Rational design of cell-permeable cyclic peptides containing a D-Pro-L-Pro motif

Abstract

Cyclic peptides are capable of binding to challenging targets (e.g., proteins involved in protein-protein interactions) with high affinity and specificity, but generally cannot gain access to intracellular targets because of poor membrane permeability. In this work, we discovered a conformationally constrained cyclic cell-penetrating peptide (CPP) containing a D-Pro-L-Pro motif, cyclo(AFΦrpPRRFQ) (where Φ is L-naphthylalanine, r is D-arginine, and p is D-proline). The structural constraints provided by cyclization and the D-Pro-L-Pro motif permitted the rational design of cell-permeable cyclic peptides of large ring sizes (up to 16 amino acids). This strategy was applied to design a potent, cell-permeable, and biologically active cyclic peptidyl inhibitor, cyclo(Y^pVNFΦrpPRR) (where Y^p is L-phosphotyrosine), against the Grb2 SH2 domain. Multidimensional NMR spectroscopic and circular dichroism analyses revealed that the cyclic CPP as well as the Grb2 SH2 inhibitor assume a predominantly random coil structure but have significant β-hairpin character surrounding the D-Pro-L-Pro motif. These results demonstrate cyclo(AFΦrpPRRFQ) as an effective CPP for endocyclic (insertion of cargo into the CPP ring) or exocyclic delivery of biological cargos (attachment of cargo to the Gln side chain).

1. Introduction

Cyclic peptides represent an exciting class of drug modality because they are capable of binding to target proteins with antibody-like affinity and specificity, even when the target protein contains no hydrophobic binding pocket and is undruggable by small molecules.^1,2 Additionally, discovery of novel cyclic peptide ligands has been greatly accelerated by the advent of several combinatorial library synthesis and screening technologies.^3–10 However, a major limitation of cyclic peptides is that they are generally impermeable to the cell membrane, preventing their applications against intracellular targets.¹¹

To improve the cell permeability of cyclic peptides, researchers have conjugated them to linear as well as cyclic cell-penetrating peptides (CPPs) by several different approaches. The first approach involves direct attachment of cyclic peptides to a CPP.^12–16 For example, conjugation of a membrane-impermeable cyclic peptide inhibitor of the Keap1-Nrf2 interaction to a cyclic CPP increased the cytosolic entry efficiency of the Keap1 inhibitor by 98-fold.¹⁶ The resulting bicyclic peptide inhibited the proteasomal degradation of Nrf2 and stimulated the expression of Nrf2-controlled genes in cell culture at nanomolar concentrations. In an alternative approach, a cyclic peptide cargo is fused to a cyclic CPP ring to form a bicycle.¹⁷ This approach has been applied to generate bicyclic peptidyl inhibitors against the NEMO-IKKß¹⁸ and Ras-Raf interactions.¹⁹ Finally, peptidyl cargoes have been inserted into the ring of a cyclic CPP and delivered into the cytosol of mammalian cells (endocyclic delivery).^20,21 However, endocyclic delivery is limited to relatively small cyclic peptides (<13 aa), as increasing the cargo length (and therefore the ring size) resulted in progressive and dramatic reduction in the cellular entry efficiency.²¹ It was hypothesized that large rings are conformationally flexible and bind poorly to the cell membrane(s), a critical step(s) during cellular entry.

Previous studies demonstrated that incorporation of a D-Pro-L-Pro motif into a cyclic peptide biases the cyclic peptide into a β-hairpin structure, with the D-Pro-L-Pro motif located at one end of the β-hairpin.^22–24 β-Hairpin peptides of this type, as well as other constrained peptide loops, have demonstrated utility in binding to biological targets such as proteins.²⁵ We hypothesize that it might be possible to design β-hairpin shaped cyclic peptides containing a cell-penetrating motif on one end and a target-binding motif on the other end (Fig. 1). Conformational constraints endowed by the β-hairpin structure (e.g., by a D-Pro-L-Pro motif) may improve the cellular entry efficiency as well as the target-binding affinity and specificity. Moreover, a β-hairpin shaped CPP may be fused with different target-binding hairpin/loop sequences to generate cell-permeable cyclic peptides against challenging intracellular protein targets. In this work, we report our discovery of a structurally constrained CPP motif and its application to the development of a cell-permeable cyclic peptidyl inhibitor against the Grb2 SH2 domain.

Fig. 1.

Design strategy for structurally constrained, cell-permeable cyclic peptides. The D-Pro-L-Pro motif constrains the CPP motif into a stable β-hairpin structure; fusion of the CPP to a target-binding peptide loop generates a double β-hairpin structure, in which one hairpin (red) imparts cellular entry and the other hairpin engages a specific target.

2. Results and discussion

2.1. Discovery of structurally constrained cyclic CPPs

Based on our previous work,^21,26 we designed a series of cyclic CPPs containing a D-Pro-L-Pro motif, which was intended to induce a β-hairpin structure to the peptides (Table 1). In addition to the D-Pro-L-Pro motif, the cyclic peptides (peptides 1-4) also contained different combinations of 3 or 4 arginine and 2 or 3 hydrophobic residues [L-phenylalanine and/or L-naphthylalanine (Nal)] surrounding the D-Pro-L-Pro motif for potential cellular entry, and a pentaalanine sequence as a mock cargo. Variation of the stereochemistry of peptide 4, which was previously shown to dramatically improve the cellular entry efficiency of cyclic CPPs,²⁶ resulted in peptides 5-8. All of the peptides also contained a Gln residue for cyclization and attachment to solid support. The peptides were labeled with naphthofluorescein (NF) at the Gln side chain through an 8-amino-3,6-dioxa-octanoyl-lysine (miniPEG-K) linker. A highly active cyclic CPP, cyclo(fΦRrRrQ) (CPP9, where f is D-phenylalanine, Φ is L-naphthylalanine, and r is D-arginine),²⁶ was used as a reference to assess the cellular entry efficiency of the new cyclic CPPs. HeLa cells were treated with 5 μM NF-labelled peptide for 2 h, and their cell uptake efficiencies were determined by flow cytometry analysis. It should be noted that, since NF is fluorescent in the neutral environments of the cytosol and nucleus but not in the acidic endosome or lysosome (pH 4.5-6.5), the cellular fluorescence as measured by flow cytometry represents the cytosolic entry efficiency of a peptide.²⁷

Table 1.

