The biology of left-handed Z-DNA.

DNA encodes biological information in at least two different ways. First, through the linear order of nucleotides, it specifies the composition of proteins. Second, through its shape, DNA can provide information that is used directly or indirectly by a variety of macromolecules to regulate the assembly of cellular machines. DNA is capable of assuming many shapes (1). One of the most dramatic changes in shape is that which occurs in going from the familiar right-handed B-DNA double helix to the slightly thinner and elongated left-handed Z-DNA conformation (2). This conformational change occurs most readily in segments with specialized sequences, favored largely by alternations of purines and pyrimidines, especially alternating deoxycytidine and deoxyguanosine residues (3–5). A requirement for specialized nucleotide sequences also appears to be true for other unusual DNA structures such as the TATA box, which undergoes a major distortion in shape when bound by the TATA binding protein (6, 7). The biological role of non-B-DNA structures is an area of active study. The aim of these investigations is to determine which alternate DNA conformations exist in vivo, how their formation is regulated, and what information they convey. Here we will review recent studies that bear on the role of Z-DNA in biologic systems.

DNA encodes biological information in at least two different ways. First, through the linear order of nucleotides, it specifies the composition of proteins. Second, through its shape, DNA can provide information that is used directly or indirectly by a variety of macromolecules to regulate the assembly of cellular machines. DNA is capable of assuming many shapes (1). One of the most dramatic changes in shape is that which occurs in going from the familiar right-handed B-DNA double helix to the slightly thinner and elongated left-handed Z-DNA conformation (2). This conformational change occurs most readily in segments with specialized sequences, favored largely by alternations of purines and pyrimidines, especially alternating deoxycytidine and deoxyguanosine residues (3)(4)(5). A requirement for specialized nucleotide sequences also appears to be true for other unusual DNA structures such as the TATA box, which undergoes a major distortion in shape when bound by the TATA binding protein (6,7).
The biological role of non-B-DNA structures is an area of active study. The aim of these investigations is to determine which alternate DNA conformations exist in vivo, how their formation is regulated, and what information they convey. Here we will review recent studies that bear on the role of Z-DNA in biologic systems.

Formation of Z-DNA in Vitro
The existence of Z-DNA was first suggested by optical studies demonstrating that a polymer of alternating deoxyguanosine and deoxycytidine residues (d(CG) n ) produced a nearly inverted circular dichroism spectrum in a high salt solution (8). The physical reason for this finding remained a mystery until an atomic resolution crystallographic study of d(CG) 3 revealed a left-handed double helix, which maintained Watson-Crick base pairing (2). The Z-DNA helix is built from a dinucleotide repeat with the deoxycytidines in the anti conformation while deoxyguanosines are in the unusual syn form. In Z-DNA, there is a single narrow groove that corresponds to the minor groove of B-DNA. There is no major groove. Instead, the "information"-rich residues that allow sequence-specific recognition of B-DNA lie exposed on the convex outer surface of Z-DNA (Fig. 1). The transition from B-DNA to Z-DNA involves "flipping" the base pairs upside down. During this process, deoxycytidine remains in the anti conformation because both the sugar and base rotate, while only the base of deoxyguanosine inverts, moving it into the syn conformation. As a consequence, the backbone follows a zigzag path, giving rise to the name Z-DNA. Z-DNA can form from B-DNA under physiological salt conditions when deoxycytidine is 5-methylated (9).
The demonstration that Z-DNA formed under conditions of negative superhelical stress raised considerable excitement as this brought the left-handed conformation within the realm of biology (3,5,10).
Stabilization of Z-DNA by negative supercoiling illustrates a number of features. First, Z-DNA is a higher energy conformation than B-DNA and will only form when plasmids are torsionally stressed. The energy necessary to stabilize Z-DNA can be determined by measuring the plasmid superhelical density at which Z-DNA formation occurs, and it is proportional to the square of the number of negative supercoils (11)(12)(13)(14). Second, sequences other than alternating purines and pyrimidines can form Z-DNA. The ease with which this occurs depends on the sequence; d(CG) n is best, d(TG) n is next, and a d(GGGC) n repeat is better than d(TA) n (13,15,16). Third, formation of B-Z DNA junctions, each of which has a free energy ⌬G near ϩ4 kcal/mol, is a significant energetic barrier to Z-DNA formation (11).
Based on these empirical findings, computer models have been developed to rank the Z-DNA-forming potential of naturally occurring sequences (15,17,18). One analysis of 137 fully sequenced human genes demonstrated that sequences which could form Z-DNA easily were present in 98 and that they were distributed nonrandomly throughout the gene; sequences were 10 times more frequent in 5Ј than in 3Ј regions (17). This fits with the expectation that the energy necessary to form Z-DNA in vivo is generated by transcription. As demonstrated by Liu and Wang (19), negative supercoils arise behind a moving RNA polymerase as it ploughs through the DNA double helix. The torsional strain generated by passage of RNA polymerases then becomes a potent source of energy to stabilize Z-DNA.

