Microbiology (2002),
148, 3-20.
© 2002 Society for General Microbiology
Immigration control of DNA in bacteria: self versus non-self
Noreen E. Murray1
Institute of Cell and Molecular Biology, Darwin Building, Mayfield Road, Edinburgh EH9 3JR, Scotland, UK1
Tel: +44 131 650 5374. Fax: +44 131 650 8650. e-mail: noreen.murray{at}ed.ac.uk
Keywords: Restriction and modification, control by proteolysis, DNA transfer, DNA modification, selfish genes
 |
Background and aims
|
|---|
Bacteria commonly endow their DNA with an identity mark. When DNA is transferred from one bacterium to another strain of the same species, DNA that lacks the identification mark of the recipient strain is recognized as ‘foreign’ rather than ‘self’. Foreign DNA is commonly degraded. The first evidence for this discriminatory process was the demonstration of a barrier, albeit incomplete, to the productive infection of Escherichia coli strain K-12 by bacteriophage 
previously propagated in either E. coli strain C or E. coli strain B (Bertani & Weigle, 1953
). Much later it was proven that the growth of phages in E. coli K-12 can be ‘restricted’ by an endonuclease, a restriction enzyme (EcoKI), which attacks foreign DNA (Meselson & Yuan, 1968 ; Linn & Arber et al., 1969). Occasionally phages escape restriction and they, like the resident bacterial chromosome, acquire a protective identification mark from a strain-specific modification enzyme that methylates defined bases within a specific target sequence (Arber & Dussoix, 1962 ; Smith et al., 1972 ). This sequence-specific modification identifies the immediate provenance of bacterial, or phage, DNA (Fig. 1).
Classically, a restriction enzyme is accompanied by its cognate modification enzyme and together the two activities comprise a restriction and modification (R-M) system. There are, however, some restriction endonucleases, so-called modification-dependent restriction enzymes, which attack DNA only when specific nucleotide sequences in the DNA are methylated. The classical R-M systems and the modification-dependent restriction enzymes share the potential to attack DNA derived from different strains and thereby ‘restrict’ DNA transfer. While the modification activity of a classical R-M system is required to protect DNA from attack by the cognate restriction endonuclease, a modification enzyme specified by one strain may impart a signal that provokes the degradative activity of a modification-dependent restriction endonuclease found in a different strain.
It is often stated, though difficult to prove, that restriction systems exist to defend bacteria against invading phages. Recently, however, it has been argued that R-M systems are ‘selfish’ elements. This hypothesis emanates from the finding that bacterial cells die if they lose the genes that specify their R-M system (Naito et al., 1995 ). It has been shown that the bacterial chromosome becomes susceptible to restriction as cell growth dilutes the modification enzyme (Handa et al., 2000 ). However, while the loss of genes that specify some simple R-M systems leads to cell death, the loss of genes that specify other, more complex, R-M systems causes no detectable viability problem (O’Neill et al., 1997 ; Kulik & Bickle, 1996 ; Makovets et al., 1998 ). In this review I wish to emphasize the different behaviour of E. coli strains dependent upon the nature, or type, of their R-M system (see also Murray, 2000 ). Some data will challenge our long-established belief that modification of DNA is essential to distinguish whether the DNA is ‘self or foreign’. Experiments show that while modification of DNA is sufficient, it is not always essential to identify resident DNA as self (Makovets et al., 1999 ; Doronina & Murray, 2001 ).
|
Types of R-M systems
|
|---|
R-M systems have been subdivided according to the complexity and cofactor requirement of the enzymes, the nature of their target sequence, and the position of the site of DNA cleavage with respect to the target sequence. Three distinct, well-characterized types of classical R-M systems have been defined (types I, II and III; Fig. 2), although a few systems do not share all the characteristics of any of these three types (for general reviews see Wilson & Murray, 1991 ; Bickle & Kruger, 1993 ; Raleigh & Brooks, 1998 ). The first R-M systems identified in E. coli K-12 and E. coli B were designated type I, but the enzymes that serve as reagents in modern biology, type II R-M systems, are very much simpler. For this reason, they are described first.
View larger version (18K):
[in this window]
[in a new window]
|
Fig. 2. The characteristics and organization of the genetic determinants and subunits of different types of R-M systems. ENase, endonuclease activity; MTase, methyltransferase activity. Modified with permission from King & Murray (1994) .
|
|
A type II R-M system comprises two separate enzymes, a restriction endonuclease and a modification enzyme, or methyltransferase. The nuclease activity is dependent on Mg2+ and the methyltransferase on S-adenosylmethionine (AdoMet) as the methyl donor. The restriction and modification enzymes recognize the same target sequence, usually a rotationally symmetrical sequence of 4–8 bp. Type II endonucleases are generally active as symmetrically arranged homodimers, an association that facilitates the co-ordinated cleavage of both strands of the DNA. The modification enzyme ensures that a specific base within the target sequence, one on each strand of the duplex, is methylated, but modification enzymes function as monomers, an organization consistent with their normal role in the methylation of newly replicated DNA (for reviews see Wilson & Murray, 1991 ; Roberts & Halford, 1993 ; Raleigh & Brooks, 1998 ; Pingoud & Jeltsch, 2001 ).
