Analysis of left-handed Z-DNA formation in short d(CG)n sequences in Escherichia coli and Halobacterium halobium plasmids. Stabilization by increasing repeat length and DNA supercoiling but not salinity.

To evaluate the relative importance of alternating d(CG) sequence length, DNA supercoiling, and salt in left-handed Z-DNA formation, plasmids containing short d(CG)n sequences (n = 3-17) with the capability of replicating in either Escherichia coli or the halophilic archaeum Halobacterium halobium were constructed. Z-DNA conformation in the d(CG)n sequences was assayed by (i) a band shift assay using the Z-DNA-specific Z22 monoclonal antibody (ZIBS assay); (ii) an S1 nuclease cleavage-primer extension assay to map B-Z junctions; and (iii) a BssHII restriction inhibition assay. Using the ZIBS assay on plasmids purified from E. coli, the transition from B-DNA to Z-DNA occurred from d(CG)4, to d(CG)5, with 20% of d(CG)4, and 90% of d(CG)5 in Z-DNA conformation. These findings were consistent with the results of S1 nuclease cleavage observed at B-Z junctions flanking d(CG)4 and d(CG)5 sequences. Resistance to BssHII restriction endonuclease digestion was observed only in supercoiled plasmids containing d(CG)8 or longer sequences, indicating that shorter d(CG)n sequences are in dynamic equilibrium between B- and Z-DNA conformations. When a plasmid containing d(CG)4, was isolated from a topA mutant of E. coli, it contained 25% greater linking deficiency and 40% greater Z-DNA conformation in the alternating d(CG) region. In plasmids purified from H. halobium, which showed 30% greater linking deficiency than from E. coli, 20-40% greater Z-DNA formation was found in d(CG)4-6 sequences. Surprisingly, no significant difference in Z-DNA formation could be detected in d(CG)3-17 sequences in plasmids from either E. coli or H. halobium in the NaCl concentration range of 0.1-4 M.

Experiments in the early 1970s showed that exposure of poly(d(CG)) DNA sequences to high salt concentrations resulted in inversion of the circular dichroism spectrum due to the formation of an unusual DNA structure (1). Subsequently, detailed structural analysis of a d(GC) hexamer in 4 M NaCl by x-ray crystallography (2) showed that the DNA forms a lefthanded helical conformation, with alternating glycosidic bonds in anti and syn conformation and zigzag tracking of the phosphate backbone (hence, the name, Z-DNA). Since these initial studies, many other investigations have confirmed the occurrence of left-handed Z-DNA using a variety of approaches, including immunological (3), enzymatic (4), chemical (5), and spectroscopic methods (for review, see Ref. 6).
The factors that influence the equilibrium between B-and Z-DNA in vitro have also been studied. The nucleotide sequence is important for Z-DNA formation (7)(8)(9); Z-DNA is generally adopted by alternating purine-pyrimidine sequences. Among these sequences, the d(CG) repeat has been shown to be the most favorable, the d(CA) repeat is intermediate, and the d(AT) repeat is the least favorable sequence for Z-DNA formation. Negative supercoiling of DNA is also important for Z-DNA formation (4,10,11), as the higher energy status of negatively supercoiled DNA may be relieved by opposite handed helix formation. Under high superhelical density, some sequences with imperfect purine-pyrimidine repeats adopt Z-DNA conformation (9,12). The presence of salts and some small molecules also affects Z-DNA formation (1). High salt concentration is thought to stabilize Z-DNA conformation by shielding the repulsion of negatively charged phosphate groups, which are closer in Z-DNA than in B-DNA. Finally, chemical modifications such as methylation (13) and bromination (14) also affect Z-DNA formation due to steric effects.
