Identification and Characterization of a New Family of Cell-penetrating Peptides

Abstract

Cell-penetrating peptides can translocate across the plasma membrane of living cells and thus are potentially useful agents in drug delivery applications. Disulfide-rich cyclic peptides also have promise in drug design because of their exceptional stability, but to date only one cyclic peptide has been reported to penetrate cells, the Momordica cochinchinensis trypsin inhibitor II (MCoTI-II). MCoTI-II belongs to the cyclotide family of plant-derived cyclic peptides that are characterized by a cyclic cystine knot motif. Previous studies in fixed cells showed that MCoTI-II could penetrate cells but kalata B1, a prototypic cyclotide from a separate subfamily of cyclotides, was bound to the plasma membrane and did not translocate into cells. Here, we show by live cell imaging that both MCoTI-II and kalata B1 can enter cells. Kalata B1 has the same cyclic cystine knot structural motif as MCoTI-II but differs significantly in sequence, and the mechanism by which these two peptides enter cells also differs. MCoTI-II appears to enter via macropinocytosis, presumably mediated by interaction of positively charged residues with phosphoinositides in the cell membrane, whereas kalata B1 interacts directly with the membrane by targeting phosphatidylethanolamine phospholipids, probably leading to membrane bending and vesicle formation. We also show that another plant-derived cyclic peptide, SFTI-1, can penetrate cells. SFTI-1 includes just 14 amino acids and, with the exception of its cyclic backbone, is structurally very different from the cyclotides, which are twice the size. Intriguingly, SFTI-1 does not interact with any of the phospholipids tested, and its mechanism of penetration appears to be distinct from MCoTI-II and kalata B1. The ability of diverse disulfide-rich cyclic peptides to penetrate cells enhances their potential in drug design, and we propose a new classification for them, i.e. cyclic cell-penetrating peptides.

Introduction

Cell-penetrating peptides (CPPs)⁶ are short peptides that overcome the barrier of the cell membrane and enter living cells. CPPs can be conjugated with a cargo (i.e. oligonucleotide, peptide sequence, or polysaccharide) and efficiently deliver it inside cells, and so they are of great importance in the field of drug delivery. The most extensively studied CPPs are the Tat peptide, derived from the HIV-1 transactivator of transcription protein (1), and penetratin, derived from the third helix of the Drosophila Antennapedia homeodomain (2). In the last 20 years, many different CPPs have been identified from a range of sources. However, only one cyclic peptide has been reported to pass through cell membranes, i.e. the cyclotide Momordica cochinchinensis trypsin inhibitor II (MCoTI-II) (3–5).

Cyclotides are head-to-tail cyclic peptides that contain ∼30 amino acids, including six conserved cysteine residues that form a cyclic cystine knot (CCK) at the core of their structure (6). The use of the stable CCK motif as a drug scaffold has emerged as an interesting field of research in recent years (7–9). In particular, the exceptional stability of the CCK motif makes it an ideal framework for molecular engineering and drug design applications (10–12). For example, the CCK peptides MCoTI-II and kalata B1 have been successfully engineered to introduce new bioactivities to the molecules (13–16). The versatility of the CCK framework together with the cell-penetrating properties of MCoTI-II makes this mini-protein of special interest for applications in drug design.

MCoTI-II has been reported to be internalized into cells by macropinocytosis (3), but the specific mechanism by which this occurs has not been determined. Furthermore, possible interactions of MCoTI-II with phospholipids or receptors in cell membranes, and the eventual fate of the molecule within cells, have not been determined. In this study, an analysis of the cellular uptake of MCoTI-II and the prototypic cyclotide kalata B1 is presented. These two peptides are representative examples of two subfamilies of cyclotides (17) and have very different sequences (Fig. 1). For comparison, another cyclic peptide, the sunflower trypsin inhibitor 1 (SFTI-1) (18), is also examined here, and all three cyclic peptides were found to penetrate cells. As the number of cell-penetrating peptides increases, it is becoming important to develop and maintain appropriate classification schemes. We propose that the cyclic peptides examined in this study constitute a new family of CPPs that we refer to as cyclic cell-penetrating peptides (CCPPs). Cyclic peptides have advantages over their linear counterparts because they have extraordinary stability (19, 20) and CCPPs appear to be promising molecules in the field of drug delivery.

FIGURE 1.

Structure and surface representations of MCoTI-II and kalata B1. A, sequence of MCoTI-II and kalata B1. Cysteines are highlighted in bold type and numbered with roman numerals. The disulfide connectivity is represented with yellow lines. The loops between the cysteine residues are numbered (loops 1–6). B and C, schematic representations of MCoTI-II (Protein Data Bank code 1IB9) (B) and kalata B1 (Protein Data Bank code 1NB1) (C). Cysteines are numbered with roman numerals, and the disulfide connectivity is represented in yellow. Loops are numbered loop 1–6 in accordance with the sequence represented in A. The surface representations are shown in two views. The surface representation on the left side is in the same orientation as the schematic. Basic residues are highlighted in blue, acidic in red, and hydrophobic in green.

EXPERIMENTAL PROCEDURES

Peptide Extraction and Purification

MCoTI-II was extracted from the seeds of M. cochinchinensis as described previously (21). Kalata B1 was extracted from the leaves of O. affinis (22). Both peptides were purified by RP-HPLC and masses analyzed using electrospray ionization mass spectrometry (ES-MS).

