TTA.TAA triplet repeats in plasmids form a non-H bonded structure.

CTG·CAG, CGG·CCG, and AAG·CTT triplet repeats proximal to or in disease genes expand by a non-Mendelian genetic process to cause several human hereditary syndromes. As part of our physical, biological, and genetic studies on the 10 possible triplet repeats, we discovered that the TTA·TAA repeat, isolated from the upstream region of the variant surface glycoprotein gene of Trypanosoma brucei, shows a propensity to adopt a non-H bonded structure under appropriate conditions. The other nine triplet repeat sequences do not exhibit this property. (TTA·TAA)n, where n = 90, 60, 30, and 18, cloned into pUC19 was studied by chemical and enzymatic probes as well as two-dimensional gel electrophoretic analyses under a variety of conditions. The helix opening was observed for all four inserts in supercoiled plasmids as a function of temperature, pH, metal ions, and buffer conditions using OsO4, diethyl pyrocarbonate, and chloroacetaldehyde probes. This unusual property of the TTA·TAA repeat suggests that it plays a different role from the other nine triplet repeats in gene expression.

Biologically unstable repeating CTG⅐CAG, CGG⅐CCG, and AAG⅐CTT triplet DNA sequences are involved in the etiology of several neurological diseases (reviewed in the accompanying paper (1)). To date, only these three types of repeating triplet sequences have been associated with human hereditary diseases; a total of 10 triplet repeats are possible. Other triplet repeats have been observed in human and mammalian microsatellites (2)(3)(4). Several long tracts of TRS have been identified in the human genome using the repeat expansion detection method (5), which is based on hybridization of short triplet repeat oligonucleotides, and fluoresence in situ hybridization (6). The only triplet repeat containing 100% A/T is TTA⅐TAA.
A survey of triplet repeats in the human genome revealed that TTA⅐TAA is the most abundant (by at least 4-fold) and is the most frequently polymorphic (4). Also, the TTA⅐TAA microsatellites are very abundant in introns but are uncommon in exons (3) and were implicated in regulation of transcription (7,8). At least one protein binds to TTA⅐TAA microsatellites (9). A 270-bp 1 tract of this sequence is found upstream of a variant surface glycoprotein gene of Trypanosoma brucei (10). Since long AAG⅐CTT tracts have very recently been implicated as the cause of Friedreich's ataxia (11), perhaps some of the other seven triplet repeats will also be associated with other hereditary diseases.
Simple repeating DNA sequences such as homo-, di-, and other types of repeats adopt non-B DNA conformations in recombinant plasmids under appropriate conditions of negative supercoil density, ionic environments, protein binding, etc. (reviewed in the accompanying paper (1)). As part of a broader investigation (1,(12)(13)(14)(15)(16)(17)(18)(19) 2 on the biochemical and biological behaviors of all 10 triplet repeat sequences (TRS), we report that the TTA⅐TAA sequence has a propensity to adopt non-paired single-stranded regions under the influence of negative supercoil density and heat.

MATERIALS AND METHODS
Plasmids-pGE117. 6, which contains (TTA⅐TAA) 90 derived from the variant surface glycoprotein gene of T. brucei (gift of Dr. David A. Campbell, University of California at Los Angeles) (10), was digested with SspI, and then the Ϫ1104 to Ϫ761 fragment (344 bp), which contained the TTA⅐TAA repeat and the neighboring (A ϩ T)-rich sequences, was recloned into the HincII site of pUC19 to give pRW3143. pRW3148, pRW3146, and pRW3145 were constructed as described (19). Briefly, pRW3143 was digested with SacI and HindIII, and regions corresponding to fragments with shorter TTA⅐TAA repeats (invisible by staining with ethidium bromide) were eluted from a 1.8% agarose gel. These components were recloned into the SacI-HindIII site of pUC19. All plasmid DNAs were isolated from Escherichia coli HB101 by the alkaline lysis method as described (20). All DNA inserts were sequenced on both strands to determine the TTA⅐TAA repeat sizes of the plasmids.
Chemical Probes-Three chemical probes, OsO 4 (Aldrich), diethyl pyrocarbonate (DEPC) (Aldrich), and chloroacetaldehyde (CAA) (Fluka), were used to modify the TTA⅐TAA repeat containing plasmids as described previously (21)(22)(23). 3 g of DNA was reacted with each reagent in 100-l reaction volume under the conditions described in the figure legends. After the reactions, all samples were purified by gel filtration through Sephadex G-50 microcolumns equilibrated with H 2 O (20) and precipitated with ethanol. Both strands of the modified plasmids were analyzed by primer extension (1,13) or chemical degradation analyses (24).
