Identification of natural ligands for SH2 domains from a phage display cDNA library1

doi:10.1006/jmbi.2000.3561 available online at http://www.idealibrary.com on

J. Mol. Biol. (2000) 297, 89±97

Identification of Natural Ligands for SH2 Domains from a Phage Display cDNA Library D. Cochrane, C. Webster, G. Masih and J. McCafferty* Cambridge Antibody Technology, The Science Park Melbourn, Cambridgeshire SG8 6JJ, UK

The cytoplasmic domain of the Fc gamma receptor IIB (FcgRIIB) can be successfully displayed on the surface of ®lamentous phage, and after phosphorylation in vitro, can interact speci®cally with the SH2 domains of SHP-2, a cytoplasmic tyrosine phosphatase. When full-length FcgRIIB is expressed on phage, however, this interaction is greatly compromised, illustrating that characteristics of the full-length sequence are not well tolerated by the phage display system. Many associations in cell physiology are driven by similar interactions involving small modular binding domains or ligands, and so a fragmented cDNA library will facilitate display of such domains free of sequences which compromise their expression. A fragmented leukocyte cDNA display library of 108 clones was constructed. This library was phosphorylated in vitro with fyn kinase and was selected against the tandem SH2 domains of SHP-2 in the search for additional ligands. A depletion strategy to remove non-speci®c clones was employed, using SHP-2 Sepharose, prior to in vitro phosphorylation and selection. This permitted the emergence of clones encoding the cytoplasmic domain of PECAM-1, another natural ligand for SHP-2. The importance of dual phosphorylation of tyrosine residues at positions 663 and 686 was con®rmed in competition ELISA experiments using phosphorylated phage and synthetic peptides. Thus, phage display of fragmented cDNA libraries permits the identi®cation and characterisation of phosphorylated ligands of modular binding domains based on their functional interaction. # 2000 Academic Press

*Corresponding author

Keywords: SHP-2 phosphatase; modular domain; PECAM-1, selection

Introduction Many eukaryotic proteins possess small modular binding domains which drive intramolecular and intermolecular protein:protein interactions. The SH3 domain is an example of a modular binding domain present on many different signalling and cytoskeletal proteins. The natural ligands often have a proline-rich sequence, and the binding interaction with the SH3 domain (e.g. abl, src, crk and grb2) can be mimicked by short synthetic pepPresent address: D. Cochrane, Actinova Ltd., Babraham Institute, Babraham, Cambridgeshire CB2 4AT, UK. Abbreviations used: CIP, calf intestinal phosphatase; ITIM, immunoreceptor tyrosine-based inhibitory motif; PECAN, platelet endothelial cell adhesion molecule; BSA, bovine serum albumin; Gst, glutathione S-transferase. E-mail address of the corresponding author: [email protected] 0022-2836/00/010089±9 $35.00/0

tides containing this sequence (reviewed by Pawson, 1995; Cohen et al., 1995). The ability of short peptides to bind speci®cally to their partner is a property shared by other modular domains, such as the WW and EH domains. This property has permitted analysis of the sequence speci®city of the interactions using synthetic peptide libraries or peptide libraries displayed on phage (Sparks et al., 1996a; Linn et al., 1997; Paoluzi et al., 1998). The binding of SH2 domains to their natural phosphorylated partners can also be mimicked by synthetic peptides (Songyang et al., 1993). The interactions of the Grb SH2 domain and the PTB domain of Shc have been mapped using peptide libraries displayed on phage (Dente et al., 1997; Gram et al., 1997). In this case, selection of the library is preceded by phosphorylation in vitro. The immunoreceptor tyrosine-based inhibitory motif (ITIM) is an SH2 binding motif identi®ed on a number of cellular receptors associated with inhibitory signalling within the immune system. The ITIM-containing Fc gamma receptor IIB # 2000 Academic Press

90 (FcgRIIB) is expressed on lymphoid and myeloid cells and is responsible for negative regulation in B cells, T cells and mast cells following ligation with the B cell receptor, the T cell receptor and the Fc epsilon receptor, respectively. This is effected by a mechanism that involves interaction of the SH2 domains of SHP-2 with the phosphorylated FcgRIIB ITIM motif (reviewed by Vely & Vivier, 1997; Vivier & Daeron, 1997). SHP-2 is a cytoplasmic tyrosine phosphatase with two tandem SH2 domains followed by a catalytic domain. SHP2 is widely expressed and is involved in signalling via cytokines and growth factors. In the absence of a phosphorylated ligand, the N-terminal SH2 domain interacts with and inhibits the catalytic site. In the presence of ligand a conformational change releases this inhibition and activates the enzyme (Hof et al., 1998). Display of ligands for modular binding domains on phage potentially provides an opportunity to identify directly binding partners from cDNA libraries. Despite the extensive use of phage display in the generation of antibodies and peptides of desired speci®city, there has been a much more limited utilisation in display and cloning of cDNA molecules. Variation in display level of different genes on phage undoubtedly contributes to this. Here, we demonstrate display of the cytoplasmic domain of the FcgRIIB on phage and its kinasedependent interaction with the tandem SH2 domains of SHP-2. This only occurs when expressed on phage in a truncated form. Based on this, a fragmented cDNA display library was constructed. Here, we searched for additional binding partners for the SH2 domains of SHP-2. We demonstrate that additional natural binding partners for SHP-2 can be identi®ed using a fragmented leukocyte cDNA library displayed on phage. The in vitro nature of the phage system permits the facile study of these interactions, as illustrated in competition ELISA using phage and differentially phosphorylated peptides.