Peptide Synthesis

The peptides synthesized in this study are summarized in Table 1. MCoTI-II single mutants (K6A, K9A, and K10A), kalata T20K (kT20K), kalata V25K (kV25K) and SFTI-1 were synthesized using a native chemical ligation strategy as previously described (23, 24). A MCoTI-II double mutant MCoKKAA (where Lys 9 and 10 are replaced with alanine residues) was assembled by Fmoc solid phase peptide synthesis in a polyethylene glycol-polystyrene (PEG) resin with a preloaded Lys (Applied Biosystmes, Australia) using an automatic peptide synthesizer (Liberty). Aspartimide formation at the DG motif was effectively suppressed by the introduction of Fmoc-2,4-dimethoxybenzyl (Merck) protection at the Gly backbone nitrogen. The linear peptide was oxidized in 0.1 m ammonium bicarbonate at pH 8 overnight and purified by RP-HPLC. The oxidized peptide was cyclized with immobilized trypsin (Promega) as previously described (25). A linear Tat peptide was synthesized by microwave-assisted solid phase peptide synthesis using an automatic peptide synthesizer (Liberty) following standard Fmoc chemistry on a 2-chlorotrityl chloride resin. The deprotected peptide was cleaved using trifluoroacetic acid. All peptides were purified using preparative RP-HPLC to purity >95%.

TABLE 1.

Peptides examined in this study

Mutated residues are underlined. The strategy applied for synthesis is specified for each peptide. Lys residues labeled with Alexa-488 are highlighted in boldface. Kalata mutants are labeled in the introduced Lys in the sequence.

Peptide	Sequence	Synthesis and cyclization method
MCoTI-II(MCo)	Cyclic-(GGVCPKILKKCRRDSDCPGACICRGNGYCGSGSD)	Plant extraction
MCoKKAA	Cyclic-(GGVCPKILAACRRDSDCPGACICRGNGYCGSGSD)	Fmoc, trypsin cyclization
MCoK6A	Cyclic-(GGVCPAILKKCRRDSDCPGACICRGNGYCGSGSD)	Boc chemistry, thioester
MCoK9A	Cyclic-(GGVCPKILAKCRRDSDCPGACICRGNGYCGSGSD)	Boc chemistry, thioester
MCoK10A	Cyclic-(GGVCPKILKACRRDSDCPGACICRGNGYCGSGSD)	Boc chemistry, thioester
Kalata B1 (kB1)	Cyclic-(GLPVCGETCVGGTCNTPGCTCSWPVCTRN)	Plant extraction
kT20K	Cyclic-(GLPVCGETCVGGTCNTPGCKCSWPVCTRN)	Boc chemistry, thioester
kV25K	Cyclic-(GLPVCGETCVGGTCNTPGCTCSWPKCTRN)	Boc chemistry, thioester
SFTI-1	Cyclic-(GRCTKSIPPICFPD)	Boc chemistry, thioester
Tat(48–57)	GRKKRRQRRR	Fmoc chemistry, linear

NMR Spectroscopy

For NMR studies peptides were dissolved in 90% H₂O, 10% D₂O at a concentration of ∼1 mm at pH 3.2. Spectra were recorded at 298K on a Bruker Avance-600 spectrometer. ¹H one-dimensional and TOCSY and NOESY two-dimensional spectra were recorded.

Biotin Labeling of Peptides

The three Lys residues of MCoTI-II and the Lys of kT20K were conjugated with a 5-fold molar excess of NHS-biotin (Quantum Scientific) in 0.1 m sodium bicarbonate, pH 8, for 2 h at room temperature. Both labeled peptides were purified by RP-HPLC using an analytical column, and the mass was confirmed by ES-MS.

Alexa-488 Labeling of Peptides

Alexa-488 was purchased from Molecular Probes. Native MCoTI-II, MCoTI-II mutants (MCoKKAA, MCoK6A, MCoK9A, and MCoK10A), kT20K, kV25K, and SFTI-1 were conjugated with Alexa-488 for cellular uptake studies. Equal moles of Alexa-488 were incubated with the peptides for 3 h at room temperature while being gently shaken. The conjugated peptides were purified by RP-HPLC, and the mass was confirmed by ES-MS. Peptides containing only one labeled Lys were selected.

Identification of Residues Labeled with Alexa-488

MCoTI-II and derivatives have several Lys residues in the sequence that could potentially contain the introduced Alexa-488. To determine which Lys was labeled, the peptides were reduced and trypsinized as described previously (26). The fragments resulting from the digestion were subjected first to MALDI-TOF-MS (Applied Biosystems) and then sequenced by ES-MS/MS using a nanospray instrument (API-QSTAR).

Peptide-Lipid Interaction Studies by Surface Plasmon Resonance

Small unilamellar vesicles with a diameter of 50 nm were prepared by extrusion as described previously (27) and used for surface plasmon resonance (SPR) studies. Briefly, the small unilamellar vesicles were prepared from lipid films using a combination of freeze-thaw fracturing and sizing by extrusion. Lipid solutions in chloroform were dried under a stream of N₂, and residual organic solvent was removed in vacuum (16 h). The lipid film was hydrated with HEPES buffer and subjected to eight freeze-thaw cycles to produce multilamellar vesicles. To produce small unilamellar vesicles, multilamellar vesicles were extruded through polycarbonate filters with a 50-nm pore size (19 times). Synthetic lipids were used to prepare model membranes as follows: palmitoyloleoylphosphatidylcholine (POPC), palmitoyloleoylphosphatidylglycerol (POPG), and palmitoyloleoylphosphatidylethanolamine (POPE). Lipid vesicles were prepared in 10 mm HEPES buffer with 150 mm NaCl, pH 7.4, for SPR studies. All measurements were performed using a BIAcore 3000 system (GE Healthcare). The interaction of the peptides with lipid bilayers was studied using different lipid mixtures; the molar ratio of each lipid in the mixture is specified in parentheses: 1 mm POPC, POPC/POPG (4:1), and POPC/POPE (4:1). The binding of the peptides to the lipid bilayers was followed by SPR as described previously (28). All solutions were prepared with 10 mm HEPES buffer, pH 7.4, containing 150 mm NaCl.

Peptide Interactions with Phosphatidylinositols and Phosphatidic Acid Using Nitrocellulose Membranes

The interaction of MCoTI-II, kalata B1, and SFTI-1 with various lipids was performed with strips of nitrocellulose membranes commercialized as PIP Strips^TM (Invitrogen) as described previously (29). The final concentration of the peptides was 1 μm.