Sequences and cellular entry efficiencies of cyclic peptides

Peptide ID	Sequencea	Cytosolic Entry (MFINF, %)b
CPP9	cyclo(fΦRrRrQ)-miniPEG-K(NF)	100
1	cyclo(AAFΦRpPRRRAAAQ)-miniPEG-K(NF)	21 ± 4
2	cyclo(AAFRRpPRRΦAAAQ)-miniPEG-K(NF)	5.7 ± 2.3
3	cyclo(AAFΦFpPRRRAAAQ)-miniPEG-K(NF)	8.3 ± 0.7
4	cyclo(AAFΦRpPRRFAAAQ)-miniPEG-K(NF)	22 ± 6
5	cyclo(AAFΦRpPrRFAAAQ)-miniPEG-K(NF)	32 ± 20
6	cyclo(AAFΦrpPRRFAAAQ)-miniPEG-K(NF)	33 ± 10
6AA	cyclo(AAFΦrAARRFAAAQ)-miniPEG-K(NF)	4.1 ± 1.6
7	cyclo(AAFΦRpPRrFAAAQ)-miniPEG-K(NF)	37 ± 6
8	cyclo(AAFϕRpPRRFAAAQ)-miniPEG-K(NF)	51 ± 25
9	cyclo(AFΦrpPRRFQ)-miniPEG-K(NF)	94 ± 19
9AA	cyclo(AFΦAAPRRFQ)-miniPEG-K(NF)	46 ± 17
10	cyclo(FΦrpPRRFAQ)-miniPEG-K(NF)	8 ± 2
11	cyclo(RAFΦrpPRRFRARQ)-miniPEG-K(NF)	34 ± 6
12	cyclo(DAFΦrpPRRFDADQ)-miniPEG-K(NF)	14 ± 6
13	cyclo(SASFΦrpPRRFSASAQ)-miniPEG-K(NF)	84 ± 24

^aUnderlined regions represent cell-penetrating sequences. Φ, L-2-naphthylalanine; ϕ, D-2-naphthylalanine; f, D-phenylalanine; r, D-arginine, p, D-proline; miniPEG, 8-amino-3,6-dioxaoctanoic acid; NF, naphthofluorescein.

^bValues represent the mean ± SD of three independent experiments and are relative to that of CPP9 (100%).

Peptides 1-4 entered HeLa cells with efficiencies 6-22% of that of CPP9. Peptide 4, which contained three hydrophobic and three arginine residues, was taken up most efficiently by HeLa cells and therefore selected for further optimization. Replacement of L-Arg at position 5 (which is immediately N-terminal to D-Pro) or position 8 with a D-arginine (r) increased the cytosolic entry efficiency by ~50% (Table 1, compares peptides 4-6). Inversion of configuration of Arg-9 or Nal-4 also substantially improved cellular entry (1.7- and 2.3-fold higher activities for peptides 7 and 8, respectively). These results provide further support for our earlier conclusion that a proper spatial arrangement of the arginine and hydrophobic side chains is crucial for high-affinity binding to the plasma and endosomal membranes of mammalian cells and consequently the cytosolic entry efficiency.²⁶

To generate a cyclic CPP without cargo, we removed the mock cargo moiety from peptide 6 to give a cyclodecapeptide, cyclo(FΦrpPRRFAA) (where p is D-proline). According to the design rules of Robinson and co-workers,²² the cyclodecapeptide should form a stable β-hairpin structure through the formation of four intramolecular hydrogen bonds (Fig. 2). To create a site for cargo attachment, we replaced one of the Ala residues with a glutamine to give peptides 9 and 10, respectively (Table 1). Peptides 9 and 10 were labeled with NF at the Gln side chain and their cytosolic entry efficiencies into HeLa cells were determined by flow cytometry. Peptide 9 is a very active CPP, with a cytosolic entry efficiency of 94% relative to that of CPP9. Surprisingly, peptide 10 is much less active (8% efficiency relative to CPP9). These data highlight the importance of a proper site for cargo attachment. Presumably, the cargo moiety attached to peptide 10 interferes with the interaction between peptide 10 and the membrane phospholipids.

Fig. 2.

Structures of conformationally constrained cyclic CPPs 9 and 10, with their amino acid residue numbers shown in red color.

The importance of the D-Pro-L-Pro motif for cellular entry was demonstrated by replacing it with Ala-Ala in peptides 6 and 9, to give peptides 6AA and 9AA, respectively (Table 1). Removal of the D-Pro-L-Pro motif reduced the cytosolic entry efficiency of peptides 6 and 9 by 8- and 2-fold, respectively. The greater effect on peptide 6 is expected, as peptide 6 has a larger ring than peptide 9 (14 aa vs 10 aa), is more flexible, and is therefore more dependent on the D-Pro-L-Pro motif for conformational constraint of the CPP sequence.

