Characterization of peptides that bind the tumor-associated Thomsen-Friedenreich antigen selected from bacteriophage display libraries1

J. Mol. Biol. (1997) 270, 374±384

Characterization of Peptides that Bind the Tumor-Associated Thomsen-Friedenreich Antigen Selected from Bacteriophage Display Libraries Elena N. Peletskaya1, Vladislav V. Glinsky2, Gennadi V. Glinsky1,2 Susan L. Deutscher1 and Thomas P. Quinn1* 1

Department of Biochemistry University of Missouri Columbia, MO 65211, USA 2

Cancer Research Center 350 Berrywood Drive Columbia, MO 65201, USA

Peptides with high af®nities and speci®cities for numerous proteins and nucleic acids have been previously identi®ed from random peptide bacteriophage display libraries. Here, random peptide bacteriophage display libraries were used to identify sequences that bound the cancer-associated Thomsen-Friedenreich glycoantigen (T antigen). The T antigen, present on most malignant cells, contains an immunodominant Galb1 ! 3GalNAca disaccharide unmasked on the surfaces of most carcinomas. This antigen has been postulated to be involved in tumor cell aggregation and metastasis. Two 15 amino acid random peptide bacteriophage display libraries were af®nity selected with glycoproteins displaying T antigen on their surfaces. Sequence analysis revealed that many of the peptides shared homology with sugar recognition sites in several carbohydrate-binding proteins. A comparison of af®nity selected sequences from both libraries yielded a common motif (W-Y-A-W/F-S-P) rich in aromatic amino acids. Four peptides, corresponding to the af®nity selected sequences, were chemically synthesized and characterized for their carbohydrate recognition properties. The synthetic peptides exhibited high speci®cities and af®nities to T antigen displayed on asialofetuin or conjugated to bovine serum albumin (Kd ˆ 5 nM for MAP-P30 binding to asialofetuin) as well as free T-antigen disaccharide in solution (Kd ˆ 10 mM for MAP-P30, 20 mM for P10). Two peptides, P30 and P10, demonstrated high af®nities and speci®cities for both asialofetuin and T antigen in solution. Iodination of a lone tyrosine residue in each sequence dramatically reduced their abilities to bind T antigen, suggesting that the tyrosine residue plays an important role in carbohydrate recognition. That these peptides are of functional signi®cance is evidenced by the ability of both P30 and P10 to inhibit asialofetuin-mediated melanoma cell aggregation in vitro and to compete with peanut lectin for binding to T antigen displayed on the surface of MDA-MB-435 breast carcinoma cells in situ. # 1997 Academic Press Limited

*Corresponding author

Keywords: phage display peptide libraries; T antigen; synthetic peptides; tumor cell binding; aggregation inhibition

Introduction

Abbreviations used: MAP, multiple antigenic peptide; FITC, ¯uorescein isothiocyanate; BSA, bovine serum albumin; T, Thomsen-Friedenreich; PBS, phosphate buffered saline (pH 7.4); TPBS, 0.5% Tween-20 (v/v) in phosphate buffered saline (pH 7.4); PNA, peanut lectin; HRP, horseradish peroxidase; GlcNAc, Nacetylglucosamine; DAB, diaminobenzidine. 0022±2836/97/280374±11 $25.00/0/mb971107

Peptides and their analogs have been successfully used in a number of important medical applications, including roles as regulators of blood pressure (Maack, 1992), cancer cell proliferation (Kvols et al., 1986), and viral replication (Collier et al., 1996). Until recently, the use of peptides and peptide analogs as diagnostic or therapeutic agents was limited to a small set of sequences that bound well characterized receptors or targets, such as # 1997 Academic Press Limited

375

T Antigen Binding Peptides

atrial natriuretic peptide (Maack, 1992), somatostatin (Bakker et al., 1991), melanotropin (Hruby et al., 1993), or HIV protease (Leonard, 1996). This limitation has been overcome by the advent of bacteriophage display library technology (Cwirla et al., 1990; Devlin et al., 1990; Parmley & Smith, 1988; Scott & Smith, 1990) and synthetic peptide libraries (Furka et al., 1991; Houghten et al., 1991; Lam et al., 1991). Random peptide libraries can be effectively screened for sequences that bind a particular antigen or receptor and are particularly useful in discovering sequences that bind molecules with poorly de®ned or understood molecular recognition properties. Combinatorial approaches have been successfully applied to the isolation of molecules that bind tumor-associated antigens with high af®nities. Random peptide libraries, either synthetic or phage borne, have yielded peptides with high af®nities for tumor antigens found on the surfaces of lymphomas (Lam et al., 1995), and colon carcinomas (Gui et al., 1996; Peletskaya et al., 1996). In addition, bacteriophage display technology has been employed to optimize binding af®nities of peptides (Li et al., 1995) for their cognate ligands. Since the number of targeted antigenic sites per cancer cell are limited, maximizing the binding af®nity of a peptide is crucial for ef®cient tumor localization. Conjugation of tumor-avid peptides and antibody fragments with radionuclides and toxin molecules have already yielded several potential anti-cancer therapeutic agents (Begent et al., 1996; Lamberts et al., 1990; Meredith et al., 1994). Cell surface carbohydrate structures are an important class of tumor antigens. Many of these molecules are believed to play crucial roles in cellcell interactions such as leukocyte recognition, cell adhesion and in¯ammatory response (Fukuda, 1994; Hakomori, 1991; Jalkanen et al., 1988; Osborn, 1990; Springer, 1995). Alterations in the composition of cell surface carbohydrates are often associated with malignant transformation (Hakomori, 1989; Springer, 1984). For example, the cancer-associated Thomsen-Friedenreich (T) antigen, a member of the blood group precursor antigens (Springer & Ansell, 1958), is expressed on the surface of most malignant tumors (Springer, 1984; Springer et al., 1995). The immunodominant portion of the T antigen is the terminal Galb1 ! 3GalNAca carbohydrate moiety (Desai & Springer, 1979; Springer & Desai, 1974). Although present on most tissues, T antigen is masked covalently or structurally and non-immunoreactive on the surfaces of healthy cells, and is exposed and immunoreactive on most human carcinomas and T cell lymphomas (Springer, 1984). The existence of T antigen-mediated cell adhesion between highly metastatic murine ESb lymphoma cells and hepatocytes is indicative of a role for this cell surface carbohydrate structure in metastatic processes (Springer et al., 1983). These properties make T antigen of paramount interest for the study of