Results Comparison of fragmented and full-length receptor displayed on phage We wished to determine if it was possible to display functional cDNA ligands for modular binding domains on phage. Initial experiments were based around the expression of the low-af®nity IgG receptor, FcgRIIB, which contains an ITIM motif. Upon phosphorylation, this binds to the SH2 domains of cytoplasmic phosphatases such as SHP-2. In one display construct (fcg-FL), the fulllength version of the receptor was by fusing to the phage coat protein encoded by gene 3. In the other (fcg-cyto), only the cytoplasmic region was displayed (Figure 1). Phage derived from vector (pCANTAB-6) without insert were used as a control.

Ligands for SH2 Domains from Phage Display

Figure 1. Structure of the FcgRIIB2 gene. The domain structure of the FcgRIIB gene is shown. Ig represents the immunoglobulin-like domains, TM represents the transmembrane region and cyto represents the cytoplasmic domain. The sequence shown begins immediately downstream of the transmembrane domain and continues to the last amino acid. There are different splice forms of this gene and the cytoplasmic sequence of the form labelled FcgRIIB2 is shown from amino acid position 203-246. This is the sequence present in the phage display construct, Fcg-cyto. The tyrosine residue at amino acid position 228, involved in the interaction with SHP-2, is shown in bold.

Following rescue of the various phage clones, it was found that the titre of fcg-FL phage was tenfold less than the titre for the fcg-cyto phage. Certain gene products can have an adverse effect on the growth of Escherichia coli, causing slower growth or cell lysis, and this may account for the lower phage titre from the fcg-FL. In fact, there was more extraneous E. coli derived material from the fcg-FL-derived culture following PEG precipitation, suggesting that some degree of E. coli lysis was occurring during cell growth. (PEG precipitation, which is used to concentrate phage after rescue, also precipitates additional E. coli-derived material.) A lower titre would represent a distinct disadvantage for the full-length clone in the context of a library, and would probably be re¯ected by an under-representation of this clone in the library preparation. Phage were phosphorylated in vitro using an excess of fyn kinase. It has previously been shown that excess enzyme and/or extended incubations reduces the sequence context speci®city of tyrosine kinases, including fyn kinase, and allows all available tyrosine residues to be phosophorylated (Dente et al., 1997). Con®rmation of phosphorylation was carried out using an anti-phosphotyrosine monoclonal antibody. Phosphorylation was also observed in the control phage, which lacked an ITIM motif. Examination of the sequences of the coat protein revealed that there are 18 tyrosine residues in the minor coat protein encoded by gene 3 (gpIII), and two tyrosine residues in the major coat protein encoded by gene 8 (gpVIII). Western blot analysis of in vitro phosphorylated phage subsequently showed that phosphorylation of tyrosine residues in gpIII does occur, con®rming this reduced speci®city of phosphorylation under excess enzyme conditions (data not shown). Phosphorylation was not observed in gpVIII, probably due to inaccessibility of the buried tyrosine residues (Terry et al., 1997).

91

Ligands for SH2 Domains from Phage Display

The tandem SH2 domains of SHP-2 expressed as a Gst fusion were coated on to ELISA wells. Binding of the various constructs after phosphorylation was determined by ELISA. The results show a kinase-dependent binding of the fcg-cyto phage with SHP-2 (Figure 2). The signal from Fcg-FL phage compared to that of Fcg-cyto phage was much lower. In fact, the signal achieved by Fcg-FL phage was the same as negative control phage, indicating no signi®cant display of the ITIM motif. Thus, ELISA results demonstrate that functional proteins derived from fragmented genes can be displayed on phage and can interact with their partner protein domains. Furthermore, it is clear that the ef®ciency of display is affected by surrounding sequences. In vitro phosphorylation and selection of a fragmented leukocyte cDNA display library The above experiment suggested that display using a fragmented cDNA library, rather than a full-length library, might have a better chance of success in the search for modular domains and their ligands. cDNA from a leukocyte library was fragmented using DNase I to give an average size of 200-500 bp. cDNA-derived peptides ranging from 66-166 amino acid residues would be expected and this would cover the size range of many modular domains such as the WW, SH3 and SH2 domains and their ligands. A library of 108 clones, with 90 % recombinants was constructed. Non-fragmented cDNA libraries in the region of 105-106 clones are considered to give adequate representation of the mRNA population. Our aim was to construct a library 100 to 1000-fold larger in size. This increased library size was required for a number of reasons. The cDNA