Primer Extension Analysis-After chemical probing, the DNA was analyzed by the procedure described previously (1,13). The DNA was dissolved in 20 l of a solution containing 0.2 M NaOH and 2 ng of 5Ј-32 P-end-labeled primer. For primer extension analysis, two primers were used, namely M13/pUC forward sequencing primer 1211 (17-mer, New England Biolabs) and M13/pUC reverse sequencing primer 1201 (16-mer, New England Biolabs) for the bottom and top strands, respectively. The plasmid was heated for 90 s at 90°C followed by incubation for 4 min at room temperature. The DNA was neutralized with 0.3 M sodium acetate (pH 5.2) and was precipitated with ethanol. The DNA was resuspended in 10 l of a solution containing 40 mM Tris-Cl (pH 7.5), 50 mM NaCl, 20 mM MgCl 2 , 10 mM dithiothreitol, 0.5 mM 2Ј-deoxynucleoside triphosphates, and 5 units of a modified T7 DNA polymerase (Sequenase Version 2.0, U. S. Biochemical Corp.), and incubated for 10 min at 37°C. After termination of the reaction by the addition of 95% formamide and 20 mM EDTA, the DNA was fractionated on a 12% denaturing polyacrylamide gel, and the bands were visualized by autoradiography.
Chemical Degradation Reactions-The modified DNA was divided into two samples, digested with either SacI and HindIII or EcoRI and PstI, and labeled with [␣-32 P]dATP using the large fragment of E. coli DNA polymerase I (Klenow fragment, U. S. Biochemical Corp.). After isolation from a polyacrylamide gel, the DNA fragments were treated with 1 M piperidine for 30 min at 90°C, lyophilized, and applied to a 12% denaturing polyacrylamide gel in parallel with sequencing markers generated by the chemical degradation method (24). The bands were visualized by autoradiography.
Generation of Topoisomers-Topoisomers of plasmids were prepared as described previously (25). Briefly, DNAs with various supercoil densities were generated by incubating 6 g of DNA in 100 l of a solution containing 10 mM Tris-Cl (pH 7.6), 50 mM KCl, 10 mM 2-mercaptoethanol, 1 mM EDTA, 0 -11 M ethidium bromide, and chicken erythrocyte topoisomerase (26) for 60 min at 37°C. The ethidium bromide and topoisomerase were removed by two phenol extractions followed by two ether extractions, and the pooled DNAs were resuspended in H 2 O after ethanol precipitation.
Two-dimensional Gel Electrophoresis-Two-dimensional gel electrophoresis was performed as described previously (27). Mixtures of topoisomer populations prepared as described above were subjected to firstdimension gel electrophoresis in 1.25% agarose at 3.3 V/cm at 25°C in 0.5 ϫ TBE buffer (pH 8.3) (0.5 ϫ TBE, 45 mM Tris borate, 1 mM EDTA) or TAE buffer (pH 8.0) (TAE, 40 mM Tris acetate, 1 mM EDTA). The gel then was soaked in 0.5 ϫ TBE buffer (pH 8.3) containing 0.7-20 M chloroquine diphosphate (Sigma) for 3 h. Electrophoresis in the second dimension was carried out at a 90°angle to the first dimension at 3.3 V/cm at 25°C in 0.5 ϫ TBE buffer (pH 8.3) containing 0.7-20 M chloroquine diphosphate. Fig. 1 shows the cloned sequences in the recombinant plasmids and a list of the plasmids used in this study. Three plasmids with shorter TTA⅐TAA repeats were produced as described (19) from the original pRW3143, which contains (TTA⅐TAA) 90 . As expected, only a part of the perfect (TTA⅐TAA) 90 repeat was deleted; inserts with 60, 30, and 18 repeats of TTA⅐TAA were produced. The plasmids were designated pRW3148, pRW3145, and pRW3146, respectively.

Plasmids Containing TTA⅐TAA Repeat Sequences-
Chemical Probing-To investigate the formation of unusual DNA (non-B DNA) structures, such as Z-DNA, triplexes, and cruciforms, chemical probes have been used (28,29) because they can specifically modify bases in single-stranded or conformationally perturbed DNA or B-Z junctions. The modified base(s) can be detected by cleavage of the modified sites with a specific reagent (i.e. hot piperidine) or the primer extension assay which detects the inhibition of elongation by DNA po-lymerase. In this study, we used OsO 4 , DEPC, and CAA to evaluate the existence of non-B DNA structures in the TTA⅐TAA repeat sequences.
