My interest in biology was first piqued by an inspiring high school teacher in England, Michael Baron, who communicated his absolute passion for the biochemical mechanisms underlying cellular metabolism. I was enthralled by the idea that one could understand how living organisms work at a fundamental level. He also had a great interest in ecology, and gave me an appreciation for the breadth of biological systems. This enthusiasm was reinforced as an undergraduate at Cambridge University, where I had the good fortune to do a project under the guidance of Tim Hunt, then working on protein synthesis, but perhaps best known for his subsequent discovery of cyclins that control the cell cycle. I still find it remarkable that Tim would spend so much energy mentoring a student and revealing the extraordinary excitement of discovering something new. At Tim’s suggestion, I went to the Imperial Cancer Research Fund in London to do a Ph.D. with Alan Smith, where I met Steven Martin. In addition to being a very funny guy, Steve had done some of the pioneering work establishing the identity of the v-src retroviral oncogene. He introduced me to the idea that in cells carrying a temperature-sensitive mutant of v-src, the transformed state could be turned on and off at will, simply by shifting the temperature of the cell culture up and down. This experiment suggested that virtually every aspect of cellular behavior, including gene expression, cytoskeletal architecture, progression through the cell cycle, metabolism, protein trafficking, and so forth, could be altered by activating a single oncogene product. By inference, the cell must have a molecular infrastructure through which a single protein can transmit a biochemical signal to many different intracellular targets in a coordinated fashion. I resolved to learn as much as possible about oncoproteins, with the hope that I could find an underlying theme to the organization of signaling pathways in animal cells.

Work from many labs, including Michael Bishop and Harold Varmus, Ray Erikson, Tony Hunter and Michael Waterfield, led to the realization that v-src encodes a protein kinase that specifically phosphorylates tyrosine, and to the stunning insight that many normal growth factor receptors possess intrinsic tyrosine kinase activity. My initial question could then be reduced to the problem of how tyrosine kinases recognize their targets. Postdoctoral work I had undertaken in Berkeley with Steve Martin (by then transplanted from the UK) and Peter Duesberg had interested me in a little-known cytoplasmic tyrosine kinase termed v-Fps, the transforming protein of Fujinami sarcoma virus. Once ensconced in my own lab at the University of British Columbia in Vancouver, Canada, we undertook to introduce multiple site-directed mutations into the v-Fps gene, in an effort to understand the regions of the protein that might be required for its transforming activity. Although this would now be a simple and mundane task, in the early 1980s this was a somewhat daunting challenge. Serendipitously, Michael Smith at UBC was in the midst of developing the technique of oligonucleotide-directed mutagenesis, and we benefited greatly from access to this new procedure. Not surprisingly, we found that the v-Fps tyrosine kinase domain was critical for its cancerous properties. 

Furthermore, in the lab, Geraldine Weinmaster found that autophosphorylation at a specific tyrosine within the v-Fps kinase domain was essential for full catalytic and transforming activity, providing formal evidence for the importance of tyrosine phosphorylation as a modification that could regulate enzymatic activity. However, Ivan Sadowski and Jim Stone noticed that mutations in a non-catalytic region N-terminal to the kinase domain also affected v-Fps kinase activity, as well as its ability to phosphorylate cellular substrates, and its capacity to transform cells. Using partial proteolysis, we also obtained evidence that this region might correspond to a folded structure. While sequence-gazing one day, I realized that Src and Abl, two other cytoplasmic tyrosine kinases, had a very similar sequence to Fps in just the same position preceding the kinase domain. Finding the description of "a region of 100 amino acids N-terminal to the kinase domain conserved between Fps, Src and Abl" somewhat unwieldy to use repeatedly in a paper, we christened the common sequence as the Src homology 2 (SH2) domain, where the kinase domain itself was implicitly SH1. 

