Spontaneous Membrane Translocating Peptides: The Role of Leucine-Arginine Consensus Motifs

Abstract

We previously used an orthogonal high-throughput screen to select peptides that spontaneously cross synthetic lipid bilayers without bilayer disruption. Many of the 12-residue spontaneous membrane translocating peptides (SMTPs) selected from the library contained a 5-residue consensus motif, LRLLR in positions 5–9. We hypothesized that the conserved motif could be a necessary and sufficient minimal motif for translocation. To test this and to explore the mechanism of spontaneous membrane translocation, we synthesized seven arginine placement variants of LRLLRWC and compared their membrane partitioning, translocation, and perturbation to one of the parent SMTPs, called “TP2”. Several motif variant peptides translocate into synthetic vesicles with rates that are similar to TP2. However, the peptide containing the selected motif, LRLLRWC, was not the fastest; sequence context is also important for translocation efficiency. Although none of these peptides permeabilize bilayers, the motif peptides translocate faster at higher peptide to lipid ratios, suggesting that bilayer perturbation and/or cooperative interactions are important for their translocation. On the other hand, TP2 translocates slower as its concentration is increased, suggesting that TP2 translocates as a monomer and is inhibited by lateral interactions in the membrane. TP2 and the LRLLR motif peptide induce lipid translocation, suggesting that lipids chaperone them across the bilayer. The other motif peptides do not induce lipid flip-flop, suggesting an alternate mechanism. Concatenated motifs translocate slower than the motifs alone. Variants of TP2 with shorter and longer arginine side-chain analogs translocate slower than TP2. In summary, these results suggest that multiple patterns of leucine and arginine can support spontaneous membrane translocation, and that sequence context is important for the contribution of the motifs. Because motifs do not make simple, additive contributions to spontaneous translocation, rational engineering of novel SMTPs will remain difficult, providing even more reason to pursue SMTP discovery with synthetic molecular evolution.

Introduction

Cell-penetrating peptides (CPPs), which are members of the large, diverse family of membrane-active peptides, are defined by their ability to cross membranes and enter cells. For several decades, CPPs have been explored as potential vehicles for the delivery of membrane-impermeant cargoes into cells (1, 2, 3, 4). Broadly, CPPs utilize a few overlapping mechanisms to access the cell interior; these can be energy-dependent or energy-independent processes (5, 6, 7). However, mechanistic classification of CPPs is a “moving target” (7) as it is strongly dependent on multiple experimental factors including peptide concentration (8, 9), experimental model system (10, 11), and cargo (12). Overall, the development of useful CPPs as tools in the clinic or in the laboratory is inhibited by the lack of rational design strategies that can be used to discover CPPs with a specific mechanism of action for translocation or CPPs that can deliver particular cargoes (3, 7).

Some highly cationic CPPs can rely on energy-dependent processes, such as endocytosis to enter cells, although this presents the further challenge of endosomal escape for successful delivery of a useful bioactive cargo (7, 13). Under some circumstances, highly cationic CPPs can also enter cells by transiently permeabilizing the plasma membrane (14, 15). Delivery, in this case, may be a threshold phenomenon, meaning that delivery without toxicity will occur under only a narrow range of conditions, limiting potential usefulness.

Some CPPs and other membrane-active peptides can cross synthetic or cellular membranes in an energy-independent manner. In some cases, translocation is driven by membrane insertion and disruption (16, 17) or by the reorganization of membrane lipids (18). In other cases, energy-independent CPPs can spontaneously cross synthetic bilayers and enter cells by translocating directly across the membrane without membrane disruption (19). We recently defined this mechanistic class of CPPs as spontaneous membrane translocating peptides (SMTPs). SMTPs may have the ability to deliver useful bioactive cargos directly into cells by a novel mechanism (1, 20, 21, 22), but only a few such peptides have been described in the literature. Thus, better rational design strategies for SMTPs are needed to expand the usefulness of this class of CPPs.

We previously developed an orthogonal high-throughput screen to select for SMTPs (19) in synthetic lipid vesicles. Over 10,000 unique sequences from a 12-residue peptide library, containing highly hydrophobic to highly cationic sequences, were screened orthogonally for membrane translocation without permeabilization. We selected the 18 most optimal peptides (19) and showed that they belong to a conserved family whose members efficiently cross both synthetic and cellular membranes with membrane-impermeant, polar fluorescent dyes attached (19, 23). The same peptides were shown to significantly improve polar cargo distribution and half-life in vivo (23).

Interestingly, many of the SMTPs selected from the screen contained a consensus LRLLR motif at position 5–9. Conversely, residues 1–4 were much more variable and had no consensus. Residues 10–12 were not varied in the library. We have noted elsewhere that concatenated LRLLR-like sequences are found in the membrane-inserted voltage sensor domains of voltage-gated ion channels (24), which undergo reversible membrane insertion, and can also translocate across bilayers (24), supporting our hypothesis that the LRLLR motif is especially well suited for membrane translocation and that it could be the ideal minimal motif for energy-independent spontaneous translocation. We further hypothesized that the motif might be a useful cassette for engineering new SMTP activities by concatenation. Here, we characterized arginine placement variants of the LRLLR motif for membrane partitioning, translocation, and perturbation using synthetic lipid bilayers to test these ideas. The simple motif sequences were compared to the parent 13-residue peptide, TP2. We also assessed the effect of peptide concentration, transmembrane potential, and membrane disordering on the translocation of both the parent peptide TP2 and the simple motif variants to test various mechanistic hypotheses.

