ABSTRACT
d(A-G)10 forms two helical structures at neutrality, at low ionic strength a single-hairpin duplex, and at higher ionic strength a double-hairpin tetraplex. An ionic strength-dependent equilibrium between these forms is indicated by native PAGE, which also reveals additional single-stranded species below 0.3 M Na+, probably corresponding to partially denatured states. The equilibrium also depends upon oligomer concentration: at very low concentrations, d(A-G)10 migrates faster than the random coil d(C-T)10, probably because it is a more compact single hairpin; at high concentrations, it co-migrates with the linear duplex d(A-G)10[middot]d(C-T)10, probably because it is a two-hairpin tetraplex. Molecular weights measured by equilibrium sedimentation in 0.1 M Na+, pH 7, reveal a mixture of monomer and dimer species at 1oC, but only a monomer at 40oC; in 0.6 M Na+, pH 7, only a dimer species is observed at 4oC. That the single- and double-stranded species are hairpin helices, is indicated by preferential S1 nuclease cleavage at the center of the oligomer(s), i.e., the loop of the hairpin(s). The UV melting transition below 0.3 M Na+ or K+, exhibits a dTm/dlog[Na+/K+] of 33 or 36oC, respectively, consistent with conversion of a two-hairpin tetraplex to a single-hairpin duplex with extrahelical residues. When [Na+/K+] >= 0.3 M, dTm/dlog [Na+/K+] is 19 or 17oC, respectively, consistent with conversion of a two-hairpin tetraplex directly to single strands. A two-hairpin structure stabilized by G-tetrads is indicated by differential scanning calorimetry in 0.15 M Na+/5 mM Mg2+, with [Delta]H of formation per mole of the two-hairpin tetraplex of -116.9 kcal or -29.2 kcal/mol of G-tetrad.
The frequency of occurrence of alternating repetitive homopurine[middot]homopyrimidine stretches in distinct functional regions, i.e., promoters, within eukaryotic genomes is much higher than predicted based upon genome complexity (1 ). Since these simple repetitive sequences have been shown to be polymorphic, it is intriguing to consider possible roles of alternative structures in gene regulation.
Several observations have shown that the repetitive sequence d(A-G)n has a potential to form self-structures. Based upon thermal melting and CD spectra, it was assumed that long strands of this sequence form a pH-independent four-stranded structure involving G- and A-tetrads (2 -4 ). However, based upon CD spectral changes, it was suggested that the same d(A-G) self-structure is promoted by either acidic pH or high monovalent salt concentrations at neutrality (5 ). Complexes at pH 4-9 in 10 mM Mg2+ of the related sequences d(G-A)7G and d(G-A)12G were interpreted as linear parallel duplexes with G[middot]G and A[middot]A base pairs (6 ). A related homodimer structure was also reported (7 ). But, flanked by self-complementary sequences, d[-(A-G)20-)] forms a hairpin stabilized by the terminal Watson-Crick base pairs (8 ). Some type of folded structure was also suggested for the repetitive sequences d(T4GAGA)4 and d(T4AGAG)4 at pH 8 (9 ); however, their electrophoretic behavior is different than that of the known fold-back tetraplex adopted by d(T4G4)4 (9 ). In attempting to utilize d(A-G)10 as a homopurine third strand directed to the target duplex d(A-G)10[middot]d(C-T)10, we were led to ascribe uncommonly weak third strand binding to its involvement in competing equilibria. Subsequently, anomolous behavior for d(A-G)10 was revealed, resulting in the identification of two pH-dependent classes of structures. Low pH and ionic strength stabilize a concentration-independent, single-stranded, [alpha]-helix-like structure for d(A-G)n <= 10 (10 -12 ). Based upon CD spectroscopy, this helix was differentiated from the other that forms at higher ionic strength at neutrality. Under such neutral pH conditions, this sequence adopts alternate conformations in a concentration- and ionic strength-dependent manner. In this report, we characterize the single- and double-hairpin conformations of d(A-G)10 at neutrality, and in the accompanying report (13 ), provide direct spectroscopic evidence for a G-tetrad H-bonding scheme in the double-hairpin conformation.
d(A-G)6, d(A-G)10 and d(C-T)10 were synthesized and purified to homogeneity (10 ); excess salts were removed and the oligomers purified as the Na+ salt using SepPak C18 (Waters) columns. Oligomer concentrations are reported on a residue basis.
