Type III restriction enzymes cleave DNA by long-range interaction between sites in both head-to-head and tail-to-tail inverted repeat

Abstract

Cleavage of viral DNA by the bacterial Type III Restriction-Modification enzymes requires the ATP-dependent long-range communication between a distant pair of DNA recognition sequences. The classical view is that Type III endonuclease activity is only activated by a pair of asymmetric sites in a specific head-to-head inverted repeat. Based on this assumption and due to the presence of helicase domains in Type III enzymes, various motor-driven DNA translocation models for communication have been suggested. Using both single-molecule and ensemble assays we demonstrate that Type III enzymes can also cleave DNA with sites in tail-to-tail repeat with high efficiency. The ability to distinguish both inverted repeat substrates from direct repeat substrates in a manner independent of DNA topology or accessory proteins can only be reconciled with an alternative sliding mode of communication.

Almost every genetic process requires protein complexes to interact simultaneously with specific DNA sites or structures that are located at-a-distance along the genome, including events in DNA replication, repair, recombination, and transcription. In many cases these so-called “long-range communications” are independent of the relative orientation of the interacting sites on the DNA (1). E.g., gene activation by remote enhancer elements is associated with random DNA looping between regulatory elements, i.e., chromatin loop formation (2). Such reactions can occur on DNA of any topology, and can even occur between sites on separate DNA molecules providing the local concentration is sufficiently elevated (1). In other cases however, a successful interaction only occurs when the sites are located on the same DNA in a specific relative orientation. The reaction can then be said to have “site-orientation selectivity,” which can be achieved using both NTP-independent or NTP-dependent pathways:

Site-specific recombinases (SSRs) have provided a mechanistic framework for NTP-independent communication (3–5). For example, the transposon-encoded resolvases have a strong preference to recombine sites in direct repeat on the same DNA (6). Site-orientation selectivity is important as uncontrolled rearrangements of DNA sequences may result in loss of function. For all SSRs studied to-date, the long-range communication occurs by the sites interacting via thermally driven three-dimensional diffusion, i.e., DNA looping (7). However, to achieve site-orientation selectivity, the geometry of the resulting site-site synapse is biased by an energetic “filter” that can be a preference for DNA substrates with a particular topology (e.g., 8) or the requirement for accessory DNA-binding factors (e.g., 9). For example, recombination by the resolvases requires the capture of three DNA nodes which are significantly favored by negative supercoiling (6, 8, 10–12).

For many processes that are NTP-dependent, site-orientation selectivity comes from directional one-dimensional motion along DNA. A classical example is transcription—only if RNA polymerase loads onto DNA in the correct direction will there be subsequent recognition of downstream transcriptional start and termination sites (13 and 14). A second example is the DNA cleavage activity of the bacterial Type I RM enzymes (15), in particular LlaGI (16 and 17). The RM enzymes play a critical role in bacterial cells in protecting against infection by foreign DNA. For LlaGI, dsDNA cleavage requires the presence of two asymmetric recognition sites in a defined head-to-head inverted repeat (HtH). Communication occurs by a helicase domain catalyzing processive, stepwise dsDNA translocation in one-dimension (16). Where there are two sites in HtH repeat, head-on collision between the converging motors activates the nuclease activity (16 and 17). Cleavage does not occur for other site arrangements, e.g., head-to-tail direct repeats (HtT) or tail-to-tail inverted repeats (TtT). Generally, site-orientation selectivity based on one-dimensional directional motor movement requires a significant input of chemical energy, in the range of at least one ATP for every base pair translocated before successful collision (16 and 18).