Genes encoding repressor-like proteins, referred to as C proteins for control, have been identified for some type II R-M systems (Ives et al., 1992 ; Tao et al., 1991 ; Tao & Blumenthal, 1992 ). The C-protein for the BamHI system has been shown to activate efficient expression of the restriction gene (Ives et al., 1992 , 1995 ). Consequently, when R-M genes are transferred to a new environment in which there is no C protein, there will be preferential expression of the modification gene, and only after the production of C protein will transcription of the restriction gene be activated.
Type I R-M systems are heterooligomeric complexes that catalyse both restriction and modification (for reviews see Murray, 2000 ; Rao et al., 2000 ; Dryden et al., 2001 ). AdoMet is the methyl donor for modification but, importantly, endonuclease activity requires both AdoMet and ATP, in addition to Mg2+. The restriction activity of type I R-M systems is associated with the hydrolysis of ATP, an activity that correlates with the bizarre characteristic of these enzymes, that of translocating DNA before they cut it at nonspecific sequences considerable distances from the target sequence (Davies et al., 1999b ). The nucleotide sequences recognized by type I enzymes are asymmetric and comprise two components, one of 3 or 4 bp and the other of 4 or 5 bp, separated by a non-specific spacer of 6–8 bp. The type I R-M enzyme binds to its target sequence in the presence of cofactors and the alternative activities of restriction or modification are determined by the methylation state of the target sequence. Hemimethylated target sequences are the substrate for modification but, if the target sequence is unmodified, the enzyme, while bound to its target sequence, translocates the DNA from both sides towards itself in an ATP-dependent manner. DNA cleavage occurs when translocation is impeded (Studier & Bandyopadhyay, 1988 ; Janscak et al., 1999a ).
The three subunits of a type I R-M system are encoded by closely linked genes: hsdR, hsdM and hsdS. The acronym hsd denotes host specificity of DNA. hsdM and hsdS are transcribed from the same promoter; hsdR is from a separate one (Loenen et al., 1987 ). The two subunits encoded by hsdM and hsdS, colloquially referred to as M and S, are both necessary and sufficient for methyltransferase activity. The third subunit, HsdR or R, is essential only for restriction. The specificity subunit, S, includes two target recognition domains (TRDs) that impart target-sequence specificity to the restriction and modification activities of the complex; the M subunits include the active site for DNA methylation and the R subunits that for nuclease activity. Two complexes are functional in bacterial cells: one comprises all three subunits (R2M2S1) and is an R-M system, and the other lacks R (M2S1) and has only methyltransferase activity (Lautenberger & Linn, 1972 ; Suri & Bickle, 1985 ; Taylor et al., 1992 ).
A separate promoter from which hsdR is transcribed suggests a means for regulating restriction activity, but experiments provide no evidence for the transcriptional regulation of any of those type I R-M systems for which data are available (Kulik & Bickle, 1996 ; Loenen et al., 1987 ; Prakash-Cheng et al., 1993 ). Evidence is accumulating for the role of post-translational regulation of restriction activity (see section on the mechanism by which restriction activity of EcoKI is controlled).
Type III R-M systems are less complex than type I systems but nevertheless share some similarities with them (see Rao et al., 2000 ). A single heterooligomeric complex catalyses both restriction and modification activities. Modification requires the cofactor AdoMet, and restriction requires Mg2+ and ATP. Recent evidence indicates that type III restriction enzymes can translocate DNA. DNA cleavage is stimulated by collision of the translocating complexes and occurs close to, but on the 3' side of, the target sequence (Meisel et al., 1995 ).
The foci of this review lecture are type I R-M systems, their extraordinary capacity for diversification and the acutely sensitive mechanisms for the control of their restriction activity: these mechanisms of control protect unmodified ‘self’ DNA from attack.
|
Distribution of R-M systems
|
|---|
Commercial catalogues document the presence of type II R-M systems in a wide variety of bacterial strains. No ‘market force’ has driven searches for type I systems: nevertheless there is biological evidence for functional type I R-M systems in Bacillus subtilis, Citrobacter freundii, Klebsiella pneumoniae, Lactococcus lactis, Mycoplasma pulmonis, Staphylococcus aureus and many strains and species of Salmonella as well as those found in E. coli (see Murray, 2000 ). Computer-based analyses of the nucleotide sequences of bacterial genomes identify numerous putative R-M systems, of all types. Potential R-M genes within completed genomic sequences have recently been tabulated by Kong et al. (2000) . Their survey indicates that >80% of the bacterial genomes for which completed sequences are available have at least one R-M system (see Table 1). Both type I and type II systems are prevalent throughout the Eubacteria and Archaea. It may be significant that strains for which screens of genomic sequences failed to identify putative R-M systems included those from very special environments, such as parasitic species in which bacterial growth may occur only within eukaryotic cells, e.g. Chlamydia, Rickettsia and Treponema pallidum, and Aquifex aeolicus, a thermophile that lives at extremely high temperatures.
