Highlights of the DNA cutters: a short history of the restriction enzymes

Abstract

In the early 1950’s, ‘host-controlled variation in bacterial viruses’ was reported as a non-hereditary phenomenon: one cycle of viral growth on certain bacterial hosts affected the ability of progeny virus to grow on other hosts by either restricting or enlarging their host range. Unlike mutation, this change was reversible, and one cycle of growth in the previous host returned the virus to its original form. These simple observations heralded the discovery of the endonuclease and methyltransferase activities of what are now termed Type I, II, III and IV DNA restriction-modification systems. The Type II restriction enzymes (e.g. EcoRI) gave rise to recombinant DNA technology that has transformed molecular biology and medicine. This review traces the discovery of restriction enzymes and their continuing impact on molecular biology and medicine.

INTRODUCTION

Restriction endonucleases (REases) such as EcoRI are familiar to virtually everyone who has worked with DNA. Currently, >19 000 putative REases are listed on REBASE (http://rebase.neb.com) (1). REases are classified into four main types, Type I, II, III and IV, with subdivisions for convenience; almost all require a divalent metal cofactor such as Mg²⁺ for activity (Table 1 and Figure 1). Type II REases represent the largest group of characterized enzymes owing to their usefulness as tools for recombinant DNA technology, and they have been studied extensively. Over 300 Type II REases, with >200 different sequence-specificities, are commercially available. Far fewer Type I, III and IV enzymes have been characterized, but putative examples are being identified daily through bioinformatic analysis of sequenced genomes (Table 1).

Figure 1.

Schematic representation of the functional roles filled in different ways in R-M systems. The functions served include the following: S, DNA sequence specificity; MT, methyltransferase catalytic activity; M, SAM binding; E, endonuclease; T, translocation. Boxes outline functions that are filled by distinct protein domains. Different colours indicate different functions, while different boxes represent distinct domains. Domains in a single polypeptide are abutted, and those in separate polypeptides spaced apart. The order of domains in a polypeptide may vary—e.g. not all Type IIS enzymes have the cleavage domain at the C-terminus. In some cases, functions are integrated with each other, e.g. S and E functions of Type IIP (striped box); in other cases, separate domains carry them out, e.g. Type IIS. See Table 2 for the complexity and diversity of Type II subtypes. The large families of Type I and Type II systems are currently subdivided in 5 and 11 different groups, respectively. The Type I families are distinguished by homology; the Type II groups are distinguished by catalytic properties rather than sequence homology. Type IV enzymes were initially identified as hydroxymethylation-dependent restriction enzymes and currently comprise a highly diverse family. Two examples are shown with and without a translocation domain.

Table 1.

Characterization and organization of the genes and subunits of the four Types of restriction enzymes

Type	Type I	Type II	Type III	Type IV
Features	Oligomeric REase and MTase complex Require ATP hydrolysis for restriction Cleave variably, often far from recognition site ‘DEAD-box’ translocating REase bipartite DNA recognition domain	Separate REase and MTase or combined REase∼MTase fusion Cleave within or at fixed positions close to recognition site Many different subtypes	Combined REase + MTase complex ATP required for restriction Cleave at fixed position outside recognition site ‘DEAD-box’ REase	Methylation-dependent REase Cleave at variable distance from recognition site Cleave m6A, m5C, hm5C and/or other modified DNA Many different types
Example	e.g. EcoKI	e.g. EcoRI	e.g. EcoP1I	No ‘typical’ example
Genes	hsdR, hsdM, hsdS	e.g. ecorIR, ecorIM	e.g. ecoP1IM, ecoP1IR	e.g. mcrA, mcrBC, mrr
Subunits	∼135, ∼62 and ∼52 kDa	∼31 and ∼38 kDa for EcoRI	∼106 and ∼75 kDa for EcoP1I	Unrelated proteins
Proteins	REase: 2R + 2M + S	Orthodox REase: 2R	REase: 1 or 2 R + 2M	Varies
	MTase: 2M + S (±2R)	Orthodox MTase: M	MTase: 2M (±2R)
REBASE	104 enzymes, 47 genes cloned, 34 genes sequenced, 5140 putatives	3938 enzymes, 633 genes cloned, 597 sequenced, 9632 putatives	21 enzymes, 19 genes cloned & sequenced, 1889 putatives	18 enzymes & genes cloned, 15 sequenced, 4822 putatives

Type I and II are currently divided in 5 and 11 different subclasses, respectively. Few enzymes have been well-characterized, but based on the current avalanche of sequence information many putative genes belonging to all Types and subtypes are being identified and listed on the restriction enzyme website (http://rebase.neb.com). The modification-dependent Type IV enzymes are highly diverse and only a few have been characterized in any detail. In each case, an example is given of one of the best-characterized enzymes within the different Types I, II and III. Note that Type II enzymes range from simple (shown here for EcoRI) to more complex systems (see Table 2 for the diversity of Type II subtypes).

REBASE count is as of 16 September 2013 (http://rebase.neb.com/cgi-bin/statlist).

Here we present a non-specialists perspective on important events in the discovery and understanding of REases. Studies of these enzymes have generated a wealth of information regarding DNA–protein interactions and catalysis, protein family relationships, control of restriction activity and plasticity of protein domains, as well as providing essential tools for molecular biology research. Discussion of the equally fascinating DNA-methyltransferase (MTase) enzymes that almost always accompany REases in vivo is beyond the scope of this review, but we note that base flipping, first discovered in the HhaI MTase (2), is not confined to these enzymes alone, but appears to be a common phenomenon that is also used by certain REases (3) and in other nucleic acid-binding enzymes (4–7).