Peptide HSPG Interaction Studies by SPR

Heparan sulfate proteoglycan (HSPG) from basement membrane (mass >400 kDa) was purchased from Sigma. The interaction of HSPG with the peptides was performed by SPR in two different experiments. In the first experiment, HSPG (0.25 mg/ml) was incubated overnight at 37 °C with POPC lipid vesicles to allow the protein to partition into the membrane. Then the lipid-HSPG system was deposited onto the L1 sensor chip surface (2 μl/min, 2400 s contact time). Peptides were injected over the lipid surface (5 μl/min, 180 s), and dissociation followed for 600 s per injection cycle. In the second experiment, the lipid vesicles were injected into the chip (2 μl/min, 2400 s contact time) followed by the injection of the HSPG (0.8 mg/ml). Peptides were then injected (5 μl/min, 180 s). Experiments were performed in 10 mm HEPES with 150 mm NaCl, pH 7.4.

Peptide HS Interaction Studies by SPR

Heparan sulfate (HS) from bovine kidney (∼14 kDa) was purchased from Sigma. HS was biotinylated via the amino groups. A solution of 5 mg/ml HS was incubated with a 15-fold molar excess of EZ-link NHS biotin (Pierce) for 3 h at room temperature. The unconjugated biotin was removed by filtration in a Centricon 3 (Amicon). The sample was recovered in 10 mm HEPES with 0.3 m NaCl, pH 7.4, for SPR measurements. The interaction of the peptides with HS was studied using an SA chip, which was derivatized with streptavidin. The biotin-labeled HS was coupled to the surface of the SA chip at two different concentrations in independent channels (0.2 and 2 μg/ml). One channel was left untreated and served as a negative control. Next, the peptide was injected over the HS surface in independent injections (5 μl/min, 180 s), and dissociation followed for 600 s per injection cycle. The experiments were performed in 10 mm HEPES with 150 mm NaCl, pH 7.4.

Cell Culture

Human MCF-7 breast cancer cells were grown in Dulbecco's modified Eagle's medium (DMEM) (Invitrogen) supplemented with 20 units/ml penicillin (Invitrogen), 20 μg/ml streptomycin (Invitrogen), 2 mm l-glutamine (Invitrogen), 0.1% insulin (Sigma), and 10% heat-inactivated fetal bovine serum (IRH Bioscience Inc.). Mouse RAW 264.7 macrophage-like cells were grown in RPMI 1640 medium (Invitrogen) supplemented with 20 units/ml penicillin (Invitrogen), 20 μg/ml streptomycin (Invitrogen), 2 mm l-glutamine (Invitrogen), and 10% heat-inactivated fetal bovine serum (IRH Bioscience Inc.). Cells were incubated at 37 °C with 5% CO₂.

Cell Viability Studies

Cell viability was evaluated by a 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide assay. MCF-7 or RAW 264.7 cells were seeded at 1 × 10⁵ cells/ml in sterile 96-well microtiter plates overnight in their supplemented media. The day after, various concentrations of the peptides were added to the plates, and the cells were incubated for 24 h. 3-(4,5-Dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide reagent (Sigma) was then added to a final concentration of 1 mg/ml and incubated for 3 h. The 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide reagent was then removed, and the precipitated formazan crystals were dissolved by adding 100 μl of isopropyl alcohol to each well. The plates were read at 590 nm using a Powerwave XS plate reader (Bio-Tek).

Cell Uptake Visualization Studies by Live Cell Imaging

To study the cellular uptake of the various peptides, 1 × 10⁵ MCF-7 or RAW 264.7 cells were seeded onto live cell imaging plates (MatTek Corp.) and let stand overnight. The following day, the cell culture medium was removed, and the cells were incubated for 1 h at 37 °C in fresh medium containing the Alexa-488-conjugated peptide at the specified concentration. After incubation, extracellular peptide was washed out, and CO₂-independent media (Invitrogen) were added. The cellular uptake was analyzed by confocal laser scanning microscopy (Zeiss LSM 510) using a 63× oil immersion objective. Cells were maintained at 37 °C during live cell imaging. For three-dimensional time-lapse imaging, confocal Z-stacks were sequentially captured and reconstructed using the LSM software (Zeiss). To stain the membrane, the wheat germ agglutinin (WGA)-Alexa-647 conjugate (Invitrogen) was used.

Cell Uptake Quantification by Flow Cytometry

MCF-7 or RAW 264.7 cells were seeded onto a 24-well plate at 1 × 10⁵ cells/ml and left to set down overnight. The following day, the cells were incubated with 5 μm Alexa-488-labeled-peptide for 1 h to allow cellular uptake. Peptide was then washed out twice with the corresponding supplemented media. A final wash with PBS was performed prior to trypsinization with 2.5 mg/ml trypsin for 10 min at 37 °C. The reaction was stopped via the addition of 4 volumes of media. Next, the cells were collected and centrifuged for 5 min at 4 °C. The pellet was washed with PBS and resuspended in 500 μl of PBS. To quench the extracellular membrane-bound peptide, the cells were treated with 25 μg/ml trypan blue and immediately analyzed by flow cytometry (BD FACSCanto II). Live cells were gated by forward/side scattering from a total of 10,000 events.

Cell Treatment with Macropinocytosis Inhibitor

MCF-7 cells were pretreated for 30 min with the macropinocytosis inhibitor ethylisopropylamiloride (EIPA) at 100 μm. Cells were then treated with Alexa-488-labeled peptide (5 μm) and incubated in either the presence or absence of EIPA for 1 h prior to flow cytometric analysis.