View this table:
[in this window]
[in a new window]
|
Table 1. Number of potential restriction systems in microbial genomes based on computational analyses of DNA sequences
|
|
|
Diversification of sequence specificity
|
|---|
Type I R-M systems appear to be better suited to evolve new specificities than are the simpler type II systems. In summary, the following points seem relevant. First, the specificity of both the restriction and modification activities of a type I R-M complex is conferred by a single specificity subunit, S; therefore a change in specificity concomitantly affects restriction and modification. Second, those type I R-M systems that have been studied are sensitive to a sophisticated mechanism that controls their endonuclease activity, thereby protecting the resident chromosome from attack. Third, a specificity subunit that comprises two TRDs, each recognizing a different target sequence, offers more scope for diversification than a classical type II restriction endonuclease which, as a dimer of identical subunits, recognizes a symmetrical target sequence (Wilson & Murray, 1991 ).
Long repeated nucleotide sequences remain in the specificity genes of some type I R-M systems as evidence of gene duplication, providing an explanation for the origin of current specificity genes encoding two TRDs (Kannan et al., 1989 ). Early type I R-M systems with the subunit composition R2M2S2 are likely to have recognized hyphenated symmetrical sequences, dictated by the symmetrical arrangement of two specificity subunits. Enzymes of this sort have been generated by deletions that truncate a specificity gene leading to an active enzyme comprising two symmetrically arranged truncated subunits (Abadjieva et al., 1993 ; Meister et al., 1993 ). Diversification of TRDs has led to the recognition of a variety of target sequences comprising 3–5 bp but always sequences within which an adenine residue is the substrate for methylation.
The evolution of a type I R-M system with a different specificity (see Fig. 3) was first witnessed by chance in the laboratory (Bullas et al., 1976 ) and later shown to be the result of recombination generating a hybrid S gene encoding a new combination of TRDs (Fuller-Pace et al., 1984 ). Similarly, a minor change in the length of the spacer sequence connecting the two TRDs was shown to alter the length of the spacer sequence separating the two components of the target sequence (Price et al., 1989 ). New combinations of TRDs can be generated experimentally quite readily, but attempts to generate new specificities as the result of changes within a TRD have been unsuccessful. The majority of many amino acid substitutions made within a TRD of EcoKI do not impair specificity (O’Neill et al., 2001 ). It seems likely that more than one amino acid substitution is necessary to change the specificity of a TRD. Even for type II R-M systems for which the structures of enzymes bound to their target sequences have been determined, it has not been possible to predict amino acid changes that lead to a new specificity. To date, BamHI has been changed so that it prefers a methylated substrate (Dorner et al., 1999 ) and EcoRV has been engineered so that its preferred target sequence is 8 rather than 6 bp (Lanio et al., 2000 ).
View larger version (51K):
[in this window]
[in a new window]
|
Fig. 3. Evolution of type I R-M systems with new specificities. (a) Recombination between hsdS genes produces hybrid genes and chimeric S polypeptides. StySPI and StyLTIII are naturally occurring type I R-M systems (see Table 2). StySQ and StySJ have hybrid hsdS genes (Fuller-Pace et al., 1984 ; Gann et al., 1987 ). The regions originating from StySPI are hatched and those originating from StyLTIII are stippled. Reassortment of the TRDs accordingly gave rise to recombinant recognition sequences (Gann et al., 1987 ; Nagaraja et al., 1985 ). Site-directed mutagenesis of the central conserved region of the StySQ hsdS gene produced StySQ*, comprising only the amino-terminal variable region from StySPI and the remainder from StyLTIII. The StySQ* target sequence confirms that the amino-terminal variable region is in fact a TRD responsible for recognition of the trinucleotide component of the sequence (Cowan et al., 1989 ). (b) Sequence specificity may also be altered by changing the length of the nonspecific spacer of the target sequence. The S polypeptides of EcoRI24I and EcoRI24II differ only in the number of times a short amino acid motif (X=TAEL) is repeated within their central conserved regions (Price et al., 1989 ), resulting in extension of the spacer in the target sequence from six nucleotides (N6) for EcoRI24I to N7 for EcoRI24II. The recognition sequence of EcoDXXI also contains a nonspecific spacer of 7 nt, corresponding to three TAEL repeats in its S polypeptide (Gubler et al., 1992 ). Chimeric S polypeptides recognize the predicted target sequences (Gubler et al., 1992 ). Modified with permission from a figure by Barcus & Murray (1995) .
