Two-step site selection for serine-integrase-mediated excision: DNA-directed integrase conformation and central dinucleotide proofreading

Abstract

Bacteriophage-encoded serine-integrases are members of the large family of serine-recombinases and catalyze site-specific integrative recombination between a phage attP site and a bacterial attB site to form an integrated prophage. Prophage excision involves a second site-specific recombination event, in which the sites generated by integration, attL and attR, are used as substrates to regenerate attP and attB. Excision is catalyzed by integrase but also requires a phage-encoded recombination directionality factor (RDF). The Bxb1 recombination sites, attP and attB, are small (<50 bp), different in sequence, and quasisymmetrical, and they give rise to attL- and attR-recombinant products that are asymmetric but similar to each other, each being composed of B- and P-type half-sites. We show here that the determination of correct excision products is a two-step process, with a presynaptic RDF-dependent step that aligns attL and attR in the correct orientation and a postsynaptic step in which the nonpalindromic central dinucleotide confers identity to attL and attR and prevents each from recombining with itself.

Temperate bacteriophages typically integrate their genomes into the host chromosome during the establishment of lysogeny and can excise for subsequent lytic growth (1). Integration and excision usually involve site-specific recombination reactions, in which specific sites on the phage (attP) and bacterial chromosomes (attB) are used as substrates for integrative recombination, and the product sites (attL and attR) act as substrates for excision. A phage-encoded integrase protein mediates catalysis of both integrative and excisive recombination reactions, and directionality is regulated by the availability of the attB, attP, attL, and attR sites, together with a recombination directionality factor (RDF), an accessory protein that determines which site pairs can productively engage in recombination (1, 2).

A subset of temperate phages use an integrase of the serine-recombinase family to catalyze integration and excision (3). Several systems have been studied, including those encoded by phages φC31 (3–5), Bxb1 (6–10), φRv1 (11, 12), A118 (13), and TP901-1 (14, 15). The enzymes have N-terminal domains homologous to the catalytic domains of transposon resolvases and DNA invertases. They use similar mechanisms for recombination, involving concerted double-strand DNA cleavages at the two recombination sites, rotation of the two helices 180° about a common axis, and religation in the recombinant configuration. Cleavage of the DNA occurs about a central dinucleotide to generate two-base 3′ extensions, and the integrase is covalently bound through a phosphoserine linkage (6).

In each of the serine-integrase systems reconstituted in vitro (4, 9, 12, 16), there is no requirement for host factors or DNA supercoiling, and the substrate sites for integrative recombination, attP and attB, are small (typically <50 bp) (3). The Bxb1 attP and attB sites differ in both sequence and size (6); each is composed of two symmetrically related half-sites (P, P′ and B, B′) surrounding a central 5′-GT dinucleotide about which strand cleavage occurs (Fig. 1A). In the Bxb1 system, it has been shown that gpInt binds to each substrate DNA as a dimer, with DNA-binding specificity conferred by the 350-residue C-terminal domain (7). Strand cleavage, and thus also synapsis, occurs with attB and attP aligned in both parallel and antiparallel orientations, suggesting that the DNA sequence symmetry within attP and attB is reflected in the overall symmetry of the gpInt–DNA complexes (7, 8). However, stand cleavage, and perhaps synapsis, is specific for attP and attB, which, in the absence of other DNA site or protein factors, is most likely determined by different conformations of gpInt when it is bound at different DNA sites (7, 8).

Fig. 1.

DNA site requirements for excisive recombination. (A) Schematic representation of Bxb1 integration and excision in which attP and attB are composed of P-type (P, P′) and B-type (B, B′) half-sites flanking a nonpalindromic central dinucleotide (5′-GT). (B and C) In vitro reactions were performed with an scDNA substrate containing either attL (B) or attR (C) and short (60-bp) linear partner DNAs as indicated (R, attR; L, attL; P, attP; B, attB). Reaction products were electrophoresed directly (Upper) or after digestion with Xmn1 (Lower). Positions of supercoiled substrates (scDNA sub), cleaved substrate (sub), linear products (Lin prod), and cleaved products (prod) are shown.