Most research interest has focused on Type I and II enzymes for historical and practical reasons, so this history is weighted to their treatment. The molecular, genetic and enzymological properties of these have been extensively reviewed [see e.g. (8–12)], and separate reviews of the Type I, III and IV systems appear elsewhere in this journal.

THE FIRST HIGHLIGHTS

Discovery of ‘host-controlled variation’

Many important scientific developments in the first half of the 20th century laid the groundwork for to the discovery of restriction and modification (R-M). These included the discoveries of radiation and to the ability to incorporate isotopes in living cells; the molecular building blocks of DNA, RNA and protein; ‘filterable agents’ (viruses); the isolation of Escherichia coli and other bacteria, and of their viruses [called (bacterio)phage] and plasmids. Enabling technical advances included development of electron microscopy, ultracentrifugation, chromatography, electrophoresis and radiographic crystallography. Key was the emerging field of microbial genetics, which flourished owing to the discovery of lysogeny, conjugation, transduction, recombination and mutation.

Preliminary descriptions of the phenomenon of R-M were published by Luria and Human (1952) (13), Anderson and Felix (1952) (14) and Bertani and Weigle (1953) (15). These reports of ‘host-controlled variation in bacterial viruses’ were reviewed by Luria (1953) (16). Host-controlled variation referred to the observation that the efficiency with which phage infected new bacterial hosts depended on the host on which they previously grew. Phage that propagated efficiently on one bacterial strain could lose that ability if grown for even a single cycle on a different strain. The loss was not due to mutation, and one cycle of growth on the previous strain returned the virus to its original state once more. R-M systems of all types initially were investigated in this same way, by comparing the ‘efficiency of plating’ (eop; number of plaques on the test host divided by the number of plaques on a permissive host) on alternate bacterial hosts (17–21). Eop values would range from ∼10⁻¹ to 10⁻⁵, thus indicating that R-M systems were effective barriers to the uptake of DNA; see (16,22–26) for early reviews.

DNA modification

A decade after these initial reports, Werner Arber and Daisy Dussoix, using phage lambda as experimental system, showed that it was the phage DNA that carried the host-range imprint (17). Different specificities could be imprinted concomitantly both by the bacterium itself (by what were later recognized to be the Type I EcoKI and EcoBI systems) and by phage P1 in its latent prophage state (the Type III EcoP1I system). Gunther Stent suggested that DNA methylation might be the basis for the modification imprint, thus prompting Arber to show that methionine was required in the growth medium to produce the imprint on the DNA (27). This important finding coincided with the discovery of RNA-methyltransferase and MTase activities in bacteria that catalysed the formation of m5C and m6A (28). Arber’s interest in the biochemical mechanisms of R-M was driven in part by insight that R-M enzymes would prove useful for analysing DNA molecules and DNA–protein interactions. He concluded a landmark 1965 review of host-controlled modification with the following words: ‘Looking toward future developments … it is to be hoped that the enzymes involved in production and control of host specificity will be isolated and characterized. Such studies, paralleled with investigations of the genes controlling R-M and of their expression, should eventually permit an explanation of the high degree of strain specificity, for example “by a mechanism of recognition of certain base sequences”. If this last idea should be correct one may further speculate that a restriction enzyme might “provide a tool for the sequence-specific cleavage of DNA” ’ (22) (our double quotes).

Sequence-specific DNA-cleavage

As chance would have it, the R-M systems studied by Arber, Type I and Type III (25), do not provide simple enzymes for the sequence-specific cleavage of DNA (see further below). However, REases with the desired sequence-specific cleavage were soon isolated, and these set the stage for the advances in gene analysis and manipulation, collectively called ‘recombinant DNA technology’, that quickly followed. The first of these new enzymes, HindII, was discovered in Hamilton (‘Ham’) Smith’s laboratory at Johns Hopkins Medical School in 1970 (29). This was subsequently termed a Type II REase as its properties were distinct from the Type I REases (25). Purified from Haemophilus influenzae serotype d, HindII (originally called endonuclease R) was found to act as a homodimer and to cleave DNA at the symmetric (though degenerate) sequence GTY’RAC (Y = C or T; R = A or G;’ indicates the cut site) (29,30). Subsequently, what was thought to be pure HindII was found to be a mixture of HindII and a second REase made by the same bacterium, HindIII. HindIII cleaved DNA at a different symmetric sequence, A’AGCTT (31,32); [see (33) for a thought-provoking discussion]. The existence of HindIII came to light during experiments to characterize the MTase activities of H. influenzae. These experiments showed that the HindII and HindIII MTases acted at the same DNA sequences as those cleaved by the REases. They modified these sequences rather than cleaving them, producing GTYRm6AC and m6AAGCTT, respectively (34–36).

The universe of enzymes in this Type II category expanded rapidly. As Smith’s work proceeded on the east coast of the USA, REases with similar behaviour but different specificity were discovered in the laboratory of Herb Boyer at the University of California, San Francisco, on the west coast. Here, PhD student Robert Yoshimori (37) benefited from the experience of Daisy Dussoix, who had moved from Werner Arber’s lab to UCSF. Yoshimori investigated restriction systems present on plasmids in clinical E. coli isolates, and purified what became known as EcoRI and EcoRII (37,38). The EcoRI REase was found to cleave G’AATTC (39,40) and the corresponding M.EcoRI MTase to modify the inner adenines in this sequence, producing GAm6ATTC (41). The EcoRII REase was found to cleave ‘CCWGG (W = A or T), and the M.EcoRII MTase to modify the inner cytosines, producing Cm5CWGG (42,43).