carbohydrate-peptide interactions as well as an important therapeutic target. Here we describe the biochemical characterization of a set of polypeptides that bind to the glycoprotein asialofetuin and the carcinoma-associated T antigen. The peptide sequences were originally identi®ed from two 15 amino acid random peptide bacteriophage display libraries af®nity selected with glycoproteins that exhibited T antigen on their surfaces. The chemically synthesized polypeptides were shown to exhibit high speci®cities and af®nities for the glycoprotein asialofetuin (5 nM to 1 mM) and T antigen (10 to 200 mM). That the peptides selectively recognized and masked the carbohydrate structures on asialofetuin was supported by the abilities of two peptides (P30 and P10) to inhibit asialofetuin mediated melanoma cell aggregation. Asialofetuin induced melanoma cell aggregation resulted from interactions between galactose-speci®c melanoma cell surface lectins and the terminal galactose residues of T antigen structures present on asialofetuin (Meromsky et al., 1986). P30 and P10 were investigated for their abilities to bind carcinoma cells that were known to display T antigen structures on their surfaces. Biotinylated P30 and P10 were shown to bind the MDA-MB-435 breast carcinoma cells directly, while a biotinylated control peptide did not. Both peptides were also able to competitively inhibit binding of T antigen speci®c peanut lectin to the breast carcinoma cells, suggesting that the peptides recognized the carcinoma-associated T antigen epitope.

Results Isolation and analysis of T antigen binding sequences Bacteriophage display libraries were screened for peptides that would bind and mask the carcinomaassociated T antigen. Bacteriophage libraries were af®nity selected with glycoproteins that displayed the immunodominant portion of the T antigen (Galb1 ! 3GalNAca) T on their surfaces, asialofetuin (Nilsson et al., 1979; Spiro & Bhoyroo, 1974) and a bovine serum albumin-T antigen conjugate (BSA-T). Two types of bacteriophage libraries were employed during the af®nity screening process. Both libraries encoded random 15 amino acid sequences, but differed in the number and location of the peptides displayed on the surface of the virions. The f3-15 library displayed ®ve copies of a random 15 amino acid peptide fused to the amino terminus of the viral coat protein III, while the f8815 library displayed approximately 150 copies of a random 15 amino acid sequence fused to the amino terminus of the major viral coat protein VIII. Aliquots of both bacteriophage display libraries were subjected to multiple selection processes that differed in antigen concentrations, wash stringencies, and elution strategies (Peletskaya et al., 1996). Clonal phage stocks were prepared from the ®nal eluates of the selection procedures. The phage

376

T Antigen Binding Peptides

Table 1. Af®nity selected peptide sequences Name P30 P89 P6 P10

a

Peptide sequence

Library

% of clones

HGRFILPWWYAFSPS RFRGLISLSQVYLSP ARVSFWRYSSFAPTY GSWYAWSPLVPSAQI

f3-15 f3-15 f3-15 f88-15

12 5.5 4 90

a The percent of individual clones that were represented in the populations of 800 f3-15 and 100 f88-15 af®nity selected and sequenced phage clones.

DNA encoding the random peptide inserts from 800 f3-15 and 100 f88-15 virions were sequenced and grouped according to frequency of occurrence in the af®nity selected phage population. The most frequently occurring peptide sequences are listed in Table 1. A putative consensus sequence of W-Y-A-W/F-S-P was present in the two most frequently occurring clones, P30 and P10. FASTA (Pearson & Lipman, 1988) and BLAST (Altschul et al., 1990) sequence analysis programs were used to examine the peptides listed in Table 1 for stretches of homology to protein sequences present in the PIR (George et al., 1996), and SWISS-PROT (Bairoch & Apweiler, 1996). The af®nity selected peptides showed homology to portions of several proteins, many of which are involved in carbohydrate metabolism or transport. Since the threedimensional structures and locations of functional sites were not de®ned in many proteins exhibiting homologies to the peptide sequences, interpretation of the biological signi®cance of these matches was impossible. Therefore, the locations of the aligned peptide sequences were only examined in the corresponding proteins that had crystal structures deposited in the Brookhaven Protein Data Bank (Bernstein et al., 1977) and possessed de®ned active sites or carbohydrate binding sites. Sequence analysis yielded homologies between the af®nity selected peptides and several proteins involved in carbohydrate binding or metabolism (Table 2). None of the peptides constituted an exact match to one of the sugar binding sites, but in several cases they were homologous to stretches of sequence in

contact with the carbohydrate. For example, a portion of the peptide sequence P30 was homologous to amino acids 57 to 64 in a-amylase. The two Trp residues and the Tyr de®ne a portion of the carbohydrate binding site in a-amylase. The pyranose rings of the carbohydrate stack on the rings of Trp59 and Tyr62. A segment of lysozyme, homologous to peptide P10, makes van der Waals contacts with the (GlcNAc)3 oligosaccharide present in the active site. While these observations do not prove that the aromatic amino acids in P30 participate in carbohydrate binding, they do suggest that this arrangement of clustered aromatic residues can stabilize carbohydrate binding. Characterization of T antigen binding peptides The polypeptide sequences P30, P89, P10, (Table 1) and a 15 amino acid control peptide were chemically synthesized in order to characterize their asialofetuin and T antigen binding properties. A MAP version of the P30 peptide (MAP-P30) was synthesized, which contained four copies of the P30 peptide linked by their carboxyl termini through a polylysine tree (Tam, 1988). The binding af®nities of the P30, MAP-P30, P89, P10 and control peptide for asialofetuin were measured directly by titrating the peptides into solutions of FITClabeled asialofetuin and monitoring the resulting ¯uorescence quenching (Connors, 1987) (Table 3). The binding isotherms for P30 and the MAP-P30 construct are shown in Figure 1. The dissociation constants calculated for P30 and MAP-P30 binding to asialofetuin were 1.2(0.2) mM and 5(0.4) nM, respectively. MAP-P30 exhibited a dissociation constant for asialofetuin of 5(0.4) nM, which is approximately three orders of magnitude less than the P30 peptide alone. The increase in af®nity of MAP-P30 for asialofetuin is apparently due to the presence of multiple binding sites which are likely to slow the off rate. The dissociation constants for P89 and P10 asialofetuin complexes were determined to be 0.7(0.2) mM, and 0.1(0.2) mM, respectively (Table 3). MAP constructs of P10 and

