Cell-PermeableBicyclic Peptide Inhibitors againstIntracellular Proteins

Cyclic peptides (and depsipeptides)exhibit a wide range of biological activities.¹ Several innovative methodologies have recently been developed tosynthesize cyclic peptides, either individually² or combinatorially,³ and screenthem for biological activity. A particularly exciting applicationof cyclic peptides is the inhibition of protein–protein interactions(PPIs),^4,5 which remain challenging targets for conventionalsmall molecules. However, a major limitation of cyclic peptides isthat they are generally impermeable to the cell membrane, precludingany application against intracellular targets, which include mostof the therapeutically relevant PPIs. Although the formation of intramolecularhydrogen bonds⁶ or N^α-methylationof the peptide backbone⁷ can improve themembrane permeability of certain cyclic peptides, alternative strategiesto increase the cell permeability of cyclic peptides are clearly needed.

Protein-tyrosine phosphatase 1B (PTP1B) is a prototypical memberof the PTP superfamily and plays numerous roles during eukaryoticcell signaling. Because of its roles in negative regulation of insulinand leptin receptor signaling, PTP1B is a valid target for treatmentof type II diabetes and obesity.⁸ A largenumber of PTP1B inhibitors have been reported,⁹ but none of them have succeeded in the clinic. Designing PTP inhibitorsis challenging because most of the phosphotyrosine (pY) isosteressuch as difluorophosphonomethyl phenylalanine (F₂Pmp)¹⁰ are impermeable to the cell membrane. Additionally,because all PTPs share a similar active site, achieving selectivityfor a single PTP has been difficult. In this work, we report a potentiallygeneral approach to the design of cell-permeable cyclic peptidyl inhibitorsagainst intracellular proteins such as PTP1B.

We recently discoveredcyclo(FΦRRRRQ) (cFΦR₄, where Φ is l-naphthylalanine) as a novel class ofcell-penetrating peptides (CPPs).¹¹ Unlikeprevious CPPs, which are typically linear peptides that are entrappedin the endosome, cFΦR₄ efficiently escapes from theendosome into the cytoplasm. Short peptide cargos (1–7 aminoacids) can be delivered into mammalian cells by incorporating theminto the cFΦR₄ ring. Encouraged by this finding,we explored the possibility of developing bifunctional cyclic peptidescontaining both cell-penetrating and target-binding sequences as cell-permeableinhibitors against intracellular proteins. To generate specific inhibitorsagainst PTP1B, we synthesized a one-bead two-compound library on spatiallysegregated ChemMatrix resin,¹² in whicheach bead displayed a bifunctional cyclic peptide on its surface andcontained the corresponding linear peptide in its interior as an encodingtag [Scheme 1 and Figure S1 in the Supporting Information (SI)]. The bifunctionalcyclic peptides all featured the CPP motif FΦR₄ (orits inverse sequence RRRRΦF) on one side and a random pentapeptidesequence (X¹X²X³X⁴X⁵) on the other side, where X² represents a 9:1(mol/mol) mixture of Tyr and F₂Pmp while X¹ andX³–X⁵ are any of the 24 amino acids thatincluded 10 proteinogenic l-amino acids (Ala, Asp, Gln, Gly,His, Ile, Pro, Ser, Tyr, Trp), five unnatural α-l-aminoacids [F₂Pmp, l-4-fluorophenylalanine (Fpa), l-norleucine (Nle), l-phenylglycine (Phg), l-pipecolic acid (Pip)], and nine α-d-amino acids [d-Ala, d-Asn, d-Glu, d-Leu, d-β-naphthylalanine (d-Nal), d-Phe, d-Pro, d-Thr, d-Val]. The library has a theoreticaldiversity of 6.6 × 10⁵. The use of the 9:1 Tyr/F₂Pmp ratio at the X² position, together with a 5-foldreduction of the surface peptide loading, reduced the amount of F₂Pmp-containing peptides at the bead surface by 50-fold, increasingthe stringency of library screening.¹³ Screening100 mg of the library (∼300 000 beads/compounds) againstTexas red-labeled PTP1B resulted in 65 positive beads, which wereindividually sequenced by partial Edman degradation–mass spectrometry(PED-MS)¹⁴ to give 42 complete sequences(Table S1 in the SI).

Scheme 1

Evolution of a Cell-PermeablePTP1B Inhibitor

Three representativehit sequences, d-Thr-d-Asn-d-Val-F₂Pmp-d-Ala-Arg-Arg-Arg-Arg-Nal-Phe-Gln(inhibitor 1), Ser-d-Val-Pro-F₂Pmp-His-Arg-Arg-Arg-Arg-Nal-Phe-Gln(inhibitor 2), and Ile-Pro-Phg-F₂Pmp-Nle-Arg-Arg-Arg-Arg-Nal-Phe-Gln(inhibitor 3), were resynthesized and purified by HPLC.All three peptides are competitive PTP1B inhibitors, with peptide 2 being the most potent (IC₅₀ = 31 ± 3 nM)(Table S2 and Figure S2). Unfortunately,inhibitor 2 showed no significant activity in cellularassays. Confocal microscopy analysis of human cells treated with fluoresceinisothiocyanate (FITC)-labeled inhibitor 2 indicated poorcellular uptake of the peptide (Figure 1a).Although disappointing, this result was not entirely unexpected. Ourprevious study showed that as the size of the cargo inserted intothe cFΦR₄ ring increased, the cellular uptake efficiencyof the cyclic peptides decreased dramatically.¹¹ We reasoned that larger rings are more conformationallyflexible and may bind less tightly to the cell-surface receptors (e.g.,membrane phospholipids) during endocytosis. The negatively chargedF₂Pmp may also interact intramolecularly with the FΦR₄ motif and interfere with its CPP function.

