Probing the Phosphopeptide Specificities of ... - Semantic Scholar

Current Protein and Peptide Science, 2002, 3, 365-397

365

Probing the Phosphopeptide Specificities of Protein Tyrosine Phosphatases, SH2 and PTB Domains with Combinatorial Library Methods Stefan W. Vetter1 and Zhong-Yin Zhang2,* 1

Department of Molecular Biology, The Scripps Research Institute, 10550 North Torrey Pines Road, La Jolla, CA 92037, USA and 2Department of Molecular Pharmacology, Albert Einstein College of Medicine, 1300 Morris Park Avenue, Bronx, NY 10461, USA Abstract: Protein tyrosine phosphatases, SH2 and PTB domains are crucial elements for cellular signal transduction and regulation. Much effort has been directed towards elucidating their specificity in the past decade using a variety of approaches. Combinatorial library methods have contributed significantly to the understanding of substrate and ligand specificity of phosphoprotein recognizing domains. This review gives a brief overview of the structural characteristics of protein tyrosine phosphatases, SH2 and PTB domains and their binding to phosphopeptides. The chemical synthesis of peptides containing phosphotyrosine or phosphotyrosine mimics and the various formats of synthesis and deconvolution of combinatorial libraries are explained in detail. Examples are given as how different combinatorial libraries have been used to study the interaction of phosphopeptides with SH2 domains and phosphatases. The intrinsic advantages and difficulties of library synthesis, screening and deconvolution are pointed out. Finally, some experimental results on the substrate specificity of protein tyrosine phosphatase 1B and the SH2 domain of the adaptor protein Grb-2 are summarized and discussed.

INTRODUCTION Proteins can be subject to many forms of posttranslational modifications and events such as proteolytic trimming, glycosylation, lipidation or phosphorylation can alter the cellular function of proteins dramatically [1-3]. The posttranslational phosphorylation of proteins has attracted a significant amount of research activity from various perspectives due to its particular role in eukaryotic cell signaling and regulation [4-9]. Tyrosine phosphorylation in particular has been recognized to be crucially involved in the transduction of extracellular signals and intracellular regulation in eukaryotic organisms by biochemists, geneticists, and physicians [8-12]. Abnormal levels of protein tyrosine phosphorylation are associated with severe health defects such as diabetes, several forms of cancer, neuro-degenerative conditions, autoimmunity and infectious diseases [13-22]. The clinical relevance of protein phosphorylation suggests protein kinases, phosphatases and phosphoprotein binding domains as potential drug targets. Protein phosphorylation in general is not a rare event, almost 30% of all cellular proteins are temporarily phosphorylated and almost any protein can be phosphorylated in vitro by a range of protein kinases. The hydroxyl groups of the amino acid side chains of serine, threonine and tyrosine

*Address correspondence to these authors at the 1Department of Molecular Biology, The Scripps Research Institute, 10550 North Torrey Pines Road, La Jolla, CA 92037, USA; Tel: (858) 784 2425; Fax (858) 784 2857; Email: [email protected] and 2Department of Molecular Pharmacology, Albert Einstein College of Medicine, 1300 Morris Park Avenue, Bronx, NY 10461, USA; Tel: (718) 430 4288; Fax (718) 430 8922; E-mail: [email protected]

1389-2037/02 $35.00+.00

are commonly identified with protein phosphorylation. The phosphorylation of other side chains, e.g. histidine and the formation of mixed acid anhydrides with aspartic and glutamic acid residues does occur, but has not been analyzed in much detail. The vast majority of phosphoproteins are phosphorylated on serine or threonine residues. Less then 0.1% of all phosphoproteins are tyrosine phosphorylated [2325]. However, tyrosine phosphorylation is particularly important as it has been recognized to play a central role for the regulation of cellular regulation and signal transduction in multicellular organisms. The yeast genome for instance encodes only for three PTPs and no bonafide PTKs are found [26, 27]. Phosphorylation based signaling is virtually absent in prokaryotes [28]. The phosphorylation of a protein alters its biophysical properties and can lead to changes in its biochemical properties [29, 30]. For instance, the addition of a phosphate group onto the hydroxyl group of a tyrosine not only increases the volume of side chain, but also introduces two negative charges. This can lead to conformational changes within the protein [29, 31, 32]. Those conformational changes in turn can result in the activation or inhibition of the biological or enzymatic activity of the protein. An alternative mechanism by which protein phosphorylation is utilized in cellular signaling is related to changes in the biophysical properties of a protein. Changes of surface properties (e.g. electrostatic potential) can effect and modulate protein-protein interactions [33]. However, it also should be mentioned that there are numerous cases of silent phosphorylations. Such phosphorylation events have no obvious effect on the protein's biochemical behavior, stability or structure and have been described in particular for the phosphorylation of serine/threonine residues [34].

© 2002 Bentham Science Publishers Ltd.

366

Current Protein and Peptide Science, 2002, Vol. 3, No. 4

The phosphorylation of tyrosine side chains, as well as the dephosphorylation of phosphotyrosine residues requires the activity of specific enzymes (the same is true for serine/threonine residues). Protein tyrosine kinases (PTKs) use ATP as phosphate donor and are responsible for the phosphorylation of specific tyrosine side chains. Their activity is counteracted by protein tyrosine phosphatases (PTPs), which catalytically hydrolyze phosphotyrosine residues to tyrosine and inorganic phosphate. The variety of PTKs and PTPs is impressive. Current estimates indicate that the human genes encode for over one hundred PTKs and more than a hundred genes encoding PTPs have been identified [8, 9, 35]. It is somewhat surprising that such a large number of PTKs and PTPs are apparently necessary for eukaryotic cells to control their state of protein phosphorylation and raises subsequently the question about the specific function of each kinase and phosphatase. Another element of protein phosphorylation mediated cell signaling has been found in a number of protein modules with specific phosphoprotein binding properties [36-39]. Among those domains are at least two, which bind specifically to phosphotyrosine residues. The first family of pTyr binding domains was initially discovered in kinases related to the proto-oncogene product Src and is hence known as SH2 (Src-homology) domains [40, 41]. A second family of phosphotyrosine binding modules is structurally distinct from SH2 domains and was initially termed phosphotyrosine interacting (PI) domain, but is now commonly referred to as PTB (phosphotyrosine binding) domain [42, 43]. The possible number of interactions between the key elements of tyrosine phosphorylation controlled signal transduction, namely: protein tyrosine kinases with cellular proteins, protein tyrosine phosphatases with phosphoproteins and PTB/SH2 domain containing proteins with phosphoproteins is very large [44, 45]. In addition, the activation of, for instance, a receptor kinase by a specific ligand leads often to an avalanche of subsequent signaling events, including phosphorylation, dephosphorylation, and secondary messenger release or gene activation. The complexity of signaling networks and the transient character of its many complexes between phosphoproteins and PTPs, PTKs, SH2 and PTB domains render it difficult to identify physiologically relevant interactions. The in vitro characterization of PTPs and phosphotyrosine binding domains appears to be a straightforward experimental approach to characterize the substrate/ligand specificity of these proteins. However, it is greatly hampered by the need for specifically and homogenously phosphorylated substrates, which are notoriously difficult to isolate from biological samples or to prepare by in vitro phosphorylation [46-49]. The ability to chemically synthesize phosphorylated peptides and the development of combinatorial library techniques has clearly promoted the systematic characterization of PTPs, SH2 and PTB domains in the last decade [50-54]. A significant portion of our current understanding on the specificity of phosphotyrosine peptide binding to SH2 and PTB domains comes from studies using phosphopeptide libraries. The same can be said for the characterization of protein tyrosine phosphatases: combinatorial

Vetter and Zhang

library methods allow detailed insights in the substrate selectivity of these enzymes. PTPs, SH2 and PTB domains have been studied over the past years with a range of combinatorial chemistry methods. Phosphopeptide libraries have been synthesized in many different fashions: as one bead - one compound libraries, as mixture based libraries or phage display libraries [55-57]. The methods used for the screening of these phosphopeptide libraries included screening on a solid support or in solution, affinity selection methods and the kinetic analysis of single substrates. The detection and deconvolution methods included direct sequencing of individual peptides, pool sequencing, encoded libraries, positional scanning or mass spectroscopy methods. The first part of this review describes briefly the characteristics of PTPs, SH2 and PTB domains. The structural characteristics are summarized and the modes of phosphopeptide binding are briefly explained based on crystal structure data from protein-phosphopeptide complexes. The second part focuses on the chemical synthesis of phosphopeptides, phosphotyrosine mimics, and various methods to generate combinatorial phosphopeptide libraries and how to deconvolute them. The third and forth part give examples on how different combinatorial phosphopeptide libraries have been used to study the interaction of phosphopeptide with SH2 domains and tyrosine phosphatases. The intrinsic advantages and limitations of each method are indicated. Finally, the experimental results from a number of studies on the substrate specificity of protein tyrosine phosphatase 1B (PTP1B) and the SH2 domain of Grb-2 are summarized and compared. PART I. STRUCTURE, FUNCTION AND LIGAND RECOGNITION OF PTPs, SH2 AND PTB DOMAINS 1.1. Protein Tyrosine Phosphatases: PTPs The first PTP was isolated and characterized in 1988 and since then more then 100 PTPs have been identified from cDNA libraries [58, 59]. Based on genomic analysis, it has been suggested in 1995 that the protein phosphatase family may contain as many as 1000 members [60, 61]. The now available sequence data from the human genome project (and many other organisms) will help to obtain a detailed picture on phosphatases and whether there are indeed 1000 phosphatases and how many of them are tyrosine specific [8]. Whatever the exact numbers might be, it is clear today that protein phosphatases comprise a large family of enzymes, which is equally complex and diverse as the protein kinase family - but less well studied. The protein phosphatase family has been subdivided based on their substrate preference into serine/threonine phosphatases and protein tyrosine phosphatases. In addition to these two subfamilies two other protein phosphatase families have to be considered. Dual specific phosphatases, which dephosphorylate phosphotyrosine as well as phosphoserine/threonine residues, and low molecular weight phosphatases that have an untypical low molecular weight below 20 kD [11, 62-65]. Sequence comparison of the catalytic domains of Ser/Thr phosphatases and PTPs has

Probing the Phosphopeptide Specificities of Protein Tyrosine Phosphatases

shown no sequence similarity [66, 67]. While both enzyme subfamilies catalyze the same chemical reaction, they have evolved different reaction mechanisms to hydrolyze the phosphomonoester. Serine/threonine phosphatases are metalloenzymes where two metal ions facilitate the direct attack of the ester bond by an activated water molecule in a single step [68]. In contrast, PTPs are metal free and use a cysteine residue as primary nucleophile to attack the phosphorous atom of the substrate [69]. The covalent phosphocysteine intermediate is subsequently hydrolyzed by water in a second reaction step [70]. Table 1.


367

The catalytic domain of PTPs is typically 230-250 amino acids in length and contains the PTP signature motif (I/V)HCXAGXGR(S/T) or more general CX5R (Table 1) [71, 72]. The cysteine and arginine residues are essential for catalysis [73]. The three dimensional structures of only a few PTPs are known, but all share a similar three-dimensional architecture [74]. A twisted β-sheet of four parallel strands flanked by five antiparallel β-strands forms the center of the domain (Fig. 1). Six α-helices surround the sheet, four on one side and two on the opposite side. The signature motif is located in the center of the domain at the bottom of the

Key Features of Protein Tyrosine Phosphatases (PTP), Src Homology 2 (SH2) and Protein Tyrosine Binding Domains (PTB)

PTP

SH2

PTB

domain length approx.

250

100

150

domain fold

Twisted β-sheet, surrounded by α-helices αββαβββββαβααα

β-sheet, sandwiched by α-helices αβββββαβ

β-sandwich, capped by α-helix βββββββα

sequence homology

moderate, signature motif CX5R conserved

high

low

activity

dephosphorylation of tyrosine phosphorylated proteins

binding to tyrosine phosphorylated proteins

binding to tyrosine phosphorylated protein and non-phosphorylated peptides

Fig. (1). PTP complexed to phosphopeptide. Ribbon diagram of the phosphatase domain of PTP1B in complex with the EGFR hexapeptide DADEpYL. The figure is based on the PDB file 1PTU.pdb, only residues 62-280 are shown.

368


substrate binding site. Four surrounding loops complete the catalytic active site and provide the residues necessary for the recognition of the phosphotyrosine residue and catalysis. The phosphate group of the bound substrate points into the core of the protein and is some 9 Å away from the protein surface. The depth of the substrate binding pocket prevents the shorter phosphoserine and phosphothreonine residues to reach the catalytic residues. Phosphopeptides are bound in an overall extended conformation with intensive contacts involving the phosphotyrosine residue itself and the peptide backbone. Residues adjacent to the phosphotyrosine resides seem to be determinants for substrate specificity [67, 69, 72, 74, 75]. The 230-250 residue catalytic domains of PTPs occur usually within the context of larger proteins. Some PTPs are transmembrane proteins and are usually referred to as receptor-like rPTPs. Some rPTPs contain two adjacent phosphatase domains, but often only one of these domains is catalytically active. Examples of receptor-like PTPs are protein tyrosine phosphatase alpha and beta (PTPα and PTPβ) and CD45 [76-78]. Their domain structures are sketched out in (Fig. 2). Both contain repeats of fibronectin type III domains in their extracellular portion, followed by a single transmembrane region. The 270 C-terminal amino acids of PTPβ form the phosphatase domain. CD45 contains two phosphatase domains in its cytoplasmic portion. However, only the first domain seems to be catalytically active, since the second domain lacks residues critical for catalysis. The size distribution of extracellular and cytosolic portion can vary significantly for different rPTPs and tissue specific isoforms result from alternative splicing. Examples for soluble PTPs include the well characterized phosphatases PTP1B and SHP-2. PTP1B is not only the first PTP to be cloned and characterized, but also the most intensively studied and best understood phosphatase. The human isoform consists of 435 residues, the catalytic domain is located between residues 40-288. The last 35 residues of the 150 residue C-terminal tail localize the protein to the endoplasmatic reticulum (Fig. 2). SHP-2 is a slightly larger protein of 548 residues (human) with its catalytic domain at its C-terminal half (276-517). The N-terminal portion contains two phosphotyrosine binding SH2 domains (residues 6-102 and 112-216) (Fig. 2). The frequent embedment of the catalytic phosphatase domains into larger multi-domain proteins makes it clear, that the biochemical and physiological functions of protein tyrosine phosphatases in vivo are greatly co-determined by those accompanying domains. 1.2. Phosphotyrosine Recognition Domains: SH2 and PTB In recent years, a number of protein domains with specific protein binding activity have been identified [7678]. The first domain with specific binding affinity for phosphotyrosine containing proteins was found in cytosolic kinases related to the src oncogene [79]. Subsequently, this type of domain was named src homology 2 domain or SH2 domain. Another domain termed SH3 (src homology domain 3) does not bind phosphopeptides, but rather proline rich