Figure 2. Interaction of SHP-2 with the phosphorylated FcgRIIB ITIM domain displayed on phage. Phage were rescued from the clones Fcg-FL (displaying the full-length FcgRIIB gene) and Fcg-cyto (displaying the cytoplasmic domain of FcgRIIB). Phage derived from pCANTAB-6 were used as a negative control. ELISA was carried out to determine the degree of binding of these phage to the SH2 domains of SHP-2 with or without in vitro phosphorylation by fyn kinase. 1010 phage particles were present in each sample.

fragments were ligated into the vector in a nondirectional format meaning that there is a 1:2 chance of the cDNA fragment inserting in the correct sense orientation. There is a 1:3 chance of inserting in frame with the signal sequence and another 1:3 chance of remaining in frame with gene 3. This means that a cDNA fragment has only a 1:18 chance of being correctly inserted in the vector. Finally, the genes have been digested into a number of fragments, so more correctly inserted clones are required to represent fully each gene. One of the advantages of a display system is that screening on plates is not required and so this larger required library size does not create any additional screening burden. Phage from the fragmented library were phosphorylated and selected against a glutathione S-transferase (Gst) fusion of the tandem SH2 domains of SHP-2 (Gst-SHP-2). A total of 192 clones from three rounds of selection were phosphorylated and screened by ELISA against GstSHP-2, Gst and bovine serum albumin (BSA). After the initial ELISA screen, 31 clones were re-screened either in a phosphorylated or non-phosphorylated state against Gst-SHP-2 and BSA. Some of these clones were subsequently found to bind control plates, while others bound SHP-2 in preference to control plates, but this interaction was not dependent on phosphorylation. All of the 31 clones were sequenced. Of these, two-thirds (21/31) were made up of 16 independent clones, all encoding the poly(A) tail region of various mRNA molecules. Since the codon for lysine is AAA, a string of lysine residues will be incorporated into the protein, regardless of reading frame. Poly-L-lysine is often used as a tissue adhesive in immunocytochemistry. This ``stickiness'' probably explains its selection in this experiment. The other ten clones from the 31 sequenced were found to consist of two separate clones. Closer examination revealed that these clones consisted of concatamers of the primers used in the initial construction of the full-length cDNA library. The two clones consisted of ®ve copies of the primer resulting in ®ve repeats of the amino acid sequence LPAYSALLPL. Although there is a tyrosine residue in the middle of the sequence, binding of this clone does not appear to be kinase-dependent. This sequence may therefore have inherent low binding af®nity for SHP-2 without phosphorylation. The fact that this sequence is repeated may give an avidity-like component to the interaction, and explain why these clones were selected. Depletion of non-specific clones from the cDNA display library It has previously been shown that excess antigen immobilised on Sepharose provides an ef®cient means of removing ``binding'' phage from a library (Jespers et al., 1997). This is probably related to the large surface area to which the library is exposed in this format. We con®rmed that Sepharose was

92 effective in the removal of phage displaying the polylysine-containing peptides described above (data not shown). To remove the sticky clones from the library and to remove any non-kinasedependent clones binding to SHP-2, library phage stock were depleted of such clones prior to phosphorylation using Gst-SHP-2 Sepharose (Figure 3). The aim was that sticky and non-kinase-dependent SHP-2 binders should remain on the Sepharose matrix. This approach is therefore used as a way of ``purging'' the library of such clones before positive selections are carried out. Following deselection, the library was phosphorylated, and three rounds of selection were carried out as before. From an initial screen of 192 clones from the third round of selection, 18 clones were re-screened for kinase-dependent binding. Of the 18 clones rescreened, ®ve clones showed speci®c binding to SHP-2 in a kinase-dependent manner. These ®ve clones were subsequently sequenced and found to be multiple isolates of the same clone. This positive clone was identical with residues 617-711 of the intracellular C terminus of human platelet endothelial cell adhesion molecule (PECAM-1 or CD31; see Figure 4). Sequence analysis of the clone (PECAM-cyto) reveals that it includes an amber stop codon (TAG) as well as a short part of the 30 untranslated region of this gene. The E. coli host strain TG1 has a supE amber suppressor genotype, and so will incorporate a glutamine residue when the TAG codon is encountered to create a fusion with the gene 3 coat protein. The clone also encodes an additional six amino acids from the 30 untranslated region, as well as an additional 14 residues encoded by primers used in the library construction. Thus, despite the fact that a stop codon is present, an in-frame fusion of the PECAM cytoplasmic domain with gene 3 is created. The clone contains two phosphotyrosine molecules which may impart extra speci®city by interacting with the tandem pair of SH2 domains on SHP2 (Ottinger et al., 1998). Synthetic 36-mer peptides representing the region underlined in Figure 4 were synthesised with either no phosphotyrosine molecule (No pTyr), a single phosphotyrosine molecule at either position 663 or 686 (pTyr-663 and pTyr-686, respectively) or one at both positions (pTyr-663, 686). The relative ability of each peptide to inhibit the interaction between PECAM-cyto phage and SHP2 was determined in competition ELISA. Figure 5 illustrates that phosphorylation of tyrosine at positions 663 and 686 is required for maximal inhibition of this interaction. Inhibition is also achieved with a peptide incorporating a single phosphotyrosine at position 663, but to a lesser extent than seen with the bi-phosphorylated peptide. No signi®cant inhibition is observed with non-phosphorylated peptide or peptide phosphorylated at position 686. Thus this system illustrates a prime role for tyrosine 663 in the interaction of SHP-2 and PECAM, but demonstrates a requirement for phosphoryl-