Z-DNA in Prokaryotic Systems
A number of experiments have been used to demonstrate that Z-DNA can form in vivo. One approach uses chemical modification. Through use of either osmium tetroxide or potassium permanganate, it can be demonstrated that plasmids containing a d(CG) n insert will form Z-DNA in vivo (20,21). UV cross-linking of bacteria treated with psoralens have confirmed these results and made possible a precise measurement of the amount of unrestrained supercoiling present within Escherichia coli (22). A more sophisticated approach has used a construct in which an EcoRI site is embedded in a Z-DNAforming sequence (23)(24)(25). In the bacterial cell, this fragment can be methylated when it is in the B-DNA conformation, but it becomes resistant to methylation while in the Z-DNA conformation. Susceptibility to methylation thus can be used as a measure of in vivo torsional strain. Results obtained with this system show that Z-DNA formation in E. coli occurs in the absence of external perturbation and is regulated by transcription, an effect that is enhanced by mutations inactivating topoisomerase I (24,25). Formation of Z-DNA, however, was not observed in Morganella, Klebsiella, or Enterobacter (25).

Z-DNA in Eukaryotic Systems
It has been difficult to directly demonstrate the existence of Z-DNA in eukaryotic systems. A number of early observations clearly suggested its existence. Unlike B-DNA, Z-DNA is highly immunogenic, and polyclonal as well as monoclonal antibodies can be made that recognize this conformation (26). The first suggestion that Z-DNA was found in eukaryotic systems came from work with humans. Analysis of sera obtained from patients with autoimmune diseases, especially lupus erythematosis, showed that these patients produced antibodies which were highly specific for Z-DNA (27). These were produced during the exacerbations of the disease, together with antibodies to many other nuclear components.
Antibodies raised in rabbits and sheep were used in staining experiments with both fixed (28) and unfixed polytene chromosomes of Drosophila (29). These produced an unusual pattern with staining in the interband regions but not in the bands. Staining was especially intense in the puffs, which are associated with high levels of transcriptional activity (reviewed in Ref. 30). Antibodies were also used in staining ciliated protozoa that have both a macronucleus and a micronucleus (31). The micronucleus is used for genetic reproduction, but the macronucleus is the site of all transcriptional activity. Here, again, the macronucleus stained exclusively, with no staining in the micronucleus. Both of these early experiments suggested somewhat indirectly a link between transcriptional activity and the presence of Z-DNA.
Analysis of intact mammalian systems has been more complicated. There are a number of limitations in these experiments. No phenotype has been associated with the presence or absence of Z-DNA-forming sequences, thus limiting the use of genetic approaches. In addition, regulation of Z-DNA is likely to be very complex. For example, what is the importance of the three RNA polymerases relative to production of Z-DNA? It is known that RNA polymerase I works on some favorable Z-DNA-forming sequences in ribosomal RNA genes. In addition, it is not known how the torsional strain in regions 5Ј to RNA polymerase II promoters is regulated. What is the influence of the TATA box sequence bound to its proteins? Are genes lacking a TATA box more topologically sensitive to the torsional strain generated by the moving polymerase? In this context, the effect of potential Z-DNA-forming sequences upstream of a promoter must be interpreted carefully; deletion or mutation of such regions, as in the case of the SV40 enhancer which has regions of alternating purine/pyrimidine repeats, may have many different interpretations (32)(33)(34)(35).
A number of experiments have been carried out using metabolically active permeabilized mammalian nuclei, which were formed by embedding intact cells in agarose using the method of Jackson and Cook (36). Here, low concentrations of detergent are used to lyse the cytoplasmic membrane and permeabilize the nuclear membrane. These nuclei have been shown to replicate DNA at 85% of the rate observed in the intact cell, and they are active in transcription (37). In these experiments the amount of Z-DNA present in the gene is measured by diffusing biotin-labeled anti-Z-DNA monoclonal antibodies into the beads (38). The amount of Z-DNA was measured initially by the amount of radioactive streptavidin that would bind within the nucleus. These experiments showed that, at low concentrations of antibody, the amount of Z-DNA measured was independent of the antibody added over a 100-fold change in antibody concentration. Furthermore, the amount of Z-DNA depended on DNA negative torsional strain. It increased dramatically as transcription increased but was largely unaffected by DNA replication (39).
It was found that individual genes could be assayed by crosslinking the antibody to DNA using a 10-ns exposure of a laser at 266 nm (40). Release of DNA fragments with attached antibody was accomplished by diffusing in restriction endonucleases and performing an in situ DNA digest. Following isolation of biotin-labeled antibody-DNA complexes with streptavidin magnetobeads, free DNA was obtained by proteolysis. These experiments made it possible to determine which regions of a gene form Z-DNA. Using hybridization or polymerase chain reaction techniques, the c-myc gene was studied in murine U937 cells (40). Three transcription-dependent Z-DNA-forming segments were identified in the 5Ј region of the gene with two of them near promoters (Fig. 2) (41). Retinoic acid, which induces the cells to differentiate into macrophages, was then used to down-regulate expression of c-myc. Loss of c-myc expression was accompanied by a rapid reduction in the amount of Z-DNA present in these three regions. In contrast, Z-DNA was detectable by polymerase chain reaction with probes for the ␤-actin gene under all the conditions tested. ␤-Actin is not down-regulated with differentiation.
In other studies with a primary liver cell line, induction of Z-DNA was measured in the corticotropin hormone-releasing gene (42). Z-DNA formation increased when the gene was up- regulated and decreased when it was down-regulated. This finding suggests that physiological events are being measured in these systems.
A major conclusion from these studies is that Z-DNA forms largely, if not exclusively, behind a moving RNA polymerase and is stabilized by the negative supercoiling generated by DNA transcription.