2.2. Cargo capacity of peptide 9

To evaluate the cargo capacity of peptide 9, we replaced the pentaalanine cargo of peptide 6 with positively charged (RARQRA, peptide 11), negatively charged (DADQDA, peptide 12), or neutral but longer peptidyl cargos (SASAQSAS, peptide 13) (Table 1). All three peptides entered HeLa cells, albeit with different efficiencies, suggesting that peptide 9 is capable of cytosolic delivery of cargos of diverse sizes and physicochemical properties. Similar to our previous observations,¹⁷ negatively charged cargos (as in peptide 12) decrease the cellular entry efficiency.

2.3. Design of a cell-permeable cyclic peptidyl inhibitor against Grb2 SH2 domain

To test whether peptide 9 can deliver biologically active cargos into the cell, we designed a cell-permeable cyclic peptidyl inhibitor against Grb2 SH2 domain. Grb2 is an adaptor protein that mediates signal transduction pathways and has been implicated in cancer cell growth and metastasis.^28,29 Grb2 has two Src homology 3 (SH3) domains that flank an SH2 domain. The SH2 domain interacts with specific phosphotyrosine (Y^p)-containing sequences in cell surface receptors, such as the epidermal growth factor receptor (EGFR), and recruits guanosine nucleotide exchange factors (e.g., SOS) to the cell surface, thereby activating the Ras protein and downstream signaling pathways.³⁰ The Grb2 SH2 domain recognizes peptides with a Y^p-X-N-X sequence, where X is typically a hydrophobic residue such as Val.^29,31 An X-ray crystal structure of Grb2 SH2 domain bound to Y^pXNX motifs revealed a β-turn conformation for the bound peptide.³¹ With this knowledge, we designed a cyclic peptidyl inhibitor, cyclo(Y^pVNFΦrpPRR) (peptide 14), by inserting the Grb2 SH2 ligand (Y^pVNF) into the cargo site of peptide 9. In our design, the phenylalanine residue (at position 2) is intended to serve dual function of cellular entry and Grb2 SH2 domain binding (Fig. 3A). We envisioned that the D-Pro-L-Pro motif would constrain the cyclic peptide into a β-hairpin structure, while the Y^pVNF would assume a β-turn conformation on the other end for binding to the Grb2 SH2 domain. We also synthesized peptides cyclo(YVNFΦrpPRR) (15) and cyclo(Y^pVNFΑapPRR) (16, where a is D-alanine) as negative controls. Peptide 15 contains a tyrosine in place of the Y^p residue and should have greatly reduced binding affinity to the SH2 domain, while peptide 16 has two critical cell-penetrating residues (Nal-arg) replaced with Ala-ala and is expected to have limited cell permeability.

Fig. 3.

(A) Structure of Grb2 SH2 domain inhibitor 14, with its cell-penetrating and Grb2-binding residues encircled in different boxes. (B) Inhibition of the binding of a fluorescently labeled peptide probe to Grb2 SH2 domain by peptides 14-16, as monitored by a fluorescence polarization assay. Data shown represent the mean ± SD of three parallel experiments and the curves were fitted to the data by using GraphPad PRISM ver.6 software.

The binding affinity of peptides 14-16 for the Grb2 SH2 domain was determined by using a fluorescence polarization (FP)-based competition assay and a fluorescein (FAM)-labeled peptidyl ligand of Grb2 SH2 domain, cyclo(AAY^pVNFFQ)-β-Ala-K(FAM) (K_D = 92 nM),³² as a probe. Peptide 14 exhibited an IC₅₀ of 0.40 μM (Fig. 3B and Table 2). As expected, peptide 15 showed no detectable binding to the Grb2 SH2 domain (IC₅₀ >10 μM), whereas peptide 16 bound to the SH2 domain with 4-fold higher affinity than peptide 14 (IC₅₀ = 0.10 μM) possibly due to removal of the steric clashes imposed by the bulky Nal residue.

Table 2.

Binding affinity and cellular entry efficiency of Grb2 SH2 domain ligands

Peptide ID	Sequencea	IC50 Value (μM)b	Cytosolic Entry Efficiency (MFINF, %)c
CPP9	cyclo(fΦRrRrQ)-miniPEG-K(NF)	-	100
14	Cyclo(FΦrpPRRYpVN)-miniPEG-K(NF)	0.40 ± 0.05	49 ± 2
15	Cyclo(FΦrpPRRYVN)-miniPEG-K(NF)	>10	34 ± 2
16	Cyclo(FAapPRRYpVN)-miniPEG-K(NF)	0.10 ± 0.02	2 ± 2

^aΦ, L-2-naphthylalanine; a, D-alanine; f, D-phenylalanine; p, D-proline; Y^p, L-phosphotyrosine; r, D-arginine; NF, naphthofluorescein.

^bValues derived from curve fitting to data derived from three sets of experiments.

^cValues represent the mean ± SD of three independent experiments and are relative to that of CPP9 (100%).

Next, we labeled peptide 14-16 with NF at the Asn side chain through a miniPEG-K linker and determined their cytosolic entry efficiencies by treating HeLa cells with the labeled peptide (5 μM) for 2 h followed by flow cytometry analysis. Peptides 14 and 15 showed cytosolic entry efficiencies of 49% and 34%, respectively, relative to CPP9 (100%) (Table 2). As expected, peptide 16 had negligible cytosolic exposure (2% relative to CPP9). Peptide 14 was also labeled at the Asn side chain with FAM and its entry into HeLa cells was examined by live-cell confocal microscopy. Cells treated with FAM-labeled peptide 14 exhibited punctate fluorescence in the cytoplasmic region (Fig. 4A), suggesting that peptide 14 entered the cells by endocytic mechanisms and a fraction of it remained entrapped inside the endosomes/lysosomes.

