The Non-Peptidic Part Determines the Internalization Mechanism and Intracellular Trafficking of Peptide Amphiphiles

Background Peptide amphiphiles (PAs) are a class of amphiphilic molecules able to self-assemble into nanomaterials that have shown efficient in vivo targeted delivery. Understanding the interactions of PAs with cells and the mechanisms of their internalization and intracellular trafficking is critical in their further development for therapeutic delivery applications. Methodology/Principal Findings PAs of a novel, cell- and tissue-penetrating peptide were synthesized possessing two different lipophilic tail architectures and their interactions with prostate cancer cells were studied in vitro. Cell uptake of peptides was greatly enhanced post-modification. Internalization occurred via lipid-raft mediated endocytosis and was common for the two analogs studied. On the contrary, we identified the non-peptidic part as the determining factor of differences between intracellular trafficking and retention of PAs. PAs composed of di-stearyl lipid tails linked through poly(ethylene glycol) to the peptide exhibited higher exocytosis rates and employed different recycling pathways compared to ones consisting of di-palmitic-coupled peptides. As a result, cell association of the former PAs decreased with time. Conclusions/Significance Control over peptide intracellular localization and retention is possible by appropriate modification with synthetic hydrophobic tails. We propose this as a strategy to design improved peptide-based delivery systems.


Introduction
Targeted delivery of macromolecular or supramolecular structures in vivo to a desired tissue, cell population or intracellular compartment constitutes a major challenge towards development of effective therapeutic and/or diagnostic modalities [1]. Peptides can serve as targeting agents for drug delivery systems [1,2] and additionally mediate intracellular delivery by efficiently crossing membrane barriers. For example, cell-penetrating peptides (CPPs) have a unique ability to induce internalization of drug formulations in a variety of cells in vitro [3]. Ideally, tissue-specific targeting and cell-specific internalization would be combined in one sequence. Recently, screening of phage libraries led to the identification of the C-terminal peptide motif R/K-XX-R/K (Cend rule or CendR motif) as a critical element in neuropilin-1 (NRP-1) mediated internalization, targeting, and vascular and tissue penetration [4,5].
The favorable tumor-homing and cell penetration properties of CendR peptides led us to explore means for their integration in nanoscale drug delivery systems via self-assembly. Peptides modified with hydrophobic, lipid-like tails known as 'peptide amphiphiles' (PAs) can be used as building blocks for the production of self-assembled nanostructures [6] or as functional coatings on preformed nanostructures [7,8]. The physicochemical properties of the hydrophobic tails and the interactions between peptide headgroups specify the supramolecular geometry [9]. For example, interposition of poly(ethylene glycol) between tissuespecific targeting peptides and a di-stearyl lipid tail favors formation of small spherical micelles [10]. Such micelles demonstrated peptide-mediated, in vivo homing to atherosclerotic plaques and to different tumors in mice [11][12][13]. However, as interactions between the PAs are physical in nature, the structures possess an inherent dynamic character that clearly poses an issue of stability. Indeed, in vitro studies have shown that in presence of albumin and lipid membranes micelle disassembly occurs within minutes [14,15]. As a consequence, PA internalization occurs following micelle disassembly and monomer insertion to the plasma membrane [15,16].
Here we studied the in vitro internalization and trafficking of PAs presenting the prototypic CendR peptide, RPARPAR [4]. Our data indicate that the lipid-anchor and not the peptide is the key determinant factor for internalization and differences in its structure result in altered subcellular trafficking of the amphiphiles. Our results have key design implications for exploiting the potential of PAs in drug delivery applications.

PAs Localize in Intracellular Vesicles
Confocal microscopy confirmed the elevated cell-association of PA 2 compared to peptide 1 ( Figure 1C). After 1-hour incubation, the majority of PA 2 exhibited an intracellular punctate fluorescence pattern indicating vesicular localization ( Figure 1C). At early time points (1 & 10 min) PA 2 localized primarily on the plasma membrane ( Figure S1) indicating that internalization occurs following PA association with the membrane. Qualitatively, there were no differences in intracellular fluorescence distribution between: i) PA 4 and the control amphiphile lacking the peptide sequence (6), ii) carboxylated (2) and amidated (3) RPARPAR PAs and iii) PAs 2 and 8, carrying a different fluorescent label ( Figure  S2). Initial imaging experiments were performed using epiflur-escent microscopy on live cells to exclude fixation artifacts ( Figure  S3). Collectively, these data show that following initial plasma membrane association a large fraction of PAs internalize into intracellular vesicles independent of peptide presence or the nature of fluorescent label.