OsO 4 reacts at the C-5ϭC-6 double bond of pyrimidines (T Ͼ Ͼ C) in the presence of tertiary amines, such as pyridine and 2,2Ј-bipyridine, and is substantially more reactive to singlestranded DNA than to double-stranded DNA (21, 30 -32). It has been used for the single-stranded regions of cruciforms (32)(33)(34) and triplexes (21,35) and for the structural perturbations at B-Z junctions (36). For inverted repeat sequences neighbored by (A ϩ T)-rich sequences, OsO 4 was used as a chemical probe to detect unpaired single-stranded DNA (32,33,(37)(38)(39). Also, AT⅐AT repeat sequences were found to form cruciform structures by detection of single-stranded regions with OsO 4 (40 -42). DEPC carboxyethylates purines (A Ͼ G) at the N-7 position by the opening of the imidazole ring in single-stranded DNA as in triplexes (21,35) and cruciforms (34,43,44) or in a syn conformation as in Z-DNA (36,45). On the other hand, CAA as well as bromoacetaldehyde reacts at the N-1 and the C-6 amino groups of adenines or the N-3 and the C-4 amino groups of cytosines to generate 1,N 6 -ethenoadenine or 3,N 4 -ethenocytosine derivatives at unpaired adenines or cytosines (46,47), and has been used for single-stranded regions of cruciforms in inverted repeats containing (A ϩ T)-rich sequences (48) and AT⅐AT repeat sequences (49), for triplexes (22,23), and for B-Z junctions (50,51). Fig. 2 shows chemical reactivities with these three chemical probes (OsO 4 , DEPC, and CAA) on the bottom strand of pRW3146, which contains 18 repeats of TTA⅐TAA, as analyzed by the primer extension assay. By this analysis, the modified bases, thymines by OsO 4 and adenines by DEPC or CAA, on the template strand were detected by the inhibition of elongation by the DNA polymerase (Sequenase) at the corresponding bases on the extended strand. Surprisingly, modifications were observed throughout the (TAA) 18 sequence by all three chemical probes as a result of inhibition of elongation by Sequenase on the top strand (TTA strand) (Fig. 2, lanes 1, 3, and 5). The (A ϩ T)-rich sequences, which are located on the 3Ј side of (TTA) 18 and contain (TTA) 5 and (TTA) 2 , also were modified by DEPC and CAA (Fig. 2, lanes 3 and 5) but not by OsO 4 (Fig. 2, lane 1). We also determined that the modification pattern of bromoacetaldehyde was identical to that of CAA (data not shown). These observations may result from the fact that DEPC was reacted at 20°C for 30 min, a longer reaction time than with OsO 4 , and CAA was used at 37°C for 30 min, a higher temperature and a longer reaction time than with OsO 4 . On the other hand, on the bottom strand (TAA strand), modifications by OsO 4 , DEPC, and CAA gave the same pattern as those of the The SspI fragment (344 bp) of pGE117.6 containing (TTA⅐TAA) 90 and the neighboring sequences was cloned into the Hin-cII site of pUC19. pRW3148, pRW3145, and pRW3146 were constructed as described under "Materials and Methods," as a result of deletions from pRW3143. top strand (TTA strand) (data not shown), indicating the existence of unpaired regions in both strands of supercoiled pRW3146.
We investigated the effect of the length of the TTA⅐TAA repeats. Fig. 3 shows the OsO 4 modification patterns for different lengths of TTA⅐TAA repeat containing plasmids under the same conditions described above. From 18 up to 90 repeats of TTA⅐TAA, the modifications were evenly distributed throughout the TTA⅐TAA repeats on both strands (Fig. 3, A and  B, lanes 1, 3, 5, and 7), indicating that unpaired regions exist in all of these lengths of TTA⅐TAA repeats.