Functional Consequences of Z-DNA Formation
In principle, Z-DNA formation could have a functional role that need not involve recognition of its shape by proteins. It has been shown that E. coli RNA polymerase does not transcribe through Z-DNA (43). Thus, the formation of Z-DNA behind (5Ј) a moving polymerase may block the following RNA polymerase from transcribing that region of a gene. This might ensure spatial separation between successive polymerases. As a consequence, processing of an RNA would then be physically and temporally removed from that of subsequent transcripts, perhaps minimizing non-functional eukaryotic trans-splicing.
Alternatively, formation of Z-DNA could facilitate recombination of homologous chromosomal domains by relieving topological strain that arises when intact duplexes are intertwined (44). The Z-DNA-forming d(CA/GT) n sequence has been shown to be recombinogenic in yeast (45) but is found to be less efficient than d(CG) n in human cells (46,47). Furthermore, there have been several reports correlating chromosomal breakpoints in human tumors to potential Z-DNA-forming sequences, although no causal relationship has yet been established (48 -52). Last, Z-DNA formation could affect the placement of nucleosomes as well as the organization of chromosomal domains (53).
Identification of proteins that bind to Z-DNA would indirectly establish the presence of Z-DNA in vivo and help establish a biological role for this shape. There has been an extensive search by a number of laboratories for Z-DNA binding proteins. Early studies were unfruitful and caused widespread skepticism that Z-DNA would be associated with any biological function. Many of the positive results reported in these studies may have been due either to artifacts or misinterpretation of data (18,54,55). However, absence of proof was confused with absence of existence.