Staggered cuts and the advent of genetic engineering

In contrast to the Type I and Type III enzymes studied in the 1960’s, EcoRI and HindIII cleave DNA within their recognition sites and, most importantly, produce staggered cuts. Since the recognition sites are symmetric, this means that every fragment is flanked by the same single-stranded extension, allowing any fragment to anneal (via the extensions) to any other fragment, thus setting the stage for recombining DNA fragments and ‘cloning’. These findings were presented at the 1972 EMBO Workshop on Restriction, organized by Werner Arber (see Supplement S1 for details of the program and attendees). Figure 2 shows a photograph of participants at this Workshop, recalled by Noreen Murray as the most exciting meeting in the history of the REases, with discussions on the impact of this vital new information on ‘sticky ends’ and the implications for novel DNA manipulation. The recently described DNA ligase (44) would allow the joining of DNA fragments with the same sticky ends. EcoRI and HindIII spurred the development of recombinant DNA work through the availability of both purified enzymes and of replicatable carriers known as vectors. Both phage lambda (45) and various plasmids (46,47) were developed into vectors into which DNA fragments generated by EcoRI and HindIII could be ligated.

Figure 2.

Photograph of the participants at the EMBO Workshop on restriction in Leuenberg (Basel), Switzerland, 26–30 September 1972, organized by Werner Arber, who took the picture (Archive Noreen Murray). Supplement S1 contains a list with names of the attendees and the program of this meeting, and puts names to faces as far as the attendees could be identified (from the archives of Noreen Murray and Werner Arber).

Fittingly, in 1978, Werner Arber was awarded the Nobel Prize together with Dan Nathans and Ham Smith in recognition for their pioneering work on R-M (www.nobelprize.org).

Emerging genetic and enzymatic complexity

While the 1972 review by Matt Meselson et al. (26) mentions only the recognition sequence of HindII, the pace soon quickened. The discovery of new restriction enzymes skyrocketed, as laborious in vivo phage-plating assays were replaced by rapid in vitro DNA-cleavage assays of cell extracts. Elucidation of differences in recognition and cutting led to the classification of additional distinct classes, or types, of restriction enzymes (25,48), which with extensions and subdivisions has stood the test of time: Type I (exemplified by EcoKI, EcoBI, EcoR124, the ‘classical’ enzymes); Type II (EcoRI, HindIII, EcoRV, the ‘orthodox’ enzymes); and Type III (EcoP1I and EcoP15I), Table 1 and Figure 1. Type IV (modification-dependent REases Mcr and Mrr) was added later (49). Sequencing and biochemistry have since led to subdivisions within the Type I (see below) and Type II systems (Table 2) [see (49,50) and http://rebase.neb.com for nomenclature and details].

Table 2.

Nomenclature of Type II restriction enzymes

Subtype	Features of restriction enzymesa	Examples
Type IIP	Palindromic recognition sequence; recognized by both homodimeric and monomeric enzymes; cleavage occurs symmetrically, usually within the recognition sequence	Prototypes EcoRI & EcoRV
Type IIA	Asymmetric recognition sequence	FokI
Type IIB	Cleavage on both sides of the recognition sequence	BcgI
Type IIC	Single, combination R-M polypeptide	HaeIV
Type IIE	Two sequences required for cleavage, one serving as allosteric effector	EcoRII, Sau3AI
Type IIF	Two sequences required for cleavage, concerted reaction by homotetramer	SfiI
Type IIG	Requires AdoMet cofactor for both R-M	Eco57I
Type IIH	Separate M and S subunits; MTase organization similar to Type I systems	BcgI
Type IIM	Require methylated recognition sequence; Type IIP or Type IIA	DpnI
Type IIS	Asymmetric recognition sequence; cleavage at fixed positions usually outside recognition sequence	FokI
Type IIT	Heterodimeric restriction enzyme.	Bpu10I, BslI
Putatives	All subtypes
Control	Control proteins of Type II restriction enzymes	C.BamHI, C.PvuII

The characteristics of the orthodox Type IIP enzymes originally distinguished this group of enzymes from the Type I and III R-M systems. Type IIP is the largest group, owing to its valuable role in molecular science and its commercial value, but the current classification and growing number of R-M systems (putatively) identified, makes it clear that Type II enzymes are highly diverse and the boundaries with the other types are beginning to blur; see also Figure 3 and text for details.

^aThese classifications reflect enzyme properties and activities, and not their evolutionary relationships. The classifications are not exclusive, and one enzyme can often belong several classes. Thus BcgI, for example, is Type IIA, B, C, G and H (see text for details).

The recombinant DNA scare

In their 1975 review (51), Nathans and Smith discuss methods for DNA cleavage and separation of the resulting fragments on gels, as well as the use of REases in other applications, e.g. the physical mapping of chromosomes, taking Simian Virus SV40 (SV40) as an example. The debate on the safety of recombinant DNA technology started soon after the 1972 EMBO Workshop and reports on the transfer of eukaryotic DNA into E. coli [documented by (52)]. The debate was extremely heated, but by 1990 many of the fears had abated as the anticipated dangers did not materialize and the advantages of DNA cloning, and the ability to produce large quantities of pharmaceutically important proteins such as insulin, hormones and vaccines became clear.