Studies directed at in vivo Z-DNA formation have provided evidence for left-handed DNA in the genomes of a variety of organisms, including plants, animals, and microorganisms (15)(16)(17)(18)(19)(20)(21)(22). No studies, however, have been directed at the genome of extremely halophilic microorganisms, for example Halobacterium halobium, which are classified as archaea and which grow at 3-5 M NaCl. It seemed likely that H. halobium may harbor a significant fraction of Z-DNA in its genome since its cytoplasm contains nearly 5 M salts. Moreover, its genomic DNA has a high GC content, 67%, and therefore, a high statistical likelihood of alternating d(CG) repeat sequences (23). We have also found, as part of this study, that plasmid DNA in H. halobium has a greater linking deficiency (and presumably negative supercoiling) than in Escherichia coli (24). These three factors together suggested that Z-DNA in the H. halobium genome may be quite prevalent and perhaps provides significant challenges and opportunities to normal genetic processes.
As an initial step in evaluating this hypothesis, we constructed a plasmid series containing short d(CG) repeats and capable of replicating in H. halobium and E. coli, and we systematically studied the importance of d(CG) repeat length, superhelical density, and NaCl concentration in Z-DNA formation. As documented in this report, increasing the length of d(CG) repeats and DNA supercoiling, but surprisingly not salt concentration, were found to promote Z-DNA formation.  (25). Culturing and transformation with Ca 2ϩ were done by standard procedures (26).
H. halobium NRC-1, a wild-type strain, was cultured in a medium containing 4.5 M NaCl at 37°C with illumination and shaking at 200 rpm (27). Transformation of H. halobium was done by the polyethylene glycol-EDTA procedure of Cline and Doolittle (28). Transformants were selected by plating on agar plates containing 10 g/ml mevinolin (a gift of Merck, Sharp, and Dohme Research Laboratory, Rahway, NJ). Mevinolin was added at the same concentration for growth in liquid culture.
Construction of d(CG) n -containing Plasmids-Polydeoxyribonucleotide d(CG) n (Sigma) was partially digested with BssHII restriction endonuclease (which cleaves at the recognition sequence GЈCGCGC) (restriction enzymes and polymerases were purchased from New England Biolabs or Bethesda Research Laboratories), and the 5Ј-overhanging ends were then converted to blunt ends by filling in with the large fragment of E. coli DNA polymerase I in the presence of dGTP and dCTP. After ligation with SmaI-digested pUC12, E. coli strain DH5␣ was transformed with the reaction mixture. Plasmids purified from transformants were digested with HindIII and EcoRI restriction endonucleases, and the approximate sizes of inserts were estimated by electrophoresis on 15% polyacrylamide gels. Several plasmids containing d(CG) inserts of different sizes were chosen for further analysis by DNA sequencing. DNA sequencing was carried out using the Sequenase kit (U. S. Biochemical Corp.) with the universal sequencing primer and dITP label mix to determine the exact length of the d(CG) stretch present in inserts. Plasmids containing various sizes of d(CG) n stretches from n ϭ 5 to 17 were obtained (29) and designated as pJKCGn, with n representing the number of d(CG) repeats. To construct a shorter d(CG) length insert, pJKCG5 was digested with BssHII, treated with S1 nuclease, and self-ligated to form pJKCG3, and pJKCG8 was digested with BssHII, and the larger fragment was selfligated to form pJKCG4. Plasmid DNAs were isolated by the alkaline lysis method followed by cesium chloride density gradient centrifugation in the presence of ethidium bromide (26).
For construction of E. coli-H. halobium shuttle vectors containing d(CG) n sequences, pUC12 and pJKCGn (n ϭ 3-7) plasmids were digested with HindIII, and the 5Ј-overhanging ends were filled in by the Klenow fragment of E. coli DNA polymerase I followed by SalI digestion. The 2.65-kb 1 fragments were isolated by electroelution after electrophoresis on a 0.8% agarose gel, ligated with SmaI-XhoI-digested pCY3 (24), and transformed into E. coli DH5␣. Plasmids isolated from E. coli transformants were analyzed by restriction mapping and transformed into H. halobium. Plasmids were purified from H. halobium by a previously published procedure (30). These plasmids were designated as pJKSUC and pJKSCG3, 4, 5, 6, and 7. Recombinant DNA procedures were carried out as described in Sambrook et al. (26).