PA Internalization Occurs Primarily through Clathrinindependent Carriers
To identify the subcellular compartment to which RPARPAR PAs are directed, PPC-1 cells were co-incubated with PAs and different internalization pathway markers (Figure 2A-D). Colocalization of PA 2 with cholera toxin subunit B (CTb) after 1hour incubation revealed substantial overlap, whereas there was only a limited overlap with transferrin and no overlap with lysotracker or mitotracker (Figure 2A-D). Co-localization of PAs with CTb was evident as early as 10 minutes after PA addition to the cells ( Figure S4). Results for PA 4 co-localization with the panel of markers were similar indicating a common internalization pathway for both types of RPARPAR PAs ( Figure S5). CTb binds the glycosphingolipid GM1 present in lipid rafts and following internalization it is trafficked through the Golgi to the endoplasmic reticulum [18]. Although CTb is often assumed to represent a specific marker for clathrin-independent uptake, there is evidence that depending on cell type multiple pathways, including clathrinmediated endocytosis, might be involved in its internalization [18,19]. Dynamin-2 is necessary for clathrin-and caveolaemediated endocytosis and we next examined its involvement in PA internalization. Imaging of PA 2 in PPC-1 cells transiently transfected with EGFP-coupled dynamin-2 showed no overlap of PAs with dynamin-2 ( Figure 2E). Moreover, PA 2 was still able to internalize in PPC-1 cells transfected with a mutant dynamin-2 dominant negative construct (Dynamin-2(K44A)-EGFP) ( Figure 2F). Similar results were obtained with PA 4 ( Figure S6). Combined these results suggested that RPARPAR PAs internalized through clathrin-independent carriers (CLICs).
CLIC formation and internalization require cholesterol in the plasma membrane [20,21]. A decrease of 48% in cell association following treatment with the cholesterol-depleting agent methylbeta-cyclodextrin (MbCD) was found for PA 2 and 84% for PA 4 ( Figure 3A). Compared to non-treated cells, MbCD-treated cells exhibited rounded morphology as previously shown [22] but their viability was not affected ( Figure 3B). The majority of fluorescence in MbCD-treated cells was associated with the plasma membrane and only a small fraction of membrane-bound PAs was internalized ( Figure 3B). The lower cell association of PA 4 with MbCD-treated cells could be due to a lower affinity of this PEGcontaining PA for the plasma membrane compared to PA 2. Despite the similarity in hydrophobic tail and peptide headgroup between the two PAs the presence of the long hydrophilic PEG chain renders PA 4 more hydrophilic. We suggest that PA 4 binds reversibly to the plasma membrane and is removed during the washing step. This hypothesis is also supported by pulse-chase experiments discussed below.
Incubation of PPC-1 cells at 4uC resulted in a 43% decrease in cell association of PA 2 ( Figure 3A). The decrease of association was apparently due to uptake inhibition as demonstrated by confocal microscopy ( Figure 3B). PAs were associated with the plasma membrane rather than being internalized indicating that PA internalization is an active process. Inhibition of clathrin mediated endocytosis by chlorpromazine and incubation with amiloride (an inhibitor of Na + /H + exchangers and macropinocytosis [23]) did not alter PA cell association compared to the controls ( Figure 3A). Collectively, our data suggest CLIC pathway as the major internalization route for PAs in PPC-1 cells.