FURTHER HIGHLIGHTS IN THE STUDY OF TYPE I R-M SYSTEMS

Type I families are defined by complementation and display sequence conservation

Type I REases were originally identified in E. coli and other enteric organisms as barriers to DNA entry. They turned out to be oligomeric proteins encoded by the three host specificity determinant (hsd) genes: a restriction (R), modification (M) and recognition (S for specificity) gene, respectively (Table 1 and Figure 1). Before the development of DNA sequencing, genetic complementation tests defined the hsdR, hsdM and hsdS genes (53,54). DNA hybridization studies, and probing with antibodies directed at EcoKI, established that EcoKI and EcoBI were more closely related to each other than to EcoAI, the Type I system in E. coli 15T⁻ [reviewed in (8)]. This approach based on biological interaction led to the division of these systems into families: the Type IA (EcoKI, EcoBI, EcoDI and Salmonella typhimurium StySPI); Type IB (EcoAI, EcoEI and Citrobacter freundii CfrAI); Type IC (EcoR124, EcoDXXI, EcoprrI) (8); and later, Type ID (StyBLI and Klebsiella pneumoniae KpnAI) (9,55,56) and Type IE (KpnBI) (57); see reviews for further details (8–10,58,59). Other organisms will have their own families, for example, Staphylococcus aureus has at least two families [(60) and unpublished DTFD results].

Preparation, cofactor requirements and structures

Landmark studies on purified enzymes in the wake of the 1962 Arber and Dussoix articles (17,61) date to 1968. Stu Linn and Werner Arber in Switzerland and Matt Meselson and Bob Yuan in the USA, respectively, purified EcoBI and EcoKI. They used restriction of phages fd and lambda as their assay for detecting the enzymes during purification, a laborious process (62,63). This was no simple matter; Bob Yuan recalls that ‘the fall flew by in deep frustration’ until he and Matt discovered that the enzyme needed S-adenosylmethionine (SAM) for activity in addition to Mg²⁺ and ATP (See Supplement S2 for his personal story). The same cofactor requirement was also found for EcoBI (62,64), reviewed in (25,26). Twenty-five years later, we have come to appreciate that SAM, like ATP, is a widely used cofactor in many metabolic reactions (65,66).

A long-awaited breakthrough did not happen until much later: The structures of the subunits and assembled Type I R-M enzymes. Two structures of S subunits appeared in 2005 and culminated in 2012 with the structure of two complete R-M enzymes containing two R subunits, two M subunits and one S subunit (67–69).

Type I enzymes cut away from the target site

In 1972, Horiuchi and Zinder showed that the DNA recognition site of EcoBI is not the cleavage site (70). They cut ³H-labelled double-strand RF DNA of phage f1 (a relative of phage M13) with EcoBI, denatured and renatured the DNA and then treated with EcoBI a second time. This resulted in a heterogeneous distribution of small DNA fragments on alkaline sucrose gradients leading to the conclusion that EcoBI cuts at a variable distance from its target site. This feature is now known to be common to all Type I restriction enzymes. It was later shown by Studier that EcoKI would preferentially cleave DNA approximately half-way between consecutive target sites (71). This feature is also common to all Type I restriction enzymes although the distribution of cleavage locations can be broad.

DNA translocation to reach the cutting sites uses molecular motors

Translocation was first observed in electron microscope (EM) studies that showed DNA looping by EcoBI and EcoKI. These were interpreted as reaction intermediates, formed by the enzymes translocating along the DNA while remaining attached to their recognition sites (72,73). In the case of EcoKI, studies with full-length phage lambda DNA and relaxed or supercoiled circular DNA showed that EcoKI translocates the DNA past itself, concomitant with a large conformational change of the enzyme, creating large bidirectional loops clearly visible in the EM. Recent studies confirm that on DNA binding, the enzyme strongly contracts from an open to a compact form (58,69,74). In contrast, EcoBI appeared to form loops in only one direction. Later studies with EcoBI did show supercoiled structures like EcoKI; however, translocation was still unidirectional, and without any apparent strand selectivity in the cleavage reaction (75,76). The translocation process would explain the cleavage observed half-way between consecutive target sites on a DNA molecule as two translocating enzyme molecules would collide roughly half-way between target sites.

The R-subunit of Type I enzymes belongs to the SNF2 helicase/translocase superfamily of proteins. These appear to be the result of an ancient fusion between nuclease and ATP-dependent RecA-like (AAA+ or ‘motor’) domains, a linkage found in many enzymes involved in DNA repair, replication, recombination and chromosome remodelling (77–88). As such, Type I enzymes could prove useful for understanding the action of SNF2 enzymes in higher organisms, including the coordinated steps of DNA scanning, recognition, binding and alteration of the helical structure, that allow other domains or subunits to move and touch the DNA. All of these steps are required to prevent indiscriminate nuclease activity (69).

The key to the functionality of the Type I REase and other SNF2 proteins is their enormous flexibility, allowing large conformational changes. First noticed for EcoKI by Yuan et al (73), and more recently for RecB and EcoR124 (69,89), large protein motions may be a general feature of SNF2 proteins. In line with such large-scale domain movement, mutational analysis of EcoR124 showed long-range effects, e.g. nuclease mutants affect the distant helicase domain leading to a reduced translocation and ATP usage rate, a decrease in the off rate, slower restart and turnover. In other words, the nuclease and motor domain together are ‘more than the sum of their parts’ (89,90).

Plasticity of type I DNA sequence recognition: hybrid specificities and phase variation

Type I enzymes recognize bipartite DNA sequences [e.g. AAC(N6)GTGC for EcoKI]. The S subunit has a duplicated organization: two ∼150 aa variable regions alternate with smaller conserved regions, which are highly similar within each of the five families. Each variable region recognizes one part of the bipartite target sequence.