Vetter and Zhang

peptides adopting a left handed polyproline type II helical conformation [80, 81]. SH2 domains are about 100 amino acids in length (Table 1) [81]. They show a high degree of sequence homology, which can help to predicted SH2 domains based on sequence data. The structures of several SH2 domains and SH2phosphopeptide complexes have been determined by nuclear magnetic resonance (NMR) or X-ray diffraction experiments. All SH2 domains show a very similar threedimensional architecture, consisting of a central β−sheet that is sandwiched by a pair of α-helices (Fig. 3). Peptide ligands are bound in an extended conformation, laying roughly perpendicular to the β-sheet plane and the helix axis. The phosphate group of the phosphotyrosine peptide interacts with basic residues located in the loop region connecting β-1 and β-2, anchoring the peptide to the protein. Ligand specificity is mainly mediated by residues C-terminal from the phosphotyrosine residue. Binding affinities of several phosphopeptide-SH2 domain pairs have been measured and are typically between 10-4 to 10-7 M. Peptide binding is strongly phosphorylation dependent and affinities decrease several orders of magnitude upon dephosphorylation. During characterization of the SH2 domain containing protein Shc a second type of phosphotyrosine peptide binding module was discovered [42]. This type of domain was initially named phosphotyrosine interacting (PI) domain or phosphotyrosine binding (PTB) domain [43]. The term PTB domain is now commonly used. PTB domains are about 150 amino acids in length (Table 1). Two PTB domain subfamilies have been suggested based on sequence homology. Those with sequence homology to the PTB domain of Shc and those with homology to the PTB domains of the insulin receptor substrate (IRS-1) [82]. Both subtypes do not share a high degree of sequence homology to each other. So far four PTB domains have been structurally characterized and they show a fold reminiscent to the architecture of PH (pleckstrin homology) domains [83-85]. Two nearly orthogonal antiparallel β-sheets form a βsandwich and a C-terminal α-helix caps the barrel-like structure on one site. The two β-sheets consist of four (β1β4) and three (β5-β7) strands respectively (Fig. 4). Peptide ligands bind along the edge of β−5 strand, interacting with the almost parallel arranged α-helix. The peptide portion C-terminal to the phosphotyrosine binds in an extended conformation and the N-terminal end of the peptide adopts a turn conformation. The ligand specificity of PTB domains seems to be mainly mediated by residues Nterminal of the phosphotyrosine residue [86-90]. In contrast to the specificity of SH2 domains, the ligand recognition of PTB domains seems not to be strictly phosphorylation dependent. Several cases are reported where nonphosphorylated peptides bind to PTB domains with affinities comparable to phosphopeptides. SH2, PTB and PTP-catalytic domains have all characteristics of independent protein modules in the sense, that they can be integrated into or fused onto larger proteins while retaining their typical fold and binding properties. Usually, SH2, PTB and PTP domains are associated with other protein domains, forming larger, multi domain proteins


(Figure 2). SH2 and PTB domains can be part of proteins with enzymatic activity, such as kinases and phosphatases, or they can be linked together with other protein binding domains, such as PDZ domains, WW domains, SH3 domains and PH domains [36, 38, 86, 91]. Proteins consisting of several binding domains without any catalytic activity are referred to as adaptor proteins. Adaptor proteins can function


369

like "molecular glue", linking proteins together, which would otherwise have no particular affinity for each other. For example, the adaptor protein Grb-2 contains two SH3 and one SH2 domain. The three domains are connected by two flexible linkers, which allow movement of the domains relative to each other [79]. The flexibility of the SH2/SH3

Fig. (2). Domain organization of PTPs and adaptor proteins. Schematic presentation of the modular organization of selected signal transduction proteins. Most proteins consist of several functional domains such as SH2, PTB, PTP, FN etc. The upper bar in the figure allows to estimate the size of the presented proteins. The abbreviations TM and ER stand for transmembrane region and endoplasmatic reticulum targeting region.

370


Vetter and Zhang

Fig. (3). SH2 domain complexed to phosphopeptide. Ribbon diagram of the SH2 domain of P56 Lck kinase in complex with the 11 residue peptide EPQpYEEIPYL. The figure is based on the PDB file 11LCJ.pdb.

binding domains relative to each other may be a prerequisite for Grb-2 to form protein complexes with various protein important for cellular signaling. The specificity of the Grb-2 SH2 domain is discussed in detail in part IV. The domain structure of several adaptor proteins mentioned in this review is shown in (Fig. 2). PART II. SYNTHESIS AND SCREENING OF COMBINATORIAL PHOSPHOPEPTIDE LIBRARIES Combinatorial libraries are collections of chemical compounds synthesized in a combinatorial manner. All compounds are different, but related to each other by a common history of chemical synthesis steps. Therefore, all

compounds can be seen as derivatives from each other. Combinatorial libraries have become increasingly popular and useful tools in many areas of research ranging from molecular biology and chemistry to material sciences and semi-conductor research over the last 10 years [92, 93]. Combinatorial chemical libraries share many aspects with real-life collections such as museums, book libraries and data banks when it comes to the point where a certain piece of information shall be extracted from the library. Obviously, the size and complexity of a library is important and larger libraries contain a more complete set of information than smaller ones. However, at the same time as the size and diversity of the library increases, it becomes more and more difficult to retrieve the information one is



371

Fig. (4). PTB domain complexed to phosphopeptide. Ribbon diagram of the PTB domain of the insulin receptor substrate 1 (Irs1) in complex with the 11 residue IL-4 receptor peptide LVIAGNPApYRS. The figure is based on the PDB file 1IRS.pdb.

looking for. Too many choices make it difficult to find the perfect answer right away (or at all). The sole presence of a certain piece of information within the library or database does not ensure by itself, that a user will be able to find it. To get a satisfying answer requires an as precisely as possible defined search string and an as well as possible organized and searchable library. Before starting to synthesize or to purchase a molecular library one has to consider a number of key questions: What type of library does one need to find the information one is looking for? How big/complex shall the library be? How does one actually search the library in order to extract the desired information? How can one identify the chemical structure of a member of the library if only minute quantities

are available? Can one expect to find a single best hit in the library, or should one expect to get a large number of positive hits? What conclusion can one draw if structurally different hits seem to be equally good answers to the original question? In the context of protein phosphatases and phosphopeptide recognizing protein domains the most frequently asked and most relevant questions are: what is the in-vivo ligand for a given SH2 or PTB domain? What is the in-vivo substrate for a certain phosphatase? These questions are of enormous importance, but also extremely difficult to answer, because they ask for the role of the specific protein within the complex environment of a living cell. However, another set of simpler questions is also of great interest, and much

372


more feasible to be answered by the in vitro screening of combinatorial libraries. For example: What synthetic molecule is the most specific and potent inhibitor/ligand for a certain phosphatase, SH2 or PTB domain? What phosphopeptide is the best substrate for a phosphatase? How selective is a SH2 or PTB domain for short peptides? What type of modifications of a peptide ligand is possible to increase binding affinity? Which residues define substrate/ligand specificity in PTPs and SH2 or PTB domains? The following section gives a short overview over methods for the synthesis and deconvolution of combinatorial libraries. Since this review focuses on the recognition of phosphopeptides by PTPs and SH2 or PTB domains we concentrate on strategies relevant for the synthesis of libraries containing tyrosine phosphorylated residues and phosphotyrosine derivatives. 2.1. Combinatorial Synthesis of Peptides As mentioned above, combinatorial libraries are collections of chemical compounds with a common synthetic history. All compounds are derivatives from each other, sharing a great deal of chemical structure and differing only in a small structural detail from compound to compound. Combinatorial libraries are most often synthesized on a solid support. Solid phase synthesis (SPS) means the compound to be synthesized is covalently linked to a macroscopic particle, most often a porous polymer bead (called resin) with a typical size between 100 and 400 mesh (150 – 40µm). The compound remains attached to the bead during the entire synthesis. Reagents are dissolved in a suitable solvent and added to the resin. Most resins swell in solvents and diffusion of reagents into the beads is the reaction ratelimiting step. The major advantage of solid phase synthesis is the ease of removing reagents and by-products compared to classical solution phase synthesis. A simple filtration step separates the solid resin from reagents, by-products and solvents. SPS therefore allows the use of large excesses of reagents, which drive the reaction to completion, and reduces reaction time. In contrast, solution phase chemistry usually requires timeconsuming procedures such as crystallization, evaporation or chromatography after each reaction step. It should also be mentioned, that combinatorial libraries can be synthesized using homogenous solution phase conditions [94, 95]. Peptides have historically a prominent position in the field of combinatorial chemistry, mostly because peptide libraries are easy to synthesize and to deconvolute. During the last 30-40 years the solid phase synthesis of peptides has been developed with great success and is the standard methodology to synthesize peptides [96]. In contrast, the majority of classical chemical transformations have been developed for homogenous solution phase conditions. The most popular strategy to create peptide libraries is the one-bead one-compound approach, also known as splitand-mix method [97, 98]. The solid phase resin is split up into portions and a chemical reaction is performed. The portions are then recombined, mixed and again distributed

Vetter and Zhang

into portions for a consecutive chemical reaction. The major advantage of this method is that each individual bead carries one defined compound. However, since the beads are mixed between chemical transformations, the identity of the compound on a particular bead is unknown. The size of onebead/one-compound libraries is limited by the number of resin beads (1g of a typical resin for SPPS contains about 1 million beads). The principle of the split and mix methods is outlined in (Fig. 5). Many of the peptide libraries used to assay SH2 domains or PTPs and discussed in part III were synthesized using the one-bead one-compound method. An alternative approach uses mixtures of reagents in a single reaction step and generates a mixture of compounds on a single bead. The advantage of this method is the ability to generate very large libraries. However, the complexity of these mixtures results in quality control issues, in particular, concerns about the relative overrepresentation of "easy to synthesize" compounds [99]. The principle of the mixture coupling method is outlined in (Fig. 6). Deconvolution denotes the process of identifying the chemical structure of a compound within the large pool of compounds of a combinatorial library. The sequence of a peptide bound to the bead from a one-bead one-compound library can be determined by direct sequencing of the peptide by Edman degradation or related methods. When direct sequencing is not possible, then a variety of alternatives exists. One possibility is to synthesize a second compound on the same bead in parallel and by doing so to encode the structure of the first compound. Several methods for reading the encoding signal have been developed. This can be done for instance by PCR sequencing when the encoding tag is an oligonucleotide, isotopic labeling and deconvolution by mass spectrometry or fluorescence dye labeling [100-103]. Libraries generated by the coupling of reagent mixtures (e.g. all 20 amino acids) require other means of deconvolution and a positional scanning strategy is frequently used. This strategy splits the original library after a certain number of coupling steps and creates a set of sublibraries. For each sub-library, one position of the peptide is defined to one single amino acid, while all other positions are randomized. Subjecting each sub-library to an activity assay establish a correlation between the known amino acid in the defined position and a measurable activity of the mixture of compounds. Defining each position in a peptide allows one to deduce a structure-function profile and to predict active compounds. The most straight forward method to know the identity of a compound is to directly keep track of every chemical transformation performed during the synthesis process. This can be done by using macroscopic reaction beads for solid phase synthesis and attaching an readable label to each bead. A type of label readable by an automatic system has been developed in the form of radio-frequency tags, which are small microchips encapsulated into an inert material and attached to the solid support. Every chemical reaction step can be recorded on the microchip attached to the bead and can later be read from the chip by non-invasive radio frequency transmission. In addition, the data can be directly fed into computer managed synthesis and assay robots



373

Fig. (5). Synthesis of combinatorial libraries by "split and mix'. 'Split and mix' methods creates combinatorial libraries where each bead contains one peptide sequence (one bead - one compound). The number of compounds increases exponentially with each additional coupling step. The sequence of the peptide on a particular bead is not known except for the last amino acid, prior to the mixing step. "One bead-one compound" libraries can be assayed either on the solid phase with the peptide still attached to the bead, or in solution after release of the peptide. The abbreviations A, C, D…symbolize Ala, Cys, Asp….

374


Vetter and Zhang

Fig. (6). Synthesis of combinatorial libraries by mixture coupling. "The mixture coupling" method can create very large libraries in the form of complex mixtures of compounds. Each beads carries many different peptide sequences at the same time. Sublibraries have to be kept separated after a position has been defined (O) by coupling only one compound. Libraries generated by this method are usually assayed in solution after cleavage of the compounds.

[104, 105]. The low cost version of direct tracking method uses a color encoding system. For each reaction step a colored (or numbered) marker is manually attached to the macroscopic solid phase support. The sequence of the markers corresponds directly to e.g. the amino acid sequence of the peptide. Direct tagging methods are particular useful for small libraries of only a few dozens to several thousand compounds. 2.2. Alanine and Inverse Alanine Scanning Another positional scanning method does not relay on randomizing non-defined positions but rather on the incorporation of a functionally neutral amino acid – alanine -

into defined positions. Individual residues of a peptide known to bind to a SH2/PTB domain or to be a PTP substrate are replaced one by one with a "neutral" alanine residue to assess the relative contribution of the replaced residue towards binding of the entire peptide. This method requires a known peptide ligand to start with and does not reveal “novel” peptide sequences. An inversion of the “classical” alanine scanning is based on the assumption that the contribution of each amino acid towards the total binding of a peptide can be analyzed individually and without regard to the adjacent peptide residues. The “inverse alanine” method uses synthetic allalanine peptides and systematically defines each position with one of the 20 (or more) amino acids. This methods


creates a set of 20 (or more) peptides for each position and requires the same number of assays as the positional scanning with randomized positions (see 2.2). A major advantage is that each assay is done with only one peptide of defined purity, rather than with complex mixtures. A potential limitation is that cooperativity effects between amino acid side chains of the peptide can not be observed when all residues have been substituted with alanine. However, the method has been used to characterize the substrate specificity of PTP1B and is discussed in part III [106]. 2.3. Phage Display Libraries Phage display libraries differ from the above described libraries by the fact that they are not chemically synthesized, but generated by living organisms. The method uses bacterial viruses (phages, bacteriophages) to display the protein of interest on the surface of the phage and at the same time to package the gene encoding for the displayed protein into the phage particle. This effectively links the displayed protein and its encoding DNA in a single small particle. Compared to chemical libraries: the phage particle corresponds to the resin bead, the displayed (fusion)protein to the chemically synthesized peptide, the DNA inside the phage particle corresponds to a readable tag and the bacterial cell, instead of a chemist, does the actual synthesis. Phage particles are released from the cell into the medium and can be easily purified under mild conditions. The purified phage are then used in a in vitro screening protocol based on, for instance, binding affinity to a target. Phage particles displaying proteins with binding affinity are then used to infect the bacterial host, thus allowing to in vivo amplify a set of proteins (phages) with particular binding properties. Phage display libraries have been particularly successful for the display of antibody libraries and a variety of different phage display formats have been developed [56, 107-110] From a technical point of view phage display libraries offer the advantage to create libraries of large peptides or proteins, whereas chemically synthesized peptide libraries rarely exceed 10-20 residues. Phage libraries can also be larger then one-bead one compound libraries, because the phage particle is much smaller than a single resin bead. Effectively, the size of phage libraries is limited by efficiency of the initial introduction of the genetically engineered DNA-plasmid library into the bacterial host to typically 10 8 members. A significant disadvantage of phage libraries for the display of phosphopeptide libraries is the inability to produce phosphorylated protein in bacterial cells, or to incorporate non-natural amino acids. However, it is possible to phosphorylate isolated phages and their displayed peptides in vitro. There are severe limitations regarding the completeness of this type of phosphopeptide libraries to consider. An example of a phage displayed phosphopeptide library is discussed in part III. 2.4. Synthesis of Phosphopeptides There are two conceptually different strategies to obtain phosphorylated peptides: enzymatic phosphorylation using