Fig. 4.

Cellular activities of Grb2 SH2 domain inhibitors 14-16. (A) Confocal microscopic images of HeLa cells after treatment with 5 μM FAM-labeled peptide 14 for 2 h (1% FBS). I, DIC; II, GFP channel; III, overlap of I and II. (B, C) Western blots showing the effect of peptides 14-16 on the Grb2/Ras signaling pathway in MDA-MB-468 cells. Cells were treated with the indicated concentrations of peptide for 2 h and stimulated with EGF (50 ng/mL) for 10 min. The cells were then lysed and the cell lysates were separated on SDS-PAGE and blotted with antibodies specific for phosphorylated and total MEK and ERK1/2. Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was used as loading controls.

Peptides 14-16 were tested for their ability to modulate the Grb2/Ras signaling pathways, by monitoring the phosphorylation of MEK and ERK1/2 kinases downstream of Ras in human breast cancer MDA-MB-468 cells, which have been used for testing other Grb2 SH2 domain inhibitors.³³ MDA-MB-468 cells were pre-treated with the peptides for 2 h and the Grb2/Ras signaling pathway was then activated by treatment with epidermal growth factor (EGF; 50 ng/mL as final concentration) for 10 min. Peptide 14 dose-dependently decreased the level of phosphorylated MEK (p-MEK), with an IC₅₀ value of ~15 μM, but had no effect on the total intracellular level of MEK (Fig. 4B). The metabolic instability of peptide 14 (dephosphorylation of Y^p by endogenous protein-tyrosine phosphatases) likely contributed to the relatively high cellular IC₅₀ value. Under the same condition, neither peptide 15 nor 16 showed detectable effect on p-MEK (Fig. 4C). The effect of peptide 14 on ERK1/2 phosphorylation was less dramatic, consistent with its further downstream position in the Grb2/Ras signaling pathway.

2.4. Structural characterization of peptides 9 and 14

We used NMR-derived distance restraints to compute an ensemble of structures of peptide 9 in a buffered aqueous solution. Backbone atoms were fully assigned excluding carbonyl carbons and proline amide nitrogens. Ninety percent of H^β were assigned stereospecifically, with Pro6 H^β protons being degenerate. All sidechain protons were assigned except for Arg H^η (at positions 4, 7, and 8). NMR spectra of HPLC-purified peptide 9 exhibited doubling of most NMR signals, indicating a slow-exchange equilibrium with a minor conformation populated at 10 ± 5%, as measured from intensities of signals in ¹H-¹³C HSQC spectra. Sequential backbone and non-stereospecific sidechain assignments were also obtained for the minor species. Assignments were made from two-dimensional (2D) ¹H TOCSY and NOESY, and 2D ¹H-¹³C HSQC and HSQC-TOCSY NMR spectra.³⁴ Distance restraints obtained from a zero-quantum filtered, perfect echo NOESY experiment (τ_m 100 ms) were limited to sequential and short range NOEs (longest being i, i+2 : Ala1 – Nal3, Ar8 – Gln10), and the resulting ensemble did not converge to a unique, well-defined structure (Fig. 5A). Nevertheless, spectral features provide insights into the structural propensities of peptides 9 and 14.

Fig. 5.

Solution structure of peptide 9. (A) Ensemble of 25 structures consistent with NOE derived distance restraints, superposed at the D-Pro-L-Pro motif. Magenta, lowest energy conformation. (B) Concentration dependent CSPs for Ala1 HB protons for the major and minor species. (C) Concentration dependent CSPs (white to red gradient) mapped onto the lowest energy member of the ensemble (magenta, panel A) with carbons, nitrogens and oxygens shown in grey. (D) Difference in proton chemical shifts at 3 mM between major (trans D-Pro5 – Pro6 peptide bond) and minor (cis D-Pro5 – Pro6 peptide bond) species (white to blue gradient) mapped onto the lowest energy member of the ensemble (magenta, panel A) with all other atoms shown in grey.

Isomerization of the D-Pro5 – Pro6 peptide bond was determined to explain the observed doubling of NMR signals. Strong NOEs between D-Pro5 H^α – Pro6 H^δ protons of the major species implies a trans-trans peptide bond between D-Pro5 – Pro6, while a strong D-Pro5 H^α – Pro6 H^α NOE in the minor species indicates isomerization of the D-Pro5 – Pro6 peptide bond to the cis conformation. The two conformers exhibit distinct backbone and sidechain chemical shifts throughout most of the peptide, with differences as large as 0.95 ppm (Fig. 5D). Sidechain protons exhibit a cis-trans chemical shift differences were limited to ~0.2 ppm. The largest sidechain chemical shift differences were observed for D-Arg4 H^N, which shifts downfield by 0.82 ppm. Pro6 H^α, Arg7 H^α, and Arg8 H^N also exhibit large chemical shifts differences of 0.54, 0.38, and 0.41 ppm, respectively. CSPs in the backbone atoms among other residues are modest. We conclude that structural consequences of isomerizing the Pro5 – Pro6 peptide bond are manifest throughout the backbone of the peptide, with limited effects on sidechain environments.

The major trans species of peptide 9 exhibits concentration dependent chemical shifts, while the minor species does not (Fig. 5B). Chemical shift perturbations (CSPs) from ¹H-¹H TOCSY spectra at varying concentrations (0.18 mM to 3.0 mM) of the major species range from 0 to 0.218 ppm over this concentration range (Fig. 5C), and were found to cluster around the Nal3 ring system (average ¹H CSP of 0.145 ppm) and the Ala1 methyl group (¹H CSP of 0.218 ppm). These observations indicate self-association of the major conformation of peptide 9, in which the hydrophobic sidechain of Nal3 and methyl group of Ala1 likely mediates self-association. The absence of concentration-dependent shifts for the minor cis conformation (Fig. 5C) suggests that it does not undergo self-association.