RESULTS

Peptide Synthesis and Characterization

The peptides examined in this study are summarized in Table 1. MCoTI-II and kalata B1 were extracted from plant material. For the sake of brevity, hereafter these peptides are referred as MCo and kB1, respectively. A double alanine mutant of MCo was synthesized by substituting Lys-9 and Lys-10 with Ala residues. The MCo double mutant (MCoKKAA) was synthesized and cyclized via an enzyme-mediated approach (25). The similarity of the αH NMR chemical shifts of MCoKKAA to those of native MCo confirmed that the mutant peptide had the same overall structure as the native peptide (supplemental Figs. S1 and S2A). The single point mutants MCoK6A, MCoK9A, and MCoK10A were synthesized by manual Boc chemistry and cyclized and oxidized in a single step. The peptides maintained the native fold based on analysis of αH chemical shifts (supplemental Fig. S2A).

For monitoring cellular uptake by flow cytometry and microscopy, peptides were labeled with Alexa-488. The labeled residues were identified and are highlighted in Table 1.

Native kB1 does not contain any Lys residues and therefore a Lys was introduced at position 20 (kT20K) to allow labeling with Alexa-488. This mutant was chosen because it has similar structure, bioactivity, and membrane-binding properties to wild-type kB1 (23, 28). An inactive mutant of kB1 (kV25K) was used as a negative control (23). Structural analysis show that both mutants maintain the fold of the native peptide (supplemental Fig. S2B).

Cyclic SFTI-1, originally derived from sunflower seeds, was synthesized by Boc chemistry, and NMR studies confirmed that it adopted the native conformation (supplemental Fig. S2C) (30).

Cellular Uptake Studies of MCo and kT20K

It has been reported that MCo penetrates cells, whereas kT20K remains bound to cell membranes (3). These studies were done in fixed cells with biotin labeling of the peptides. Recent evidence suggests that fixation of cells can lead to artifactual redistribution of membrane-bound peptides into the cytoplasm and nucleus (31). Therefore, it was of interest to reevaluate the cell penetration of cyclic peptides by live cell imaging using Alexa-488-labeled peptides. Hereafter Alexa-488 will be referred to as Alexa.

Two cell lines were examined, the macrophage cell line RAW 264.7 and the breast cancer cell line MCF-7. The two cell lines were used to allow comparison with the previous studies of cellular uptake of MCo and kT20K (3). To choose a nontoxic concentration to perform the assays, the cell viability was evaluated in the presence of 5 and 15 μm concentrations of labeled peptides. The cytotoxicities of MCo-Alexa and kT20K-Alexa conjugates were tested in RAW 264.7 and MCF-7 cells. In accordance with previous studies (3), kT20K-Alexa was toxic to cells at concentrations >5 μm, whereas MCo-Alexa was not toxic at the concentrations tested (supplemental Fig. S3). Consequently, the experiments with MCo- and kT20K-Alexa conjugates were carried out at 5 μm.

To study cellular uptake into MCF-7 or RAW 264.7 cells, 5 μm Alexa-labeled MCo or kT20K was incubated with the cells for 1 h. Live cell imaging by confocal microscopy showed that both MCo- and kT20K-Alexa conjugates were present inside MCF-7 and RAW 264.7 cells (Fig. 2).

FIGURE 2.

MCo-Alexa and kT20K-Alexa cellular uptake in RAW 264.7 and MCF-7. Alexa-labeled MCo and kT20K (5 μm) were incubated with RAW 264.7 and MCF-7 cells for 1 h, washed out, and analyzed by confocal microscopy. MCo-Alexa and kT20K-Alexa are represented in green. MCo-Alexa and kT20K-Alexa are present in punctate compartments in RAW 264.7 and MCF-7 cells. For visualization of peptides that were interacting with the plasma membrane, the plasma membrane of MCF-7 cells was labeled with WGA-Alexa 647 (MCF-7 + WGA) and represented in red. MCo-Alexa is not found attached to the membrane, whereas kT20K-Alexa is found in endosomal compartments and bound to the cell membrane (white arrow). Images are from a single confocal plane. Bar, 5 μm.

Confocal microscopy images suggested that MCo-Alexa and kT20K-Alexa were located in endosomal compartments (Fig. 2), sized from 1 to 2 μm. Endosomes larger than 0.2 μm are defined as macropinosomes (32, 33), and therefore it is probable that MCo-Alexa and kT20K-Alexa were present in macropinosomes. Macropinosomes containing MCo-Alexa and kT20K-Alexa were more abundant in RAW 264.7 than in MCF-7 cells (Fig. 2). This result is consistent with the enhanced macropinocytic activity of macrophages as part of their surveillance role in the immune response (32). The presence of kT20K in endosomal compartments was surprising because kT20K interacts with the plasma membrane, and therefore, it was expected that it would enter the cells by direct membrane interaction and be located in the cytoplasm.

To evaluate if MCo-Alexa and kT20K-Alexa colocalize with the plasma membrane in living cells, wheat germ agglutinin conjugated with Alexa-647 (WGA-Alexa-647) was added to MCF-7 cells to label the plasma membrane. WGA is a lectin that binds to N-acetylglucosamine and N-acetylneuraminic acid (34). Binding of MCo-Alexa to the cell membrane was not observed (Fig. 2), but kT20K-Alexa was partly located in endosomes (Fig. 2) and partly attached to the cell membrane (Fig. 2, white arrowhead).

Cellular Uptake of MCo and kT20K Is Inhibited at 4 °C

Endocytosis is inhibited at low temperatures (35); therefore, to corroborate that MCo-Alexa and kT20K-Alexa follow an endocytic pathway, internalization experiments were conducted at 4 °C. As expected, MCo-Alexa was not present inside RAW 264.7 or MCF-7 cells at 4 °C (Fig. 3A). Similarly, kT20K-Alexa did not enter cells at 4 °C. This suggests that both peptides penetrate cells by endocytosis. By contrast, peptides that penetrate cells by membrane translocation have been reported to penetrate cells with the same efficiency at 4 and 37 °C (36, 37).

FIGURE 3.