PA Trafficking is Determined by the Non-peptidic Part of the Molecule
We next examined the intracellular fate of PAs and in particular whether they are retained inside cells or are exocytosed with time. Pulse-chase experiments revealed that the amount of PAs inside PPC-1 cells following a 1-hour pulse and different chase periods was dependent on the non-peptidic part ( Figure 4A). PA 2 cell association remained unchanged from 1 to 3 hours. At the same time, a loss of approximately 75% in PA 4 cell association occurred within one hour of chase and 80% at 3 hours. After a 24hour chase, association of PA 2 with cells also decreased by 50%, but this most likely reflects dilution of the label during cell division since the total amount of fluorescence remained constant ( Figure 4A-inset).
Confocal micrographs of PAs after different chase times showed qualitatively the same intracellular distribution despite the quantitative differences in cell association ( Figure 4B). After 1hour incubation (chase time = 0), PAs localized in the plasma membrane and in vesicles scattered around the cytoplasm. Chase for 1 hour and 24 hours showed substantial loss of membrane fluorescence and gradual accumulation of PA-containing intracellular vesicles at a perinuclear site ( Figure 4B). However, upon closer inspection a difference between the two PAs was observed after a 24-hour chase: whereas PA 4 could be also detected on the plasma membrane, PA 2 was only observed in intracellular vesicles ( Figure S7).
Pulse-chase was next combined with co-localization experiments of PAs 8 (diC 16 tail) and 4 (DSPE-PEG 2000 tail) ( Figure 5). The two PAs stained the same intracellular vesicles when co- incubated for 1-hour ( Figure 5A) as well as when chased for an additional hour ( Figure 5D), indicating that they share the same initial internalization and trafficking routes. PPC-1 cells that were pulsed for 1 hour with PA 4, washed and subsequently incubated for an additional hour with PA 8 did not show overlap of fluorescence demonstrating that PA trafficking after 1 hour has not reached a steady state ( Figure 5B). When the order of PA addition was reversed with PA 8 incubated first, some colocalization of PAs was evident ( Figure 5C) suggesting that after 1hour chase PA 8 still resided in intracellular vesicles that PA 4 are trafficked through. When PA 8 chase time was increased to 3 hours, co-localization was reduced indicating that the two PAs colocalized transiently ( Figure 5E).
We hypothesized that quantitative differences in PA retention observed in the pulse-chase experiments were due to a combination of altered PA trafficking and differences in membrane affinity of the two PAs. We postulated that both PAs are recycled to the plasma membrane, where the more hydrophilic PA 4 would be washed away whereas PA 8 (or 2) would be firmly anchored to the plasma membrane and thus be re-internalized. This hypothesis is consistent with a reduction of PA 4 intracellular levels with time. It also explains why higher co-localization was observed when PA 8 was chased for 1 hour and PA 4 was pulsed for 1 hour compared to when the order was reversed. To test our hypothesis, we transiently transfected PPC-1 cells with Rab4, Rab7, and Rab11 and studied the co-localization of the PAs with Rab-positive vesicles [24,25] (Figure 6). Both PA 2 and 4 co-localized with Rab4-positive vesicles confirming that a fraction of internalized PAs is recycled via the 'short-loop' recycling route to the plasma membrane ( Figure 6A&B). PA 4 additionally showed low levels of co-localization with Rab11-positive vesicles suggesting that recycling occurred also via the 'long-loop' recycling ( Figure 6F). PA 2 on the other hand did not co-localize with Rab11 ( Figure 6E). Both PAs co-localized with the late-endosomal marker Rab7 indicating that they are trafficked towards lysosomes ( Figure 6C&D) even though they had not reached these organelles after 1 hour   Figure 7). However, whereas nearly all PA 2-positive vesicles were identified as lysosomes, PA 4 showed substantial co-localization with CTb-positive vesicles. Collectively, these results show that differences in PA chemical structure lead to differences in their intracellular trafficking.

