The Genom of Homo sapiens.pdf

Annotation of Novel Proteins Utilizing A FunctionalGenome Shotgun Coupled with High-ThroughputProtein Interaction MappingJ.A. MALEK,* J.M. WIERZBOWSKI,* G.A. DASCH, † M.E. EREMEVA, ‡P.J. MCEWAN,* AND K.J. MCKERNAN**Agencourt Bioscience Corporation, Beverly, Massachusetts 01915; † Centers for Disease Control and Prevention,Atlanta, Georgia 30332; and ‡ University of Maryland, Baltimore, Maryland 21201It is quoted frequently that the amount of genomic sequencedata is increasing at a tremendous rate while functionalmethods for studying proteins have not kept pace.Many groups have attempted to address this issue by useof microarrays, yeast two-hybrid screens, and proteincomplex purification with subsequent identification. Theunderlying theme of these approaches is their use of“guilt-by-association” (Oliver 2000) methods for annotationof proteins of unknown function. The association ofa protein of unknown function with proteins of knownfunction is used to derive a potential function for the proteinof unknown function. Although large-scale microarrayexperiments have increased dramatically and are carriedout in numerous large and small laboratories,proteome-wide two-hybrid experiments have only beencarried out on yeast (Uetz et al. 2000; Ito et al. 2001) withsome large studies in Caenorhabditis elegans (Walhoutet al. 2000), and Helicobacter pylori (Rain et al. 2001),among others. Numerous review papers have been writtenon these large-scale two-hybrid studies, analyzing thedata, testing the data’s validity, and using the data to trainin silico protein interaction prediction software. The needfor further, validated, protein interaction information isclear. Among the challenges in generating proteomewideinteraction data are the lack of fully automated processes,the sheer amount of screening necessary to completeone map for one organism, and an incomplete graspof what constitutes a true, physiologically important proteininteraction. It is our belief that using comparative interactiondata will allow deciphering of what interactionsare physiologically valid. Validation of interactions hascentered around comparisons to databases of individuallyobtained and presumably more verified interactions, thepresence of interactions among proteins with similar expressionprofiles, and the frequency of interactionsamong proteins sharing similar biological processesand/or cellular compartments (Deane et al. 2002). Althoughthese methods of verification may add a level ofsignificance to any interaction, their absence should notper se be used to subtract from an interaction’s validity.Observing similar expression profiles between two proteinsmay suggest they are functionally related but doesnot mean that they physically interact. Observing interactionsamong proteins of different biological processesmay reveal a gap in our knowledge more than an incorrectinteraction. Observation of interactions among proteinsfrom different cellular compartments is less meaningfulin organelle-free microbes. Physiologically significantinteractions with a wide range of strengths have been observed,therefore a significance cutoff based on interactionstrength cannot be set at present.Automated DNA sequencing technology was heavilydeveloped during the Human Genome Project into a robustand cheap process. It would be of benefit to use thesedevelopments in advancing proteome-wide interactiondata. We have attempted to improve the ease with whichsuch studies can be carried out by adopting a strategy thatrelies on whole-genome shotgun sequencing and a bacterialtwo-hybrid system (Dove et al. 1997; Shaywitz et al.2000). The whole-genome shotgun method generatescloned overlapping fragments of genomic DNA which, ifcloned in the proper orientation and frame, can be expressedas a protein. The use of peptide fragments ratherthan full-length proteins has been shown to reduce falsenegatives (Ward et al. 2002) while offering the opportunityto localize the domain of a protein responsible for aninteraction. Use of the bacterial two-hybrid system allowsintegration into standard sequencing pipelines. The twovectors used in the system are standard sequencing vectorsthat are transformed together into an essentially standardcloning strain of Escherichia coli. The system, similarto various yeast two-hybrid systems, relies onrecruitment of transcriptional machinery to promoters upstreamof reporter genes. Briefly, a protein of interest isfused to the λcI protein which binds a λ operator on thereporter construct (Fig. 1a). A second protein of interestis fused to the RNA polymerase α-subunit. An interactionbetween the proteins of interest stabilizes the transcriptionalmachinery at a weak promoter upstream of the reporterconstruct (Fig. 1b). Interactions are observed as acolony able to grow in the presence of an antibiotic andthe absence of any carbon source other than lactose.Colonies can enter a standard sequencing pipeline at thispoint through the automated colony pickers. Sequencingof the bait of prey fragment is conducted with primersspecific for either vector.Cold Spring Harbor Symposia on Quantitative Biology, Volume LXVIII. © 2003 Cold Spring Harbor Laboratory Press 0-87969-709-1/04. 331