The six sample conditions (all containing 0.01 M Tris-HCl, pH 7.4) included: (i) 0.01 M Na+, (ii) 0.01 M Na+/5 mM Mg2+, (iii) 0.15 M Na+, (iv) 0.15 M Na+/5 mM Mg2+, (v) 0.6 M Na+ or 0.6 M Na+/5 mM Mg2+. Except where otherwise indicated, all cations were included as the chloride salt. Concentrations of d(A-G)10 were varied as indicated; d(C-T)10 and d(A-G)10[middot]d(C-T)10 were 6 * 10-1 mM. Each oligomer was labeled, purified (12 ) and an equal concentration of the appropriate 5'-32P-end-labeled oligomer (6 * 10-4 mM) was included in all mixtures. Samples were heated to 75oC, cooled slowly to 25oC, and equilibrated at 4oC overnight. At 4oC, aliquots (4.5 [mu]l) of each sample were added to 20% Ficoll-400 (1.5 [mu]l), mixed briefly, and loaded onto a pre-electrophoresed 16% native polyacrylamide gel. Gels and electrophoresis buffers contained 0.089 M Tris-borate, pH 7.4 in addition to each of the aforementioned ionic conditions. Gels (0.4 mm * 31.0 cm * 38.5 cm) were run at 5-7 V/cm at 4oC with recirculation until bromophenol blue in a marker lane had migrated 16 cm; the samples contained no dye. Autoradiography of the wet gels was carried out for 18 h at 4oC.
Solutions of d(A-G)10 in 0.01 M cacodylate, pH 7 with 0.6 M Na+ or 0.15 M Na+/5 mM Mg2+ at loading concentrations of 0.0042, 0.0083, 0.016, 0.05 and 0.13 mg/ml were centrifuged at 30 000 and 40 000 r.p.m. at 4oC. The ultracentrifuge and solution loading were as described (11 ). The quantities of solution and solvent used produced solution columns ~5 mm in height. Appropriate combinations of solution concentration, cell pathlengths of 4 and 12 mm, and wavelengths of light of 290, 280 and 260 nm were used to measure the concentration gradients by absorption. The oligomer in 0.01 M cacodylate, pH 7, with 0.1 M Na+ was studied similarly at both 1 (30 000 and 40 000 r.p.m.) and 40oC (40 000 r.p.m.), with loading concentrations of 0.005, 0.015, 0.05 and 0.13 mg/ml.
Data analysis was performed (11 ) using a non-linear least squares program (nonlin, 14 ) which fits all the data (speeds, wavelengths and optical paths) globally to various assumed association models. A good fit produces the reduced molecular weight M"= [M(1 - [Phi]'[rho])], where [Phi]'is the specific volume at constant free energy (15 ) and [rho] is the solvent density, the values of the other constants related to the model (i.e., second virial coefficient, B; equilibrium constants for association, K; etc.), and the value of the r.m.s. error.
Solvent densities were estimated from density tables to be 1.004 and 0.999 g/ml at 1 and 40oC respectively for the 0.1 M Na+ solvent, and 1.008 and 1.026 g/ml respectively for the 0.15 M Na+/5 mM Mg2+ and 0.6 M Na+ solvents at 4oC.
Reaction mixtures contained 0.01 M Tris-HCl, pH 7.4, 0.15 M Na+, 1 mM Zn2+ and 5 U S1 nuclease (US Biochemicals). Those with 0.64 mM d(A-G)10 included 7-8 * 10-3 mM 5'-32P-end-labeled oligomer, with and without 5 mM Mg2+; 6 * 10-4 mM samples contained only labeled oligomer. Mixtures were heated to 90oC and slowly cooled to 25oC; after 5 min at 5, 15 or 40oC (as indicated), digestion was initiated by adding S1 nuclease and Zn2+. While pH 4.0-4.3 is optimal for enzyme activity, digestions were performed at pH 7.4, where the activity is dramatically diminished (16 ), allowing detection of the earliest and most sensitive cleavage sites. Digestions were halted by removing 5 [mu]l aliquots at 0.5, 1, 2, 5, 10, 15, 30 and 60 min, adding them to 4 [mu]l loading dye (70% formamide, 0.1% xylene cyanole, 0.1% bromophenol blue, 57 mM EDTA, pH 7.5), and freezing. Frozen samples were heated at 90oC for 5 min prior to loading on a 22% denaturing gel, and electrophoresed at 50 V/cm until the bromophenol blue migrated 20 cm. Products were visualized by autoradiography.