In vitro excision reactions have been established in the Bxb1 and φRv1 systems (8, 11), and both require a phage-encoded RDF in addition to gpInt. Although the φRv1 and TP901–1 RDFs have sequence similarity to some Xis proteins that function with tyrosine-integrases (11, 17), the Bxb1 RDF (gp47) is a protein unrelated to other RDFs (8). Bxb1 gp47 does not bind to DNA, and although the φRv1 RDF binds DNA weakly, DNA-binding activity is not required for excisive recombination (8). Thus, these RDFs may function to determine alternate synaptic determinants by altering gpInt–DNA conformations, such as to promote attL × attR synapsis while disturbing attP × attB synapsis.

The quasisymmetry of gpInt–attP and gpInt–attB complexes raises a question as to what prevents them from giving rise to products with incorrect polarities (i.e., joining B and B′ and P and P′). In both the Bxb1 and φC31 systems, this decision is made postsynaptically and postchemically through the use of nonpalindromic central dinucleotides. Thus, although incorrect site alignment leads to synapsis, strand cleavage, and rotation, the mismatched central dinucleotide prevents religation of incorrect half-sites (5, 6). This role of the central dinucleotide is illustrated by topological relaxation of a supercoiled DNA (scDNA) containing an attP site in the presence of a short linear attB DNA containing a mutant central dinucleotide, in which each loss of a DNA supercoil results from two 180°C rotations and religation in the substrate configuration (6).

The attL and attR sites are substrates for the excision reaction, and each is composed of one B-type half-site and one P-type half-site. Given the requirements for the integration and excision reactions, two key questions arise: (i) Does DNA sequence asymmetry lead to asymmetry of dimer gpInt–attL and gpInt–attR complexes determining a unique alignment at synapsis; and (ii) what distinguishes between attL and attR complexes to ensure attL × attR recombination, rather than attL × attL and attR × attR reactions? We show here that synapsis for excision is indeed alignment-specific, and that the identity of attL and attR is solely determined by their nonpalindromic central dinucleotides, such that an attL mutant with a symmetrical dinucleotide is fully capable of recombination with itself, but only joins attB and attP half-sites.

Results

Excisive Recombination Occurs Only Between attL and attR.

Although the conditions for Bxb1 integrase-mediated excisive recombination have been reported previously (8), the combinations of sites supporting recombination in the presence of both Bxb1 gpInt and its RDF (gp47) have not. To determine the site preferences, recombination reactions were performed in which attL was part of an scDNA plasmid (although note that supercoiling is not required for the reaction) (8), and each of the four possible sites (attP, attB, attL, and attR) was provided as a linear DNA partner (Fig. 1B). Recombination generates a linear product containing attP and attB, whose identity is confirmed by digestion of the reaction products (Fig. 1B). In these reactions, recombinant products are formed when attR is provided, and both gpInt and gp47 are required as expected (Fig. 1B). Likewise, if the scDNA contains attR, recombinants are generated only when attL is present (Fig. 1C). However, in both sets of experiments, there is some apparent smearing of the scDNA when either no partner DNA is added or attP or attB are added, which is dependent on gpInt and gp47. Because no recombinant products are formed in these reactions (Fig. 1 B and C), this smearing likely results from topological relaxation of the scDNA substrate and/or formation of catenanes and knots. When the noncognate partner is the same as the scDNA site (i.e., attL in Fig. 1 B and attR in Fig. 1C), there is a gp47-dependent reduction of the scDNA substrate, and slower migrating bands are generated. These molecules are presumably also topological variants of the scDNA substrate because digestion shows only substrate DNA bands (Fig. 1 B and C). Thus, the formation of excision products (attP and attB) is strongly restricted to the attL × attR substrate pair, although the noncognate pairs, attL × attL and attR × attR, are not completely inert.

Synapsis Does Not Discriminate Between attL and attR.

The ability of attL × attL and attR × attR reactions to support topological changes in the scDNA substrate suggests that these site pairs can undergo synapsis and initiate strand exchange. To address this issue, we analyzed the patterns of protein binding and complex formation by using an attL suicide DNA substrate, in which a centrally located nick leads to the accumulation of synaptic complexes (Fig. 2). When gpInt, gp47, and attR DNA are incubated with an attL suicide substrate, a slow-migrating band is formed (Fig. 2A); its identity as a trapped synaptic complex has been described previously (8). Interestingly, when the attR DNA is omitted (so that only the suicide attL DNA is present) (Fig. 2A), a trapped synaptic complex containing covalent protein–DNA linkages (Fig. 2C) is readily formed. Similar results were observed by using an attR suicide substrate by itself (data not shown). The simple interpretation is that there is no discrimination between attL and attR at the stage of synapsis and that all three combinations of sites (attL × attR, attL × attL, and attR × attR) undergo synapsis and strand-transfer reactions. In contrast, in similar experiments with integrative recombination conditions, only attP/attB and no other site combination forms a trapped synaptic complex (ref. 7 and data not shown).