Ligands for SH2 Domains from Phage Display

Figure 3. Depletion of non-speci®c cDNA clones from the library. A Direct selection on SHP-2. In the standard selection, phage are phosphorylated before adding to the plates coated with immobilised SH2 domains (phosphate groups are represented as ``p''). Three types of binding can be envisioned and are depicted: a kinasedependent binding of ligands in the cDNA library to the SH2 domains; b non-speci®c binding of ``sticky'' clones to the matrix; c kinase-independent binding to the immobilised protein. B Depletion on SHP-2 Sepharose prior to selection. Interactions of types b and c dominated the ®rst attempts at selection on the SH2 domains. To circumvent this, the library was depleted of such clones by passing over Sepharose loaded with the SH2 domains prior to phosphorylation. Clones where binding is dependent on phosphorylation are recovered in the ¯ow-through. The recovered population is recovered, phosphorylated and selected on the SH2 domains as before.

Ligands for SH2 Domains from Phage Display

93

Figure 4. Structure of the PECAM-1 gene. The domain structure of the PECAM-1 gene is shown. Ig represents the immunoglobulin-like domains, TM represents the transmembrane region and cyto represents the cytoplasmic domain of the gene. The sequence shown is is the cytoplasmic domain of human PECAM-1 beginning immediately downstream of the transmembrane domain. The sequence in upper case represents the sequence present within the clone PECAM-cyto derived from the library. The asterisk represents the end of the PECAM gene, i.e. the position of the amber stop codon. In the E. coli strain used for display, this encodes a glutamine residue and the remaining sequence (TARPDA) is encoded by a short segment of the 30 untranslated region of the gene. The tyrosine residues involved in the interaction with SHP-2 are shown in bold. Peptides representing the underlined region were used for competition studies.

ation at both positions for maximal binding. This cytoplasmic region of PECAM-1 has previously been recognised as a binding partner for SHP-2. The importance of tyrosine phosphorylation at positions 663 and 686 has also been shown in transfected cells expressing mutant PECAM genes (Jackson et al., 1997).

Discussion Phage display has proven to be a very powerful method for isolating genes encoding peptides, antibodies and other proteins with desired binding properties. While parallels have been made between the phage system and the yeast twohybrid system in the past (Allen et al., 1995) the reality is that there are few examples of identi®cation of natural binding pairs with the phage system. The yeast two-hybrid system on the other

hand has made a signi®cant impact on the discovery of many physiological binding partners. The two-hybrid system relies on heterologous protein:protein interactions in vivo to restore the activity of a trans-activator, which causes expression of a reporter gene. The main requirement of the twohybrid system is that both partners can reach and function in the nucleus. In the phage system there is a requirement that one of the partners is transportable to the periplasm and remains functional when displayed on a phage. While a number of different types of protein have been displayed on phage, there is great variation in expression levels of different proteins, and there is a selective advantage for those proteins which are well-tolerated in this expression system. This sequence-dependent variation in expression may, in part, explain the paucity of successful examples of cDNA cloning via phage display. To minimise the effect of

Figure 5. Inhibition of PECAM-cyto binding with phosphorylated peptides. The 36-mer peptides representing the underlined region of Figure 4 were synthesised and allowed to compete with phosphorylated phage for SHP-2 binding. Peptides either have no phosphotyrosine molecule (No pTyr), phosphotyrosine at position 663 (pTyr-663), phosphotyrosine at position 686 (pTyr-686) or phosphotyrosine residues at position 663 and 686 (pTyr-663, 686). Control phage represents PECAM-cyto phage without phosphorylation.