Samples contained 0.16 or 0.64 mM d(A-G)10 in 0.01 M cacodylate, pH 7.0, a total Na+ or K+ concentration of 0.05, 0.10, 0.15, 0.30, 0.60 or 1.2 M, and where indicated 5 mM Mg2+. Solutions were equilibrated at 25oC for 1 h followed by 10 min at 3oC prior to thermal melting in 0.1 or 1.0 cm pathlength cuvettes. After melting, samples in the Na+ series were brought to 25oC and equilibrated for 4 weeks, followed by 10 min at 3oC and remelted. Profile reproducibility was excellent.
Absorbances were recorded using a computer-driven AVIV 14DS spectrophotometer (AVIV Associates, Lakewood, NJ) equipped with a thermoelectrically controlled cuvette holder. Spectra recorded between 320 and 230 nm at 1 nm intervals every 2oC from 3 to 89oC were used to generate melting profiles and their derivatives at desired wavelengths. Oligonucleotide spectra were corrected for solvent and cuvette contributions at each temperature. Melting profiles were corrected for thermal expansion of water, smoothed by a least-squares polynomial fit (third order) and normalized to an absorbance of 1.0 at 3oC.
Measurements were made as described (11 ). A baseline was obtained between 10 and 60oC with the solvent (0.01 M cacodylate, pH 7, 0.15 M Na+, 5 mM Mg2+) in both reference and sample cells, and used to correct the DSC plot for the sample. The sample cell with 1.59 mM d(A-G)10 in solvent and the reference cell with solvent alone were scanned from 10 to 60oC. After slow cooling to 10oC, and equilibration overnight, they were re-scanned.
In contrast to the single-stranded helix without base stacking formed by this sequence in 0.01 M Na+ at acidic pH (10 ), preliminary experiments on this oligomer in 0.15 M Na+, 5 mM Mg2+ at pH 7 revealed cooperative UV melting attended by a large hyperchromic change and oligomer concentration-dependence of Tm. Also, at neutrality a unique intense CD spectrum characteristic of a helical structure was observed (Figure 10 in ref. 14 ).
Native PAGE was utilized to determine how oligomer concentration influences formation of the d(A-G)10 helical structure(s) at pH 7, using the duplex d(A-G)10[middot]d(C-T)10 and single-stranded random coil d(C-T)10 as mobility standards of known size and conformation. [While d(C-T)n has been reported to exist as a duplex at very high concentrations (17 ), we have found that in the concentration and ionic strength range studied here, it does not form a multi-stranded stacked structure. Thus, at pH 7.4, d(C-T)10 migrates equally in 0.15 M Na+ with and without Mg2+ at both 6 * 10-3 and 1 mM oligomer on native PAGE, i.e., in a concentration-independent manner. Also, d(C-T)10 exhibits no UV hyperchromism with increasing temperature.] Figure 1 shows the electrophoretic mobility of d(A-G)10 over a 1000-fold oligomer concentration range, from 6 * 10-4 to 6 * 10-1 mM (residues) and under a range of ionic conditions, in comparison with the two standards. The relative mobilities reveal two salient features: (i) there is oligomer concentration and ionic strength dependence of structure formation, and (ii) there is an equilibrium between multiple monomer (i.e., single-stranded) species and one dimer species of d(A-G)10. These features are manifest in several ways. In general, d(A-G)10 migrates as very broad band(s), indicative of continual re-equilibration among structures as oligomer is diluted during the course of electrophoresis. There are a minimum of three types of species with different mobilities, the equilibrium between them depending on oligomer concentration and ionic conditions: a duplex species migrating like d(A-G)10[middot]d(C-T)10, a random coil single-stranded species migrating like d(C-T)10, and still faster migrating species.
Equilibrium sedimentation analysis was used to obtain molecular weights directly and thereby the number of oligomer units associated with the three types of structures revealed by native PAGE. For this purpose, sedimentation experiments were performed under three ionic conditions, in one case at two temperatures.
(i) In 0.6 M Na+ at 4oC, where the gels reveal a high propensity for d(A-G)10 to form a stable duplex structure (Fig. 1 A), 10 sets of sedimentation data covering a range from ~10 to ~0.6 mg/ml oligomer were fitted. An ideal single species (ISS) model fit the data, producing a molecular weight of 6.583 * 103 g/mol, corresponding to a dimer, assuming [Phi]"is ~0.56 ml/g.