The NMR data suggest that peptide 9 does not adopt a unique conformation and only samples discrete β hairpin-like structures transiently. Absence of spectral signatures consistent with stable β-hairpin structures, such as cross-strand sidechain-sidechain, H^α-H^α, H^N-H^α, or H^N-H^N NOEs (Fig. S1), or amide proton chemical shifts greater than 8.5 ppm, indicates that a stable β-hairpin structure is not adopted. Nevertheless, ³J_Nα coupling constants measured from a DQF-COSY spectrum (recorded at 3 mM) show three bond couplings greater than 8 Hz for the major species residues Nal3, D-Arg4, Arg7, Arg8, and Phe9, consistent with an extended β-strand conformation; however, broad line widths (≥ 6 Hz) for the major species, likely result in overestimated the three bond couplings. Likewise, the C^α and C^β chemical shifts fall in the random coil region and are not generally consistent with a β-hairpin structure. The minor cis species also does not exhibit observable cross-strand sidechain-sidechain, H^α-H^α, H^N-H^α, or H^N-H^N NOEs, or amide proton chemical shifts greater than 8.5 ppm. Comparing the major trans and minor cis species, C^α chemical shifts become more negative and C^β chemical shifts becoming more positive (Fig. 6A); ignoring possible confounding effects from the presence of two D amino acids, this suggests greater β-strand like character in the minor cis species. Residues D-Arg4 and Arg7, on either side of the D-Pro5 – Pro6 motif, exhibit coupling constants of 8.8 and 8.2 Hz, respectively (Fig. 6E), which are suggestive of β-sheet character, but no cross-strand NOEs are observed between residues Phe2 and Phe9, Nal3 and Arg8, or D-Arg4 and Arg7.³⁵ Thus, while portions of the peptide exhibit some NMR signatures consistent with a β hairpin structure, we conclude that such well-defined structure is only transiently sampled.

Fig. 6.

Peptides 9 and 14 exhibit spectral features consistent with partial/transient β hairpin formation. (A-D) Deviation of C_α and C_β chemical shift values from those of random coils in peptide 9, major and minor species, and peptide 14, major and minor species, respectively. (E) ³J_Nα coupling constants for the major and minor species of peptides 9 and 14. Color coded for alpha helix character (< 6 Hz, red), random coil (6 – 8 Hz, green) and beta strand character (>8 Hz, blue).

Peptide 14 also exhibited doubling of most NMR signals. A strong D-Pro5 H^α – Pro6 H^α NOE in the minor species supports an analogous trans to cis peptide bond isomerization. The major and minor species were found to exist in a 7:1 ratio as determined from peak volumes of the ¹H-¹³C HSQC spectra. Sequential backbone and non-stereospecific sidechain assignments were completed for both the major and minor species from TOCSY and NOESY spectra. As for peptide 9, peptide 14 also lacks NOE patterns consistent with formation of a stable β-hairpin (Fig. S2). The minor species residues D-Arg4 and Arg7 show ³J_Nα couplings greater than 8 Hz (Fig. 6E). Likewise, in the minor species, most residues exhibit C^α chemical shifts that are more negative and C^β chemical shifts more positive, indicative of increased β-strand character (Fig. 6D). For the minor species, D-Arg4 has large sidechain chemical shift differences (0.4 – 0.8 ppm) and D-Arg4 H^N shifts downfield by 0.44 ppm. Pro6 H^α, Arg7 H^N, Arg7 H^α, and Arg8 H^N also experience chemical shifts changes of 0.47, 0.28, 0.31, and 0.55 ppm, respectively.

The chemical shifts, ³J_Nα couplings and NOE data indicate that neither peptide 9 nor 14 adopts a stable, classical β-hairpin structure in buffered aqueous conditions. However, chemical shift data suggest that both peptides (especially their minor species) have significantly greater β-strand character than the more populated trans conformations. This interpretation is supported by the circular dichroism (CD) spectrum of peptide 14, which features a negative band at 208 nm and a positive band at 190 nm (Fig. 7). A well-defined β-hairpin peptide typically has a strong minimum at 215 nm and a positive signal at 190 nm, whereas a random coil is characterized a strong minimum at 198 nm.^36,37 The CD spectrum of peptide 14 is most consistent with a mixture of random coil and β-strand components. There is also a small increase in signal near 227 nm, which has previously been observed in peptides containing an L-2-naphthylalanine.³⁸

Fig. 7.

Circular dichroism of peptide 14 in aqueous solution (pH 7.4).