MCo-Alexa and kT20K-Alexa cellular uptake at 4 °C and in the presence of a macropinocytosis inhibitor. A, MCo and kT20K Alexa-labeled peptides (5 μm) were incubated for 1 h at 4 °C; the peptide was washed out, and the cells were analyzed by confocal microscopy. The uptake of MCo-Alexa and kT20K-Alexa in RAW 264.7 and MCF-7 cells at 4 °C is represented. None of the peptides was detected inside cells at 4 °C. Bar, 5 μm. B, cellular uptake of MCo-Alexa (5 μm) and kT20K-Alexa (5 μm) into MCF-7 cells treated with the macropinocytosis inhibitor EIPA (100 μm) was quantified by flow cytometry. Relative cellular uptake is plotted in the absence (−) or presence (+) of EIPA. Data are the average (± S.E.) of two experiments (n = 4). Statistical significance between the cellular uptake of MCo-Alexa with and without EIPA treatment (*) was analyzed by analysis of variance with Newman-Keuls post hoc test (p < 0.0001).

Cellular Uptake of MCo Is Reduced after Treatment with a Macropinocytosis Inhibitor

To determine whether the peptides entered cells by macropinocytosis, MCF-7 cells were treated with the macropinocytosis inhibitor EIPA, which is derived from amiloride. Differences in cellular uptake between MCo-Alexa and kT20K-Alexa in EIPA-treated and control cells were quantified by flow cytometry. EIPA decreased the uptake of MCo-Alexa by ∼40%, whereas kT20K-Alexa uptake was not affected (Fig. 3B). These data indicate that MCo-Alexa and kT20K-Alexa follow different internalization routes. MCo-Alexa penetrated MCF-7 cells to a high extent by macropinocytosis, whereas kT20K-Alexa most likely penetrated MCF-7 cells through direct peptide membrane interaction. However, the reason why kT20K-Alexa appeared in macropinosome-like compartments is unknown.

Dissecting the Important Residues for Cell Penetration in MCo and kB1

The distribution of residues on the surface of MCo and kB1 is very different (Fig. 1). kB1 has a hydrophobic patch that is implicated in its membrane binding (Fig. 1C) (28, 38), whereas MCo is more hydrophilic, with a positive patch on the surface that could be responsible for its cell-penetrating properties (Fig. 1B). The MCo mutants, MCoKKAA, MCoK6A, MCoK9A, and MCoK10A, and the kB1 mutants, kT20K and kV25K, were used to determine important residues for cell penetration.

Prior to the cell penetration experiments, the toxicity of the mutants was tested at 5 and 15 μm. The MCo mutants were not cytotoxic to RAW 264.7 and MCF-7 cells at the tested concentrations (supplemental Fig. S3). The inactive mutant of kB1, kV25K, was also noncytotoxic at the same concentrations (supplemental Fig. S3), in agreement with previous reports, which demonstrated that this mutant is not hemolytic, and it does not disturb lipid bilayers (23, 28).

The cellular uptake of Alexa-labeled MCo, kB1, and the mutant peptides was quantified by flow cytometry. Consistent with the live cell images, the cellular uptake of MCo-Alexa and kB1-Alexa was higher in RAW 264.7 cells (Fig. 4A) than in MCF-7 cells (Fig. 4B).

FIGURE 4.

Cellular uptake of MCo mutants and kB1 mutants in RAW 264.7 and MCF-7 cells. The relative fluorescence of Alexa-labeled peptides was measured by flow cytometry. The amount of peptides taken up by cells was normalized to the results of the control cells as a standard. Alexa-labeled peptides were incubated for 1 h at 5 μm. Data represent the average (± S.E.) of three experiments (n = 6). Statistical significance between the control and the rest of the samples (#), between MCo-Alexa and its mutants (*), and between kT20K-Alexa and kV25K-Alexa (‡) was analyzed by analysis of variance with Newman-Keuls post hoc test (p < 0.05).

The decrease in the cellular uptake of Alexa-labeled MCo mutants in RAW 264.7 cells was significant relative to native MCo. The cellular uptake of MCoKKAA-Alexa decreased by 50% relative to MCo-Alexa (Fig. 4A), whereas the decrease in cellular uptake of mutants MCoK6A-Alexa, MCoK9A-Alexa, and MCoK10A-Alexa was lower. For the three single mutants, the cellular uptake decreased by ∼30% relative to MCo-Alexa (Fig. 4A). This decrease suggests that the positive charge of MCo is important for its cellular uptake.

In MCF-7 cells, the cellular uptake of MCoKKAA-Alexa decreased by 25% relative to MCo-Alexa (Fig. 4B), in contrast to the 50% decrease observed in RAW 264.7. The lowest level of cellular uptake in MCF-7 cells was observed for MCoK10A-Alexa, and its cellular uptake decreased by 50% compared with MCo-Alexa. In contrast, MCoK9A-Alexa uptake was only ∼18% lower than MCo-Alexa. Interestingly, the uptake of the single mutant MCoK6A-Alexa into MCF-7 cells increased by 20% in comparison with MCo-Alexa.

The uptake of MCo-Alexa in RAW 264.7 cells was 1.5-fold higher than kT20K-Alexa (Fig. 4A). By contrast, kT20K-Alexa uptake in MCF-7 cells was 1.5-fold higher than MCo-Alexa. The uptake of kT20K-Alexa and kV25K-Alexa was similar in RAW 264.7 cells (Fig. 4A) but strikingly different in MCF-7 cells (Fig. 4B). In fact, the uptake of kV25K-Alexa was negligible in MCF-7 cells. This finding suggests that kT20K-Alexa uptake into MCF-7 is dependent on its affinity for membranes, as the membrane-inactive analog (28) did not penetrate MCF-7 cells.

Interaction of MCo with Model Membranes

kB1 is a hemolytic peptide (24) that has a strong affinity for membranes (28, 39), and indeed most of its bioactivities can be attributed to membrane interactions (38, 40). However, no information is available on the membrane affinity of MCo. Insight into the interaction of MCo with membranes is crucial to understand its cellular uptake mechanism. For this reason, model membranes of different compositions were examined. kB1 was also studied for comparison under similar conditions.