(ii) In 0.15 M Na+/5 mM Mg2+ at 4oC, where the gels reveal the same predominant molecular species (Fig. 1 C), 10 sets of data were fitted covering a range from <10 [mu]g/ml to >0.6 mg/ml. The best fit was obtained by the ISS model, producing a molecular weightcorresponding to a dimer, assuming a [Phi]"of ~0.59 ml/g.
(iii) At 1oC in 0.1 M Na+, 10 sets of data were fitted spanning a range of a few [mu]g/ml to ~0.18 mg/ml, close to that in the gel studies. The fit of the data to the ISS model was poor, giving a Z average molecular weight between that of the monomer and dimer. The best fit was to a monomer-dimer model wherein different equilibrium constants are allowed for each loading concentration and speed. This fit produced a molecular weight corresponding to the actual monomer, assuming a [Phi]"of 0.59 ml/g. The range of values for lnK2 (l/g) was from 3.4 to 12.9 (avg. ~6), indicating heterogeneity in association tendencies within the predominant monomer-dimer equilibrium. The range of values for the equilibrium constant indicates that >= 80% of the strands are present as dimers at 1 mg/ml (3.15 mM) at 1oC.
(iv) Seven data sets spanning a range up to ~0.8 mg/ml in 0.1 M Na+ at 40oC, best fitted a non-ideal single species model, with a molecular weight corresponding to the single chain oligomer if [Phi]"is 0.55 ml/g, and a small positive virial coefficient, probably resulting from a small effective charge on the molecule under these conditions. The ISS model fit the data almost as well, however, resulting in a molecular weight corresponding to the monomer if [Phi]"=0.56 ml/g. These results indicate that under these conditions the molecule exists as a single chain with a small effective charge.
The values of [Phi]"reported for a DNA double helix (19 ) are ~0.53 and 0.55 ml/g respectively in 0.15 and 0.6 M Na+. Those for d(A-G)10 at neutrality are somewhat different; at low temperatures and higher salt concentrations, where it is present as a hairpin dimer in a tetraplex structure (see below), [Phi]"= 0.57-0.59 ml/g, and for the monomer strand at 40oC, the value is 0.55-0.56 ml/g. Moreover, the single-stranded d(A+-G)10 helix at low temperature and its random coil form at elevated temperature exhibit a [Phi]"of 0.54 ml/g (11 ). Apparently, structure and base composition affect [Phi]", though exactly how is not obvious.
These direct observations of the molecular weights of the d(A-G)10 structures at neutrality completely corroborate the interpretations made on the basis of native PAGE.
To distinguish between hairpin and linear structures for the putative helical conformations indicated by the native PAGE and sedimentation equilibrium studies, the earliest sites of S1 nuclease cleavage were determined under various conditions of ionic strength, oligomer concentration and temperature under which those conformations had been identified. Because of the dynamic equilibrium between the different structures, it was necessary to find digestion conditions that would produce cleavage patterns characteristic of the predominant species. Realizing this goal was complicated because conditions for optimal structure stability do not correspond to those for optimal activity of the nuclease (see Materials and Methods). Moreover, near 0oC, where the helical conformations are most stable, the longer digestion time required to generate specific cleavage products also contributes to the background of observable products.
Figure 2 A shows the results in 0.64 mM d(A-G)10 in 0.15 M Na+, where the dimer helix is the predominant species; and Figure 2 B shows the results in 6 * 10-4 mM d(A-G)10 in 0.15 M Na+, which favors the monomer helix. A cleavage pattern essentially the same as for Figure 2 A was obtained when 5 mM Mg2+ was added to 0.64 mM d(A-G)10 in 0.15 M Na+ (data not shown). For both of these structures, the major nuclease sensitive sites are at the center of the molecule. S1 nuclease appears to cleave more readily after dG than after dA residues, with clearly preferred cleavage sites at G10 and G12, as established by comparison with the undigested d(A-G)6 and d(A-G)10 chainlength markers. Though the kinetics of digestion are obviously faster at higher temperatures, parallel experiments at 15 and 40oC or 5 and 15oC for each oligomer concentration reveal the same cleavage sites. Cleavage at the 5' and 3' ends of the strand(s) is also observed, in keeping with the propensity of helices to `fray' at their termini. However, at lower temperatures, cleavage at the termini is preferentially diminished relative to cleavage at the center of the molecule.
