The FokI endonuclease is a monomeric protein with discrete DNA-recognition and catalytic domains. The latter has only one active site so, to cut both strands, the catalytic domains from two monomers associate to form a dimer. The dimer involving a monomer at the recognition site and another from free solution is less stable than that from two proteins tethered to the same DNA. FokI thus cleaves DNA with two sites better than one-site DNA. The two sites can be immediately adjacent, but they can alternatively be many hundreds of base pairs apart, in either inverted or repeated orientations. The catalytic domain of FokI is often a component of zinc finger nucleases. Typically, the zinc finger domains of two such nucleases are designed to recognize two neighbouring DNA sequences, with the objective of cutting the DNA exclusively between the target sequences. However, this strategy fails to take account of the fact that the catalytic domains of FokI can dimerize across distant sites or even at a solitary site. Additional copies of either target sequence elsewhere in the chromosome must elicit off-target cleavages.
- DNA looping
- gene therapy
- protein–protein interaction
- recognition sequence
- restriction enzyme
Symmetrical recognition sites
Type II restriction endonucleases recognize specific DNA sequences, typically 4–8 bp long, and cleave both stands at fixed positions relative to the sequence, in reactions that usually require Mg2+ ions as a cofactor . The recognition site is often a palindrome, with the same 5′–3′ sequence in both strands, and the enzymes that act at such sites, now known as Type IIP systems , are commonly dimers or tetramers of identical subunits [3,4]. The dimers interact symmetrically with their palindromic targets, with each subunit making identical contacts with half of the duplex. They also cleave the DNA symmetrically: both strands are cut at the same phosphodiester bond within the 5′–3′ sequence. The active site in one subunit hydrolyses the scissile linkage in one strand, while the other active site acts likewise on the complementary strand. Many well-known restriction enzymes fall into this category: viz., BamHI, EcoRI, EcoRV and PvuII. However, the palindromic site need not be a continuous sequence: BglI, for example, recognizes an interrupted palindrome, GCCnnnn↓nCGG (n denotes any base and ↓ the point of cleavage), but acts symmetrically at this site in terms of both its contacts to the specified elements and its cleavage reactions [5,6].
The tetrameric restriction enzymes that act at palindromic sites normally employ two subunits to bind to and cut both strands at one copy of the sequence, while the other two subunits bind and cut a second copy . The dimeric units boundto each copy resemble a dimeric Type IIP enzyme bound to an individual site. However, a striking feature of the tetrameric enzymes is that, when only one dimeric unit is bound to cognate DNA, it is virtually inactive on that DNA. Instead, the tetramer becomes fully active only after binding two copies of the specific sequence, one to each of the dimeric units . They then cut in four parallel reactions both strands at both sites . The restriction enzymes that bind and cleave two target sites are called Type IIF enzymes . Another set of Type II enzymes, the Type IIE systems, also bind two DNA sites, but these cleave only one of the bound segments; the other binds to a regulatory domain in the protein, which enables the catalytic functions .
Proteins that bind two DNA sites prefer sites in cis, in the same molecule of DNA, over sites in trans, in separate molecules, simply because the distance between loci is limited to the length of the intervening DNA for sites in cis but has no upper limit for sites in trans . In the former case, the protein bound to both sites holds the intervening DNA in a loop while, in the latter, it has to hold together the two DNA molecules. The restriction enzymes that need two cognate sites for full activity thus cleave two-site substrates more rapidly than DNA with a single site . The Type IIF enzymes can cleave a two-site substrate directly to give the final product cut at both sites, while the Type IIE enzymes first cut one of the two sites at a high rate and then the residual site at the same low rate as one-site DNA.
Asymmetrical recognition sites
Not all Type II restriction enzymes recognize palindromic sequences. Many recognize instead asymmetric non-palindromic sites and, while these still cleave both DNA strands at fixed positions relative to the site, they typically cleave the DNA asymmetrically, at non-identical positions in the two strands, often several base pairs away from the site . The enzymes that act in this manner are called Type IIS enzymes , where S indicates that they cleave DNA at loci shifted away from one side of the recognition site. For example, the Type IIS restriction enzyme, FokI, recognizes the following sequence: and cleaves each strand at the position indicated: 9 nt downstream of the site in the ‘top’ strand; 13 nt away in the ‘bottom’ strand.
While a palindromic sequence can be recognized by the two subunits of homodimeric enzyme each contacting one-half of the duplex, the same mechanism cannot be applied to a non-palindromic site. Instead, an asymmetric sequence must be identified by either a monomeric protein in touch with the complete sequence or by a heterodimeric protein (or larger oligomer) composed of two (or more) different subunits that each contact part of the sequence. While each subunit of a homodimeric or tetrameric enzyme carries an active site positioned to cleave one strand of the DNA [3,4], a monomeric protein may possess only one active site and, if so, would be expected to cleave only one strand. Many Type IIS nucleases, including FokI [13,14], are monomers in solution , yet almost all of them cut both strands . Indeed, most of these monomeric enzymes act not only on both strands but also at two separate sites in the DNA . On the other hand, a heterodimeric enzyme has the potential to cut both strands, as each subunit could carry a strand-specific active site. Although not a Type IIS enzyme, the BbvCI endonuclease illustrates exactly this behaviour [16,17].