Peptide-lipid interactions were monitored by SPR using various lipid compositions. POPC, the major component of eukaryotic cell membranes, was used to mimic the neutral charge properties of eukaryotic cells (41). kB1 has been reported to bind to POPE (28), and therefore, POPC/POPE mixtures were also studied. In addition, POPC/POPG lipid mixtures were used to mimic the charge properties of bacterial membranes. Fig. 5A shows that MCo does not bind to any model membrane tested, including POPC, POPC/POPG, or POPC/POPE. In contrast, kB1 had high affinity for the POPC/POPE model membrane (Signal ∼7000 response units (RU)), consistent with published data (28, 39).

FIGURE 5.

MCo and kB1 phospholipid interaction. A, MCo and kB1 affinity for model membranes of different composition. Peptides were tested at 50 μm. The affinity of peptides for POPC, POPC/POPG (4:1), or POPC/POPE (4:1) was studied by SPR. Signal in response units (RU). B, MCo-biotin (1 μm) and kT20K-biotin (1 μm) binding to other lipids was tested using a PIP Strips^TM membrane. Peptides were labeled with biotin to allow detection of the interaction.

Interaction of MCo with Phosphatidylinositides (PIs) and Phosphatidic Acid (PA)

The membrane lipid phosphatidylinositol can be phosphorylated either singly or in combination to produce distinct lipids known as phosphoinositides (PIs). Each phosphatidylinositol derivative has a distinct set of biological activities and displays its own discrete subcellular localization (42, 43). As MCo did not interact with the model membranes tested, it was of interest to study whether PIs are involved in the cellular uptake of MCo.

To study peptide-PI interactions, commercially available nitrocellulose membranes that contain eight different PIs were used. kT20K was used as a control peptide, and biontylinated MCo and kT20K were incubated on the membranes at 1 μm. The results obtained for both peptides are illustrated in Fig. 5B. MCo-biotin showed a strong interaction with PI(3,4,5)P₃ and with PA. MCo-biotin showed a weaker interaction with other PIs, including PI(3)P, PI(4)P, PI(5)P, PI(3,4)P₂, PI(3,5)P₂, and PI(4,5)P₂ (Fig. 5B). In agreement with the SPR data, MCo-biotin did not interact with POPC or POPE.

kT20K-biotin only interacted with POPE, consistent with the SPR results. MCo-biotin and kT20K-biotin both showed weak affinity for phosphatidylserine (PS). PS is located in the inner leaflet of the plasma membrane of healthy cells, and it is only found in the outer membrane in cells that undergo apoptosis (44). Consequently, it is unlikely that this phospholipid takes part in the cell penetration of the peptides.

Interaction of MCo with Cell Receptor HSPG and HS

Membrane-associated proteoglycans, including the HSPG, have been reported to be crucial for internalization of CPPs (45, 46). The protein moiety of HSPG is inserted in the membrane, whereas the glycosaminoglycan chains stretch outward from the membrane surface (47). It has been postulated that the attachment of cationic CPPs with negatively charged HS on plasma membranes via electrostatic interactions is an initial step for their internalization (48). To understand the cellular uptake of MCo and kB1, it was of interest to study whether HSPG could interact with them. The well studied CPP derived from the sequence of the HIV-1 transcriptional activator protein, known as the Tat peptide, was used as a control (see sequence in Table 1).

As a first approach, the phospholipid POPC was deposited on the L1 chip. HSPG was injected over the lipid bilayer, and peptides were injected after HSPG incorporation in the membrane. MCo did not show affinity for the membrane-HSPG complex (Fig. 6A). kB1 gave a slightly higher response in the presence of HSPG receptor in comparison with MCo. The control peptide Tat showed high affinity for HSPG (2000–3000 RU).

FIGURE 6.

Binding of peptides to HSPG and HS. A, sensograms for MCo, kB1, and Tat binding to POPC containing HSPG (0.8 mg/ml) without preincubation. Peptides were tested at 50 μm. B, peptide binding to POPC in the presence of HSPG (0.25 mg/ml) with preincubation overnight at 37 °C. C, binding of MCo (50 μm) to biotinylated HS immobilized on SA sensor chip. MCo does not bind to HS at any of the tested concentrations. D, binding of Tat (50 μm) to biotinylated HS immobilized on SA sensor chip. Tat binds to HS with a corrected intensity of ∼150 RU.

In a second experiment, HSPG was incubated overnight with POPC lipid vesicles to allow the protein to partition into the membrane. No differences were detected between preincubation overnight of POPC-HSPG and direct injection over lipid surface on the chip (Fig. 6B). The results were confirmed with different concentrations of HSPG with and without overnight preincubation.

In another approach, the polysaccharide part of the HSPG receptor, HS, was used. HS chains extend out from the outer membrane surface (47) and are therefore accessible for peptide interactions. The experiment was carried out only with MCo, because it was expected that the positive patch of MCo would be more likely to interact with the negatively charged HS. By contrast, kB1 has a neutral charge, and an interaction with HS is unlikely. For MCo-HS binding experiments, biontylinated HS was attached to a chip coated with streptavidin, and the affinity of the peptides for HS was monitored by SPR. MCo did not bind to HS at any of the concentrations tested (Fig. 6C).

The Tat peptide has been reported to bind to HS and was used as a positive control (48). Indeed, Tat bound to HS in a concentration-dependent manner (Fig. 6D). Tat showed a nonspecific interaction with the control channel where no HS was injected. This kind of nonspecific interaction has been reported for other CPPs using this methodology (49). Overall, the intensity of the response in the presence of HS was high, and it clearly shows the tendency of the Tat peptide to interact with HS with a corrected intensity of ∼150 RU (Fig. 6D).