Recently, we observed that CTG⅐CAG repeat sequences derived from the myotonic dystrophy gene and CGG⅐CCG from the fragile X gene, even without chemical modification, caused pausings of DNA polymerases, namely Sequenase, the Klenow fragment of E. coli DNA polymerase I, and human DNA polymerase ␤ (13). This behavior was attributed to the formation of a new type of non-B conformation by this triplet repeat sequence. Parallel studies were performed on the TTA⅐TAA triplet repeats described herein. No pausing of Sequenase was found on both strands of all TTA⅐TAA inserted plasmids (Fig. 3, A and B,  lanes 2, 4, 6, and 8), even with preincubation (at 37 or 50°C for 10 min) before primer extension, a condition which strongly enhanced the pausings for CTG⅐CAG repeat sequences (13). No chemical probes were employed in this study. Also, there was no pausing with the Klenow fragment in the absence of chemical modifications, even under the preincubation conditions (data not shown). Hence, we conclude that TTA⅐TAA repeats do not adopt unusual DNA conformations that inhibit the progression of DNA polymerase, in contrast to the behavior with CTG⅐CAG and CGG⅐CCG repeats.
Influence of Negative Supercoil Density-The chemical probing described above revealed that TTA⅐TAA repeat sequences formed unpaired regions at the supercoil density of Ϫ0.06. Some unusual DNA structures (left-handed Z-DNA, cruciforms, triplexes) are known to be induced by negative superhelical density in recombinant plasmids (21,35,(52)(53)(54)(55). Accordingly, the effect of superhelical density on OsO 4 modification on these TTA⅐TAA sequences was investigated. Fig. 4 shows the quantitation of the amount of OsO 4 modification for pRW3146 as a function of average negative supercoil densities (Ϫ). At low supercoil density (Յ0.025), few OsO 4 modification sites were observed. In linearized pRW3146, no sites were observed by chemical degradation analyses (data not shown). As the torsional stress was increased, however, the degree of OsO 4 modification increased. The reactivity relative to supercoiling shown in Fig. 4 is broad compared with the all-or-none transitions found for triplexes or cruciforms. 80% of the transition took place over 0.07 density units, whereas only ϳ0.01 was required for the other conformational changes (35,52,56). Thus, the gradual transition ( Fig. 4) indicates that a larger duplex region is unpaired as more supercoil density is added to the system. Two-dimensional Gel Electrophoresis-Supercoil-dependent alternate structures of DNA, such as cruciforms, triplexes, and Z-DNA, have been detected by two-dimensional gel electrophoresis (38,39,(53)(54)(55). The chemical probe analyses described above suggested that unpaired DNA regions exist throughout the TTA⅐TAA repeat and in a part of the (A ϩ T)-rich neighboring sequence, depending on the conditions employed. We used two-dimensional agarose gel electrophoresis to further investigate the unwound DNA regions revealed by the chemical probing results. The topoisomer populations of pUC19, pRW3146, and pRW3145 were analyzed as shown in Fig. 5. pRW3146 and pRW3145 exhibited an initial relaxation of 3.5 and 5 supercoil turns, respectively, at topoisomer numbers Ϫ10 and Ϫ11. This corresponded to relaxation of 37 and 53 bp at supercoil densities of Ϫ0.037 and Ϫ0.041, similar to the superhelical density at which the initiation of OsO 4 modification occurred (Fig. 4 for pRW3146 and data not shown for pRW3145). The relaxation continued until topoisomer numbers Ϫ17 and Ϫ18 (superhelical density Ϫ0.061 and Ϫ0.063, respectively). The total relaxation for pRW3146 and pRW3145 was 10.  1, 3, and 5, respectively). Lane 1, supercoiled pRW3146 was incubated in TE buffer (pH 7.4) at 20°C for 20 min and then modified by 1.1% OsO 4 in the presence of 2% pyridine at 20°C for 10 min. Lane 3, the DNA was incubated under the same condition described above and then modified by 10% DEPC at 20°C for 30 min. Lane 5, the DNA was incubated under the same condition described above and modified by 2% CAA at 37°C for 30 min. After chemical probing, the modified sites were mapped using the primer extension method as described under "Materials and Methods." The control plasmids (lanes 2, 4, and 6) were treated similarly except no modifying chemicals were added. The G, A, T, and C sequencing lanes are indicated for the top strand of unmodified pRW3146 using the same primer as for the modified samples. The corresponding 2Ј,3Ј-dideoxynucleoside triphosphates (U. S. Biochemical Corp.) were added to the reaction mixtures. The sequence of the top strand is indicated at the left side of the gel.
below. No transition was observed for the control pUC19. The electrophoresis was repeated on the same samples in TAE buffer (pH 8.0). Interestingly, no transitions were observed. This result will be discussed in the next section.