Materials and Methods

Peptide synthesis and labeling

Peptides were synthesized on TentaGel-S-Ram polystyrene resin beads (Rapp Polymere, Tübingen, Germany). FMOC-protected amino acids were dissolved in DMF with N,N,N′,N′-Tetramethyl-O-(1H-benzotriazol-1-yl)uronium hexafluorophosphate, hydroxybenzotriazole, and diisopropylethylamine. Couplings were done at a fourfold molar excess relative to the stated loading capacity of the resin, and were continued for 1 h at room temperature. Fmoc removal was done with 30% v/v piperidine in DMF for 15 min at room temperature. Ninhydrin colorimetric assays were performed after each coupling and Fmoc removal step to monitor completion. Peptide deprotection and cleavage from the resin was done with TFA/phenol/water/TIPS (88:5:5:2) % v/v (Reagent B) for 4 h, starting at 0°C and warming to room temperature. Crude peptide was lyophilized from glacial acetic acid and purified via reverse phase HPLC. Peptide identity was verified with MALDI-TOF mass spectrometry. Labeling of peptides with nitrobenzoxadiazole (NBD) was achieved by dissolving purified peptide in DMF with 2 molar excess of Iodoacetamide-NBD for 2 h at room temperature. Labeled peptides were then purified via HPLC using NBD fluorescence.

Vesicle production

A quantity of 1-palmitoyl-2-oleoyl-3-sn-glycerophosphatidylcholine (POPC; Avanti Polar Lipids, Alabaster, AL) was dried from chloroform and resuspended in buffer (10 mM HEPES, 45 mM NaCl, 1 mM EDTA pH 7.4) containing 1 mg/mL TLCK-treated chymotrypsin (Sigma-Aldrich, St. Louis, MO). Initial lipid concentration was 30–40 mM. The lipid/protease solution was frozen and thawed 10 times using liquid nitrogen and a warm water bath (40°C) followed by extrusion (25) through a polycarbonate membrane with 0.1-μm-sized pores (Whatman, Maidstone, UK), giving large unilamellar vesicles (POPC-LUVs) with entrapped chymotrypsin (POPC-CVs). A Sephadex G-75 size exclusion column was used to separate vesicles, which elute in the void volume, from free chymotrypsin. To reduce the residual external chymotrypsin even further, vesicles are placed in a 100,000 Da molecular mass cutoff spin filter (Vivaspin; https://www.sartorius.com/) at 4°C and allowed to filter under gravity flow overnight. Each flowthrough sample is tested for the presence of chymotrypsin using an EnzChek protease detection kit (Thermo Fisher Scientific, Waltham, MA). Fresh buffer is added to the POPC-CVs to replace the amount of flowthrough collected and the process is repeated until the flowthrough has no detectable chymotrypsin. The POPC-CVs are then removed from the spin filter, resuspended in buffer, and the lipid concentration is measured using a Stewart assay (26).

Translocation assays

POPC-CVs were diluted in buffer (10 mM HEPES, 45 mM NaCl, 1 mM EDTA pH 7.4) to achieve 1 mM final lipid concentration. A two- to threefold excess of α-1-antitrypsin (Sigma-Aldrich) over entrapped chymotrypsin was added to the POPC-CVs and allowed to incubate for 30 min at room temperature before addition of peptide. By extrapolating from lower concentrations of chymotrypsin, where cleavage rates can be measured, we estimate that the halftime of cleavage of the peptides studied here by the 1 mg/mL entrapped chymotrypsin is 50–200 ms. Furthermore, cleavage rates are the same whether peptides are free in solution or mostly bound to lipid vesicles (data not shown). Membrane binding does not decrease cleavage rate. Therefore, any peptide that translocates into a chymotrypsin vesicle will be cleaved by the entrapped chymotrypsin almost instantly.

For translocation measurements, purified NBD-labeled peptides were added to HPLC vials with POPC-CVs and antitrypsin. At each time point, a sample was eluted over a C18 reverse phase HPLC column, using fluorescence monitoring of the NBD probe. Peak areas for intact peptide and cleavage product (H₃N-Cys(NBD)-amide, for all peptides) were used to calculate the fraction of uncleaved peptide remaining at each time point. To show that the entrapped chymotrypsin is active and that the external antitrypsin concentration is sufficient to inhibit all chymotrypsin contained in the POPC-CVs, we lysed POPC-CVs with 0.1% v/v reduced Triton X-100 (Sigma-Aldrich), and then added peptide. In the absence of antitrypsin, rapid and complete peptide cleavage was observed. In the presence of antitrypsin, no cleavage was observed.

Permeabilization

To measure vesicle permeabilization, we prepared POPC-LUVs with entrapped ANTS, a fluorophore, and its quencher, DPX (ANTS/DPX POPC-LUVs) using an established protocol (27). Peptides were added to wells of a 96-well plate at a fixed concentration. ANTS/DPX POPC-LUVs were serially diluted in a separate 96-well plate and then added to the plate containing the peptides. The peptides were incubated with the ANTS/DPX POPC-LUVs for 4 h at room temperature. Detection of ANTS fluorescence was performed on a plate reader (BioTek Synergy II; Bio-Tek, Winooski, VT). All leakage values are compared to 100% leakage established by the maximum ANTS fluorescent signal after treatment of ANTS/DPX POPC-LUVs with detergent (0.1% v/v reduced Triton X-100).