Z-DNA Immunoband Shift (ZIBS) Assay-This assay utilized the well characterized Z22 monoclonal antibody to detect left-handed Z-DNA (a gift of Prof. B. D. Stollar, Tufts University Medical School, Boston) (31). The standard procedure involved incubation of plasmid (0.1 g) with 0.2 g of Z22 antibody for 2 h at 37°C in 100 l of 50 mM Tris⅐HCl (pH 7.5), 1 mM EDTA, and 0.1 M NaCl. After sequential addition of 2.25 l of formaldehyde and 2 l of glutaraldehyde (32) and incubation on ice for 15 min each, the reaction mixture was passed through a Sephadex G-50 spin column and lyophilized to remove aldehydes. The DNA-antibody complex was incubated with 10 units of HinfI in a 20-l reaction volume at 37°C for 2 h. The 3Ј-recessed ends were labeled by the addition of 10 l of 1 nmol of dGTP, dCTP, and dTTP; 1 pmol of [␣-32 P]dATP (10 mCi/ml, 3,000 Ci/mmol; Amersham Corp.); and 0.2 unit of Klenow fragment of E. coli DNA polymerase I. After incubation at 37°C for 1 h, the labeled restriction fragments were resolved by electrophoresis on 4% polyacrylamide gels and detected by autoradiography on x-ray film. The bands containing Z-DNA were shifted mostly to the origin of the gel, and the decrease of band intensity was used as the criterion for Z-DNA formation. Band intensity was quantified using a PDI densitometry system (Huntington Station, NY) running on a SUN Sparcstation IPC (Mountain View, CA). These results were confirmed in duplicate or triplicate. S1 Nuclease Assays-For S1 nuclease assays, plasmid DNAs (1 g) were treated with 0.1 unit of S1 nuclease in 20-l reactions containing 50 mM sodium acetate (pH 4.6), 50 mM NaCl, and 1 mM zinc acetate at 37°C for 10 min. After extraction with phenol twice and with chloroform once, DNA was concentrated by ethanol precipitation. DNA was dissolved in 10 l of deionized water, denatured by the addition of 10 l of 0.8 M NaOH and incubation at room temperature for 10 min, neutralized by the addition of sodium acetate (pH 5.2) to 0.45 M, and precipitated with ethanol. Primer extension was carried out at 37°C for 1 h in 20-l reaction mixtures containing 0.5 pmol of primer (either the forward universal primer or the reverse primer, previously labeled at the 5Ј-end with T4 polynucleotide kinase and [␥-32 P]ATP), 40 mM Tris-HCl (pH 7.5), 20 mM MgCl 2 , 50 mM NaCl, 0.05 mM dNTPs, and 0.5 unit of Klenow fragment. Reaction products were fractionated on 8% polyacrylamide, 8.3 M urea gels and subjected to autoradiography.
Restriction Inhibition Assay-For the standard restriction inhibition assay, 1.5 g of pJKCGn plasmid was incubated with 6 units of BssHII restriction endonuclease at 37°C in a 50-l reaction volume. Aliquots were removed at various time intervals, and the reaction was stopped by the addition of 50 mM EDTA and 0.1% sodium dodecyl sulfate. The digestion mixtures were fractionated by electrophoresis on 0.7% agarose gels and visualized by staining with ethidium bromide (26). The intensity of supercoiled DNA bands was quantitated using the PDI densitometry system.