A High Affinity Z-DNA Binding Protein with Enzymatic Activity
Recent results give cause for optimism. An assay that by its design detects only proteins with high affinity for Z-DNA has revealed that one type of double-stranded RNA adenosine deaminase (dsRAD) 1 called DRADA binds Z-DNA in vitro (56 -58). The dissociation constant of the Z-DNA binding domain is nanomolar, making it likely that this interaction is of biological relevance. The domain maps to a region separate from the three copies of the RNA binding motif present in the protein and also from the catalytic domain (59 -61).
DRADA is an example of a family of deaminases, the dsRAD family, that modify mRNA by catalyzing the hydrolytic deamination of adenine to inosine in regions that are double-stranded (62)(63)(64)(65). RNA shape is important in this reaction as neither single-stranded RNA nor DNA are substrates for this reaction. The efficiency of editing in vitro is influenced by the length of dsRNA, with maximum efficiency seen in a synthetic substrate about 100 base pairs long (66). It is likely that different members of the dsRAD family will obtain specificity from the recognition of different RNA shapes (67). Members of the dsRAD family are ubiquitous in metazoa, suggesting that this activity is of great evolutionary significance (59,68,69). These enzymes may be an important source of phenotypic variation as they have the potential to significantly alter the linear flow of information from DNA to RNA (59). For example, inosine is translated as guanine, so that editing of a codon can result in the substitution of one amino acid for another. An illustrative example of the type of reaction that a dsRAD may catalyze is editing of the GluR-B receptor; whether DRADA or another member of the dsRAD family, such as RED-1, is involved in this reaction is currently a matter of debate (70). Editing of the second transmembrane domain of the GluR-B receptor RNA results in the substitution of an arginine (CGG) for glutamine (CAG), changing the electrophysiological properties of the assembled receptor (71). The double-stranded RNA substrate that is modified by the enzyme is formed by folding the 3Јintron back onto the exon to base pair with the site that is edited (72). In this case, the involvement of introns requires that editing occurs soon after transcription of RNA and before splicing.
The potential involvement of introns in creating the substrates for editing provides a number of rationales for the recognition of Z-DNA by DRADA. As discussed above, Z-DNA in vivo is a transcription-dependent structure and will form when appropriate sequences are present behind (5Ј to) a moving RNA polymerase. This transcription-induced Z-DNA may serve to localize DRADA to a particular region of a gene where editing is to occur, and it may also prevent indiscriminate editing of other regions (Fig. 3). What is important is that the Z-DNA binding domain of the editing enzyme would target only transcribing genes and allow DRADA to act before the splicing apparatus removes the intron. In addition, recognition of Z-DNA by DRADA may block the gene from further transcription until editing of the RNA is complete. Currently there is no direct evidence in vivo that the Z-DNA binding domain influences the catalytic function of the enzyme. However, DRADA is present as a complex inside cells, and interactions may be mediated through other proteins. It should be possible to examine the role of Z-DNA recognition by DRADA in vivo by using UV cross-linking to identify the regions of genes that are bound to DRADA and correlate these with sites of dsRNA editing. It will also be of interest to determine whether other members of the dsRAD family are present within this complex. It is also possible that Z-DNA is not the only transcription-dependent structure recognized by this family of enzymes. FIG. 3. In vivo, Z-DNA is thought to be stabilized by the negative supercoiling generated by an RNA polymerase moving through a gene. Transcription also gives rise to regions of doublestranded RNA (dsRNA), formed when a nascent RNA transcript folds back on itself. An RNA editing enzyme, dsRNA adenosine deaminase (DRADA), has been shown to bind both Z-DNA and dsRNA with nanomolar affinity. Each nucleic acid is bound by DRADA through a separate domain. This enzyme then catalyzes the hydrolytic deamination of adenine within the dsRNA to form inosine. Inosine is subsequently translated as guanine. Several editing sites may exist in a single pre-mRNA. DRADA thus utilizes the structural information encoded by DNA and RNA shapes to change the message read from a gene.
Other proteins may exist that bind to Z-DNA with lower affinity than DRADA. It has been demonstrated that peptides in which every second residue is lysine will stabilize Z-DNA in vitro at micromolar concentrations (73). This provides a simple protein motif with which to recognize Z-DNA. This motif exists in a number of proteins, but it remains to be shown that such proteins interact with Z-DNA. In addition, evidence has been presented to show that topoisomerase II from Drosophila, humans, and calf thymus recognizes a number of different DNA shapes, including Z-DNA (74 -76). However, the domain interacting with these shapes has not yet been biochemically defined.

Future Prospects
While a role for Z-DNA in vivo has not yet been firmly established, the recognition of this shape by DRADA provides a promising lead. Currently there are many unresolved questions. 1) How many of the potential Z-DNA-forming regions in a genome actually form Z-DNA in vivo? 2) Do some of the Z-DNA-forming regions represent sites of action of DRADA and are they recognized by other members of the dsRAD family? 3) What is the nature of the complex that includes DRADA and how do the components interact? Is this complex an editosome containing other RNA editing enzymes? 4) Since DRADA has a modular structure, are there other proteins that have a Z-DNA binding domain but a different enzymatic function? 5) Are there other families of Z-DNA binding proteins? The answers to these questions are of great interest and will lead to new insights on how nucleic acid shape affects the linear flow of information from DNA to RNA to protein.