|
|
|
The immigration control region and the family concept
|
|---|
In E. coli K-12 the genes specifying EcoKI are flanked by genes that encode methylation-dependent restriction endonucleases (Mrr and McrBC). The segment of the genome that specifies these three endonucleases has been referred to as the immigration control region (Raleigh, 1992 ). Genetic analyses of other strains of E. coli and Salmonella enterica indicated considerable allelic diversity within, or close to, this region long before the era of genomic sequences (Boyer & Roulland-Dussoix, 1969 ; Bullas et al., 1980 ). E. coli strains K-12 and B, and S. enterica serovars typhimurium LT2 and potsdam, have alleles that specify type I R-M systems with different specificities. The respective enzymes (EcoKI, EcoBI, StyLTIII and StySPI) differ from each other in one or both of their TRDs (Gough & Murray, 1983 ; Fuller-Pace et al., 1984 ). Of fundamental influence in our understanding of type I R-M systems has been the demonstration that these enzymes can be considered as members of a family within which the subunits of different enzymes are interchangeable. It came as a surprise, however, that alleles at this locus, in particular those specifying EcoAI in E. coli strain 15T-, encode sufficiently dissimilar type I R-M systems to warrant their separation into a different family (Murray et al., 1982 ). The initial evidence came from hybridization screens of bacterial DNAs and serological screens of bacterial extracts. As expected, the nucleotide sequences of hsd genes for EcoKI and EcoBI would hybridize to each other and antibodies raised against EcoKI reacted with EcoBI, but in contrast, DNA probes comprising the EcoKI genes failed to hybridize with those of E. coli 15T-; similarly antibodies against EcoKI did not react with EcoAI. At least three families of type I R-M systems (IA, IB and ID) are encoded by alternative genes within the immigration control region of enteric bacteria (Fig. 4); currently these identify at least 16 specificities (Barcus et al., 1995 ; Thorpe et al., 1997 ; Titheradge et al., 2001 ). The sequence of the genome of E. coli O157 (Perna et al., 2001 ) identifies a type IB system.
View larger version (13K):
[in this window]
[in a new window]
|
Fig. 4. Alternatives at the hsd locus of E. coli. The diagrams identify the hsd genes within the immigration control region of E. coli K-12, E. coli 15T-, E. coli R9 and E. coli C.
|
|
An additional family (type IC), headed by EcoRI24I, was recognized initially via plasmid-encoded members (Glover et al., 1983 ), but genes for a chromosomally encoded representative have been identified in an E. coli strain at a location distinct from the immigration control region (Tyndall et al., 1994 ). While the only major difference between two enzymes within the same family is confined to their TRDs, the subunits of members of different families share only limited identity (15–35%) when their amino acid sequences are aligned. Sequence comparisons of the putative type I R-M systems predicted from genomic sequences suggest that family affiliations extend across the Eubacterial kingdom (Titheradge et al., 2001 ).
It seems likely that all type I R-M systems derive from a common ancestor (Sharp et al., 1992 ), but systems allocated to different families are now so dissimilar that little evidence of homology remains at the level of gene sequences. One interesting exception is the 5' part of the specificity genes of StyLTIII (type IA) and EcoAI or EcoEI (type IB); these specify a TRD that recognizes the same trinucleotide target sequence (Table 2
). An examination of the target sequences of the type I R-M systems (see Table 2) indicates that the evolution of different families of enzymes has enhanced the scope for diversification by varying the distance between those adenine residues within the target sequences that are the substrates for methylation. In the target sequences for members of the IB family, the adenine residues are separated by 9 bp, in the IA family by 8 bp, in the IC family by 7 or 8 bp and in type ID by only 6 bp; the variability in the IC family is dependent on whether a tetrapeptide sequence (TAEL) within the central conserved region is present in duplicate or in triplicate. The importance of the correct spacing between the adenine residues is illustrated by the target sequences for EcoRI24I
and EcoDXXI (see Table 2). These are the systems that comprise symmetrically arranged truncated S polypeptides and their target sequences require an additional base pair in the spacer to maintain the distance between the adenine residues.
In summary, diversity of specificity in type I systems, where two TRDs are present within the specificity subunit, not only depends on diversification of TRDs but is enhanced by different spacing between the TRDs and new combinations of TRDs.
|
Is restriction an effective barrier to the acquisition of foreign DNA?
|
|---|
DNA in which the target sequences lack the correct identification mark is generally sensitive to restriction irrespective of whether the DNA enters the cell in single-or double-stranded form. Phage or plasmid DNA that enters in a single-stranded form becomes susceptible to restriction after the synthesis of the second strand. The fragmentation of foreign DNA reduces the efficiency of productive, or lysogenic, infection by phages, and the frequency of acquisition of conjugative plasmids. DNA fragments, particularly those that share sequence similarity with the resident chromosome, may be rescued by recombination. Early experiments in which gene transfer was monitored when unmodified donor DNA from an Hfr strain entered a restriction-proficient recipient showed the acquisition of early markers to be inefficient and linkage much reduced (Boyer, 1964 ; Pittard, 1964 ; Arber & Morse, 1965 ). However, many phage and conjugative plasmids, but not the well-known F factor, have the means of moderating their susceptibility to R-M systems. They may modify their DNA in unusual ways, or produce proteins that interfere with restriction, e.g. phage T7 or plasmid ColIb.