Table 2. Results from sequence homology searches of T antigen peptides to carbohydrate binding proteins with known crystal structures P30 a-Amylase NAc-b-glucosaminyl arginine amidase

(1PNF)

P6 Glucoamylase

(3GLY)

ARVSFWRYSSFAPTY qSFWgsyilA

224±233

(1LMP) (1EDT)

GSWYAWSPLVPSAQI GaWvAW YAWnPyygtwQ

106±111 190±200

P10 Lysozyme 1,4-Endo-b-glucanase a

(1PPI)a

HGRFILPWWYAFSPS PWWerYqPb RFItPyW

57±64c 80±86

Brookhaven Protein Data Bank identi®er code. Capital letters stand for the direct matches in the sequence, conservative changes are italicized, and amino acids with no homology are in lower case. c Amino acid numbers were derived from the protein structures deposited in the Brookhaven Protein Data Bank. b

377

T Antigen Binding Peptides Table 3. Dissociation constants for synthetic peptide binding to selected glycoproteins and T antigen Asialofetuin-FITC Fetuin-FITC Transferrin-FITC BSA-FITC T-Antigen

P30

MAP-P30

P89

P10

Cntrlc

1.2  0.2 mMa 20  2 mM 20  1 mM n.b. 200  20 mM

5  0.4 nM 100  10 nM 100  10 nM 2  0.2 mM 10  1 mM

0.7  0.2 mM 0.7  0.2 mM 0.7  0.2 mM 0.7  0.2 mM n.b.

100  20 nM 1  0.3 mM n.b. n.b. 20  2 mM

n.b.b n.b. n.b. n.b. n.b.

a Dissociation constants were determined from binding curves generated from ¯uorescence quenching versus addition of peptide or T antigen. b n.b. stands for no detectable binding. c The control peptide (Cntrl) sequence was RNVPPIFNDVYWIAF.

P89 were not synthesized due to their limited solubility. The control peptide did not bind to the FITC-labeled asialofetuin. The abilities of the peptides to bind FITC labeled BSA and two glycoproteins fetuin and transferrin were also examined. Fetuin is identical to asialofetuin except that its carbohydrate structures contain terminally linked sialic acid moieties (Spiro & Bhoyroo, 1974). The carbohydrate structures on transferrin also contain sialic acid, terminally linked a(2-6) to galactose (Finne & Krusius, 1979). The presence of terminal sialic acid moieties on fetuin reduced P30 and MAP-P30 af®nity 20-fold. A similar reduction in peptide af®nity for transferrin was also demonstrated. P30 did not bind BSA, while MAP-P30 did display weak af®nity for BSA which was 500 times less than asialofetuin. The weak af®nity of MAP-P30 for BSA is probably non-speci®c and resulted from an increase in the molecule's hydrophobicity. P89 exhibited similar binding af®nities to FITC labeled asialofetuin, fetuin, transferrin and BSA indicating that it bound the proteins nonselectively. The P10 peptide exhibited a tenfold reduction in af®nity for fetuin compared to asialofetuin. P10 did not bind transferrin or BSA, demonstrating speci®city for asialofetuin. None of the peptides bound FITC (data not shown). These data demonstrate that peptides with

Figure 1. Binding pro®les of P30 ( & ), MAP-P30 (~), and a control peptide (*) to FITC labeled asialofetuin determined by ¯uorescence quench titrations at 25 C. Calculated Kd values are listed in Table 3.

high af®nities and selectivities for the glycoprotein asialofetuin were selected from the random peptide libraries. The random peptide libraries were screened with two different glycoproteins which displayed the T antigen carbohydrate structure in an attempt to identify peptides that selectively bound the disaccharide epitope. Fluorescence quench titration analysis was used to determine if P30, P89, and P10 were able to bind to the T antigen disaccharide. The abilities of peptides to bind other monoand disaccharides including, D-galactose, D-glucose, D-mannose, D-lactose, N-acetylgalactosamine, N-acetylglucosamine and N-acetylmannosamine were also examined. P30 and P10 exhibited speci®c and saturable binding to the T antigen disaccharide and not to any of the other mono- and disaccharides listed above. The dissociation constants for P30 and P10 to the T antigen disaccharide were calculated to be 200(20) mM, and 20 mM, respectively (Table 3). The MAP-P30 construct displayed the highest af®nity for T antigen with a dissociation constant of 10 mM (Table 3). These dissociation constants are lower than the published dissociation constant for T antigen binding PNA (Lotan et al., 1975), suggesting that the peptides may be able to inhibit some forms of lectin/carbohydrate mediated cell adhesion. P89 did not exhibit speci®c or saturable binding to the T antigen disaccharide or any of the other mono- and disaccharides listed above. P89 did not show speci®city for asialofetuin or af®nity for the T antigen disaccharide, indicating that the peptide lacked glycoprotein or carbohydrate speci®city. P89 was also one of the few peptides that did not contain a consensus sequence rich in aromatic residues like P30, P10, and P6. The lack of a consensus sequence rich in aromatic amino acids is also indicative of a different binding speci®city. The control peptide did not display speci®c or saturable binding to any of the mono- and disaccharides including T antigen. The binding data above suggest that the two peptides that contained a consensus sequence rich in aromatic residues, P30 and P10, were able to bind the T antigen disaccharide. To examine the possibility that one of these aromatic amino acids was involved in sugar recognition, the lone tyrosine residues in P30 and P10 were iodinated. The abilities of the iodinated peptides to bind T antigen were determined by ¯uorescence quench titrations.

378

T Antigen Binding Peptides

Figure 2. Effect of peptide iodination on T antigen binding. Binding isotherms of P30 (^), P10 (~), iodinated P30 (^) and iodinated P10 (!) with T antigen determined by ¯uorescence quench titration at 25 C.