Similar studies were performed on pRW3143 which contains 90 repeats of TTA⅐TAA with 20 M chloroquine in the second dimension. Whereas the resolution of the gel did not permit an accurate determination of the total number of topoisomers relaxed, at least 26 were detected indicating an unpairing of at least 270 bp.
Effect of Environmental Conditions-Temperature and salt concentration have been shown to affect the unpairing of (A ϩ T)-rich sequences as well as AT⅐AT repeat sequences in a supercoiled molecule (37-39, 42, 54, 57, 58). (A ϩ T)-rich sequences were observed to induce adjacent inverted repeats to form cruciform structures via the C-type extrusion formed by large-scale helix opening, and this was dependent on supercoil density, temperature, and salt concentration (37). Bowater and co-workers (39) observed that large-scale opening of (A ϩ T)rich regions within a supercoiled molecule was suppressed by salt. It is also known that AT⅐AT repeat sequences form cruciform structures at low temperature or in the presence of metal ions (Na ϩ , Mg 2ϩ , K ϩ , Ca 2ϩ , [(NH 3 )Co] 3ϩ , etc.) (40,42). Thus, we investigated the effect of temperature and metal ions for the TTA⅐TAA repeat sequence. On the presumption that T⅐T and A⅐A bp might be formed in addition to A⅐T pairs in the TTA⅐TAA sequence, a TTA⅐TAA repeat could form a cruciform structure. Fig. 6 shows the effect of temperature on OsO 4 modification of supercoiled pRW3146. As the temperature was increased, OsO 4 modification intensified in the (TAA) 18 region. pRW3146 contains an (A ϩ T)-rich region, 87.5% A/T content, adjacent to a perfect (TAA) 18 . At 37°C, the (A ϩ T)-rich region was strongly modified. This modification pattern was identical to that found with CAA (Fig. 2, lane 5). However, at lower temperatures, 0 or 10°C, there was little modification and the patterns did not reveal the presence of cruciforms.
The reversibility of the temperature effect was investigated. After preincubation at one temperature (i.e. 37, 20, 10, or 0°C) for 20 min, the chemical probe reaction was conducted at a second temperature (i.e. 0, 10, 20, or 37°C) for 10 min. The resulting OsO 4 modification pattern was the same as that found when the second temperature was used for both the preincubation and the reaction (data not shown). Thus, this result indicates that the helix opening in the (TAA⅐TTA) 18 repeat and the neighboring (A ϩ T)-rich sequences is formed reversibly as a function of temperature.
The effect of metal ions on the OsO 4 modification patterns also was investigated for supercoiled pRW3146. In the case of NaCl up to 75 mM, the modification was observed throughout the (TAA⅐TTA) 18 tract at 20°C, but at 150 mM NaCl, pH 7.4 and 20°C, no modification was observed (data not shown). Under physiological conditions, at 150 mM NaCl, pH 7.4 and 37°C, although the modification was observed throughout the (TAA⅐TTA) 18 region, the intensity of OsO 4 modification was much weaker than with no NaCl (data not shown). Incidentally, for the range 0.1-150 mM NaCl, no cruciform-like structures were observed. On the other hand, in the presence of 5 mM MgCl 2 or higher, no modification was observed at 20°C, pH 7.4, while there was modification throughout the (TAA⅐TTA) 18 region up to 1 mM MgCl 2 (data not shown). Thus, metal ions (Na ϩ , Mg 2ϩ ), especially at higher concentration, inhibited helix opening.
