Studies of DNA dumbbells. I. Melting curves of 17 DNA dumbbells with different duplex stem sequences linked by T 4 endloops: Evaluation of the nearest‐neighbor stacking interactions in DNA

Seventeen DNA dumbbells were constructed that have duplex sequences ranging in length from 14 to 18 base pairs linked on the ends by T 4 single‐strand loops. Fifteen of the molecules have the core duplexes with the sequences 5′ G‐T‐A‐T‐C‐C‐(W‐X‐Y‐Z)‐G‐G‐A‐T‐A‐C 3′ , where (W‐X‐Y‐Z) represents a unique combination of A · T, T · A, G · C, and C · G base pairs. The remaining two molecules have the central sequences (W‐X‐Y‐Z) = A‐C and A‐C‐A‐C‐A‐C. These duplex sequences were designed such that the central sequences include different combinations of the 10 possible nearest‐neighbor (n‐n) stacks in DNA. In this sense the set of molecules is complete and serves as a model system for evaluating sequence‐dependent local stability of DNA. Optical melting curves of the samples were collected in 25, 55, 85, and 115 m M [Na + ], and showed, regardless of solvent ionic strength, that the transition temperatures of the dumbbells vary by as much as 14° for different molecules of the set. Results of melting experiments analyzed in terms of a n‐n sequence‐dependent model allowed evaluation of nine independent linear combinations of the n‐n stacking interactions in DNA as a function of solvent ionic strength. Although there are in principle 10 possible different n‐n interactions in DNA, these 10 are not linearly independent and therefore can not be uniquely determined. For molecules with ends, there are 9 linearly independent combinations, as opposed to circular or semiinfinite repeating copolymers where only 8 linear combinations of the 10 possible n‐n interactions are linearly independent. The n‐n interactions are presented as combinations of the deviations from average stacking for the 5′‐3′ base‐pair doublets, δ G i , and reveal several interesting features: (1) Titratable changes in the values of δ G i , with changing salt environment are observed. In all salts the most stable unique combination is δ G 4 = (δ G GpC + δ G CpG ,)/2, and the least stable is the GpG/CpC stack, δ G 2 = δ G GpG / CpC . (2) The χ 2 values of the fits of the evaluated δ G i 's to experimental data increased with decreasing [Na + ], suggesting that significant interactions beyond nearest neighbors become more pronounced, particularly at 25 m M Na + . (3) In 85 and 115 m M Na + , where the n‐n approximation seems to be most valid, the absolute value of δ G i for any n‐n stack or average of two n‐n stacks is not more than ∼ 220 cal/mole, indicating that deviations from average stacking due to n‐n interactions represent about 15% of the total stability of a base pair. The overall thermodynamic stability of DNA is predominantly determined by the sequence content (%G · C). Even though the contribution of n‐n interactions to overall stability are intrinsically small, reliable predictions of DNA transition temperatures de novo from sequence can be significantly compromised by cumulative errors in the δ G i 's. (4) Comparisons of our set of n‐n linear combinations evaluated in 115 m M Na + with various published sets evaluated from melting experiments of long restriction fragments, synthetic polymers, and short oligomers, and those obtained from a reanalysis of published melting data of synthetic polymers, are presented. The analysis reveals a major consensus agreement between n‐n free energies evaluated from melting data of restriction fragments and long synthetic repeating copolymers. In contrast, only a minor consensus agreement is obtained between our n‐n set and these values or those obtained from melting analysis of a combination of short oligomers and long polymers or those theoretically calculated. Results of these comparisons suggest the values of n‐n interactions evaluated from DNA melting curves depend on the length of the melted duplex regions of the DNA molecules that comprise the sample set.