375

protein kinases and chemical phosphorylation of synthetic peptides. The enzymatic phosphorylation of peptides has been widely used in the past due to a lack of alternatives. However, there are serious intrinsic problems. Kinases are substrate specific enzymes and it is generally not possible to phosphorylated every tyrosine residue within every given peptide sequence. Another difficulty is the possibility of uncontrolled multiple phosphorylations within a peptide with several potential phosphorylation sites. It will often be impossible to achieve the selective phosphorylation of one defined tyrosine residue [111-113]. On the other side, enzymatic phosphorylation can be very helpful (and often is the only way) to phosphorylate proteins, which are too large to be chemically synthesized. Enzymatic phosphorylation of synthetic peptide libraries is generally not used when a synthetic alternative exists. The chemical synthesis of tyrosine, serine and threonine phosphorylated peptides has become routine in recent years and is conveniently possible using either Boc- or Fmocsolid phase peptide synthesis (SPPS). A difficulty to obtain relatively pure phosphorylated peptides without chromatographic purification is related to the limited stability of phosphate esters under strong acidic or basic conditions required during peptide synthesis. Free and protected phosphate groups of phosphorylated amino acids tend to partially decompose via β-elimination (serine/threonine residues) upon long exposure to strong acids or bases needed for Boc or Fmoc SPPS [114-119]. The synthesis of phosphoserine and phosphothreonine peptides is more problematic than phosphotyrosine peptides. The possible loss of the phosphate group has been taken into consideration when libraries are synthesized. From the synthetic point of view one can discriminate two strategies to synthesize phosphopeptides: the global phosphorylation approach and the building block approach. The global phosphorylation approach can be compared with a posttranslational modification. The peptides are synthesized using standard Boc- or Fmoc-SPPS chemistry and are then phosphorylated on the solid support after complete chain assembly [50, 120-122]. The amino acids that shall be phosphorylated are incorporated either as side chain unprotected derivatives or as selectively deprotectable amino acids. The free hydroxyl group is subsequently phosphinylated with a suitable reagent and the initially resulting P(III) compound is oxidized to the more stable P(V) compound (Fig. 7). [52, 121, 123]. The global phosphorylation process has the advantage that splitting of the peptide resin prior to the phosphorylation allows one to obtain both, the phosphorylated and the nonphosphorylated peptides from a single synthesis. A major disadvantage in respect to peptide libraries, however, is the fact that the phosphorylation efficiency often varies with the peptide sequence and it is not possible to provide complete phosphorylation of all library members. In addition, the oxidation of the P(III) to the P(V) compound carries the risk of oxidative side reactions on methionine and tryptophan side chains [124]. The possibility, that the phosphopeptide libraries generated by global phosphorylation contain contaminations by non-phosphorylated and side chain

376


Vetter and Zhang

Fig. (7). Synthesis of phosphopeptides. Comparison of two possible strategies to synthesize phosphorylated peptides. The global phosphorylation approach involves additional synthetic steps, but allows to obtain phosphorylated and non-phosphorylated peptide from a single peptide synthesis. The use of pre-phosphorylated building blocks avoids the risk of incomplete phosphorylation and sidechain modification during the phosphinylation/oxidation steps and is the preferred methods for library generation.

oxidized peptides should be avoided by using prephosphorylated building blocks for library synthesis. The building block approach uses preformed phosphoamino acid derivatives to assemble phosphopeptides. The

phosphoamino acids is incorporated into the growing peptide chain like any other regular amino acid and does not require any changes to the regular SPPS protocol Fig. (7) [54, 125]. The phosphoderivative can be incorporated quantitatively into the peptide, this minimizes the chance of contaminations


of the library by non-phosphorylated peptides. Phosphoderivatives for serine, threonine and tyrosine are commercially available and make the building block approach very straight forward. The use of preformed building blocks is certainly the method of choice for the synthesis of phosphopeptide libraries, a selection of commercially available phosphotyrosine building blocks is shown in (Fig. 8, comp. 2-4). 2.5. Phosphotyrosine Mimics Phosphorylated peptides are perfectly stable compounds in the absence of phosphatases. However, a broad variety of phosphatases is capable of quickly dephosphorylating phosphopeptides. The limited life-time of phosphopeptides in-vivo makes it difficult to use phosphopeptides for the screening of cell extracts or in cell based assays. Phosphotyrosine mimics need to resemble as closely as possible the key properties of phosphotyrosine with respect to its interaction with phosphotyrosine binding domains and protein tyrosine phosphatases. At the same time they have to be resistant towards degradation by phosphatases and modifications by other enzymes. Key features of phosphotyrosine are the aromatic portion of phenol ring and the negative charges associated with the phosphate monoester group. The phosphate group forms hydrogen bonds and electrostatic interactions with the protein in a defined three-dimensional geometry and the phenyl ring is important for making aromatic and hydrophobic contacts. These interactions occur in defined distances and angles relative to the peptide backbone. All these features have to be retained by a good phosphotyrosine mimic. Obvious modifications to increase hydrolytic stability of phosphotyrosine involve the replacement of the bridging oxygen. If the bridging oxygen is eliminated at all, then the phosphorus atom is directly bonded to the carbon atom of the phenyl ring in the resulting phenyl phosphonate (Fig. 8, comp. 5). Consequently, the phenyl phosphonate compound is about 1.6 Å shorter than tyrosine phosphate and the pK values for the two acidic protons are different. To avoid the reduced distance between the aromatic ring and the phosphonate group it is possible to replace the bridging oxygen of the tyrosine phosphate by a methylene group (Fig. 8, comp. 6). In this benzyl phosphonate the distance between the phosphorous atom and the carbon atom of the phenyl ring is comparable to the distance in tyrosine phosphate [126]. To further improve the hydrogen bonding properties of the methylene group and to resemble the properties of the bridging oxygen, the methylene hydrogen atoms have been replaced by fluorine atoms (Fig. 8, comp. 7) [127, 128]. The hydrogen bonding properties of the resulting difluoromethylphosphonophenylalanine derivative are very favorable. An alternative approach replaces the non-bridging oxygen atoms with sulfur atoms (Fig. 8, comp. 8) [129]. For certain applications, the high acidity of the phosphate and phosphonate monoesters are desired. Under physiological condition the phosphate/phosphonate groups are deprotonated and the negative charge reduces the membrane permeability of the molecules and their bioavailability. One type of phosphotyrosine mimics has been developed with the intent to increase their membrane


377

permeability and to make them more suitable for cells based experiments. To achieve this, the phosphate group has been replaced with a malonic acid moiety Fig. (8, comp. 9) [130, 131]. The carboxylic groups can additionally be esterified to temporarily mask the carboxyl groups. The malonyl esters are deprotected in vivo by hydrolysis through esterases, regenerating the two acidic groups. A variation of the Omalonyltyrosine derivative has the malonyl group replaced by an acetic acid group and a second carboxyl group in ortho position (Fig. 8, comp. 10). Both compounds are structurally and chemically clearly different from phosphotyrosine, but have been successfully used for the design of SH2 ligands and PTP inhibitors [132]. PART III. SPECIFICITY DETERMINATION OF SH2 AND PTB DOMAINS Binding specificity in protein-ligand interactions can be approached from different angles and answers regarding the specificity of PTPs, SH2 and PTB domains depend on the initial question and partially on the experimental approach chosen to address the problem. Too often, selectivity is simply interpreted as high affinity binding. Tight binding of a ligand to a specific protein is often an element of selectivity, but neither strictly sufficient nor necessary for selectivity. The most selective ligand is not necessary the one with the highest binding affinity. Selectivity of a proteinligand pair has to be seen from the perspectives of both partners, the protein and the ligand, in the complex environment of a living cell. For example, a phosphatase (e.g. PTP1B) may select a certain peptide Pep1 (e.g. AspGlu-Pmp-Leu) from a library containing several thousand peptides based on tight binding. The phosphatase may now be considered selective for this peptide. However, the same peptide Pep1 may bind to a variety to other PTPs (e.g. PhosA, PhosB, PhosC) as well and can therefore not be considered specific for either protein PTP1B, PhosA, PhosB or PhosC. Chances are that the peptide library also contained another peptide Pep2 that has a lower affinity for PTP1B than Pep1, but does not bind to PhosA, PhosB or PhosC. Peptide Pep2 would therefore be specific for PTP1B. One should keep this in mind if one wishes to identify specific substrates or inhibitors. It is important to design the library, the assay and screening conditions carefully in order to obtain results relevant to the problem under investigation. The first aspect to consider is the design of the library itself. What kind of ligand is desired? Peptide or small organic molecule? Is any information about other ligands or binding partners available and can they be used as starting point in the library design? Shall the library be as much randomized as possible or how much bias is acceptable? How shall the library be synthesized? One-bead one-compound method, mixture coupling or phage display? Shall the assay be done in solution or on the solid support? Shall the screening be done with single compounds or by positional scanning? There is no single optimal procedure to investigate the specificity of PTPs, SH2 and PTB domains. We present in the following paragraph a selection of approaches that have been applied in the past to gain a better understanding of phosphopeptide recognition.

378


Vetter and Zhang

Fig. (8). Phosphotyrosine and derivatives. Structures of tyrosine and tyrosine derivatives useful for the synthesis of phos-phopeptides (compounds 2-4) or suitable for the incorporation as phosphotyrosine mimics into peptides (compounds 5-10).

3.1. Affinity Enrichment from Peptide Mixtures and Pool-Sequencing One of the first systematic studies addressing the specificity of SH2 domains was published in 1993 by

Songyang et al [40, 133]. The authors used a highly biased library of 5832 peptides and enriched a pool of peptide sequences employing an affinity chromatography method (Fig. 9). At this time more than 20 SH2 domains from cytosolic proteins were known and it was clear that the SH2



379

Fig. (9). Screening and deconvolution of soluble libraries. Schematic presentation of two strategies used to screen soluble combinatorial libraries. The complex between protein and peptide is formed either free in solution of with the protein immobilized. Free, unbound peptides are removed and the protein/peptide complexes dissociated. The peptide pools are analyzed by either mass spectrometry of peptide sequencing. The methods are not limited to SH2 binding peptides, but can be applied to other proteins, such as PTPs or enzymes in general.

domains of proteins such as Crk, PLC-γ, Src, GAP and Abl bind specifically to tyrosine phosphorylated peptides. It also was clear that the binding affinity does somehow depend on the amino acid sequence surrounding the phosphotyrosine residue. In particular the C-terminal motif pTyr-Met-XxxMet had been noticed to be recognized by some SH2 domains. The authors chose to create a library with three invariant N-terminal residues Gly-Asp-Gly-pTyr based on the observation that an acid residue is often found Nterminally to native tyrosine phosphorylation sites. The last five C-terminal residues were fixed to the sequence Ser-ProLeu-Leu-Leu, firstly to allow the formation of a bend in the Ser-Pro region and secondly to increase the peptides hydrophobicity in order to bind the peptides to a membrane for peptide sequencing.

The library had a total of three randomized positions GDGpYXXXSPLLL, which were degenerated with all possible amino acids except cysteine and tryptophan. This resulted therefore in 183 = 5832 different peptide sequences. The library was synthesized in the split and recombine format, leading to a presumably equimolar mixture of all peptides. However, amino acid analysis and sequencing of the entire library indicated some variability in molar amounts up to 2 and 3-fold, respectively. 14 different SH2 domains were expressed as GST fusion proteins and immobilized on gluthathione-agarose beads. The beads were packed into small columns, each carrying about 600 µg fusion protein and 2 mg of the peptide library (about 0.2 nmol of each peptide) were applied to the column. After two short washes the bound ligands were eluted with sodium phenylphosphate,

380


which would specifically replace the phosphotyrosine moiety of the peptide from the SH2 domain's binding pocket. The eluted peptides were concentrated, sequenced and the relative content of each amino acids in each of the three randomized positions was calculated. The various SH2 domains clearly selected different sets of peptides, indicating that the retention of peptides was indeed due to specific binding of the phosphopeptides to the SH2 domains. Domains from members of the Src family (Src, Fyn, Lck and Fgr) selected peptides with the sequence pTyr-Glu-Glu-Ile, whereas the SH2 domains of Abl, Crk and Nck preferred Pro in the +3 position and hydrophilic residues in +1 and +2 position. The SH2 domains of p85, PLC-γ and SHP2 in contrast selected the general motif pTyr-hydrophobic-no preference-hydrophobic. This early study confirmed that SH2 domains indeed display selective binding properties towards phosphotyrosine peptides and that combinatorial library techniques can be used to identify preferred amino acid compositions. However, the method did not allow the identification of individual peptide sequences. In addition, the selection for individual amino acids was not absolute and in many cases equal levels of selection for different amino acids were observed. For instance, the selection by the SH2 domain of Crk showed almost equal preference for Asp and Lys in the +1 position or for Arg and Phe in position +2. These amino acids are so different in the physicochemical properties that it is not immediately obvious how they can possibly substitute for each other. Results obtained from library methods as described above (affinity selection and pool sequencing) should always be considered as preliminary and further evaluation will be necessary. This can be done either by synthesizing a second generation library with fewer members, leaving out amino acids identified as non-favorable from the first generation library or by directly testing individual peptide sequences. 3.2. Affinity Selection on Bead Bound Libraries and Direct Sequencing from Beads Pool sequencing of peptide mixtures fails to provide information about individual peptide sequences, which are much more valuable than averaged amino acid ratios. To obtain individual peptide sequences it is necessary to have sufficient amounts of peptide to allow direct sequencing or to be able to deconvolute the peptide sequence by other means. A feasible and popular method uses libraries synthesized in the "one-bead one-compound" format, where the peptides are deprotected, but remain attached to the resin. The solid phase resin used for “on bead” assays should swell in aqueous buffer. Polyethylenglycol or polyacryl amide based resins can be used. When the beads are larger then about 100 µm in diameter, then they can be seen and handled with the unaided eye, making the isolation of individual beads very easy. The general assay strategy of bead bound phosphopeptide libraries is based on the interaction of the free, soluble phosphotyrosine binding domain with phosphotyrosine peptides immobilized on the resin beads. SH2 or PTB

Vetter and Zhang

domains will bind to beads carrying peptides that are recognized by the respective domain. Washing of the resin removes non- and weakly bound protein. However, beads carrying peptides with high affinity (slow dissociation) for the SH2- or PTB domain will retain the domains on the bead. Beads containing SH2/PTB domains on their surface can subsequently be identified by immunological detection or direct labeling (Fig. 10). The SH2/PTB domain can be labeled prior to use in the binding assay with a fluorescent dye, which allows identification of beads based on their fluorescence under a UV-light source. Biotinylation of the domain enables to detect positive beads using (strept)avidin conjugated to a reporter enzyme such as alkaline phosphatase or peroxidase. If an insoluble chromogenic enzyme substrate is used, then the beads themselves become stained and are easily visible. Immunological methods use either a primary antibody directed against the SH2/PTB domain itself, a protein fused to the domain (e.g. GST), or a peptide tag (e.g. Flag, Myc, Streptag) attached to the SH2/PTB domain. The primary antibody can be labeled by itself, or used in conjunction with a labeled secondary antibody. The assay and screening conditions have to be chosen such that only a small fraction of all beads will finally appear as positive (labeled). These beads can then be isolated, washed and stripped off the bound proteins and/or dyes. A single bead can then be transferred between the membranes of an automated peptide sequencer and sequenced like a regular membrane adsorbed peptide (Fig. 10). In order to obtain clean sequencing results it is important to ensure that only a single bead is placed between the membranes. Some beads get damaged during the synthesis and screening process. Some break into smaller fragments which can stick to other beads, resulting in several parallel peptide sequences with relative signal intensities corresponding to the volume of the bead fragments. The advantage of obtaining single peptide sequences using one-bead one-compound libraries has to be paid for by the time consuming process of sequencing a large number of single beads. However, preferences for certain amino acids in a given position become often obvious after sequencing a reasonable number of beads. The results at the single sequence level also allow one to identify details that are not apparent in averaged data from pool sequencing experiments. 3.3. Fluorescence-Activated Bead Sorting (FABS) and Pool-Sequencing A combination of the two methodologies described above has been used by Müller et al. [134]. This approach used a one-bead one-compound library on very small beads. The binding assay was performed by incubating soluble SH2 domains with the bead-bound peptide library. Beads presenting peptides recognized by the SH2 domain were then separated in a machine designed for fluorescence activated cell sorting (FACS). Pools of beads were subsequently sequenced (Fig. 10). The authors synthesized a library of 636,804 individual peptide sequences in the format



381

Fig. (10). Screening and deconvolution of bead bound libraries. The screening of bead bound peptide libraries requires the identification and isolation of bead that contain peptides with affinity for the protein under investigation. Isolated beads can then by analyzed either individually, leading to single peptide sequences information, or in pools of peptides.