Fig. 2.

Synapsis of attL × attR and attL × attL sites. (A and B) A 60-bp radiolabeled suicide attL (A) or normal attL (B) substrate (attL*) was incubated with Bxb1 gpInt, gp47, and a 200-bp attR partner DNA (attR₂₀₀), and the products were analyzed by native gel electrophoresis. The attL suicide substrate forms a slow-migrating complex in the absence of attR (A) and proteolytic digestion. (C) A second dimension of electrophoresis shows that this complex contains cleaved attL DNA; thus, attL synapses with itself. When a normal attL DNA substrate is used, attL and attR recombine to generate attP and attB, to which gpInt and gp47 can bind (B). The positions of the putative synaptic complexes (SC) in A and B are indicated. In the absence of attR, the primary complex is a gpInt–attL complex (cmplx I), although a small amount of a supershifted complex containing gp47 also is observed. Two bands corresponding to the cleaved attL may result from incomplete digestion by proteinase K. (D) Position of the nick in the attL suicide substrate.

The Nonpalindromic Central Dinucleotide Is the Sole Determinant of attL Versus attR Identity.

The ability of attL × attL and attR × attR to synapse and initiate strand exchange raises the question as to whether there is discrimination between the two possible alignments (i.e., parallel vs. antiparallel configurations) (Fig. 3A). If parallel alignment of the sites occurs, then the central dinucleotides would be matched and could conceivably religate in the product configuration. Although synapsis in this parallel site alignment cannot be excluded, we note that neither attL × attL nor attR × attR reactions yield recombinant products (which would correspond to B-B and P′-P′ sites or B′-B′ and P-P sites) (Fig. 2). If attL × attL or attR × attR aligns in the antiparallel alignment, then the central dinucleotides will not be matched at the religation stage (Fig. 3A). This explanation could account for the ability of attL × attL and attR × attR pairs to promote topological changes while not yielding attP/attB-recombinant products.

Fig. 3.

Role of the central dinucleotide in Bxb1 excisive recombination. (A) Schematic representation of the alignment of an attL site with itself or a mutant attL(AC) site in the parallel and antiparallel orientations. (B) Excisive recombination reactions were performed as in Fig. 1, but by using linear substrates with altered central dinucleotides, WT (5′-GT), 5′-GC, or 5′-AC as indicated. The products were separated by electrophoresis either directly (Upper) or after Xmn I digestion (Lower). attL(AC) recombines with WT attL to generate a linear recombinant product (Lin prod), whereas other mismatched pairs promote superhelical relaxation. (C) Recombination reactions (as in B) were performed with scDNA substrates containing either WT attL or an attL(GC) mutant site as indicated. Linear DNAs containing WT attL or an attL(GC) mutant were added in increasing concentrations (4.5, 22.5, 90, and 180 ng/μl). Products were analyzed by electrophoresis either directly (Upper) or after Xmn I digestion (Lower). The scDNA-attL(GC) DNA can recombine with itself, which generates a larger band after digestion.

To test whether the central dinucleotide is the only determinant preventing attL × attL and attR × attR recombination, we constructed a mutant attL site with an inverted central dinucleotide (5′-AC), which, if aligned with attL in the antiparallel configuration, could religate to form recombinant products and thus behave as an attR-like partner (Fig. 3A). This attL(AC) substrate indeed recombines well with attL DNA (Fig. 3B), showing that the central dinucleotide is sufficient to distinguish attL from attR. This finding also suggests that the sites align primarily in the antiparallel configuration. We propose that in this alignment, whenever the central dinucleotides are mismatched [including attL(WT) × attL(WT) and attR(WT) × attR(WT)], the reaction proceeds normally, except that religation in the product configuration cannot occur and an additional 180° DNA rotation occurs, leading to loss of a DNA supercoil when one site resides in scDNA (Figs. 1 and 3B). This situation is formally analogous to the partner DNA-dependent topological relaxation of an scDNA substrate in Bxb1 integrative recombination (6).

