Binding of Two Different DNA Sequences by Conformational Switching

Proteins can recognize varied structures by making use of conformational flexibility. For example, DNA-binding proteins can have affinity for multiple sequences using a single binding pocket, by changing conformation in response to varied structures. To date, we know of no synthetic ligand which can recognize multiple sequences in this way. We now wish to report the design and synthesis of a macrocyclic ligand which can bind to two distinct DNA sequences in a mutually exclusive way by changing conformation on binding.

We constructed the 36-base oligonucleotide 1 by a template-directed cyclization of the linear precursor, 5′-dTCTCTTTTTTTTTTTCTCTCTCTTTTTTTTTTTCTCp.^1,2 The template oligomer 5′-dAAAGAGAGAGAAA (1 equiv) was used to align the reactive ends in an aqueous esterification reaction, as described previously.^3,4 The yield of purified circular 1 was 48%.^5,6 Also prepared were two nine-base purine oligomers, dAAAAAAAAA (2) and dAGAGAGAGA (3) (Figure 1).

Figure 1.

Sequence of macrocycle 1 and the conformational changes involved in the binding of 1 to the sequences 2 and 3. Dotted lines demarcate the four domains in 1. When both sequences are present, the macrocycle binds to one of the two as shown, but cannot bind to both simultaneously.

As expected from its non-self-complementary sequence, compound 1 alone shows no significant temperature-dependent melting behavior when examined at 260 nm in a pH 7.0 buffer containing 100 mM NaCl and 10 mM MgCl₂. In the presence of 1 equiv of the oligomer dA₉ (2), however, there is a sharp, apparently two-state transition showing a hyperchromicity of 23% and a melting temperature (T_m) of 48.0 °C (Figure 2 (inset)).⁷ This is a considerably higher T_m than that observed for the duplex dA₉·dT₉ (T_m = 28.7 °C) under identical conditions. This high affinity is consistent with the bimolecular triple helical structure expected for this complex,⁴ with the purine oligomer held between opposing dT₉ domains in the macrocycle.

Figure 2.

Mixing plots of mole fraction vs absorbance at 260 nm for circle 1 with purine complementary strands. Total DNA strand concentration was 3 μM, with molar ratios being varied as shown. Plots show 1 mixed with dA₉ (▲), 1 with d(AG)4A (■, offset 0.1 AU), and 1 with a 1:1 mixture of dA₉ and d(AG)₄A (●, offset 0.3 AU). From the plots, mole fractions for full complexation are 0.52, 0.53, and 0.47, respectively. (Inset: Melting profiles of hyperchromicity vs temperature for 1:1 complexes of circle 1 with dA₉ (T_m = 48.0 °C) and 1 with d(AG)₄A (T_m = 44.6 °C) at pH 7.0. Experiments were carried out as described.⁷

Significantly, the macrocycle displays almost identical binding behavior with a second, distinctly different sequence. When added to the oligomer dAGAGAGAGA (3), a melting transition having a 22% hyperchromicity and a T_m of 44.6 °C is seen at pH 7 (Figure 2). Once again, this is significantly higher than the corresponding duplex d(AG)₄A·d(TC)₄T, which has a T_m of 32.7 °C under these conditions.

Several observations support the involvement of two opposing binding domains, resulting in triplexes, for the two complexes. Mixing plots carried out at 25 °C between the circle and 2 (Figure 2) show that the stoichiometry of the complex is 1:1; if both dT₉ domains were involved in duplexes, the expected stoichiometry would be 2:1 dA₉:circle. Similarly, a mixing plot for the circle and dAGAGAGAGA reveals a stoichiometry of 1:1 as well (Figure 2). In addition, the denaturation of complex 1·2 shows a significant hyperchromicity of 13% at 284 nm, consistent with reported properties of T_n·A_n·T_n triple helices and not duplexes.⁸ Finally, the melting transition for complex 1·3 is pH dependent, with a higher T_m of 56.2 °C being observed at pH 6.0; this is again consistent with a triplex protonated at cytosines in the Hoogsteen strand.⁹

The formation of these two different complexes requires that the macrocycle adopt two opposite conformations in the binding of 2 and 3 (Figure 1). In forming a complex, two opposed nine-base domains move inward and align themselves with the complementary purine strand on opposite sides, forming two or three hydrogen bonds on each side of a base. The unbound nine-base domains then act as loops for bridging the bound domains. In order to bind the other sequence, the pairs of domains reverse roles, with the former binding domains acting as bridging loops and vice versa.

Interestingly, a further mixing experiment reveals that this complexation is mutually exclusive: the act of binding to one sequence precludes the binding of the other, even though the necessary bases are unpaired in the loops. When the circle is mixed at varied molar ratios with a one-to-one mixture of the two sequences (Figure 2), the mixing plot shows a stoichiometry of 1:1. This indicates that a given complex cannot bind the second sequence with its loops at 25 °C. While this macrocycle at all times carries the bases necessary for forming complexes with the two sequences, the act of binding causes a conformational switch, holding the loops in an inaccessible or unproductive conformation.¹⁰

To confirm that the macrocycle can recognize these two sequences within a longer strand of DNA, we synthesized a 33-nucleotide oligomer having the sequence 5′-dCACAAGAGAGAGAATCCCTAAAAAAAAAAACAC (4). This oligomer contains the two recognition sequences separated by seven bases. When mixing experiments are performed (pH 7.0, 25 °C) with this oligomer and the macrocycle 1, a plot of absorbance versus mole fraction (not shown) clearly demonstrates a binding stoichiometry of two macrocycles to 1 equiv of 4; this confirms that the unusual binding property is also seen with two sites in one strand.

The results demonstrate a new strategy for the binding of two or more specific nucleic acid sequences with a single ligand. Such a strategy may be useful in the design of multifunctional biosensors or of therapeutics directed to multiple medicinally important targets.