The iodinated P30 and P10 peptides did not bind T antigen while their unmodi®ed counterparts did (Figure 2). These results suggested that the Tyr residues and possibly the other surrounding aromatic amino acids may play an important role in recognition of the T antigen epitope. Inhibition of tumor cell aggregation Cellular adhesion, an important step in metastasis, is often mediated by carbohydrate-lectin interactions (Hakomori, 1991; Meromsky et al., 1986; Springer et al., 1983). Studies suggest a positive correlation between the homotypic and heterotypic cell to cell adhesion propensity of tumor cells and their metastatic potential. B16-F1 murine melanoma cells do not display T antigen on their surfaces but do exhibit b-galactoside-speci®c cell surface lectins that bind T antigen (Lotan et al., 1985). Asialofetuin induces homotypic aggregation of murine B16 melanoma cells through interactions with the b-galactoside-speci®c cell surface lectins (Inohara & Raz, 1995; Meromsky et al., 1986). The ability of P30, P89 and P10 to inhibit asialofetuin-mediated melanoma cell aggregation was examined in vitro. Addition of asialofetuin to the murine melanoma cells greatly enhanced the number of cells present in aggregates, which could be inhibited by the addition of P30 and P10 but not P89 or the control peptide (Figure 3a). The presence of P30, P89, P10 or the control peptide alone did not effect the basal level of melanoma cell aggregation. The ability of P30 and P10 to inhibit asialofetuin induced melanoma cell aggregation was examined at various peptide concentrations (Figure 3b). Asialofetuin induced melanoma cell aggregation was inhibited in a dose dependent manner by the addition of P30 and P10, which was indicative of a speci®c inhibitory interaction. Inhibition of asialofetuin induced melanoma cell aggregation by P30 and P10

Figure 3. Cell aggregation assay. a, The percentage of melanoma cells in aggregates in samples containing 0.5  106 B16-F1 (B16) cells alone, with the addition of 0.1 mg/ml asialofetuin (AF), or with the addition of 0.1 mg/ml asialofetuin (AF) ‡ 0.15 mg/ml of the designated peptide (P). An irrelevant control peptide (Crtl) was also included. b, P30 and P10 inhibition of T antigen-induced B16-F1 melanoma cell aggregation. Various concentrations of P30 (^) and P10 ( & ) were titrated into samples containing 0.1 mg/ml asialofetuin (AF) and 0.5  106 B16-F1 cells. The sequence of the control peptide was RNVPPIFNDVYWIAF.

suggests that these peptides are masking the terminal galactose containing carbohydrate structures on asialofetuin. Human breast carcinoma cell binding The abilities of P30, P10 and MAP-P30 to bind the human breast carcinoma cell line MDA-MB-435 (Price, 1996; Zhang et al., 1991) were examined in situ. T antigen has been shown to be present on most breast carcinoma cells (Springer, 1984; Springer et al., 1980). The presence of T antigen on the surfaces of the MDA-MD-435 breast carcinoma cells was con®rmed by PNA-HRP lectin binding (Figure 4a). PNA-HRP did not bind murine melanoma B16-F1 cells which do not express T antigen (data not shown). Biotinylated P30 and P10 bound the breast carcinoma cells (Figure 4b and c) while the biotinylated control peptide did not bind the tumor cells (data not shown). The biotinylated MAP-P30 molecule was very insoluble under phys-

379

T Antigen Binding Peptides

Figure 4. Direct breast carcinoma cell binding assay. a, The presence of T antigen on the MDA-435 cells was con®rmed by PNA-HRP binding and subsequent development with DAB. Biotinylated P30 (b) and P10 (c) bound to the MDA-MB-435 breast carcinoma cells and were visualized by HRP-streptavidin reaction with DAB.

iological buffer conditions which prevented its use in the direct binding assay. To determine if the peptides bound the same carbohydrate antigen recognized by the PNA lectin, competition binding assays were performed with P30, P10, and MAPP30. The PNA-HRP conjugate was not competed off the MDA-MB-435 cells with the control peptide (Figure 5a), but could be competed off with P30, MAP-P30, and P10 (Figure 5b, c and d). These

results suggest that P30, P10 and MAP-P30 were able to recognize the T antigen glycoepitope displayed on the breast carcinoma cells.

Discussion Here, peptides that bound the cancer-associated T antigen were identi®ed from 15 amino acid random peptide display bacteriophage libraries. These

Figure 5. Competitive breast carcinoma cell binding assay. The T antigen speci®c PNA-HRP lectin complex was incubated with the MDA-MB-435 breast carcinoma cells in the presence of (a) an irrelevant control peptide, (b) P30, (c) MAP-P30 and (d) P10. PNA-HRP binding was visualized by development with DAB.

380 peptides were examined for their abilities to inhibit carbohydrate mediated cell adhesion and bind to T antigen in vitro and in vivo. The bacteriophage display libraries were screened by alternating rounds of selection with two glycoproteins, asialofetuin and a BSA-T conjugate, both of which display T antigen on their surfaces. Two T antigen containing glycoproteins were employed in the selection scheme to reduce the isolation of sequences that bound the proteins non-speci®cally. Peptide sequence data obtained from the two 15 amino acid peptide display libraries were analyzed for clones preferentially enriched by the af®nity maturation process. Both libraries underwent the same selection regiment. The most common sequences represented approximately 4 to 12% of the total f3-15 clones sequenced, while the most common sequence from the f88-15 library was present in 90% of clones. These sequence distributions were somewhat unexpected. The f88-15 library displays approximately 150 copies of a particular peptide in contrast to the ®ve copies displayed by the f3-15 library. It was anticipated that weaker binders from the f88-15 library might make it through the af®nity maturation process due to the high copy number magnifying af®nity. This was not the case, and suggested that the peptide's local environment, accessibility, and perhaps ¯exibility in¯uenced the biopanning outputs to a greater extent than just the number of peptides displayed per virion. Despite these differences, it was possible to develop the consensus sequence W-Y-A-W/F-S-P from the most common sequence isolated from each library. BLAST and FASTA sequence searches revealed that the af®nity selected peptide sequences had varying degrees of homologies to proteins involved in carbohydrate binding or metabolism. Where possible, the crystal structures of the protein-carbohydrate complexes were examined for the roles of individual amino acid residues in sugar binding. It was evident from the crystal structure analysis of a-amylase and lysozyme that stretches of sequence homologous to the af®nity selected peptides played important roles in the carbohydrate binding sites, usually by forming extensive van der Waals contacts with pyranose rings. Conversely, it has been demonstrated that lectins (Scott et al., 1992) and anti-carbohydrate antibodies (Valadon et al., 1996) exhibit speci®cities for peptides that are rich in aromatic residues. Together, these results con®rm the importance of aromatic ring interactions in carbohydrate recognition as well as illustrate the ability of aromatic amino acid side-chains to mimic pyranose rings. It appears that the common property shared in both instances is the ability of the ring constituents to stack and/or form extensive van der Waals contacts with each other facilitating molecular recognition and stabilizing binding. The peptides P30, P89, P10 and MAP-P30 were chemically synthesized and examined for their abilities to bind asialofeuin and T antigen. The dissociation constants determined for P10, P30, and