EPXpYXXXXX. The randomized positions X were degenerated with a limited set of amino acids, using six amino acids in position - 1 and + 5, seven in + 2 and + 4 and 19 in position + 1 and + 3 (6x19x7x19x7x6 = 636,804 members total). The library was synthesized in the split and recombine format on 2.1x10 9 small beads (1.2x10 6 beads per mg). The beads had a size of 10 µm diameter, which is optimal for use in an automated cell sorter. At the same time the small size did permit only a peptide load of about 8 fmol per bead, which was not enough for direct peptide sequencing from single beads. The two SH2 domains from Grb-2 and Syk, were expressed as GST-fusion proteins and incubated with the peptide library. The beads were collected by centrifugation, washed and incubated with a FTIC-labeled anti-GST antibody. The beads were applied to the (FACS) sorter and settings were chosen to select for beads with highest fluorescence. After several rounds, 8000 beads were isolated

out of 6x107 (0.013%). The beads were collected and peptide sequencing was performed under standard sequencing condition on a commercial sequencer. Analysis of the pool sequencing data was based on the relative amount of each amino acid found for each sequencing cycle. The two SH2 domains of Grb-2 and Syk gave different sequencing results, demonstrating selectivity in their binding to the peptide library. The SH2 domain of Grb-2 selected peptides with the motif FpYEN(D/E)(D/P)D, while the Syk-SH2 domains selected EpYEELDD. Certain position displayed greater selectivity than others. For instance, Syk selected exclusively Leu out of 19 possible amino acid in position +3, but showed only a slight preference for Glu in position +2 where only seven amino acids had been used for randomization. For Grb-2 the situation was the opposite with a broad acceptance of substitutions in position +3 and exclusive selectivity for Asn in position +2.

382


The use of a fluorescence activated cell sorter (FACS) enables screening of very large numbers of beads in a very short time. Furthermore, it allows at the same time to analyze beads quantitatively for fluorescence intensity. Both are tremendous advantages compared to manual selection of beads based on visual inspection. A drawback however is the limitation on the size of the beads, which ultimately prohibited sequencing of single beads. 3.4. Parallel Synthesis and Screening by Affinity Competition Enzyme-Linked Immunosorbant Assay (ELISA) Peptide libraries can be easily deconvoluted by simple peptide sequencing and, in principle, most linear polyamides can be analyzed by sequential degradation. However, it can cause substantial efforts to establish suitable conditions to sequence peptides containing non-natural amino acids. In addition, branched peptides, cyclic peptides, N-terminally modified peptides and small molecule libraries can not be deconvoluted by classical peptide sequencing. In these cases it is necessary to track individual compounds through the entire synthesis process in order to know the identity of the displayed compound. A feasible way to synthesize libraries of small to moderate size and to keep track of each compound is to perform the synthesis in small, individual synthesis compartments. Commercially available multi-well plates offer one possibility to perform reactions in parallel. A typical example for the use of multi-well plates for the synthesis of a small combinatorial library in an academic laboratory setting is described by Lee and Lawrence [135, 136]. Their aim was to find novel ligands for SH2 domains and parallel synthesis in multi-well plates was used to generate 900 individual compounds. The library was based on the core peptide pYEEI, which was known to bind to the SH2 domains of Fyn and Lck. The position immediately Nterminal to the phosphotyrosine (position -1) was degenerated X-pYEEI by the coupling of 900 different commercially available carboxylic acids. The core peptide pYEEI-SS-@ was synthesized on a thiol cleavable disulfide linker (-S-S-) using standard solid phase peptide chemistry. The resin was then distributed into 96 well plates designed for solid phase chemical synthesis (solvent resistant, filter bottom for solvent removal, etc.) and each of the 900 carboxylic acids was coupled to the resin in one well, completing the synthesis (Fig. 11). The peptides were deprotected and then released from the solid support with an aqueous sample buffer, containing 10 mM DTT to reduce the disulfide linker. The individually recovered cleavage solutions contained the free, deprotected peptide and were directly used in the following SH2-domain binding assays (Fig. 11). The binding activity of each compound to the SH2 domains was assayed by affinity competition with a reference peptide of known binding affinity. This biotinylated phosphopeptide (biotin)-EPYpYEEIPIYL was immobilized on streptavidin coated ELISA plates. The library compounds and the SH2 domain were added to the immobilized reference peptide, where the soluble library

Vetter and Zhang

compound and the immobilized peptide compete for binding to the SH2 domain. If the library compound has a much weaker affinity for the SH2 domain than the immobilized reference peptide, then the SH2 domain will be immobilized to the ELISA well. If the library compound has a much stronger binding affinity for the SH2 domain than the immobilized peptide, then the soluble ligand will bind to the SH2 domain and prevent immobilization of the SH2 domain to the ELISA well. The amount of immobilized SH2 domain was detected taking advantage of the a GST-SH2 fusion protein construct. A primary anti-GST antibody and a secondary peroxidase labeled antibody were used to quantify the amount of the retained SH2 domain. The authors were able to identify a novel ligand for the Lck-SH2 domain (binding affinity 35 nM) and a ligand with an affinity of 60 nM for the Fyn-SH2 domain. The Lck ligand had 7-hydroxy-coumarin-4-acetic acid at the –1 position, and the Fyn ligand had 5-sulpho-isophthalic acid at the –1 position. In addition, the compounds showed selectivity for either Fyn over Lck and visa versa. Although the selectivity was moderate (3-4 fold) it presents a first example for selective ligands between members of the Src kinase SH2 family [135, 136]. This example demonstrates the feasibility to generate several hundred compounds with simple and inexpensive equipment. Obviously, much larger libraries with several randomized positions and tens of thousands of compounds, would require more logistic effort. 3.6. Phage Display of Phosphopeptides Phage display techniques present an alternative to chemically synthesized peptide libraries. However, the display of non-genetically encoded amino acids is (usually) not possible and the display of posttranslationally modified amino acids is difficult. The bacterial phage host does not normally support posttranslational modifications such as lipidation, glycosylation or phosphorylation in vivo. On the other hand, it is possible to modify phage displayed peptides in vitro. The applicability of phosphopeptide phage display to assess SH2 domain specificity has been demonstrated by Gram et al. [137]. The authors generated a phage library with seven randomized positions surrounding a central tyrosine residue E-L-E-X-X-X-Y-X-X-X-X-A. All 20 amino acids were incorporated into the randomized positions, resulting in a theoretical library size of 207 = 1.28 x 109 peptide sequences. The phage particles were produced in E. coli, isolated and then phosphorylated in vitro using a mixture of three protein tyrosine kinases (c-Src, Blk and Syk) [137, 138]. The phosphorylation of the displayed peptide increases the complexity of the library to 21 possible amino acids. Leading to 217 = 1.8 x 109 possible sequences under the assumption that every tyrosine can either be phosphorylated or non-modified. However, kinases have intrinsic sequence preferences and only peptides containing the kinase consensus sequence will be efficiently phosphorylated [138]. This limitation reduces the number of actually phosphorylated peptides within the library in a non-controllable manner. It also has to



383

Fig. (11). Parallel synthesis of libraries. Parallel synthesis of several hundred compounds can be done conveniently in multiwell plates. The compounds can be released into multiwell plates for storage or directly into plates suitable for activity assays. Multiwell plate reader allow quantitative spectroscopic analysis of e.g. 96 compounds simultaneously. Since the identity of every compounds A1, A2….is known, it is straight forward to establish structure-activity relationships.

384


be taken into consideration that the coat proteins of the phage themselves can also be potentially phosphorylated, which may compromise infectivity of the phage. The in-vitro phosphorylated phage particles prepared by Gram et al. were subjected to an affinity enrichment by panning against the immobilized SH2 domain of Grb-2. The SH2 domain had been expressed as GST a fusion protein, covalently immobilized on a Sepharose matrix and packed like a chromatographic column. Absorbed phages were released from the column by shifting the pH to 2.3. Eluted phages were used to infect fresh bacterial cells and to amplify the selected phage population. The cycle of phage production, in vitro phosphorylation, affinity panning, and phage amplification was repeated several times. The authors could demonstrate after four rounds of panning that individual phage clones showed a phosphorylation dependent binding to Grb-2-sepharose. Sequencing of the phage DNA revealed the sequences of the displayed peptides. A total of 24 sequences showed an apparent exclusive selectivity of the Grb-2 SH2 domain for Asn in position +2. Position +1 showed preference for either Met or Glu, position +3 was selective for aromatic amino acids, mainly Trp. Proline was frequently encountered in positions -2 and +1. Position -1 preferred hydrophobic residues Ile, Leu and Val (Table 2). The pattern of preferred substitutions surrounding the central pTyr agreed with data from synthetic peptide libraries with respect to the pivotal role of Asn in position +2, but also suggested a clear preference for Trp/Phe in position +3, which had not been revealed from previous screens of synthetic libraries (for discussion see Part V). PART IV. ASSESSMENT OF SUBSTRATE SPECIFICITY OF PROTEIN TYROSINE PHOSPHATASES The affinity selection techniques described in Part III for probing the specificity of SH2 and PTB domains can not directly be applied to PTPs, because the dephosphorylation activity of PTPs does not permit the isolation of stable PTPphosphopeptide complexes. Incubation of an active phosphatase with a phosphopeptide library leads to the dephosphorylation of substrate peptides and subsequently to the rapid dissociation of the enzyme-product complex. Theoretically, it is possible to design experiments where the active phosphatase is incubated for a short time with a phosphopeptide library and to analyze the resulting mixture of phosphorylated and dephosphorylated peptides. The dephosphorylated peptides might be isolated and the amino acid sequence of these substrates could be obtained by sequencing or other means. However, this would be a laborious process because the separation of small amounts of dephosphorylated peptide from a large pool of excess phosphopeptides is difficult. One possible approach to avoid pTyr dephosphorylation depends on the use of catalytically impaired mutant PTPs. It then becomes possible to apply all the methods described for SH2/PTB domains (Part III) for the characterization of PTP substrate specificity. Two different approaches can be used to prevent PTPs from substrate hydrolysis. Firstly it is possible to inactivate

Vetter and Zhang

the phosphatase by mutating residues within the catalytic site, for instance replacing the catalytic cysteine residue by serine, or secondly to use hydrolysis resistant phosphotyrosine mimics (pseudosubstrates) for the library synthesis. Examples for the use of phosphotyrosine mimics to characterize PTPs are given below, which also describe strategies to identify peptide ligands that can be easily applied to SH2/PTB domains as well 4.1. Inhibition of PTP Activity and Deconvolution by Positional Scanning A peptide containing a phosphotyrosine mimic functions as a pseudosubstrate, binding into the PTP catalytic site and therefore as a competitive inhibitor. This makes it possible to determine binding affinities of non-hydrolysable phosphopeptide analogs by measuring their inhibitory effect on the rate of dephoshosphorylation of a reference substrate. This method has been used to screen a set of libraries in the positional scanning format by Pellegrini et al. [139]. The authors used the EGFR peptide DADEpYLIP and scanned the three N-terminal positions. The library was synthesized using the split and recombination method to randomize positions with 20 amino acids (no cysteine, but hydroxyproline). The first library consisted of 20 sub-libraries of the format DOXXmYLIP, where O designates the defined position and each sub-library contained 400 (OXX= 1x20x20) different peptide sequences mY refers to the nonhydrolyzable phosphotyrosine mimic O-malonyl tyrosine. The sub-libraries were cleaved from the resin and used for a phosphatase inhibition assay in solution. The release of inorganic phosphate from a reference peptide was measured (malachite green assay) and the activity of PTP1B in the presence of the inhibitor peptides was determined. The scanning of position -3 identified Asp as the most effective substitution in this position (68% inhibition). However, structurally unrelated residues, such as Gly, Trp or hPro also caused considerable inhibition (58-55%) and showed better efficiency than Glu (53% inhibition). The second generation of libraries had the format DDOXmYLIP, defining position -2 and randomizing the -1 position. The screening process was repeated and the most favorable amino acids were the acidic residues Glu and Asp as well as Ser. In the last step DDEOmYLIP was used and Trp and Tyr were identified from the screen for the -1 position (Table 2). The inhibitory activity increased only slightly when positions -2 were defined from an IC50 of 8.6 µM (DDXXmYLIP) to an IC50 of 6.8 µM (DDEXmYLIP), but increased 10 fold when the position closest to the mTyr residue was defined (DDEWmYLIP) to an IC50 of 0.7 µM. The positional scanning method in combination with the split and recombine method for randomizing positions limits the direction of the scanning. Peptides are synthesized from the C-terminus to the N-terminus and the split and recombine method would make it laborious to generate a peptide library of the format DXOXmYLIP. The split and recombination method requires to split the resin into 20 fraction to introduce the defined positions O. These sublibraries (OXmYLIP) have to be kept separated and can not be mixed in order to know the identity of the defined



385

Table 2. Selectivity of Grb-2 SH2 Domain

Study

Library Format

position +1

position +2

position +3

position +4

Gram et al. [137]

E-L-E-X-X-X-pTyr-X-X-X-X-X-A

Glu (33%) Met (25%) Gln (12%) Tyr (12%) Leu, Phe, Trp

Asn (100%)

Trp (54%) Phe (25%) Ile, Leu, Tyr

Pro (37%) Ala, Ile, Leu, Met, Phe, Ser, Thr, Tyr Val.

Müller et al. [134]

E-P-X-pTyr-X-X-X-X-X

Glu (8.3) Gln (4.3) Val (2.0) ::::::::::::: Tyr (0.5)

Asn (7.0)

Asp (3.2) Glu (3.0) Val (2.3) ::::::::::::::: Tyr (1.7) Gln (1.8) Phe (1.2)

Pro (2.4) Asp (2.3)

Olingino et al. [145]

(cylco)-C-X-X-X-X-X-X-X-X-X-C

Glu

Asn

Val

Gly

Songyang et al. (Drk) [40, 133]

G-D-A-pTyr-X-X-X-S-P-L-L-L

Tyr (2.3) Ile (1.4) Val (1.3)

Asn (7.1)

Phe (1.7) Leu (1.3) Ile (1.3) Val (1.3)

Songyang et al. (Sem-5) [40, 133]

G-D-A-pTyr-X-X-X-S-P-L-L

Leu (2.3) Val (2.0) Ile (1.9), Met (1.8)

Asn (4.6)

Val (1.7) Pro (1.7)

Songyang et al. [40, 133]

G-D-A-pTyr-X-X-X-S-P-L-L-L

Gln (3.0) Tyr (2.3) Val (1.7)

Asn (5.8)

Tyr (2.2) Gln (2.1) Phe (1.8)

Wards et al. [146]

O-O-O-O-O-pTyr-O-O-O-O-O O-O-O-O-O-pTyr-O-O-O-O-O O-O-O-O-O-pTyr-O-O-O-O-O O-O-O-O-O-pTyr-O-O-O-O-O " " " " " " " " " " " " " " " " " " " " " "

Ile (3) Val (3) Met (3) Gln (2) Ser (2) Leu (2) Ala, Asp, Glu, His, Tyr

Asn (10) Ile (2) Asp (2) Ala, Gln, Glu, Leu, Ser, Tyr

Pro (4) Gln (2) Thr (2) Met (2) Ala (2) Val (2) Asn, Asp, Glu, Ile, Ser, Tyr

Glu (4) Val (3) Asp (2) Lys (2) Ser (2) Arg, Asn, His, Leu, Phe, Pro, Tyr

Amino acid substitutions identified as the most favorable and unfavorable for peptide binding to Grb-2 from five studies. The values behind amino acid represent a measure for selectivity, but are not comparable in their numeric values between methodological different studies. Songyang et al. and Müller et al. report "preference values" for individual amino acids, corresponding to the relative enrichment of the amino acid over the expected value for no preference at all. Gram et al. list the relative occurrence of a certain amino acid in a position in %, a random incorporation rate would correspond to an averaged occurrence of 5%. An occurrence of 33% can therefore be seen as a 6.6 fold preference and is somewhat comparable to Songyang et al. and Müller et al. The study by Wards et al. used 256 individual peptides and the given number reflect how often a certain amino acid was found among the top 20 binding peptides. The study form Olingino et al. identified only one peptide (see text for details). Positions represented by X were randomized, positions marked with O were defined by one single amino acid.

position. To randomize the next position (XOXmYLIP) using the split and recombine method would make it necessary to split each of the 20 sub-libraries into again 20 sub-libraries and to perform effectively 400 independent amino acid coupling. This substantial effort can be circumvented by abandoning the split and recombination method and instead to couple a mixture of amino acids (see also 2.1, Figs. 5 & 6, and 4.2). Inhibition of PTP activity and positional scanning provides as straight forward methodology for the identification of inhibitors. It can be used for the screening of modified peptides, small organic molecules or any other compound library, but can not be applied to PTB/SH2 domains because they lack catalytic activity.