Discussion
Our data show that PAs associate with cells in vitro by virtue of their hydrophobic tail and subsequent internalization and trafficking are dependent on the chemical structure of the tail and linker to the peptide. RPARPAR PAs did not rely on the CendR mechanism for cell association since i) RPARPAR PAs with amidated (PAs 3&5) and carboxylated (PAs 2&4) C-terminus were associated with PPC-1 cells at the same levels [4] and ii) high PA association was observed with a cell line lacking the CendR receptor NRP1. Nevertheless, the peptide sequence significantly influenced cell association of PAs. PA 2 association was higher than that of PA 7 (diC 16 tails) and PA 4 was associated to a greater extent than amphiphile 6 (DSPE-PEG 2000 tails). It is not yet clear whether the peptide headgroup dependence was an effect of altered PA hydrophilicity, a consequence of non-specific cell binding or a combination of both. Arginine-rich peptides are known to bind strongly to the cell membrane and subsequently internalize into cells, with many CPPs based on arginine [3,26]. At the same time, the higher hydrophilicity of the charged peptide headgroup is expected to increase its desorption rate in an aqueous environment [27], thus accelerating PA transfer between plasma membrane, proteins and self-assembled objects [10,28].
A CLIC internalization pathway of RPARPAR PAs following their anchoring to the plasma membrane was common for the two different tested lipid tails. RPARPAR PAs co-localized after 1hour pulse and 1-hour chase and exhibited identical staining patterns with intracellular markers. The saturated diC 16 -and DSPE-tails of both PAs used in this study are expected to associate with cholesterol-rich, ordered-lipid domains known as detergentresistant membranes or lipid rafts [29]. Co-localization with cholera toxin subunit B (CTb) supported this prediction. Cholesterol depletion significantly inhibited internalization of both RPARPAR PAs; interestingly however, cholesterol depletion did not abolish PA binding to the plasma membrane of MbCDtreated cells (Figure 3), indicating its role in vesicle budding and not in promoting membrane anchoring. Clathrin-mediated endocytosis was not a major uptake mechanism of RPARPAR PAs, since their internalization was independent of dynamin and not inhibited by chlorpromazine; however, we cannot exclude some minor contribution of this pathway. Structurally related amphiphilic lipid-like molecules have previously been shown to employ CLIC pathways, often in combination with clathrin-mediated uptake (CME). A CLIC pathway is employed by the naturally occurring, membrane bound glycophosphatidylinositol (GPI) proteins, which position on the exoplasmic leaflet of the membrane via a glycolipid anchor [21,30]. Synthetic analogs, including a fluorescent lipid molecule and a PEG-lipid conjugate co-localized with both transferrin (marker for CME) and folate (marker for CLIC pathway); a slight preference for the PEG-lipid to internalize by CLIC pathways was noted [31]. Finally, the diC 16 p53 PA used in this study as a control was previously shown to mainly utilize CME for cell entry in SJSA-1 osteosarcoma cells [15]. Therefore, it appears that PAs and similar amphiphiles do not necessarily utilize identical internalization pathways, depending on the cell type and nature of the hydrophilic headgroup.
The fate of PAs following uptake is important for intracellular targeting applications. Our findings demonstrate that the chemical structure of the non-peptidic part is responsible for dissimilar trafficking of the two RPARPAR PAs studied. PA 4 was present in Rab4 and Rab11 positive vesicles suggesting that it is recycled to the membrane using both the short and long recycling loops [24,25]. After 24 hours, a small fraction of this PA was still found in the plasma membrane while most of it was localized in lysosomes and endocytic vesicles, in which CTb accumulates after 1-hour incubation. On the other hand, PA 2 co-localized predominantly with Rab4-positive vesicles at early time points and at 24 hours the majority of the PA was found in lysosomes with plasma membrane devoid of it.
We have summarized our results in a hypothetical model for PA trafficking in Figure 8. Both types of RPARPAR PA utilize the same routes for cell entry and co-localize with Rab4, a marker for 'short loop' recycling endosomes. However, diversion of PA 4 from the lysosomes towards the Rab11-dependent recycling to the plasma membrane results in dissimilar trafficking at longer time points. The increased hydrophilicity and consequent increased desorption rate is responsible for the decrease in its cell-association during chase experiments: each time the PA undergoes an exocytosis-endocytosis cycle some of it is desorbed from the plasma membrane.
The idea that trafficking of lipid-like amphiphiles is determined by the chemistry of their hydrophobic tails was first proposed by Mukherjee et al. [32]. The authors proposed that distribution of amphiphilic dyes in membrane regions of differing fluidity determine their localization in sorting endosomes and therefore determine their ensuing trafficking. In our case, the difference in 2 methylene units/acyl chain is not expected to significantly influence the PA membrane distribution since both PAs partition in lipid-ordered domains at physiological temperature and pH [29,33]. On the other hand, the contrast in headgroup size and steric effects might be the determining factor. This argument, introduced by Bhagatji et al. for explaining how GPI-anchored proteins are excluded from protein-dense clathrin-coated pits [31] could also influence sorting in recycling endosomes.
The findings of our study provide important considerations on peptide amphiphile design. The reversible cell association of the PEG-containing PA is expected to enhance their tissue penetration: internalized PAs are not trapped in the first cell they encounter but can exit and diffuse to neighboring cells. Another potential advantage of the PEG-containing lipids is their routing away of the lysosomes. Avoiding the harsh environment of these organelles should protect peptide drugs from degradation. On the other hand, the lower stability of PEG-lipids supramolecular constructs compared to the shorter, more hydrophobic, 'classical' PAs raises concerns for their in vivo stability and biodistrubution. Future work should focus on in vivo tumor targeting and tissue penetration of the two different PA architectures and correlate the findings to the in vitro results described here. In conclusion, we have shown that PA architecture is a way to control intracellular trafficking and retention of a model peptide.

