The Genom of Homo sapiens.pdf

FUNCTIONAL SHOTGUN SEQUENCING 333Figure 4. An interaction map of the screening. Results of thescreening are displayed using Cytoscape (Ideker et al. 2002)with the bait proteins displayed in red. Interactions between twoproteins are represented as edges between the two nodes. Interactionsoccurring with multiple subunits of the T4SS are displayedin the inner circle and subunit-specific interactions aredisplayed in the outer circle.teraction. 96% of a randomly selected set of interactionswas capable of reconstitution: Purification of plasmid andreintroduction into fresh cells reconstituted the interaction.Approximately 40% of intra-complex interactions reportedin a study of A. tumefaciens T4SS orthologs usingthe yeast two-hybrid (Ward et al. 2002) were observed inour study of R. sibirica subunits using the bacterial twohybridsystem. This number may improve as we clone andscreen regions in the subunits not screened due to the factthat they were missing from our baits. Earlier analysis ofthe overlap between any two studies of the same interactionsusing the yeast two-hybrid revealed an ~16–35%overlap (Matthews et al. 2001) showing that our results aretoward the upper end of this observation. We observed interactionsamong subunits not observed in the Ward andcolleagues study, and these are being followed up withfurther investigation. Among the proteins interacting withmultiple subunits were proteins involved in cellwall/membrane biogenesis, defense mechanisms, lipidtransport and metabolism, and uncharacterized proteins.Strong interactions were also observed with some proteinsinvolved in DNA replication or repair. One potential falsepositive was observed as a strong interaction betweenDnaK, a chaperone, and most of the subunits.DISCUSSIONUsing protein interactions, it is possible to begin assigningfunctional context to previously unannotated pro-teins. We have focused our future studies on the proteinsinteracting with the T4SS in our search for rickettsial hosteffectors. The use of multiple methods such as microarrays,two-hybrid systems, and large-scale gene knockdownswill be necessary to begin functionally annotatingthe rapidly increasing number of novel proteins.On the basis of various studies, the average protein isexpected to have ~3 interacting partners. Our study revealssignificantly more interactions than this average. Apossible explanation is that proteins we investigated arepart of a transmembrane complex known to transport variouseffectors thus increasing the number of proteins theywould interact with. This may only be a partial explanation.We believe that previous studies have underestimatedthe average number of interacting partners per protein.A potential reason for this is the lack of depth ofscreening, or the stringency used in carrying out interactions,which could have eliminated valid, weaker interactions.If this is true, then the complexity of interactionsoccurring in any given cell could be immense.Interactions between apparently unrelated proteins arequite likely meaningless to a protein’s direct function, althoughthey may have a critical role in cell physiologyand intracellular movement. We believe the approach ofcapturing as many interactions as possible in a screen isimportant. In “casting a wide net,” the protein interactionspace is narrowed without loss of sensitivity. Further validationthrough other systems such as yeast two-hybridand protein complex studies would then further narrowwhat is considered to be of importance. We stress againthat comparative protein interaction mapping of organismswill offer a critical method for validating interactions.The opposite of protein–protein interactions, protein–proteinrepulsions, may play as important a role inthe forming of localized protein complexes in the cell.This is a relatively unexplored possibility, and studyingthe combination of protein–protein interactions and repulsionsmay give more insight into cell physiology.We have reported a method to improve the rate atwhich necessary proteome-wide interaction studies canbe generated. By combining the methods of shotgun sequencingand a bacterial two-hybrid system, we hope thatthe number of organisms with interaction maps will improveand bring this field into the era of comparativestudies.REFERENCESDeane C.M., Salwinski L., Xenarios I., and Eisenberg D. 2002.Protein interactions: Two methods for assessment of the reliabilityof high throughput observations. Mol. Cell. Proteomics5: 349.Dove S.L., Joung J.K., and Hochschild A. 1997. Activation ofprokaryotic transcription through arbitrary protein-proteincontacts. Nature 386: 627.Ideker T., Ozier O., Schwikowski B., and Siegel A.F. 2002. Discoveringregulatory and signaling circuits in molecular interactionnetworks. Bioinformatics 18: S233.Ito T., Chiba T., Ozawa R., Yoshida M., Hattori M., and SakakiY. 2001. A comprehensive two-hybrid analysis to explore theyeast protein interactome. Proc. Natl Acad. Sci. 98: 4569.Matthews L.R., Vaglio P., Reboul J., Ge H., Davis B.P., Garrels