Analysis of Intermediate Resolution Structures (AIRS) - NCMI - Baylor College of Medicine

Analysis of Intermediate Resolution Structures (AIRS)

Wen Jiang, Matthew Baker, Steve Ludtke and Wah Chiu.
Collaborators are G Lu, A Bajaj and ZH Zhou

Advances in electron cryomicroscopy have made it possible to determine the structures of macromolecular complexes with high internal symmetry to 7-9 � resolution without using a crystal. The mining data from these large macromolecular complexes can yield significant structural and/or biochemical information but requires both advanced graphics and computational feature extraction techniques.

Due to the large volume of density maps, it is often difficult to visualize structural elements of an entire macromolecular complex using interactive graphics tools. A more practical approach in visualizing internal structure of large complex is to computationally segment individual proteins or smaller protein components. When dealing with these smaller subsets, it is possible not only to accurately segment domains but also to recognize secondary structure elements, such as individual a helices and b sheets, although individual strands cannot be identified. At this resolution, helices appear as cylinders of densities ~5-6 � in diameter, while b sheets are continuous and flat densities (Figure 1, outer shell protein P8 subunit extracted from the 6.8 � structure of Rice Dwarf Virus(Reference 1) (RDV)). Visual recognition of these features can be subjective and a quantitative method of analysis can make their identification unambiguous.

We have developed and tested software to address these aforementioned issues (Reference 2). Helixhunter identifies the locations of the helices within a density map objectively and automatically using a multi-step process of cross correlation, density segmentation, quantification, helix identification and an explicit description of the helices. For visualization purposes, final helices can be represented as cylinders (Figure 2) described by six parameters (three for center, two for orientation, and one for length). However, the identified helices do not contain any information on their sequence correspondence. To map the amino acid segments to the observed helices, it is possible to correlate sequence-based secondary structure predictions. In the outer shell protein P8 of RDV, where nine helices were predicted in the termini, it was possible to match the lengths of the helices identified in the 3D density map to those predicted from a consensus secondary structure analyses as shown in Figure 2. The connections between the helical densities can be seen in the lower domain of P8, allowing us to establish a rough backbone model for the lower domain of P81.

To assess the accuracy of such a prediction, DejaVu and COSEC can be used to perform ?spatial fold recognition?. A successful match in the helix arrangement between the helixhunter results of a structure in the PDB may suggest a possible fold homologue. 6 of the 9 helices within the lower domain of RDV P8 subunit were matched to bluetongue virus VP7 (centroid RMS < 5�). These proteins have only remote sequence similarity (<20%), however both proteins are from the same virus family and have similar structures and function in their respective capsids.

Additionally, individual protein or portion of a protein within the macromolecular complex may have a homologous structure. Once a homologous fold has been identified, it is necessary to localize this structure back to the entire complex. Foldhunter, a template-based cross-correlation tool, automatically searches all possible rotations and translations for the best fit of the homologous structure to the density map of the macromolecule. In RDV P8, it was possible to identify the middle sequence segment to be structurally similar to the b sheet domain of VP7 of the bluetongue virus. Figure 2 shows the localization of the putative jelly-roll b-sandwich fold of bluetongue virus to the density in the upper domain of RDV P8. Together with the earlier helixhunter results, a model for the entire structure of P8 can be obtained.

Such tools have now allowed a combination of intermediate resolution (7-9 �) electron cryomicroscopy and bioinformatics approaches to obtain a quasi-atomic model. This imaging technology is not restricted to viruses and may be extend to other macromolecular complexes with less or no symmetry. With increasing information and improving technology, we can now begin to use this information to support the structural analysis to obtain a quasi-atomic model for large complexes from intermediate structures.

Acknowledgement: This research has been supported by NIH (P41RR02250, AI38469, AI43656), Agouron Institute, W.M. Keck Center for Computational Biology and Robert Welch Foundation.

Selected Images

Click on image for larger version

Figure 1. Monomer of RDV P8

Figure 2. Interpreted secondary structure elements and their connectivity of RDV P8.

Selected Publications

	Zhou, Z. H., Baker, M. L., Jiang, W., Dougherty, M., Jakana, J., Dong, G., Lu, G. and Chiu, W. (2001). Electron cryomicroscopy and bioinformatics suggest protein fold models for rice dwarf virus. Nature Struct. Biol. 8: 868-873.
	Jiang, W., Baker, M. L., Ludtke, S. J. and Chiu, W. (2001). Bridging the information gap: Computational tools for intermediate resolution structure interpretation. J. Mol. Biol. 308:1033-1044.
	Chiu, W., Baker, Matthew L., Jiang, W. & Zhou, Z. H. (2002) Deriving folds of macromolecular complexes through electron cryomicroscopy and bioinformatics approaches. Curr Opin Struct Biol, 12:263-269.