Alternatively, NTP-dependent site-orientation selectivity can be achieved independent of long-range directional motion and with greatly reduced NTP consumption, e.g., in mismatch repair (19) and by Type III RM enzymes (20), of which EcoPI, EcoP15I, and PstII are prototypical members. The Type III enzymes recognize short asymmetric sequences and cleave the nonspecific DNA 25-28 nucleotides 3′ to the site (Fig. 1A). As with LlaGI, nuclease activity requires at least two sites in specific orientations on the same DNA molecule and ATP hydrolysis by the Superfamily 2 helicase domain(s) (21–27). Based on their similarity to the Type I family, a unidirectional loop translocation scheme identical to that for LlaGI was originally proposed (15 and 22). Additional embellishments to the model including one-dimensional translocation without looping (28) and additional three-dimensional diffusive looping (29), have also been suggested. More recently, a single-molecule study provided unambiguous evidence that both passive and active DNA looping are not on-pathway to cleavage and that communication occurs bidirectionally in one-dimension along the DNA contour (30). Together with the low ATPase rates of the Type III enzymes (1,000-fold lower than for Type I enzymes, 22, 23, 30), which realistically rules out substantial motor-driven movement, a new model was developed in which motion along DNA occurs by thermally-driven one-dimensional diffusion, i.e., DNA sliding (20 and 30).

Fig. 1.

Efficiency of cleavage of tail-to-tail linear DNA is improved by end-capping. (A) Orientation of the Type III DNA recognition sequences. Cleavage loci are indicated by the red arrows. (B) Cleavage of TtT DNA by EcoPI or EcoP15I. DNA (without or with biotin ends, as indicated by the absence or presence, respectively, of a terminal cross), was incubated at 25 °C for 60 min with enzyme, streptavidin (Strp) and/or AdoMet, as indicated. DNA samples were separated by gel electrophoresis and the relative amount of dsDNA cleavage quantified by gel densitometry. For comparative gels with HtH and HtT DNA, refer to Fig. 3D and Fig. S3C in ref. 30. (C) Cleavage of biotinylated HtH, TtT, and HtT DNA by PstII. DNA was incubated for 10 min with PstII and/or streptavidin as indicated. Samples were analyzed as in B. Distances shown are the base pairs between the sites.

Based on previous studies it has been suggested that Type III enzymes, like LlaGI, cut sites which are in a HtH configuration with a high selectivity over sites in either HtT or TtT orientation (21, 22, 27). Defining these preferences properly is the primary step in formulating models for site-orientation selectivity. The experiments presented here show that Type III enzymes can in fact cut both HtH and TtT oriented sites with high efficiency. These observations can only be explained in the framework of the sliding model and are consistent with, and sufficient for, the biological role of Type III communication, which is to prevent cleavage of nonmethylated sites following replication (21). Our observations give insight into a unique logic for site-orientation selectivity, which provides conceptual input for understanding other processes, like mismatch repair, where directional communications also occur as a result of relatively low levels of ATP hydrolysis (19 and 20).

Results

DNA End Capping Stimulates Cleavage of Tail-to-Tail Oriented Pairs of Sites.

In earlier studies of Type III enzymes, where the strong preference for HtH oriented sites was concluded, the DNA substrates chosen had some crucial limitations. Firstly, the early EcoP15I studies (21 and 22), were carried out using multisite DNA where competing interactions could occur and where inefficient TtT cleavage could not be unambiguously assigned. Secondly, in some cases plasmids were used (25 and 26), where communication between HtH and TtT orientations cannot be distinguished. Thirdly, the use of simple linear DNA substrates resulted in significantly reduced cleavage compared to plasmid DNA (31).

Recently we have shown that cleavage efficiency of linear HtH DNA can be dramatically restored by binding physical roadblocks, such as streptavidin, at the DNA ends, which we term “end-capping” (30). In contrast, the poor cleavage activity on substrates with sites in direct (HtT) repeat was unaffected by end-capping. Thus, capped DNA supports optimum cleavage efficiency and specificity, similar to plasmid DNA, but importantly allows one to unambiguously distinguish differences in the cleavage kinetics for HtH and TtT orientations.

To provide a clear-cut verification of how Type III enzymes communicate, we generated linear DNA substrates with two sites in TtT repeat and with biotin-labeled ends, and tested the influence of end-capping with streptavidin. We first tested the highly related enzymes EcoPI and EcoP15I (Fig. 1 A,B). Similar to results seen for HtH substrates (30), end-capping had, for both enzymes, a striking effect on the product yield in the absence of the methyl donor S-adenosyl methionine (AdoMet). Addition of 100 μM AdoMet had an inhibitory effect, which was mild for EcoPI but more notable for EcoP15I (see below). The extent of cleavage of the capped TtT DNA was similar to that observed using capped HtH DNA; in contrast capped HtT DNA is not cleaved (30, see below). The distribution and location of the cleavage sites on the TtT DNA are the same as observed on HtH DNA. Therefore, we propose that the same principal endonuclease mechanism is utilized to cleave DNA with sites in either orientation.