DNA Supercoiling Assays-DNA supercoiling assays were carried out using pCY1, a 5.3-kb plasmid that had been constructed by ligation of two fragments, the mev gene fragment obtained from pNGMEV100 as a 3.5-kb BamHI fragment (33) and the SalI fragment of a 1.8-kb Halobacterium plasmid pGRB1 (34), and transformation of H. halobium. Plasmid DNA was first purified by CsCl ethidium bromide gradient centrifugation and agarose gel electrophoresis. Plasmid topoisomers were generated by treatment of pCY1 with wheat germ topoisomerase I (Promega, Madison, WI) in the presence of ethidium bromide. The reactions contained 4 g of pCY1, 4 units of topoisomerase I, 50 mM Tris-HCl (pH 7.6), 50 mM NaCl, 10 mM MgCl 2 , 0.1 mM EDTA, and 0 -12 g/ml ethidium bromide and incubated for 6 h at 30°C. The reaction mixtures were extracted three times with (10 mM Tris⅐HCl (pH 7.5) and 1 mM EDTA saturated with phenol and twice with chloroform, ethanol precipitated, and dissolved in H 2 O. Plasmid was electrophoresed on 1.0% agarose gels containing 0 -6 M chloroquine (35,36). Electrophoresis was carried out for 6 -18 h at 3-4 volts/cm with circulation of buffer (0.1 mM Tris ⅐base, 0.1 M boric acid, and 2 mM EDTA), and gels were stained with 0.5 g/ml ethidium bromide and destained before photography under uv illumination.

ZIBS Assay and Effect of d(CG) n Length on Z-DNA Formation in Supercoiled DNA
Isolated from E. coli-For quantitative analysis of Z-DNA formation within alternating d(CG) sequences, a plasmid series containing d(CG) n stretches of various lengths was examined by the ZIBS assay. For this assay, plasmids were incubated with Z22 monoclonal antibody specific for Z-DNA, and antibody-DNA complexes were cross-linked by treatment with aldehydes. The plasmids were then digested with HinfI restriction endonuclease, end-labeled using the Klenow fragment of DNA polymerase I and [␣-32 P]dATP, and electrophoresed on polyacrylamide gels. If there was Z-DNA formation in d(CG) stretches, antibody-HinfI fragments of 218 Ϯ d(CG) n -bp size complexes were shifted up to the origin of the gel. Reduction of band intensity was quantified by densitometry, and the change was used as a criterion for Z-DNA formation. Intensities of unshifted restriction fragments were used as controls for quantitation.
Using the ZIBS assay, we examined Z-DNA formation in d(CG) n stretches in plasmids isolated from E. coli DH5␣ ( Detection of B-Z Junctions Using S1 Nuclease-To rule out the possibility that Z22 antibody is promoting Z-DNA formation (37,38) in the ZIBS assay, we utilized an independent method involving S1 nuclease. S1 nuclease is known to cleave at single-stranded regions occurring at B-Z junctions (4), and the site of cleavage can be mapped at the nucleotide level by primer extension analysis. S1 nuclease-treated pJKCGn plasmid DNAs were denatured, annealed with 32 P-labeled sequencing primer (or reverse sequencing primer; data not shown), and then primers were extended using the Klenow fragment of DNA polymerase I and dNTPs (Fig. 2). Synthesized cDNAs were resolved by denaturing polyacrylamide gel electrophoresis together with sequencing reaction mixtures generated using the same primers. S1 cleavage was detectable at the ends of d(CG) n sequences in pJKCG4 and to a greater extent in pJKCG5 and pJKCG6 (lanes 3-5), reflecting Z-DNA formation. The degree of Z-DNA formation in the S1 nuclease-primer extension experiment was consistent with the ZIBS assay result, confirming that a d(CG) 4 stretch is the minimum length required for Z-DNA formation and that the Z-DNA formation observed in the ZIBS assay is not a result of induction or stabilization of left-handed conformation by antibody.
