INTRODUCTION

Brief segments of relaxed double-stranded DNA (dsDNA)¹ can assume conformations, such as bending, which are either intrinsic, i.e. imposed solely by primary sequence or the result of protein binding. Intrinsic bending has been implicated in a variety of biological processes such as transcriptional regulation, recombination, and replication (1, 2, 3, 4, 5, 6, 7). A characteristic feature of intrinsically bent DNA is a series of several helically phased A_4-6 tracts. In a kinetoplast DNA bend (K-DNA) (8, 9) four A_5-6 tracts cause intrinsic curvature of 17-21° per tract resulting in 68-84° overall curvature (9, 10). Similarly, in the intrinsically bent phage lambda DNA origin of replication (-ori), four A_4-5 tracts are helically phased. The properties of intrinsically bent DNA embedded in larger restriction fragments have been studied primarily by physical methods, such as gel electrophoresis and electron microscopy, whereas the conformation of small synthetic molecules has been characterized by x-ray crystallography. Although these studies have deciphered some of the sequence requirements for intrinsic bending, as well as some of its structural features, the precise conformation of intrinsically bent DNA remains in question. In particular, it is not known to what extent bending involves DNA strand separation, a feature that may facilitate protein binding. Furthermore, most of the published evidence indicates that bending occurs incrementally between consecutive base pairs of the A_4-6 tracts (see Refs. 11 and 12 for reviews), whereas x-ray studies of synthetic molecules indicate a uniformly rigid conformation for the A_4-6 tracts as a block (13, 14, 15).

If DNA bending causes localized strand separation, the resulting non-canonical DNA conformation should be a preferred target for single-stranded DNA-specific endonucleases such as S1 and T7 endonuclease I. These enzymes have been extensively used as probes of localized structures such as cruciforms and Z-DNA under supercoil stress both in vitro and in vivo (7). However, the usefulness of these enzymes is limited by a background nucleolytic activity toward nearly random sites (16, 17, 18). With supercoiled DNA, the fast rate of cleavage of the prominent cruciform or Z-DNA targets overcomes this limitation. In contrast, background activity is mostly observed with relaxed DNA, and the location of the few preferred sites is difficult to determine. An exception is an engineered endonuclease (H144) consisting of Lac repressor (lacI) and T7 endonuclease I. This enzyme maintains the specific repressor and nuclease activities of its components both in vitro and in vivo (17, 18). H144 cleaves cruciforms on supercoiled DNA with specificity and kinetics essentially identical to those of T7 endonuclease I. However, unlike T7 endonuclease I, H144 preferentially binds at lacO and cleaves relaxed dsDNA at sites located within a 100-1,500-bp range. The cleavage preference by H144 at a particular site depends on the prominence and location of the target relative to lacO. This specificity is practically lost, reverting to that of T7 endonuclease I, when the operator is not present on the same fragment as the targets. Therefore, the net effect of the lacI domain in H144 is to lower the background activity by sequestering the T7 endonuclease I domain to the vicinity of lacO. Here, we report that the five non-canonical DNA structures probed by H144 on relaxed dsDNA coincide with DNA regulatory sequences and intrinsically bent structures.

MATERIALS AND METHODS

Plasmids pCP144 and pCP82 have been described (16, 17). Plasmid pCP64 carries a kanamycin resistance gene and a region encompassing base pair coordinates 37,401 and 41,740 of bacteriophage lambda DNA that includes the origin of replication (-ori). These plasmids served as the parents of the plasmids used in these studies.

The K-DNA bend from Leishmania tarentolae (L. tarentolae) (8, 9) was chemically synthesized and polymerase chain reaction-amplified before subcloning into pCP82 to generate pCP230. The -ori bend (19) was prepared by splicing the -ori region from pCP64 into pCP82. The correctness of the constructs was determined by restriction analysis and DNA sequencing. An exact description of genetic engineering manipulations is provided elsewhere (18).

