Determination of the DNA bend angle induced by the restriction endonuclease EcoRV in the presence of Mg2+.

We have used the method of Zinkel and Crothers (Zinkel, S.S., and Crothers, D.M. (1990) Biopolymers 29, 29-38) to determine the degree of bending induced by the binding of the restriction endonuclease EcoRV to its recognition sequence (-GATATC-). A set of four calibration DNA fragments was constructed that contained zero, two, four, or six phased A-tracts in their centers and an EcoRV site at the 5'-end to account for the electrophoretic influence of the bound protein. The mobilities of these calibration molecules complexed with EcoRV were compared to that of a test DNA containing a central EcoRV site also complexed with EcoRV. The EcoRV-induced bend angle was found to be 44 degrees +/- 4 degrees. These experiments were performed with a catalytically inactive EcoRV mutant that still binds DNA specifically in the presence of Mg2+. In the absence of Mg2+, which is necessary for specific binding, there is no difference in the mobilities of the fragments with a peripheral or a central EcoRV site complexed with EcoRV, indicating that nonspecific binding on average does not lead to measurable DNA bending.

We have used the method of Zinkel and Crothers (Zinkel, S. S., and Crothers, D. M. (1990) Biopolymers 29,29-38) to determine the degree of bending induced by the binding of the restriction endonuclease EcoRV to its recognition sequence (-GATATC-). A set of four calibration DNA fragments was constructed that contained zero, two, four, or six phased A-tracts in their centers and an EcoRV site at the 5'-end to account for the electrophoretic influence of the bound protein. The mobilities of these calibration molecules complexed with EcoRV were compared to that of a test DNA containing a central EcoRV site also complexed with EcoRV. The EcoRV-induced bend angle was found to be 4 4 O f 4 O . These experiments were performed with a catalytically inactive EcoRV mutant that still binds DNA specifically in the presence of M 8 + . In the absence of M 8 + , which is necessary for specific binding, there is no difference in the mobilities of the fragments with a peripheral or a central EcoRV site complexed with EcoRV, indicating that nonspecific binding on average does not lead to measurable DNA bending. "The distortion of DNA from the straight, regular (and essentially mythical) double-helical structure beloved of textbooks is a ubiquitous feature of protein-DNA complexes" (Travers, 1990). Curvature can be an intrinsic feature of the DNA itself (for recent reviews, see Crothers et al. (1990), Hagerman (1990Hagerman ( , 1992, and Trifonov (1991)), either due to a specific sequence or as a consequence of supercoiling, but can also be caused by bending induced by protein binding. The bendability of a given DNA sequence presumably plays a substantial role in the DNA recognition by repressors, activators, and DNA-processing enzymes and in the packaging of DNA into nucleosomes (Travers, 1987(Travers, ,1991. Protein-induced DNA bending has been demonstrated to accompany many protein-DNA interactions. For the catabolite activator protein, for example, bending of the DNA has been shown both crystallographically in the protein-DNA complex (90") (Weber and Steitz, 1987) and in solution by gel shift experiments (100") (Zinkel and Crothers, 1990). It has been proposed that a possible role of this bending is the formation of additional contacts of the DNA upstream of the catabolite activator protein binding sequence with the RNA polymerase (Zinkel and Crothers, 1991). A crystal structure of a 434-repressor fragment with a 20-mer reveals bending and overwinding of DNA in the central AT-rich region that is accompanied by a dramatic narrowing of the minor groove * This work was supported by Deutsche Forschungsgemeinschaft Grants Pil22,2-3 and Pi122,5-2. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. (Aggarwal et at., 1988). DNA bending by Cro protein binding has been demonstrated by Lyubchenko et at. (1991) in DNA circularization experiments in the absence and presence of Cro protein.
Also for DNA-processing enzymes such as restriction endonucleases, bending might be important in facilitating specific binding and/or catalysis. For the restriction endonucleases EcoRI (Kim et al., 1990;Rosenberg, 1991) and EcoRV (Winkler, 1992), crystal structures of their specific DNA complexes have been solved. In both these complexes, the DNA is distorted from its "normal" B-form. For the EcoRI DNA sequence -GAATTC-, there is evidence that the DNA itself already shows deviation from the B-form. This has been shown by NMR (Nerdal et al. 1989), by anomalies in gel migration (Diekmann and McLaughlin, 1988), and by x-ray crystallography (Wing et al., 1980;Dickerson and Drew, 1981;Frederick et al., 1988). Upon binding of EcoRI, this deviation is reinforced (Thomas et al., 1989), as has been shown by unwinding experiments (Kim et al., 1984) and gel electrophoretic mobility shift experiments (Thompson and Landy, 1988;Douc-Rasy et al., 1989). Similar results have been reported for the isoschizomer of EcoRI, RsrI (Aiken et al., 1991).
DNA containing an EcoRV site has not been as thoroughly studied so far. No NMR or x-ray data are available for an oligodeoxynucleotide containing the sequence -GATATC-. However, the x-ray structure of the specific EcoRV-DNA complex in the absence of M$+ (Winkler, 1992) shows that the DNA is bent in the complex. Gel shift experiments with the catalytically inactive EcoRV mutant D90A that still binds DNA specifically in the presence of Mg2+ (Thielking et al., 1992) indicate that, also in solution, bending occurs in the complex.
To investigate this result in more detail and to examine the role of Mg2+ in the bending process, we determined the bend angle of the DNA induced by EcoRV in the absence and presence of Mg2+ using the D90A mutant.

