We therefore kept isoleucine as the pY-1 residue and replaced the Ala residue at the pY+1 position with Abu, Ile, Val, D-Thr, Gln, Asn, norleucine (Nle), Leu, or tert-leucine (Tle) (Table 1, peptides 13C21). for anticancer drug design.4 T-cell PTP (TCPTP) is involved in haematopoiesis and cytokine response and is linked to the development of several inflammatory disorders including type 1 diabetes, Crohns disease, and rheumatoid arthritis.5 Selective inhibitors against PTPs would provide novel therapeutic agents as well as valuable chemical probes for investigating their physiological and pathological roles. However, designing isoform-specific PTP inhibitors has been challenging, because all of the PTPs share a highly conserved active site structure, which is positively charged. High-affinity binding to the PTP active site requires a negatively charged species which is, however, generally impermeable to the cell membrane. To overcome this problem, we recently developed a bicyclic peptide approach, featuring a short cell-penetrating peptide (CPP; e.g., FRRRR where is usually L-2-naphthylalanine) in one ring for endocytic cellular uptake and a target-binding motif in the second ring.6 Application of this method to PTP1B resulted in a cell-permeable, potent, and selective inhibitor for PTP1B (KD = 37 nM). Importantly, the bicyclic peptidyl inhibitor displayed a 17-fold selectivity over TCPTP, which shares 90% sequence identity CCI-006 with PTP1B within the active site. Presumably, the peptidyl inhibitor achieves isoform selectivity by engaging in additional interactions with the less conserved protein surfaces beyond CCI-006 the active site. The bicyclic approach was subsequently shown to be effective for delivering a wide variety of peptide sequences including negatively charged phosphopeptides into the cytosol of mammalian cells.7 In this study, we set out to test whether the bicyclic peptide approach might be applied to generate isoform-specific inhibitors against other members of the PTP superfamily. We chose TCPTP as the target, because it has been difficult to develop Capn2 inhibitors with selectivity for TCPTP over PTP1B.8 To this end, we designed a bicyclic peptide library featuring a degenerate peptide sequence, X1-X2-X3-F2Pmp-X4 [where F2Pmp is L-(phosphonodifluoromethyl)phenylalanine and X1-X4 are any of the 24 amino acid building blocks], in the first (or N-terminal) ring and 12 different amphipathic CPP motifs in the second (or C-terminal) ring (Fig. 1). F2Pmp is a non-hydrolysable analogue of phosphotyrosine (pY), which binds to the active site of PTPs with modest affinity and selectivity.9 The 24 building blocks included 10 proteinogenic L-amino acids (Ala, Ser, Pro, Ile, Asp, Gln, His, Tyr, Trp, and Gly), 5 unnatural -L-amino acids [norleucine (Nle), 2-aminobutyruc acid (Abu), 4-fluorophenylalanine (Fpa), phenylglycine (Phg), and pipecolic acid (Pip)], and 9 -D-amino acids [D-Ala, D-Pro, D-Val, D-Thr, D-Leu, D-Asn, D-Glu, D-Phe, D-2-naphthylalanine (D-Nal)]. The 12 CPP sequences consisted of different combinations of two or three aromatic hydrophobic residues (e.g., Phe and Nal) and three or four L- or D-arginine residues (Table S1).10 The bicyclic peptide library has a theoretical diversity of 3.98 106 and was synthesized on 2 g of TentaGel microbeads (~90 m; ~2.8 106 beads/g), as detailed in Supporting Information. Briefly, each library bead was topologically segregated into two different layers, with a unique bicyclic peptide synthesized in the surface layer and a linear peptide of identical sequence prepared in the inner core as an encoding tag.6,11 To increase the stringency of library screening, the ligand loading density of the surface layer (but not the inner layer) was reduced by 10-fold.12 In addition, during the coupling of F2Pmp, a 9:1 (mol/mol) ratio of Fmoc-Tyr/Fmoc-F2Pmp was used, resulting in another 10-fold reduction in the F2Pmp content. Altogether, the loading density of F2Pmp-containing bicyclic peptides around the bead CCI-006 surface was decreased by 100-fold, relative to that of the linear encoding peptides inside the beads. We have previously exhibited that reduction of the surface ligand density greatly reduces nonspecific binding caused by simultaneous conversation of a single protein molecule with multiple ligand molecules around the solid support.12 Open in a separate window Fig. 1 Structures of the bicyclic peptide library, the initial hit selected for optimization, and the optimization product (peptide 25). Residues modified during optimization are highlighted in red. Approximately 400 mg of the bicyclic peptide library (~1.1 106 compounds) was subjected to two rounds of screening for binding to TCPTP. During the first round, the library was incubated with biotinylated TCPTP (300 nM) and subsequently a streptavidin-alkaline phosphatase conjugate. After removal of any unbound proteins, the protein-bound beads were stained by incubating with a solution of 5-bromo-4-chloro-3-indolyl phosphate (BCIP).13 Turquoise colored beads (189 beads) were separated into intensely colored (49 beads) and medium colored categories (140 beads) and exhaustively washed to remove the bound proteins and the indigo dye. The two pools.