The mismatch of the central dinucleotide in an attL × attL or attR × attR reaction arises because it is nonpalindromic. Thus, we would predict that an excision substrate with a palindromic central dinucleotide would have full potential to recombine with itself, and, indeed, an attL(GC) substrate does so (Fig. 3C). The Bxb1 excision reaction can thus be reduced to a three-component process requiring just a single DNA substrate [attL(GC), gpInt, and gp47]. The nonpalindromic central dinucleotide is thus the sole determinant of attL versus attR identity.

The Central Dinucleotide Also Is the Sole Determinant of attL Versus attR Identity in φRv1 Excision.

Because all of the characterized serine-integrase systems use small recombination substrates with quasisymmetric attP and attB sites, the use of a nonpalindromic central dinucleotide should be a common feature that not only ensures integrative recombination yields recombinant products with the sites aligned in the appropriate configuration, but also distinguishes between attL and attR for excisive recombination. An in vitro excisive recombination reaction has been described for the Mycobacterium tuberculosis prophage-like element φRv1 (11), in which the central dinucleotide is 5′-TG. We have tested whether dinucleotide asymmetry plays a similarly important role to that in the Bxb1 system.

As shown in Fig. 4, the WT pair of substrates attL(TG) attR(TG) gives rise to recombinant products, and changing the attR central dinucleotide to either 5′-AG or 5′-CA prevents product formation, but promotes the topological relaxation of the scDNA substrate. However, attL(CA) recombines efficiently with WT attL(TG) DNA, illustrating that changing the central dinucleotide is necessary and sufficient to convert attL into attR (Fig. 4). Similarly, φRV1 attL can be converted into attR by simply switching the central dinucleotide to 5′CA (Fig. 4). Thus, the nonpalindromic central dinucleotide is the sole determinant of attL versus attR identity in both the Bxb1 and φRv1 systems.

Fig. 4.

Role of the central dinucleotide in φRv1 excisive recombination. In vitro recombination reactions (11) were performed by using a scDNA substrate containing either attL (Left) or attR (Right) and linear DNA substrates containing central dinucleotides as indicated. attL and attR recombine to form a linear product (lin prod), whereas sites with mismatched central dinucleotides promote topological relaxation. Inverting the central dinucleotide [i.e., attL(CA) or attR(CA)] converts attL into an attR-like site and vice versa.

Antiparallel Alignment of DNA Sites Is Required for Excisive Synapsis and Recombination.

attL and attR are asymmetric in sequence (because the B- and P-type half-sites are different), but it is unclear whether gpInt-bound complexes reflect this asymmetry; if not, they could synapse initially in both parallel and antiparallel configurations (see Fig. 5). We have used two approaches to address whether synapsis for excision is indeed orientation-selective. First, we took advantage of the ability of an attL(GC) substrate to recombine with itself (Fig. 3) to ask whether products are generated corresponding to both parallel and antiparallel synaptic alignments. When two attL(GC) substrates of different lengths are used, recombinants arising from only the antiparallel alignment are observed (Fig. 6).

Fig. 5.

A two-step model for control of serine-integrase catalyzed excision. In the first presynaptic step of excision control, conformation of gp–Int complexes prevents synapses in the parallel alignment, and we propose that gpInt protomers will only interact if one is bound to a B-type half-site and the other to a P-type half-site. However, attL and attR cannot be distinguished presynaptically. In the second step, the nonpalindromic central dinucleotides prevent attL × attL (or attR × attR) synapsis from completing recombination, providing a proofreading role in excision. Thus, an attL(AC) central dinucleotide mutant becomes functionally equivalent to attR.

Fig. 6.

Antiparallel alignment is favored for synapsis in excisive recombination. (A) Schematic representation of the experimental scheme in which two attL DNA substrates of different lengths, both with a palindromic (5′-GC) central dinucleotide, could potentially synapse in either parallel or antiparallel alignment, and the resulting products are indicated. Only the smaller (attL-75) substrate is labeled, and observed products could arise from attL-75 recombining with itself (attL-75 × attL-75) or with the longer substrate (attL-75 × attL-90). (B) Excisive recombination reactions using the substrates shown in A were analyzed by SDS gel electrophoresis. Only the products derived from antiparallel alignment of the sites are observed. (C and D) Schematic representation of the substrates used for in-gel FRET are shown in C and electrophoresis of the reactions in D. attL DNA was doubly labeled with ³²P and fluorescein (F) as shown. Suicide substrates contained a nick on either the top or bottom strand and were either unlabeled or labeled with Black Hole Quencher (C). The levels of fluorescence quenching were determined as the differences between the quencher-labeled and unlabeled suicide substrates normalized for complex formation. Two experiments were run, each with three separate reactions, and a representative set of reactions is shown in D. (C Lower) Levels of quenching calculated from the differences of the means of the three values in each of the two experiments.