94 adverse sequences on the ability to display cDNA on phage, we adopted an approach whereby the cDNA is deliberately fragmented. It is clear that many genes have small self-contained modular domains or ligands which are not affected by gene fragmentation outside their coding region. It follows that functional binding domains can be separated from potentially problematic sequences in the same gene. Phage displaying cDNA fragments were phosphorylated in vitro using an excess of fyn kinase. It is possible that the speci®city of the kinase used for phosphorylation could in¯uence the outcome of library selection depending on substrate speci®city of the enzyme. According to Dente et al. (1997), however, under conditions of excess enzyme and extended reaction times, ``virtually all phage with a tyrosine can be phosphorylated irrespective of amino acid content.`` This reduced speci®city is borne out here, where phosphorylation was not only observed with phage displaying full-length and truncated FcgRIIB, but was also observed equally with phage derived from the vector pCANTAB6. Western blot analyses con®rm that coat proteins are indeed phosphorylated. Phosphorylation of phage was con®rmed on blots using an anti-phosphotyrosine antibody (data not shown). We demonstrated the validity of the cDNA fragmentation approach by comparing expression and binding of full-length FcgRIIB with expression from the ITIM containing the C-terminal fragment of the same gene. While the fragment displayed on phage was shown to interact with the SH2 domains of SHP-2, binding of the phage expressing the full-length receptor was barely detectable above the background. Since phosphorylation of full-length FcgRIIB phage has been con®rmed, it is likely that low expression of the receptor on phage is the problem, and in particular it is likely that the transmembrane domain of the full-length receptor contributes to its poor expression on phage. In this study we have pulled out multiple isolates of a clone encoding the C terminus of platelet endothelial cell adhesion molecule-1 (PECAM-1). There are a number of lines of evidence that identify SHP-2 as a natural binding partner of PECAM (Masuda et al., 1997; Jackson et al., 1997; Hua et al., 1998). The isolated clone incudes a pair of tyrosine residues proposed to be involved in the interaction with SHP-2. Thus using the cDNA display approach described here, ligands can be identi®ed on the basis of ``functional'' homology, rather than sequence homology. This is important in situations where there is poor sequence conservation but some structural/functional conservation between domains. This principle is well illustrated by the recent discovery of a distant member of the SH2 family which was identi®ed, not on the basis of sequence homology, but by structural homology (Meng et al., 1999). The SH2 binding domains identi®ed here would have been dif®cult to spot as physiological ligands for SHP-2 purely on the basis

Ligands for SH2 Domains from Phage Display

of sequence homology. This study has focussed on the SH2 domain but is equally applicable to the isolation of cDNA partners for other phosphopeptide recognition domains (e.g. PDB domains) as well as non-phosphotyrosine binding domains (e.g. WW, EH and SH3 domains). As with the Fcg-cyto construct described above, the isolated clone (PECAM-cyto) represents the C-terminal domain of the gene. With fragmented libraries there is a potential for loss of clones, particularly when the ligand site is present on short cytoplasmic domains. It is likely that transmembrane regions will cause problems in phage display and so successfully displayed cDNA fragments of intracellular domains are likely to begin downstream of the transmembrane region. In situations where the cytoplasmic domain is short, there is an increased likelihood of including the stop codon when using cDNA inserts with an average size of 200 ± 500 bp. In the clone identi®ed here, an amber stop codon (TAG) and part of the 30 untranslated region is included. In the supE suppressor strain used, the amber codon encodes a glutamine residue and so a fusion with gene 3 is created. In cases where stop codons other than TAG are used, such clones will be lost. Crameri & Suter (1993) describe a phage cDNA expression system which utilises the association between fos and jun leucine zipper and overcomes problems associated with stop codons in cDNA molecules. The jun zipper is expressed as a fusion with gene 3. The protein encoded by the cDNA is produced as a fos fusion and associates with the phage particle which encodes the gene, via the fos:jun interaction. Thus the association of the cDNAderived protein, with its encoding phage particle, occurs even if a stop codon is present. This feature also removes the requirement for cloning in frame with a downstream fusion partner (as in the case of direct gene 3 fusions). A combination of this indirect display system with the fragmented cDNA approach described here will have advantages in the search for ligands present in short C-terminal regions close to transmembrane domains. There are a number of potential methods to recognise interactions involving phosphorylated ligands. Using a modi®cation of the two-hybrid, Osbourne et al. (1995) identi®ed a novel gene containing an SH2 domain. Their ``yeast tribid'' system relies on the co-expression of a tyrosine kinase, an SH2 domain and a tyrosine-containing substrate. Several in vitro systems are available to identify binding partners. Zozulya et al. (1999), described selection of binding domains displayed on T7 bacteriophage using phosphotyrosine-containing peptide ligands. Sparks et al. (1996b) described a method for using ligand mimics derived from peptide display libraries to clone modular binding domains, by probing bacterial libraries cloned into lgt11 expression libraries (cloning of ligand targets, COLT). Thus with COLT, the phage system is used indirectly to identify novel cDNA molecules. Selection of phos-

95

Ligands for SH2 Domains from Phage Display

phorylated ligands or their binding domains have not yet been described. An advantage of in vitro systems is the facility to manipulate the library (e.g. phosphorylation in vitro) and the selection process (e.g. in excluding certain undesired or dominating interactions by competition or depletion (Parsons et al., 1996; Jespers et al., 1997; and see Results). In addition, binding partners isolated in these systems are immediately amenable to af®nity and speci®city studies. This is illustrated in this study using synthetic peptides to identify and characterise residues important in the interaction of PECAM and SHP-2. In a similar way, mutagenesis of the resultant phage clones could be used to carry out mapping of interactions or to generate higher af®nity variants for biochemical and cellular studies. The application of phage display has been utilised to great effect in study of antibody:antigen interactions and in the isolation of peptides, antibodies and other proteins with desired binding activities. Approaches described here will facilitate a greater use of phage display in the identi®cation of interactions between modular binding domains and their ligands directly from cDNA libraries. The in vitro nature of the phage system makes the resultant clones immediately amenable to sequence:function studies.

