PREDICTIVE MODELS FOR TRANSCRIPTIONAL ENHANCERS

Abstract

Our objective for the project is to develop a predictive method for studying and understanding the interactions between an enhancer element and it's targeted binding protein, Based on homoloFy modeling approach, we have developed a method for predicting structures o l both the protein and DNA in the complex. Using homeodomain-DNA complex as our model system, we have shown that the prediction method is capable of generating structures that compare favorably with experimental results. The results from our study also provide insights to the sequence-specific recognition of the DNA by its targeted protein as well as the differences in binding due to site-directed mutations in the protein. Our prediction method is general, therefore, can be used for predicting structures of various protein-DNA complexes. Background and Research Objectives Being able to identify control elements such as transcriptional enhancers is an important step toward the understanding of gene expression. This process requires the knowledge of how these elements interact specifically with the corresponding proteins. The homeodomain is a small conserved protein motif that binds specifically to DNA and plays a central role in gene regulation. The homeodomain/DNA complex provide an ideal system for studying the interactions between gene control elements, such as enhancers, and their binding proteins. To accomplish this goal, we developed a method that is capable of predicting structures of both protein and DNA in a protein-DNA complex. Using this method, we studied the specific interactions between several homeodomain-DNA complexes. Importance to LANL's Science and Technology Base and National R&D Needs from experiments and genome sequence data for predicting structures and functions of unknown proteins. We have furthered the homology modeling method to include the prediction of both the protein and DNA in iI protein-DNA complex. Our prediction method is both novel and general and can be used for predicting structures of various protein-DNA complexes. This method should be very useful in the structural genome initiative. Homology modeling provides the crucial step in using the known structures derived *Principal Investigator email: gxg @lanl.gov Scientific Approach and Accomplishments Homology modeling is a method that predicts thestructure of a protein (target) based on the known structure of a protein (template) that has homologous sequence [l]. This modeling approach has gained rapid popularity in recent years. Up-to-date, homology modeling remains the only method that can provide models with a root-mean-square error lower than 2 A [2]. €allowing the basic concept of homology modeling, we have developed a method for predicting structures of both protein and DNA. Using the engrailed homeodoniain protein-DNA complex structure (1 HDD in Protein Data Bank[3]) as the template (central molecule in Fig. la), we modeled the structures of the complex structures of both the antennapedia (left molecule in Fig. la) and the Mata2 (right molecule in Fig. 1 a) homeodomains. The sequence identity between the engrailed and the antennapedia homeodoamins is higher (50%) than that between the engrailed and the Mata2 homeodomains (30%). The root-mean-square errors for antennapedia homeodomain-DNA complex are 2.8 A for protein (1.1 A for C,) and 1.72 8, for DNA. The corresponding values are 3.0 A (2.1 A for C,) and 2.0 8, for Mata2 homeodomain-DNA complex. These results show that our modeled structures compare favorably with those determined experimentally. Based on the structure of the engrailed homeoclomain protein-DNA complex (1HDD in Protein Data Bank), we modeled the binding of the protein to different locations of the DNA. The molecule in the center of Fig. lb shows the protein binds to the protein recognition sequence TTAATG (the corresponding base pairs are depicted as six horizontal lines). In our study, the DNA is moved 1 or 2 base pairs up and down with respect to the protein. With each of the different binding, modes, the structure of the complex is energy minimized using AMBERC.11. The resultant structures are shown in Fig. 1 b. The total number of hydrogen-bonds (H-bonds) between the protein and the DNA, as well as the H-bonds between the protein and the bases (listed in parenthesis), are listed under each of the complexes. While the total number of H-bonds between DNA backbone and protein remains unchanged, the number of H-bonds between DNA bases and protein is reduced by 2-3 when protein is moved from its optimum DNA binding position. This allows the sequence discrimination of the DNA binding for the protein.