MATERIALS AND METHODS
Bacterial Strains and Plasmids-For the amplification of plasmids, Escherichia coli strain LK111(X) was used. Plasmid pAT153 (Twigg and Sherratt, 1980) was the vector for the construction of the different bending fragments. HinfI digests of plasmid pTR54, a derivative of pBR322,' served as a standard for the calibration of DNA length.

Construction and Purification of EcoRV Mutants with His Tags-
The construction of the DSOA mutant (Thielking et al., 1992), the introduction of an NHz-terminal Hiss tag, and the purification of tagged enzymes (Wende et al., 1991) have already been described.
Construction of DNA Substrates-A set of DNA molecules of identical length containing calibration bends and reference and test bends was constructed as follows. Double-stranded oligodeoxynucleotide cassettes containing the phased standard bending A-tracts (Bend I = 36', Bend I1 = 72", and Bend I11 = lot?"), the EcoRV sequence, or a "stuffer fragment" with a random sequence were ligated T. Ruter, unpublished data. primers that were used to obtain the five different constructs: 418(center), 418(end), and Bends 1-111. Primers were chosen such that the length of all fragments was exactly the same and such that the A-tracts (for Bends 1-111) and the EcoRV site (for construct 4Wcenter)) were always perfectly centered. Primers for Bends I1 and 111 introduced a new EcoRV site at the 5'-end of the fragment; the primer for construct 418(center) deleted the natural EcoRV site of pAT153. The locations of the PCR primers with respect to the DNA to be amplified are indicated.
into the BarnHI site of pAT153. The oligodeoxynucleotides had compatible BarnHI overhangs, but did not restore the site, thus providing a selectable restriction marker. Oligodeoxynucleotide Bends 1-111 also had an NsiI site (Fig. L4). After transformation of ligated plasmids, positive clones were identified by restriction digest and verified by sequencing. The plasmids then served as PCR' templates for the production of DNA fragments of identical length (418 bp). PCR primers were chosen such that the bending A-tracts (Bends I-111), the EcoRV site (construct 418(center)), and the stuffer fragment (construct 418(end)) were always positioned at the exact center of the fragment and such that the length of the fragment was always exactly the same (Fig.   1B).
In the PCR reaction, an EcoRV site was introduced 5 base pairs from the 5'-end of both the bending standards and the fragment containing the stuffer sequence (construct 418(end)) with an appropriate PCR primer. The fragment containing the EcoRV site in the center (construct 418(center)) had no EcoRV site at the end. The different constructs are shown in Fig. 2.
Polymerase C h i n Reaction-PCR was carried out in a TPS*5.4 thermocycler (version 3.2; Landgraf, Hannover, Germany). For radioactive labeling of the fragments during PCR, [w3'P]dATP (Amer- The abbreviations used are: PCR, polymerase chain reaction; bp, base pair(s).
sham, Braunschweig, Germany) was used. The following reaction conditions were chosen: a 50-pl volume of 4.21 nM template DNA, 0.4 FM concentration of each primer, 0.2 mM dNTP mixture, 5 p1 (lox) of Taq buffer, 3 units of Taq polymerase (Amersham), and 15 pCi of [(II-~'P]~ATP. Cycle 1 was for 390 s at 92 "C; cycle 2 for 90 s at 92 "C, 90 s at 54 "C, and 210 s at 72 "C; and cycle 3 for 360 s at 72 "C and 30 s at 0 "C. Cycles 1 and 3 were run once; cycle 2 was repeated 30 times.