Figure 1

(a) Live-cell confocalmicroscopy images (same Z section) of A549 lung cancercells after treatment for 2 h with5 μM FITC-labeled inhibitor 2 (top panel) or 4 (bottom panel) and the endocytosis marker dextran^Rho (1.0 mg/mL). (b) Lineweaver–Burk plot showing competitiveinhibition of PTP1B by inhibitor 4 at concentrationsof 0 (○), 28 (●), 56 (□), and 112 nM (▲).(c) Sensitivity of various PTPs to inhibitor 4 (the y axis values are pNPP hydrolysis rates for the PTPs relativeto that in the absence of inhibitor). The data shown in (b) and (c)are representative data sets.

To improve the cell permeability of inhibitor 2, weexplored a bicyclic system in which the CPP motif is placed in onering whereas the target-binding sequence constitutes the other ring(Scheme 1). The bicyclic system keeps the CPPring to a minimal size, which according to the previously observedtrend¹¹ should result in more efficientcellular uptake. The bicyclic system should be able to accommodatecargos of any size because incorporation of the latter does not changethe size of the CPP ring and therefore should not affect the deliveryefficiency of the cyclic CPP. The use of a rigid scaffold (e.g., trimesicacid) may also help keep the CPP and cargo motifs away from each otherand minimize any mutual interference. The smaller rings of a bicyclicpeptide compared with its monocyclic counterpart should result ingreater structural rigidity and improved metabolic stability.

To convert the monocyclic PTP1B inhibitor 2 into abicyclic peptide, we replaced the Gln residue (used for attachmentto the solid support and peptide cyclization) with (S)-2,3-diaminopropionic acid (Dap) and inserted a second Dap residueat the junction of the CPP and PTP1B-binding sequences (C-terminalto His) (Scheme 1). Synthesis of the bicyclewas accomplished by the formation of three amide bonds between a trimesicacid and the N-terminal amine and the side chains of the two Dap residues(Scheme 2).⁵ Briefly,the linear peptide was synthesized on Rink amide resin using standardFmoc chemistry and N^β-alloxycarbonyl(Alloc)-protected Dap. After removal of the N-terminal Fmoc group,the exposed amine was acylated with trimesic acid. Removal of theAlloc groups with Pd(PPh₃)₄ followed by treatmentwith PyBOP afforded the desired bicyclic structure. To facilitatelabeling with fluorescent probes, a lysine was added to the C-terminus.The bicyclic peptide (peptide 4) was deprotected usingtrifluoroacetic acid (TFA) and purified to homogeneity by HPLC.

Scheme 2

Solid-Phase Synthesis of Inhibitor 4

Reagents: (a) standard Fmoc chemistry;(b) trimesic acid, HBTU; (c) Pd(PPh₃)₄, N-methylaniline; (d) PyBOP; (e) TFA.

Bicyclic peptide 4 acts as a competitive inhibitorof PTP1B with a K_I value of 37 ±4 nM (Figure 1b). It is highly selective forPTP1B. When assayed against PTP1B and TCPTP using p-nitrophenyl phosphate (pNPP) as substrate (500 μM), inhibitor 4 had IC₅₀ values of 30 ± 4 and 500 ±250 nM, respectively (Figure 1c and Table S3). It exhibited minimal inhibition ofany of the other PTPs tested (≤10% inhibition of HePTP, SHP-1,PTPRC, PTPH1, or PTPRO at 1 μM inhibitor concentration). Gratifyingly,inhibitor 4 has greatly improved cell permeability overpeptide 2, as detected by live-cell confocal microscopyof A549 cells treated with FITC-labeled inhibitor 4 (Figure 1a). The treated cells showed diffuse fluorescencethroughout the cytoplasm and nucleus as well as fluorescence puncta,indicating that a fraction of the inhibitors reached the cytoplasmand nucleus while the rest was likely entrapped in the endosomes.Incubation of inhibitor 4 in human serum for 24 h at37 °C resulted in ∼10% degradation, whereas 91% of inhibitor 2 was degraded under the same conditions (Figure S3). Overall, inhibitor 4 compares favorablywith the small-molecule PTP1B inhibitors reported to date⁹ with respect to potency, selectivity over thehighly similar TCPTP (17-fold), and cell permeability (Table S4).