Restriction Inhibition Assay for Z-DNA Conformation-Activities of several restriction and modification enzymes are known to be sensitive to DNA conformation (39,40). It was shown that BssHII cleavage of a d(CG) 7 sequence is inhibited when present on a highly negatively supercoiled plasmid but not when relaxed (39), indicating that BssHII is sensitive to DNA conformation. Z-DNA formation in the pJKCGn plasmid series was examined by an inhibition assay using BssHII. For the standard restriction inhibition assay, 1.5 g of pJKCGn plasmids containing different lengths of d(CG) segments were incubated with 6 units of BssHII. Aliquots were taken after various time periods and electrophoresed in an agarose gel (data not shown). The amount of supercoiled molecules remaining was quantitated by densitometry and plotted as a function of time (Fig. 3). Under these conditions, pJKCG4 and pJKCG5 were both completely cleaved, but pJKCG4 was cleaved more slowly (30 min) than pJKCG5 (15 min). Partial resistance to cleavage was observed for pJKCG6, and the inhibition of BssHII cleavage increased for plasmids with increasing d(CG) n length. Plasmids containing d(CG) 8 or longer stretches were cleaved only partially even when excess enzyme was present Effect of NaCl Concentration on Z-DNA Formation-Earlier studies showed that high salt concentration induced B-Z transition in poly d(CG) nucleotides (1) and plasmids containing long stretches of d(CG) sequences (10,41). To determine the effect of NaCl on shorter stretches of d(CG) repeats, we treated plasmids that in the absence of salt were not in Z-DNA form (pJKCG3), or only partially (20%) in Z-DNA form (pJKCG4), with various concentrations of NaCl (0.1, 1, 2, 3, and 4 M) and carried out the ZIBS assay. The result in Fig. 4A showed that no detectable change in Z-DNA formation could be observed with increasing NaCl concentrations in either plasmid (lanes 1-10). The possibility of an inhibitory effect of salt on Z22 antibody binding was ruled out by the lack of reduction in antibody binding to pJKCG11 (Fig. 4B, lanes 1-5) (and also in pJKCG5 and pJKCG7; data not shown) at the higher NaCl concentrations. The effect of NaCl on more negatively supercoiled DNA isolated from the E. coli topA Ϫ mutant strain also showed no distinguishable salt effect on Z-DNA formation (see below). These results indicate that NaCl does not have a measurable effect on Z-DNA formation in plasmids at the superhelical densities occurring in E. coli.
To examine the effect of salt on d(CG) repeats in linear DNA fragments, pJKCGn plasmids were digested with HinfI before antibody binding, incubated with various concentrations of NaCl, and then assayed for Z-DNA formation using the band shift assay (Fig. 4B). None of the d(CG) n sequences up to n ϭ 17, including the d(CG) 11  Effect of DNA Supercoiling on Z-DNA Formation-Plasmid DNA of different superhelical densities were obtained from an E. coli topA ϩ strain, JTT, and an isogenic topA Ϫ strain, RS2 (25,42). Differences in linking deficiency of plasmids were observed in agarose gel electrophoresis containing 5 M chloroquine, indicating that plasmids isolated from the topA RS2 strain were 25% more negatively supercoiled than from the JTT strain (Fig. 5C). Plasmids pUC12 and pJKCG3 isolated from each strain showed no difference by the ZIBS assay (data not shown); however, 40% more Z-DNA formation was observed in pJKCG4 isolated from RS2 (Fig. 5, A and B, lanes 5-8) than JTT (lanes 1-4). These results confirmed the importance of DNA supercoiling on Z-DNA formation. However, the addition of NaCl up to 4.5 M concentration (lanes 4 and 8) did not promote the formation of Z-DNA in either pJKCG3 or pJKCG4, indicating that salt does not affect the B-Z equilibrium.
The effect of supercoiling on Z-DNA formation in the pJKCGn plasmids was also analyzed by BssHII restriction inhibition analysis. Plasmids isolated from E. coli JTT (topA ϩ ) and RS2 (topA Ϫ ) strains were treated with BssHII, and the intensity of supercoiled plasmid DNA was plotted against time (Fig. 6). Plasmids isolated from the topA Ϫ RS2 strain showed more resistance to cleavage than the same plasmids isolated from the topA ϩ JTT strain. Plasmid pJKCG4 isolated from the topA Ϫ strain showed a similar level of restriction inhibition as pJKCG6 isolated from the topA ϩ strain. Restriction inhibition of pJKCG5 from the topA Ϫ strain was inhibited to an extent similar to that of pJKCG8 isolated from the topA ϩ strain (29).