Partitioning

POPC LUVs were produced by freeze-thaws followed by extrusion through a membrane with 0.1-μm-sized pores. Equilibrium dialysis experiments were set up using a Teflon dialyzer (Chemours, Wilmington, DE) separated by a semipermeable cellulose membrane with 100,000 Da molecular mass cutoff (Harvard Apparatus, Holliston, MA). POPC-LUVs were diluted to 10 mM and added to one side of the dialyzer whereas purified NBD-labeled peptide diluted in buffer was added to the other side of the membrane. The dialyzer remained on a rocker at room temperature for 48 h for equilibration. Samples were taken from each side of the dialysis membrane, placed in separate vials and analyzed by HPLC. Peak areas for each sample were then used to calculate the mole fraction partition coefficients, as described in White et al. (28).

Membrane potential

POPC-CVs were created as outlined above except that vesicles were made in a potassium-containing buffer, 10 mM HEPES, 128 mM KCl, 0.1 mM EDTA pH 7.4. After extrusion, the external buffer was replaced by gel filtration with a sodium-containing buffer, 10 mM HEPES, 128 mM NaCl, 0.1 mM EDTA, pH 7.4. After incubation with 2 μM antitrypsin for 30 min, 4 nM valinomycin was added. After 10 min at room temperature, the valinomycin POPC-CVs were used as described above for translocation measurements.

Induced lipid flip-flop

Asymmetric POPC vesicles with 1 mol % C6-NBD-lipid in the outer monolayer were prepared by incubating premade vesicles with 1-palmitoyl-2-{6-[(7-nitro-2-1,3-benzoxadiazol-4-yl)amino]hexanoyl}-sn-glycero-3-phosphocholine (C6-NBD-PC). Vesicles with 1 mol % symmetric C6-NBD-PC were made by mixing all lipids in chloroform before extrusion. To measure peptide-induced flip-flop, we added unlabeled peptide to the asymmetric vesicles and incubated at room temperature. The peptide-lipid mixture was then brought up in 100 μL buffer. Initial NBD fluorescence (Ex. = 470 nm; Em. = 540 nm) was measured on a steady-state fluorometer for at least 1 min. In the meantime, a fresh stock of 0.6 M sodium dithionite was prepared by dissolving sodium dithionite with chilled buffer (10 mM Tris, pH 10.5). Quenching of exposed NBD-PC was measured by adding 45 mM fresh sodium dithionite from the concentrated stock and measuring the loss of NBD fluorescence over 5 min. Time course of intensity decay was fit with an exponential decay curve to determine the protected fraction from the asymptote.

Results

LRLLR motif variants

The 13-residue peptide TP2 (Table 1), a representative SMTP, was selected from a rational peptide library using synthetic vesicles (19). TP2 and the other SMTPs selected from the same library can deliver an impermeant polar fluorescent dye (e.g., TAMRA) and other small polar cargoes across synthetic bilayers and into cells, without endocytosis (19, 23, 29). TP2 and many of the SMTPs discovered along with it (19) have a conserved LRLLR sequence in positions 5–9. Here we test the hypothesis that the LRLLR motif is a minimal motif required for spontaneous translocation and that the placement of arginines in the LRLLR motif is optimal for translocation.

Table 1.

The Peptides Studied in this Work

Peptide	Sequence	Symbol
TP2	PLIY–LRLLR–GQWC
ONEG	PLGR–PQLRR–GQWC
Motif	LRLLR–WC
Motif Variants	RLRLL–WC	▲
	LRLRL–WC
	LLRLR–WC
	RLLLR–WC	▼
	RRLLL–WC
	LLLRR–WC	+

TP2 and ONEG are observed positive and negative peptides for membrane translocation that were selected in a high throughput screen (19). LRLLR is a subsequence of TP2. The six variants all have three leucines and two arginines. All peptides studied have a C-terminal tryptophan-cysteine-amide sequence and a free amino group on the N-terminus. The cysteine sulfhydryl group is labeled with the dye NBD for most experiments. The dashes separate segments of the sequences that appear to have functional significance. The symbols and colors for TP2, ONEG, and the motif are used in all figures herein.

Arginine placement along the peptide backbone has been shown to affect the activity of some arginine-rich CPPs by allowing more favorable arginine-membrane interactions (30). Further, the motif ϕRϕϕR, where ϕ is any hydrophobic residue, is found concatenated three times in potassium channel voltage-sensor domains, which likely interact with and insert into bilayers in response to changes in membrane potential (24). To test the contribution of the LRLLR motif for translocation, we made seven simplified variants with identical amino acid compositions, but varied arginine placement and spacing (Table 1), including arginines with 0, 1, 2, and 3 intervening leucines. In place of the –GQF-amide sequence on the C terminus of the original SMTPs, all motif peptides contained –WC-amide; a tryptophan for membrane binding and ease of quantification and a C-terminal cysteine to enable labeling with a small fluorescent probe, NBD, which enables detection of chymotrypsin cleavage of the C-terminal cysteine. In all cases, we compared motif translocation to that of the parent, 13-residue SMTP TP2 with a C-terminal cysteine, which served as a positive control. For a negative control, we used ONEG, an observed negative peptide for translocation selected in the original SMTP screen.

