction proteins encoded by a Deinococcus radiodurans plasmid (ArdU protein; Meima and Lidstrom, 2000), by genomic DNA of a filamentous nitrogen-fixing cyanobacterium (Kaneko et al., 2001) and by the conjugative transposon Tn916 of Enterococcus faecalis (ORF18 protein; Flannagan et al., 1994). ArdA proteins specifically inhibit restriction by type I R-M systems. ColIb ArdA functions in conjugation to alleviate restriction of the transferring plasmid, but the protein is without effect in transfers by bacterial transformation. Evidence for a role in conjugation includes co-regulation of ardA expression with the plasmid transfer genes, location of ardA in the leading region defined as the first sector of the plasmid to enter the recipient cell (Chilley and Wilkins, 1995) and transient conjugative induction of ardA after its entry into the recipient cell (Althorpe et al., 1999). Current evidence indicates that this induction reflects the activity of a novel type of promoter (ssi3) that forms in the T-strand to allow transcription of the ardA sequence from the incoming single-stranded DNA (Bates et al., 1999; Fig. 1). According to this model, ardA is rapidly expressed before and until the T-strand is converted into duplex DNA, the substrate for restriction endonucleases. Expression of the ardA gene and the function of plasmid DNA primase in ColIb-mediated conjugation. A. The T-strand is transferring by the rolling circle mechanism with the 5′ leading terminus covalently linked to the DNA relaxase (hexagon). The ssi3 promoter forms in the single strand of DNA as an ≈ 300 nt region of imperfectly paired secondary structure. The promoter potentiates E. coli RNA polymerase to express ardA in a burst of transcription in the newly infected cell. B. Plasmid DNA primase (ovals) is transferred from the donor cell to make short RNA primers for the discontinuous synthesis of the complementary DNA strand. Synthesis of the second strand eliminates the secondary structure that forms the ssi3 promoter, and transcription of ardA is silenced. SSB is single-stranded DNA-binding protein (for further details, see Bates et al., 1999; Wilkins and Thomas, 2000). Enterobacterial plasmids can carry other antirestriction genes, as shown by the presence of ardB on the IncN plasmid pKM101, or R46 (Belogurov et al., 1993). The locus is non-homologous to ardA but is located nearby in the plasmid leading region. ArdB consists of 141 amino acids and is only slightly acidic. The protein confers strong protection against type I restriction, weaker protection against type II restriction and no effect on modification reactions. The closely linked ardA and ardB genes on pKM101 are controlled by conserved upstream elements containing a strong promoter. These promoters are regulated by two plasmid proteins, ArdK and ArdR, with at least ArdK acting as a repressor. The arrangement is thought to control expression of antirestriction and other leading region genes on pKM101 during conjugation (Delver and Belogurov, 1997). The ColIb leading region contains a putative ardB homologue called ycbG. The locus encodes a predicted 142-amino-acid product showing 39% positional similarity to ArdB. Other potential ardB homologues are orf141 located in the F plasmid leading region and klcA of IncP plasmids (Larsen and Figurski, 1994; Manwaring et al., 1999). It has been postulated by these investigators that ardB, klcA and orf141 represent a second family of antirestriction genes that are active in bacterial conjugation to allow restriction avoidance. Indeed, carriage of the loci might explain the considerable resistance of a ColIb ardA mutant and the F-like R100 plasmid to restriction by type I and II systems (Table 1). It is noted that klcA is located outside the IncP plasmid leading region, and there is no evidence that KlcA protein is active against type I restriction systems (Larsen and Figurski, 1994), which is consistent with the sensitivity of transfer of IncP plasmids to restriction (Table 1). A simplistic model is that, as the products of symbiotic genetic elements, Ard proteins transiently block type I restriction systems in the conjugatively infected cell while allowing the immigrant plasmid to acquire protective ­modification. This hypothesis is inconsistent with reports that, under non-physiological conditions causing overproduction of the protein, ColIb and pKM101 ArdA inhibited modification by EcoKI in addition to restriction (Delver et al., 1991; Read et al., 1992; Belogurov et al., 1992). The effects of ArdA on EcoKI have been re-evaluated using the ardA homologue carried by the IncB plasmid R16, which shares 95% nucleotide positional identity with the ColIb equivalent. The results show that R16 ArdA protein inhibited the restriction activity of EcoKI, leaving significant levels of methyltransferase activity under conditions in which restriction was almost completely prevented (A. T. Thomas and B. M. Wilkins, unpublished). Thus, at least one ArdA protein selectively targets the restriction activity of type I enzymes. It is unclear how ArdA and ArdB enzymes combat type I restriction systems. The polypeptides contain a 14-residue domain that is common to the gene 0.3 antirestriction (Ocr) protein of the virulent phage T7. 