SFTI-1 Cellular Uptake Studies

The cellular uptake of the cyclic peptide SFTI-1 was studied (see structure in Fig. 7). Studies using live cell imaging with MCF-7 cells indicated that SFTI-1-Alexa can also penetrate cells and is located in endosomal compartments (Fig. 8A). The cellular uptake of SFTI-1-Alexa in MCF-7 was quantified by flow cytometry. SFTI-1-Alexa cellular uptake at 5 μm (Fig. 8B) was comparable with MCo-Alexa (Fig. 4B).

FIGURE 7.

SFTI-1 sequence and structure representation. A, sequence of SFTI-1. Cysteines are highlighted in bold type and numbered with roman numerals. The disulfide connectivity is represented with a gray line. The loops between the cysteine residues are numbered (loops 1 and 2). B, schematic representation of SFTI-1 (Protein Data Bank code 1JBL). Cysteines are numbered with roman numerals, and the disulfide connectivity is represented. Loops are numbered loop 1–2 in accordance with the sequence represented in A.

FIGURE 8.

SFTI-1 cellular uptake and phospholipid interaction. A, SFTI-1-Alexa (15 μm) cellular uptake into MCF-7. SFTI-1-Alexa is represented in green. SFTI-1-Alexa is located in endosomal compartments in MCF-7 cells. Bar, 5 μm. B, SFTI-1-Alexa (5 μm) cellular uptake in MCF-7 quantified by flow cytometry. SFTI-1-Alexa (5 μm) was added to MCF-7 and incubated for 1 h, and the intracellular peptide was quantified by flow cytometry. Error bars represent ± S.E. of three different experiments (n = 6). C, SFTI-1 membrane interaction studies by SPR. SFTI-1 (50 μm) affinity for POPC, POPC/POPG (4:1), and POPC/POPE (4:1) was studied by SPR immobilized on the surface of L1 chip. D, biotin-labeled SFTI-1 (1 μm) was used for phospholipid interaction studies by using a PIP Strips^TM membrane (see lipid spot position in Fig. 5B). SFTI-1 did not interact with any of the phospholipids.

The membrane affinity of SFTI-1 was also tested. SFTI-1 did not have affinity for the model membranes POPC, POPC/POPG, or POPC/POPE (Fig. 8C). Furthermore, biotin-labeled SFTI-1 did not bind to any other phospholipids present in the nitrocellulose membrane strips (Fig. 8D) MCo-biotin was used as a positive control in these studies (Fig. 5B).

DISCUSSION

This study has demonstrated that the disulfide-rich head-to-tail cyclic peptides MCo, kB1, and SFTI-1 enter cells, and we propose that they can be regarded as a new class of cell-penetrating peptides that we define here as CCPPs. Although they all have a cyclic backbone, they enter cells via different mechanisms, and elucidation of their membrane and lipid binding properties has provided insights into these mechanisms.

MCo Enters Cells Mainly by Macropinocytosis Probably by Interacting with PIs and PA

Cellular uptake experiments carried out in living cells demonstrated that MCo-Alexa penetrates RAW 264.7 and MCF-7 cells, but it penetrates the RAW 264.7 cells more efficiently. This observation is not unexpected because macropinocytosis appears to be the main route of internalization of MCo (3), and RAW 264.7 cells display high levels of constitutive macropinocytosis (32). By contrast, MCF-7 cells have a lower propensity for macropinocytosis than RAW 264.7 cells. However, when MCF-7 cells were treated with the macropinocytosis inhibitor EIPA, the cellular uptake of MCo-Alexa decreased by ∼40%, which suggests that MCo-Alexa also penetrates MCF-7 cells to a high extent by macropinocytosis.

A patch of positively charged residues on the surface of MCo could be important for macropinocytotic uptake in RAW 264.7 cells because mutants with Lys residues substituted with Ala entered cells less efficiently than the native peptide. Interestingly, MCoK6A-Alexa uptake increased by 20% relative to MCo-Alexa in MCF-7 cells, indicating that positively charged residues alone do not trigger cellular uptake and that the location of the positive residues is important. Lys-6 is the active site residue for trypsin inhibition, and studies on the related peptide MCoTI-I have shown that mutating the equivalent residue to alanine prevents binding to trypsin and trypsin inhibitory activity (50). Therefore, analogs of MCo without trypsin inhibitory activity can still be internalized into cells.

MCo did not have affinity for model membranes, for the proteoglycan HSPG, or for the glycosaminoglycan HS. The interaction of CPPs with proteoglycans has been reported to be important for the induction of macropinocytosis (46). Recent data suggest that glycosaminoglycans participate in Tat transduction but are not mandatory for the cellular uptake of Tat (51). Consequently, it has been postulated that a different mechanism is responsible for macropinocytotic stimulation by CPPs. This hypothesis is supported by the results presented here, because the stimulation of macropinocytosis by MCo was independent of the interaction with glycosaminoglycan.

In vitro interaction studies showed that MCo-biotin binds to several PIs. From the PIs that show interaction with MCo-biotin, PI(3,4,5)P₃, PI(4,5)P₂, and PI(3,4)P₂ are present predominantly in the plasma membrane (52). It appears likely that in vivo MCo binds to PI(4,5)P₂ because it is the most abundant PI in the plasma membrane (53), and it takes part in endocytotic processes and reorganization of the actin cytoskeleton (42). It has previously been found that proteins with weak affinity for membranes can bind to PI(4,5)P₂ through clusters of basic residues (54, 55), and the PI assists in the recruitment of the proteins to the cell membrane and subsequent internalization (56). Therefore, the binding of MCo to PI(4,5)P₂ could assist in the recruitment of MCo to the plasma membrane, with the subsequent translocation of MCo into the cell. The hypothesis that PI(4,5)P₂ assists in MCo cellular uptake is supported by the recent finding that the binding of the HIV-1 transactivating Tat protein to PI(4,5)P₂ is strictly required for Tat secretion (57). The Tat protein includes 86 amino acids and binds to PI(4,5)P₂ through the Tat basic domain sequence, Tat (48–57).