Translocation rates of the LRLLR motif variants

As described in Marks et al. (19) and He et al. (24), we entrapped chymotrypsin at 1 mg/mL in large unilamellar vesicles made from POPC, to make chymotrypsin vesicles (POPC-CVs), which were used to measure translocation rates for each of the peptides in Table 1. We added α1-antitrypsin outside the vesicles to inhibit any external or released chymotrypsin. The sufficiency of the antitrypsin was verified by lysing the vesicles with reduced Triton X-100 and observing a lack of cleavage of TP2-NBD. The entrapped concentration of protease is high enough to digest translocated peptides in seconds, even when peptide is mostly bound. For the peptides studied here, partitioning is moderately strong, thus free and bound peptides will be in rapid equilibrium on the experimental timescales. Further, inherent cleavage rates are very similar for all peptides because all peptides have a chymotrypsin cleavage site at the C terminus tryptophan (WC-NBD-amide).

To measure translocation rate, peptides were incubated with POPC-CVs and treated with α1-antitrypsin at room temperature. Samples of the peptide/vesicle mixture were analyzed by reverse phase HPLC at various time points to measure the peak area of the intact peptide peak and its degradation product peaks by NBD fluorescence (Fig. 1 A). ONEG served as a negative control for translocation, and little or no cleavage of ONEG was observed. The positive control for translocation, TP2, showed similar rates of translocation into these POPC-CVs as previously reported with vesicles made with 90% POPC and 10% anionic phosphatidylglycerol lipids (19). We note that spontaneous translocation of peptides into 100 nm LUVs is much slower than translocation into giant vesicles and cells (19, 23), such that half-times of 1–2 h in LUVs translate to half-times of 10 min in giant vesicles and cells (19, 23).

Figure 1.

Spontaneous membrane translocation. (A) Given here are example translocation measurements at P/L = 1:100. Lipid vesicles, at 0.5 mM, with entrapped chymotrypsin and external α1-antitrypsin, were incubated with 5-μM NBD-labeled peptide. At intervals, samples were eluted over C18 reverse phase HPLC. The areas of the cleavage product and uncleaved peptides were used to determine the fractional cleavage, as shown. All curves were fit to a single exponential decay. (B) Shown here are the mean and SD of translocation rates in units of h⁻¹. Values are from at least three independent experiments. In all figures and tables, red denotes TP2, and green denotes the LRLLR motif peptide. To see this figure in color, go online.

All of the short motif peptides translocated measurably into POPC-CVs. Changes to the arginine placement in the LRLLR motif resulted in modest changes in translocation rates (Fig. 1 B). At a P/L of 1:100, there was a 17-fold difference between the fastest (RLRLL) rate (1.5 h⁻¹) and the slowest (LRLRL) (0.09 h⁻¹) rate. The parent LRLLR motif peptide translocated at a similar rate to the parent peptide, TP2 (0.9 h⁻¹). The two asymmetric motif variants with one intervening leucine, RLRLL and LLRLR, showed even higher translocation rates than that of TP2 or the LRLLR motif. Variants with adjacent arginines, RRLLL and LLLRR, were slower than LRLLR, whereas the symmetric variants, RLLLR and LRLRL, had the slowest rates of translocation.

Concentration-enhanced translocation has been reported for multiple CPPs in cells (6, 31, 32) and in synthetic vesicles (10). Increased translocation rates coupled with increasing P/Ls suggest that translocation is due to a cooperative effect that could involve peptide self-assembly, membrane perturbation, or both. It is not ideal for SMTPs to rely on self-assembly as this may lead to unintended toxic membrane perturbations and will narrow useful concentration windows. We have previously reported evidence suggesting that TP2 is unusual in that it translocates noncooperatively as a monomer at low concentration. To test whether translocation of TP2 and the motif peptides is cooperative, we measured translocation rates at P/L = 1:100 (as in Fig. 1) and 1:500. We also measured translocation at 1:1000 for TP2. The results in Fig. 2 A show that translocation rates of the motif peptides decrease significantly when P/L is decreased from 1:100 (1 mol % peptide) to 1:500 (0.2 mol % peptide). The LLRLR motif, for example, translocates >500-fold slower at 1:500 (compared to 1:100) whereas the RRLLL motif changes only approximately threefold. The rank order for translocation rates at P/L = 1:500 is different than that at a P/L of 1:100. RLRLL is fastest at P/L = 1:100 and RRLLL is fastest at P/L = 1:500. The parent LRLLR motif is the third fastest of seven at 1:100 and the third slowest at 1:500. These results suggest strongly that translocation of the motif peptides is influenced either by membrane disruption or by cooperative peptide-peptide interactions, or by both. As expected, because of their short length, the motif peptides have overall random coil secondary structure in bilayers (data not shown), but it remains possible that the low abundance transition state during translocation requires a specific structure (33). Further, the results suggest that the various motif peptides do not have exactly the same mechanism of translocation as they are affected differently by concentration.

Figure 2.

Effect of P/L on activity. (A) Translocation rates were measured as described in Fig. 1 at P/L = 1:100 and P/L = 1:500. For TP2 we also measured the translocation rate for P/L = 1:1000. Peptides for which no bar is shown have rates <0.001 h⁻¹. (B) Shown here is membrane permeabilization across a wide range of P/L for the translocating peptides and motifs. POPC vesicles contained entrapped ANTS and DPX: a fluorophore quencher pair (27). The fluorescence of ANTS was used to quantitate permeabilization. No release of ANTS/DPX was observed for any of the translocating peptides. The pore-forming peptide melittin is a positive control and releases ANTS/DPX at relatively low concentrations. To see this figure in color, go online.