Purification of PCR Products-PCR products were separated by gel electrophoresis on a 6% polyacrylamide gel and visualized by autoradiography. The DNA fragments of interest were excised and eluted with water. The purified fragments were used for gel electrophoretic mobility shift experiments.
Gel Electrophoretic Mobility Shift Experiments-Gel electrophoretic mobility shift experiments with EcoRV have been described in detail by Thielking et al. (1992). In the experiments reported here, the concentration of the DNA fragments was 1.25 nM, and the concentration of the His-tagged EcoRV mutant D90A (Selent et al., 1992;Wende et al., 1991) was 25 nM. Bands were visualized by autoradiography.
Measurement of Gel Electrophoretic Mobilities-The mobilities of the different DNA constructs and the protein-DNA complexes were measured from the autoradiographs of the gels. An RL value was computed for the uncomplexed DNA species and for the protein-DNA complexes. RL is the apparent length of the molecule, deter- shifted from both the 5'-and 3'-ends toward the center of the fragment to keep the length of all fragments identical. This leads to a loss of the natural (pAT153-derived) EcoRV site in Bends 1-111, which is therefore introduced via the PCR primers. In construct 418(center), the natural EcoRV site is deleted with the PCR primer. The expected degrees of bending for the different constructs are indicated. mined from its mobility relative to a standard, divided by its actual length. A plot of RL uersus the square of the number of A-tracts was used as a calibration curve to determine the bend angle of the protein-DNA complex. Alternatively, a direct plot of RF uersus degree of bending (one A-tract = 18") was used. The RF was determined by dividing the migration distances of the Bend 1-111 fragments by that of construct 418(end).

RESULTS
The DNA fragments 418(center) (with the EcoRV recognition sequence in the center), 418(end) (with the EcoRV recognition sequence 5 base pairs from the 5'-end), and a stuffer cassette in the center (Bend I (two A-tracts), Bend I1 (four A-tracts), and Bend I11 (six A-tracts)) were constructed as described under "Materials and Methods" (Fig. 2). Bends 1-111 serve as standards for the degree of DNA bending. Atracts phased at 10.5-bp intervals are especially suited as standards as their bending propensities have been studied extensively with different methods, i.e. electric dichroism (Levene et al., 1986), cyclization kinetics (Zahn and Blattner, 1987;Koo and Crothers, 1988;Koo et al., 1990), and crystallographic studies (Nelson et ul., 1987). According to these independent methods, one A-tract induces a bending of 18". Therefore, Bends 1-111 have been assigned 36", 72", and 108" of bending, respectively.
The A-tracts were placed in the center of the DNA molecule under study, and an EcoRV recognition site was placed 5 base pairs from the 5'-end to account for the contribution of bound protein. The DNA that served to determine the EcoRVinduced bending was a fragment of identical length with an EcoRV site instead of A-tracts in the center. Construct 418(end) (with the EcoRV sequence at the 5'-end) was chosen as a reference. Binding of EcoRV to the end of the DNA fragment should not influence the mobility, except for the increase in molecular weight of the complex, as the end-toend distance of the DNA remains more or less the same. This approach is similar to the one described by Zinkel and Crothers (1990) for the catabolite activator protein.

