Oncogene (1998) 17, 1469 ± 1474 1998 Stockton Press All rights reserved 0950 ± 9232/98 $12.00 http://www.stockton-press.co.uk/onc From Src Homology domains to other signaling modules: proposal of the `protein recognition code' Marius Sudol The Department of Biochemistry, Mount Sinai School of Medicine, One Gustave Levy Place, New York, NY 10029, USA The study of oncogenes has illuminated many aspects of domains have identi®ed a set of common features cellular signaling. The delineation and characterization which aids in their description. The domains are of protein modules exempli®ed by Src Homology composed of 40 ± 150 amino acids which fold to domains has revolutionized our understanding of the generate one or more ligand-binding surfaces, known molecular events underlying signal transduction path- as `recognition pockets'. Generally, conserved residues ways. Several well characterized intracellular modules of the domain are directly involved in mediating which mediate protein-protein interactions, namely SH2, contacts with the amino acids of the ligand and in SH3, PH, PTB, EH, PDZ, EVH1 and WW domains, maintaining the structure of the domain. Ligands are directly involved in the multitude of membrane, interact with their complementary domains through cytoplasmic and nuclear processes in multicellular and/or short sequence motifs (also called core motifs), unicellular organisms. The modular character of these composed of only 3 ± 6 amino acids (Figure 1). The protein domains and their cognate motifs, the univers- sequence of the core determines the selection of the ality of their molecular function, their widespread domains (Songyang and Cantley, 1995). The most occurrence, and the speci®city as well as the degeneracy `important' amino acid within the core is traditionally of their interactions have prompted us to propose the designated `0' and the remaining amino acids are concept of the `protein recognition code'. By a parallel designated `-1', `-2', and `1', `2', toward the amino- analogy to the universal genetic code, we propose here and carboxy-terminal directions, respectively. With that there will be a ®nite set of precise rules to govern one or two exceptions (Songyang et al., 1997), the and predict protein-protein interactions mediated by core motifs are surrounded by ¯anking amino acids at modules. Several rules of the `protein recognition code' the amino-(fN) and carboxy-terminal (fC) regions which have already emerged. dictate the speci®city of interaction within a given family of modules (Rickles et al., 1995). As expected, Keywords: protein modules; protein motifs; genetic variabilities in primary structures of domains also code; phosphotyrosine; polyproline contribute to the speci®city of interaction with the ligand. We propose to use a term, `epsilon determi- nant(s)' (e from the Greek word for speci®city: eidikotZta, pronounced edekohteta) for those amino The SH2 domain is a paradigm of small conserved acids within domains which determine ligand predilec- protein modules that mediate the formation of tions. The e determinant is represented by one or intracellular complexes in vivo and which participate several amino acid positions located mainly within the in a diverse array of signaling events (Pawson, 1995). conserved structure of the domain, and usually in the With the discovery and characterization of the SH2 ligand binding interface. Hydrophobic amino acids are domain, two important features became immediately represented by c (Psi) and aromatic amino acids by z apparent: First, its widespread occurrence and func- (Zeta). Any amino acid or an amino acid from a tion, and second, its speci®city of interaction with subset that can confer speci®city are indicated by x. In cognate protein ligands (Schlessinger, 1994). Subse- the latter case the x amino acids correlate with the e quent delineation of other intracellular signaling determinant (see Figures 2 and 3, and SH2 rules modules including SH3, PTB, PDZ, EVH1, and WW below). With the aid of this simple glossary, we will domains, as well as their binders, con®rmed these two brie¯y summarize the emerging rules of protein- characteristics (Cohen et al., 1995; Sudol, 1996a). protein interaction mediated by modules, using Intracellular protein modules are frequently referred several of the best characterized domains as exam- to as biological `Velcro', or as molecular adhesives. ples. Implicitly, the proposed formalization is meant However, the interactions that these modules mediate to unify several diverse descriptions currently in use in with their cognate ligands do not resemble a uniform order to facilitate discussion in the rapidly developing structure or a mechanism implicit in the analogy to ®eld of signaling by protein modules. molecular glue. Rather, each of the domains represents a unique three-dimensional structure that is comple- mentary to a speci®c sequence motif of its ligand (Kuriyan and Cowburn, 1997). SH2 rules The SH2 (Src Homology 2) domain is widely General rules distributed among cytoplasmic proteins, and the discovery of