Limitations of computer modeling
To produce the large-scale bends in DNA, we adjusted the roll angles using an approximation of 10 bp/turn. In experiments, biochemical analyses have shown that straight DNA in solution has 10.4 ± 0.1 bp/turn, rather than 10 bp/turn as observed in the solid state.34 In vivo , supercoiling of DNA leads to a slight unwinding with the result of 11.1 bp/turn.35Because the corresponding parts of I1 and I2 half sites are 21 bp apart, it suggests that in this region of DNA, the twist is closer to 10.5 bp/turn to have the DBDs bind on the same side of the DNA. For 10, 10.4, and 11.2 bp/turn, the total twist over the 38 bp length of I1-spacer-I2 is three full helical turns and 288°, 235°, and 152° of the fourth helical turn, respectively. The incomplete fourth turn would affect which side of the DNA the DBDs could bind to and the relative rotation of the two DBDs. For producing tertiary and quaternary models of protein/DNA complexes, the value of 10 bp/turn was adequate for bending the DNA in a narrow step range (< 10 steps) but it introduces inaccuracies in the hydrogen bonding within DNA, the plane of the bend, and the relative orientation of the two DBDs.
To create the full model of the AraC/DNA complex, docking software was used as one stage of the process to form protein-DNA and protein-protein complexes. Our docking procedure however, cannot produce reliable results for residue/base, base/base, and residue/residue interactions. Some docking packages support the DNA sequence as input.17 However, we found that when the DNA structure is generated by a computer from a sequence, the DBD could not be satisfactorily docked onto DNA. We had to divide the sequence into segments and construct each segment from experimentally-determined coordinates. While the DNA segment method initially positioned the DBD on the DNA, it was still insufficient to achieve satisfactory docking in that both HTHs of the DBD did not sit fully in the major groove of DNA. Experiments have shown that protein and DNA both experience conformational changes upon binding, and that when protein bends DNA, there is an energy cost associated with distortion such that the DNA is not in its lowest energy configuration.14 Docking software should ideally allow for DNA flexibility and protein backbone and side chain flexibility, which are difficult features for any software package to possess.36, 37 We addressed this issue partially by manually adjusting the bend in the DNA and submitting the 3D structure to the docking software. This process was repeated until satisfactory surface complementarity was achieved. However detailed manipulations on the atomic scale were not conducted that would result in the true energy of the DNA or the protein. We also observed that upon docking, the two HTHs of the DBD can sit in the DNA major groove in various ways. We had to make a judgment on which model was the best; inaccuracies could have been introduced at this stage. Finally the docking approach in this paper did not take into account effects of solvation and salt.39 As a step towards determining tertiary and quaternary structure of the AraC/DNA complex, docking was useful; however conclusions about structure and interactions on an atomic scale must be drawn carefully.