4.2. Affinity Selection and Deconvolution by Mass Spectrometry Affinity selection of pseudosubstrates/inhibitors of PTP1B and deconvolution by mass spectrometry was used by Huyer et al. and was also part of the study of Pellegrini et al. (see 4.1) [139, 140]. Both studies used peptides containing hydrolysis resistant phosphotyrosine mimics. Huyer et al. used difluoromethylphosphonophenylalanine (F2Pmp) (Fig. 8, comp. 7) as phosphotyrosine mimic and a peptide library of the format DADE(F2Pmp)L. Pellegrini et al. used malonyl tyrosine (mY) (Fig. 8, comp. 9) as pTyr mimic and their library was based on the peptide

386


DADEmYLIP. The sequence DADEpYLIP is a well characterized PTP1B substrate and is derived from the autophosporylation site Y992 of the epidermal growth factor receptor (EGFR) [141]. Pellegrini et al. had screened a library of the format DXXX(mY)LIP by positional scanning in an PTP1B inhibition assay and identified the sequence DDEWmYLIP as the peptide with the highest inhibitory activity (see 4.1) [139]. To cross-validate these results a second set of experiments based on affinity selection as performed. The 400 compound library of the format DDXX(mY)LIP was subjected to affinity selection using PTP1B immobilized on a chromatographic support. The retained peptides were released from PTP1B and analyzed by electrospray massspectroscopy (ES-MS) and tandem mass spectroscopy (MS/MS). The deconvolution method is based on measuring the exact mass of peptides. This allows them to identify the two amino acids in the randomized positions (XX) of most peptides, but it does not yield information about the amino acid sequence. However, it is possible to select peptides with defined mass within the mass spectrometer and then subject the peptides to conditions where the peptides breaks into fragments. The masses of these fragments are measured in a second mass spectrometer. The entire process is done in a single instrument known as electrospray-tandem mass spectrometry (ES-MS/MS). Analyzing the masses of the peptide fragments, in combination with the mass of the nonfragmented peptide allows the determination of the amino acid sequence of the parent peptide. Using this methodology the authors found that peptides with acidic residues (Glu, Asp) in position -2 and Tyr in position -1 were selected from the 400 member library ((DDXX(mY)LIP). The PTP1B activity inhibition assay had suggested Trp as most favorable in position -1 (see 4.1), but using the method described here Trp was not seen at all (for discussion see part V). Huyer's et al. used a positional scanning approach to characterize all five positions in the peptide DADE(F 2Pmp)L and a pseudosubstrate/inhibitor trapping method [140]. Five libraries were synthesized by coupling a mixture of 19 amino acids (all natural amino acids except cysteine) in the randomized position, while keeping the nondegenerated positions unaltered. For example, to scan the position -4 the library would have the format XADE(F2Pmp)L, for position +1 it would be DADE(F2Pmp)X. The amino acid mixture used to randomize positions contained the amino acid in ratios according to their relative reactivity (isokinetic mixture). This should lead to an equimolar incorporation of all amino acids. However, amino acid analysis indicated that Val, Leu, Ile and Thr were over-represented, while Gly, Ser and Pro residues were under-represented. The peptide libraries were incubated with PTP1B. Free peptides were separated from PTP1B/peptide complexes by size exclusion chromatography. The eluted PTP1B/peptide complex was dissociated, the phosphatase was then removed by ultrafiltration and the free peptides collected. The pool of peptides was characterized by electrospray ionization mass spectroscopic (ES-MS). The peak intensity was correlated with the relative amount of peptides using an internal standard. Since each library contained only 19 compounds it

Vetter and Zhang

was possible to rapidly identify individual peptides. Huyer et al. found a clear preference for the acidic residues (Glu, Asp) in positions -4, -3, -2 and -1. The preference for Tyr/Trp over Glu in position -1 suggested by the study of Pelligrin et al. was not seen, but selection of Tyr, Phe and Ile/Leu was comparable to Asp. The position C-terminal to the F2Pmp residue (position +1) showed little selectivity for any particular amino acid, but indicated unfavorable binding with Arg, His, or Trp (Table 2) [140]. Mass spectroscopic deconvolution offers several advantages. The high sensitivity requires less material than needed for amino acids sequencing and is much faster. Most importantly, it can be used with modified peptides, nonnatural peptides or small organic molecules. An intrinsic disadvantage is that isobaric compounds can not be distinguished. Isoleucine and leucine have identical atomic composition and can not be distinguished by MS. In addition, lysine and glutamine have very similar masses and their corresponding MS peaks superimpose on each other. If libraries contain more then one degenerated position the number of isobaric amino acid pairs increase quickly which renders unambiguous identification of amino acid sequences difficult. 4.3. Immobilized PTP Substrate Deconvolution by Encoding

Libraries

and

An interesting strategy that uses one-bead one-compound libraries of phosphopeptides and the dephosphorylation of peptides by an active phosphatase has been described by Cheung et al. [142]. Incubation of a phosphopeptide library with a PTP will result in gradual dephosphorylation of the peptides. These peptides need to be identified, if possible quantitatively, among thousands of beads. The assay used by Cheung was based on the selective proteolytic cleavage of de-phosphorylated peptides, followed by fluorescence labeling for detection and reading of an encoding tag for deconvolution (Fig. 12). The protease α-chymotrypsin cleaves amide bonds preferentially on the C-terminal site of aromatic residues, in particular tyrosine. Tyrosine phosphorylation altered the substrate properties of the peptide by a factor of 100 in favor for cleavage of the non-phosphorylated form. The following screening scheme was developed: a one-bead one-compound library Fmoc-XAXXpYLIPQQG-@ was synthesized based on the EGFR peptide DADEpYLIPQQG. It is important to note, that the N-terminus was reversibly protected by the Fmoc group. The positions -1, -2 and -4 were degenerated with eight amino acids (D/E/G/V/S/M/Q/P), resulting in a 83 = 512 member library. The phosphotyrosine building block was incorporated as a mixture with 30% glycine, generating a second peptide strand in parallel on the same bead. This strand has the same amino acid sequence as the phosphotyrosine containing strand, except that it contained a Gly instead of pTyr and functioned as an encoding strand. The library was incubated with the PTP LAR and subsequently exposed to proteolytic digestion with α-



387

Fig. (12). Deconvolution by encoding. A strategy to analyze the substrate preference with bead-bound substrate peptides is schematically shown. The encoding strand is resistant to dephosphorylation dependent proteolysis by a protease. Digested beads can be labeled with a fluorescent compound and isolated for sequencing of the encoding strand. The limitations of the methods are discussed in the text.

388


Vetter and Zhang

chymotrypsin. Dephosphorylated peptides were cleaved by the protease, leaving a peptide fragment with a free Nterminal amino group H2N-LIPQQG-@ on the resin bead. These beads were identified by fluorescence labeling of the free amino terminus with fluorescein isothiocyanate. Fluorescent beads were then manually removed from the bulk of non-labeled beads. To deconvolute the original peptide sequence the encoding peptide strand was used. The Fmoc-group of the non-proteolyzed (glycine containing) encoding peptide strand was removed and the bead was subjected to amino acid sequencing, revealing the entire sequence of the original pTyr containing peptide (Fig. 12). This screening method identified peptides with acidic residues in positions -1, -2 (Asp, Glu) and -4 (Asp/Asn, Glu), therefore basically iterating the sequence of the EGFR peptide. The described method has several limitations, which restrict its applicability significantly. A number of amino acids must remain excluded from the library, in particular aromatic residues, which would initiate cleavage by chymotrypsin in the coding strand, and also lysine can not be incorporated, because it would interfere with the fluorescence labeling step. 4.4. Soluble PTP Substrates and Deconvolution by Inverse Alanine Scanning

format requires the synthesis of 20 single peptides to scan one position. The small number of peptides allows analytical characterization of each peptide, an advantage over conventional combinatorial methods where the large number of peptides renders a detailed characterization of the library difficult or impossible. A library of 153 peptides was used to scan a total of eight positions, four C- and four N-terminally to a central phosphotyrosine. For example, the peptides used to scan the -4 position had the format OAAApYAAAA, where O represents any of the 20 standard amino acids. Each peptide was subjected to a PTP1B substrate assay and the rate of dephosphorylation was measured. The obtained kinetic parameter kcat/Km furnished information about the contribution of individual amino acids with respect to substrate recognition for each peptide. It was found that indeed every peptide was dephosphorylated by PTP1B but with considerable difference in kcat/Km values. The outer positions had less impact on the substrate properties than positions adjacent to the central pTyr residue. Acidic residues were preferred in N-terminal position -4 and -2, Leu was selected for -3 and position -1 selected Phe or Tyr. The C-terminal position selected Met in position +1, an acidic residue in +2 and aromatic residues Phe or Tyr in +3. Combining the most favorable substitution in each position yielded a novel consensus substrate peptide ELEFpYMDYE (Table 3).

Inverse alanine scanning has recently been used to characterize the substrate preference of PTP1B [106]. This approach takes advantage of the relative broad substrate acceptance of PTP1B. The method uses a set of peptides composed entirely of alanine residues, except for a phosphotyrosine residues and one defined position. This

A nice feature of the inverse alanine scanning approach is the possibility to analytically characterize every peptide and directly measure the kinetics of the enzymatic dephosphorylation. Limitations may include that not every PTP will accept alanine rich peptide as substrates and that potentially important interactions with amino acid side

Table 3.

Selectivity of PTP1B

Study

Library format

position -4

Position -3

Position -2

position -1

position +1

Huyer et al. [140]

X-A-D-E-pTyr-L D-X-D-E-pTyr-L D-A-X-E-pTyr-L D-A-D-X-pTyr-L D-A-D-E-pTyr-X

Glu (3.0) Asp (1.9) Phe (1.3) Asn (1.2) :::::::: Arg (0.4)

Glu (3.5) Asp (2.2) Ala (1.2) Ser (1.1) :::::::::: Arg (0.2)

Glu (6.6) Asp (4.1) Pro (1.3) Gly (1.1) :::::::::: Arg (0.5)

Glu (2.8) Tyr (1.8) Asp (1.6) Phe (1.5) ::::::::::: Arg (0.3)

Met (1.8) Gly (1.7) Asn (1.5) Gln (1.5) ::::::::::: Arg (0.3)

Pellegrini et al. [139]

D-O-X-X-pTyr-L-I-P D-X-O-X-pTyr-L-I-P D-X-X-O-pTyr-L-I-P

Asp (68%) Gly (58%) Trp (55%) hPro 55%) ::::::::::::: Lys (5%)

Glu (65%) Asp (55%) Ser (52%) Gln (38%) :::::::::: Lys (3%)

Trp (55%) Tyr (45%) Phe (35) Leu (30%) :::::::::: Lys (8%)

Vetter et al. [106]

O-A-A-A-pTyr-A-A-A-A A-O-A-A-pTyr-A-A-A-A " " " " " " " " " A-A-A-A-pTyr-A-A-O-A A-A-A-A-pTyr-A-A-A-O

Leu (27.4) Gln (26.0) Glu(24.2) Tyr (22.0) ::::::::::::: Lys (3.6)

Glu (26.0) Asp(24.4) Pro (23.0) Gly (19.1) ::::::::::::: Lys (5.6)

Phe (48.4) Tyr (44.5) Leu (37.9) Asp(35.0) ::::::::::::: Lys (3.6)

Glu (27.0) Tyr (19.3) Trp (18.9) Leu (18.1) :::::::::: Lys (5.8)

Met (38.8) Cys (30.6) Gln (26.2) Leu (17.0) ::::::::::::: Pro (0.5)

Results from three studies assessing the substrate specificity of PTP1B. The numeric values in brackets indicate relative preference of amino acids in a each position. Shown are the four "best" and the "worst" substitution. The values allow comparison between the "best" and "worst" substitution in each position, but are not comparable among the different studies (see text for details).


chains of the peptide itself (cooperativity effects) are suppressed due to the substitution with "inert" alanine residues. PART V. RESULTS AND CONCLUSIONS FROM ASSESSMENTS OF PTB1B AND GRB-2 WITH COMBINATORIAL STRATEGIES The research on phosphotyrosine recognition by SH2/PTB domains or protein tyrosine phosphatases is driven by several motives. The structural biologist is focused on the molecular recognition event and the interactions between the protein and the pTyr containing peptide. The molecular pharmacologist attempts to generate compounds that bind to a PTP or SH2/PTB domains with the intend to specifically interrupt or to modulate its function. Others wish to understand the biological function of PTPs or SH2/PTB domain containing proteins in a cellular context. For those researchers it becomes important to identify the in-vivo substrates of a certain PTP or to find the target proteins of a PTB or SH2 domain. At the same time, it can be helpful to have specific antagonist and inhibitor at hand in order to probe and to understand the cellular function of a protein. Specific PTP inhibitors and SH2/PTB antagonists, which would block these proteins specifically at a defined time point of the cell cycle, are still missing today, but would certainly be appreciated by many researchers. In this context it becomes worthwhile for everybody involved in protein phosphorylation, phosphatases and phosphorylation mediated signal transduction to take a critical look at combinatorial library methods and to ask what they can do for you. How useful are screenings of peptide libraries and what kind of conclusions can be drawn from them? Are they largely an academic exercise, or can they identify relevant in-vivo substrates. Can libraries guide the path to the development of selective inhibitors? What type of library shall one actually use? Do different screening and deconvolution methods influence the results? How well do results from combinatorial library screens agree with results from biochemical experiments? And last but not least, how coherent are results from different combinatorial methods? The last part of this review can obviously not address all these points. Instead we compare a few results from studies involving PTP1B and the SH2 domain of Grb-2. The combinatorial methods employed in these studies have been mostly described in part III and IV. Both proteins are considered potential drug targets and have subsequently attracted considerable interest, in particular with respect to the development of selective antagonists. Grb-2 is crucially involved in receptor tyrosine kinase initiated signaling, because is transduces the primary (auto-)phosphorylation signal of transmembrane receptors to Ras controlled signaling cascades. The SH2 domain of Grb-2 binds on the one side to the receptor tyrosine kinase (RTK), while its SH3 domains on the other side facilitates binding to SOS (Grb-2 also binds to a number of other proteins important in cellular regulation). SOS activates Ras, which in turn activates a cascade of Ser/Thr kinases, ultimately transducing the signal into the nucleus and leading to gene regulation. Therefore, a specific Grb-2 SH2 inhibitor would intercept RTK initiated