A key event in understanding the significance and mechanism of variation of sequence specificity was the discovery of a brand-new specificity resulting from a genetic cross (91,92). As a result of crossing-over in the conserved central region between the two variable regions, hybrid specificities were found. This change in specificity was found to occur in vivo and in vitro and was first noted for Salmonella species (91–94). An extensive treatment of this topic is found in the accompanying Type I review.

A variety of genetic processes promote variation in S subunits proteins of Type I and Type III enzymes. In Lactococcus lactis, entering plasmids may bring hsdS genes with them (95). The segmented organization of hsdS, with two DNA recognition domains, lends itself to variation by DNA rearrangement. Site-specific recombination leads to expression of S proteins with alternative recognition domains in Mycoplasma pulmonis, thus generating combinatorial variations of recognition sequence (96). Such plasticity of restriction specificity is also inferred in Bacteroides fragilis (97) and other species.

FURTHER HIGHLIGHTS IN THE STUDY OF TYPE II R-M SYSTEMS

Subdivisions of type II enzymes

Type II REases are defined rather broadly as enzymes that cleave DNA at a fixed position with respect to their recognition sequence, and produce distinct DNA-fragment banding patterns during gel electrophoresis. These REases are extremely varied and occur in many structural forms. The Type II classification was used originally to define the simplest kind of REases, exemplified by HindII and EcoRI, that recognize symmetric DNA sequences and require Mg²⁺ ions for cleavage activity (25). Enzymes of this sort generally act as homodimers and cleave DNA within their recognition sequences. In vivo, they function in conjunction with a separate modification MTase that acts independently as a monomer (Table 1). The first distinction made among Type II enzymes concerned REases such as HphI and FokI that recognize asymmetric sequences and cleave a short distance away, to one side. These were designated Type IIS (98).

As the number of REases producing distinct fragments grew, it became clear that many unrelated proteins were included in the category (99). Rather than dividing these into further Types based on their phylogenies, it was agreed that ‘Type II’ should become a utilitarian classification that reflected enzymatic behaviour rather than evolutionary relatedness, and for convenience, a number of Type II groups, corresponding to particular enzymatic behaviours, were defined (49) (Table 2). Each of these groups, A, B, C, E, F, G, H, M, P, S and T, should be thought of, not as an exclusive subdivision, but rather as an icon that signifies some specific property. Enzymes may exhibit more than one salient property and thus belong to more than one group. HindIII and EcoRI remain simple; they are members of just the one, Type IIP, group (‘P’ for Palindromic). BcgI, in contrast, is complicated since it recognizes an asymmetric DNA sequence (= Type IIA); cleaves on both sides of that sequence (= Type IIB); and comprises a fused endonuclease-methyltransferase subunit (= Type IIC) plus a Type I-like DNA-specificity subunit (= Type IIH). BcgI, thus, is a member of multiple groups (100–102). DpnI (Gm6A’TC) is a Type IIM REase, which cleaves its recognition sequence only when the sequence is methylated (103). DpnI is also a member of the Type IIP group since its recognition sequence is palindromic sequence and cleavage is internal and symmetric (Figure 3 and http://rebase.neb.com).

Figure 3.

Type II restriction enzymes grouped by cleavage properties. ‘Orthodox’IIP enzymes (e.g. EcoRI, EcoRV) cut at the recognition site. Type IIS cut away from the site (e.g. FokI, BfiI). Type IIB require two recognition sites and cut on the outside (e.g. BplI). Type IIE require two recognition sites, and one of the two sites acts as allosteric effector (e.g. EcoRII). Type IIF require two sites and cut at both sites as a tetramer after bringing the two regions together by looping the DNA (e.g. SfiI). Enzymes such as BcnI act as a monomer, in contrast to most Type II REases that act as dimers. See Table 2 and text for further details.

Type IIP (‘orthodox’) REases such as EcoRI and HindIII were crucial to the development of recombinant DNA technology. Certain ‘unorthodox’ enzymes have also been widely used. Sau3AI (’GATC) is a monomeric Type IIE REase that dimerizes on the DNA, inducing DNA loops. Two recognition sites must be bound for activity; one is cleaved while the other acts as allosteric effector (104). EcoRII is somewhat similar, and many REases are now known to cleave only as dimers of dimers bound to two separate sites.

Predicting enzyme families: sequences, structures and bioinformatics

Early amino acid sequences of Type II enzymes (e.g. EcoRI, EcoRV, PvuII and BamHI) showed them to be almost completely unrelated (105–109). When crystal structures appeared (110–114), commonalities began to emerge. The motif PD-(D/E)XK was identified as a common feature (115,116). This motif also appeared in other nucleases, e.g. lambda exonuclease (117) and the Tn7 transposase protein TnsA (118). This motif is the catalytic core of a Mg²⁺-dependent nuclease.

REBASE, the Restriction Enzyme Database set up by Rich Roberts to keep track of RM specificities and indicate how to acquire the enzymes, made possible the next phase of understanding. First on paper (119,120), then via the nascent Internet by File Transfer Protocol and finally on the World Wide Web (1,121) this resource makes available a focused organized data set allowing computational analysis of sequences and structures as well as access to individual topics of interest [e.g. (99,122–129)].

Recently, Sau3A (104) and several other REases proved able to cut DNA/RNA hybrids (130). The rarity of this property (6 of 223 surveyed) suggests that any biological roles for this ability will be specialized, but the property could be used to study the ubiquitous small RNA molecules that regulate expression in all domains of life (131).