0.3 protein contains 116 residues and is very acidic with a net charge of −28; it can block all functions of type I R-M enzymes. Recent biophysical evidence indicates that 0.3 protein acts as a dimer to mimic both the size and the shape of a bent DNA molecule of ≈ 20 nt and the arrangement of the negative charges along the phosphate backbone of B-form DNA (Walkinshaw et al., 2002). A similar domain is also found in the HsdS (specificity) polypeptide of type I enzymes. The presence of the domain and the remarkable acidity of the plasmid and phage antirestriction proteins led Belogurov and Delver (1995) to propose that Ard proteins function similarly to 0.3 protein by binding the type I R-M enzyme and disrupting the functional integrity of the complex, ­perhaps to inhibit DNA recognition. Such a hypothesis fits uneasily with the evidence that cells overproducing R16 ArdA protein retain significant type I modification activity (see above) and that ArdB has anti-restriction but not antimodification activity (Belogurov et al., 1993). One appealing speculation is that Ard proteins target the R2M2S1 form of the type I enzyme but not the M2S1 complex, which is active as a methyltransferase. The IncW plasmid pSa encodes a third type of restriction avoidance protein called ArdC (Belogurov et al., 2000). ArdC is specific for type I restriction systems. The protein binds single-stranded DNA and is predicted to consist of 297 residues; it shows no significant sequence similarity to ArdA and ArdB except for possessing the 14-amino-acid ‘antirestriction’ domain. Remarkably, the protein has some 38% positional identity to the N-terminal portion of the TraC1 product of RP4. TraC1 is a plasmid DNA primase and is very unusual among bacterial proteins in being transferred abundantly into the cytoplasm of the recipient cell during conjugation. Here, the plasmid ­primase can generate RNA primers to initiate complementary strand synthesis on the T-strand (Rees and Wilkins, 1990; Fig. 1). The N-terminal portion of TraC1 is thought to be required for binding to the T-strand or for passage of the protein across the cell envelope during conjugation. It was proposed that ArdC is transported through the conjugative pore to the recipient to confer immediate ­protection on the incoming T-strand (Belogurov et al., 2000). Thus, ArdC is a putative member of a dedicated group of plasmid proteins that are transmitted into the recipient by conjugation. Protein and DNA transfer by conjugation-like processes is a specialized mechanism described as type IV secretion. This secretion pathway is important in the transport of specific virulence effectors from bacterial pathogens to human cells (Wilkins and Thomas, 2000; Christie, 2001). An interesting parallel to ArdC exists in bacteriophage P1 infections in which phage Dar antirestriction proteins are delivered with the DNA to inhibit type I restriction of the infecting genome (Iida et al., 1987). Some plasmids have undergone selection for elimination of restriction enzyme recognition sites. IncP plasmids are remarkable for their deficiency of cleavage sites targeted by commonly used type II restriction enzymes relative to the frequencies predicted from the mononucleotide composition of the plasmids. Paradoxically, IncP plasmids contain an unusually high density of sites for type II enzymes that recognize certain rotationally symmetric palindromes containing G+C residues (Wilkins et al., 1996; Thorsted et al., 1998). The under- and over-represented palindromes are located in the plasmid backbone, defined as the sectors responsible for replication, maintenance stability, transfer and central control. The biased frequencies presumably reflect a combination of mutational and selection pressures exerted on IncP genomes by the evolutionary hosts of the plasmids. This inference is based on the finding that the frequencies of all 256 tetranucleotide motifs in the RP4 (IncPα) backbone are similar to those in Pseudomonas aeruginosa, a candidate evolutionary host, with the notable exception that many 6 bp palindromes are under-represented in the plasmid. The overabundant