389

gene expression at an early stage, making Grb-2 antagonists interesting reagents in cancer research [143]. PTP1B on the other hand has long been believed to be intricately involved in insulin mediated signaling and fat/carbohydrate metabolism. Recently, studies on PTP1B knock-out mice have demonstrated that these animals display enhanced insulin sensitivity and withstand a high fat diet without becoming obese or losing sensitivity towards insulin, which would lead to type II diabetes. These results obviously suggest PTP1B inhibitors as interesting reagents for researchers working on diabetes, dietary caused obesity and associated health problems [144]. Due to the large amount of publications on PTP1B and Grb2 the studies selected here are just examples to illustrate the principles and by no means a complete coverage of the subject. 5.1. Specificity of the SH2 domain of Grb2 The first systematic study to define the specificity of SH2 domains with combinatorial methods was conducted by Songyang et al. in 1993 (see part III 3.2) [40, 133]. A very small phosphopeptide library consisting of only 5832 members was screened against a panel of 22 SH2 domains. This study was important because it demonstrated for the first time that the screening of a combinatorial peptide library can be used to define families of SH2 domains with similar selectivity. Among the screened SH2 domains was the murine form of Grb-2. The screen revealed a very clear preference for an asparagine residue in +2 position relative to pTyr. Asn was 7 fold preferred over the background and clearly selected over the structurally related residues Asp, Glu and Gln. However, selectivity was much less pronounced for positions +1 and +3, where preferences for a certain amino acid were only 3.0 to 1.7 fold. The authors concluded that Gln (3.0 fold), Tyr (2.3 fold) and Val (1.7 fold) are preferred in position +1, and Tyr (2.2 fold), Gln (2.1 fold) and Phe (1.8 fold) are favorable in position +3 (Table 2). Also included in the screening were the Grb-2 homologues from two other species, Drk from Drosophila malanogaster and Sem-5 from Candida elegans. The SH2 domains of these three proteins are not identical, but share a significant degree of sequence homology. In addition, the RTK-Grb2/SOS-Ras-Raf signaling pathway is remarkably conserved among the different organisms and Grb2/Drk/Sem-5 are functional homologues. Considering the conserved sequence and cellular function it would not be surprising, if the three domains would have also a closely related ligand selectivity. Indeed, all three SH2 domains show a very strong preference for peptide ligands with an asparagine residue in +2 position (7.1 and 4.6 fold over background). Preferences for the neighboring positions +1 and +3 were generally weaker then for the +2 position, and more importantly differed among the three homologues. The preference for Gln was not seen in either Sem-5 or Drk. Sem-5 instead preferred Leu (2.3) and Val (2.0) in position +1 and Val (1.7) or Pro (1.7) in position +3. Drk selected Tyr (2.3) in position +1 and Phe (1.7) in position +3 (Table 2). Another interesting piece of information comes from amino acids that were not selected and strongly disfavored. Basic,

390


small and hydrophobic residues were frequently disfavored in any of the three screened positions. In summary, the study from Songyang et al. identified clearly Asn in position +2 as an important motif in Grb-2 ligands and suggested preference for hydrophobic residues in +1 and +3 position over basic, very small or hydrophilic residues. A few years later Müller et al. screened a 100 times larger library (6.4x105 members) against Grb-2 [134]. The study analyzed the positions from -3 to +5, but here only positions +1 to +3 are considered to allow comparison with the other studies. Müller et al. also found a very clear preference for Asn in +2 position, a result that is in agreement with the earlier study and findings from biochemical experiments. However, the selected amino acids in position +1 and +3 were quite different compared to the results from Songyang et al. The selectivity of hydrophobic residues (Phe, Tyr, Val) was not seen, instead the acidic residues Glu and Asp seemed to make up particular good ligands (Table 2). Again, certain residues were identified as particular disfavored in +1 and +3 positions and these were again basic, small or hydrophilic residues. A study published in the following year used a phage display method to present phosphorylated peptides [137](see II.1). This study found Asn in +2 position in all sequenced clones. Position +1 selected either Glu or Met, which is somewhat comparable to the two other studies mentioned here. However, for the +3 position something new was discovered. Over 50% of the sequenced clones had a Trp residue in +3 position, and 25% Phe residues. This finding would suggest a strong preference for an aromatic residue in this position. A hydrophobic/aromatic selectivity was already observed by Songyang, however favoring Tyr or Phe instead of Trp. On the other side this study agreed with the findings of Müller et al. concerning the selectivity for the +4 position, where both studies found preference for Pro (Table 2). The difficulty to define an optimal peptide ligand for Grb-2 got further underlined by another phage display approach also reported in 1997 by Olingion et al. [145]. In this case a library of non-phosphorylated, cyclic peptides was screened. These experiment identified a peptide that could indeed bind into the phosphotyrosine binding cleft without actually being a phosphopeptide. The peptide bound with an IC50 of 10-25 µM, which is comparable to typical binding affinities of phosphorylated peptide ligands. The peptide sequence contained an Asn residue in +2 position relative to the (non-phosphorylated) Tyr residue. Position +1 was occupied by Glu and position +3 by Val. The sequence Tyr-Glu-Asn-Val does correspond quite well to the preferred substitutions found in the study of Müller et al. (Table 2). However, the peptide was restricted in its conformation due to the disulfide bridge between the two cysteine residues in positions -3 and +7. It can therefore not be directly compared to linear peptides used in other studies. On the other side, the finding of Asn in position +2 was in agreement with all previous studies and once again confirmed the significance of the Tyr-Xxx-Asn motif. The last study mentioned here used a quite different approach. Wards et al. based their experiment on the fact

Vetter and Zhang

that Grb-2 binds to the cytosolic portion of tyrosine phosphorylated receptor tyrosine kinases [146]. The authors chose to analyze the epidermal growth factor receptor (EGFR), the platelet derived growth factor receptor (PDGFR), the basic fibroblast growth factor receptor (bFGFR), the insulin receptor (IR) and insulin receptor substrate I (IRS1) for potential Grb-2 binding sites. Each tyrosine residue of the proteins was considered a potential binding site and peptides corresponding to each site were generated. A total of 256 synthetic 11 residue peptides, each containing a central phosphotyrosine residue flanked by 5 residues corresponding to the sequence found in the receptor were tested. The peptides were biotinylated and immobilized on streptavidin coated multi-well plates. A Grb-2-GST fusion protein construct was incubated with the immobilized peptides and the relative binding affinity of each peptide to Grb-2-GST was measured using labeled anti-GST antibodies. The study was overall in good agreement with biochemical data concerning Grb-2 binding sites to these proteins. All known binding sites were found and additionally new binding regions were identified. Importantly, the 10 peptides with highest affinity all had the pTyr-Xxx-Asn motif, with variable substitution in +1, +3 and +4 position (Table 2). The study yielded also some surprising results. The peptides ranked 11-20 (out of 256) in binding affinity to Grb-2 did not contain the Asn residue in +2 position, but instead a variety of other residues (Table 2). Moreover, the suggested rule that small, hydrophilic and basic residues are disfavored positions +1 and +3 was violated and peptides with residues such as Ser, Ala, His and Arg in these positions were found have decent binding affinity to Grb-2. Looking at the above mentioned studies in retrospect gives the impression that it is difficult to define the specificity of Grb-2 with one simple consensus sequence. Further, the amino acid sequence preferences are partially contradictory and it becomes difficult to conclude much more than the pTyr-Xxx-Asp motif, where Xxx shall not be a small, basic or hydrophobic amino acid. This raises several questions: Did some of the studies fail to identify the right ligand? Is one methodological approach better then another? How long should a peptide ligand be to identify a consensus sequence? Is there actually a true consensus peptide for Grb2? Keeping this in mind one has to see that the five studies described here presented the potential Gr-2 ligands in quite different formats. The study by Songyang et al. randomized three positions within the context of a defined 12-mer peptide. In contrast, the study of Müller used a library of 9mer peptide with 6 randomized positions and Wards et al. used set of completely defined 12-mer peptides. Further, the other two studies utilized phage display systems and either cyclic non-phosphorylated peptides or enzymatically phosphorylated peptides. Taking these differences into account, one can conclude that the presentation of the ligand library influences partially the outcome of the preferred ligand. In addition, it suggests that there is not "one true" consensus peptide for Grb-2 binding. This indicates that the in-vivo binding partners of Grb-2 are not solely defined by a few amino acids immediately adjacent to a pTyr, but that also residues further away are important.


5.2. Novel Grb-2 Ligands Evolved from the pTyr-IleAsp Motif Looking back at the above paragraph it might be tempting to conclude that combinatorial library studies are of little value, because they do not yield a unique consensus and results can vary from study to study. However, such a judgment is premature. Every identified binding peptide contains information concerning the plasticity of molecular recognition between protein and ligand. This is particularly obvious when structural data from NMR or crystallographic experiments for different protein-peptide complexes are available. To illustrate this we will briefly present a series of consecutive experiments where an existing Grb-2 ligand was systematically modified and evolved into a novel high affinity ligand [143, 147-154]. The three-residue peptide pTyr-Ile-Asn (Fig. 13, comp. 11) was the starting point for a Novartis research team. The motif pTyr-Xxx-Asn had been suggested as consensus motif for Grb2 binding (see above) and the three residue peptide was a known Grb2-SH2 ligand. The low binding affinity (57 µM) of the peptide rendered it of little value for the inhibition of Grb2. Ligands with much higher affinity and a longer in-vivo half-live time were sought. The peptide was stepwise modified in a systematic manner and every position was screened for possible substitutions that would lead to higher affinity. Initially, it was observed that the coupling of an anthranilic acid onto the N-terminus of H2N-pTyr-Ile-Asnamide caused an increase in affinity to Grb-2. The systematic evaluation led to the identification of 3-aminobenzyloxycarbonyl group (3-amino-Z) as a modification that increased binding affinity of the peptide 870 fold from IC 50 of 57µM to 65nM [148] (Fig. 13, comp. 13). This dramatic increase in binding affinity could be rationalized by modeling the new peptide ligand into the Grb-2 binding side. It turned out that the aromatic portion of the (3-amino)-Z-group formed stacking interactions with a conserved arginine side chain of the binding pocket and the amino group established a hydrogen bond to the phosphate moiety, both energetically favorable interactions that cause tighter binding. Equally, the systematic coupling of various amines onto the C-terminus and screening of a small library led to a derivative with a 27 fold increased binding (Fig. 13, comp. 13 and 14) [150]. Besides optimizing the terminal position, the internal positions were also targeted. The Ile residue in +1 position was investigated with a series of cyclic hydrocarbon amino acids. Ile could not only be functionally replaced by 1-amino-cyclohexanecarboxylic acid (Ac6c), but also the binding affinity increased 65 fold (Fig. 13, comp. 15 and 19) [149, 152]. It seemed challenging to attempt replacement of the Asn residue after the pTyr-Xxx-Asn motif had been proposed as core motif for Grb-2 binding. Analyzing crystallographic data identified the key requirements for the recognition of Asn in +2 position. The peptide ligand forms a type I β-turn and the carboxamide group of Asn forms a hydrogen bond with the peptide backbone [147, 151]. Based on this structural information suitable mimics were designed. (1S,2R)-2-amino-cyclohex-


391

3-ene carboxylic acid was found to be able to replace Asn with only a minor loss in affinity (1.6 fold, IC50 1.6 nM) (Fig. 13, comp. 16). Another interesting compound contained a pTyr mimic to increase stability towards dephosphorylation in vivo. This compound also had a novel C-terminal hydroxy indol modification (Fig. 13, comp. 18) [153]. This ligand showed IC50 values in the sub nano molar range (IC50 0.3 nM) and the affinity is more than 5 orders of magnitude increased compared to the starting peptide H2N-pTyr-Ile-Asn-amide (IC50 57 µM vs. IC50 0.3 nM). While the compound is still an oligoamide, it differs substantially from "native" peptides. This makes it less susceptible to in vivo degradation and more suitable for biochemical and pharmacological applications. This example demonstrates nicely that the consensus motif identified from combinatorial library screens could be evolved into a highly potent Grb-2 antagonist using again combinatorial methods. However, the crystal structures of ligands complexed to the Grb-2 SH2 domain had been fundamental for the design of these libraries. The identification and understanding of crucial binding interactions within the protein-ligand complex helped greatly in the design of next generation libraries and ligands. 5.3. Specificity of PTP1B Protein tyrosine phosphatase 1B is the workhorse of phosphatase research and a very well studied example of the protein tyrosine phosphatase family. One of the early studies towards PTP1B substrate specificity used a set of peptides derived from the autophosphorylation site of the EGF receptor 988-998 [155]. The peptide DADEpYLIPQQG was subjected to a classical alanine scanning in order to identify amino acids important for the recognition by PTP1B. The substitution of each amino acid by an Ala generally reduced the affinity of the peptide to the protein. However, the degree of loss in affinity was clearly position dependent. The most important residues was found to be the Glu residue in position -1, causing an increase in Km from 2.6 µM to 12.4 µM. Also of importance were the proline residue in +3 position, Ile in +2 and Asp in -2 position. This study demonstrated two features. First, it showed a general greater importance of residues close to the pTyr residue compared the residues further away. Secondly it demonstrated a preference for acidic residues N-terminally to the pTyr, as well as importance of the Ile-Pro sequence in positions +2 and +3. A subsequent investigation into the size requirements for substrate recognition by PTP1B used truncated forms of the EGFR peptide [141]. None of the truncated peptides showed a better binding affinity than the template peptide, however, some came close. Striking findings were that residues Nterminal of the pTyr were more important than residues Cterminal to the tyrosine. The peptide pYLIPQQG showed an Km of 377 µM, whereas the peptide DADEpY had a Km of 16 µM. It was also seen that acetylation of the N-terminus and amidation of the carboxy-terminus always increased binding affinity compared to non-modified termini.

392


Vetter and Zhang

Fig. (13). Evolution of Grb-2 SH2 domain lignads. Combinatorial library methods in conjunction with X-ray structure based design led to a series of Grb-2 SH2 ligands shown in the figure. The starting peptide pTyr-Ile-Asn 11 had only moderate affinity for Grb-2. A series of substitution (marked with dashed boxes) were discovered and resulted in dramatic increases affinity for Grb-2. Compound 16 has an more than 35.000 fold increased IC50 compared to the parent compound 11 while residues Ile and Asn could be replaced with non-natural amino acids. In compound 18 the pTyr residue is replaced by a PTPase resistant mimic (compound 6, figure 9) and the IC50 of 0.3 nM makes it a very potent Grb-2 antagonist.