Because this experiment relies on product formation, it does not rule out the possibility of nonproductive alignment in the parallel orientation. To monitor the alignment of sites directly in the synaptic complex, we used an in-gel FRET assay with fluorescently labeled DNA substrates, where fluorescence quenching should differ as a function of synaptic site alignment (Fig. 6). An attL substrate labeled with fluorescein at one end and with radioactively at another end was incubated with gpInt, gp47, and a 10-fold molar excess of a suicide substrate (substrate 1 or 2, with nicks on the bottom and top strands, respectively) (Fig. 6C), which was either unlabeled or labeled with a fluorescence quencher at one end (Fig. 6C). The formation of synaptic complexes in the presence of both gp47 and gpInt was then detected by radiography, and the in-gel fluorescence levels were normalized to the amounts of the complexes. We observed substantial quenching with suicide substrate 2 when antiparallel alignment of the sites places the fluorescein label and the quencher on the same side of the protein–DNA complex and relatively little with suicide substrate 1 when parallel alignment position them on the same side (Fig. 6). This experiment suggests that there is an orientation preference favoring an antiparallel alignment.

Discussion

The serine-integrase site-specific recombination systems differ from the tyrosine-integrase systems (such as phage λ), in that they satisfy the same biological requirements for prophage integration and excision, but with considerably less molecular information (i.e., there are no host factor proteins, and they use smaller DNA substrates with no binding sites for accessory proteins, including RDF). The serine-integrases also are distinct from their serine DNA invertases and transposon resolvases counterparts, where the sites are identical (or near identical) in sequence, and synaptic alignment is achieved through accessory-binding sites (either for the recombinase, as in the resolvase systems, or for accessory proteins, as in the invertase systems), together with the influences of DNA topology (18, 19).

The Bxb1 integration reaction is unusual in that there is no selection for the correct alignment of sites at synapsis, and DNA cleavage and strand rotation occur regardless of the alignment. The sole determinant that prevents the formation of incorrect products (i.e., the B half-site joined to the B′ half-site, and the P and P′ half-sites joined) is the nonpalindromic central dinucleotide. The central dinucleotide plays a central role in all of the serine-recombinase systems, but only in these integration systems is it the sole arbiter of product polarity. Most important, these observations show that site symmetries are fully reflected in the symmetries of the gpInt–attP and gpInt–attB complexes. However, because the B- and P-type half-sites have distinctly different sequences, the excision sites attL and attR are asymmetric with regard to their sequences. Bxb1 gpInt binds as a dimer to attL and attR, and the sequence asymmetries are apparently reflected in asymmetries of these protein complexes, conferring an orientation-selectivity in gpXis-dependent synapsis that is absent in the integration reaction.

These observations reveal a remarkable aspect of the serine-integrase reactions, in that, although synapsis is protein-directed, the DNA sequence information directly determines which synaptic events can occur. Thus, two simple rules can account for the site specificities of both integration and excision: (i) a gpInt protomer bound to a B-type half-site can only interact with a gpInt protomer bound to a P-type half-site, and (ii) synapsis requires two such interactions (Fig. 5). These rules suggest that gpInt adopts distinctly different conformations when bound to B- and P-type half-sites and that the two gpInt protomers within a bound dimer act cooperatively in synapsis.

These two rules explain synaptic selectivity, but not product selectivity. We have shown here that product selection also is determined by the central dinucleotide, which plays a critical role in both Bxb1 excision as well as integration. However, whereas the nonpalindromic central dinucleotide is the sole determinant of correct product polarity for integration, in excision it is the sole arbiter of the identities of the excision substrates attL and attR. Inverting the central dinucleotide of one site is sufficient to give it the identity of the other, and just a single attL site with a palindromic central dinucleotide efficiently recombines with itself. For both integration and excision, the central dinucleotide role is at a late stage in the reaction, preventing religation of incorrect half sites and enabling a further 180° rotation to regenerate the substrate sites (Fig. 5).