DNA Superhelical Density Is Highly Negative in H. halobium-The linking deficiency of plasmids was previously shown to be much more negative in H. halobium than in E. coli on average, but in several cases individual topoisomers were not resolved due to the large sizes of these plasmids (24). Therefore, we used the smallest available plasmid with the ability to replicate in H. halobium, pCY1, as a reporter of superhelical density inside H. halobium. Topoisomers of pCY1 were generated by treatment of the plasmid with topoisomerase I in the presence of various concentrations of ethidium bromide, and they were compared with natural pCY1 topoisomers isolated from NRC-1, by electrophoresis in agarose gels containing 0 -6 M chloroquine (Fig. 7). The results showed that pCY1, which is 5.3 kb in size, was composed of topoisomers with linking deficiency (Ϫ⌬Lk) of 36 -38, which would be manifested as 36 -38 negative superhelical turns. Assuming that pCY1 is entirely in B-form, the linking number (Lk 0 ) is 5,300/ 10.4 ϭ 510 when relaxed. Therefore, the total (constrained plus unconstrained) superhelical density, , is as follows.  (Fig. 7). To compare Z-DNA formation in plasmids of H. halobium with those of E. coli, plasmids capable of replication in both microorganisms and containing d(CG) n stretches were constructed. When Z-DNA formation of plasmids containing d(CG) 3 , d(CG) 4 , d(CG) 5 , and d(CG) 6 from both E. coli and H. halobium (Fig. 8, A and B, lanes 2-5 and 7-10) were compared by ZIBS assays followed by densitometric analysis (Fig. 7C), 20 -40% more Z-DNA formation was observed in d(CG) 4 , d(CG) 5 , and d(CG) 6 sequences from H. halobium, a result consistent with the importance of DNA supercoiling in promoting Z-DNA formation. DISCUSSION In this study, the requirements for Z-DNA formation were systematically examined using a plasmid series containing short stretches of alternating d(CG) sequence in supercoiled plasmids. First, we determined the minimum d(CG) n length required for Z-DNA formation using the ZIBS and other assays (Fig. 1). Our results showed that Z-DNA did not form in plasmids pUC12 and pJKCG3. About 20% of the d(CG) 4 insert in pJKCG4 and 90% of the pJKCG5 insert were in Z-DNA conformation. Longer d(CG) n stretches were found to be essentially completely in Z-DNA form. Our systematic analysis showed that the length of d(CG) n sequences is critical for Z-DNA formation and that d(CG) 4 is the minimum length required for Z-DNA formation at natural superhelical densities. These findings are consistent with previously published results showing that a 10-bp d(CG) sequence in a supercoiled plasmid from E. coli (44) and an 8-bp alternating purine-pyrimidine sequence in SV40 (45) can adopt Z-DNA conformation. They provide a framework for predicting Z-DNA formation in natural sequences (46).
Second, we determined that DNA supercoiling is necessary for Z-DNA formation in short alternating d(CG) sequences. Although d(CG) n sequences 10 bp or longer in supercoiled plasmids were largely in the Z-form, no Z-DNA formation could be detected when these plasmids were linearized, even in d(CG) n sequences up to 34 bp long ( Fig. 1 and data not shown). The effect of DNA supercoiling on Z-DNA formation was also analyzed using plasmids isolated from E. coli strains with wildtype and defective topoisomerase I (topA) genes (Fig. 3). No Z-DNA forming capacity was observed for pUC12 and pJKCG3 isolated from either strain (data not shown). However, a significant increase in Z-DNA formation was detected in more negatively supercoiled pJKCG4 isolated from the topA Ϫ strain than in pJKCG4 isolated from the topA ϩ strain (Fig. 5). These results demonstrated and confirmed (4, 10, 11) that DNA supercoiling plays a critical role in Z-DNA formation.