As described above, when two-dimensional gel electrophoresis was performed in TAE buffer (pH 8.0) instead of 0.5 ϫ TBE buffer (pH 8.3) in the first dimension, no transitions were observed, although a positive control, pRW1561, containing a potential Z-DNA sequence (55) exhibited the expected transitions in both buffers (data not shown). To further investigate this result, we evaluated the effect of buffer conditions (pH and salt concentration) on the OsO 4 modification of supercoiled pRW3146. There were distinct differences in the modification patterns as influenced by salt and pH. In 0.5 ϫ TBE buffer (pH 8.0), the entire 94-bp (A ϩ T)-rich regions (the (TTA⅐TAA) 18 repeat and the 40 bp of 87.5% A/T sequence on the 3Ј side) were modified (Fig. 7, A and B, lane 1), in agreement with the two-dimensional gel results. On the other hand, in TAE buffer (pH 8.0), the OsO 4 modification was observed throughout the (TTA⅐TAA) 18 repeat and the neighboring (TTA⅐TAA) 5 repeat sequences, although the intensity of the modification in the (TTA⅐TAA) 2 repeat region was much less than that in 0.5 ϫ TBE buffer (pH 8.0). While this result seems inconsistent with that of the two-dimensional gel electrophoresis described above, it is known that OsO 4 reactions are more sensitive than the two-dimensional gel electrophoresis analysis because the FIG. 5. Two-dimensional agarose gel electrophoresis of topoisomers of pUC19, pRW3146, and pRW3145. A tracing of each of the gels is shown to the right of the photographs. The DNA preparation and the gel conditions were described under "Materials and Methods." Chloroquine concentration in the second dimension was 0.7, 0.75, and 0.8 M for pUC19, pRW3146, and pRW3145, respectively. 20 M chloroquine also was used to increase resolution of pRW3145 (data not shown). The first dimension direction is from top to bottom, and the second dimension direction is from left to right. The bright spot labeled N in the upper left-hand corner of each tracing corresponds to nicked plasmid DNA, and the spot in the middle labeled L corresponds to linear DNA. gels measure an equilibrium state, whereas the OsO 4 reactions can detect a transitory relaxation. Since the ionic strength of the two buffers is similar, it appears that the borate ions favor the helix opening, whereas the acetate ions may preserve the hydrogen bonded B-DNA conformation.
The modification also was influenced by pH. The (TTA⅐TAA) 5 repeat sequence was modified more strongly at pH 8.0 than at pH 7.4 in TE buffer (TE, 50 mM Tris-Cl, 1 mM EDTA) (Fig. 7, A  and B, lanes 3 and 5), whereas at acidic pH (pH 4.5) in TAE buffer (Fig. 7, A and B, lane 9) the modification pattern was similar to that at pH 7.4. Hence, these results indicate that the helix opening in the (TTA⅐TAA) 18 repeat and the neighboring (A ϩ T)-rich sequences of pRW3146 are influenced not only by pH, salt, and metal ion concentration but also by the type of anion. DISCUSSION We demonstrate that a TTA⅐TAA triplet repeat sequence shows the unusual propensity of adopting a non-H bonded structure in plasmids under appropriate conditions of supercoiling, temperature, pH, salt, and metal ion concentrations. Compared with the other nine TRS, only this TRS exhibits this behavior. The SspI fragment of pGE117.6, which contains 90 repeats of TTA⅐TAA and the neighboring (A ϩ T)-rich se-quences, was inserted into pUC19 to give pRW3143. Kang et al. (19) constructed several lengths of CTG⅐CAG repeats in plasmids by deletion of the original 130-repeat sequence derived from the myotonic dystrophy gene (19), and similar procedures worked effectively for other triplet repeats (1,14,15). Using this method, we constructed three other plasmids with shorter TTA⅐TAA repeats (60, 30, and 18 triplets) from pRW3143 as a result of deletions of a portion of the (TTA⅐TAA) 90 tract. Attempts to obtain expansions (1,14,19) of (TTA⅐TAA) 90 were unsuccessful.
The opened helix structure was observed in all four inserts by chemical probe (OsO 4 , DEPC, and CAA) and two-dimensional gel electrophoretic analyses. All thymines and adenines in both strands of the TTA⅐TAA repeats reacted with the reagents and were detected by the primer extension assay as a result of the inhibition of elongation by the DNA polymerase (Sequenase). This result shows that non-H bonded singlestranded regions exist throughout the repeats. These singlestranded regions may be disoriented random coils or may be highly ordered, stacked structures; our data do not enable more definitive conclusions from a structural standpoint. Fig. 8 shows a model of the unpairing of the TTA⅐TAA repeats and the neighboring (A ϩ T)-rich sequences in pRW3146. This model also pertains to other longer TTA⅐TAA repeats since the chemical reactivities by OsO 4 were similar to that of pRW3146. This indicates that the helix opening of the TTA⅐TAA repeats and the neighboring (A ϩ T)-rich sequences is not influenced by the repeat lengths. The opening was dependent on negative supercoil density and was not observed for linear DNA. The unpairing occurred gradually over a broad range of supercoil density, especially as compared with that found for triplexes or cruciforms. Hence, the transition is not a cooperative, all-or-none process.