Mobility of Different Bend Constructs on Polyacrylamide
Gels-To show the different mobilities of the five constructs in the absence of protein, they were separated by electrophoresis on both EDTA-and M$+-containing polyacrylamide gels. The position of the EcoRV site on the DNA fragment does not influence the mobility of the free DNA (data not shown). As shown in Fig, 3 (top), an increasing number of A-

bp
tracts leads to an increase in gel retardation. The relative mobility of the different constructs is independent of the presence of Mg+. It should be noted that MP-containing gels need a longer run time to achieve the same separation of the different fragments, but the relative distances between the bands representing the different fragments remain the same.
Gel Electrophoretic Mobility Shift Assays-For the gel electrophoretic mobility shift assays, the EcoRV mutant D90A was chosen. This mutant still binds to DNA specifically in the presence of Mg+, but does not catalyze the cleavage reaction. In the absence of M F , however, the D90A mutant (Thielking et al., 1992), like the wild-type enzyme (Taylor et al., 1991), binds to DNA largely nonspecifically. In this study, the experiments were carried out with a His-tagged D90A mutant. It must be emphasized that NH2-terminal His tags leave DNA binding and DNA cleavage activities of the wildtype EcoRV as well as the EcoRV mutants unaffected (Wende et al., 1991).3 In Fig. 3 (top), binding of D90A to the different DNA constructs in the absence of M$+ is shown. All constructs yield several band shifts, although only one EcoRV-binding site is present, indicating nonspecific binding (Thlelking et al., 1992). Due to their intrinsic bends, Bends 1-111 show a different mobility compared to the DNA without A-tracts both in the free and the enzyme-bound DNAs. Constructs

418(center) (lane b ) and 418(end) (lam a )
show exactly the same mobility, which is reduced to the same extent when the DNA is complexed with the D90A mutant (lanes g and f ) . In Fig. 3 (bottom), a gel electrophoretic mobility shift assay of D90A with the different constructs in the presence of Mg2+ is shown. Upon specific binding as it occurs in the presence of M e , only a single band shift occurs. The test bend (construct 418(center)) (laneg) now shows a markedly different mobility than construct 418(end) (lane f ) due to the bending by EcoRV. In the gel electrophoretic mobility shift experiments carried out in the presence of M e (Fig. 3,  bottom), but not, however, in its absence (Fig. 3, top), the main shifted band is accompanied by a minor band of higher mobility. We have observed this phenomenon before with the EcoRV mutant D90A (Thielking et al., 1992); we believe that the satellite band is due to a conformational isomer of EcoRV.  (relative curvature), where C is a constant). In the top panel, the stippled boxes represent the experimental RL values ( i e . the apparent length of construct 418(end) or Bend I1 or I11 when complexed with D90A, divided by its actual length the electrophoretic mobility of Bend I, 11, or I11 when complexed with D90A, divided by the electrophoretic mobility of construct 418(end) when complexed with D90A). The D90A-418(center) complex has an RF value that is slightly smaller than the one for the DWA-Bend I complex, which means that D90A induces a bend of 44".