T Antigen Binding Peptides

P89 binding to asialofetuin were 0.1 mM, 1.2 mM, and 0.7 mM, respectively. The dissociation constants of P10 and P30 for the T antigen disaccharide were determined to be 20 mM and 0.2 mM, respectively. While P10 and P30 only displayed micromolar binding af®nities for T antigen disaccharide in solution, they were highly speci®c for T antigen and did not bind other mono- or disaccharides. The two peptides that bound both asialofetuin and the T antigen disaccharide contained stretches of aromatic amino acids which formed a potential sugar recognition site. Aromatic amino acids often play important roles in carbohydrate binding sites. The aromatic amino acid rich potential sugar binding sites in P30 and P10 contained a lone tyrosine residue. The tyrosine residues in both peptides were iodinated to determine if the modi®cation would alter the peptides' binding af®nities for T antigen. Neither iodinated P30 nor P10 were able to bind the T antigen disaccharide indicating that the tyrosine residues speci®cally and the conserved aromatic sequences in general are important for recognition of the carbohydrate epitope. The ability of peptides to selectively bind carbohydrate structures with af®nities comparable to lectins was supported by studies from Heerze et al. (1992). They reported that a 20 amino acid peptide corresponding to the carbohydrate binding site of pertussis toxin subunit 2 was able to bind sialic acid and inhibit the binding of the toxin to the glycoprotein fetuin. These results demonstrated that a peptide, corresponding to a natural carbohydrate binding site, was capable of binding its cognate sugar. In light of this report and the results presented, it seems that peptides which bind any number of speci®c carbohydrate structures, can be identi®ed from random peptide phage display libraries. In an attempt to increase the binding af®nities of the peptides even further, four P30 peptides were coupled together into a single molecule via a polylysine tree. The resulting MAP-P30 molecule exhibited dissociation constants for asialofetuin and the T antigen disaccharide of 5 nM and 10 mM, respectively. It appeared that the presence of multiple binding sites greatly enhanced the binding af®nity of the MAP-P30 construct for asialofetuin. The strategy of placing more than one binding site into a molecule to increase its binding af®nity is common in nature and exempli®ed by the bidentate structures of antibodies (Alzari et al., 1988) and the multiple sugar binding domains present in many lectins (Taylor et al., 1992). That the peptides were able to disrupt carbohydrate mediated cell adhesion was demonstrated by the ability of P10, P30 and MAP-P30 to inhibit asialofetuin induced melanoma cell aggregation. Melanoma cellular aggregation results from the interaction of the terminal galactose residue of carbohydrate structures on asialofetuin with a galactosespeci®c lectin on the surface of the melanoma cells. P89 did not bind the T antigen disaccharide or inhibit asialofetuin despite its ability to non-speci®-

381

T Antigen Binding Peptides

cally bind asialofetuin. This result suggested a fundamental difference in the molecular recognition properties of P89 compared with P30 and P10 and showed that any peptides which bound asialofetuin did not necessarily inhibit asialofetuinmediated cell aggregation. Titration of P10 and P30 into the asialofetuin-melanoma cell aggregation assay demonstrated that the peptides inhibited cellular aggregation in a dose dependent manner which is indicative of a speci®c and saturable binding interaction. These results are consistent with the peptides masking the carbohydrate moieties of the T antigen epitopes on asialofetuin and preventing them from being bound by the lectin molecules on the melanoma cells. Only the two peptides that exhibited binding af®nities for the T antigen disaccharide in solution (P10 and P30) were able to inhibit asialofetuin induced melanoma cell aggregation. Fluorescence binding studies revealed that P30 and P10 did not bind galactose or N-acetyl-galactose suggesting that the disaccharide structure of T antigen was required for binding. It appears that the T antigen speci®cities of the peptides were responsible for binding to asialofetuin, masking the carbohydrate structures and inhibiting lectin recognition and cell aggregation. T antigen is a common tumor-associated glycoepitope present on most breast carcinomas (Springer, 1984; Springer et al., 1980, 1995). The abilities of P30 and P10 to bind cell surface T antigen epitopes present on the human breast carcinoma cell line MDA-MB-435 was examined. Peptides P10 and P30 were shown to bind to breast carcinoma cells in situ and were able to competitively inhibit peanut lectin binding. Peanut lectin binds selectively to the carbohydrate moiety of T antigen. Inhibition of peanut lectin binding to the breast carcinoma cells suggested that the peptides recognized cell surface T glycoepitopes. While the biotinylated MAP-P30 peptide construct was too insoluble for direct binding analysis, it was able to compete off peanut lectin binding to the breast carcinoma cells. Since P30 and P10 appear to recognize and bind T antigen structures in situ, they may be useful as probes to study tumor cell aggregation inhibition or as vehicles for targeting radionuclides to tumors for diagnostic imaging or therapy.