Analysis of thermal melting profiles for d(A-G)10 over a range of ionic conditions is not straightforward. Although determined over the entire near-UV range, profiles are shown only at 260 and 280 nm. Since 280 nm is isosbestic for the unstacking of dA residues (19 ), changes in the stacking of only the dG residues are observable there; and at 260 nm, the hyperchromic contribution from unstacking of dA residues is approximately twice that from dG residues (19 ). These hyperchromic changes can be due to unstacking of bases that may or may not be H-bonded. Only if the bases are in helices involving two or more strands, will the changes be responsive to ionic strength. Against this background, and with the knowledge of the single- and double-hairpin helices identified above at low temperatures, the melting profiles are qualitatively interpretable.
In 0.6 M Na+ or K+, or 0.15 M Na+/5 mM Mg2+, where the prevailing species at low temperature is the two-hairpin tetraplex, the melting profiles are quite similar (Fig. 3 B, D and E). The 260 and 280 nm profiles essentially coincide up to 25oC, suggesting that the small absorbance increase constitutes some `pre-melting' adjustment of tetraplex conformation. Between ~25 and 55oC, the dA residues apparently become extrahelical and the dG-tetrads begin to melt. These processes are not coincidental, based upon the profiles in Figure 3 B, D and E; and there is no reason why they should be. For, with tetrad melting, the liberated dG residues can stack with their dA nearest neighbors, and these dA-dG stacks will then melt with increasing temperature. Above ~60oC, the further melting of dA and dG residues must overlap sufficiently, since all that is observed is a gradual hyperchromic change at 280 nm.
The foregoing observations, together with those of Mukerji et al. (13 ), indicate that d(A-G)10 forms a bimolecular hairpin at low temperature, stabilized by alternating G-tetrads and intercalated dA residues. Since the complex exhibits a reversible UV melting profile, values for the thermodynamic parameters of the complex in 0.15 M Na+/5 mM Mg2+ (Fig. 3 E) were first obtained by assuming a two-state cooperative transition (28 ,29 ). Shape analysis of this thermal melting profile gave an equilibrium constant for complex formation at Tm (36.7oC) of 1.25 * 105, and thermodynamic values for formation of the complex of [Delta]HvH(shape) = -73.0 kcal/mol, [Delta]G298 = -3.4 kcal/mol and [Delta]S = -233.4 cal/mol[middot]K. However, the apparent breadth of the melting profile and wavelength dependence of Tm (see below) suggest that complications are introduced in assuming a two-state transition, even at higher ionic strength, as presupposed by the van't Hoff analysis.
DSC was therefore used to directly obtain thermodynamic parameters. The calorimetric transition curve (Fig. 6 ), corresponding to excess heat capacity of the sample versus temperature, is almost symmetrical and centered about a Tm of 43.97oC (higher than the value from the UV profile due to higher oligomer concentration), with flat and reliable (data not shown) pre- and post-transitional baselines. The area under the curve indicates a favorable enthalpy of formation ([Delta]Hcal) of -116.9 kcal/mol complex. The stability of the complex at 25oC is reflected in the value of [Delta]G = -7.0 kcal/mol, while the large entropy of formation, [Delta]Scal = -368.8 cal/mol[middot]K, is in keeping with the requirement for the association of four dG residues to form tetrads in the complex.
A variety of techniques have been used to provide evidence that at neutrality and low temperature d(A-G)10 is involved in multiple equilibria between monomer and dimer molecularities, whose principal conformations are a single hairpin duplex and a two-hairpin tetraplex. The equilibria depend upon oligomer and salt concentrations in addition to temperature. Under a variety of conditions, native PAGE reveals fast-, slow- and co-migrating species relative to d(C-T)10, a random coil of identical chainlength. The fast species is most likely accounted for by the compact nature of a fully formed single hairpin. The slow-migrating species, which approximately co-migrates with d(A-G)10.d(C-T)10, is best explained by some type of duplex. These molecularities are confirmed by sedimentation equilibrium measurements. In 0.1 M Na+ at 1oC, the oligonucleotide monomer is in equilibrium with the dimer; at higher concentrations, larger aggregates may be present to a very small extent. However, at 40oC in this buffer, the oligonucleotide is present as monomer only. In 0.15 M Na+/5 mM Mg2+ at 4oC, the oligonucleotide is present as a dimer. In 0.6 M Na+ at 4oC, the oligonucleotide is present as a dimer, with a very small amount of higher aggregates (possibly linear tetraplex).