3. Conclusion

We discovered a highly efficient cyclic CPP (peptide 9) containing a D-Pro-L-Pro motif. Cyclic CPPs 9 and 14 do not adopt a stable β-hairpin structure in aqueous solution. However, NMR data suggest the peptides do transiently sample β-hairpin conformations immediately surrounding the D-Pro-L-Pro motif, where the CPP sequence is located. This conformational plasticity may contribute to its relatively high cytosolic entry efficiency, which rivals that of cyclic CPP9, one of the most active CPPs reported to date.²⁶ It is presently unclear whether peptide 9 adopts the β-hairpin structure upon binding to the cell membrane and/or when a cargo molecule is inserted into its ring. Nevertheless, peptide 9 has demonstrated the capacity of effectively delivering cargo molecules into the cytosol of mammalian cells by either endocyclic (insertion of peptidyl cargos into the CPP ring) or exocyclic approach (e.g., attachment of a fluorescent dye to the Gln side chain). Importantly, peptide 9 is able to deliver either positively charged, neutral, or negatively charged peptide sequences into the cell, including a Y^p-containing peptide. Finally, we have shown that peptide 9 can be leveraged to rationally design a relatively potent, cell-permeable, and biologically active cyclic peptidyl inhibitor (peptide 14) against the Grb2 SH2 domain. However, it should be noted that the cargo can profoundly impact the cellular entry efficiency of peptide 9, especially in the case of endocyclic delivery, by neutralizing the positive charge of the CPP, perturbing the conformation of the CPP, and etc. In fact, other investigators have previously introduced β-hairpin-inducing motifs (e.g., D-Pro-L-Pro) into linear CPPs and obtained mixed results. Schneider and colleagues found that incorporation of D-Pro-L-Pro induced linear CPPs to form β-hairpin structures and improved the membrane activity of the peptides,³⁹ whereas Safa et al. showed that introduction of β-hairpin structures into arginine-rich CPPs greatly decreased their cellular entry.⁴⁰ We have also previously observed that excessive conformational rigidity, or more specifically, constraining a CPP into nonproductive conformations, greatly reduces the cellular entry efficiency of the CPP.⁴¹

4. Experimental section

4.1. Materials

Fmoc-protected L/D-amino acids for peptide synthesis were purchased from Advanced ChemTech (Louisville, KY), NovaBiochem (La Jolla, CA), or Aapptec (Louisville, KY). Fmoc-protected phsophotyrosine (Fmoc-pTyr) was purchased from CreoSalus (Louisville, KY). Rink amide resin (100-200 mesh, 0.3-0.6 meq/g) was purchased from Chem-Impex (Wood Dale, IL). O-Benzotriazole-N,N,N’,N’-tetramethyluronium hexafluorophosphate (HBTU) and 1-hydroxybenzotriazole hydrate (HOBt) were purchased from Aapptec. Fluorescein isothiocyanate (FITC), 5-(and-6)-carboxynaphthofluorescein succinimidyl ester (NF-NHS) were purchased from Sigma Aldrich (St. Louis, MO), and 5(6)-carboxy-tetramethylrhodamine was purchased from Chem-Impex (Wood Dale, IL). All solvents and other chemical reagents were obtained from Sigma-Aldrich, VWR (West Chester, PA), or Fisher Scientific and were used without further purification unless noted otherwise. Cell culture media, fetal bovine serum, penicillin-streptomycin, Dulbecco’s phosphate-buffered saline (DPBS) (2.67 mM potassium chloride, 1.47 mM potassium phosphate monobasic, 137 mM sodium chloride, 8.06 mM sodium phosphate dibasic), and 0.25% trypsin-EDTA were purchased from Invitrogen (Carlsbad, CA). Dulbecco’s Modified Eagle’s medium (DMEM), 0.25% trypsin-EDTA solution, was purchased from Sigma-Aldrich, Dulbecco’s phosphate-buffered saline (DPBS) was purchased from Thermo Fisher. Fetal bovine serum (FBS) was purchased from VWR. Antibodies for MEK1/2 (#9126), phosphor-MEK1/2 (Ser217/221) (#9154), GAPDH (#5174), and anti-rabbit IgG HRP (#7074) were purchased from Cell Signaling Technology.

4.2. Peptide synthesis and labeling

Peptides were synthesized on Rink amide resin (100-200 mesh, 0.3-0.6 meq/g) using standard Fmoc chemistry. The typical coupling reaction contained 5 equivalents of Fmoc-amino acid, 5 equivalents of HBTU, 5 equivalents of HOBt and 10 equivalents of diisopropylethylamine (DIPEA) and was allowed to proceed with mixing for 10 min twice. After the addition of the last (N-terminal) residue, the allyl group on the C-terminal Glu residue was removed by treatment with Pd(PPh₃)₄, phenylsilane (0.1 and 10 equivalents, respectively) in anhydrous DCM (3 × 15 min). The N-terminal Fmoc group was removed by treatment with 20% piperidine in DMF and the peptide was cyclized by treatment with benzotriazole-1-yl-oxy-tris-pyrrolidino-phosphonium hexafluorophosphate (PyBOP)/HOBt/DIPEA (5, 5, and 10 equivalents) in DMF for 3 h. Peptide were treated with 82.5:5:5:5:2.5 (v/v) TFA/thioanisole/water/phenol/ethanedithiol for 3 h for deprotection and cleavage from the resin. After evaporation of solvents by a N₂ stream, peptides were triturated with cold ethyl ether (3x) and purified by reversed-phase HPLC on a C18 column and eluted with linear gradients of acetonitrile (containing 0.05% TFA) in ddH₂O (containing 0.05% TFA). To label peptide with FITC, purified and lyophilized peptide (~1 mg) was dissolve in 200 μL of 1:1:1 (v/v) DMSO/DMF/150 mM sodium bicarbonate (pH 8.5) and mixed with 10 μL of FITC in DMSO (100 mg/mL). After 30 min reaction at room temperature, the mixture was subjected to reversed-phase HPLC on a semi-preparative C18 column to isolate the FITC-labeled peptide. To generate rhodamine/NF/FAM-labeled peptides, peptide and dye (1:1 mol/mol) were dissolved in 150 μL 1:1 (v/v) DMF/150 mM sodium bicarbonate (pH 8.5) and incubated for 2 h at room temperature. The mixture was subjected to reversed-phase HPLC on a semi-preparative C18 column to isolate the fluorescent-labeled peptide. The authenticity of peptides was confirmed by MALDI-TOF mass spectrometry. For unlabeled peptides, concentration was determined based on absorbance at 280 nm and the following molar absorptivity: Trp ε₂₈₀: 5560 M⁻¹ cm⁻¹, Nal ε₂₈₀: 3550 M⁻¹ cm⁻¹, and Tyr ε₂₈₀: 1500 M⁻¹ cm⁻¹.