Third, the importance of salt concentration in promoting Z-DNA formation in alternating d(CG) sequences in supercoiled plasmids was studied. We focused on pJKCG4 and pJKCG5, which were known to contain sequences that partly adopt Z-DNA conformation at a low salt concentration and wild-type superhelical density. However, no enhancement of Z-DNA in supercoiled pJKCG4 and pJKCG5 was detected in NaCl concentrations up to 4 M using the ZIBS assay (Fig. 4). The lack of a measurable salt effect is not likely due to inhibition of Z22 antibody binding (6). No salt-enhanced Z-DNA formation was detected in highly supercoiled topoisomers of  1-4) and topA Ϫ strain RS2 (lanes [5][6][7][8] were analyzed by the ZIBS assay after incubation at NaCl concentrations of 0.1 M (lanes 1 and 5), 1.5 M (lanes 2 and 6), 3 M (lanes 3 and 7), and 4.5 M (lanes 4 and 8). Panel B, densitometric analysis of the d(CG) 4 -containing band, with lanes corresponding to lanes in panel A. Panel C, measurement of supercoiling density of pUC12 (lanes 1 and 2) and pJKCG4 (lanes 3 and 4) isolated from E. coli JTT (lanes 1 and 3) and RS2 (lanes 2 and 4) was compared by electrophoresis in 1% agarose gels containing 5 M chloroquine at 50 mV for 15 h. The lack of a measurable effect of salt on Z-formation was surprising given that d(CG) homopolymer (47) and d(CG) 3 nucleotide (48) have long been known to adopt Z-DNA conformation at high salt concentration. Moreover, salt is also known to increase twist, which will contribute to increasing DNA supercoiling and Z-DNA formation in covalently closed circular molecules (49). The lack of a salt effect on Z-DNA formation in short d(CG) n sequences in plasmids is at least partly due to the presence of flanking B-DNA stretches, which increase the activation energy necessary for Z-DNA formation (12). An additional factor may be that a long stretch of Z-DNA is stabilized by cooperative charge shielding between phosphate groups in the DNA backbone by salt, but the cooperativity is absent or insignificant in short sequences.
The effect of polycations on Z-DNA formation in linear polymers (50) was also examined in this study using the ZIBS assay (data not shown). However, we found no measurable increase in left-handed conformation in d(CG) n sequences in supercoiled plasmids by spermidine, spermine, and hexamine cobalt chloride. These results indicate that the presence of monovalent or polyvalent cations was not sufficient to convert short stretches of d(CG) n sequences to Z-DNA when they are flanked by B-DNA.
Since the ZIBS assay was carried out in the presence of a 10-fold molar excess of Z22 antibody, we were concerned about the antibody inducing or stabilizing Z-DNA conformation in pJKCGn plasmids. To determine if Z-DNA can form in the absence of antibody, several alternate Z-DNA detection methods not involving Z22 antibody were used. S1 nuclease (Fig. 2) and OsO 4 (data not shown) were used to assay for singlestranded DNA regions at predicted B-Z junctions flanking d(CG) n sequences. The finding of cleavage by S1 nuclease and OsO 4 at both ends of alternating d(CG) sequences in pJKCG4 and pJKCG5 was consistent with Z-DNA formation. Greater cleavage of pJKCG5 than pJKCG4 was consistent with the results of ZIBS assays.