MCo also interacted strongly with PA. PA has been reported to have a key role in vesicle trafficking and endocytosis (58). Proteins can bind to PA through their polybasic motif (59) and consequently be tethered to cell membranes (60). The binding of positive residues in MCo to this lipid might therefore also play a role in its cellular uptake. Maurocalcine, a scorpion toxin peptide that penetrates cells by macropinocytosis, has also been reported to interact with PA and PIs (29, 61). Despite being a linear peptide, maurocalcine is otherwise similar to MCo, including 33 amino acids and three disulfide bonds that form a cystine knot motif (61). Mutational studies indicated that maurocalcine mutants that entered cells more efficiently also interacted strongly with PA and PIs and vice versa. This implies that PA and PIs interaction is directly correlated with cellular uptake efficiency (29).

kT20K Induces Endocytosis, Probably by Changing Membrane Curvature

It was earlier reported that kT20K does not enter cells but attaches to the outer cell membrane (3). In this study, using improved methodology, we have demonstrated by live cell imaging that Alexa-labeled kT20K can indeed enter cells. kT20K-Alexa colocalized to a significant extent with the membrane, but some peptide was located inside the cell in endosome-like compartments. Differences in peptide labeling and cell imaging methodologies appear to account for the differing results. In the previous report, peptides were labeled with biotin, and the detection was performed with a streptavidin-FITC conjugate after cell fixation. Using this methodology, MCo could not be detected at 5 μm, and therefore, the working concentration was increased to 50 μm (3). Alexa is brighter and more photostable than FITC (62), and it is likely that the biotin-streptavidin system was not sensitive enough to detect kT20K inside cells. Additionally, the live cell system has been shown to be more reliable than fixed cells (31). Consequently, the live cell imaging reported here indicates that kT20K can enter RAW 264.7 and MCF-7 cells, in contrast to the previous study using fixed cells.

Model membrane studies recently showed that kT20K binds to liposomes rich in POPE (28). Based on the strong affinity of kT20K for the membrane, it would be expected that the peptide would enter the cell by direct membrane interaction and therefore be located in the cytosol. Interestingly, kT20K-Alexa was found in endosome-like compartments in RAW 264.7 and MCF-7 cells. Incubation at 4 °C stopped translocation of kT20K-Alexa, which suggests that its cellular uptake is dependent on endocytosis. Nevertheless, there is evidence that kT20K has weak binding to rigid membranes⁷; therefore, we propose that the lack of internalization at 4 °C is due to increased membrane rigidity at low temperatures. Inhibition of cellular uptake at 4 °C has also been reported for the CPP oligoarginine, which enters cells by direct membrane interaction (63).

The relevance of membrane affinity in the cellular uptake of kT20K-Alexa is supported by the significant decrease in uptake in MCF-7 cells of the mutant kV25K-Alexa, which does not have affinity for the membrane (28). By contrast, kT20K-Alexa and kV25K-Alexa cellular uptake in RAW 264.7 cells is comparable. Most likely, kV25K-Alexa and kT20K-Alexa enter macrophages by extracellular liquid engulfing in a nonspecific way. In contrast, MCF-7 cells are not as active in macropinocytosis, and uptake relies on peptide-membrane interaction. In addition, kT20K-Alexa cellular uptake was not affected by the macropinocytosis inhibitor EIPA in MCF-7 cells, suggesting macropinocytosis does not take part in the MCF-7 uptake of kT20K.

kB1 has been reported to induce the outward movement of PE-phospholipids in model membranes (28). Such lipids are mainly restricted to the inner leaflet of the plasma membrane (64) but can have a role in endocytosis. For example, translocation of POPE has been reported to accelerate endocytosis in mammalian cells (65) by producing membrane asymmetry and consequently membrane bending (66–68). Membrane bending can be further promoted by the binding of hydrophobic proteins to the plasma membrane (69–71). Recently, it has been suggested that a facilitated “physical endocytosis” mechanism exists, based on the ability of some peptides to bind to the membrane and produce vesicle formation (72). A similar mechanism has been reported for the peptides duramycin, cinnamycin (73, 74), and melittin (75).

Based on the information above, it can be speculated that kT20K induces endosome formation in two ways as follows: (i) by translocating POPE to the outside of the membrane and therefore creating membrane asymmetry and membrane bending; because PE headgroups are smaller, there will be a positive curvature toward the inside of the cell, favoring budding into the cell of an endosome; and (ii) by inserting its hydrophobic patch into the membrane and creating more bending, causing the membrane to bulge toward the inside of the cell. The proposed model for the cellular uptake of kT20K is supported by the evidence that kB1 and kT20K insert into model membranes and form pores (38).

SFTI-1 Is Also Taken Up into Cells

To test whether other cyclic peptides are internalized, the cellular uptake of SFTI-1 was studied. SFTI-1 was indeed internalized into MCF-7 cells, although it did not interact with any of the phospholipids tested.

In conclusion, the results presented here suggest that MCo enters cells to a high extent by macropinocytosis, but it is likely that other endocytotic routes coexist. The cellular uptake of MCo is possibly assisted by the phospholipids PI(4,5)P₂ and PA. By contrast, kB1 probably produces membrane deformation upon binding to the plasma membrane. This binding appears to be the driving force for endosome-like formation in the cell. Finally, SFTI-1 penetrates cells by a mechanism independent of any tested phospholipid, and the mechanism of penetration remains unresolved at this stage. Overall, the results show that diverse cyclic peptides are readily taken up by cells. Because of their special topology and cyclic backbone, they constitute a new family of cell-penetrating peptides that we refer to as CCPPs. Such peptides have several advantages over linear CPPs, as they are more stable and potentially more bioavailable and are thus promising molecules for drug design. The fundamental studies provided here underpin the potential application of these peptides. Finally, whether or not a wider range of other cyclic peptides can also penetrate cells is not known, but this seems likely given the favorable biopharmaceutical properties of small cyclic peptides such as the immunosuppressive agent cyclosporin.