In the year following to the above mentioned studies the crystal structure of peptide Ac-DADEpYL-amide bound to the catalytically inactive Cys215Ser mutant of PTP1B was published and revealed key-features of the peptide binding mode [156]. Besides crucial hydrogen-bonds between the phosphate group and the protein, a surprisingly large number of hydrogen bonds involving the peptide backbone was seen, but only a limited number of interactions involved peptide side-chains was found. The most prominent interactions were found between the carboxyl group of Glu and Asp in 1/-2 position of the peptide with the guanidinium group of Arg 47 of PTP1B. Several studies on PTP1B substrate specificity involving large combinatorial libraries have been published since 1998 and these provide some interesting new insights into the plasticity of protein - ligand interaction in PTP1B. The study of Huyer et al. used an affinity selection strategy involving functional PTP1B and the hydrolysis resistant pTyr mimic

F2Pmp (see part IV 4.2) [140]. In essence, the study found a clear preference for the acidic residues Glu or Asp for all four positions N-terminal to the pTyr residue. In addition, position -1 also accepted the aromatic residues Tyr and Phe with selectivity values comparable to substitutions with Glu and Asp (Table 3). The position immediately C-terminal to pTyr (+1) did not show a strong selectivity, but the residues Gly or Met appeared to be somewhat preferred. On the other side, the study showed clearly, that the basic residue Arg was the worst possible substitution in any position (Table 3). The study conducted by Pellegrini et al. focused on the N-terminal positions -3 to -1 and confirmed the preference for Asp and Glu in position -2 [139]. However, these authors found a unexpected selectivity for aromatic residues in positions -3 and -1 (Table 3). Indeed, position -1 clearly selected Trp, Phe and Tyr over Glu or Asp. Position -3 appeared to accept various different substitution equally well, discriminating only poorly between Asp, Glu, Gly, Trp


and hydroxyproline (Table 3). Again, basic residues were unfavorable substitution, however, in this study Lys was found to be worse then Arg. The preference for aromatic residues in position -1 was also seen in an "inverse alanine scanning” approach [106]. In this study the hydrophobic/aromatic residues Phe, Tyr and Leu were identified to be better substitutions then Asp or Glu (Table 3). This study also scanned the four C-terminal position and found a clear preference for Met in position +1. In addition, acidic residues Glu/Asp were selected for position +2, the aromatic residues Phe and Tyr in position +3 and position +4 showed only a slight selectivity for acidic residues (Table 3). The deduced consensus peptide ELEFpYMDYE showed a very high kcat/Km (2.2x107 M -1s-1) and was equally well accepted as a substrate as the EGFR peptide DADEpYLIPQQG. This is remarkable, because the two peptides differ significantly in size and amino acid sequence. The three combinatorial library studies described above used experimentally very different approaches to assess the specificity of PTP1B for phosphopeptide recognition and led to similar, but not identical results. By looking not only at the "best" substitution for each position in a PTP1B substrate peptide, but also considering the "top four" substitutions makes it difficult to identify a single "best" sequence (Table 3). This suggests, that there might not be a single "true" consensus sequence and raises an interesting question: how is it possible that peptides with quite different amino acid sequences can be equally good as substrates? The crystal structure of the PTP1B - DADEpYL peptide complex had clearly shown that the carboxyl group in -1 position formed a hydrogen bond to a non-conserved arginine residue (Arg47) [156]. This interaction was one of a few interactions involving the side chains of the peptide ligand and was considered as a determinant for substrate recognition. Now, to the contrary, combinatorial library screens suggest unexpectedly that a Phe/Tyr residue can functionally replace the Glu reside in -1 position. A recent study compared the crystal structures of three PTP1B-peptide complexes in order to address this question [157]. The structure of the EGFR-peptide DADEpYL-amide was compared with the peptides containing aromatic amino acids (Phe and p-benzoylphenylalanine(Bpa)) in position -1 (ELEFpYMDYE and DAD(Bpa)pYLIPQQG respectively). In essence, all three peptides were similarly oriented within the substrate binding pocket and the positioning of the pTyr residue and the peptide backbone between positions -1 and +1 were basically identical in all three structures. The key element for substrate plasticity was found to be the side chain of the Arg 47 residue. If positions -1 and -2 are occupied by acidic residues (EGFR-peptide), then both carboxylic groups are hydrogen bonded to the guanidinium head group of Arg 47. But, if position -1 is occupied by an hydrophobic/aromatic side chain, than the Arg 47 side chain is repositioned (change in χ by 120°) and the guanidinium group moves by 7Å. The hydrocarbon chain of Arg 47 now interacts with the aromatic rings at –1 position and shows hydrophobic interactions with Leu at the –3 position. This example demonstrates that combinatorial libraries can be used to probe the substrate/ligand plasticity of PTP1B and


393

protein in general. The initially unexpected possibility to accommodate acidic, as well as aromatic residues in position crucial for substrate recognition could be rationalized once structural data were available. CONCLUSION "You get what you screen for" can be seen as the golden rule of combinatorial library screenings - it defines the limitations and opportunities. Combinatorial methods can be extremely useful tools for the detailed investigation of protein-ligand interactions. A plethora of information about binding site plasticity can be extracted from a careful analysis of screening results. The first step of any library experiment should concern the decision about the kind of compounds to be synthesized: peptides, modified peptides, small molecules etc. This will depend on the question under investigation: substrate profile, inhibitor design or in vivo ligand. Subsequently the library has to be synthesized and one has to make compromise between library size, library diversity, quality control, and labor effort. A moderately sized library without uncertainties about completeness and quality is probably a better starting point than a large library synthesized under conditions that do not allow any type of quality control. The screening assay and the identification of positive hits is obviously crucial. Individual sequence information are more valuable than averaged data from pool sequencing experiments. Analysis and interpretation of screening results should include all information and not only focus on the single "best" substitution for a given position. Proper analysis of screening results can be a key factor for the successful design of second generation libraries. Structural information or detailed kinetic data of the ligand recognition can guide the design of second generation libraries in the right direction. This is particularly important for the development of enzyme inhibitors and protein antagonists. Finally, a certain degree of care should be taken when conclusions regarding potentially biological relevant protein-protein interactions are drawn from the screening of simple peptide libraries. The binding of a short peptide to an protein is an enormous oversimplification of the actual protein-protein interaction and it is often not a reliable enough model to speculate about in vivo interactions. ACKNOWLEDGEMENTS This work was supported in part by National Institutes of Health Grant AI48506, the G. Harold and Leila Y. Mathers Charitable Foundation, and the Irma T. Hirschl Foundation. ABBREVIATIONS ATP

= Adenosine triphosphate

Boc

= Tert.-butoxycarbonyl

Bpa

= p-Benzoylphenylalanine

ELISA = Enzyme linked immunosorbent assay F2Pmp = Difluoromethylphosphonophenylalanine

394


FHA

= Forkhead associated (domain)

Fmoc

= Fluorenylmethoxycarbonyl

FTIC

= Fluorescein thioisocyanate

Grb-2

= Growth factor bound protein 2

GST

= Glutathion S-transferase

IRS

= Insulin receptor substrate

Vetter and Zhang

[10]

Neel, B. G.; Tonks, N. K. (1997) Curr. Opin. Cell Biol., 9, 193-204.

[11]

Su, X.-D.; Taddel, N.; Stefani, M.; Ramponi, G.; Nordlund, P. (1994) Nature, 370, 575-578.

[12]

Streuli, M. (1996) Curr. Opin. Cell. Biol., 8, 182-188.

[13]

Bliska, J. B.; Guan, K. L.; Dixon, J. E.; Falkow, S. (1991) Proc. Natl. Acad. Sci. USA, 88, 1187-1191.

[14]

Cantley, L. C.; Auger, K. R.; Carpenter, C.; Duckworth, B.; Graziani, A.; Kapeller, R.; Soltoff, S. (1991) Cell, 64, 281302.

[15]

Goldstein, B. J.; Ahmad, F.; Ding, W.; Li, P. M.; Zhang, W. R. (1998) Mol. Cell. Biochem., 182, 91-99.

[16]

Malchiodi-Albedi, F.; Petrucci, T. C.; Picconi, B.; Iosi, F.; Falchi, M. (1997) J. Neurosci. Res., 48, 425-438.

[17]

Maxham, C. M.; Malbon, C. C. (1996) Nature, 379, 840844.

[18]

Michielis, T.; Cornelis, G. (1988) Microbiol. Pathogen., 5, 449-459.

[19]

Myers, M. G. J.; White, M. F. (1996) Annu. Rev. Pharmacol. Toxicol., 36, 615-658.

[20]

Shultz, L. D.; Rajan, T. V.; Greiner, D. L. (1997) TIBTECH, 15, 302-307.

[21]

Yakura, H. (1994) Crit. Rev. Immunol., 14, 311-336.

[22]

Bolin, I.; Wolf-Watz, H. (1988) Mol. Microbiol., 2, 237-245.

[23]

Hunter, T.; Sefton, B. M. (1980) Proc. Natl. Acad. Sci. USA, 77, 1311-1315.

[24]

Roach, P. J. (1991) J. Biol. Chem., 266, 14139-14142.

[25]

Swanson, R. V.; Alex, L. A.; Simon, M. I. (1994) TIBS, 19, 485-491.

[26]

Castellanos, R. M.; Mazon, M. J. (1985) J. Biol. Chem., 260, 8240-8242.

[27]

Adamikova, L.; Resnick, R. J.; Tomaska, L. (1996) Yeast, 12, 833-838.

[28]

Shi, L.; Potts, M.; Kennelly, P. J. (1998) FEMS Microbiol. Rev., 22, 229-253.

[29]

Barford, D.; Hu, S. H.; Johnson, L. N. (1991) J. Mol. Biol., 218, 233-260.

[30]

Johnson, T.; Quibell, M.; Owen, D.; Sheppard, R.C. (1993) J. Chem. Soc. Chem. Commun., 369-372.

[31]

Johnson, L. N.; Barford, D. (1994) Protein Sci., 3, 17261730.

ESi-MS = Electro spray ionization mass spectrometry PH

= Pleckstrin homology (domain)

PI

= Phosphotyrosine interacting (domain)

PTB

= Phosphotyrosine binding (domain)

PTP

= Protein tyrosine phosphatase

PTK

= Protein tyrosine kinase

pTyr

= Phosphotyrosine

mTyr

= O-Malonyl tyrosine

SH2

= Src homology 2 (domain)

SH3

= Src homology 3 (domain)

SPPS

= Solid phase peptide synthesis

TFA

= Trifluoroacetic acid

WW

= Tryptophan-tryptophan (domain)

@

= Symbolizes a solid support in SPPS

REFERENCES [1]

Graves, D. J.; Martin, B. L.; Wang, J. H. (1994) Co- and post-translational modifications of proteins. Chemical principles and biological effects., Oxford University Press: New York, Oxford.

[2]

Wold, F. (1981) Annu. Rev. Biochem., 50, 783-814.

[3]

Wold, F.; Moldave, K. Eds. (1984). Methods in Enzymology, Academic Press Vol. 106 and Vol. 107.

[4]

Fantl, W. D.; Johnson, D. E.; Williams, L. T. (1993) Annu. Rev. Biochem., 62, 4535-481.

[5]

Avruch, J.; Zhang, X.; Kyriakis, J. M. (1994) Trends Biochem. Sci., 19, 279-283. [32]

Barford, D.; Johnson, L. N. (1989) Nature, 340, 609-616.

[6]

San, H.; Tonks, N. (1994) Trends Biochem. Sci., 19, 480385.

[33]

Forest, K. T.; Dunham, S. A.; Koomey, M.; Tainer, J. A. (1999) Mol. Microbiol., 31, 743-752.

Sun, H.; Tonks, N. K. (1994) Trends Biochem. Sci., 19, 480-485.

[34]

Johnson, L. N.; Barford, D. (1993) Annu. Rev. Biophys. Biomol. Struct., 22, 199-232.

[35]

Consortium, I. H. G. S. (2001) Nature, 409, 860-921.

[7] [8]

Hunter, T. (1995) Cell, 80, 225-236.

[9]

Tonks, N. K. (1996) Adv. Pharmacol., 36, 91-119.



395

[36]

Barinaga, M. (1999) Science, 283, 1247-1248.

[61]

Hunter, T. (1996) Biochem. Soc. Trans., 24, 307-327.

[37]

Cohen, G. B.; Ren, R.; Baltimore, D. (1995) Cell, 80, 237248.

[62]

Ramponi, G.; Stefani, M. (1997) Biochim. Biophys. Acta, 1341, 137-156.

[38]

Kuriyan, J.; Cowburn, D. (1997) Annu. Rev. Biophys. Biomol. Struct., 26, 259-288.

[63]

Guan, K. L.; Broyles, S.; Dixon, J. E. (1991) Nature, 350, 359-361.

[39]

Pawson, T. (1995) Nature, 373, 573-579.

[64]

[40]

Songyang, Z.; Shoelson, S. E.; Chandhuri, M.; Gish, G.; Pawson, T.; Haser, D. M.; King, F.; Roberts, T.; Ratnofsky, S.; Lechleider, R. J.; Neel, B. G.; Birge, R. B.; Fajardo, J. E.; Chou, M. M.; Hanafusa, B.; Schaffhausen, L. C.; Cantley, L. C. (1993) Cell, 72, 767-778.

Henderson, L. E.; Copeland, T. D.; Sowder, R. C.; Smythers, G. W.; Oroszlan, S. (1981) J. Biol. Chem., 256, 8400-8406.

[65]

Zhang, Z.-Y.; Van Etten, R. L. (1990) Arch. Biochem. Biophys., 282, 39-49.

[66]

Charbonneau, H.; Tonks, N. K.; Kumar, S.; Diltz, C. D.; Harrylock, M.; Cool, D. E.; Krebs, E. G.; Fischer, E. H.; Walsh, K. A. (1989) Proc. Natl. Acad. Sci. USA, 86, 52525256.

[67]

Cohen, P. (1989) Annu. Rev. Biochem., 58, 453-508.

[68]

Barford, D. (1996) Trends Biochem. Sci., 21, 407-412.

[69]

Zhang, Z.-Y. (1997) Curr. Top. Cell Reg., 35, 21-68.

[70]

Pannifer, A. D. B.; Flint, A. J.; Tonks, N. K.; Barford, D. (1998) J. Biol. Chem., 273, 10454-1062.

[71]

Zhang, Z.-Y.; Wang, Y.; Wu, L.; Fauman, E.; Stucky, J. A.; Schubert, H. L.; Saper, M. A.; Dixon, J. E. (1994) Biochemistry, 33, 15266-15270.

[41]

Waksman, G.; Kominos, D.; Robertson, C.; Pant, N.; Baltimore, R. B.; Cowburn, D.; Hanafusa, H.; Mayer, B. J.; Overduin, M.; Resh, M. D.; Rios, C. B.; Silverman, L.; Kuriyan, J. (1992) Nature, 358, 646-653.

[42]

Eck, M. J. (1995) Structure, 3, 421-424.

[43]

Margolis, B. (1996) J. Lab. Clin. Med., 128, 235-241.

[44]

Hunter, T. (2000) Cell, 100, 113-127.

[45]

Bhalla, U. S.; Iyengar, R. (1999) Science, 283, 381-387.

[46]

Becker-Ursic, D.; Davies, J. (1976) Biochemistry, 15, 22892296.

[47]

Glass, D. B.; Masaracchia, R. A.; Feramisco, J. R.; Kemp, B. E. (1978) Anal. Biochem., 87, 566-575.

[72]

Barford, D. (1995) Curr. Opinion Struc. Biol., 5, 728-734.

[48]

Glenney, J. R. (1991) Methods Enzymol., 201, 92-100.

[73]

[49]

Williams, R. O. (1972) Biochem. Biophys. Res. Commun., 47, 671-678.

Zhang, Z.-Y. (1998) Crit. Rev. Biochem. Mol. Biol., 33, 152.

[74]

de Bont, H. B. A.; van Boom, J. H.; Liskamp, R. M. J. (1990) Tetrahedron Lett., 31, 2497-2500.

Barford, D.; Das, A. K.; Egloff, M. (1998) Annu. Rev. Biophys. Biomol. Struct., 27, 133-164.

[75]

Jia, Z. (1997) Biochem. Cell Biol., 75, 17-26.