Interestingly, TP2 has the opposite behavior; increasing P/Ls caused decreased rates of translocation (Fig. 2 A). The effect is moderate, with a 20-fold decrease in translocation rate when P/L is increased from 1:1000 to 1:100. This result agrees with our previous conclusion that TP2 translocates as a monomer (34). In addition, it suggests that lateral interactions between peptides, which we have been reported to result in β-sheet structure (34) at high peptide concentration, may inhibit translocation of TP2. The opposite concentration dependencies of TP2 and the motif peptides translocation cannot currently be explained. Future experimental and computational work will be needed to address this fascinating observation.

Membrane leakage with LRLLR motif variants

The effect of P/L on translocation of the motif peptides suggested that their translocation requires some degree of membrane disruption or lateral peptide-peptide interactions in the membrane, or both. CPPs that rely on relatively high peptide concentrations for cooperative translocation likely use a mechanism of translocation that involves membrane disruption (15, 35). Here, we tested the ability of TP2 and the motif peptides to permeabilize PC lipid vesicles by measuring the release of the entrapped fluorophore ANTS and its quencher DPX (27). No significant leakage of ANTS/DPX was observed with either TP2 (as reported previously) or the motif peptides over a wide range of P/L ratios, including those used in translocation experiments (Fig. 2 B). The bee venom peptide, melittin, is a positive control that causes leakage at the expected concentrations. We conclude that neither the motif peptides nor TP2 cause enough bilayer disruption to enable release of the 350-Da small molecules ANTS and DPX. The motifs thus also do not cause enough disruption for self-promoted passage through peptide-induced pores in the membrane. The concentration dependences of TP2 and the motif peptides likely arise from lateral interactions in the membrane. These interactions apparently inhibit translocation of TP2, and promote translocation of the motif peptides.

Next, we measured the mole fraction partition coefficients (K_x) for all NBD-labeled peptides to test the hypothesis that differences in translocation rates reflect differences in the fraction of peptide that is bound to vesicles at equilibrium. The measured K_x values range from 5 × 10⁴ to 4 × 10⁵ (Fig. 3), therefore at 1 mM lipid, the fraction of bound peptide ranges from 48 to 88% (28). This less than twofold variation in fraction bound is not enough to account for a 10- to 100-fold difference in translocation rates (Figs. 1 B and 2 A). In any case, the translocation rates are not correlated with partition coefficient within this series of peptides (p > 0.05, r² = 0.011; see Fig. 3).

Figure 3.

Lack of correlation between mole fraction partition coefficient (K_x) and translocation rate (K_T). Translocation rate is plotted against mole fraction membrane partition coefficient (K_x), which was measured by fluorescence titration (54) and equilibrium dialysis (28). Linear regression gave a p value of 0.8 for correlation. To see this figure in color, go online.

The measured partitioning ΔG values for the motif peptides and TP2 range from −6.8 to −7.6 kcal/mol, in reasonable agreement with the value of −6.3 and −7.1 kcal/mol estimated for the motif peptides and TP2, respectively, using the Wimley-White interfacial hydrophobicity scale (36). For this calculation, we estimate that the difference in ΔG between Cys(SH) and Cys(NBD) is –2 kcal/mol. This agreement supports our conclusion that these peptides are mostly unstructured and interfacially bound.

The role of transmembrane potential in spontaneous translocation

Living eukaryotic cells maintain a net negative-inside electrochemical potential across the plasma membrane that could drive translocation of SMTPs such as TP2, or other CPPs. In fact, such an effect has possibly been observed for some CPPs, including penetratin (37, 38). The rate of translocation of SMTPs will depend on the highest energy transition state encountered during translocation. The vast majority of CPPs, including the ones studied here, are cationic (39, 40, 41), which means that the transition state probably occurs when arginine or lysine residues are in the middle of the bilayer (33, 42). In silico molecular modeling was previously used to investigate the translocation of TP2 across an implicit slab membrane (33). A possible semistable intermediate state for TP2 was suggested in which the two arginines of the motif bridge the inner and outer leaflets whereas the remainder of TP2 forms a compact secondary structure in the bilayer core (33). Based on this structure, which shows that the arginines cross one at a time, it is reasonable to hypothesize that the highest energy transition state would be when the first arginine crosses the bilayer hydrocarbon core.

If this idea is correct, an electrochemical potential across the membrane could exert a driving force for spontaneous translocation. We therefore tested the effect of a transmembrane potential across POPC-LUVs on the translocation rates for TP2 and the motif variants. To make vesicles with an initial transmembrane potential, we prepared POPC vesicles with 128 mM KCl inside and 128 mM NaCl outside. Valinomycin, a potassium-specific ionophore, was then added to create a negative-inside membrane potential (37) by specific transport of potassium. An increase in translocation rate after valinomycin treatment would suggest that the spontaneous translocation of peptides is coupled to the electrochemical potential gradient across the membrane. However, as we show in Fig. 4, translocation rates for the motif peptides in LUVs treated with valinomycin (which have a membrane potential) were not markedly different from the same vesicles without valinomycin (which do not have a membrane potential). Neither TP2 nor the motif peptides translocate measurably faster in the presence of a negative inside electrochemical potential.