DISCUSSION
The crystal structure of a specific EcoRV-DNA complex has demonstrated that the DNA is highly distorted compared to regular B-DNA. Most conspicuous is the bending with a bend angle of -55" (Winkler, 1992). It is tempting to assume that this distortion, which drives the DNA into a rather unfavorable conformation, is needed for specific binding and catalysis in as much as it maximizes protein-nucleic acid contacts, provides a binding site for the essential cofactor M$+ , increases the reactivity of the DNA, and positions catalytically relevant groups of the protein and the DNA (Jeltsch et al., 1992). On the other hand, the specific EcoRV-DNA complex analyzed crystallographically was obtained in the absence of M F to prevent cleavage, i.e. under conditions in which EcoRV binds all DNA sequences (specific and nonspecific) with similar affinity (Taylor et al., 1991). Only in the presence of M$+ does EcoRV show a strong preference in binding specific DNA sequences (Thielking et al., 1992). Hence, the question arises as to what extent the distortion of the DNA seen in the crystal structure is also present in solution, in particular, whether it is a characteristic feature of the recognition complex formed in the presence of Mg+. We Zinkel and Crothers (1990) for the determination of the DNA bending induced by the catabolite activator protein. To be able to carry out these experiments in the absence and presence of Mg+, we have employed a mutant of EcoRV, D90A, which carries a substitution in the active site, is catalytically inactive, but retains its sequence-specific binding capacity (Selent et al., 1992). The experiments were performed with a NHz-terminally €&-tagged EcoRV mutant. As the NHz terminus of EcoRV is far away from the DNAbinding site or the subunit-subunit interface, no interference of the His6 tag with the enzymatic function of EcoRV was expected. Indeed, a comparison of the DNA binding and DNA cleavage activities of wild-type EcoRV and His6-EcoRV, as well as several mutants, without and with His6 tags, shows that the 6 extra His residues at the NH2 terminus have no effect on these activities. For the work described here it was important to demonstrate that D90A and His6-D90A with the same concentration dependence produce multiple band shifts in the absence of M$+ and a single band shift in the presence of M$+ in a gel electrophoretic mobility shift assay with a 377-bp fragment (Thielking et al., 1992). 4 Our results show that in the absence of M$+, EcoRV does not lead to bending of the DNA as in the gel retardation assays, no differences were detected among the mobilities of the complexes of the D90A mutant with DNAs containing a centrally or a peripherally located EcoRV site. This finding suggests that in contrast to the crystal structure, there is no bending in the absence of the cofactor. However, from Fig. 3  (top) it is clear that binding is nonspecific in the absence of the cofactor M$+ because instead of just one band, multiple band shifts appear. This means that even in the 1:1 complex, the enzyme probably does not occupy the cognate site, but rather can be located at any site. As the 418-bp-long DNA T. Stover, unpublished data. fragment employed in this study contains -400 noncognatebinding sites and as the preference for the cognate site is low in the absence of M$+, the proportion of complexes with the enzyme occupying the cognate site is too small to be seen in a gel shift assay. Therefore, with the approach chosen, nothing can be concluded for the DNA bending induced by EcoRV when occupying a cognate site in the absence of M P . It is clear, however, that averaged over time and space (i.e. when EcoRV associates to and dissociates from a noncognate site in order to associate with another noncognate site, only to dissociate again, etc.), no bending occurs.
In the presence of M2+, when EcoRV forms a specific complex with DNA, DNA bending by 44" occurs, as our gel electrophoretic mobility shift experiments demonstrate. This bend angle is similar to the one determined from the crystal structure for a specific complex formed in the absence of M P . The small difference in bend angles may be due to crystal packing forces or may be a consequence of the presence or absence of M P . We conclude from the similarity of the two values that the specific EcoRV-DNA complex obtained in the absence of Mg2+ has most of the features of a true recognition complex, which requires M 2 + bound to the active site. This conclusion is supported by the finding that M P soaked into a co-crystal does not lead to major structural changes. ' We envisage the process by which EcoRV "finds" its recognition site on a macromolecular DNA substrate as random association-dissociation events, possibly facilitated by linear diffusion along the DNA. Conformational fluctuations of the protein and the DNA serve to probe for optimal interactions as they are formed when EcoRV occupies a cognate site. These lead to further conformational changes, including bending of the DNA, which in turn allows M e to associate with the complex and to initiate catalysis.