The modification of DNA by glucosylation, as in T-even phages, is effective against most restriction systems, while proteins that interfere with the activity of the enzyme may be specific to one enzyme, or one type of system. Phage T5 can inhibit the activity of EcoRI (Davison & Brunel, 1979 ) and a variety of host enzymes that modify DNA (see McCorquodale & Warner, 1988 ), but most of the anti-restriction functions currently identified are directed against type I systems. It seems unlikely that this bias towards functions that protect against type I systems simply reflects the fact that most work has been done with E. coli K-12 and E. coli B; E. coli strains specifying EcoRI have been in common laboratory use for 30 years. The bias could reflect the prevalence of type I systems in natural strains of E. coli enhanced by the fact that some feature common to type I R-M systems, or the conformation of their DNA substrates, permits the evolution of anti-restriction proteins that are able to combat all members of one family or even the members of different families of type I R-M systems.
The 0.3 gene products of phages T3 and T7 are the only anti-restriction functions available in significant quantities for detailed molecular analyses. These proteins, sometimes referred to as Ocr (overcoming classical restriction), bind to type I restriction complexes, both the R-M complex and the modification enzyme, and prevent them from binding to DNA (Atanasiu et al., 2001 ). The T3 product also destroys the cofactor AdoMet. It has been suggested that the 0.3 gene product, or Ocr, mimics the DNA substrate, thereby neutralizing the R-M complexes (Bandyopadhyay et al., 1985 ). Recent evidence based on the structure of Ocr supports this model: the protein is an elongated dimer that reflects both the size and shape of a bent DNA molecule (Atanasiu et al., 2001 ; M. Walkinshaw & D. Dryden, personal communication). An alternative proposal for the Ard (alleviation of restriction of DNA) proteins of conjugative plasmids, based on their acidic nature, is that an acidic surface mimics sequences of the specificity subunits of type I systems and the Ard proteins can displace the specificity subunit from the active R-M complex (Belogurov & Delver, 1995 ). Both the 0.3 gene product (C. Atanasiu & D. Dryden, personal communication) and ArdA (Read et al., 1992 ) are active against members of different families of type I R-M enzymes.
The efficacy of anti-restriction functions poses the critical question of how a protein specified by the unmodified DNA of a transmissible agent is able to act before the sequence that encodes it is attacked by the restriction enzyme. Bacteriophage P1 solves the problem by co-transfer of the protein with its DNA. In contrast, the 0.3 genes of T3 and T7 are transcribed early, prior to the internalization of the remainder of the genome. The ard genes of transmissible plasmids, like the 0.3 gene of T3 or T7, are located in the leading end of the DNA, but for conjugative plasmids it is single-stranded DNA that is transferred (5' to 3'). Current evidence for IncI1 and ColIb supports a regulatory model in which the genes in the leading region of the DNA are transcribed from special promoters recognized within secondary structures of single-stranded DNA (Bates et al., 1999 ). This allows transcription of ard genes and the accumulation of anti-restriction protein before the transferred strand is converted into duplex DNA (see Fig. 5).
View larger version (8K):
[in this window]
[in a new window]
|
Fig. 5. The ORFs in the leading region of ColIb. The direction of transfer from nic is from left to right. All ORFs (shown as arrows) are transcribed from right to left. ardA identifies the ORF specifying the anti-restriction protein. The regions identified as ssi are presumptive promoters for leftward transcription of the transferred strand of DNA. Reproduced with permission from Bates et al. (1999) .
|
|
|
Conserved sequences and active sites in type I R-M systems
|
|---|
While the S subunit confers sequence specificity to both the R-M (R2M2S1) and modification (M2S1) complexes, the M subunit contributes the active site for modification. Modification enzymes, whether type I, II or III, include motifs characteristic of methyltransferases. The type I systems transfer methyl groups to adenine residues and their M subunits include the sequence N/DPPF/Y/W as motif IV rather than the PC motif characteristic of cytosine methyltransferases (see Dryden, 1999 ). For EcoKI, amino acid substitutions within motif IV have been made that block the catalytic activity without impairing the binding of AdoMet, the methyl donor (Willcock et al., 1994 ). In contrast, substitutions in motif I prevent binding of the methyl donor, a cofactor essential for restriction as well as modification.
