(311f) Computational Predictions of Protein-Protein Structures and Interfaces within a Messenger RNA Degradation Machine | AIChE

(311f) Computational Predictions of Protein-Protein Structures and Interfaces within a Messenger RNA Degradation Machine


Ovacik, M. A. - Presenter, Rutgers University
Knapinska, A. - Presenter, UMDNJ-Robert Wood Johnson Medical School
Brewer, G. - Presenter, Rutgers University
Androulakis, I. - Presenter, Rutgers - The State University of New Jersey

Gene expression spans DNA transcription to protein processing and localization, and its regulation is central to proper cell function. Though multiple mechanisms control gene expression, they can be analyzed as a network of coordinated processes. One control mechanism, mRNA stability, has recently received increasing attention, since it is a major determinant of protein levels. This is particularly true for regulatory proteins involved in embryogenesis, immune responses, and diseases such as cancer. The stability of an mRNA is determined by molecular interactions between RNA-binding proteins and sequence elements within the mRNAs that they control. One of the most studied families of regulatory mRNA sequences is the A+U-rich element (ARE), located within the 3'-untranslated region of as many as 8% of mRNAs within the human transcriptome. Interaction of an ARE with the protein AUF1 targets the mRNA for rapid degradation. Numerous studies have examined its RNA-binding mechanism, but it is unknown how AUF1 recognizes AREs. The 3-D structure of AUF1 is not solved yet except for one region of the protein that contributes to its ARE-binding function. This region is known as an RNA-recognition motif (RRM), and AUF1 bears two, nonidentical RRMs that are both required for ARE binding. Characterization of mRNA binding sites within the AUF1 polypeptide will be very helpful toward understanding the details of their intermolecular interactions. In vivo experiments demonstrated that AUF1 forms complexes with several proteins, including translation initiation factor eIF4G, heat shock proteins Hsc/Hsp70, and poly(A)-binding protein. These proteins act in concert to recruit the mRNA degradation machinery to mRNAs bearing AREs. Thus, modeling AUF1-ARE interactions, as well as interactions of AUF1 with these auxiliary proteins, will permit us to link protein-protein and protein-RNA interactions to posttranscriptional control of gene expression(5). Protein 3-D structure coupled with predictive procedures for protein-protein interactions are becoming more successful each year, as judged by validation of predictions with experimental data. We have the advantage of utilizing predictive methods with the guidance of experimental results. Consequently, we can use predictive methods to formulate hypothesis-driven experimental studies. In this study, our goal is to demonstrate that experimental and computational methods can be employed simultaneously to understand mRNA and AUF1 interactions. The Robetta server, introduced by the University of Washington, is widely used and fully automated for predictions of protein structure (1) . The server merges template-based and de novo structure prediction techniques to produce native structures of the proteins. Utilizing this server, we have obtained five different models of AUF1 structure for each of its four protein isoforms: p37, p40, p42 and p45. ClusPRO is the first fully automated web server for predictions of protein-protein interactions. The required input is structures of proteins known to interact. The docking algorithm produces billions of presumed complexes of rigid bodies and a filtering method is applied to cluster complexes with good electrostatic and desolvation free energies(2). Moreover, a user can block or select certain regions in the protein. Specific algorithms for homo-multimers are also included in the server. Experimental results show that AUF1p37 forms homodimers in vitro and this requires the N-terminal 29 amino acids of the protein. This alanine-rich region is located at the N-terminus of all four protein isoforms of AUF1. By selecting the N-terminal 29 residues and employing a homo-multimer algorithm, we have generated predicted 3-D structures of each homodimer. Although 3-D structure of the C-terminal RRM is solved by NMR, the structure of RNA bound to the complex could not be determined because of the high flexibility of unfolded mRNA. However, the U1A-RNA complex, the structure and length of which are similar to the C-terminal RRM of AUF1, can be used as a starting point to model mRNA-AUF1 binding sites(4). Additionally, the thermodynamically optimal hairpin structure of one ARE-RNA was found to regulate mRNA decay (3). We have obtained a similar structure for the hairpin by performing a 5 ns, molecular dynamics simulation. High Ambiguity Driven protein-protein DOCKing (HADDOCK) will be used to obtain the AUF1-mRNA complex, because RNA has a flexible structure and rigid body docking will not be suffucient to explain the binding characteristics between AUF1-mRNA complexes. An additional modeling goal for AUF1 is to understand the effect of its phosphorylation on homodimer formation for the p40 isoform, as this posttranslational modification alters both its ARE interaction dynamics and protein-induced changes in RNA structure. Structure for AUF1p40 phosphorylated at ser83 and ser87 will be obtained using Molecular Operating Environment (MOE). Comparison of mRNA binding to phosphorylated and unphosphorylated protein will help as to characterize effects of phosphorylation on mRNA stability. As noted above, AUF1 forms a complex with several auxiliary proteins and the ARE-RNA to trigger mRNA degradation. Obtaining structures for this complex is a challenge, because the structures of the individual subunits of this complex are not known. However, we can eliminate a considerable number of experiments by introducing several predictions for potential binding sites of these auxiliary proteins to the AUF1-mRNA complex.

1. Chivian D, Kim DE, Malmstrom L, Bradley P, Robertson T, et al. 2003. Automated prediction of CASP-5 structures using the Robetta server. Proteins 53 Suppl 6: 524-33

2. Comeau SR, Gatchell DW, Vajda S, Camacho CJ. 2004. ClusPro: a fully automated algorithm for protein-protein docking. Nucleic Acids Res 32: W96-9

3. Fialcowitz EJ, Brewer BY, Keenan BP, Wilson GM. 2005. A hairpin-like structure within an AU-rich mRNA-destabilizing element regulates trans-factor binding selectivity and mRNA decay kinetics. J Biol Chem 280: 22406-17

4. Kajita Y, Nakayama J, Aizawa M, Ishikawa F. 1995. The UUAG-specific RNA binding protein, heterogeneous nuclear ribonucleoprotein D0. Common modular structure and binding properties of the 2xRBD-Gly family. J Biol Chem 270: 22167-75

5. Knapinska A, Irizarry-Barreto P, Adusumalli S, Androulakis I, Brewer G. 2005. Molecular mechanisms regulationg mRNA stability: physiologcal and pathological signifigance. Current Genomics 6: 471-86