Thus in 1986, we proposed that the SH2 domain, while not required for catalysis per se, was nonetheless involved in the regulation of kinase activity, and in the recognition of tyrosine kinase targets in the cell. This led to the idea that the molecular infrastructure we were seeking, which could organize intracellular signaling pathways, might be composed of non-catalytic protein modules that could facilitate the interactions of proteins with one another.

Thanks to work by Hidesaburo Hanafusa and John Knopf, we came to appreciate that there were other conserved non-catalytic domains involved in tyrosine kinase signaling, particularly the SH3 domain, which could be found in the same proteins as SH2 domains. Strikingly, the v-Crk oncoprotein discovered by Bruce Mayer in the Hanafusa lab contained only an SH2 and an SH3 domain. In addition, an increasing number of proteins involved in normal intracellular signaling, such as phospholipase C-γ, turned up with SH2 and SH3 domains. To my great excitement, we found that Ras GTPase activating protein, which Frank McCormick had shown to have two SH2 domains, was a substrate for tyrosine phosphorylation and was also associated with two other phosphotyrosine-containing proteins. This strongly suggested that SH2 domains were the hallmark of cytoplasmic proteins involved in tyrosine kinase signaling, and further argued that physical protein complexes, mediated by modules such as the SH2 domain, were involved in transmitting signals from growth factor receptors to the Ras pathway.

Meanwhile, in the lab, Michael Moran (now based in Toronto) had been making TrpE-SH2 fusion proteins in bacteria to use as immunogens to raise antibodies to SH2-containing proteins. Making use of these reagents, we decided to see what we could pull down from lysates of cells stimulated with growth factors, or transformed by tyrosine kinase oncogenes. Remarkably, Mike, together with Deborah Anderson, found that the SH2 domains of proteins such as PLC-γ, GAP and Src could bind directly to activated receptor tyrosine kinases, but only when the receptor had been modified by autophosphorylation. These results, coupled with data emerging from the Hanafusa laboratory, led to the proposition that receptor autophosphorylation created binding sites for the SH2 domains of cytoplasmic proteins, which as a consequence of their physical recruitment to the activated receptor were able to stimulate intracellular signaling pathways.

The last decade has seen a tremendous expansion of this rather simple notion. We now appreciate that there are a large number of protein modules that mediate protein-protein and protein-phospholipid interactions. We know not only of domains that recognize phosphotyrosine, but also of modules that specifically recognize phosphothreonine/serine and acetylated motifs. Other interaction domains bind proline-rich sequences, or motifs at the C-termini of proteins, or mediate protein oligomerization. Thanks primarily to the work of Lewis Cantley, we understand in some detail the specificities of individual domains for distinct peptide sequences, and have a growing ability to predict the interaction partners of a particular protein based purely on primary amino acid sequence information. These domains are found in proteins that control signaling pathways, the cell cycle, gene expression, cytoskeletal architecture, cell metabolism, protein degradation, and protein trafficking. Furthermore, it is common to find multiple distinct interaction domains in a single polypeptide, and to observe such proteins in increasingly complex networks of associated proteins. It therefore appears that the wiring diagram of the cell is built up through the reiterated use of a limited set of interaction domains.

Our lab is currently interested in the involvement of protein interaction modules in a variety of systems, especially in cell movement and polarity, and in the development of cells and complex structures in an animal. Particularly, we are fascinated by the idea that one can define the functions of the products of human genes to a first approximation by identifying their interaction partners, a concept we have termed "guilt by association." New techniques such as biological mass spectrometry lend themselves to the rapid and sensitive analysis of protein complexes and modifications, and to the ultimate derivation of a map of the cell. As a final, and perhaps fanciful idea, if we can understand in detail how cells are wired together through the assembly of protein complexes, we may be able to devise ways of re-wiring the cell to undertake new functions, or at the least to re-route signals down novel pathways. If applied to aberrantly active signaling pathways, as found in many disease states, such approaches might have a broad range of therapeutic applications.