The R subunits are essential for restriction but not modification. A type I R subunit includes motifs characteristic of ATP-binding proteins (Loenen et al., 1987 ), consistent with the ATP-dependence of restriction. In addition, they include conserved sequences indicating the presence of motifs characteristic of ATP-dependent helicases (Gorbalenya & Koonin, 1991 ; Murray et al., 1993 ; Titheradge et al., 1996 ). It has been suggested that these motifs, the DEAD-box motifs, define an ‘engine’ that powers DNA translocation (Hall & Matson, 1999 ). Analyses of mutations in the hsdR gene of E. coli K-12 (Fig. 6) demonstrated that each of the seven DEAD-box motifs of EcoKI is essential for a restriction-proficient phenotype and for the DNA-dependent ATPase activity of the enzyme (Davies et al., 1998 , 1999a ). Of special relevance was the finding that these restriction-deficient mutants lack DNA translocation activity (Davies et al., 1999a ). This activity was assayed by monitoring the EcoKI-dependent transfer of the T7 genome from the phage capsid to the bacterial cell (Fig. 7), an assay that relies on the inhibition of RNA polymerase activity, the normal means of DNA transfer, and the presence within the leading region of the T7 genome of a single target for EcoKI (Garcia & Molineux, 1999 ). The EcoKI complex bound to the unmodified EcoKI target can mobilize the 39 kb of T7 DNA at the rate of 
100 bp s-1.
View larger version (9K):
[in this window]
[in a new window]
|
Fig. 6. Domains, motifs and amino acid substitutions in the HsdR subunit of EcoKI. The N- and C-terminal regions are omitted. The two domains that include the DEAD-box motifs correlate with domains IA and 2A, as determined for structures of DNA helicases (see Davies et al., 1999b). Substitutions for an underlined amino acid confer a restriction-deficient phenotype. These changes identify the restriction-deficient strains analysed for DNA translocation, ATPase and endonuclease activities. Reproduced with permission from Murray (2000) .
|
|
Additional conserved sequences in the N-terminal part of the R subunits of type I R-M systems (Titheradge et al., 1996 ) show similarities with those motifs associated with DNA nicking in other nucleases (Davies et al., 1999b ). Site-directed mutagenesis proved the relevance of this motif to the endonuclease activity of EcoAI (Janscak et al., 1999b ) and EcoKI (Davies et al., 1999a , b ). Experiments in vitro for EcoAI (Janscak et al., 1999b ) and in vivo for EcoKI (Davies et al., 1999a , b ) showed that changes within the endonuclease motif do not block the ATPase and translocase activities of the R-M complex.
|
Mechanism of action of type I restriction enzymes
|
|---|
Our current understanding of the mode of action of a type I restriction enzyme is essentially as outlined in the ‘collision’ model of Studier & Bandyopadhyay (1988) . According to this model (Fig. 8), an enzyme binds to its target sequence and while remaining bound to this sequence it pulls in the DNA from both sides, simultaneously, in a process dependent upon the hydrolysis of ATP. When translocation is impeded, as for example by the collision of two translocating complexes, endonuclease activity is stimulated.
Representatives of three families of type I R-M systems have been studied in vitro (see Szcelkun, 2000 ). Each endonuclease is dependent upon AdoMet, ATP and Mg2+, and all are believed to function in a similar way. For EcoKI, the addition of either ATP or a non-hydrolysable analogue in the presence of AdoMet allows tight binding of the enzyme to unmodified target sequences. DNA footprints demonstrate a conformational change that precedes the hydrolysis of ATP (Powell et al., 1998 ). Enzymes with substitutions in DEAD-box motifs remain capable of the conformational change associated with target recognition, despite their failure to hydrolyse ATP and translocate DNA (Davies et al., 1998 , 1999a ). Enzymes with conservative substitutions within the endonuclease motif retain their ability to translocate DNA, but these enzymes fail to hydrolyse phosphodiester bonds (Davies et al., 1999a , b ).
The in vivo and in vitro consequences of mutations in the hsdR gene of E. coli K-12 separate the restriction pathway into a series of steps in which AdoMet and ATP are required as cofactors for specific binding to the target sequence, while ATP hydrolysis is essential for the DNA translocation that precedes the eventual breakage of phosphodiester bonds in a Mg2+-dependent reaction (Fig. 9). Known mutations in hsdR apparently fail to prevent the binding of ATP and they block either the second or the third step in the pathway. AdoMet binds to the M rather than the R subunit; a substitution in motif I of the M subunit of EcoKI, which prevents the binding of AdoMet (Willcock et al., 1994 ), results in an enzyme incapable of either modification or restriction (Doronina & Murray, 2001 ). This defect is consistent with the predicted block in the first step of the restriction pathway. In contrast, a substitution in motif IV, which blocks methyltransferase activity but has little effect on the binding of AdoMet, leaves a complex able to translocate and break DNA. The expected consequence of this mutation in vivo would be fragmentation of the bacterial chromosome. However, recent experiments contradict this expectation (Makovets et al., 1999 ; Doronina & Murray, 2001 ; Cromie & Leach, 2001 ). It would appear that when modification fails, the bacterial cell is endowed with the means of causing the restriction pathway to abort before the enzymes break the DNA. This effective control of the restriction activity of type I complexes is in stark contrast to the cell death that follows the concomitant loss of the genes that encode type II R-M system.