To verify that cleavage of both inverted repeats is not exclusive to EcoPI and EcoP15I, we carried out similar experiments using PstII (Fig. 1 A,C), a more distantly related Type III enzyme (27). As for EcoPI and EcoP15I, we observed that TtT sites were cut when the DNA ends were capped, albeit less efficiently than the capped HtH DNA (Fig. 1C). Cleavage of the HtT DNA was not observed within the detection limit (Fig. 1C). Therefore, we also propose that cleavage of both HtH and TtT sites can be considered a general property of Type III enzymes.

Rates of Cleavage of Tail-to-Tail Repeats.

To compare cleavage of HtH and TtT substrates more quantitatively, we characterized the EcoPI and EcoP15I endonuclease kinetics using two independent assays: a highly parallel magnetic tweezers assay (30 and 32) that allows the detection of DNA looping and cleavage on stretched DNA; and bulk solution gel electrophoresis, where cleavage occurs in a random coil configuration of the DNA. The principle of the single-molecule experiments is shown in Fig. 2A (inset). Here we used an applied force of F = 1.5 pN, which stretches the DNA to ∼89% of its contour length. Fig. 2A shows a typical experiment for EcoP15I using a TtT substrate. For both enzymes, efficient cleavage of the TtT substrates was observed. In addition, none of the cleavage profiles recorded showed any evidence of loop translocation events prior to cleavage.

Fig. 2.

Comparison of DNA cleavage rates on inverted sites using single-molecule and bulk solution assays. (A) Highly-parallel magnetic tweezers assay (30). (inset) DNA molecules are bound at one end to the flow cell surface and at the other end to magnetic microspheres. A pair of permanent magnets are used to apply a stretching force. DNA length is determined by tracking the microsphere position using video microscopy. An example cleavage experiment using EcoP15I and a TtT DNA is shown, with the simultaneous tracking of 19 DNA molecules. z is uncorrected for the incomplete DNA stretching at F = 1.5 pN. During the period marked “flush”, EcoP15I and ATP were introduced into the flow cell. Subsequent DNA cleavage events are readily identified by the apparent rapid DNA lengthening which represents loss of microsphere tracking. The expected change in z if a loop were formed between the sites is indicated by the red arrows—none of the profiles showed this change. (B) Comparison of DNA cleavage kinetics for the HtH, TtT, and HtT substrates illustrated, obtained using magnetic tweezers or bulk solution measurements. Single-molecule HtT data is taken from ref. 30, and is shown for comparison. For the magnetic tweezers, data was normalized by the total number (n) of molecules measured with n = 62 (EcoPI HtH), n = 45 (EcoPI TtT), n = 78 (EcoP15I HtH), n = 49 (EcoP15I TtT), n = 68 (EcoPI HtT), and n = 39 (EcoP15I HtT). For the bulk solution reactions the same HtH and TtT DNA as used for the magnetic tweezers was labeled at both ends with biotin. DNA was then incubated with enzyme and streptavidin, samples taken at the times shown and the extent of DNA cleavage ascertained by gel electrophoresis and densitometry. Base pairs between the sites were: 1,097 (HtH), 2,450 (TtT), and 1,074 (HtT) for EcoPI; and 946 (HtH), 2,598 (TtT), and 925 (HtT) for EcoP15I.

The DNA cleavage kinetics, obtained from the individual cleavage times of many molecules (Fig. 2B) was compared to cleavage data obtained under identical reaction conditions in bulk solution using capped linear DNA. For both enzymes and both HtH and TtT orientations, the single-molecule and bulk solution data coincide, within experimental error. Given the different DNA configurations in the tweezers (DNA stretched by applied force) and in bulk solution (no applied force and DNA as a random coil), the similar kinetics provide further evidence for force-independent one-dimensional communication, as noted previously for HtH DNA using a broad range of forces (30). In comparison, neither of the two enzymes produce any significant cleavage of HtT DNA; the comparable data from ref. 30 in Fig. 2B shows cleavage confined to < 10% of molecules and with a drastically slower rate.