The Type IIS restriction enzymes have evolved various strategies to cut both DNA strands at an asymmetric site . In some instances, such as Mva1269I and BtsCI, the protein exists as a monomer in solution but with two catalytic centres within the one polypeptide: mutagenesis studies revealed that each centre acts on one particular strand of the DNA [19,20]. Enzymes that operate in this manner were often identified initially from the observation that, unlike most Type IIS systems , they cleaved one- and two-site substrates at the same rate. Another strategy is represented by BfiI: a homodimeric protein with a DNA-recognition domain in each subunit, both of which must bind the cognate sequence before catalysis can occur, but BfiI has only one active site, at the subunit interface. The active site is used first to cleave the bottom strand at one site but the same active site is subsequently employed to cut the top strand . The switch between strands requires not only a 180° rotation around the helical axis of the DNA but also a 180° rotation perpendicular to the DNA axis, to address the opposite polarities of the two strands . The heterodimer scheme, noted above for BbvCI, also applies to some Type IIS enzymes . For example, BspD6I consists of a large subunit that recognizes the target sequence and cuts the top strand 4 nt downstream of the site, and a small subunit that, when bound to the large subunit, cleaves the bottom strand 6 nt away from the site .
Action at two asymmetrical sites
The most-studied Type IIS enzyme is the FokI endonuclease. It is generally considered the archetype of this group, although how representative it is remains uncertain: MboII is one of very few that has actually been shown to function like FokI . Nevertheless, three schemes were noted above for cutting both strands at an asymmetric DNA sequence: a monomeric protein with two active sites; a single active site switching between strands; a heterodimeric enzyme with a strand-specific active site in each subunit. All three of these ought in principle to be fully capable of acting at a solitary site on DNA (although the single active site in BfiI becomes accessible only after specific DNA is bound to both of its regulatory domains ). Yet, most Type IIS endonucleases tested to date cleave DNA with two or more target sites more readily than DNA with a single site [15,18,24–26]. The mode of action of FokI (Figure 1) accounts for how a monomeric enzyme can be more active at two copies of an asymmetric sequence than at a single site, and so this scheme may be applicable to many Type IIS enzymes.
By both gel filtration and analytical ultracentrifugation, the FokI restriction enzyme is shown to be a monomer in solution and when bound to DNA in the absence of divalent metal ions [13,14]. The monomer features two discrete domains, as shown by limited proteolysis  and then by X-ray crystallography : an N-terminal DNA-recognition domain that makes all of the contacts with the target sequence; and a C-terminal catalytic domain with one active site that cleaves DNA non-specifically. To cut both strands at its recognition sequence, the catalytic domains from two separate monomers associate to form a dimer with two active sites [14,29–31]. The catalytic domain of the protein bound to the recognition sequence engages the target phosphodiester bond 13 nt away in the bottom strand but by itself has no activity; it cannot even nick the DNA . However, it becomes active on forming a dimer with a catalytic domain from a second monomer of the FokI protein. The latter engages the scissile bond 9 nt away in the top strand and the dimeric unit then proceeds to cut both strands.
The second catalytic domain can come from a monomer of FokI in free solution (Figure 1a). However, the dimerization interface between catalytic domains covers only a small surface area , with the result that the dimer formed between one DNA bound and one free monomer of FokI is unstable : it has an equilibrium dissociation constant of >100 nM, far higher than the concentration of enzyme normally used for its reactions in vitro . Hence, on a DNA with one cognate site for FokI, only a small fraction of the enzyme bound to that site will be in the dimeric state at any given time, with the result that the one-site DNA is cleaved at a low rate : the actual rate will be the intrinsic rate for DNA cleavage by FokI divided by the ratio of monomer to dimer, and the latter will be approx. 100:1 at 1 nM protein.
On a DNA with two or more recognition sites for FokI (Figure 1b), the dimer can alternatively be formed from two monomers of the FokI protein, each bound to a specific site in the same DNA molecule [31,33]. The two monomers bound in cis are held physically close to each other; they cannot be further apart than the contour length of the intervening DNA, while a monomer in free solution can be any distance away from the DNA-bound monomer . Consequently, the FokI protein at its recognition site associates more readily with a second FokI bound elsewhere in the same chain of DNA than to a monomer free in solution. As a result, FokI cleaves DNA substrates with two copies of its recognition sequence more rapidly than DNA with one copy [15,31]. The two-site substrate is, however, cleaved rapidly in both strands at just one site: the residual site left after the first scission is cleaved at the low rate characteristic of the reaction of FokI on a one-site substrate [15,31]. In this respect, FokI acts like a Type IIE restriction enzyme, the sort that uses one DNA site to activate cleavage at another [2,9,11]. There seems to be no systematic selection for which of the two sites is cleaved first. The catalytic domains from the two DNA-bound monomers can thus associate at either site, with essentially equal probability.