H144 was purified by ammonium sulfate fractionation of a cleared lysate of induced Escherichia coli harboring pCP144 as described (17) with the following modifications of the chromatography steps. H144 was loaded on a CM-cellulose cartridge (CM1010, Millipore) in buffer A (5 mM EDTA, 0.5 mM DTT, 5% glycerol, 50 mM MOPS, pH 7.2) and eluted at 200 mM NaCl using a salt gradient. Fractions containing H144 were identified by assaying for preferential and specific binding and cleavage of a DNA fragment carrying lacUV5 in a mixture of restriction fragments, as described (17). H144 was 50% pure, as judged by SDS-PAGE stained by Coomassie Brilliant Blue R-250 or Silver. Fractions containing H144 were pooled and further fractionated on a phosphocellulose P11 column (Whatman), in buffer P11 (40 mM sodium phosphate, pH 7.0, 40 mM NaCl, 5 mM EDTA, 0.5 mM DTT, 5% glycerol), using a salt gradient. H144 eluted at approximately 450 mM NaCl and was approximately 90% pure. As a final purification step, hydrophobic interaction chromatography was performed on a 1-ml phenyl-Superose FPLC column (Pharmacia Biotech, Inc.). Pooled fractions were adjusted by progressive addition of granular ammonium sulfate to 1 M. A linear reverse gradient was applied to 20 mM NaCl in buffer A. H144 eluted at 0.5 M ammonium sulfate and was better than 90% pure, as judged by SDS-PAGE followed by Coomassie staining (data not shown). The peak fractions were dialyzed against 500 volumes of 50 mM Tris, pH 7.0, 50% glycerol, 20 mM NaCl, 2 mM EDTA, 0.1 mM DTT, and stored at 20 °C. Three mg of purified H144 were obtained from 40 g of induced cells.

The concentration of the H144 preparation used in this study was 0.84 mg/ml (16 µM) as determined by the Bradford protein assay (Bio-Rad). H144 was analyzed by gel filtration chromatography on a Superdex-75 column (Pharmacia) in 50 mM Tris, pH 8.0, 100 mM NaCl. H144 eluted as a single peak corresponding to an apparent molecular weight of 58,000 (data not shown). This figure is close to 52,500 calculated for an H144 monomer from its amino acid composition. Higher order multimers or aggregates were not detected. Thus, at least when not bound to lacO DNA, H144 is monomeric.

H144 was also analyzed by reverse phase chromatography. H144 (24 µg) was loaded on a 1-ml C-8 column (Poros IIR-M) in 0.1% trifluoroacetic acid, 10% acetonitrile and was eluted using an acetonitrile gradient. H144 eluted as a single peak at 42% acetonitrile (data not shown). No other peaks were detected, indicating that H144 is more than 95% pure and free of proteolytic fragments. H144 recovered from the C-8 column was analyzed by mass spectrometry, where it displayed a molecular size of 52,430 ± 104. This figure is within experimental error of the 52,511 value calculated from the amino acid composition (data not shown).

Reactions with H144 and DNA were performed in 50 mM Tris, pH 7.5, 50 mM NaCl at 37 °C. Where indicated 1 mg/ml BSA was also included. The reactions were initiated by the addition of MgCl₂ to 10 mM and terminated either by the addition of 0.2 M EDTA, pH 8.0, to 20 mM final concentration, or by the addition of 0.1 volume of Stop buffer (50 mM Tris, pH 8.5, 50 mM borate, 125 mM EDTA, 43.5% glycerol, 0.05 mg/ml bromphenol blue). When desired, stopped reactions employing H144 were heated to 75 °C for 5 min to dissociate hybrid protein from lacO-DNA.

In order to map H144 target sites with base pair resolution, restriction fragments were specifically radiolabeled at the 3 ends by DNApol large fragment (Klenow, New England BioLabs) in the presence of the appropriate -³⁵S-dNTP (DuPont NEN) or at the 5 ends with polynucleotide kinase (New England BioLabs) in the presence of [-³²P]ATP (DuPont NEN). Fragments labeled with polynucleotide kinase were first dephosphorylated with alkaline phosphatase (Boehringer Mannheim). The radiolabeled DNA was reacted with H144, analyzed on 6% PAGE, and PAGE, 7 M urea strand-denaturing gels, and visualized by autoradiography. An M13 sequencing ladder size marker was generated as recommended (Sequenase 2.0; U. S. Biochemical Corp.).