FURTHER HIGHLIGHTS IN THE STUDY OF TYPE III R-M SYSTEMS

Type III enzymes have properties that are intermediate between Types I and II (Table 1 and Figure 1). In general, Type III enzymes recognize asymmetric sequences, cleave 25–27 nucleotides away from their recognition site and use ATP and SAM as cofactors, although they do not have an absolute requirement for the latter. Particularly interesting topics include control of the phage-borne R.EcoP1I REase activity following infection and how newly replicated DNA can be protected when only one strand of the recognition sequence is protected by methylation (132–139).

An early result showed that two copies of the target site were required for DNA cleavage but that these sites had to be in a head-to-head orientation (135,140). A head-to-tail orientation prevented cleavage. How this communication between the two target sites was achieved when ATP hydrolysis was insufficient for DNA translocation like the Type I enzymes (59) has provoked much discussion (141). It appears that DNA looping may have a role in bringing the sites together (142,143), but recent single-molecule analyses (144,145) show strong evidence for enzyme diffusion along the DNA triggered by an ATP-dependent conformational change as a novel mechanism for bringing two copies of the enzyme together to give cleavage, see also (83,146). The long-awaited atomic structure of a Type III R-M enzyme should resolve many of the complexities of these enzymes [AK Aggarwal, personal communication (147,148)].

FURTHER HIGHLIGHTS IN THE STUDY OF TYPE IV RESTRICTION SYSTEMS

Modification-dependent restriction was first observed with populations of phage T4 that contained hydroxymethylcytosine (hm5C)-substituted DNA (13), reviewed in (149,150). This original discovery relied on the fortuitous use of Shigella dysenteria SH as permissive host: it lacks both of the E. coli K-12 hm5C-targeted endonucleases and also the donor for the protective modification, glucosylation. This allowed glucoseless phage to be propagated in Shigella, while picking apart the E. coli K-12 set of restricting and modifying genes.

Key advances in the early years lay in determining the nature of the modifications in T-even phage DNA and the genes that enable them. hm5C is incorporated into the DNA during synthesis, and then glucose residues are added in different configurations. The host provides the glucosyl donor (151,152), while the phage provides the glucosyltransferase enzymes (153–155). With these genetic tools in hand, the host genes mediating the phage restriction activity were identified (156). These were named rglA and rglB (restricts glucoseless phage) because they mediate restriction of hmC-containing phage that lack the further glucose modification.

In the 1980s, the focus switched to other modifications, particularly m5C, with efforts to clone Type II MTases and eukaryotic DNA into E. coli (157–159). The m5C-specific functions mcrA and mcrB were mapped (160) and were shown to be identical to the rglA and rglB genes (161). A third modification–dependent enzyme was found to recognize m6A as well as m5C (162). Using the genetic tools described above, glucose-specific activity was identified (163,164). Most recently, a newly described DNA modification (165) has provided new targets (166) for Type IV enzymes: phosphorothioate linkages in the phosphodiester backbone.

The utility of all these discoveries was, at first, the ability to avoid them (167–169): these restriction systems were found to underlie difficulties encountered in the introduction of foreign MTases into E. coli (157,158,170). On the positive side, use of Type IV restriction in vivo also allowed enrichment of clone libraries for active eukaryotic genomic sequence, since much transcriptionally silenced DNA is heavily methylated [e.g. (171)].

Type IV enzymes have aroused considerable interest in recent years following the rediscovery of hm5C in the DNA of higher eukaryotes (172–175). This finding could portend the discovery of further, as yet unknown or neglected, DNA modifications. The ability of Type IV enzymes to distinguish between C, m5C, hm5C and other molecular variations of cytosine implicates these enzymes as useful tools for studies of epigenetic phenomena; the commercially available enzyme McrBC has been used for the study of such modification patterns (176,177).

Much history may remain to be written. The accompanying review focuses on structural and enzymatic properties of the systems that are known, and sketches some of the evolutionary pressures faced by restriction systems as they compete with each other and with invading replicons.

CONTROL OF RESTRICTION

Double-stranded cleavage of cellular DNA is extremely deleterious to the host cell, even when it can be repaired. Early in the study of restriction systems, the ease of moving systems among strains with differing systems by conjugation or transduction was noted. This suggested that regulation must be present to enable exchange of activities. More recently, the sporadic distribution of R-M systems in genomes of closely related strains strongly suggests that acquisition of a new system is a relatively frequent event in nature as well. Thus, coordination of expression or activity of the R-M activities is a key research topic. Transcriptional or translational control of Type I systems has not been documented, despite efforts to find it (8,178). However post-translational control is exerted at several levels and is described in the accompanying review on Type I R-M systems. The control of Type II R-M systems recapitulates the mechanisms for other regulatory systems and is described here.

Transcriptional control of Type II enzyme expression

In contrast to Type I enzymes, transcriptional control has been found for Type II enzymes. Most of the Type II systems that have been examined have the problem of integrating control of the modification and restriction activities separately, since they are embodied in separate proteins. Once again, the introduction of these genes into a naïve host is of special interest.