For EcoPI, DNA cleavage was found to be equally fast and efficient for both HtH and TtT substrates. For EcoP15I however, while the efficiencies of cleavage were similar, cleavage of the TtT substrate occurred almost 10-fold more slowly compared to the HtH substrate, which is still substantially faster than the limited HtT cleavage (Fig. 2B). This difference suggests that despite the similarity between the two enzymes, EcoP15I has more difficulty in communicating between TtT pairings. One possible explanation for the disparities observed in Fig. 2B is that the intersite spacing is different for the HtH and TtT substrates and a limited processivity of the EcoP15I motion could reduce the efficiency of communication for the longer site spacings. However, a comparison of HtH and TtT cleavage rates on a variety of different substrates showed that the kinetics are independent of intersite distance (SI Text), as noted previously (30 and 31).

To address if the ATP consumption per cleavage was different between HtH and TtT substrates, we measured the rate of the initial phase of ATP hydrolysis using EcoPI (SI Text). We found that the ATPase rates were, within experimental error, similar on the HtH and TtT substrates and were independent of end-capping. As found previously (30), the amount of ATP hydrolyzed per cleavage event (∼150 to ∼175 ATP per cleavage event, ∼100 to ∼115 if we correct for background activity; see SI Text) is below that expected for true dsDNA translocases (16 and 18).

The Effect of Roadblocks on Long-Range Communication Between Sites in Inverted Repeats.

To confirm that the mode of communication between both HtH and TtT sites uses the same one-dimensional route, we investigated the effect of protein or DNA molecules (“roadblocks”) bound in the path of the enzymes. Previously this roadblock effect was investigated for the Type III enzymes by binding Lac repressor between a pair of HtH sites only (22). We extended this analysis to both HtH and TtT DNA by using Tn21 resolvase as a roadblock (33; Fig. 3A). A Tn21 res site (a specific sequence for the binding of three resolvase dimers) was placed in either the HtH or TtT arcs of an EcoPI plasmid. End-capped linear HtH or TtT substrates were also generated. On both circular DNA (Fig. 3A), binding of resolvase had little effect (the moderate reduction in efficiency was due to altered buffer conditions). In contrast, on both linear DNA, resolvase binding completely inhibited cleavage (Fig. 3A). These results are only consistent with one-dimensional bidirectional motion: on circular DNA, there are two routes for communication, via the HtH arc and via the TtT arc. If one route is blocked by resolvase, the enzyme can use the alternative route. On linear DNA in contrast, there is only one route and resolvase completely blocks communication and cleavage.

Fig. 3.

The effect of protein and DNA roadblocks on DNA communications. (A) Tn21 resolvase binding blocks Type III communication in one-dimension. Circular or linear DNA were prepared with two inverted EcoPI recognition sites and a single Tn21 res site in either the HtH (left) or TtT (right) DNA arc. Linear DNA substrates were labeled with biotin and streptavidin added. The DNA was then incubated for 5 min at 25 °C with EcoPI and Tn21 resolvase, as indicated, and samples separated by gel electrophoresis. CCC is Covalently Closed Circular DNA, FLL is Full Length Linear DNA and OC is Open Circle DNA (nicked). (B) DNA triplex binding does not block Type III RM enzyme communication. Circular or linear DNA were prepared with two inverted EcoPI sites and two independent Triplex Forming Oligonucleotides (TFO#1 and TFO#2) bound in either the HtH or TtT DNA arcs, as indicated. The circle on each TFO indicates the 5′ end. Linear DNA was labeled with biotin and streptavidin. DNA was bound with 1.5-fold molar excess of one or both triplexes. Cleavage reactions were incubated for 5 min, with EcoPI where indicated. Reaction products were separated by agarose gel electrophoresis and analyzed by gel densitometry.