RESULTS

To test the activity of H144 toward linear intrinsically bent DNA, a mixture of restriction fragments from plasmid pCP230 that carries lacO and an intrinsic bend from kinetoplast DNA (K-DNA) (8, 9) was reacted with H144. Restriction digestion of pCP230 by NheI plus PvuII generated three fragments, the smallest of which (681 bp) carried K-DNA ~130 bp away from lacO (Fig. 1a). Incubation of these fragments with H144 for increasing periods led to preferential cleavage of the 681-bp fragment to two fragments, 650- and 80-bp apparent size, respectively (Fig. 1b). Two of the substrate fragments (2311 and 681 bp) had been ³⁵S-labeled selectively at the NheI site. Of the products generated by H144 only the 80-bp fragment was visible by autoradiography (Fig. 1c). Since filling-in at the NheI site ³⁵S labels only one strand, further analysis by electrophoresis on high resolution denaturing gels revealed that H144 generated an 85-bp ³⁵S-labeled single-stranded DNA fragment by nicking 85 bp away from the filled-in NheI site (Fig. 1d). When the opposite strand was ³²P-labeled with polynucleotide kinase, a single major cleavage product was observed, the result of nicking 79 bp away from the NheI site (data not shown). Since the two major nicks on the upper and lower strands were staggered by 6 bp, combined cleavage at both sites must have generated the double-stranded DNA products observed on nondenaturing gels.

Assignment of the nicking sites on the DNA sequence showed that the two nicks that H144 introduces on the 681-bp fragment fall within the K-DNA bend (Fig. 1e). The nick on the upper strand occurs within one of the four helically phased A-tracts, whereas the nick on the lower strand is located at the position corresponding to the proposed physical center of the bend (8).

The Intrinsically Bent Locus of the  Origin of Replication Is Also Cleaved Preferentially by H144

To confirm and better understand the preference of H144 for intrinsic bends, a second DNA that displays intrinsic curvature was tested. The bending locus of the lambda origin of replication (-ori) was chosen, because it does not share common sequences with the K-DNA bend apart from the bend-inducing A_4-5 tracts (Fig. 2a). Moreover, the physical center of bending for this sequence is also known (20).

Preferential cleavage by H144 at -ori and at a T7 late promoter.

Fig. 2, panels b and c, shows the activity of H144 toward three restriction fragments of pCP212, one of which (577 bp) carries -ori approximately 200 bp downstream of lacO. Four prominent new fragments were generated by H144, and their apparent sizes indicated fragmentation of the 577-bp substrate fragment at three sites (Fig. 2b). Two of these products, 280 and 150 bp in length, were also visible by autoradiography when the substrate fragments had been ³⁵S-labeled at the NheI site (Fig. 2c). By further restriction analysis of end-labeled molecules, Southern hybridization, and high resolution denaturing gel electrophoresis, the exact positions of the nicks that generated these fragments were determined, as described for the K-DNA bend (data not shown). Assignment of the nicking sites on the DNA sequence confirmed that H144 cleaves the 577-bp fragment at three sites. The first two are located within -ori. The nicks at positions 265 and 278 on the upper strand and positions 268 and 278 on the lower strand fall within the GC-rich region of iterons III and IV (Fig. 2e). This region encompasses the proposed physical center of bending at -ori (20). Thus, H144 preferentially cleaves a second intrinsically bent DNA sequence at sites that correspond to the physical center of bending. Parenthetically, two less prominent bands are also seen in Fig. 2c, lane 2, corresponding to approximately a 580- and a 380-bp labeled fragment. The targets that give rise to these fragments must be located very near the EcoRV site, but their exact locations have not been determined.

The third site at which the 577-bp fragment is cleaved by H144 is not located within the -ori region. Instead, inspection of the DNA sequence showed that cleavage occurred 130 bp away, within a region specifying a T7 late promoter. Nicking occurred at two positions at coordinates 139 and 149 of the upper strand and at a single position at coordinate 145 of the lower strand (Fig. 2d).