3GalNAca-O-BSA), Peanut lectin-horseradish peroxidase conjugate and diaminobenzidine (DAB), ¯uorescein isothiocyanate (FITC) were purchased from Sigma Chemical Co (St. Louis, MO). The glycoproteins asialofetuin, fetuin, carboxypeptidase Y and transferrin were purchased from Boehringer Mannheim (Indianapolis, IN). Biotinylation of peptides and proteins was performed using the biotin-N-hydroxysuccinimide ester (Bethesda Research Laboratories). All other chemicals were purchased from Fisher Scienti®c (St. Louis, MO). Affinity selection The random 15 amino acid phage display libraries were af®nity selected against asialofetuin and BSA-T antigen conjugant. The af®nity selection procedure was based on the biopanning methods previously described by Smith & Scott (1993). In several independent selection procedures performed we used alternating rounds of selection for binding to immobilized biotinylated asialofetuin and BSA-T to ensure speci®city to T antigen (Peletskaya et al., 1996). Wells of the microtiter plates were ®rst coated with streptavidin, washed with TPBS (phosphate buffered saline, pH 7.4, 0.5% (v/v) Tween-20), and blocked with 3% (w/v) BSA prior to addition of biotinylated antigens. Phage were incubated with antigen prior to washing and elution. In the last two rounds of some procedures, phage were pre-incubated with the biotinylated antigens before the streptavidin capture. The ®nal elution step was performed with free T antigen (1 mM). Lactose (10 mM), and BSA (1mg/ml) were added to the wash steps to reduce non-speci®c binding. DNA and peptide sequence analysis Individual phage isolates were sequenced by a modi®ed dideoxy sequencing methodology (Haas & Smith, 1993), utilizing a 32P-labeled oligonucleotide primer located 15 nucleotides downstream of the pIII gene cloning site. Sequence analysis was performed by ALIGN from the PANALIST program package (Novosibirsk, Russia), FASTA (Pearson & Lipman, 1988) present in the GCG (Genetics Computer Group, Inc. Madison, Wisconsin) software package and BLAST (Altschul et al., 1990) program provided by Human Genome Center (Baylor College of Medicine). Selected X-ray models of the carbohydrate binding proteins with homologies to the peptide regions of interest available from the Brookhaven Protein Data Bank (Bernstein et al., 1977) were examined using SYBYL software package (TRIPOS, Inc. St. Louis, MO). Peptide synthesis and purification

Materials and Methods Phage libraries and bacterial strains The 15 amino acid random phage display library constructed in the fUSE 5 vector (Scott & Smith, 1990), the 15 amino acid random library with the insert in the p8 coat protein gene of the phage (f88), and the Escherichia coli strain K91BlueKan were generously provided by Dr George Smith (Univ. MO). Reagents T antigen (Galb1 ! 3GalNAc), asialofetuin, bovine serum albumin (BSA), BSA-T conjugate (Galb1 !

T antigen-binding peptides were chemically synthesized on the Applied Biosystems peptide synthesizer 431A using FMOC-based chemistry and puri®ed to homogeneity on a C-18 reverse-phase HPLC column (ISCO, Corp.). Synthesis of biotin-peptide and protein conjugants Peptides and asialofetuin, BSA-T and BSA were modi®ed with NHS-biotin. Peptide (0.5 mM) or protein (100 mM) were stirred in 0.1 M NaHCO3 buffer (pH 8.5) with 10 to 50-fold molar excess of NHS-biotin for one hour at 25 C. The reaction was stopped by the addition of 0.1 M NH4OH. Labeled peptides were puri®ed on a

382 RP-HPLC C18 column (ISCO), while labeled glycoproteins were puri®ed by Centricon ®ltration.

T Antigen Binding Peptides non-labeled peptides (50 mM) for one hour at 25 C then washed with PBS, stained with DAB, mounted and examined with a microscope.

Synthesis of fluorescent peptide and protein conjugants Peptides and asialofetuin were modi®ed with ¯uorescein-isothiocyanate (FITC). Peptide (0.5 mM) and FITC (5 mM) or asialofetuin (100 mM) and FITC (1 mM) was stirred in 0.1 M NaHCO3 buffer (pH 8.5) for one hour at 25 C. The reaction was stopped by the addition of 0.1 M NH4OH. Labeled peptides were puri®ed on a RP-HPLC C18 column (ISCO), while labeled glycoproteins were puri®ed on a G10 Sephadex column. Synthesis of the iodinated peptide derivatives Peptides were modi®ed with iodine using Iodogen (Pierce). Peptides were dissolved in 20 mM Na-PO4 buffer (pH 7.0) and stirred in the Iodogen-covered Eppendorf tube with threefold molar excess of KI for 20 minutes at room temperarure and transferred to a fresh tube to stop the reaction. I-labeled peptides were puri®ed on a RP-HPLC C18 column (ISCO). Fluorescence titrations Fluorescence measurements were performed on a SLM Aminco ¯uorimeter equipped with dual detectors. The ¯uorimeter was interfaced to a DEL 433/L PC running SLM Aminco 8100 series 2 software. Fluorescence intensity of FITC was monitored at 518 nm (excitation wavelength, 490 nm). Tryptophan ¯uorescence was monitored at 350 nm with excitation wavelength 280 nm for the unmodi®ed peptides and 290 nm for their iodinated analogs. Titrations were performed at 20 C in 20 mM NH4COOH (pH 7.0) or phosphate buffered saline (pH 7.4). Melanoma cell aggregation assay Asialofetuin induced B16-F1 melanoma cell aggregation assays were used to investigate the ability of P30 or P89 to inhibit tumor cell aggregation in vitro. The concentration of asialofetuin necessary to yield 50% of B16F1 cells in aggregates was determined to be 0.1 mg/ml. Inhibition of asialofeuin- induced cell aggregation assays were performed by incubating 0.5  106 B16-F1 cells with 0.1 mg/ml asialofetuin and 0.15 mg/ml peptide in PBS for two hours at 37 C. The experiments were performed in triplicate with three samples quanti®ed for each experiment according to the procedure decribed by Meromsky et al. (1986) Samples containing 0.5  106 B16F1 cells and 0.1 mg/ml asialofetuin were also titrated with various amounts of P30. Each point was reported as an average of three experiments. Binding of peptides to breast carcinoma cells The breast carcinoma cell line MB-435 was grown in the tissue culture ¯asks to approximately 80% con¯uence, trypsinized, washed with PBS and incubated in the 2% formaldehyde-PBS solution for 30 minutes at 25 C to ®x the cells before placing them onto the glass slides. The slides were washed with PBS, incubated with the biotin-labeled peptide followed by PBS wash and incubation with HRP-streptavidin (0.5 mM), HRP-PNA (5 mM) in PBS or the mixture of HRP-PNA (5 mM) and

Acknowledgements We thank Dr George Smith for the gift of the peptide libraries and he and his laboratory members for their helpful discussions. We also thank Dr I. Fider for the B16-F1 melanoma cell line and Dr Janet Price for the MDA-MB-435 breast carcinoma cell line. This research was supported by grants from the Department of Energy (DOE DE ER93-60661) to T.P.Q. and the National Institutes of Health (GM 47979) to S.L.D.