Control of restriction of Type II enzymes: the case of EcoRI

Expression of the MTase gene and methylation of the host DNA before synthesis of the REase is the obvious solution and the so-called ‘Hungarian trick’ was the basis for the cloning of many of the first restriction enzymes (179). The lab of Ichizo Kobayashi investigated the regulation of the EcoRI gene, ecoRIR (180–182). This gene is upstream of the modification gene, ecoRIM. The M gene has its own promoters embedded in ecoRIR and no transcription terminator between the genes, so ecoRIM can be transcribed with and without ecoRIR. Using primer extension to locate the start sites and gene fusions to assess expression, two adjacent promoters for ecoRIM as well as two reverse promoters were found within ecoRIR. These convergent promoters negatively affect each other [as in lambda (183)]. Transcription from the reverse promoter is terminated by the forward promoters and generates a small antisense RNA. The presence of the antisense RNA gene in trans reduced lethality mediated by cleavage of under-methylated chromosomes after loss of the EcoRI plasmid (post-segregational killing) (182,184).

Dual transcription control by C proteins

The Blumenthal laboratory provided the first evidence for temporal control in the plasmid-based PvuII system of Proteus vulgaris (108,185). A similar open reading frame with similar function was also found contemporaneously in the BamHI system (186,187). In the PvuII system, the MTase is expressed without delay from an independent promoter and protects the host DNA. The REase gene is in an operon with that for an autogenous activator/repressor protein, C.PvuII. Low basal expression from the pvuIIC promoter leads to accumulation of the activator, thereby boosting transcription of the C and REase genes (108,185) (Figure 4).

Figure 4.

Intricate control of restriction in the operons of the Type II R-M systems of PvuII and Esp1396I by controlling C proteins. A small C gene upstream of, and partially overlapping with, R is coexpressed from p_res1, located within the M gene, at low level with R after entry of the self-transmissible PvuII plasmid into a new host, while M is expressed at normal levels from its own two promoters p_mod1 and p_mod2 located within the C gene. A similar C protein operates in Esp1396I, but in this case the genes are convergently transcribed with transcription terminator structures in between, and M is expressed from a promoter under negative control of operator O_R, when engaged by C protein in a manner similar to that of the PvuII system. Briefly, the C protein binds to two palindromic sequences (C boxes) defining operator sites O_R and O_L upstream of the C and R genes. After initial low-level expression of C.PvuII protein from the weak promoter p_res1, positive feedback by high-affinity binding of a C protein dimer to the distal O_L site later stimulates expression from the second promoter p_res, resulting in a leaderless transcript and more C and R protein. The proximal site O_R is a much weaker binding site, but C protein bound at O_L enhances the affinity of O_R for C protein, and at high levels of C protein, the protein-O_R complex downregulates expression of C and R. In this way, C protein is both an activator and negative regulator of its own transcription. In addition, it is a negative regulator of M, which makes sense as overmethylation of DNA may also be harmful to the cell (see text for further details). C.Esp1396I controls O_R, O_L and O_M in a similar manner as described above. In this way, C proteins keep both R-M under control, and have been tentatively identified in >300 R-M systems (Table 2).

The C protein binds to palindromic DNA sequences (C boxes) defining two sites upstream of its gene: O_L, associated with activation, and O_R, associated with repression. The C protein activates expression of its own gene as well as that of the REase (188). The regulation is similar to gene control in phage lambda: differential binding affinities for the promoters in turn depend on differential DNA sequence and dual symmetry recognition. C proteins belong to the helix-turn-helix family of transcriptional regulators that include the cI and cro repressor proteins of lambdoid phages.

In the wake of PvuII and BamHI, other R-M systems were discovered that were controlled by C proteins, including BglII (189), Eco72I (190), EcoRV (191), Esp396I (192) and SmaI (193). Currently, Rebase lists 19 documented C proteins, as well as 432 putatives based on sequence data (16 September 2013, http://rebase.neb.com). The organization of the genes in the system and regulatory details differ from system to system (108,185,194). There is no published evidence addressing the question of whether R-M systems as a whole evolve in concert with the C proteins. An interesting system would be one homologous to a C-regulated system but without the C gene.

Structure of C proteins

The first structures of C-proteins appeared in 2005: C.AhdI from Geoff Kneale’s laboratory (195), and C.BclI from a consortium of workers (196). The structures were solved without bound DNA, and while they confirmed the close relationship between C-proteins and helix-turn-helix DNA-binding proteins in general, they did not reveal details of the interactions between C-proteins and their C-box binding sites in DNA (195,197–204). That came 4 years later with the crystal structure of C.Esp1396I bound to DNA (205). This structure, coupled with experimental investigations, revealed the mechanics of the genetic switch and the nature of the sequence-specific and non-specific interactions with the promoters controlling the C/R and M genes (205–208). C.Esp1396I bound as a tetramer, with two dimers bound adjacently on the 35-bp operator sequence O_L + O_R (206). This cooperative binding of dimers to the DNA operator controls the switch from activation to repression of the C and R genes.

Biological consequences of transcriptional regulation

The existence of C proteins explains why it was difficult to introduce some R-M genes in E. coli. For instance, the BamHI system of Bacillus amyloliquefaciens could only be maintained in E. coli when the REase and MTase were present on one plasmid with an additional copy of the MTase on a second plasmid (209). Further analysis suggested that in Bacillus subtilis, a host more closely related to the original expression of R-M was even more stringently regulated (109). C.BamHI enhanced activity of the REase 100-fold in E. coli, but at least 1000-fold in B. subtilis. In E. coli, the C protein repressed expression of the MTase 15-fold. The B. subtilis vegetative RNA polymerase is known be more stringent in its promoter sequence requirements than that of E. coli (210), possibly accounting for the difference in behaviour in the two species.