One explanation for the approximate co-migration with the Watson-Crick duplex is a linear d(A-G)10[middot]d(A-G)10 model, as suggested by Rippe et al. (6 ); another possibility is a hairpin duplex in which the two strands interact to form a tetraplex. Although a hairpin duplex is more compact than a linear duplex, the homopurine nature and unusual stacking in this complex might make its mobility relative to the linear Watson-Crick duplex hard to predict. Hence, on the basis of gel migration patterns, neither the linear nor the hairpin models could be ruled out. Discrimination between these conformations was sought using dimethylsulfate (DMS) modification to determine the presence of any preferred cleavage site (data not shown). However, this approach only afforded inconclusive results, probably because methylation at the hairpin loops and termini causes a shift in the conformational equilibrium. A more selective conformational probe was found in S1 nuclease, which under conditions previously determined to stabilize the monomer or dimer complex, preferentially cleaves at positions G10 and G12 (the center of the molecule). This result establishes that the strand(s) of both the monomer and dimer species are hairpins, in one case forming a duplex, in the other a tetraplex.
A two-hairpin tetraplex could be interlocked, with the loops at opposite ends, as shown in Figure 7 A. This type of complex has been observed using 1H NMR spectroscopy in a structure containing G-tetrads and T residues at the turns (33 ). However, such an interlocked tetraplex should melt in an all-or-none fashion, which is not what is observed (Fig. 3 ). This absence of a two-state transition in the melting profiles at either low or even at higher ionic strength is most clearly demonstrated by the inequality of the transition enthalpies, [Delta]HvH(shape) and [Delta]Hcal. Therefore, side-by-side arrangements appear most likely.
Dimeric side-by-side and monomeric hairpins are schematically depicted in Figure 7 B. Homogeneous associations of single hairpins a, f and k are shown in horizontal rows b-e, g-j and l-o, respectively. Heterogenous associations of hairpins f and k are shown in horizontal rows p-s. Since dissimilar melting profiles at wavelengths more selective for A or G residues were indicated by both UV absorption and UVRR spectroscopies (13 ), the lack of evidence for A[middot]G base pairs would seem to eliminate structures a-e.
The preferred cleavage of both monomers and dimers at G10 and G12 can most easily be explained by single hairpins f and k and their homogeneous associations in g-j and l-o, respectively; however, heterogeneous associations of hairpins f and k, as in p-s, might yield similar results. While the earliest timepoints at which cleavage could be observed indicate equal cleavage at G10 and G12, the resolution of the assay is not good enough to allow discrimination between these models.
Based upon fluorescence energy transfer studies of d(G-A)7G and d(G-A)12G under similar ionic conditions, Rippe et al. (6 ) concluded that these oligomers form linear parallel duplexes stabilized by A[middot]A and G[middot]G base pairs. However, their results are not incompatible with our hairpin model, so long as the two hairpins are parallel. Moreover, the possibility exists that the presence of fluorescent tags at the 5' termini may shift the equilibrium from the two-hairpin tetraplex we have observed towards their suggested linear duplex. This interpretation is supported by the absence of a specific cleavage pattern when methylation by DMS was used to probe the conformation of d(A-G)10 in this study, or of d(G-A)7G and d(G-A)12G in the work of Rippe et al. (6 ). Using unmodified d(A-G)10, UV absorption and UVRR data (13 ) support only the interaction of G-residues.
G-tetrad helices have been recognized for a long time (38 -40 ), and recently hairpin helices containing G-tetrads with T-loops have been described and possibly implicated in telomere physiology (25 ,27 ,33 ,37 ). This work, together with that of Mukerji et al. (13 ), demonstrates that G-tetrads separated by non-H-bonded intercalated A residues provides another variant of this type of four-stranded helical structure.
Promoter regions comprised of GAGA[middot]TCTC duplex repeats have been shown to be specifically recognized by transcriptionally related proteins (38 ,39 ), and may play a role in nucleosomal organization (40 ). The types of hairpin structures described here seem worthy of consideration in developing mechanisms to explain recognition by transcriptional proteins of DNA sequences containing GAGA[middot]TCTC repeats, and how such repeats may play a role in nucleosomal organization.
This work was supported by grants to J.R.F. from the National Institutes of Health (GM42936) and to E.H.B. from the National Science Foundation (BIR9318373). M.C.S. was supported in part by a Sterling postdoctoral fellowship. We thank D. Zhu for help with the ultracentrifugation measurements, and R. Friedman and Y. Kim for preliminary thermal melting studies and helpful discussions.
*To whom correspondence should be addressed. Tel: +1 609 258 3927; Fax: +1 609 258 6730; Email: jrfresco@princeton.edu
+Present address: Department of Biochemistry, University of Capetown, South Africa
REFERENCES
Return