Figure 4.

The effect of transmembrane potential on translocation. POPC vesicles were made with entrapped chymotrypsin and entrapped 128 mM KCl. In the exterior solution, there was α1-antitrypsin and 128 mM NaCl. Valinomycin was added to some samples to shuttle potassium across the membrane and generate a negative inside transmembrane potential of −110 mV. Translocation experiments, as described above, were performed in the absence and presence of 4 nM valinomycin, to create 0 and −110 mV membrane potential, respectively. To see this figure in color, go online.

This result suggests that the highest energy transition state of the translocating species is neutral for the motif peptides. Because arginine likely remains protonated at all times in membranes (43, 44), translocation may thus require a transient complex between the guanidine group and a counteranion. The guanidinium group of arginine can potentially form multiple hydrogen bonds with the lipid phosphate located in the interfacial domain of membranes (45, 46, 47, 48). In this scenario, hydrogen bonding between the guanidinium groups of arginines and the phosphate of POPC lipids could neutralize cationic charges, and a neutral complex may be the species that generates the highest energy transition state in the middle of the bilayer during translocation (45, 48).

The role of lipid disorder in spontaneous translocation

Lipid disorder and lipid transbilayer equilibration (i.e., flip-flop) have been associated with the so-called adaptive translocation mechanism for spontaneous translocation (48), which recognizes the importance of hydrogen bonding between arginines and interfacial phosphates of phospholipids. Complex formation leads to lipid molecules adapting (i.e., chaperoning) peptides across the bilayer (46). In support of this idea, one group has shown that fatty acids can efficiently chaperone cationic peptides across bilayers (14), presumably because fatty acids disorder bilayers with detergent-like properties while also presenting carboxyl groups to interact with arginines. If the translocating species is a transient complex between peptides and lipids, bridged by guanidine-phosphate interactions, we hypothesized that introducing membrane disorder would increase the rate of translocation. To test this idea, we used alamethicin, a potent pore-forming peptide from the fungus Trichoderma viride, which has been shown by several groups to promote rapid lipid flip-flop even at very low concentration (49). Previous observations in our lab have shown that alamethicin incubated with LUVs at P/L ratios <1:1000 induces rapid exchange (flip-flop) of dye-labeled lipids from the inner leaflet of the bilayer to the outer leaflet (50), an effect that occurred even when the concentration of alamethicin was too low to cause membrane permeabilization.

Although we do not know the molecular mechanism of this activity of alamethicin, it must require localized disruption of lipid packing, or perhaps transient, localized nonbilayer structures in the membrane. Here we tested the idea that lipid disruption caused by alamethicin, which promotes translocation of lipids, can also promote the translocation of peptides or perhaps peptide-lipid complexes. To find the concentration of alamethicin that did not allow rapid peptide entry into vesicles by pore formation, we incubated the negative control for translocation, ONEG, with POPC-LUVs with entrapped chymotrypsin in the presence of varying concentrations of alamethicin. The degradation by entrapped chymotrypsin was measured at 30 min (Fig. 5 A). At P/Ls above 1:2000, alamethicin created membrane defects large enough for rapid peptide entry as shown by rapid ONEG degradation. At or below P/L = 1:2000, we observed little ONEG degradation on this timescale. As a result, we used 1:6000 P/L alamethicin to test the correlation between lipid disorder and peptide translocation. Even at this very low concentration of alamethicin (<20 peptides per vesicle), we observed an increase in rates of translocation for all motifs in POPC LUVs at P/L = 1:100. Translocation of the motifs increased by 3- to 10-fold in the presence of alamethicin, and the peptides with the slowest rates had the largest relative increase.

Figure 5.

Effect of membrane disorder on translocation. (A) Shown here is the effect of the pore-forming peptide alamethicin on the cleavage of ONEG by chymotrypsin vesicles. Cleavage was measured at 30 min. Below P/L 1:2000, cleavage does not occur, which indicates that peptides cannot enter vesicles through alamethicin pores at P/L ≤ 1:2000. (B) Given here is the translocation rate of peptides at P/L = 1:100 in the absence and presence of alamethicin at P/L = 1:6000, measured as described above. (C) Given here is the translocation rate for TP2 at different concentrations in the presence and absence of alamethicin at P/L = 1:6000. To see this figure in color, go online.

Translocation of TP2 at P/L 1:100 was also increased moderately by alamethicin (Fig. 5 B). However, we showed earlier that the translocation of TP2, unlike the motifs, is inhibited at P/L = 1:100 compared to lower P/L (Fig. 2 A). Therefore, we also studied the effect of alamethicin on TP2 translocation at P/L = 1:500 and 1:1000. The results in Fig. 5 C show that the effect of alamethicin is much larger at P/L = 1:100 (sixfold increase) than at P/L = 1:500 or at P/L = 1:1000, where only a 1.3-fold increase was observed.

Although we do not know exactly how alamethicin increases disorder, we conclude from these results that increasing the amount of lipid disorder creates more favorable pathways for peptide-lipid complexes to translocate across bilayers. For TP2, the disturbance caused by alamethicin has a greater effect at high P/L where we hypothesized earlier that peptide self-assembly decreases translocation rate.

