The RNA kink-turn is an important recurrent RNA motif; an internal loop with characteristic consensus sequence forming highly conserved three-dimensional structure. Functional arrangement of RNA kink-turns shows a sharp bend in the phosphodiester backbone. Among other signature interactions, kink-turns form A-minor interaction between their two stems. Most kink-turns possess extended A-minor I (A-I) interaction where adenine of the second A center dot G base pair of the NC-stem interacts with the first canonical pair of the C-stem (i.e., the receptor pair) via trans-sugar-edge/sugar-edge (tSS) and cis-sugar-edge/sugar-edge (cSS) interactions. The remaining kink-turns have less compact A-minor 0 (A-0) interaction with just one tSS contact. We show that kink-turns with A-I in ribosomal X-ray structures keep G=C receptor base pair during evolution while the inverted pair (C=G) is not realized. In contrast, kink-turns with A-0 in the observed structures alternate G=C and C=G base pairs in sequences. We carried out an extended set (similar to 5 mu s) of explicit-solvent molecular dynamics simulations of kink-turns to rationalize this structural/evolutionary pattern. The simulations were done using a net-neutral Na+ cation atmosphere (with similar to 0.25 M cation concentration) supplemented by simulations with either excess salt KCl atmosphere or inclusion of Mg2+. The results do not seem to depend on the treatment of ions. The simulations started with X-ray structures of several kink-turns while we tested the response of the simulated system to base substitutions, modest structural perturbations and constraints. The trends seen in the simulations reveal that the A-I/G=C arrangement is preferred over all three other structures. The A-I/C=G triple appears structurally entirely unstable, consistent with the covariation patterns seen during the evolution. The A-0 arrangements tend to shift toward the A-I pattern in simulations, which suggests that formation of the A-0 interaction is likely supported by the surrounding protein and RNA molecules. A-0 may also be stabilized by additional kink-turn nucleotides not belonging to the kink-turn consensus, as shown for the kink-turn from ribosomal Helix 15. Quantum-chemical calculations on all four A-minor triples suggest that there is a different balance of electrostatic and dispersion stabilization in the A-I/G=C and A-I/C=G triples, which may explain different behavior of these otherwise isosteric triples in the context of kink-turns.