We previously ruled out a stepwise motor mechanism for Type III enzymes based on a triplex displacement assay (30), in which motion is inferred from the displacement of a specifically bound DNA roadblock at a set distance from the initiation site (34). Using circular substrates, we found no evidence for triplex displacement and also no effect on DNA cleavage, and concluded that the enzyme can bypass triplexes during sliding (30). However, as we have now demonstrated (Fig. 3A), if one DNA arc of a plasmid is blocked, communication could occur via the other arc. We therefore repeated our triplex assays using a circular DNA with two triplex binding sites in the TtT arc only, and also on linear end-capped HtH and TtT substrates with two triplex binding sites between the Type III recognition sequences (Fig. 3B). Each DNA was efficiently cleaved by EcoPI irrespective of the presence of one or other, or both, triplexes. Depending on the triplex used, background levels of displacement were 7%–36%, but were not increased in the presence of EcoPI. Therefore we can rule out that triplex displacement is on pathway to cleavage. Additionally, the results on linear DNA now provide direct evidence that Type III enzymes can efficiently bypass triplexes in stark contrast to true dsDNA motors (16 and 18).

Discussion

Only a Sliding Mode Can Rationalize the Site-Orientation Selectivity of Type III Restriction Enzymes.

The data presented here proves that whilst Type III enzymes can communicate between distant sites both up- and downstream of their recognition sequences, cleavage only occurs at inverted repeats. Therefore, communication is a one-dimensional process that must be bidirectional and yet retain a selective bias for site orientation. How could this occur? The “topological filter” proposed for the SSRs (Fig. 4A; 4 and 8), cannot explain our results as there is no requirement for DNA supercoiling or looping (Fig. 2; 30). Alternative ATP-driven directional translocation schemes have an additional problem beyond the insufficient ATPase rates of Type III RM enzymes, namely that directional motion can only communicate between one relative arrangement (Fig. 4B, 16); for example, if HtH sites are cut, by definition TtT sites should never be cut as the translocating enzymes would diverge. To account for our observations we must now add DNA sliding, i.e., bidirectional random movement, as a third way of explaining site-orientation selectivity during long-range communication (Figs. 4 C,D) (20).

Fig. 4.

Alternative schemes for site-orientation selectivity. (A) Topological filter for site-specific recombinases (4 and 8). In this example (e.g., Tn21 resolvase), three-dimensional diffusion results in the capture of two binding sites and the trapping of three negative DNA nodes (10 and 11). DNA supercoiling is required to stabilize the complex. Strand transfer results in the formation of a single-interlinked DNA catenane. (B) dsDNA loop translocation as demonstrated for LlaGI (16) and proposed for the Type III enzymes (22). The enzyme remains bound to the site while dsDNA translocation by the helicase domain extrudes a DNA loop. Convergent collision results in the formation of an active nuclease and thus DNA cleavage. (C) Type III sliding mechanism on HtH DNA (20 and 30). Capped DNA ends are shown as crosses and the Type III enzymes as bilobed ellipsoids (where M is the dimeric methyltransferase subunit and R is the dimeric helicase-endonuclease subunit). The change in color of the Type III enzymes represents an ATP-dependent conformational switch from site-specific DNA binding (blue) to random bidirectional DNA sliding (green). Upon binding of the second enzyme, only those enzymes that are sliding in the leftward region are available for direct head-on interaction to produce DNA cleavage. (D) Type III sliding mechanism on TtT DNA. Upon binding of the second enzyme, only those enzymes that are sliding in the rightward region are available for direct head-on interaction to produce DNA cleavage.

The role of limited ATP hydrolysis by the Type III enzymes may seem superfluous as one-dimensional diffusion is driven by thermal energy. However, as sliding can also play a role in the initial search for a site (35), the Type III enzymes must have a way to distinguish between initiation from specific and nonspecific sites. We speculate that this is where ATP plays a critical role. Only those enzymes that have already visited a site are in the correct nucleotide-bound state to induce cleavage upon collision with a second enzyme bound at another site. The low ATPase activity then represents loading of sliding enzymes at a site and not the stepwise motion along DNA thereafter. This role is reminiscent of ATPases in loading sliding clamps onto DNA (36). Moreover, ATP may play a wider role in initiation of DNA sliding, for example in mismatch repair (19 and 20). While in theory as little as two ATP could be required (i.e., two loading events) for DNA cleavage by Type III enzymes, finite lifetimes of sliding and/or site association would lead to some sliding events falling off the DNA before collision making multiple rounds of loading-sliding-dissociation necessary before a successful collision-cleavage event occurs.