Materials and Methods Preparation of leukocyte cDNA Leukocytes were obtained from Ficoll gradient centrifugation of blood derived from two healthy donors. mRNA from the leukocytes was prepared using Quick prep Micro mRNA puri®cation (Pharmacia). Full-length cDNA from 10 mg of mRNA was obtained using the SMART PCR cDNA kit (Clontech). Display of the full-length and fragmented Fcgg RIIB gene The full-length receptor FcRgIIb and the cytoplasmic region of the receptor were cloned into the phage display vector pCANTAB-6 (McCafferty et al., 1994). Three primers were designed. Primer 1 hybridises to the 50 end of the full-length gene, primer 2 hybridises to the 50 end of the cytoplasmic region of the receptor gene and primer 3 hybridises to the 30 end of the receptor gene. Primers 1 and 2 contain S®I restriction enzyme sites, and primer 3 contains a NotI restriction enzyme site for directional cloning into the vector pCANTAB-6: primer 1, GCT ACC GCG GCC CAG CCG GCC ATG GCC GCT CCC CCA AAG GCT GTC CTG; primer 2, GCT ACC GCG GCC CAG CCG GCC ATG GCC AGG AAA AAG CGG ATT TCA GCC; primer 3, GAA TAG TCT AGA TGC GGC CGC AAT ACG GTT CTG GTC ATC AGG. Primers 1 and 3 were used in PCR to amplify the full-length gene, and primers 2 and 3 were used to amplify the cytosolic region of the receptor. PCR reactions were carried out in 50 ml of reaction mix using the leukocyte cDNA referred to above. PCR products were digested and ligated using a commercial ligation system (Amersham International) into S®I/NotI-digested pCANTAB-6. Following transform-

ation into E. coli strain TG1, bacteria were plated out on 2  TY agar supplemented with 2 % (w/v) glucose and 100 mg/ml ampicillin (TYAG) and grown overnight at 30  C. Single colonies were picked and their sequences con®rmed. Preparation of fragmented leukocyte cDNA display library Fragmentation of the leukocyte cDNA described above was carried out by digestion with DNase I (Sigma). Digestion of the cDNA was carried out at 30  C for 30 minutes. A total of 0.02 unit of DNase I was used in 100 ml of reaction mixture containing 10 mg of leukocyte cDNA in 50 mM Tris-HCl (pH 7.4), 1 mM MnCl2, 100 mg/ml BSA. At the end of 30 minutes the reaction was stopped by the addition of EDTA (5 mM ®nal concentration), and cleaned up by phenol/chloroformextraction and ethanol-precipitation. Fragments of cDNA ranging from 200-500 bp were excised from agarose gels and puri®ed using Wizard Clean Up resin (Promega). The cDNA fragments were then treated with phage T4 DNA polymerase to repair the ends, and were puri®ed with Wizard PCR Clean Up (Promega). NotI - XmnI non-palindromic adapters (New England Biologicals) were ligated using T4 DNA ligase (New England Biologicals) for incorporation into the NotI site of the vector pCANTAB-6. Before insert ligation, the pCANTAB-6 vector was treated with CIP (calf intestinal phosphatase; New England Biologicals) to prevent vector re-ligation. Ligation of fragmented cDNA inserts into CIP-treated pCANTAB6 was then carried out using a commercial ligation system (Amersham). Ligated DNA was transformed into E. coli strain TG1 and plated on to TYAG agar, at 30  C. Resultant library colonies were scraped into 40 ml of TYAG and 15 % (v/v) glycerol for storage at ÿ70  C as a library stock. In vitro phosphorylation of fragmented cDNA Phage particles were rescued as described by McCafferty & Johnson (1996). Phosphorylation of cDNA expressed on the surface of the phage was carried out using fyn kinase (Upstate Biotechnology) as described by Dente et al. (1997). Phosphorylation was carried out using 6.6 units of fyn kinase at 37  C for six hours in a 100 ml reaction mix which consisted of 1  1011 phage in 50 mM Tris (pH 7.4), 50 mM MgCl, 0.5 mM NaVO4, 3 mM ATP. After three hours, fresh fyn kinase and ATP were added as before. The phosphorylation reaction was stopped by incubation at room temperature for 30 minutes with the kinase inhibitor PP2 (Calbiochem) at a ®nal concentration of 600 nM. Effective phosphorylation was con®rmed by dot blotting phage and detecting with an anti-phosphotyrosine mAb (PY99, Santa Cruz). Selection of library on SHP-2 Maxisorb plates (Nunc) were coated overnight at 4  C with 10mg/ml of a Gst fusion containing the two tandem SH2 domains of SHP-2 (Santa Cruz). After overnight incubation, wells were washed three times with PBS and blocked at 37  C for two hours with 3 % (w/v) skimmed milk protein (Marvel) in phosphate-buffered saline (PBS/M). Phosphorylated library phage stocks were preincubated in PBS/M for 30 minutes prior to selection at room temperature. After blocking, wells were washed three times with PBS and the library phage was added