Lipid flip-flop and translocation

To test the hypothesis that lipids can chaperone SMTPs across bilayers, we measured peptide-induced lipid translocation (flip-flop) directly using a published technique (51, 52). POPC bilayers were incubated with 1 mol % C6-NBD-PC lipid with one short chain, creating asymmetric vesicles. These vesicles were then incubated with peptides, or buffer. Lipid flip-flop is normally very slow but peptide-induced flip-flop will create a pool of NBD lipids on the inner monolayer, where their NBD moieties will be protected from the membrane impermeant quencher dithionite. The results in Fig. 6 show that the parent peptide TP2 induced significant protection of NBD lipids, and thus is driving translocation of lipids. Lipid flip-flop was not observed after incubation with buffer or ONEG. The LRLLR motif also induced substantial lipid flip-flop. Interestingly, the other motif peptides caused no measurable lipid flip-flop. It is possible that the placement or the arginines in the LRLLR motif promote lipid-chaperoned translocation and that peptides with the selected motif, LRLLR, may have a special mechanism of translocation. If so, there is a second mechanism that does not involve lipid flip-flop, which was used by the other motifs. In either case, peptide arginines must be able to cross the bilayer. We suggest that TP2, aided by the LRLLR motif, does this as a monomer by being chaperoned by lipid phosphates, and the other motifs translocate without concomitant lipid translocation. Their arginines may yet interact with lipid phosphates, but do so in a way that does not require lipid translocation to accompany peptide translocation. We speculate that peptide arginines may be handed off from one lipid phosphate to another during translocation. Translocation of some peptides may require that the motifs interact with other peptides or other counterions in the membrane, rather than, or in addition to, lipid phosphates.

Figure 6.

Peptide-induced flip-flop. Asymmetric POPC vesicles with 1 mol % C6-NBD PC lipids in the outer monolayer were incubated with peptide at a P/L of 1:25 overnight. A higher P/L was used here to compensate for the weaker binding of the NBD-free peptides compared to translocation experiments done with labeled peptides. A concentration of 45 mM of fresh sodium dithionite was used to quench the accessible NBD on the outer membrane of the vesicles after incubation with buffer, SMTP, or ONEG. Treatment of symmetric vesicles shows that approximately half of the NBD is protected as expected. Each column is the mean ± SD of at least four experiments. By ANOVA, TP2 and LRLLRWC cause highly significant flip-flip (p < 0.0001 and p < 0.001, respectively). The other motif peptides are statistically indistinguishable from blank (p > 0.05). To see this figure in color, go online.

Concatenated motif variants

Ultimately, we would like to be able to engineer novel SMTPs with specific properties or SMTPs that can deliver specific cargoes. We showed above that some of the simple motif peptides translocate across bilayers with rates that are similar to the parent peptide, TP2. This raises the following questions: Are the properties of these simple peptides additive? Can the motifs be used as cassettes that can be combined to engineer SMTPs with novel properties? We have begun to test these ideas by measuring translocation of peptides made by the concatenation of two-motif variants. We used the motifs that showed the highest (RLRLL) and lowest (LRLRL) rates of translocation alone to make the concatenated peptides RLRLL-RLRLL-WC and LRLRL-LRLRL-WC to test if concatenation has an additive effect on translocation. By lengthening the sequences with this modular idea in mind, we predicted that the peptides would have the opportunity for more interactions with the membrane, which could enhance their translocation abilities. They would also have more opportunity for secondary structure (28), although we cannot predict whether increased secondary structure of the transition state will increase or decrease translocation rates because we observed both effects, as discussed above.

Translocation rates for both concatenated motif variants decreased when compared to their respective single motifs and to TP2 (Fig. 7). In particular, the slowest translocating concatamer was made from the fastest motif, and vice versa. Simple additive modularity does not appear to be a reliable feature of the translocation motifs.

Figure 7.

Translocation of concatenated motifs. The fastest (RLRLL) and slowest (LRLRL) motifs were concatenated to form new 12-residue peptides. Both peptides have an RLLR motif at the junction between the motifs. The novel peptides translocated slower than TP2. Translocation rates are not an additive or predictable property in concatenated peptides. To see this figure in color, go online.

Arginine variants of TP2

Based on the computational transition state model described above (33), we hypothesized that the specific location of the guanidine groups of TP2, with respect to the interfacial domains of the membrane, may be a crucial determinant for spontaneous translocation. To test this hypothesis, we synthesized two TP2 variants in which we replaced the native arginines with derivatives that vary in length. Homoarginine contains a longer alkyl side chain with the addition of an additional methylene carbon, whereas norarginine contains a shorter side chain with one fewer methylene group. We hypothesized that these modifications to the arginines of TP2 would change the ability of the guanidinium group to “snorkel” up toward the interfacial domain of the membrane during the transition state. The model suggested that we would observe an impact of side-chain length on translocation. Experimentally, translocation rates for the two arginine variants of TP2 were both lower than TP2 with arginine (Fig. 8). Ultimately, the data suggest that the unique positions of the guanidinium groups of arginines in the context of TP2 are ideal for the peptide to adopt the most favorable conformation during a translocation event. However, we also note that the efficient translocation of short peptides with different arginine placement suggests that there is some flexibility in the guanidine geometry. Ultimately, we do not have a molecular explanation for this observation, but we note that TP2 and the LRLLR motif was selected from a library that contained only arginine. It may be enlightening to revisit this issue by screening libraries containing other Arg variants as well.