Both the Type III enzymes and LlaGI (Fig. 4 B,C,D) (16 and 17), require indirectly repeated binding sites to cleave DNA. This selection logic is important in their biological context, since for both types of enzymes only one strand of the recognition sequence is methylated. While “hemimethylation” of a site protects the host DNA from auto-cleavage, following replication there is a transient period when an unmethylated site arises on one daughter DNA (16 and 21). By requiring the interaction of pairs of sites in an inverted repeat, at least one site of the two will always be hemimethylated (21). It is perfectly acceptable to use HtH and TtT arrangements to achieve this, as both provide the same “logic gate.” It is very remarkable that LlaGI and Type III enzymes have similar helicase domains, similar biological roles and similar requirements for directional communication, and yet use completely different mechanisms for communication.

How DNA Binding Lifetimes Can Influence the Relative Outcome on Different DNA Substrates.

While EcoPI cleaves HtH and TtT DNA with equal rates and efficiencies, both EcoP15I and PstII are less productive on TtT DNA. This behavior can be understood by considering that protein binding can act as a roadblock/reflecting barrier to Type III motion (Fig. 3; 22 and 30) To cleave TtT sites, a Type III enzyme must bypass an oppositely oriented, distant site (Fig. 4D). The specific binding of a Type III enzyme at this site would block bypass and inhibit the reaction. In contrast, cleavage of HtH DNA can occur without bypass of the second site (Fig. 4C). The reduction in efficiency of TtT cleavage is thus determined by the dynamics of the individual reaction steps, including enzyme-site-association, initiation of sliding, sliding itself, and enzyme dissociation.

To illustrate this effect we carried out Monte-Carlo simulations of DNA cleavage based on the sliding model (Fig. 4 C,D), in which the initiation rate was varied relative to the site-association rate and the dissociation rate (Fig. 5). These rates where chosen to be slow compared to the sliding rate, in agreement with the observation that the cleavage kinetics does not depend on the intersite spacing (SI Text; 30). While both HtH and TtT cleavage occur at similar rates when initiation is fast, TtT cleavage becomes more strongly disfavored as the residence time at the site increases (Fig. 5). Allowing the enzymes to dissociate from the DNA during sliding provides a similar trend but with an even more strongly disfavored TtT cleavage. Thus, if we assume a residence time with the order PstII > EcoP15I > EcoPI, we can readily explain our experimental observations within the framework of the sliding model. Moreover, anything that enhances DNA binding, e.g., AdoMet (37), will also enhance inhibition of TtT cleavage relative to HtH cleavage, exactly as observed for EcoP15I (Fig. 1; 30). In future work it is necessary to measure these rates directly and to confirm the suggested role of ATP as a molecular switch.

Fig. 5.

How sliding can explain differences in the efficiency of cleavage of head-to-head and tail-to-tail DNA. The model shows a Type III enzyme that binds to its site with a rate k_bind, hydrolyses ATP and switches to a sliding mode with a rate k_ini and then dissociates during sliding with a rate k_off. Leftward or rightward stepping during sliding occurs at a rate 1 × 10⁶-fold faster than k_bind. Monte Carlo simulations were undertaken on DNA arrays with two sites in either HtH or TtT arrangement, with 50 sliding steps between the sites and 100 sliding steps from each site to the capped DNA end. Attempts to step off the DNA ends were reflected back and multiple enzymes could be bound simultaneously. Once an enzyme left a site, another enzyme was then allowed to associate, such that multiple enzymes could be loaded onto the DNA. While for the HtH arrangements a low average number of enzymes loaded (< 3) was required to achieve cleavage, for the TtT arrangements typically more enzymes (up to 10) were required. The number of model cycles until two enzymes collided in the correct cleavage orientation (Fig. 4 C,D) was then recorded, which in turn reflects the cleavage time. The ratio of these values for the HtH versus TtT simulations was calculated. Values are the mean from 150 iterations of the model.

Materials and Methods

Proteins and DNA.