Inhibitor 4 wasnext tested for its ability to perturbPTP1B function during cell signaling. Treatment of A549 cells withinhibitor 4 (0–5 μM) resulted in a dramaticand dose-dependent increase in the pY levels of a large number ofproteins, consistent with the broad substrate specificity of PTP1B¹⁵ (Figure 2a). Analysisof the same samples by Coomassie blue staining showed similar amountsof proteins in all of the samples (Figure 2b), indicating that the increased pY levels reflected increased phosphorylation(or decreased PTP reaction) instead of changes in the total proteinlevels. Remarkably, the increase in tyrosine phosphorylation was alreadyapparent at 8 nM inhibitor 4. Interestingly, furtherincreases in the inhibitor concentration beyond 1 μM reversedthe effect on tyrosine phosphorylation, an observation that was alsomade previously by Zhang and co-workers with a different PTP1B inhibitor.¹⁶ To obtain further evidence that intracellularPTP1B was inhibited by peptide 4, we monitored the pYlevel of insulin receptor (IR), a well-established PTP1B substratein vivo,⁸ by immunoblotting with specificantibodies against the pY¹¹⁶²pY¹¹⁶³ site. Again,treatment with inhibitor 4 caused a dose-dependent increasein IR phosphorylation up to 1 μM inhibitor, and the effect leveledoff at higher concentrations (Figure 2c,d).Taken together, our data indicate that bicyclic inhibitor 4 efficiently entered mammalian cells and inhibited PTP1B in vivo.The decreased phosphorylation at higher inhibitor concentrations mayhave been caused by nonspecific inhibition of other PTPs (which mayin turn down regulate protein tyrosine kinases). It may also reflectthe pleiotropic roles played by PTP1B, which can both negatively andpositively regulate the activities of different protein kinases.¹⁷

Figure 2

Inhibition of PTP1B in vivo. (a) Anti-pY immunoblot ofglobal pYprotein levels in A549 cells after treatment with 0–5 μMinhibitor 4 for 2 h. (b) SDS-PAGE analysis (Coomassieblue staining) of the same samples from (a) showing uniform sampleloading in all lanes. (c) Effect of inhibitor 4 on insulinreceptor phosphorylation at Tyr¹¹⁶² and Tyr¹¹⁶³ sites. HepG2 cells were treated with the indicated concentrationsof inhibitor 4 for 2 h, stimulated with insulin (100nM) for 5 min, and analyzed by SDS-PAGE and immunoblotting with anti-IRpY¹¹⁶²/pY¹¹⁶³ antibody. (d) Quantitation of IR pYlevels from (c) (data shown are means ± SD from five independentexperiments).

To test the generalityof the bicyclic approach, we applied itto the design of cell-permeable inhibitors against peptidyl-prolylcis–trans isomerase (Pin1), a potential target for treatmentof a variety of human diseases, including cancer,¹⁸ for which potent, selective, and biologically active inhibitorsare still lacking.¹⁹ Thus, we fused cFΦR₄ with the previously reported monocyclic peptide 5, which is a potent inhibitor against Pin1 in vitro (K_D = 258 ± 66 nM) but membrane-impermeable²⁰ (Scheme 3). In addition,we replaced the l-Tyr at the pThr + 3 position with Arg toimprove the aqueous solubility. The resulting bicyclic peptide 6 bound Pin1 with K_D = 130 ±44 nM (Table S5 and Figure S4). Insertionof d-Ala at the pThr + 5 position to increase the separationbetween the Pin1-binding and cell-penetrating motifs improved theinhibitor potency by ∼2-fold (K_D = 72 ± 21 nM for inhibitor 7). Inhibitor 7 competed with FITC-labeled inhibitor 5 forbinding to Pin1 (Figure S5), indicatingthat they both bind to the Pin1 active site. Substitution of d-Thr for d-pThr of inhibitor 7 reduced itspotency by ∼10-fold (K_D = 620 ±120 nM for inhibitor 8; Table S5), whereas further replacement of the pipecolyl residue with d-Ala abolished the Pin1 inhibitory activity (peptide 9). As expected, bicyclic inhibitors 7–9 are cell-permeable (Figure S6). Treatment of HeLa cells with inhibitor 7 resultedin dose-dependent inhibition of cell growth (45% inhibition aftertreatment for 3 days at 20 μM inhibitor 7), whereasthe impermeable inhibitor 5 and inactive peptide 9 had no effect (Figure S7). Peptide 8 also inhibited cell growth, but to a lesser extent thaninhibitor 7. Finally, treatment of HeLa cells with inhibitor 7 dramatically increased the cellular levels of promyelocyticleukemia protein (PML), an established Pin1 substrate destabilizedby Pin1 activity (Figure S8).²¹

Scheme 3

Conversion of Impermeable Pin1 Inhibitor 5 into Cell-PermeableBicyclic Inhibitor 7

In conclusion, we have developed a potentially generalapproachfor the design of cell-permeable bicyclic peptides against intracellulartargets. Our preliminary studies show that replacement of the PTP1B-bindingmotif with other peptide sequences having different physicochemicalproperties also resulted in their efficient delivery into culturedmammalian cells.²² The availability ofa general intracellular delivery method should greatly expand theutility of cyclic peptides in drug discovery and biomedical research.