|
Guarding the bacterial chromosome against DNA breakage
|
|---|
DNA modification marks and protects the chromosome of a restriction-proficient bacterium, but there are situations where unmodified targets could become exposed to a restriction enzyme and thereby jeopardise the integrity of the bacterial chromosome. An obvious example of this problem is encountered when a bacterium acquires genes that encode a different R-M system from any already present within the cell. One simple solution is to delay production of the restriction enzyme until the modification enzyme has had time to modify all the targets in the bacterial chromosome (Prakash-Cheng & Ryu, 1993 ). This process, however, takes many generations following the acquisition of the genes specifying EcoKI, because unmethylated DNA is a very poor substrate for modification (Makovets, 1999 ). For type II R-M systems, transcriptional control of gene expression is well documented (see Raleigh & Brooks, 1998 ), but transcriptional control has not been found to be relevant for any type I or type III system that has been investigated (Loenen et al., 1987 ; Prakash-Cheng et al., 1993 ; Kulik & Bickle, 1996 ; Redaschi & Bickle, 1996 ). The dependency of type II R-M systems on transcriptional regulation would explain why E. coli can cope with the acquisition of type II systems but is sensitive to their loss; following gene loss, transcriptional control is no longer possible and residual endonuclease will attack unmodified targets within the bacterial chromosome (Handa et al., 2000 ). The loss of genes encoding type I R-M systems is not associated with any loss of viability (O’Neill et al., 1997 ; Makovets et al., 1998 ). This may reflect loss of restriction activity by the dissociation of the R subunits of EcoKI to yield a complex (M2S1) with only modification activity.
The early experiments of Bertani & Weigle (1953) showed that the restriction proficiency of E. coli K-12 was alleviated following UV irradiation. Many experiments now document this phenomenon for type I systems, but not, so far, for any type II system. A similar response has been demonstrated for a variety of agents that damage DNA, including mutagens such as the base analogue 2-aminopurine (2-AP), and defects in some genes that affect DNA metabolism, e.g. dam, topA and mutD (dnaQ) (Efimova et al., 1988a , b ; Thoms & Wackernagel, 1984 ; Makovets et al., 1999 ). DNA damage may generate unmodified target sequences as a consequence of the repair of double-strand breaks by homologous recombination (see Fig. 10), or directly by mutations that create target sequences. The original genetic evidence for the creation of vulnerable target sequences by mutation (Makovets et al., 1999 ) is now supported by the demonstration of breaks in the bacterial chromosome when E. coli K-12 is treated with 2-AP. The breaks are dependent on EcoKI and, as predicted if they arise by base substitutions, their generation requires two rounds of replication (Cromie & Leach, 2001 ).
View larger version (16K):
[in this window]
[in a new window]
|
Fig. 10. DNA damage can induce the alleviation of restriction. The diagram illustrates how unmodified target sequences could be generated following DNA damage. Methylated strands of DNA are shown as thick lines and unmethylated strands are shown as thin lines. Homologous recombination, involved in the repair of double-strand breaks or postreplicative repair, can generate regions of unmethylated double-stranded DNA via annealing of two unmethylated strands (regions within boxes). In addition, the SOS mutagenesis pathway leads to new (unmodified) target sequences as the result of base changes. 2-AP, a base analogue, is believed to create new target sequences as the result of base substitutions. Reproduced from Murray (2000) with permission.
|
|
Diversification of sequence specificity appears to be the hallmark of type I R-M systems and the control of restriction activity could facilitate the generation of new specificities. In Mycoplasma pulmonis, site-specific inversions of sequences within the specificity gene can ‘switch’ the sequence specificity of resident systems (Dybvig et al., 1998 ). This finding prompts the question of whether most cells that acquire an enzyme with a new specificity die, or whether the restriction potential of the new enzyme is controlled by a mechanism other than transcriptional regulation.
Finally, in the context of the generation of new specificities, it seems likely that the evolution of a TRD that recognizes a different nucleotide sequence will require a series of amino acid changes, some of which may initially impair the efficiency of modification. Our recent experiments show that even a modest drop in modification activity, one so small that the mutant strain still scores as modification proficient, elicits the modulation of restriction activity and this modulation is essential for the bacterium to survive (O’Neill et al., 2001 ). In this case a mutation in hsdS is associated with a restriction-deficient, modification-proficient phenotype!
|
ClpX and ClpP are needed to modulate the restriction activity of some type I R-M systems
|
|---|
The efficient transmission of the genes encoding EcoKI requires some function specified by the recipient cell, if the recipient is modification deficient (Prakash-Cheng et al., 1993 ). Given the heterooligomeric nature of the R-M complex (R2M2S1), an obvious way of alleviating the restriction activity within the recipient cells would be to destroy, or sequester, the R subunits of the complex. Energy-dependent proteases are known to play important regulatory roles in bacteria (see Gottesman, 1999 ), therefore mutants deficient in proteases were screened to check whether they might identify the unknown function. These experiments implicated the protease ClpXP (Makovets et al., 1998 ), which comprises two components, ClpX and ClpP. In the absence of either ClpX or ClpP, acquisition of hsd genes specifying either EcoKI (type IA) or EcoAI (type IB) led to the death of modification-deficient recipients (Makovets et al., 1998 ). Together, ClpX and ClpP form a large, but hollow, complex (see Gottesman, 1999 ); ClpX serves to recognize and unfold its substrate so that the polypeptide can be transported to the chamber within the complex where it becomes the target for degradation by ClpP. The alleviation of restriction in response to treatment with UV light, nalidixic acid or 2-AP, and to mutations in dam, topA or mutD, is dependent on ClpXP (Makovets et al., 1999 ). Similarly, survival of mutants in which methyltransferase activity is blocked (Makovets et al., 1999 ; Doronina & Murray, 2001 ), or even slightly impaired (O’Neill et al., 2001 ), requires ClpXP. The ClpXP protease provides a mechanism for controlling the restriction activity of type IA and IB systems, but it is not relevant to the control of all type I systems (see Murray, 2000 ).