4.3. Cell culture

HeLa cells and MDA-MB-468 cells were maintained in medium consisting of DMEM, 10% FBS and 1% penicillin/streptomycin. H1299 cells were grown in RPMI-1640 supplemented with 10% FBS and 1% penicillin/streptomycin. Cells were cultured in a humidified incubator at 37 °C with 5% CO₂.

4.4. Protein expression and purification

Escherichia coli BL21 (DE3) cell harboring pGEX-Grb2SH2 plasmid was grown at 37 °C in Luria broth (LB) supplemented with 0.05 mg/ml ampicillin to an OD₆₀₀ of 0.50. The expression of protein was induced by addition of 0.1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) for 4 h at 30 °C. The cells were harvested by centrifugation at 2,000 rpm for 30 min. The cell pellet was suspended in 20 mL of buffer A (1x PBS pH 7.4, 2 mM DTT, 0.2 mg/mL lysosome, 2 tablet of protease inhibitor cocktail, 100 μg/ml phenylmethylsulfonyl fluoride (PMSF) and 1% tween 20) and stirred at 4 °C for 30 min. After subjected to brief sonication (2 × 10 s pulses), the slurry was supplied with 0.5% protamine sulfate and stirred for another 20 min at 4 °C. The crude lysate was clarified by centrifugation at 15,000 rpm in a SS-34 fixed angle rotor for 30 min. The supernatant was loaded onto a glutathione-Sepharose column, after 1 h incubation the column was washed with buffer A to remove unbounded protein, and elute with buffer B (1x PBS pH 7.4, 2 mM DTT, 0.2 mg/mL lysosome, 2 tablet of protease inhibitor cocktail, 100 μg/ml PMSF, 1% tween 20 and 10 mM glutathione (GSH)). Fractions containing GST-Grb2 SH2 protein were combined and concentrated in an Amicon Ultra-15 cellulose filter. Protein concentration was determined by Bradford assay using bovine serum albumin as the standard.

4.5. Fluorescence polarization

FP experiments were performed by incubating 100 nM peptide cyclo(AAY^pVNFFQ)-β-Ala-K(FAM) with varying concentrations of Grb2 SH2 protein (0-5 μM) in PBS (2.67 mM potassium chloride, 1.47 mM potassium phosphate monobasic, 137 mM sodium chloride, 8.06 mM sodium phosphate dibasic, pH 7.4) for 1 h at room temperature. The FP values were measured on a Tecan infinite M1000Pro, with excitation and emission wavelengths at 470 and 535 nm, respectively. Equilibrium dissociation constants (K_D) were determined by plotting the FP values as a function of Grb2 SH2 concentration. The titration curves were fitted to the following equation

$$Y=\frac{\left(A_{min}+\left(A_{max}\times \frac{Q_b}{Q_f}-A_{min}\right)\left(\frac{\left(L+x+K_d\right)-\sqrt{\left({\left(L+x+K_d\right)}^2-4Lx\right)}}{2L}\right)\right)}{\left(1+\left(\frac{Q_b}{Q_f}-1\right)\left(\frac{\left(L+x+K_d\right)-\sqrt{\left({\left(L+x+K_d\right)}^2-4Lx\right)}}{2L}\right)\right)}$$

where Y is the measured FP at a given Grb2 SH2 concentration x; L is the peptide concentration; Q_b/Q_f is the correction factor for dye-protein interaction; A_max is the maximum polarization when all the peptides are bound to Grb2 SH2, while A_min is the minimum polarization. Competition assay was performed by incubating 100 nM cyclo(AAY^pVNFFQ)-β-Ala-K(FAM), 0.18 μM Grb2 SH2 protein, and varying concentrations (0-20 μM) of unlabeled peptides in PBS (pH 7.4) for 2 h. The FP values were measured as described above. IC₅₀ values were determined by plotting the FP values as a function of peptide concentration and fitted to a four-parameter logistic curve with GraphPad PRISM ver.6 software.

4.6. Flow cytometry

HeLa cells were seeded in 12-well plates (1.5 × 10⁵ cells per well) for 24 h. On the day of experiment, the cells were incubated with 5 μM FITC-, FAM-, or NF-labeled peptide in DMEM containing 1% or 10% FBS at 37 °C for 2 h in the presence of 5% CO₂. At the end of incubation, the peptide medium was removed, and the cells were washed with DPBS twice. The cells were detached from the plate with 0.25% trypsin, diluted into DPBS, pelleted at 200g for 5 min, washed twice with DPBS, resuspended in 200 μL of DPBS, and analyzed on a BD FACS LSR II flow cytometer. For the FITC/FAM-labeled peptides, a 488-nm laser was used for excitation and the fluorescence was analyzed in the FITC channel. For NF-labeled peptides, a 633-nm laser was used for excitation and the fluorescence emission was analyzed in the APC channel. All flow cytometry experiments were gated for 10,000 live cell events.