The unpairing of the neighboring (A ϩ T)-rich sequence was sensitive to environmental conditions also. As the temperature was increased, the (A ϩ T)-rich sequence was modified strongly by OsO 4 . At 37°C, the TTA⅐TAA repeats and the neighboring (A ϩ T)-rich sequences were entirely opened, but both regions were not opened at 0°C (Fig. 8). The other neighboring region (5Ј side of the TTA⅐TAA repeats) was not unpaired even at 37°C. Hence, high A/T content sequences were easily opened. The unpaired structure formed reversibly as a function of temperature. This temperature-sensitive type of helix opening has been found in other (A ϩ T)-rich sequences (37-39, 42, 54, 57). Different modification patterns were found between OsO 4 (reaction time of 10 min) and DEPC (reaction time of 30 min) at 20°C, indicating that longer incubation times caused further unpairing.
In addition, the concentration of metal ions affected the helix opening. Na ϩ (150 mM) and Mg 2ϩ (5 mM) inhibited the unpairing at 20°C. Under physiological conditions (37°C, 150 mM NaCl, pH 7.4), the intensity of the OsO 4 modification was low, suggesting that in vivo the opening may not be occurring. Bowater et al. (39) suggested that some proteins will be required to facilitate the helix unpairing in order to regulate gene expression in vivo. Umek and Kowalski (59) also suggested that an (A ϩ T)-rich sequence destabilizes helical structure and promotes unwinding for the initiation of DNA replication. The original (TTA⅐TAA) 90 and the neighboring (A ϩ T)-rich sequences are located in the region upstream of the transcription start site, suggesting that the region may play a role in transcriptional initiation. However, the biological function(s) of the TTA⅐TAA tracts is unknown.
Furthermore, the opening was influenced by pH and buffer conditions. OsO 4 modifications differed in the (TTA⅐TAA) 5  modifications between pH 4.5 and 8.0 in TAE buffer. Twodimensional agarose gel electrophoresis showed relaxations in 0.5 ϫ TBE buffer (pH 8.3) but not in TAE (pH 8.0), although the ionic strength of the two buffers is similar. On the other hand, OsO 4 reactivities were observed in both buffers although there was a difference between the two buffers in the (TTA⅐TAA) 2 tract. This suggested that the OsO 4 reaction is more sensitive than the two-dimensional gel electrophoresis; the gels measure an equilibrium state between relaxed and unrelaxed molecules, whereas the OsO 4 reactions can detect a transient singlestranded region. This electrophoresis phenomenon is peculiar to the (A ϩ T)-rich sequence in this study because plasmids containing a Z-DNA forming sequence showed relaxation in both buffers. These results also indicate that the borate ions favor the helix unpairing more than the acetate ions. Sodium acetate has been shown in previous work (60) on studies with left-handed Z-DNA to have an unconventional effect on DNA conformation.
Kang et al. (13) observed strong pausings of DNA polymerases (Sequenase, the Klenow fragment, and human DNA polymerase ␤) at specific locations within CTG⅐CAG and CGG⅐CCG repeats depending on the number of repeats and pretreatment of the duplex template, indicating that appropriate lengths of the repeats adopted a non-B DNA conformation(s) which caused the pausings. We tested the capacity of the TTA⅐TAA repeats to cause pausing in the primer extension assay; 90 repeats of TTA⅐TAA, as well as the shorter inserts ( Fig. 1), did not show any pausings of the DNA polymerases as had been found for CTG⅐CAG and CGG⅐CCG repeats. By comparison, pausings had been observed for 80 repeats of CTG⅐CAG (13). This and other studies (1, 13, 15) 2 show that the TTA⅐TAA and the other nine repeats have quite different biochemical and genetic properties.
Although the mechanism of the expansion of the CTG⅐CAG, CGG⅐CCG, and AAG⅐CTT repeats is not yet known (1,12,14,15,19,61), slippage-mediated DNA structures may cause expansion during DNA replication (12,(62)(63)(64). Schlötterer and Tautz (65) showed that the synthesis rate for the TTA⅐TAA repeat in vitro is the fastest of the 10 triplet repeats, indicating that slippage correlated with the A/T content of the sequences. These results, along with the case of base pair disruption, may explain why polymerase pausing was not found with the TTA⅐TAA repeat. Since the TTA⅐TAA repeat sequences have several physical, biochemical, and genetic properties that differ significantly from those observed for the other nine repeats (1, 12-16), 2 we conclude that this sequence may play different roles in gene regulation.