[50] [51]

Kitas, E. A.; Perich, J. W.; Wade, J. D.; Johns, R. B.; Tregear, G. W. (1989) Tetrahedron Letters, 30, 6229-6232.

[76]

Hall, L. R.; Streuli, M.; Schlossmann, S. F.; Saito, H. (1988) J. Immunol., 141, 2781-2787.

[52]

Otvos, L.; Elekes, I.; Lee, V. M.-Y. (1989) Int. J. Pept. Prot. Res., 34, 129-133.

[77]

[53]

Sakaguchi, K.; Roller, P. P.; Appells, E. (1996) Genetic Engineering, 18, 249-278.

Levy, J. B.; Canoll, P. D.; Silvennoinen, O.; Barnea, G.; Morse, B.; Honegger, A. M.; Huang, J. T.; Cannizzaro, L. A.; Park, S. H.; Druck, T. (1993) J. Biol. Chem., 268, 10573-10581.

[54]

Wakamiya, T.; Saruta, K.; Yasuoka, J.; Kusumoto, S. (1994) Chem. Lett., 1099-1102.

[78]

Symons, A.; Willis, A. C.; Barclay, A. N. (1999) Protein Eng., 12, 885-892.

[55]

Lam, K. S.; Salmon, S. E.; Hersh, E. M.; Hruby, V. J.; Kazmierski, W. M.; Knapp, R. J. (1991) Nature, 354, 8284.

[79]

Cowburn, D. (1995) Structure, 3, 429-430.

[80]

Feng, S. C.; James, K.; Yu, Hongtao; Simon, Julian A.; Schreiber, Stuart L. (1994) Science, 266, 1241-1247.

[56]

Smith, G. P.; Petrenko, V. A. (1997) Chem. Rev., 97, 391410.

[81]

Kuriyan, J.; Cowburn, D. (1993) Curr. Opin. Struct. Biol., 3, 828-837.

[57]

Houghten, R. A.; Pinilla, C.; Blondelle, S. E.; Appel, J. R.; Dooley, C. T.; Cuervo, J. H. (1991) Nature, 354, 84-86.

[82]

Margolis, B.; Borg, J.-P.; Straight, S.; Meyer, D. (1999) Kidney Int., 56, 1230-1237.

[58]

Tonks, N. K.; Diltz, C. D.; Fischer, E. H. (1988) J. Biol. Chem., 263, 6722-6730.

[83]

[59]

Tonks, N. K.; Diltz, C. D.; Fischer, E. H. (1988) J. Biol. Chem., 263, 6731-6737.

Dhe-Pagnon, S.; Ottinger, E. A.; Nolte, R. T.; Eck, M. J.; Shoelson, S. E. (1999) Proc. Natl. Acad. Sci. USA, 96, 8378-8383.

[84]

Li, S.-C.; Lai, K.-M. V.; Gish, G. D.; Parris, W. E.; van der Geer, P.; Forman-Kay, J.; Pawson, T. (1996) J. Biol. Chem., 271, 31855-31862.

[60]

Charbonneau, H.; Tonks, N. K. (1992) Annu. Rev. Cell. Biol., 8, 463-493.

396

[85]


Li, S.-C.; Zwahlen, C.; Vincent, S. J. F.; McGlade, C. J.; Kay, L. E.; Pawson, T.; Forman-Kay, J. D. (1998) Nat. Struct. Biol., 5, 1075-1083.

Vetter and Zhang

[108] Scott, J. K. (1992) Trends Biochem. Sci., 17, 241-245. [109] Pini, A.; Viti, F.; Santucci, A.; Carnemolla, B.; Zardi, L.; Neri, P.; Neri, D. (1998) J. Biol. Chem., 34, 21769-21776.

[86]

Cowburn, D. (1997) Curr. Opin. Struct. Biol., 7, 835-838.

[87]

Forman-Kay, J. D.; Pawson, T. (1999) Curr. Opin. Struct. Biol., 9, 690-695.

[88]

van der Geer, P.; Wiley, S.; Gish, G. D.; Lai, V. K.-M.; Stephens, R.; White, M. F.; Kaplan, D.; Pawson, T. (1996) Proc. Natl. Acad. Sci. USA, 93, 963-968.

[89]

Yajnik, V.; Blaikie, P.; Bork, P.; Margolis, B. (1996) J. Biol. Chem., 271, 1813-1816.

[90]

Zhou, Y.; Abagyan, R. (1998) Folding & Design, 3, 513522.

[91]

Lu, P.-J.; Zhou, X. Z.; Shen, M.; Lu, K. P. (1999) Science, 283, 1325-1328.

[92]

Thompson, L. A.; Ellmann, J. A. (1996) Chem. Rev., 96, 555-600.

[116] Arendt, A.; Palczewski, K.; Moore, W. T.; Caprioli, R. M.; McDonwell, J. H.; Hargrave, P. A. ( 1989) Int. J. Pept. Prot. Res., 33, 468-476.

[93]

Service, R. F. (1997) Science, 277, 474-475.

[117] Perich, J. W. (1991) Methods Enzymol., 201, 234-245.

[94]

Han, H.; Wolfe, M. M.; Brenner, S.; Janda, K. D. (1995) Proc. Natl. Acad. Sci. USA, 92, 6412-6423.

[118] Paquet, A.; Johns, M. (1990) Int. J. Pept. Prot. Res., 36, 97103.

[95]

Gravert, D. J.; Janda, K. D. (1997) Chem. Rev., 97, 489509.

[119] Lacombe, J. M.; Andriamanampiosa, F.; Pavia, A. A. (1990) Int. J. Pept. Prot. Res., 36, 275-280.

[96]

Fields, G. B. (1990) In Solid Phase Synthesis, Epton, R. Ed.; SPCC Ltd. Birmingham, pp. 241-260.

[120] Perich, J. W.; Johns, R. B. (1988) Tetrahedron Lett., 29, 2369-2372.

[97]

Furka, A.; Sevenstyen, F.; Asgedom, M.; Dibo, G. (1991) Int. J. Prot. Pept. Res., 37, 487-493.

[121] Bannwarth, W.; Trzeciak, A. (1987) Helv. Chim. Acta, 70, 175-186.

[98]

Lam, K. S.; Lebl, M.; Krchnak, V. (1997) Chem. Rev., 97, 411-448.

[122] Andrews, D. M.; Kitchin, J.; Seale, P.W. (1991) Int. J. Peptide Protein Res., 38, 469-475.

[99]

Boutin, J. A.; Gesson, I.; Henlin, J.M.; Bertin, S.; Lambert, P. H.; Volland, J. P.; Fauchere, J. L. (1997) Molecular Diversity, 3, 43-60.

[123] Valerio, R. M.; Alewood, P. F.; Johns, R. B.; Kemp, B. E. (1989) Int. J. Pept. Prot. Res., 33, 428-438.

[100] Youngquist, R. S.; Fuentes, G. R.; Lacey, M. P.; Kenough, T. (1995) J. Am. Chem. Soc., 117, 3900-3906. [101] Ohlmeyer, M. H. J.; Swanson, R. N.; Dillard, L. W.; Reader, J. C.; Asouline, G.; Kobayashi, R.; Wigler, M.; Still, W. C. (1993) Proc. Natl. Acad. Sci. USA, 90, 1092210926. [102] Nielsen, J.; Brenner, S.; Janda, K. D. (1993) J. Am. Chem. Soc., 115, 9812-9813. [103] Geysen, H. M.; Wagner, C. D.; BOdnar, W. M.; Markworth, C. J.; Parke, G. J.; Schoenen, F. J.; Wagner, D. S.; Kinder, D. S. (1996) Chem. Biol., 3, 679-688. [104] Armstrong, R. W.; Tempest, P. A.; Cargill, J. F. (1996) Chimia, 50, 258-260. [105] Nicolaou, K. C.; Xiao, X. Y.; Parandoosh, Z.; Senyei, A.; Parke, G. J.; Nova, M. P. (1995) Angew. Chem. Int. Ed. Engl., 34, 2289-2291. [106] Vetter, S. W.; Keng, Y.-F.; Lawrence, D. S.; Zhang, Z.-Y. (2000) J. Biol. Chem., 275, 2265-2268. [107] Smith, G. P. (1991) Curr. Opin. Biotechnol., 2, 668-673.

[110] Hoess, R. H. (2001) Chem. Rev., 101, 3205-3218. [111] Kemp, B. E.; Bylund, D. B.; Huang, T.-S.; Krebs, E. G. (1975) Proc. Nat. Acad. Sci. USA, 72, 3448-3452. [112] Kemp, B. E.; Pearson, R. B. (1991) Methods Enzymol., 200, 121-134. [113] Graves, D. J. (1983) Methods Enzymol., 99, 268-278. [114] Perich, J. W.; Valerio, R. M.; Johns, R. B. (1986) Tetrahedron Lett., 27, 1377-1380. [115] Perich, J. W.; Johns, R. B. (1989) J. Org. Chem., 54, 17501752.

[124] Bodansky, M.; Martinez, J. (1981) Synthesis, 333-356. [125] Tian, Z.; Gu, C.; Roeske, R. W.; Zhou, M.; van Etten, R. L. (1993) Int. J. Pept. Prot. Res., 42, 155-158. [126] Burke, T. R.; Russ, P.; Lim, B. (1991) Synthesis, 10191020. [127] Burke, T. R.; Smyth, M. S.; Otaka, A.; Roller, P. P. (1993) Tetrahedron Lett., 34, 4125-4128. [128] Chen, L.; Wu, L.; Otaka, A.; Smyth, M. S.; Roller, P. P.; Burke, T. R.; den Hertog, J.; Zhang, Z.-Y. (1995) Biochem. Biophys. Res. Commun., 216, 976-984. [129] de Bonte, D. B. A.; Moree, J. W.; van Boom, J. H.; Liskamp, J. (1993) J. Org. Chem., 58, 1309-1317. [130] Ye, B.; Akamatsu, M.; Shoelson, S. E.; Wolf, G.; GiorgettiPeraldi, S.; Yan, X.; Roller, P. P.; Burke, T. R. (1995) J. Med. Chem., 38, 4270-4275. [131] Ye, B.; Burke, T. J. (1995) Tetrahedron Lett., 37, 47334736. [132] Burke, T. R. J.; Yao, Z.-J.; Zhao, H.; Milne, G. W. A.; Wu, L.; Zhang, Z.-Y.; Voigt, J. H. (1998) Tetrahedron, 54, 9981-9994.


[133] Songyang, Z.; Shoelson, S. E.; McGlade, J.; Olivier, P.; Pawson, T.; Bustelo, X. R.; Barbacid, M.; Sabe, H.; Hanafusa, B.; Yi, T.; Ren, R.; Baltimore, D.; Ratnofsky, S.; Feldman, R. A.; Cantley, L. C. (1994) Mol. Cell. Biol., 14, 2777-2785. [134] Müller, K.; Gombert, F. O.; Manning, U.; Grossmuller, F.; Graff, P.; Zaegel, H.; Zuber, J. F.; Freuler, F.; Tschopp, C.; Baumann, G. (1996) J. Biol. Chem., 271, 2. [135] Lee, T. R.; Lawrence, D. S. (1999) J. Med. Chem., 42, 784787. [136] Lee, T. R.; Lawrence, D. S. (2000) J. Med. Chem., 43, 1173-1179. [137] Gram, H.; Schmitz, R.; Zuber, J. F.; Baumann, G. (1997) Eur. J. Biochem., 246, 633-637. [138] Schmitz, R.; Baumann, G.; Gram, H. (1996) J. Mol. Biol., 260, 664-677. [139] Pellegrini, M. C.; Liang, H.; Mandiyan, S.; Wang, K.; Yuryev, A.; Vlattas, I.; Sytwu, T.; Li, Y.-C.; Wennogle, L. P. (1998) Biochemistry, 37, 15598-15606. [140] Huyer, G.; Kelly, J.; Moffat, J.; Zamboni, R.; Jia, Z.; Gresser, M. J.; Ramachandra, C. (1998) Analy. Biochem., 258, 19-30. [141] Zhang, Z.-Y.; Maclean, D.; McNamara, D. J.; Sawyer, T. K.; Dixon, J. E. (1994) Biochemistry, 33, 2285-2290. [142] Cheung, Y. W.; Abell, C.; Balasubramanian, S. (1997) J. Am. Chem. Soc., 119, 9568-9569. [143] Gay, B.; Suarez, S.; Caravatti, G.; Furet, P.; Meyer, T.; Schoerpfer, J. (1999) Int. J. Cancer, 83, 235-241. [144] Elchebly, M.; Payette, P.; Michaliszyn, E.; Cromlish, W.; Collins, S.; Lee Loy, A.; Normandin, D.; Cheng, A.; Himms-Hagen, J.; Chan, C.-C.; Ramachandran, C.; Gresser, M. J.; Tremblay, M. L.; Kennedy, B. P. (1999) Science, 283, 1544-1548. [145] Oligino, L.; Lung, F.-D.; Sastry, L.; Bigelow, J.; Cao, T.; Curran, M.; Burke, T. R.; Wang, S.; Krag, D.; Roller, P. P.; King, C. R. (1997) J. Biol. Chem., 272, 29046-29052.


397

[146] Wards, C. W.; Gough, K. H.; Sahke, M.; Wan, S. S.; Tribbick, G.; Wang, J.-X. (1996) J. Biol. Chem., 271, 56035609. [147] Rahuel, J.; Gay, B.; Erdmann, D.; Strauss, A.; GarciaEcheverria, C.; Furet, P.; Caravatti, G.; Fretz, H.; Schoepfer, J.; Gruetter, M. G. (1996) Nature Struct. Biol., 3, 586-589. [148] Furet, P.; Gay, B.; Garcia-Echeverria, C.; Rahuel, J.; Fretz, H.; Schoepfer, J.; Caravatti, G. (1997) J. Med. Chem., 40, 3551-3556. [149] Garcia-Echeverria, C.; Furet, P.; Gay, B.; Fretz, H.; Rahuel, J.; Schoepfer, J.; Caravatti, G. (1998) J. Med. Chem., 41, 1741-1744. [150] Furet, P.; Gay, B.; Caravatti, G.; Garcia-Echeverria, C.; Rahuel, J.; Schoepfer, J.; Fretz, H. (1998) J. Med. Chem., 41, 3442-3449. [151] Furet, P.; Garcia-Echeverria, C.; Gay, B.; Schoepfer, J.; Zeller, J.; Rahuel, J. (1999) J. Med. Chem., 42, 2358-2363. [152] Caravatti, G.; Rahuel, J.; Gay, B.; Furet, P. (1999) Bioorg. Med. Chem. Lett., 9, 2915-2920. [153] Schoepfer, J.; Fretz, H.; Gay, B.; Furet, P.; GarciaEcheverria, C.; End, N.; Caravatti, G. (1999) Bioorg. Med. Chem. Lett., 9, 221-226. [154] Caravatti, G.; Rahuel, J.; Gay, B.; Furet, P. (1999) Bioorg. Med. Chem. Lett., 9, 1973-1978. [155] Zhang, Z.-Y.; Thieme-Sefler, A. M.; MacLean, D.; McMamara, D. J.; Dobrusin, E. M.; Sawyer, T. K.; Dixon, J. E. (1993) Proc. Natl. Acad. Sci. USA, 90, 4446-4450. [156] Jia, Z.; Barford, D.; Flint, A. J.; Tonks, N. K. (1995) Science, 268, 1754-1758. [157] Sarmiento, M.; Puius, Y. A.; Vetter, S. W.; Keng, Y.-F.; Wu, L.; Zhao, Y.; Lawrence, D. S.; Almo, S. C.; Zhang, Z.Y. (2000) Biochemistry, 39, 8171-8179.