Z-DNA conformation in d(CG) n -containing plasmids was also analyzed by the BssHII restriction inhibition assay. BssHII cleaves at d(GC) 3 sequences in the B-form but not in the Z-form (39,40). Plasmids that are partly in Z-DNA conformation such as pJKCG4 and pJKCG5 were completely cleaved by BssHII, suggesting that the alternating d(CG) sequences rapidly isomerize from Z-to B-form. Interestingly, faster cleavage was observed for pJKCG5 than for pJKCG4, in spite of more Z-DNA formation in the former, as revealed by by ZIBS and S1 nuclease. This may be due to the presence of two overlapping BssHII recognition sites in pJKCG5 which provides a larger target for the restriction enzyme compared with pJKCG4. However, with longer stretches of alternating d(CG) sequences, we observed greater inhibition of BssHII cleavage, confirming the occurrence of more Z-DNA conformation. Using the restriction inhibition assay, we also compared Z-DNA formation in pJKCG5 and pJKCG6 isolated from isogenic E. coli topA Ϫ and topA ϩ strains. No difference in Z-DNA formation had been detectable by the ZIBS assay as both were almost completely in Z-DNA conformation. However, a very significant difference in cleavage rates, and therefore, in the rates of Z-to B-DNA isomerization could be detected by the restriction inhibition assay. In fact, more negatively supercoiled pJKCG5 plasmid isolated from the topA Ϫ strain showed greater resistance to cleavage than less negatively supercoiled pJKCG6 plasmid isolated from the topA ϩ strain. This result suggested that the greater supercoiling in the topA Ϫ strain, about 25% greater linking deficiency compared with the topA ϩ strain, stabilizes Z-DNA conformation more than one extra d(CG) dinucleotide.
Restriction inhibition analysis of pJKCG4, pJKCG5, and pJKCG6 provided a rough estimate of the rates of B-Z transitions. The approximate time for 50% cleavage of the plasmids is in the range of 5 min to 2 h, which is consistent with the rate of B-Z transitions measured in vivo by Lukomski and Wells using a methylation assay (51). In contrast, the resistance to cleavage shown by plasmids containing longer than d(CG) 8 sequences indicated that these sequences are rarely in B-form. Thus, it may be concluded that DNA sequences in the range from d(CG) 4 to d(CG) 7 can exist in both B-and Z-forms at E. coli topA ϩ superhelical density, and sequences in the range from d(CG) 4 to d(CG) 6 can exist in both forms at more negatively supercoiled condition present in the E. coli topA Ϫ strain and H. halobium (Fig. 7). d(CG) stretches longer than 14 bp are probably frozen in Z-DNA conformation. We would predict that natural alternating d(CG) sequences within the 8 -14-bp range may have been recruited for functions involving structural isomerization between B-and Z-DNA. In this context, it is interesting to note that a large number of natural sequences capable of forming Z-DNA-like structure have been found in the H. halobium genome, 2 including an 11-bp alternating pu-2 J.-K. Kim and S. DasSarma, submitted for publication. A) and H. halobium (panel B) by the ZIBS assay. Lanes 1-10 contain pJKSUC (lanes 1 and 6), pJKSCG3 (lanes 2 and 7), pJKSCG4 (lanes 3 and 8), pJKSCG5 (lanes 4 and 9), and pJKSCG6 (lanes 5 and 10), respectively. Panel C shows the relative intensity of the d(CG) ncontaining band in corresponding lanes. rine-pyrimidine sequence centered 23 bp 5Ј to the bop gene transcription start site which may be required for promoter activity (24). A search of the bacterial portion of GenBank showed the occurrence of more than 3,000 sequences of d(CG) 4 length, about 200 sequences of d(CG) 5 , 10 sequences of d(CG) 6 , and no sequences of d(CG) 7 or longer.

FIG. 8. Comparison of Z-DNA formation in d(CG) n -containing E. coli-H. halobium shuttle plasmids isolated from E. coli (panel
In summary, we have carefully examined the relative role of d(CG) n length, DNA supercoiling, and salt concentration on Z-DNA formation. Among these factors, effects of repeat length and DNA supercoiling on Z-DNA formation were much more more pronounced than salt concentration. Therefore, high salt concentration present in H. halobium may not contribute directly to the stabilization of Z-DNA in natural sequences. However, slightly more Z-DNA formation can be observed in plasmids isolated from H. halobium than in E. coli, most likely reflecting the greater negative DNA supercoiling density in H. halobium. Our findings provide a better framework for understanding Z-DNA formation in a wide variety of organisms, especially those with d(CG)-rich genomes.