Peptides & Peptide Amphiphiles (PAs)
Peptides were synthesized using standard solid phase synthesis methods either on a Rink Amide resin or a Wang resin. diC 16    DSPE-PEG 2000 -RPARPAR PAs were synthesized by coupling RPAR peptides bearing a cysteine to 1,2-distearoyl-sn-glycero-3phosphoethanolamine-N-maleimide(polyethylene glycol) 2000 (Avanti Polar Lipids) at 1:1 molar ratio in PBS 10 mM at room temperature for 1 hour. Peptides and PAs were purified using high-performance liquid chromatography (HPLC; Shimadzu Corporation) and identity was verified using matrix-assisted laser desorption/ionization (MALDI) time-of-flight (TOF) mass spectrometry. Materials of purity greater than 95% were stored dry at 220uC until used. A list of the peptides and PAs used is presented in Figure 1. PA solutions were prepared as follows: PAs were dissolved in a 1:1 mixture of chloroform and methanol and solvents were evaporated under N 2 flow to form a PA film on a glass vial, which was then dried in vacuum. PA films were then hydrated at 60uC for 1 hour. PA solutions were stored at 4uC and used within 1 week of preparation.

Cell Culture
PPC-1 and M21 cell lines [4] were cultured as exponentially growing, sub-confluent monolayers in DMEM cell culture medium (ATCC) supplemented with 10% v/v calf bovine serum (ATCC) and 0.1% v/v Penicillin/Streptomycin (Gibco). Cells were grown at 37uC, humidified atmosphere and 5% CO 2 . Cells were detached following a 5-minute incubation with a 2 mM EDTA solution in PBS. For all in vitro studies PPC-1 and M21 cells were seeded at a density of 2610 4 cells/cm 2 . Typically cells were allowed to attach on the surfaces overnight (12-16 h) prior to experiments.

Cell Association Studies
Association of peptides and PAs with adherent cells was determined using fluorescence spectroscopy measurements at 37uC, unless otherwise noted. Cells seeded in 12-well plates were incubated with fluorescent peptides or PAs, washed once with PBS 10 mM, detached and placed in 15 ml centrifuge tubes. Cells were centrifuged for 5 minutes at 1400 rpm, supernatant was discarded and 2 ml PBS 10 mM was used to resuspend the cells. Following a second centrifugation and removal of supernatant, 0.5 ml of triton X-100 in water (1% v/v) were added and the solutions briefly vortexed to lyse the cells. Three 100 ml aliquots were used to measure fluorescence intensity using a Tecan Infinite M200 and two 50 ml aliquots were used with the BCA protein assay kit in order to calculate total protein content. Fluorescence intensity was converted to peptide/PA concentration using calibration curves constructed with pure material in triton X-100 aqueous solutions.

Pulse-Chase Experiments
In pulse-chase experiments, cells were incubated for a first time period (pulse) with a PA, washed with PBS and subsequently incubated for a second time period (chase) in cell culture medium or in presence of a different PA.

Imaging
For microscopy studies, cells were seeded in Lab-Tek chambered coverglass slides (Nalge Nunc). Following incubation with formulations under set conditions cells were washed two times with sterile filtered PBS (10 mM, pH 7.4). In some cases, Hoechst 33342 was added 10 minutes prior to washing to stain cell nuclei. For live cell imaging, cells were visualized in supplemented cell culture medium using a Nikon Eclipse TE-200 microscope equipped with a 10x and 100x objective and a 100 W mercury arc lamp. Images were processed and false color was added using ImageJ software. Alternatively, cells were fixed using 4% parafolmadehyde solution in phosphate buffer (100 mM) for 30 minutes at room temperature.
After fixation, cells were washed and imaged in PBS using a Zeiss Laser Scanning Confocal microscope (LSM 700) equipped with 20x and 63x objectives and 405, 488 and 555 nm solid-state lasers for fluorophore excitation. Image processing was performed using the Zen software provided by Zeiss.

Inhibition, Co-localization and Transfection Studies
A number of different conditions were used in order to gain insight on cell internalization mechanisms. Cells were incubated with 10 mg/ml chlorpromazine or 1 mM amiloride or 5 mM methyl-b-cyclodextrin (MbCD) or in absence of serum and antibiotics, 30 minutes prior to addition of PAs. Inhibitors were present also during PA incubation. Cell association was performed at 4uC to assess energy requirements. Co-localization studies were performed by incubating cells with PAs and 1) 0.