96 and incubated for two hours at room temperature. Wells were then washed 20 times with PBS and bound phage eluted with 100 mM triethylamine. Phage eluate was neutralised with an equal volume of 1 M Tris HCl (pH 7.4) and infected into TG1 cells. Resultant colonies were scraped and phage rescued for further rounds of phosphorylation and selections. ELISA screening After three rounds of selections, single clones were screened in ELISA for SHP-2 binding. Individual phage preparations were phosphorylated as stated above and blocked with PBS/M for 30 minutes at room temperature prior to being used. Clones were also screened against Gst and BSA, to check their binding speci®city. ELISA plates were coated overnight at 4  C with either Gst-SHP-2, Gst or BSA at 10 mg/ml in PBS. After overnight incubation, wells were washed three times with PBS and then blocked at 37  C for two hours with PBS/ M. After blocking, ELISA plates were washed three times with PBS and the phosphorylated phage preparations were added and incubated for one hour at room temperature. For the inhibition experiment, phosphorylated phage and peptide at the indicated concentration were mixed before addition to the ELISA wells. Wells were then washed three times with PBS, 0.1 % Tween 20, and then three times with PBS. Phage binding was detected by anti-M13-HRP (Pharmacia) diluted 1:5000 in PBS/M, and incubated for one hour at room temperature. Wells were then washed three times in PBS, 0.1 % Tween 20 and three times with PBS. TMB was added for colorimetric detection. The reaction was stopped with a half-volume of 0.5 M H2SO4 and the plates were read at a wavelength of 450 nm. Deselection of non-specific clones from library Bacterial clones carrying the SH2 domains of SHP-2 fused to Gst were induced with 1 mM IPTG for four hours and the expressed protein extracted. Gst Sepharose was mixed with the bacterial extract and washed according to the manufacturer's instructions to prepare the SHP-2 Sepharose matrix used for depletion. A 50 ml packed volume of SHP-2 Sepharose was added to 100 ml of library phage and incubated for two hours at room temperature. After incubation, the non-bound phage were recovered and the process repeated. The ®nal deselected phage population was phosphorylated and used for the ®rst round selections. Peptide synthesis Peptides were made by solid phase synthesis on an Advanced ChemTech 396 Biomolecular synthesiser using Fmoc chemistry. Peptides were characterised using a Micromass Platform LC-MS system. Peptides were separated using a C8 column and analysed by electrospray mass spectrometry.

References Allen, J. B., Walberg, M. W., Edwards, M. C. & Elledge, S. J. (1995). Finding prospective partners in the library: the two-hybrid system and phage display ®nd a match. Trends Biol. Sci. 20, 511-516.