EcoPI and EcoP15I (30), PstII (27) and Tn21 resolvase (33) were produced as previously described. Plasmids are detailed in Table S1. The lengths of the DNA segments in each of the Figures are detailed in Table S2. For the magnetic tweezers, plasmids were cut with suitable Type II restriction enzymes to generate cohesive ends. Biotin- or digoxigen-modified attachment handles (∼0.6 kbp) were prepared by cutting PCR fragments with the same restriction enzymes, and subsequently ligating to the digested plasmid. For bulk solution assays, plasmids were cleaved with suitable Type II restriction enzymes and the DNA ends biotinylated by using either, Klenow polymerase and biotin-dUTP (30), or terminal transferase and biotin-16-ddUTP (Roche).

Magnetic Tweezers.

The basic parallel magnetic tweezers protocol has been described (30). In brief, biotin/digoxigen-labeled DNA constructs were mixed with 1 μm streptavidin-coated superparamagnetic microspheres (Invitrogen) and the DNA-attached beads coupled to the surface of a flow cell treated with antidigoxigen and maintained at 25 °C. Subsequently, 15 nM EcoPI or EcoP15I, and 4 mM ATP in Buffer R+ (50 mM Tris-Cl, pH 8.0, 50 mM KCl, 10 mM MgCl₂, 1 mM DTT, 100 μg/mL BSA) were flushed into the cell, the flow stopped, and multiple microspheres observed simultaneously. The cleavage time was taken as the period from the start of the flush to the loss of the microsphere tracking.

Bulk Solution Cleavage Measurements.

Unless indicated otherwise, cleavage assays contained 2 nM DNA (supercoiled or linear), 4 mM ATP and 15 nM EcoPI or EcoP15I in buffer R+. Where included, AdoMet was at 100 µM and streptavidin at 100 nM. For the roadblock assays, 80 nM Tn21 resolvase was also added. Reactions were started by adding Type III enzyme and then incubating at 25 °C for the times indicated. Reactions were stopped with 0.5 volumes of Stop Buffer I [0.1 M Tris (pH 7.5), 0.2 M EDTA, 40% (w/v) sucrose, 0.4 mg/mL bromophenol blue]. For reactions with streptavidin, biotin was added to a final concentration of at least 82 μM and the samples heated at 80 °C for at least 5 min. Samples were then analyzed by agarose gel electrophoresis and the percentage of DNA in each band per lane ascertained by gel densitometry. Reactions using PstII were carried out as above except that the assays contained 5 nM DNA and 142 nM PstII in NEBuffer4+ [20 mM Tris-acetate, pH 7.9, 50 mM potassium acetate, 10 mM magnesium acetate, 1 mM DTT, 0.01% (v/v) Triton-X100]. Mean and standard deviation values where quoted are from at least two repeat measurements.

Triplex Displacement Assays.

DNA triplexes were formed by incubating 50 nM DNA with 75 nM ³²P-labeled triplex forming oligonucleotide #1 (TFO#1, 5′-TTCTTTTCTTTCTTCTTTCTTT-3′) and/or TFO#2 (5′-TTTCTTCTTCTTTTCTTTTCTT-3′) overnight at 20 °C in 10 mM MES (pH 5.5) and 12.5 mM MgCl₂ (34). 5 nM triplex-bound DNA and 37.5  M EcoPI were mixed in buffer R+ supplemented with 4 mM ATP and, where required, 250 nM streptavidin. The reactions were incubated at 25 °C for 5 min and quenched with either 0.25 volumes of Stop Buffer II [15% (w/v) glucose, 3% (w/v) SDS, 250 mM MOPS pH 5.5, 0.4 mg/mL bromophenol blue] or 0.5 volumes of Stop Buffer I plus at least 82 μM biotin. To separate the free and bound triplex forms, the Stop Buffer II samples were loaded directly onto acid agarose gels as described (34). Displaced triplex values were corrected for the free (unbound) TFO present at the start of the reaction and for reactions with two triplexes, both contribute to the free and displaced pool. To separate the substrate and cleavage products, the STEB samples were heated for at least 5 min at 80 °C and loaded on a standard agarose gel. Mean and standard deviation values where quoted are from at least two repeat measurements.