4.7. Confocal microscopy

To monitor peptide internalization, 1 mL of HeLa cell suspension (5 × 10⁴ cells) was seeded in a 35-mm glass-bottomed microwell dish (MatTek) and cultured overnight. Cells were gently washed with DPBS twice and treated with 5 μM FAM-labeled peptide in phenol red-free DMEM containing 1% FBS for 37 °C in the presence of 5% CO₂. After 2 h of incubation, the medium was removed, and the cells were gently washed with DPBS twice and phenol red-free DMEM was added. The cells were imaged on a Nikon A1R live-cell confocal microscope equipped with 100X oil objective in a heated chamber at 37 °C and 5% CO₂. Images were captured under the same parameters. NIS-Elements AR was used to denoise the images and add scale bars. Images presented in this report were generated using the instruments and services at the Campus Microscopy and Imaging Facility, The Ohio State University.

4.8. Western blot analysis

MDA-MB-468 cells were maintained in DMEM supplemented with 10% FBS and 1% Abs. The cells were seeded in 6-well plate until reaching ~80% confluence. The cells were starved in serum-free DMEM for 20 h. Cells were washed with DPBS once and then treated with indicated concentrations of peptide in serum-free medium for 2 h, following by stimulation with EGF (50 ng/mL) for 10 min. The cells were harvested immediately and lysed in RIPA buffer (Thermo, #89900) supplemented with 2x protease inhibitor cocktail (Millipore, #539137), 2x phosphatase inhibitor (Roche, # 04906845001) and 5 mM EDTA. Protein concentrations were measured by BCA Protein Assay Kit (Thermo, #23235), and equal amounts of total protein (~20 μg) were separated on a 10% SDS-PAGE gel and electrophoretically transferred to nitrocellulose membranes. After blocked with 5% nonfat dry milk in TBST for 1 h at room temperature, the membranes were then incubated with corresponding primary antibodies at 4 °C overnight. The membranes were washed with TBST for 5 min (three times) and incubated with secondary antibody for 2 h at room temperature. The membranes were washed three times with TBST, and the signals were detected by using Chemiluminescent HRP Antibody Detection Reagent (Denville, E-2500) following the manufacturer’s protocol.

4.9. NMR spectroscopy

Fractions of HPLC-purified peptide were combined, lyophilized, and resuspended in 10 mM potassium phosphate, 147 mM NaCl, 3 mM KCl, pH 5.0 (PBS). The resulting sample was divided evenly, lyophilized, and then resuspended with PBS containing either 10% or 99.96% D₂O (Cambridge Isotope Laboratories, Inc.) resulting in concentrations of 2.95 mM for peptide 9 and 1 mM for peptide 14 as measured by UV absorbance at 280 nm using a NanoDrop 1c. The molar extinction coefficient of 3936 M⁻¹cm⁻¹ was used for L-napthylalanine.⁴² All datasets were recorded at 298 K on Bruker Avance III 600 MHz and 800 MHz spectrometers equipped with 5-mm TXI ¹H-detect cryoprobes with z-axis gradients. Data were processed using NMRFx and NMRPipe and were analyzed/visualized with NMRViewJ.^34,43,44

Backbone resonance assignments were obtained from a combination of two-dimensional (2D) ¹H-¹H NOESY (τ_m 350 ms) and ¹H-¹H TOCSY (τ_m 70 ms, DIPSI2, 10 kHz nominal B1 field) spectra by “walking” along the backbone for sequential assignment of H^N and H^α resonances.⁴⁵ A 2D ¹H-¹⁵N HSQC spectrum was also obtained (at natural abundance) to aid in the assignment of amide protons. Sidechain assignments were carried out by recording a 2D ¹H-detected, ¹³C-resolved HSQC-NOESY (τ_m 400 ms) for the aliphatic and aromatic proton regions, and a 2D ¹H-¹³C-HSQC-TOCSY (τ_m 60 ms, DIPSI2, 10 kHz nominal B₁ field) of the aliphatic regions.^46,47 Aromatic sidechain resonances were assigned based on strong H_β-H_δ correlations in the 2D ¹H-¹H NOESY spectrum. Intramolecular distance restraints were obtained from a zero quantum-filtered perfect-echo ¹H-¹H NOESY using a mixing time of 100 ms (acquisition times of 283 ms (direct) and 55 ms (indirect)).⁴⁸ Distance restraints were categorized as strong, medium, and weak with upper bounds set to 2.5, 3.7, and 5 Å, respectively.

For peptide 9, the chemical shift difference between the major (trans D-Pro5 – Pro6 peptide bond) and minor (cis D-Pro5 – Pro6 peptide bond) species were calculated by taking the difference of the major and minor species proton chemical shift at 3 mM. Concentration-dependent chemical shift perturbations (CSPs) were measured by obtaining the difference between shifts in the 0.2 mM and 3 mM samples.

Using the Python interface to Xplor-NIH⁴⁹ and custom napthylalanine parameter files derived from HIC-Up,⁵⁰ an extended structure of BHT1 generated in Xplor was first subjected to a simulated annealing protocol using NOE derived distance restraints. The lowest energy structure was then subjected to a round of refinement, generating 100 structures, from which 25 lowest energy conformers were chosen for analysis. NMRPipe and Xplor-NIH were accessed and used through the NMRBox platform.⁵¹

4.10. Circular dichroism

CD spectra were obtained using a Jasco J-815 circular dichroism spectrometer using a 1 mm quartz cell (Starna, Inc). Peptide samples were dissolved in 10 mM phosphate buffer (pH 7.4) for a final concentration of 150 μM. Parameters set for the experiment are as follow: wavelengths from 300 to 190 nm were measured in a scanning speed of 100 nm/min with data pitch of 0.5 nm; bandwidth of 1 nm; temperature was kept at 25 °C; baseline was corrected by subtract the blank; and CD data were output in ellipticity measured in millidegrees (mdeg) with an average of three scans.