Crosstalk among the C genes of similar specificity can allow exclusion with R-M systems of different sequence specificities because of the premature activation of the R gene. The pvuIIC and bamHIC genes define one incompatibility group of exclusion, whereas ecoRVC defines another (211). Entry of a second R-M system thus becomes lethal, a phenomenon called ‘apoptotic mutual exclusion’ (211).

THE IMPACT OF RESTRICTION ENZYMES

The technical ingenuity applied to the use of restriction enzymes warrants a separate detailed Survey and Summary or indeed an entire book. For instance, their use led to the production of insulin from recombinant bacteria and yeast by Genentech, thus greatly increasing the supply for diabetics and the production of a recombinant vaccine for Hepatitis B by Biogen to treat the hundreds of millions of people at risk of infection by this virus. More recently they have been redesigned to create artifical nucleases, the Zinc-finger nucleases and the TAL-effector nucleases, which have potential for gene targeting and gene therapy (212). Here, we limit ourselves to a few other examples with significant scientific or public impact.

Genetic engineering

Type II enzymes yielded many practical benefits, as E. coli K12, its genes and its vectors became the workhorses of molecular biology in the 1970s for cloning, generation of libraries, DNA sequencing, detection and overproduction of enzymes, hormones, etc [e.g. (45,213–224)]. The applications of Type II enzymes continued to expand, especially after the arrival of synthetic DNA, in vitro packaging of DNA in phage particles and improved bacterial hosts and vectors for overexpression and stabilization of proteins [see e.g. (225–232)].

A historical perspective on the above topics is beyond the scope of this review. However, a couple of vignettes illustrate how use of REases enabled the research community to leverage a store of understanding to create tools for new advances.

The lacZ gene, which had EcoRI sites suitable for early vectors, and its encoded enzyme, beta-galactosidase, had a long history of investigation. Its utility in the creation of cloning vectors relied on identification of domains within the encoded protein, namely a large catalytic domain and a small multimerization domain. This was discovered by the Muller-Hill group in 1974 (233,234). The 25 N-terminal residues of the small domain can be replaced by peptides of any size and origin without destroying the ability of the multimerization domain to interact with the catalytic domain (233,234). As a result, vectors with short stretches of DNA carrying multiple restriction sites could be created (235–237). Cloning into these sites interrupted the translation of the small domain, destroying its ability to interact with the separately expressed large one. This made possible rapid screening of bacterial colonies on an agar plate for those lacking the activity of LacZ using a colour assay.

In addition, vectors carrying the intact gene but with multiple cloning sites allowed EcoRI-based DNA constructs for transcriptional and translational fusions to the lacZ gene (238–250). The majority (90%) of such LacZ-fusion proteins are stable, allowing purification of chimeric antigens, as well as detection of positive clones with colour assays (238,251). Mutagenesis studies in the laboratory of Jeffrey Miller used the lacZ gene in phage f1, allowing the rapid detection of spontaneous or induced base substitutions and frameshifts (252–254). This resulted in e.g. LacZ-transgenic mice for studies on DNA damage in different organs and tissues in mammalian cells (255,256).

DNA fingerprinting

Restriction enzymes are tools for monitoring Restriction Fragment Length Polymorphisms, allowing the location of mutations, generation of human linkage maps, identification of disease genes (such as sickle cell trait or Huntington disease), and last, but not least, the DNA fingerprinting technique developed by Alec Jeffreys (257–267). DNA fingerprinting (268) allows the solution of paternity cases, the identification of criminals and their victims and the exoneration of the falsely accused. The use of REases in this system enabled the creation of suitable procedures for such identification, although PCR has largely displaced REases in this application.

REases have also proven useful for identifying pathogenic bacterial strains, most recently of S. aureus sp with antibiotic-resistance and virulence factors mediated by mobile genetic elements, e.g. the methicillin-resistant S. aureus (MRSA) bacteria (269). Such strains pose a great threat to humans and animals (270).

FINAL THOUGHTS

In 1977, Werner Arber proposed that REases might have additional functions in the cell (271), and this is an idea to keep in mind given that much of the study of restriction enzymes has been aimed at creating tools rather than a basic study of their behaviour in their natural hosts.

For example, the actions of translocating enzymes such as the Type I and IV enzymes at a replication fork or other variant structure are one such possibility (272,273). This activity may seem of arcane interest, but a broader understanding especially of the translocating enzymes could further understanding of genome stabilization activities in all domains of life. Applications to genome manipulation or medicine could emerge. Action at aberrant structures is a major topic of interest in medicine (274).

Lastly, it is interesting to speculate on the condition of molecular biology and all of its associated sciences at the present day if the simple experiment of spreading bacteria and phage on agar plates to follow the restriction-modification phenomenon (13–15) had not been pursued. It is clear that a multi-billion dollar biotechnology industry would not have been spawned, medical diagnostics and the treatment of many diseases would have been severely retarded, genomics and genome sequencing projects would have been difficult if not impossible and their support of bioinformatics and evolutionary studies would also not have been possible, thus greatly diminishing our current appreciation of the spectacular diversity of life on earth.

FUNDING

Funding for open access charge: New England Biolabs.

Conflict of interest statement. None declared.

ACKNOWLEDGEMENTS

We thank the many colleagues who generously provided information for this review. Attendees of the 1972 EMBO Workshop helped identify the people in Figure 2. Special thanks to Werner Arber, Joe Bertani and Rich Roberts for historical material, reprints from their archives and helpful emails over the past decade.

Author notes

^†Noreen Murray passed away during the early stages of preparing this review, which is fondly dedicated to her memory.

The authors wish it to be known that, in their opinion, the first three authors should be regarded as Joint First Authors.