|
The mechanism by which the restriction activity of EcoKI is controlled
|
|---|
The fate of the subunits of EcoKI is readily monitored when restriction is alleviated in response to a DNA-damaging agent or because of a defect in modification activity. When E. coli was treated with 2-AP, a reduction in the concentration of the R polypeptide was observed if the cells were clp+ but not if they were clpX- (see Fig. 11); the concentration of M appeared to be unaffected. In the absence of ClpX, the stability of the R polypeptide is enhanced in cells treated with 2-AP. These results are consistent with the activation of a control pathway in which R becomes susceptible to ClpXP-dependent proteolysis (Makovets et al., 1999 ). This susceptibility to proteolysis was found only when the R subunit was part of a functional restriction complex; neither a wild-type R subunit in the absence of M or S nor a defective R subunit in the presence of wild-type M and S was susceptible to proteolysis in vivo. These findings suggest that control of the restriction activity requires that the R-M complex can recognize its substrate and thereby embark on the restriction pathway. Such a mechanism would provide a remarkably specific control process that becomes effective only after the restriction pathway is initiated, but is able to act before damage is inflicted. An EcoKI complex with a substitution in motif IV of HsdM that blocks methyltransferase activity but not the endonuclease activity (Doronina & Murray, 2001 ), should initiate the restriction pathway on the resident DNA thereby making the R subunits vulnerable to ClpXP-dependent degradation. An examination of this mutant strain (hsdMF269G) revealed the predicted depletion of the wild-type R subunit, but depletion did not occur when the complex was impaired by a missense mutation in hsdR (Makovets et al., 1999 ). A modification-deficient EcoKI complex leaves a bacterial chromosome with around 600 unmodified target sequences. According to our model (Makovets et al., 1999 ), these targets will provide a powerful stimulus for the ClpXP-dependent alleviation of restriction by the degradation of R.
View larger version (97K):
[in this window]
[in a new window]
|
Fig. 11. Treatment with 2-AP leads to a Clp-dependent deficiency of HsdR. The figure depicts a series of assays for HsdR and HsdM of EcoKI, following treatment with 2-AP. Panel (a) shows extracts from clp+ bacteria, panel (b) from clpX bacteria. The polyclonal antibody used in the Western blots fails to detect HsdS, but detects some other E. coli proteins in addition to HsdR and HsdM. In the absence of 2-AP (data not shown), the assays for clp+ and clpX bacteria were indistinguishable from those seen in (b). Taken with permission from Makovets et al. (1999) .
|
|
The available missense mutations in the hsdR gene of E. coli K-12 block either the ATP-dependent DNA translocation or the later step of DNA breakage. Both classes of mutants are defective in restriction. Are the R subunits of both classes of mutants refractory or susceptible to ClpXP-dependent proteolysis? A series of double mutants was made in which a mutation in hsdR was combined with the mutation (hsdMF269G) that provokes degradation of HsdR. Each double mutant was monitored for the presence of the R subunit (Fig. 12). The R subunit of restriction-deficient mutants in which the ATP-dependent translocation activity was retained (Davies et al., 1999a ) remained sensitive to proteolysis (as in track 3), but no depletion of R (see track 7) was observed in mutants where ATP-dependent translocation was blocked (Doronina & Murray, 2001 ).
View larger version (34K):
[in this window]
[in a new window]
|
Fig. 12. The effect of mutations in hsdR on the degradation of HsdR in response to a mutation in hsdM (substitution F269G) that blocks methyltransferase activity of the EcoKI complex. The bacteria in tracks 2, 4, 6 and 8 have no mutation in hsdM, and hence no stimulus to alleviate restriction, those in tracks 3, 5, 7 and 9 have the substitution F269G in HsdM. Degradation of HsdR correlates with the ATPase activity of the complex. Each of seven mutations that block ATPase activity, like that shown in track 7, prevents the degradation of HsdR; the two mutations that block endonuclease activity but have no effect on ATPase activity, like the one shown in track 3, had no effect on the degradation of HsdR (Doronina & Murray, 2001 ).
|
|
The finding that ClpXP-dependent proteolysis p
TOP
Background and aims
Types of R-M systems