Figure 8.

Effect of arginine side-chain length on translocation. TP2 (Table 1) and variants with shorter and longer arginine side chains were assayed in parallel for translocation as described above. All side chains have a guanidine group, but differ in the number of methylene (-CH₂-) groups in the side chain. To see this figure in color, go online.

Discussion

Sequence context in spontaneous membrane translocation

We previously described how a family of 12-residue SMTPs, discovered using an orthogonal screen of a peptide library, frequently contained a conserved LRLLR motif at positions 5–9 (19). Each of the nine varied positions in the original library contained a possible arginine, and every position contained a possible hydrophobe: leucine (2, 4, 5, 7, 8, 9), proline (1), isoleucine (3), or phenylalanine (6). Therefore, the library contained many different arginine-hydrophobe motifs throughout the sequence, yet the conserved LRLLR motif was found in positions 5–9 in 10 of 18 SMTPs. Two other peptides had related motifs LQLLR and RFLLR in positions 5–9. We also noted that the residues 1–4 in the selected peptides were much more variable; some SMTPs have four hydrophobes (e.g., PLIY in TP2) whereas others have as many as two basic residues (e.g., RRIL, in one selected peptide).

In this work, we tested the hypothesis that the conserved LRLLR motif in the TP2 family could be a minimal sequence for spontaneous membrane translocation, and that the specific placement and/or spacing of the arginines within the motif is necessary for efficient translocation. All motif variants that we studied translocate across synthetic bilayers, although arginine spacing and position in the sequence influences translocation rate in a concentration-dependent manner. Although the motif observed in the screen, LRLLR, translocated across PC bilayers, it was not the fastest translocating variant. At P/L = 1:100, peptides with RLRLL and LLRLR have the fastest rates whereas peptides with LRLRL, RRLLL, and LLLRR have the slowest. Interestingly, the slowest translocating peptide LRLRL is a one-residue circular permutation of the two fastest, RLRLL and LLRLR. We conclude that the composition of these peptides supports translocation, generally. The arginine spacing and placement influence translocation, but the observed spacing and position of the arginines in the LRLLR motif is not the only arrangement that supports efficient translocation.

Sequence context appears to be an important factor in determining the rates of translocation. For example, the simple motif patterns that translocated the fastest (and caused no permeabilization) were not often selected in the library, in the context of the 12-residue library members, despite the fact that they were available to be selected. Furthermore, the selected peptides frequently had the conserved LRLLR motif in residues 5–9 when it could have been selected elsewhere in the sequence. The molecular reasons for these preferences are unknown.

Mechanistic insights

Specific interactions between arginine guanidine groups and lipid phosphates or other counterions have been proposed to be important in membrane insertion and/or translocation of voltage sensor domains, antimicrobial peptides, cell-penetrating peptides, and membrane translocating peptides (44). Specifically, it has been proposed that lipid phosphates can dramatically reduce the energetic cost of inserting arginine into the membranes (45) and that lipids effectively chaperone arginine across the bilayer (43). Several of our results further support this mechanistic model for SMTP translocation: 1) a transmembrane potential did not affect translocation, suggesting that the transition state arginine has a counterion; 2) even a trace amount of alamethicin, a peptide that induces lipid disorder, increases translocation rates; and 3) TP2 and the LRLLR motif peptide directly increase translocation rates (flip-flop) of lipids.

Pathways to new SMTPs

Ultimately, SMTPs will need to be further developed and refined if they are to deliver useful bioactive molecules directly to the interior of cells, especially in vivo. Most of the data available for the TP2 family, and indeed for most CPPs in Kauffman et al. (7), are only for the delivery of fluorescent dyes. We have yet to learn how to routinely deliver most other classes of cargoes.

Based on the work reported here, SMTP behavior is not fully predictable, is nonadditive, and is dependent on sequence context in ways we do not yet understand. Because of this, the rational engineering of novel SMTPs will remain difficult, providing even more reason to pursue SMTP discovery with synthetic molecular evolution, a process defined by iterative cycles of library design and screening (53). Our previous peptide library (i.e., the first generation of SMTPs) was designed to broadly span the physical properties of CPPs from different classes. Thus, the sequences ranged from highly hydrophobic to highly basic, and did not vary in length. The studies described here will enable us to more intelligently design the next generation of SMTP library. We now know that many peptides with a ratio of around three leucines (or other hydrophobes) to two arginines can translocate without permeabilization, and that multiple arginine spacings and placements will support translocation. Peptides as short as seven residues will translocate.

Importantly for our next generation library and screen, we found here that increased concentration can either promote or inhibit translocation, showing that there are multiple mechanisms. Some SMTPs (e.g., TP2) translocate fastest as monomers at low concentration, whereas others translocate better at high concentration when they might be self-assembling in the membrane. Although we cannot currently predict monomeric translocation, we can readily design a screen for it in the next generation screen. For example, we can screen in parallel at P/L = 1:1000 and at P/L = 1:100 to select for monomer translocation that does not require, or is not inhibited by, peptide interactions in the membrane. The observations made here are being used to guide the design of refined, next generations of SMTP libraries and screens to enable selection of SMTPs with novel or improved function and improved cargo delivery capabilities.

Author Contributions

W.C.W. and T.F. designed experiments that were conducted and analyzed by T.F. Both authors wrote the manuscript.

Editor: Andreas Engel.