restriction sites in RP4 are explicable on the basis that the plasmid contains overlapping copies of CGGC/GCCG, which is identified as the most abundant tetranucleotide in the plasmid. Compared with the transfer of IncI1 plasmids, that of IncP plasmids is unusually sensitive to type I and type II restriction systems (Table 1), implying that IncP plasmids lack a dedicated antirestriction gene and, hence, have experienced particularly strong selection for loss of restriction targets. Interestingly, Pseudomonas spp. of rRNA homology group I, the putative evolutionary hosts of the plasmid backbone, specify many type II enzymes that target 6 bp palindromes (Wilkins et al., 1996). Some R-M systems are specified by natural conjugative and mobilizable plasmids themselves. The plasmids may determine type I (e.g. EcoR124I) or type II (e.g. EcoRI and PvuII) systems. This property raises questions about the nature of safeguard mechanisms that permit the horizontal transfer of such plasmids without subsequent degradation of unprotected chromosomal DNA in a naïve recipient cell lacking cognate modification. Protection against the potentially lethal effects of the acquisition of a type II system may be regulated by various mechanisms. These include preferential expression of the MTase at the transcriptional and translational levels and a passive mechanism in which the MTase polypeptides are active early as monomers, whereas the REase polypeptides require dimerization to provide a functional enzyme. There is also evidence that transient low levels of DNA cleavage can be tolerated (Redaschi and Bickle, 1996). Transfer of the hsd determinants of a type I system to a naïve recipient cell is associated with post-translational regulation of restriction activity (Kulik and Bickle, 1996; Makovets et al., 1998). Intriguingly, transfer of the EcoKI genes was observed to be associated with prompt expression of the modification phenotype, but a delay in the onset of restriction by as many as 15 cell generations (Prakash-Cheng et al., 1991). Efficient transmission of the EcoKI determinants requires a two-component protease encoded by the chromosome of the recipient cell. The protease consists of ClpX active as a molecular chaperone, which prepares the polypeptide substrate for hydrolysis by ClpP as a serine protease. The ClpXP complex is thought to target the HsdR subunit for degradation after the REase has bound to unmodified chromosomal DNA targets and before DNA cleavage (Makovets et al., 1998). The ClpXP restriction evasion mechanism may have evolved to allow vegetative bacteria to survive the production of unmodified target sequences generated by errors in replication or recombinational repair of damaged DNA; similar chromosomal damage may trigger the response when a cell acquires a novel hsd system (Murray, 2000; 2002). Importantly, the ClpXP response is not induced by the entry of unmodified plasmids or phages, as the DNAs of phage M13 and phage λ as well as plasmid F are recognized as foreign in Clp+ and ClpX– recipients ­(Doronina and Murray, 2001; Murray, 2002). A striking observation is that the known plasmid antirestriction enzymes preferentially target type I restriction systems. This bias might reflect the variety and prevalence of type I R-M systems in the evolutionary hosts of the plasmids. Alternatively, evolution of an anti-type I restriction function would confer the advantage of protecting against a number of different enzymes, as type I but not type II systems group into families in which the subunits are functionally interchangeable. Further understanding of the mode of action of Ard proteins is likely to require the purification of the enzymes for biochemical tests. Conversely, proof that ArdB and ArdC proteins operate in conjugation will require appropriate genetic tests. Another challenge is to elucidate the discriminatory process whereby ClpXP-mediated restriction alleviation is caused by the generation of unmodified targets in the chromosome, but not in immigrant DNA. Cellular compartmentalization is unlikely to be the answer, as replication of entrant single-stranded DNA occurs in the cytoplasm. A final unanswered question concerns the widely varied sensitivities of different plasmids to restriction when ­allowance is made for the number of specificity sites per plasmid (Table 1). Which is the more common phenotype: the relative resistance of the narrow-host-range IncI1 plasmids or the sensitivity of IncP plasmids widely regarded as the paradigm of promiscuous genetic ­elements? I thank William Brammar and Noreen Murray for stimulating discussions of the biology of restriction. This work was supported by Wellcome Trust grant 060852.