The finding of preferential cleavage within a T7 promoter prompted us to examine more closely the locations of the targets for H144 that had been previously assigned within the region of the -lactamase promoter (16). Mapping these targets within a 475-bp restriction fragment (Fig. 3a) was accomplished by DNA labeling and gel analysis (Fig. 3, b and c). The precise location of these targets was determined by high resolution denaturing gel electrophoresis using the methodology described above for the targets located within the K-DNA bend (data not shown). The results show that H144 nicks the substrate DNA at five major locations (Fig. 3d). Two nicks separated by 4 bp on opposing strands span the leftward transcription initiation nucleotide; two others separated by 6 bp on opposing strands are located within the 35 element of the -lactamase promoter at positions 40 and 45 (21); the fifth prominent nick is in the exact same position as that introduced by gyrase (22). Interestingly, in contrast to the nicks mapped within the T7 late promoter, no nicks were mapped at the 10 region (TATA box) of the -lactamase promoter, where strand separation in response to protein binding and the formation of the transcription initiation complex occurs (see ``Discussion'').

Mapping of H144 targets in the -lactamase promoter region.

In supercoiled dsDNA, localized non-canonical conformations are more or less preferred targets for single-strand specific endonucleases. The preferential cleavage of dsDNA by H144, a nuclease that has a marked preference for ssDNA, suggested the presence of single-strandedness at the targets. Even though the results described above were obtained with relaxed rather than supercoiled dsDNA, the target sequences were examined for high A + T content, inverted repeats, alternating purine/pyrimidine and pyrimidine stretches that specify ``breathing'' sites, cruciforms, Z-DNA, and triple-stranded DNA, respectively. No such motifs were found, indicating that even if single-strandedness is the structural feature recognized by H144, it does not result from such structures.

In a search for other common patterns, the sequences (20 bp) surrounding the 14 nicking sites were aligned (Table I). As seen in the lower part of Table I, no individual nucleotide is conserved at any particular position, with 8/14 hits being the most frequent occurrence. Even though 5 of the 14 sequences encompassing the targets share the degenerate pentanucleotide CTA(A/T)(A/T) immediately 3 to the nick (positions 11-13), the same sequence in other locations of the substrate DNA was not nicked at detectable levels (18). Therefore, the particular DNA sequence responsible for preferential nicking by H144 is not evident, even though sequences analogous to the CTA(A/T)(A/T) pentanucleotide may be favored. Additional sites would have to be identified and analyzed, in order to decipher the DNA sequence context(s) that dictates the target conformation.

Table I. Alignment of H144 target sequences The 10 nucleotides on each side of the slash (/) introduced by H144 at each of the 15 major targets are compared A, T, G, or C refer to the total number of each particular nucleotide occurring at each position. Nucleotide sequence, 5  3 1 G A A G C A T T T A / T C A G G G T T A T 2 G A C A A T A A C C / C T G A T A A A T G 3 A T T T G A A T G T / A T T T A G A A A A 4 G T T T A T T T T T / C T A A A T A C A T 5 A T A C G A C T C A / C T A T A G G A G A 6 C T A T A G G A G A / A C C T T A A G G T 7 T A A G G T T C T C / C T A T A G T G A G 8 A G A A A A A T A A / A C A A A T A G G G 9 A A A T G T C A A A / A A A T A G G C A A 10 C A T T T T T T G C / C T A T T T T T T G 11 C C T C A A A A C G / A G G G A A A A T C 12 G G G G A T T T T C / C C T C G T T T T G 13 G G A A A A T C C C / C T A A A A C G A G 14 C T C G T T T T A G / G G G A T T T T C C  A 4 5 7 3 7 6 4 4 3 5 / 5 1 8 5 8 4 6 4 6 3  T 1 5 4 5 2 7 7 8 4 2 / 1 7 2 6 4 5 5 4 4 3  G 5 3 1 4 4 1 1 0 3 2 / 1 2 3 2 2 5 2 4 3 6  C 4 1 2 2 1 0 2 2 4 5 / 7 4 1 1 0 0 1 2 1 2

Finally, to eliminate the possibility that the DNA during preparation was somehow selectively modified in vivo at or near the nicking sites, we used as substrates for H144 DNA fragments synthesized in vitro by polymerase chain reaction. Identical results were obtained with these synthetic substrates (data not shown), indicating that DNA modification does not play a role in target recognition.