References Altschul, S. F., Gish, W., Miller, W., Myers, E. W. & Lipman, D. J. (1990). Basic local alignment search tool. J. Mol. Biol. 215, 403±410. Alzari, P. M., Lascombe, M. B. & Poljak, R. J. (1988). Three-dimensional structure of antibodies. Annu. Rev. Immunol. 6, 555± 580. Bairoch, A. & Apweiler, R. (1996). The SWISS-PROT protein sequence data bank and its new supplement TrEMBL. Nucl. Acids Res. 24, 21± 25. Bakker, W. H., Krenning, E. P., Reubi, J. C., Breeman, W. A. P., Setyono-Han, B., de Jong, M., Kooij, P. P. M., Bruns, C., van Hagen, P. M., Marbach, P., Visser, T. J., Pless, J. & Lamberts, S. W. J. (1991). In vivo application of [111In-DTPA-D-Phe]-octreotide for detection of somatostatin receptor-positive tumors in rats. Life Sci. 49, 1593± 1601. Begent, R. H., Verhaar, M. J., Chester, K. A., Casey, J. L., Green, A. J., Napier, M. P., Hope-Stone, L. D., Cushen, N. A. K. P., Johnson, C. J., Hawkins, R. E., Wilson, A. J. & Robson, L. (1996). Clinical evidence of ef®ecient tumor targeting based on single-chain Fv antibody selected from a combinatorial library. Nature Med. 2, 979± 984. Bernstein, F. C., Koetzle, T. F., Williams, G. J., Meyer, E. E., Brice, M. D., Rodgers, J. R., Kennard, O., Shimanouchi, T. & Tasumi, M. (1977). The protein data bank: a computer-based archival ®le for macromolecular structures. J. Mol. Biol. 112, 535 ± 542. Collier, A. C., Coombs, R. W., Bassett, D. A., Baruch, A. & Corey, L. (1996). Combination therapy with zidovudine, didanosine and saquinovir. Antiviral Res. 29, 99. Connors, K. A. (1987). Binding Constants: The Measurement of Molecular Complex Stability, Wiley, New York. Cwirla, S. E., Peters, B. A., Barrett, R. W. & Dower, W. J. (1990). Peptides on phage: a vast library of peptides for identifying ligands. Proc. Natl Acad. Sci. USA, 87, 6378± 6382. Desai, P. R. & Springer, G. F. (1979). Biosynthesis of human blood group T-, N- and M-speci®c immunodeterminants on human erythrocyte antigens. J. Immunogenet. 6(6), 403± 417. Devlin, J. J., Panganiban, L. C. & Devlin, P. E. (1990). Random peptide libraries: a source of speci®c protein binding molecules. Science, 249, 404± 406.

T Antigen Binding Peptides Finne, J. & Krusius, T. (1979). Structural features of the carbohydrate units of plasma glycoproteins. Eur. J. Biochem. 102, 583 ± 588. Fukuda, M. (1994). Cell Surface Carbohydrates: Cell-type Speci®c Expression. Frontiers in Molecular Glycobiology (Fukuda, M. & Hindsgaul, O., eds), vol. 1, Oxford University Press, Oxford. Furka, A., Sebestyen, F., Asgedom, M. & Dibo, G. (1991). General method for rapid synthesis of multicomponent peptide mixtures. Int. J. Peptide Res. 37, 487± 493. George, D. G., Barker, W. C., Mewes, H.-W., Pfeiffer, F. & Tsugita, A. (1996). The PIR international protein sequence database. Nucl. Acids Res. 24, 17 ± 20. Gui, J., Moyana, T. & Xiang, J. (1996). Selection of a peptide with af®nity for the tumor-associated TAG72 antigen from a phage-displayed library. Biochem. Biophys. Res. Commun. 218, 414± 419. Haas, S. J. & Smith, G. P. (1993). Rapid sequencing of viral DNA from ®lamentous bacteriophage. Biotechniques, 15(3), 422 ± 4, 426 ± 8, 431. Hakomori, S.-I. (1989). Abberant glycosylation in tumors and tumor-associated carbohydrate antigens. Advan. Cancer Res. 52, 257± 331. Hakomori, S.-I. (1991). Possible functions of tumor-associated carbohydrate antigens. Curr. Opin. Immunol. 3, 646±653. Heerze, L. D., Chong, P. C. S. & Armstrong, G. D. (1992). Investigation of the lectin-like binding domains in pertussis toxin using synthetic peptide sequences. J. Biol. Chem. 267, 25810± 25815. Houghten, R. A., Pinilla, C., Blondelle, S. E., Appel, J. R., Dooley, C. T. & Cuervo, J. H. (1991). Generation and use of synthetic peptide combinatorial libraries for basic research and drug discovery. Nature, 354, 84± 86. Hruby, V. J., Sharma, S. D., Toth, K., Jaw, J. Y., alObeidi, F., Sawyer, T. K. & Hadley, M. E. (1993). Design, synthesis, and conformation of superpotent and prolonged acting melanotropins. Ann. New York Acad. Sci. 680, 51± 63. Inohara, H. & Raz, A. (1995). Functional evidience that cell surface galectin-3 mediates homeotypic cell adhesion. Cancer. Res. 55, 3267± 3271. Jalkanen, S., Bargatze, R., Jalkanen, M., Lewinsohn, D., Streeter, P., Lakey, E. & Butcher, E. C. (1988). Lymphocyte migration molecules. Advan. Expt. Med. Biol. 237, 21± 29. Kvols, L. K., Moertel, C. G., O'Connell, M. J., Rubin, J. & Hahn, R. G. (1986). Treatment of malignant carcinoid syndrome: evaluation of a long-acting somatostatin analogue. N. Engl. J. Med. 315, 663 ± 666. Lam, K. S., Salmon, S. E., Hersh, E. M., Hruby, V. J., Kazmierski, W. M. & Knapp, R. J. (1991). A new type of synthetic peptide library for identifying ligand-binding activities. Nature, 354, 82 ± 84. Lam, K. S., Quiang, L., Zhao, Z.-G., Chen, M.-L., Pleshko, E. & Salmon, S. E. (1995). Idiotype speci®c peptides bind to the surface immunoglobulins of two murine B-cell lymphoma lines, inducing signal transduction. Biomed. Peptides Nucl. Acids, 1, 205± 210. Lamberts, S. W. J., Bakker, W. H., Reubi, J.-C. & Krenning, E. P. (1990). Somatostatin-receptor imaging in the localization of endocrine tumors. N. Engl. J. Med. 323, 1246± 1249. Leonard, K. M. (1996). Perspectives in HIV protease inhibitors. Advan. Expt. Med. Biol. 394, 319 ± 325.