Although the serine-recombinase family of proteins may use common mechanisms of nucleophilic attack, DNA cleavage, phosphotransfer reactions, stand rotation, and religation, the serine-integrases appear to differ from other systems in that DNA–protein conformation, rather than DNA topology, plays a key role in synapsis (18, 19). However, the conformationally determined rules for synapsis do not impose as complete a set of constraints as DNA topology does, and they do not prevent incorrect alignment of attB and attP for integration or select attL × attR partners for excision. The nonpalindromic central dinucleotide plays a crucial role in these reactions in proofreading the integration and excision reactions to prevent recombination between improperly synapsed configurations (Fig. 5). Interestingly, resolvase mutants that function without the accessory resolvase-binding sites and without the need for DNA supercoiling (20, 21) catalyze site I × site I recombination, and the palindromic central dinucleotide (5′-AT) permits strand exchange in both parallel and antiparallel site alignments (22). Because none of the serine-integrase systems appears to use accessory DNA sites or proteins, we predict that these systems will commonly use a two-step level of control, with protein–DNA complex conformations preventing most incorrect synaptic configurations and the nonpalindromic central dinucleotide discriminating productive and unproductive synaptic events.

Materials and Methods

DNA and Proteins.

Plasmids pMOS-attL(60), pMOS-attL(367), pMOS-attR(60), and pMOS-attR(376) were generated by cloning 60 and ≈370 bp of Bxb1 attL and attR into the EcoRV site of pMOS-Blue (Amersham). Linear partner attP, attB, attL, and attR DNAs were obtained by annealing complementary oligonucleotides of the required sequence. A suicide DNA substrate of attL was obtained by annealing a 60-bp oligonucleotide corresponding to the bottom strand and 27-bp (5′-phosphorylated) and 33-bp oligonucleotides corresponding to the top strand. Bxb1 gpInt and gp47 were purified after overexpression with minor modifications of previously published methods (8, 9).

In Vitro Recombination Assays.

In vitro excisive recombination was carried out between indicated sizes of attL or attR in pMOS-attL or pMOS-attR and linear partner DNA (14 ng/ml or as indicated) in a recombination buffer containing 20 mM Tris·HCl (pH 7.5), 10 mM EDTA, 25 mM NaCl, 10 mM spermidine, and 1 mM DTT with or without the addition of purified proteins (gpInt and gp47 as required). The reactions were carried out at 25°C for 2 h, separated on a 0.8% agarose gel, and visualized by ethidium bromide staining. Product identities were confirmed by digestion with Xmn1. Substrate DNA yields one 3-kbp fragment, and product fragments are 1,984 and 1,023 bp. In vitro reactions with φRv1 integrase were performed as described previously (11).

DNA-Binding Assays.

Briefly, ≈5 ng of labeled DNA (suicide substrate or normal DNA) was incubated with the 175 nM gpInt and/or 5.3 μM gp47 in a buffer containing 20 mM Tris·HCl (pH 7.5), 10 mM EDTA, 25 mM NaCl, 10 mM spermidine, 1 mM DTT, and 1 mg of calf thymus DNA in a total volume of 10 μl and incubated at 25°C for 1 h. Partner DNA was added where indicated. Protein–DNA complexes were separated from free DNA on a 5% nondenaturing polyacrylamide gel at 10°C. Where required, the reaction products were treated with 1 mg/ml proteinase K and 0.5% SDS and separated on a 10% polyacrylamide gel containing 0.05% SDS.

For 2D analysis of complexes, a vertical strip of gel containing the complex was excised, soaked in buffer containing 0.5% SDS (wt/vol) and 1 mg/ml proteinase K, layed on a 8% polyacrylamide gel containing 0.05% SDS, and subjected to electrophoresis.

In-Gel FRET Assay.

An 87-bp donor attL DNA was obtained by annealing a fluorescein-labeled attL oligonucleotide with a ³²P-labeled antiparallel oligonucleotide. Two suicide substrates containing a nick on either the top or the bottom strand at the core position (designated suicide substrate 2 and 1, respectively) were used. The DNAs were labeled at the 3′ end as shown with the Black Hole Quencher (IDT). Mobility shift assays were carried out as described for DNA Binding Assays, and the gel was scanned by using a fluorescence imager and a phosphoimager, and the intensities were quantified by using Multigauge (Fujifilm).