its ligands had a great impact on the Structural and functional studies of the domain-ligand understanding of cellular signaling (Sadowski et al., complexes for SH2, SH3, PTB, PDZ and WW 1986; Pawson, 1995). The domain functions primarily Protein-protein interaction by modules MSudolet al 1470 Figure 1 Schematic representation of widely distributed modular protein domains and their cognate ligands. Protein domains are structurally coherent, three-dimensional units that have the ability to fold independently (Siddiqui and Barton, 1995; Bork et al., 1996). They are composed of a minimum of 35 ± 40 amino acids. These protein domains include: SH2, PTB, PDZ, SH3, EVH1 and WW. They mediate a vast array of interactions by binding to unique sequences located in target proteins. The cognate sequences in the ligands are represented by short, conserved amino acid cores containing, for example, phospho-tyrosines (p- Y) or polyprolines. The formation of these complexes is the foundation of the intracellular network of signaling which Figure 2 Formalization of terms to describe protein-protein governs cell proliferation, dierentiation, cytoskeleton organiza- interaction mediated by modules using the SH2 domain and its tion, endocytosis, controlled protein degradation, and apoptosis ligand as an example by regulating various cellular events including enzyme For SH2 domains that belong to the same group, the e activity, substrate recruitment and protein localization determinant may reside in the EF1 position (the ®rst (Pawson, 1995). SH2 domains bind ligands containing amino acid in the loop between E [®fth] and F [sixth] b- phosphotyrosine residues within a speci®c sequence strands). A single amino acid change in the EF1 position (Schlessinger, 1994; Pawson, 1995). High anity of the Src SH2 domain from threonine to tryptophan binding is provided by the phosphotyrosine residue was shown to switch the ligand-binding speci®city to itself, and by residues carboxyterminal to it, p-Yxxc resemble that of the SH2 domain of Grb2 (Growth (Songyang and Cantley, 1995; Figure 2). Factor Receptor-Binding Protein 2) adaptor protein. Examination of several crystal and NMR (Nuclear The binding speci®city of the mutated domain correlated Magnetic Resonance spectroscopy) structures of SH2 well with the biological activity as the mutated Src SH2 domain-ligand complexes, as well as extensive analysis domain substituted the SH2 domain of the Grb2 protein of binding preferences of a variety of SH2 domains in activation of the Ras pathway in vivo (Marengere et using molecular repertoires of phosphopeptides, helped al., 1994). These two examples support the argument to classify the SH2 domain into four groups (Songyang that there is a relatively simple relationship between the e and Cantley, 1995). The e determinant for SH2 domains determinants and binding speci®city. is represented by the bD5 position (the notation The SH2 rules are being re®ned as more structural indicates ®fth amino acid in the fourth b-strand; D and functional data become available, but even in their indicates the fourth strand). The majority of known present form and with several well-documented SH2 domains fall into group I, which prefer hydrophilic exceptions, they provide a valuable tool to predict residues in the +1 position of the ligand. The e cognate pairs for SH2 domain-ligand complexes. determinant of the group I domains, bD5 position, is represented by aromatic residues, tyrosine or phenyla- lanine. Group III SH2 domains prefer hydrophobic PTB rules amino acids in the +1 position of the ligand and isoleucine or cysteine in the e determinant of the In general, PTB (Phosphotyrosine-binding) domains play domain. The e determinant for group II SH2 is a similar role to SH2 domains (Kavanaugh and Williams, threonine, whereas for group IV, it is methionine or 1994; Blaikie et al., 1994; Gustafson et al., 1995). Based on valine. ligand recognition, the PTB domains fall into two groups: The importance of the bD5 residue in determining Group I represents those PTBs which bind to ligands speci®cities of SH2 domains was documented experi- containing NP6p-Y (N-asparagine, P-proline) cores mentally by replacing the aliphatic residue of the group where tyrosine is phosphorylated (Van der Geer and III SH2 domain (phosphoinositide 3-kinase SH2 Pawson, 1995). The PTB domains of Shc (Src Homo- domain) with tyrosine of the group I SH2. The logous and Collagen) and IRS-1 (Insulin Receptor resulting mutant acquired the speci®city of the group Substrate-1) belong to this class and bind ligands with I SH2 domain in terms of selectivity for phosphopep- c6NP6p-Yc and ccc66NP6p-Y consensus se- tide ligands (Songyang et al., 1995) quences, respectively (Van der Geer et al., 1995; He et Protein-protein interaction by modules MSudolet al 1471 al., 1995; Isako et al., 1996). The group II PTBs bind to the fC sequences include prolines, or the positively ligands with NP6Ycores where tyrosine is not necessarily charged amino acids lysine and/or arginine (Chen et phosphorylated (Zambrano et al., 1997).