383 Li, B., Tom, J. Y., Oare, D., Yen, R., Fairbrother, W. J., Wells, J. A. & Cunningham, B. C. (1995). Minimization of a polypeptide hormone. Science, 270(5242), 1657± 1660. Lotan, R., Skutelsky, E., Danon, D. & Sharon, N. (1975). The puri®cation, composition, and speci®city of the anti-T lectin from peanut (Arachis hypogaea). J. Biol. Chem. 250(21), 8518± 8523. Lotan, R., Lotan, D. & Raz, A. (1985). Inhibition of tumor cell colony formation in culture by a monoclonal antibody to endogenous lectins. Cancer. Res. 45, 4349± 4353. Maack, T. (1992). Receptors of atrial natriuretic factor. Annu. Rev. Physiol. 54, 11 ± 27. Meredith, R. F., Bueschen, A. J., Khazaeli, M. B., Plott, W. E., Grizzle, W. E., Wheeler, R. H., Schlom, J., Russell, C. D., Liu, T. & LoBuglio, A. F. (1994). Treatment of metastatic prostate carcinoma with radiolabeled antibody CC49. J. Nucl. Med. 35(6), 1017± 1022. Meromsky, L., Lotan, R. & Raz, A. (1986). Implications of endogenous tumor cell surface lectins as mediators of cellular interactions and lung colonizations. Cancer Res. 46, 5270± 5275. Nilsson, B., Norden, N. E. & Svensson, S. (1979). Structural studies on the carbohydrate portion of fetuin. J. Biol. Chem. 254, 4545±4553. Osborn, L. (1990). Leukocyte adhesion to endothelium in in¯amation. Cell, 62, 3 ± 6. Parmley, S. F. & Smith, G. P. (1988). Antibody-selectable ®lamentous fd phage vectors: af®nity puri®cation of target genes. Gene, 73(2), 305± 318. Pearson, W. R. & Lipman, D. J. (1988). Improved tools for biological sequence comparison. Proc. Natl Acad. Sci. USA, 85, 2444± 2448. Peletskaya, E. N., Glinsky, G. V., Deutscher, S. L. & Quinn, T. P. (1996). Identi®cation of peptide sequences that bind the Thomsen-Friedenreich cancer-associated glycoantigen from bacteriophage peptide display librariues. Mol. Diversity, 2, 13 ± 18. Price, J. E. (1996). Metastasis from human breast cancer cell lines. Breast Cancer Res. Treat. 39, 93± 102. Smith, G. P. & Scott, J. K. (1993). Libraries of peptides and proteins displayed on ®lamentous phage. Methods Enzymol. 217, 228±257. Scott, J. K. & Smith, G. P. (1990). Searching for peptide ligands with an epitope library. Science, 249(4967), 386± 390. Scott, J. K., Loganathan, D., Easely, R. B., Gong, X. & Goldstein, I. J. (1992). A family of concanavalin Abinding peptides from a hexapeptide epitope library. Proc. Natl Acad. Sci. 89, 5398± 5402. Spiro, R. G. & Bhoyroo, V. D. (1974). Structure of the Oglycosidically linked carbohydrate units of fetuin. J. Biol. Chem. 249, 5704±5717. Springer, G. F. (1984). T and Tn, general carcinoma autoantigens. Science, 224(4654), 1198 ±1206. Springer, G. F. & Ansell, N. J. (1958). Inactivation of human erythrocyte agglutinogens M and N by in¯uenza viruses and receptor destroying enzymes. Proc. Natl Acad. Sci. USA, 44, 182± 189. Springer, G. F. & Desai, P. R. (1974). Common precursors of human blood group MN speci®cities. Biochem. Biophys. Res. Commun. 61(2), 470± 475. Springer, G. F., Murthy, M. S., Desai, P. R. & Scanlon, E. F. (1980). Breast cancer patient's cell-mediated immune response to Thomsen-Friedenreich (T) antigen. Cancer, 45(12), 2949± 2954.

384

T Antigen Binding Peptides

Springer, G. F., Cheingsong-Popov, R., Schirrmacher, V., Desai, P. R. & Tegtmeyer, H. (1983). Proposed molecular basis of murine tumor cell-hepatocyte interaction. J. Biol. Chem. 258(9), 5702± 5706. Springer, G. F., Desai, P. R., Ghazizadeh, M. & Tegtmeyer, H. (1995). T/Tn pancarcinoma autoantigens: fundamental, diagnostic, and prognostic aspects. Cancer Detect. Prev. 19(2), 173± 182. Springer, T. A. (1995). Traf®cing signals on edothelium for lymphocyte recirculation and leukocyte emigration. Annu. Rev. Physiol. 57, 827 ± 872. Tam, J. P. (1988). Synthetic peptide vaccine design: synthesis and properties of a high-density multiple

antigenic peptide system. Proc. Natl Acad. Sci. USA, 85(15), 5409± 5413. Taylor, M. E., Bezouska, K. & Drickamer, K. (1992). Contribution to ligand binding by multiple carbohydrate-recognition domains in the macrophage mannose receptor. J. Biol. Chem. 267, 1719± 1726. Valadon, P., Nussbaum, G., Boyd, L. F., Margulies, D. M. & Scharff, M. D. (1996). Peptide libraries de®ne the speci®city of anti-polysaccharide antibodies to cryptococcus neoforms. J. Mol. Biol. 261, 11± 22. Zhang, R. D., Fidler, I. J. & Price, J. E. (1991). Relative malignant potential of human breast carcinoma cell lines established from pleural infusions and a brain metastasis. Invasion Metastasis, 11, 204± 215.

Edited by J. A. Wells (Received 2 January 1997; received in revised form 17 April 1997; accepted 17 April 1997)