It has been suggested that the catalytic functions of FokI require a pair of recognition sequences in close proximity to each other . Indeed, two FokI sites in head-to-head orientation separated by 21 bp possess a single set of cleavage points and might thus allow for the direct interaction of the catalytic domains from the monomers at each site, along the DNA (Figure 2a). However, the enhanced activity of FokI at two recognition sites in cis does not require direct interactions between immediately adjacent proteins in a particular orientation. Instead, the two FokI sites can be separated by hundreds rather than tens of base pairs, and can be present in either head-to-head, head-to-tail or tail-to-tail orientations [15,31]. When bridging distant sites, FokI traps the intervening DNA in a loop [31,33,35]. In the characteristic manner for a DNA–looping interaction , the stability of the synaptic complex spanning two FokI sites varies cyclically as the length of DNA between the sites is varied from 170 to 200 bp, with a periodicity that matches the helical repeat of DNA . The stability of the synaptic complex also depends on the relative orientation of the two sites. Nevertheless, these variations in synapse stability do not affect the initial reaction rate of FokI on two-site substrates.
ZFNs (zinc-finger nucleases)
The recognition sites for Type II restriction endonucleases are generally 4–8 bp long . Such sites occur too often in genomic DNA for any individual gene to carry a unique restriction site, a site that is not found elsewhere in the genome. The ability to cleave chromosomal DNA within a gene of interest is, however, a desirable goal [36,37]. It can lead to the inactivation of that gene, owing to damage caused by non-homologous end-joining [38,39]. It can also promote recombination at that site with a homologous DNA provided in trans. The latter is potentially a strategy for gene therapy, the replacement of a mutant gene with a correct copy [36,37]. On statistical grounds, a unique site in the human genome is likely to be ≥16 bp long. Endonucleases that cleave DNA specifically at sequences longer than any restriction site have been constructed from the FokI enzyme. The modular architecture of this protein has allowed for the replacement of its DNA-recognition domain that contacts a 5-bp sequence, with the specificity domain from other DNA-binding proteins, most commonly from zinc finger proteins . The resultant proteins are called ZFNs.
The zinc finger domains fused to the catalytic domain of FokI usually contain three adjacent fingers that each contact three consecutive base pairs, thus resulting in a 9-bp recognition sequence [36,41]. Since the catalytic domain of FokI is only active as a dimer, the ZFN requires two such sequences. If the two sites have to be in close proximity and in head-to-head orientation, as has been suggested , then they constitute an 18-bp target sequence, which could potentially be unique in the human genome (Figure 2b). Moreover, the two sites within the pair need not have the same sequence. A particular advantage of zinc fingers is that they can be tailored to recognize a sequence of choice . Consequently, by utilizing a ZFN that recognizes one particular 9 bp sequence in conjunction with a second that had been tailored to the adjacent (inverted) sequence, it is theoretically possible to develop a system that is capable of cleaving any 18 bp sequence that one wishes to cut. In a system of this type, the two ZFNs can be forced to form the heterodimer, as opposed to either forming a homodimer, by the appropriate mutations at the dimerization interface of FokI .
Pairs of ZFNs designed to act at two adjacent sequences have been widely advocated as potential tools for targeting individual genes and for gene therapy [43–45]. Nevertheless, while ZFNs constructed along these principles can act efficiently at the sites they were designed to cleave, they almost always cause substantial levels of additional ‘off-site’ cleavages [36,45]. But because the catalytic domains of FokI do not require their associated recognition sites to be near each other along the DNA, or to be present in any particular orientation, a pair of ZFNs tailored to recognize two adjacent 9 bp sequences is absolutely bound to give a large number of off-site cleavages. If either of the two 9 bp sequences occurs elsewhere in the chromosome, the catalytic domains of the ZFNs will be able to span a pair of distant rather than adjacent sites (Figure 2c). Moreover, the long-range interactions of FokI are essentially uniform across a wide range of inter-site spacings, and, unlike short-range interactions, independent of their relative orientation. In addition, a solitary copy of either 9 bp sequence will elicit significant levels of DNA cleavage by the designed ZFN. Even though FokI requires two sites for full activity, its activity at an isolated site is still readily detected.
It thus seems unlikely that a nuclease can ever be constructed from the catalytic domain of FokI that cleaves DNA specifically at a unique sequence within a eukaryotic genome. The off-site cleavages that are unavoidable with an endonuclease of this type may be minor concerns in many situations, such as in the manipulation of lower organisms, like Caenorhabditis elegans or Drosophila, or mammalian cell lines. However, the inbuilt errors of ZFNs may well limit their applications as tools for gene therapy in clinical medicine.
The work in this laboratory was funded by the Wellcome Trust [grant number 074498].
NACON VIII: 8th International Meeting on Recognition Studies in Nucleic Acids: An Independent Meeting held at The Edge, University of Sheffield, Sheffield, U.K., 12–16 September 2010. Organized by Mike Blackburn, Mark Dickman, Jane Grasby, David Hornby, Chris Hunter, John Rafferty, Jim Thomas, David Williams and Nick Williams (Sheffield, U.K.).
Abbreviations: ZFN, zinc-finger nuclease
- © The Authors Journal compilation © 2011 Biochemical Society