Overall, three functional DNA sites, a bacterial promoter, a phage promoter, and a gyrase contact site, were found to be preferred targets for H144, in addition to the two intrinsic bends, at least one of which (-ori) is also a functional site for DNA replication. Since only a few target sequences were detected per several hundred bp, their common non-canonical structural feature is not encountered very frequently on dsDNA.

DISCUSSION

H144 was found to cleave relaxed dsDNA at sites that are known to interact with particular proteins. The -lactamase and T7 late promoters interact with their respective polymerases, and the gyrase site is nicked by bacterial gyrase. Intrinsic bends are also thought to be preferred sites for topoisomerase action (23, 24), and the -ori interacts with O-protein. Therefore, all the sequences targeted by H144 share a common property; they interact with DNA-binding proteins. The fact that the sites cleaved by H144 are also contact sites for various regulatory proteins suggests that the target structure which is present in linear DNA, in the absence of supercoil pressure, is of physiological significance. Altered (non-B) sequence-dependent conformations at these functional sites may serve as recognition signals and/or entry points for these proteins.

T7 endonuclease I cleaves single-stranded DNA much faster than relaxed, dsDNA. This enzyme also shows strong specificity toward cruciforms, branched DNA, and even single base pair mismatches, i.e. structures that expose small localized single-stranded regions in dsDNA. T7 endonuclease I cleaves cruciform DNA 2-4 bp away from the single-stranded loop and base regions (25).² In branched dsDNA, cleavage by T7 endonuclease I and other single-stranded endonucleases occurs 1-10 bp away from the branch point at sites that do not conform to a strict pattern with respect to distance from the branch point, strand, or sequence preference (25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35). The same pattern emerges from probing synthetic, mismatched DNA, where strand scission occurs at variable lengths from the mismatch (36). Therefore, cleavage by these enzymes does not occur at the single-stranded regions expected to be exposed by these localized structures but, instead, a few bp away.

One explanation for this phenomenon is that the binding and catalytic sites of T7 endonuclease I are physically separated, so that binding occurs at single-stranded regions and cleavage ensues a few bp away at sites that may only be partially single-stranded. Alternatively, however, it is also possible that the enzyme recognizes and cleaves its specific substrate conformation through a single binding and catalytic site. In this case, single-stranded DNA may assume the substrate conformation by freedom of rotation, whereas the constraints of localized single-strandedness in cruciforms, branched, and mismatched DNA may instead expose the substrate conformation at a nearby site. The published evidence is consistent with either one of these mechanisms. Therefore, the exact points of cleavage (targets) by the T7 endonuclease I domain of H144 are not necessarily single-stranded, even though they are most likely located nearby a region that can easily assume single-strandedness. The nicks mapped within the T7 late promoter occur within a region of strand separation in response to T7 RNA polymerase binding (37, 38). In contrast, no nicks were mapped at the 10 region (TATA box) of the -lactamase promoter, where strand separation occurs in response to protein binding and the formation of the transcription initiation complex.

Regardless of the exact conformation of the targets, the evidence presented here shows that in linear dsDNA, protein contact sites and bends are selectively cleaved by H144. In bent DNA, the existence of single-strandedness has been postulated by molecular modeling (39) and, more recently, deduced from biophysical measurements (40). The results obtained with an enzymatic probe (H144) are consistent with strand opening within intrinsic bends.

Other enzymes, such as the mung bean, P1, and S1 nucleases that preferentially cleave ssDNA, have been used to probe for single-stranded regions within intrinsically bent DNA, without success (19, 41, 42, 43). In addition, the ssDNA probe bromoacetaldehyde also failed to detect an intrinsic bend (42). Even T7 endonuclease I does not cleave bent DNA with sufficient preference to allow detection over the background (18). Similarly, H144 loses its specificity for bent DNA in the absence of lacO. Thus, the usefulness of H144 as an enzymatic probe of localized non-canonical structures stems from the fact that the LacI domain directs the protein to lacO and, thus, limits nonspecific interactions. As a result, H144 preferentially cleaves intrinsic DNA bends, as well as other specific targets that coincide with DNA sites where proteins act. An important aspect of the evidence presented here is that preferential cleavage occurred in the absence of supercoiling or cognate protein binding. This finding implies that DNA/protein contact sites share a structural feature common to bent DNA that may serve as a signal for protein binding.