Ligands for SH2 Domains from Phage Display Cohen, G. B., Ren, R. & Baltimore, D. (1995). Modular binding domains in signal transduction proteins. Cell, 80, 237-248. Crameri, R. & Suter, M. (1993). Display of biologically active proteins on the surface of ®lamentous phages: a cDNA cloning system for selection of functional gene products linked to the genetic information responsible for their production. Gene, 137, 69-75. Crameri, R. & Suter, M. (1995). Display of biologically active proteins on the surface of ®lamentous phages: a cDNA cloning system for selection of functional gene products linked to the genetic information responsible for their production: addendum. Gene, 160, 139. Dente, L., Vetriai, C., Zucconi, A., Pelicci, G., Lanfrancone, L., Pelicci, P. G. & Cesareni, G. (1997). Modi®ed phage peptide libraries as a tool to study speci®city of phosphorylation and recognition of tyrosine containing peptides. J. Mol. Biol. 269, 694703. Gram, H., Schmitz, R., Zuber, J. F. & Baumann, G. (1997). Identi®cation of phosphopeptide ligands for the Src-homology 2 (SH2) domain of Grb2 by phage display. Eur. J. Biochem. 246, 633-637. Hof, P., Pluskey, S., Dhe-Paganon, S., Eck, M. J. & Shoelson, S. E. (1998). Crystal structure of the tyrosine phosphatase SHP-2. Cell, 92, 441-450. Hua, C. T., Gamble, J. R., Vadas, M. A. & Jackson, D. E. (1998). Recruitment and activation of SHP-1 protein tyrosine phosphatase by human platelet endothelial cell adhesion molecule-1 (PECAM-1): identi®cation of immunotyrosine based inhibitory motif like binding motifs and substrates. J. Biol. Chem. 273, 2833228340. Jackson, D. E., Kupcho, K. R. & Newman, P. J. (1997). Characterization of phosphotyrosine binding motifs in the cytoplasmic domain of platelet/endothelial cell adhesion molecule-1 (PECAM-1) that are required for the cellular association and activation of the protein-tyrosine phosphatase, SHP-2. J. Biol. Chem. 272, 24868-24875. Jespers, L., Jenne, S., Lasters, I. & Collen, D. (1997). Epitope mapping by negative selection of randomized antigen libraries displayed on ®lamentous phage. J. Mol. Biol. 269, 704-718. Linn, H., Ermekova, K. S., Rentschler, S., Sparks, A. B., Kay, B. K. & Sudol, M. (19??). Using molecular repertoires to identify high-af®nity peptide ligands of the WW domain of human and mouse YAP. Biol. Chem. 378, 531-537. Paoluzi, P., Castagnoli, L., Coda, I. L., Fre, S., Confalonieri, S., Pelicci, P. G., Di Fiore, P. P. & Cesareni, G. (1998). Recognition speci®city of individual EH domains of mammals and yeast. EMBO J. 17, 6541-6550. Parsons, H. L., Earnshaw, J. C., Wilton, J., Johnson, K. S., Schueler, P. A., Mahoney, W. & McCafferty, J. (1996). Directing phage selection towards speci®c epitopes. Protein Eng. 9, 1043-1049. Masuda, M., Osawa, M., Shigematsu, H., Harada, N. & Fujiwara, K. (1997). Platelet endothelial cell adhesion molecule-1 is a major SH-PTP2 binding protein in vascular endothelial cells. FEBS Letters, 408, 331-336. McCafferty, J. & Johnson, K. S. (1996). Construction and screening of antibody display libraries. In Phage Display of Peptides and Proteins (Kay, B. K., Winter, J. &

Ligands for SH2 Domains from Phage Display McCafferty, J., eds), pp. 79-111, Academic Press, London. McCafferty, J., Fitzgerald, K. J., Earnshaw, J., Chiswell, D. J., Link, J., Smith, R. & Kenten, J. (1994). Selection and rapid puri®cation of murine antibody fragments that bind a transition state analogue by phage display. Apll. Biochem. Biotechnol. 47, 157-173. Meng, W., Sawasdikosol, S., Burakoff, S. J. & Eck, M. J. (1999). Structure of the amino-terminal domain of Cbl complexed to its binding site on ZAP-70 kinase. Nature, 398, 84-89. Osbourne, M. A., Dalton, S. & Kochan, J. P. (1995). The yeast tribid system-genetic detection of transphosphorylated ITAM-SH2 interactions. Biotechnology, 13, 1474-1478. Ottinger, E. A., Bot®eld, M. C. & Shoelson, S. E. (1998). Tandem SH2 domains confer high speci®city in tyrosine kinase signaling. J. Biol. Chem. 273, 729-735. Pawson, T. (1995). Protein modules and signalling networks. Nature, 373, 573-580. Songyang, Z., Shoelson, S. E., Chaudhuri, M., Gish, G., Pawson, T., Haser, W. G., King, F., Roberts, T., Ratnofsky, S., Lechleider, R. J., Neel, B. G., Birge, R. B., Fajardo, J. E., Chou, M. M., Hanafusa, H., Schaffhausen, B. & Cantley, L. C. (1993). SH2

97 domains recognize speci®c phosphopeptide sequences. Cell, 72, 767-778. Sparks, A. B., Rider, J. E., Hoffman, N. G., Fowlkes, D. M., Quilliam, L. A. & Kay, B. K. (1999). Distinct ligand preferences of Src homology 3 domains from Src, Yes Abl, Cortactin, p53bp2, PLCg, Crk, and Grb2. Proc. Natl Acad. Sci, USA, 93, 1540-1544. Sparks, A. B., Hoffman, N. G., McConnell, S. J., Fowlkes, D. M. & Kay, B. K. (1996b). Cloning of ligand targets: Systematic isolation of SH3 domaincontaining proteins. Nature Biotechnol. 14, 741-744. Terry, T. D., Malik, P. & Perham, R. N. (1997). Accessibility of peptides displayed on ®lamentous bacteriophage virion: susceptibility to proteinases. Biol. Chem. 378, 523-530. Vely, F. & Vivier, E. (1997). Conservation of structural features reveals the existence of a large family of inhibitory cell surface receptors and non-inhibitory/ activatory counterparts. J. Immunol. 159, 2075-2077. Vivier, E. & Daeron, M. (1997). Immunoreceptor tyrosine-based inhibition motifs. Immunol. Today, 18, 286-291. Zozuulya, S., Lioubin, M., Hill, R. J., Abram, C. & Gishizy, M. L. (1999). Mapping signal transduction pathways by phage display. Nature Biotechnol. 17, 1193-1198.

Edited by J. A. Wells (Received 21 September 1999; received in revised form 20 January 2000; accepted 27 January 2000)